# ETHYLENE BIOLOGY AND BEYOND: NOVEL INSIGHTS IN THE ETHYLENE PATHWAY AND ITS INTERACTIONS

EDITED BY : Dominique Van Der Straeten, Angelos K. Kanellis, Caren Chang, Mondher Bouzayen, Panagiotis Kalaitzis, Jin-Song Zhang and Autar Krishen Mattoo PUBLISHED IN : Frontiers in Plant Science

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-678-5 DOI 10.3389/978-2-88963-678-5

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

## ETHYLENE BIOLOGY AND BEYOND: NOVEL INSIGHTS IN THE ETHYLENE PATHWAY AND ITS INTERACTIONS

Topic Editors:

Dominique Van Der Straeten, Ghent University, Belgium Angelos K. Kanellis, Aristotle University of Thessaloniki, Greece Caren Chang, University of Maryland, College Park, United States Mondher Bouzayen, National Polytechnic Institute of Toulouse, France Panagiotis Kalaitzis, Mediterranean Agronomic Institute of Chania, Greece Jin-Song Zhang, Institute of Genetics and Developmental Biology (CAS), China Autar Krishen Mattoo, Agricultural Research Service (USDA), United States

Citation: Van Der Straeten, D., Kanellis, A. K., Chang, C., Bouzayen, M., Kalaitzis, P., Zhang, J.-S., Mattoo, A. K., eds. (2020). Ethylene Biology and Beyond: Novel Insights in the Ethylene Pathway and its Interactions. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-678-5

# Table of Contents

*05 Editorial: Ethylene Biology and Beyond: Novel Insights in the Ethylene Pathway and Its Interactions*

Dominique Van Der Straeten, Angelos Kanellis, Panagiotis Kalaitzis, Mondher Bouzayen, Caren Chang, Autar Mattoo and Jin-Song Zhang


Likai Wang and Hong Qiao


Hua Qin, Lina He and Rongfeng Huang


Sudarat Khunmuang, Sirichai Kanlayanarat, Chalermchai Wongs-Aree, Shimon Meir, Sonia Philosoph-Hadas, Michal Oren-Shamir, Rinat Ovadia and Mantana Buanong


Yi Chen, Valérie Rofidal, Sonia Hem, Julie Gil, Joanna Nosarzewska, Nathalie Berger, Vincent Demolombe, Mondher Bouzayen, Beenish J. Azhar, Samina N. Shakeel, G. Eric Schaller, Brad M. Binder, Véronique Santoni and Christian Chervin

*111 Biochemical Characterization of the* Fusarium graminearum *Candidate ACC-Deaminases and Virulence Testing of Knockout Mutant Strains*

Thomas Svoboda, Alexandra Parich, Ulrich Güldener, Denise Schöfbeck, Krisztian Twaruschek, Marta Václavíková, Roland Hellinger, Gerlinde Wiesenberger, Rainer Schuhmacher and Gerhard Adam


Carolin Seyfferth, Bernard A. Wessels, András Gorzsás, Jonathan W. Love, Markus Rüggeberg, Nicolas Delhomme, Thomas Vain, Kamil Antos, Hannele Tuominen, Björn Sundberg and Judith Felten

*162 Super-*Agrobacterium *ver. 4: Improving the Transformation Frequencies and Genetic Engineering Possibilities for Crop Plants* Satoko Nonaka, Tatsuhiko Someya, Yasuhiro Kadota, Kouji Nakamura and Hiroshi Ezura

### *174 Red to Brown: An Elevated Anthocyanic Response in Apple Drives Ethylene to Advance Maturity and Fruit Flesh Browning* Richard V. Espley, Davin Leif, Blue Plunkett, Tony McGhie, Rebecca Henry-Kirk, Miriam Hall, Jason W. Johnston, Matthew P. Punter, Helen Boldingh, Simona Nardozza, Richard K. Volz, Samuel O'Donnell and Andrew C. Allan

### *189 The Ethylene Precursor ACC Affects Early Vegetative Development Independently of Ethylene Signaling*

Lisa Vanderstraeten, Thomas Depaepe, Sophie Bertrand and Dominique Van Der Straeten

# Editorial: Ethylene Biology and Beyond: Novel Insights in the Ethylene Pathway and Its Interactions

Dominique Van Der Straeten<sup>1</sup> \*, Angelos Kanellis <sup>2</sup> , Panagiotis Kalaitzis <sup>3</sup> , Mondher Bouzayen<sup>4</sup> , Caren Chang<sup>5</sup> , Autar Mattoo<sup>6</sup> and Jin-Song Zhang<sup>7</sup>

<sup>1</sup> Laboratory of Functional Plant Biology, Department of Biology, Ghent University, Ghent, Belgium, <sup>2</sup> Laboratory of Pharmacognosy, Department of Pharmaceutical Sciences, Aristotle University of Thessaloniki, Thessaloniki, Greece, <sup>3</sup> Department of Horticultural Genetics and Biotechnology, Mediterranean Agronomic Institute of Chania, Chania, Greece, 4 INRA Génomique et Biotechnologie des Fruits (GBF), Toulouse, France, <sup>5</sup> Department of Cell Biology and Molecular Genetics, University of Maryland, College Park, MD, United States, <sup>6</sup> Beltsville Agricultural Research Center, Agricultural Research Service (USDA), Beltsville, MD, United States, <sup>7</sup> Institute of Genetics and Developmental Biology, Chinese Academy of Sciences (CAS), Beijing, China

Keywords: ethylene, ACC (1-aminocyclopropane-1-carboxylic acid), stress, ACC synthase (ACS), ACC oxidase (ACO), gold standard

**Editorial on the Research Topic**

**Ethylene Biology and Beyond: Novel Insights in the Ethylene Pathway and Its Interactions**

This Research Topic presents selected contributions to **Ethylene 2018**, the **XI International Symposium on the Plant Hormone Ethylene**, held in Chania, Greece, on 2nd−6th June 2018, covering exciting new discoveries in the ethylene field.

This issue brings novel insights in ethylene signaling in bacteria, algae, and lower plants as well as evidence supporting a specific role for the ethylene precursor 1-aminocyclopropane-1-carboxylic acid (ACC) in plants.

Ethylene receptors were initially thought to be specific to plants. Interestingly, ethylene receptor homologs have been found in cyanobacteria. It was previously shown that in Synechocystis sp. an ethylene receptor regulates phototaxis and biofilm formation. In this issue, the same group (Allen et al.) demonstrates that another cyanobacterium, the filamentous species Geitlerinema sp. which possesses two ethylene receptor homologs, similarly displays ethylene-dependent alterations in phototaxis, suggesting that such signaling could be prevalent in cyanobacteria. Both species are highly sensitive to ethylene although their ethylene binding characteristics resemble that of plants, thus suggesting signal amplification in cyanobacteria.

In higher plants, subfamily-I of the Arabidopsis receptors (ETR1 and ERS1) and their interactions with downstream players have been extensively studied, while information on subfamily-II (ETR2, ERS2, and EIN4) remains sparse. Here, Berleth et al. demonstrate that ETR2 displays comparable affinities for CTR1 and EIN2 to that previously reported for ETR1, suggesting similar protein-protein interaction-mediated signal transfer for both subfamilies. In addition, the authors show enhanced stability of type-II receptor homomers and type-II:type-I heteromers as compared to type-I homomers, emphasizing the importance of type-II receptors.

Wang and Qiao present a mini-review on the transcriptional regulation of the ethylene response in Arabidopsis. Ethylene signaling involves numerous regulatory steps leading to a diversity of responses in plant growth and development. This review discusses our current understanding of the transcriptional regulation as a major control point in ethylene signaling, focusing on recent insights into the role of chromatin modification in repressing transcription.

Edited and reviewed by:

Gerrit T. S. Beemster, University of Antwerp, Belgium

#### \*Correspondence:

Dominique Van Der Straeten dominique.vanderstraeten@ugent.be

#### Specialty section:

This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science

Received: 21 January 2020 Accepted: 18 February 2020 Published: 12 March 2020

#### Citation:

Van Der Straeten D, Kanellis A, Kalaitzis P, Bouzayen M, Chang C, Mattoo A and Zhang J-S (2020) Editorial: Ethylene Biology and Beyond: Novel Insights in the Ethylene Pathway and Its Interactions. Front. Plant Sci. 11:248. doi: 10.3389/fpls.2020.00248

**5**

Dolgikh et al. summarize recent advances on the molecular mechanisms that underlie EIN3/EIL1-directed ethylene signaling in Arabidopsis, and focus on the role of EIN3/EIL in tuning transcriptional regulation of ethylene response in time and space. Furthermore, they consider the role of EIN3/EIL1-independent regulation of ethylene signaling.

Qin et al. review hormonal crosstalk of ethylene in primary root growth of Arabidopsis and rice. Based on the proposed model, ethylene restricts primary root growth by governing auxin biosynthesis, transport, and signaling through EIN3/EIL1, ERF1, and HB52 interactions in Arabidopsis. ABA and CKs constrain primary root growth by controlling the posttranscriptional regulation of ACS resulting in stimulated ethylene synthesis. GA and ethylene antagonistically regulate stability of DELLA proteins, which act as growth repressors. Low levels of BRs hinder ethylene synthesis by BZR1 and BES1 suppressed ACS gene expression, while high levels of BRs increase ACS stability. Through a different mechanism, ethylene restricts primary root growth in rice, by augmenting auxin and ABA biosynthesis. Understanding light-dependent differences in ethylene synthesis and signaling is essential to expand our insight into the roles of ethylene in growth and development across the plant life cycle.

Harkey et al. performed a meta-analysis of multiple transcriptomic datasets to uncover responses to ethylene that are both light-dependent and light-independent. A set of 139 transcripts with robust and consistent responses to elevated ethylene across root-specific datasets was identified. This "gold standard" group of ethylene-regulated transcripts includes numerous genes encoding proteins that function in ethylene signaling and synthesis. The study further reveals a number of previously uncharacterized factors that may contribute to ethylene response phenotypes.

Plants synthesize ethylene in a two-step reaction starting with the conversion of S-adenosyl-L-methionine (SAM) to 1-aminocyclopropane-1-carboxylic acid (ACC). Vanderstraeten et al. demonstrated that ACC affects early vegetative development independently of ethylene signaling. Using ethylene biosynthesis and signaling inhibitors, as well as mutants, ACC-specific ethylene-independent growth responses in both dark- and lightgrown Arabidopsis seedlings were revealed. Hence, researchers employing ACC as an ethylene precursor should be mindful of putative ACC effects confounding ethylene responses in vegetative growth. The exact mechanism underlying the ACC response remains to be identified.

While ACC may play an ethylene-independent role in plant growth and development, the three-membered ring amino-acid is probably also of paramount importance in the balanced coexistence of plants and beneficial micro-organisms around them, surmounting attacks of pathogens.

Though the genome of the pathogenic fungus Fusarium graminearum, which infects wheat, seems to lack genes encoding ethylene biosynthetic enzymes, Svoboda et al. identified two ACC deaminase genes, one of which encodes an active ACC deaminase. Considering that ethylene insensitive wheat exhibited higher resistance to Fusarium and reduced mycotoxin content, Fusarium knockouts of both genes were generated. No differences in the fungal infection and mycotoxin content were detected indicating that ethylene might not affect full Fusarium virulence.

Nonaka et al. used ACC deaminase in a Super-Agrobacterium strain to enhance transformation efficiencies of recalcitrant species. Super-Agrobacterium was updated to version 4 by introducing both an ACC deaminase and GABA transaminase gene, encoding enzymes that degrade ACC and GABA, respectively. The use of this modified strain resulted in substantial reduction of the amount of time and labor required for transformations, hence providing a more powerful tool for plant genetic engineering and functional analysis.

While most studies have focused on the conversion of SAM to ACC by ACC synthases, there has been increasing interest in the subsequent step, carried out by ACC oxidases (ACOs). Houben and Van de Poel provide a comprehensive and historical perspective on our understanding of the ACO enzyme, including its discovery, activity, evolution, expression, regulation and potential biotechnological applications.

A thorough analysis of players in ethylene signaling enabled identification of key regulators in fruit development and ripening. Not unexpectedly, translational regulation of ethylene biosynthesis and signaling appears to be fundamental in the control of ethylene effects, besides transcriptional regulation.

Mitalo et al. show that many aspects of kiwifruit ripening can be triggered by ethylene treatment or by low temperature, with the exception of aroma volatile production, a major ripening feature contributing to fruit quality, that is lacking in cold-triggered ripening. The data indicate that the production of aroma compounds is strongly ethylene-dependent and support the notion that in kiwifruit ethylene and low temperature-induced ripening may involve two different regulatory mechanisms.

Chen et al. used a modification of parallel reaction monitoring, a targeted mass spectrometry proteomic method, to study ETR receptor abundance in tomato fruit. Focusing on single peptides of rare proteins, they compared the abundance of ETRs in WT and in the NR mutant. Pearson correlations between mRNA and protein profiles were used as indicators to discriminate the two genotypes and reveal changes over fruit development. It is proposed as an approach to study ETR subfunctionalization across the plant kingdom in development as well as in plant-microorganism interactions.

Regarding the role of ethylene in pigmentation, Khunmuang et al. demonstrate that ethylene-mediated petal color fading in cut Vanda "Sansai Blue" flowers results from degradation of anthocyanidins, cyaniding, and delphinidin, and is not related to the levels of flavonols, such as kaempferol. The endogenous anthocyanin degradation process stimulated by ethylene appears related to increased peroxidase activity but was independent of flower senescence. Espley et al. investigated the incidence of internal browning flesh disorder (IBFD) in high anthocyanin red apples (Malus x domestica). The study, using the highly pigmented "Royal Gala" apple cultivar over-expressing MYB10, revealed that the anthocyanin-related transcription factor is associated with the undesirable fruit disorder. The MYB10 transgenic fruit had a high incidence of IBFD compared to wild type. Interestingly, MYB10-expressor apple had higher expression of ethylene-related genes ACS, ACO, and ERF. Prematurely induced ethylene can advance fruit maturity and, as shown by these authors, may lead to adverse effects on storage of high-anthocyanin fruits.

Hormonal pathways are not functioning as independent routes, but rather contribute to an information web, connecting internal signaling to transduction pathways of external cues. Ethylene is involved in such an intricate network which fine-tunes development from cell and organ specification to senescence and abscission, but also incorporates external signals enabling plasticity in an ever changing environment, and controlling reactions to adverse conditions, such as those encountered as a result of global climate change.

External factors such as wind, hale or snow, and/or growth on uneven terrain, induces tension wood formation in order to reorient and uplift the stem toward its original growth position. Tension wood is characterized by the presence of gelatinous, cellulose-rich (G-)fibers with its microfibrils oriented parallel to the fiber cell axis. Seyfferth et al. demonstrated that ethylene regulates transcriptional responses related to the amount of G-fiber formation and their properties, including chemistry and cellulose microfibril angle, in hybrid aspen. The quantitative and qualitative changes in G-fibers are likely to contribute to uplifting of stems.

Given the importance of ethylene in agricultural applications and the need for sustainable crop production, on the field as well as in post-harvest control of fruits and vegetables, recent research has also focused on improved control of ethylene release and sensitivity, as well as on long-term storage conditions of crop products.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Van Der Straeten, Kanellis, Kalaitzis, Bouzayen, Chang, Mattoo and Zhang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# 1-Aminocyclopropane-1-Carboxylic Acid Oxidase (ACO): The Enzyme That Makes the Plant Hormone Ethylene

Maarten Houben and Bram Van de Poel\*

Molecular Plant Hormone Physiology Laboratory, Division of Crop Biotechnics, Department of Biosystems, KU Leuven, Leuven, Belgium

The volatile plant hormone ethylene regulates many plant developmental processes and stress responses. It is therefore crucial that plants can precisely control their ethylene production levels in space and time. The ethylene biosynthesis pathway consists of two dedicated steps. In a first reaction, S-adenosyl-L-methionine (SAM) is converted into 1-aminocyclopropane-1-carboxylic acid (ACC) by ACC-synthase (ACS). In a second reaction, ACC is converted into ethylene by ACC-oxidase (ACO). Initially, it was postulated that ACS is the rate-limiting enzyme of this pathway, directing many studies to unravel the regulation of ACS protein activity, and stability. However, an increasing amount of evidence has been gathered over the years, which shows that ACO is the rate-limiting step in ethylene production during certain dedicated processes. This implies that also the ACO protein family is subjected to a stringent regulation. In this review, we give an overview about the state-of-the-art regarding ACO evolution, functionality and regulation, with an emphasis on the transcriptional, posttranscriptional, and post-translational control. We also highlight the importance of ACO being a prime target for genetic engineering and precision breeding, in order to control plant ethylene production levels.

Keywords: ethylene biosynthesis, 1-aminocyclopropane-1-carboxylate oxidase, transcriptional and posttranslation regulation, phylogeny, physiology

### INTRODUCTION ON ETHYLENE BIOSYNTHESIS

Ethylene was the first gaseous hormone to be discovered in plants. It is an important regulator of many developmental and physiological processes such as seed dormancy, germination, vegetative growth, flowering, climacteric fruit ripening, and senescence. Additionally, ethylene was shown to play an important role in the plant's defense against biotic and abiotic stress factors (Lin et al., 2009; Van de Poel et al., 2015; Wen, 2015).

The general precursor of the ethylene biosynthesis pathway is the amino acid methionine (**Figure 1**; Lieberman et al., 1966). In a first, but general reaction, methionine is converted into S-adenosyl-L-methionine (SAM) by SAM synthetase using ATP (Adams and Yang, 1977). The subsequent reaction steps are unique to the ethylene biosynthesis pathway. First, SAM is converted into 1-aminocyclopropane-1-carboxylic acid (ACC) and 5<sup>0</sup> -methylthioadenosine (MTA) by ACCsynthase (ACS) (Murr and Yang, 1975; Adams and Yang, 1979; Boller et al., 1979). ACS is a member

#### Edited by:

Caren Chang, University of Maryland, College Park, United States

#### Reviewed by:

Gyeong Mee Yoon, Purdue University, United States Donald Grierson, University of Nottingham, United Kingdom

\*Correspondence: Bram Van de Poel Bram.vandepoel@kuleuven.be

#### Specialty section:

This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science

Received: 08 April 2019 Accepted: 09 May 2019 Published: 29 May 2019

#### Citation:

Houben M and Van de Poel B (2019) 1-Aminocyclopropane-1-Carboxylic Acid Oxidase (ACO): The Enzyme That Makes the Plant Hormone Ethylene. Front. Plant Sci. 10:695. doi: 10.3389/fpls.2019.00695

**8**

of the pyridoxal-5<sup>0</sup> -phosphate (PLP) dependent aminotransferases, which use PLP as a co-factor (Boller et al., 1979). The side product MTA is recycled back to methionine by the Yang cycle to avoid a depletion of methionine during high rates of ethylene production (Murr and Yang, 1975). More details on the different steps of the Yang cycle are presented in Bürstenbinder et al. (2010) and Pommerrenig et al. (2011). In a second step, ethylene is released from ACC by ACC-oxidase (ACO) (Hamilton et al., 1990; Ververidis and John, 1991), a reaction that requires molecular oxygen (Burg and Burg, 1965). Alternatively, ACC can be converted to malonyl-ACC (MACC; Amrhein et al., 1981), γ-glutamyl-ACC (GACC; Martin and Saftner, 1995), and jasmonyl-ACC (JA-ACC; Staswick and Tiryaki, 2004). An in-depth review on the derivatization of ACC is given by Van de Poel and Van Der Straeten (2014).

### THE DISCOVERY OF ACO AND ITS REACTION MECHANISM

For a long time it remained extremely difficult to purify ACO (formerly named the Ethylene Forming Enzyme, EFE) and determine its in vitro activity, mainly because it was thought that ACO was a membrane bound protein that lost its activity upon homogenization (Kende, 1989). Some residual or partial in vitro ACO activity was retained in membrane preparations of pea (Guy and Kende, 1984; Porter et al., 1986), bean (Guy and Kende, 1984; Mayne and Kende, 1986), Sprenger's asparagus (Porter et al., 1986) and kiwi fruit (Mitchell et al., 1988), which was only a fraction (5–0.5%) of the total in vivo ethylene production capacity. A breakthrough was made when the clone pTOM13 was characterized to code for a putative ACO gene of tomato (Hamilton et al., 1990). The elucidation of the protein sequence of this first ACO allowed Ververidis and John to find sequence similarity with a flavonone 3-hydroxylase of snapdragon (Antirrhinum majus). This homology made them realize that both iron and ascorbic acid could be essential for ACO enzyme activity. This insight made Ververidis and John (1991) the first to successfully extract and quantify in vitro ACO activity from melon fruit tissue.

Iron, in the form of Fe(II), is an essential metal cofactor, which is required for ACO enzyme activity (Bouzayen et al., 1991). Iron participates by coordinating the binding of the amino group of ACC to H177 and the carboxylate group of ACC to D179, which are two critical ACO residues in the reaction center (Zhang et al., 2004; Tierney et al., 2005; Brisson et al., 2012). The ascorbate cofactor is used as a reductant to catalyze the opening of the ACC-ring (Zhang et al., 2004; Murphy et al., 2014). The ACO reaction mechanism also uses molecular oxygen and bicarbonate as activators in order to catalyze the conversion of ACC into ethylene (Adams and Yang, 1981; Peiser et al., 1984). During this reaction, an unstable intermediate cyanoformate ion [(NCCO2) <sup>−</sup>] is formed, which rapidly decomposes in CO<sup>2</sup> and CN<sup>−</sup> (Murphy et al., 2014). The reactive cyanide ion (CN−) is subsequently detoxified into β-cyanoalanine (Peiser et al., 1984; Dilley et al., 2013; Murphy et al., 2014).

ACC-oxidase is a member of the 2-oxoglutarate-dependent dioxygenase (2OGD) superfamily of non-heme iron-containing proteins (Kawai et al., 2014). The 2OGD superfamily is one of the largest enzyme families in plants, with most of its members being active in oxygenation and hydroxylation reactions (Kawai et al., 2014). Nonetheless, 2OGD enzymes can have more diverse roles and participate for example in demethylations, desaturations, ring closure, ring cleavage, epimerization, rearrangement, halogenation, and demethylenation reactions in plants (Farrow and Facchini, 2014). Characteristic for all 2OGDs is the double-stranded β-helix (DSBH) core fold, which contains a typical 2-His-1-carboxylate motif required for iron binding, also encountered in ACO. This motif consists of two His residues and the carboxylate group from an Asp or a Glu residue, and is responsible for the ligation of Fe(II) in the enzyme catalytic site, and thus critical for ACC binding (Aik et al., 2015; Martinez and Hausinger, 2015; Murphy et al., 2014).

Despite the fact that 2OGD enzymes are typically localized in the cytosol (Kawai et al., 2014), the exact subcellular localization of ACO remains a matter of debate. Some studies have suggested that ACO is localized at the plasma membrane (Rombaldi et al., 1994; Ramassamy et al., 1998), as originally postulated (Kende, 1989). However, other studies have shown that ACO is localized in the cytosol (Peck et al., 1992; Reinhardt et al., 1994; Chung et al., 2002; Hudgins et al., 2006), matching the general localization of 2OGD enzymes. Other studies have measured ACO activity both for membrane/apoplast and intracellular preparations (Bouzayen et al., 1990). All these studies used immunolocalization or activity assays in combination with (sub)cellular fractionations, and perhaps these techniques did not provide sufficient resolution to elucidate the exact ACO localization. A recent study tagged a safflower (Carthamus tinctorius) ACO with a GFP (green fluorescent protein) and performed an ectopic localization in onion epidermis cells. Their results showed that CtACO1 localizes in the cytosol (potentially linked with membranes) and in the nucleus, but their images lacked markers for these organelles (Tu et al., 2019). All studies combined are not conclusive about the exact ACO localization, and thus the actual subcellular site of ethylene production.

#### ACO PHYLOGENY AND RESIDUE ANALYSIS

The plant 2OGD superfamily can be categorized into three subclasses (DOXA, DOXB, and DOXC) based on amino acid sequence similarity (Kawai et al., 2014). ACO is a part of the DOXC subclass, the largest and most diverse group, containing over 400 different 2OGDs, mainly linked to the specialized metabolism (Kawai et al., 2014). Kawai et al. (2014) further subcategorized the DOXC subclass and classified ACO as part of the DOXC53 subclade. This subclade has 2OGD members, which are typically retrieved in all angiosperms. ACO is a unique member of the plant 2OGD superfamily, because it uses ascorbate as a catalyst instead of 2-oxoglutarate (Kawai et al., 2014).

A small phylogenetic study using a limited amount of ACO sequences from tomato (Solanum lycopersicum), potato

(Solanum tuberosum), bonnet pepper (Capsicum chinense), petunia (Petunia hybrida), and tobacco (Nicotiana tabacum) classified the ACO protein family in three distinct phylogenetic groups (Jafari et al., 2013). A more detailed phylogenetic analysis of putative ACOs from mosses, lycophytes, gymnosperms, monocots, and dicots showed that ACO got more diversified after the monocot-dicot split (Clouse and Carraro, 2014). They also observed that there are three main clusters of ACOs and that monocot and dicot ACOs diverged together from a common pre-gymnosperm ancestor (Clouse and Carraro, 2014).

Because not many ACOs have been shown to be functional ACO enzymes that can convert ACC into ethylene, it remains questionable if putative ACOs used in phylogenetic analyses are in fact functional ACOs. Trivial protein sequence similarity searches may lead to false or incorrect ACO annotations in genome and protein databases. In fact, there are only a few studies that have purified recombinant ACOs for functional characterization. This was done for tomato (SlACO1-3; Solyc07g049530, Solyc12g005940, Solyc07g049550; Bidonde et al., 1998), petunia (PhACO1; Zhang et al., 2004), apple (MdACO1; MDP0000195885; Dilley et al., 2013), and Arabidopsis (AtACO2; AT1G62380; Sun et al., 2017). The study of Clouse and Carraro (2014) used annotated, but functionally unverified, ACO protein sequences as queries to identify novel ACO sequences in other species, without performing reciprocal BLAST searches. This approach resulted in the identification of false ACOs, leading to an overestimation of the size of the ACO protein family in certain species (e.g., 13 ACO members for Arabidopsis thaliana instead of 5). Therefore, we have performed a novel sequence similarity search using only the tomato ACO1 (Solyc07g049530) as search query, because this protein has been shown to be a true ACO with a confirmed activity (Bidonde et al., 1998). BLASTp jobs were done for 21 species using the Phytozome (v12.1.) database and Gymno plaza 1.0 (Proost et al., 2014), and top hits were only retained after a positive reciprocal BLAST search. **Table 1** lists all the putative ACOs for some agriculturally important crops and Arabidopsis, while **Supplementary Table 1** lists all the ACOs for the other plant species used in our phylogenetic analysis. We were able to identify 5 ACO members for Arabidopsis, 7 for tomato, 7 for apple, 9 for rice, and 13 for maize (**Table 1**). All putative ACO sequences were used to build a phylogenetic tree (see **Supplementary Figure 1**), which clearly shows a cluster of "ancient" ACOs within the clade of non-seed land plants and algae. This ancient clade most likely originated from an evolutionary distant algal 2OGD that gradually diverged into a functioning ACO during seed plant evolution. A more detailed phylogenetic tree of a selected amount of agriculturally important angiosperms shows 3 clusters of ACOs (**Figure 2**). Therefore, we suggest dividing the ACO family in three types: Type I, Type II, and Type III ACO. These three clusters are also observed in the larger phylogenetic tree of **Supplementary Figure 1**. Our analysis also shows that the gymnosperm ACOs group within the Type III ACO cluster of angiosperms, and that monocot and dicot ACOs diverged separately for each individual type. Our phylogenetic analysis indicates that the 3 types of ACOs diverged in parallel from a shared non-seed plant ancestral ACO or 2ODG.

A detailed residue analysis of the ACO alignment of Arabidopsis, tomato and apple presented in **Figure 3** further confirms the existence of 3 types of ACO. The important 2-His-1 carboxylate Fe(II) binding motif is conserved in all ACOs. Shaw et al. (1996) provided some first experimental insight that this motif is composed of the H177-D179-H234 triad in MdACO1 and that it is essential for ACO activity. This was confirmed in other studies in apple (Kadyrzhanova et al., 1999; Yoo et al., 2006) and for a petunia (Zhang et al., 2004) and tomato ACO (Brisson et al., 2012) (see also **Supplementary Table 2**).

Furthermore, a thorough mutagenesis study of MdACO1, identified other important residues essential for ACO activity: C28, T157, K158, R175, Q188, K199, K230, R244, S246, K292, E294, E297, R299, F300, and E301 (Dilley et al., 2013). Some residues (R175, R299, and K158) have been proposed to coordinate bicarbonate binding (Zhang et al., 2004; Brisson et al., 2012; Dilley et al., 2013), while other residues (K292, K158, and F300) are proposed binding sites for ascorbate

TABLE 1 | List of ACO sequences used for construction of the maximal likelihood phylogenetic tree of Arabidopsis thaliana, Solanum lycopersicum, Malus domestica, Oryza sativa, and Zea mays.


Sequences were retrieved from Phytozome (v12.1) and top hits were only retained after a positive reciprocal BLAST search. Numbering of the different ACOs was done according to previously published work (if available), otherwise a new numbering is proposed. Putative splice variants of the same gene are represented by α, β, and γ .

(Dilley et al., 2013). Besides these six amino acids, two additional residues (R244 and S246; highly conserved and part of the so-called RXS motif) complete the ACC/bicarbonate/ascorbate binding site of ACO (Kadyrzhanova et al., 1999; Seo et al., 2004; Zhang et al., 2004; Brisson et al., 2012; Dilley et al., 2013). Interestingly, the three ACO types can be classified based on the intermediate residue present in the conserved RXS-motif. This motif consists of R-M-S for type I ACOs, R-L/I-S for type II ACOs, and R-R-S for type III ACOs. All the residues considered important for ACO activity according to Dilley et al. (2013), are conserved in the three types of ACO, except for E294, E297, and E301. E294 is not well conserved in the three ACO types, while E297 is replaced by glycine only in the type II ACOs and E301 is not conserved in type III ACOs. It remains to be investigated whether or not the 3 types of ACOs actually have differences in functionality related to for example enzyme activity and/or protein stability.

FIGURE 2 | Maximal likelihood phylogenetic tree for ACO protein sequences of Arabidopsis thaliana (AT), Tomato (Solanum lycopersicum; Solyc), Apple (Malus domestica; MDP), Rice (Oryza sativa; Os), and Maize (Zea mays; Zm) retrieved from Phytozome (v12.1). Protein sequences were aligned in Geneious (v10.2.2) using the MUSCLE alignment plugin. The phylogenetic tree was build using RAxML (v8.2.11) for best-scoring maximum likelihood tree with rapid bootstrapping (1000 bootstrap replicates). Bootstrap values for the main branches are depicted on the tree. Type I ACO is shown in blue, Type II ACO is shown in red, and Type III ACO is shown in green.

red, and Type III ACO is shown in green.

#### THE REGULATION OF ACO

ACC-oxidase is expressed to a variable degree in all vegetative and reproductive tissues, which led to the belief that ACO proteins are always present and ready to produce ethylene. Furthermore, treating plant tissue with ACC typically results in a rapid production of ethylene. Therefore, it has been proposed that not ACO, but ACS is the rate-limiting enzyme in ethylene biosynthesis (Adams and Yang, 1979). This hypothesis has been readily absorbed by the community, leading to an abundance of studies focusing on unraveling the regulation and function of ACS in relation to its prime role in ethylene production (Argueso et al., 2007; Booker and DeLong, 2015; Yoon, 2015). However, there is an increasing amount of evidence demonstrating the importance of ACO, and not ACS, in controlling ethylene production in plants. For example ACO is the rate limiting step during flooding of tomato (English et al., 1995) and Rumex palustris (Vriezen et al., 1999). ACO activity, and not the availability of ACC, has been shown to be crucial during the formation of tension wood in poplar trees (Andersson-Gunneras et al., 2003; Love et al., 2009). Ethylene induced cotton fiber cell elongation has also been linked to a strong upregulation of its respective ACO genes (Shi et al., 2006). ACO has also been shown to be rate limiting during the post-climacteric ripening of tomato fruit (Van de Poel et al., 2012; Grierson, 2014; Van de Poel et al., 2014a,b). Even more recently, a key role for ACO during the sex determination of cucumber flowers was discovered (Chen et al., 2016). These studies indicate that ACO can sometimes be rate limiting and thus controls ethylene production, indicative of a stringent regulatory mechanism that controls ACO expression, stability and/or activity.

#### Transcriptional Regulation of ACO

Despite the fact that ACO expression has been observed to be temporally and spatially regulated (e.g., during tomato flower and fruit development; Barry et al., 1996; Blume and Grierson, 1997; Nakatsuka et al., 1998; Van de Poel et al., 2012), only a few transcription factors have been identified that are known to control ACO expression (see **Table 2**). In tomato, SlHB-1, a homeodomain-leucine zipper (HD-Zip) class-I transcription factor was shown to interact with the tomato ACO1 (Solyc07g049530) promoter using gel retardation assays (Lin et al., 2008). Furthermore, experiments using virus-induced gene silencing showed that a repression of HB-1 expression resulted in a decrease in ACO1 transcript levels (Lin et al., 2008). Additionally, Lin et al. (2008) predicted that HB-1 could also target other ripening-related genes such as ACO2 (Solyc12g005940), PG1, RIN, and NOR (Lin et al., 2008). Martel et al. (2011) reported that the master ripening regulator RIN could interact with the promoter of HB-1, placing HB-1 downstream of RIN during tomato fruit ripening. Later it was shown that RIN itself can interact directly with the CArG box in the promoter region of ACO4 (Solyc02g081190) (Li et al., 2017). Besides HB-1 and RIN, different NAC transcription factors have also been observed to play an important role in the control of ethylene biosynthesis in tomato. Specifically, SNAC4 and SNAC9 have been shown to influence tomato fruit ripening by interacting with the promoters of ACS2, ACS4, and ACO1 (Kou et al., 2016). Silencing SNAC4 and SNAC9 dramatically reduces the expression of these genes, inhibiting fruit ripening. Furthermore, silencing of ERF2, ACS4, and ACO1 also reduces the expression of both SNAC4 and SNAC9, which suggests the existence of a tightly controlled feedback mechanism (Kou et al., 2016).

Ethylene response factors (ERFs) have been shown to be an integral part of the ethylene signaling and response pathway. ERFs are transcription factors that can bind with cis-acting elements such as GCC-box motifs and dehydration-responsive elements (DREs) (Ohme-Takagi and Hideaki, 1995; Müller and Munné-Bosch, 2015). Zhang et al. (2009) showed that the tomato ERF2 (and a homolog allele TERF2) was able to interact with the DRE in the promoter of SlACO3 to activate transcription. They observed a significant increase in ethylene production of the ERF2/TERF2 overexpression lines and a decrease in the ERF2/TERF2 antisense-lines compared to the wild type, suggesting that these ERFs are positive regulators of ACO3 expression in tomato (Zhang et al., 2009).

In banana (Musa acuminata), the transcription factor ERF11 was shown to interact directly with the GCC-box motif in the promoter region of ACO1 and repress ACO1 expression (Han et al., 2016). Han et al. (2016) also demonstrated that ERF11 can physically interact with the histone deacetylase HDA1, which in turn reinforces the ERF11-induced repression of ACO1. Furthermore, the MADS-box transcription factor MADS7 was also shown to interact directly with the promoter of ACO1 in banana using a yeast one-hybrid (Y1H) system and a transient GUS-reporter activation assay in tobacco (Liu et al., 2015). MADS7 is only expressed in banana fruit and its expression is stimulated by ethylene and inhibited by 1-MCP. Ectopic overexpression of MaMADS7 in tomato fruit resulted in a 10-fold increase of SlACO1 expression compared to wild-type fruit, and resulted in an enhanced ethylene production level (Liu et al., 2015).

Another transcription factor that controls ACO expression was also identified in melon fruit (Cucumis melo). Huang et al. (2010) reported that EIN3-like proteins EIL1 and EIL2 induce the expression of ACO1 by interacting with different cis-acting elements of the ACO1 promoter. It was hypothesized that both EIL proteins are targeted for proteolysis by EBF1/EBF2 (similar as in Arabidopsis) in the absence of ethylene, however, upon ethylene release they are stabilized and elevate the biosynthesis of ethylene by inducing the transcription of ACO1 and thus promote ripening (Huang et al., 2010).

In cucumber (Cucumis sativus), the transcription factor WIP1 can regulate flower sex determination by directly binding the promoter of ACO2 and inhibiting its expression (Chen et al., 2016). Evidence was provided using a dual luciferase activation assay in tobacco, Y1H, ChIP-qPCR, and EMSA to validate the interaction between WIP1 and the ACO2 promoter (Chen et al., 2016). Chen et al. (2016) also demonstrated that the melon (Cucumis melo) homolog of WIP1 can interact with the promoter of CmACO3, and similarly as in cucumber, negatively influence CmACO3 expression.

In Arabidopsis thaliana, the NAC transcription factor Speedy Hyponastic Growth (SHYG) was shown to interact with the



promoter region of ACO5 (AT77330; Rauf et al., 2013). When SHYG was overexpressed using an inducible promoter, the expression of ACO5 was shown to be strongly induced (Rauf et al., 2013).

#### Differential ACO Expression Profiles

In order to get a better insight in the differential expression of the ACO gene family, we have summarized the tissue-specific and developmental expression profiles for Arabidopsis and tomato using the eFP browser (http://bar.utoronto.ca/efp/cgibin/efpWeb.cgi; Waese et al., 2017) and the Tomato Expression Atlas (http://tea.solgenomics.net/), respectively.

#### **Differential ACO expression in Arabidopsis thaliana**

**Figure 4** demonstrates that the Arabidopsis thaliana ACO genes show a distinct tissue-specific expression pattern. ACO1 (AT2G19590; Type II) is upregulated during the torpedo and walking-stick stage of embryogenesis, is highly expressed in imbibed seeds and upregulated during the first stages of germination, mainly in the radicle. Furthermore, ACO1 is also strongly expressed in the roots, where it might be involved in lateral root formation (Park et al., 2018). ACO2 (AT1G62380; Type I) was shown to be involved in germination, where it participates in ethylene production to control endosperm cap weakening and endosperm rupture (Linkies et al., 2009; Linkies and Leubner-Metzger, 2012). ACO2 is mostly expressed in the emerging seedling hypocotyl, where it is involved in the formation of the apical hook (Raz and Ecker, 1999). ACO2 is also highly expressed in the phloem and companion cells of the roots (not shown in **Figure 4**; Brady et al., 2007). ACO2 is also upregulated during flower opening and specifically during anther, stamen and petal development (van Es et al., 2018). ACO3 (AT1G12010; Type I) is highly expressed during embryogenesis and during further seed maturation. Furthermore, ACO3 is expressed in the root, more precisely in the phloem and companion cells (Brady et al., 2007). ACO4 (AT1G05010; Type I) is mostly expressed in vegetative tissue such as the cotyledons, the rosette, cauline leaves, sepals, and the petiole of senescing leaves. ACO5 (AT177330; Type III), is mainly expressed in the root (of both seedlings and adult plants), especially in the root apex and the root cap (Brady et al., 2007).

#### **Differential ACO expression during tomato fruit development and climacteric ripening**

**Figure 5** summarizes the differential and tissue-specific expression of the seven ACOs during tomato fruit development and climacteric ripening, based on the Tomato Expression Atlas (TEA). ACO1 (Solyc07g049530; Type I) is already expressed shortly after anthesis, and expression levels increase moderately throughout fruit development (system I). At the onset of ripening (system II), ACO1 expression increases strongly and correlates well with the autocatalytic rise in ethylene production (Blume and Grierson, 1997; Nakatsuka et al., 1998; Alexander and Grierson, 2002; Van de Poel et al., 2012). ACO1 expression appears to be strongest in the pericarp, septa and columella of orange to red fruit, matching the tissue-specific ACO in vitro activity reported by Van de Poel et al. (2014b). ACO2 (Solyc12g005940; Type I) is expressed at anthesis, however, expression drops to a basal level during further fruit development and ripening (Van de Poel et al., 2012). Expression of ACO3 (Solyc07g049550; Type I) is high during anthesis, but readily drops during initial fruit development. At the onset of ripening, ACO3 expression strongly increases again. These observations are contradictory to the qPCR data observed by Van de Poel et al. (2012), who showed that ACO3 expression declines after the breaker stage. ACO4 (Solyc02g081190; Type I) expression is high during initial fruit development (mainly pericarp tissue) and declines thereafter to a basal expression level. However, there is a temporal increase in ACO4 expression during the breaker stage, mainly in the columella and placenta tissue. Nakatsuka et al. (1998) showed that ACO4 expression increases during fruit ripening, but perhaps the use of degenerate primers in the Nakatsuka study could not discriminate between ACO4 and another ACO homolog. ACO5 (Solyc07g026650; Type II) expression increases slightly after anthesis and remains at a similar level during further fruit development and ripening. However, qPCR analysis by Van de Poel et al. (2012) indicated that ACO5 follows an expression pattern similar to that of ACO3. ACO6 (Solyc02g036350; Type I) is strongly expressed at anthesis, followed by a low expression during fruit development and a

temporal high expression during the breaker stage, followed by a gradual decline during further ripening. ACO6 expression is strongest in the pericarp tissue, which is also the tissue that showed the highest in vivo ethylene production (Van de Poel et al., 2014b). ACO7 (Solyc06g060070; Type III) is only basally expressed during fruit development and ripening. The ACO6 and ACO7 genes have not yet been characterized during tomato fruit development and ripening.

## Post-transcriptional Regulation of ACO

MicroRNAs (miRNAs) are often involved in the posttranscriptional regulation of diverse processes in plant growth

and development. MiRNAs constitute a class of regulatory, small, non-coding RNA molecules of 20 to 24 nucleotides long, which can intervene in gene expression by cleaving mRNA transcripts in a sequence specific way (Liu et al., 2017, 2018; Yu et al., 2017). MiR396b was identified as a cold-responsive miRNA in the cold hardy citrus variety Ponciferus trifoliata (Zhang et al., 2014). When the precursor of this miRNA (MIR396b) was overexpressed in lemon (Citrus lemon), it led to an increase in cold tolerance. Interestingly, overexpression of this miRNA reduced the expression of ACO compared to the wild type lemon. Zhang et al. (2016) showed that miR396b directs the cleavage of ACO transcripts, consequently inhibiting ethylene biosynthesis.

Recently, a second miRNA was identified, which affects the expression of ACO (Wang et al., 2018). In tomato, miR1917 directs the cleavage of a specific CTR4 splice variant in tomato leading to an altered ethylene response. Overexpression of miR1917 significantly enhanced the expression of ACS2, ACS4, ACO1 (Solyc07g049530) and ACO3 (Solyc07g049550), leading to specific ethylene response phenotypes such as the triple response in etiolated seedlings, an increase in epinastic curvature of leaf petioles, an increased pedicel abscission rate and an accelerated fruit ripening (Wang et al., 2018).

### Post-translational Regulation of ACO

An in vitro phosphorylation assay using protein extracts of pre- and post-climacteric apple fruit and an ectopically expressed His-tagged MdACO1 (MDP0000195885) hinted for the first time that ACO protein-protein interactions could exist (Dilley et al., 1995). It was observed that apple ACO1 could interact with unidentified proteins and resulted in the phosphorylation of these associated proteins (and possibly also ACO1 itself) in both pre-and post-climacteric samples (Dilley et al., 1995). Later Dilley et al. also speculated about a possible cysteine protease activity of ACO, unrelated to its role in catalyzing the conversion of ACC to ethylene. This hypothesis was based on sequence similarity between ACO

and a tomato cysteine protease (Matarasso et al., 2005), and the strongly conserved C28 as a key residue in the active site of this protease (Dilley et al., 2013). Using in silico predictions based on sequence similarity, other motifs for ACO protein-protein interactions, serine/threonine kinases, tyrosine kinases, and glycosylation were identified (Aitken, 1999; Dilley et al., 2013). Despite these predictions, only a few post-translational modifications of ACO have currently been experimentally observed.

Arabidopsis ACO2 (AT1g62380) was identified as a target in a broad proteomics screen that characterized stress-induced protein S-glutathionylation, suggesting that ACO2 is post-translationally modified by glutathionylation (Dixon et al., 2005). This was later confirmed in a dedicated S-glutathione pull-down assay with ACO2, where C63 was identified as the target residue for glutathionylation (Datta et al., 2015).The relevance of this post-translational glutathionylation and its involvement in ethylene production remains unknown.

Another thiol-residue modification of ACO was discovered in a large cysteine S-sulfhydration screen in Arabidopsis, which showed that ACO4 can get sulfhydrated (Aroca et al., 2015). This S-modification of ACO was recently confirmed in tomato by Jia et al. (2018), who demonstrated that H2S production increases in stomata upon a prolonged ethylene treatment, and that this leads to the S-sulfhydration of C60 (equivalent to C63 of AtACO2) of SlACO1 (Solyc07g049530) and SlACO2 (Solyc12g005940). This ACO sulfhydration resulted in a significant drop in ACO activity and consequently ethylene production, unmasking a direct crosstalk between ethylene and H2S production. It was argued that this sulfhydration was necessary to protect the plant from the detrimental effects of prolonged ethylene production and exposure (e.g., during senescence and programmed cell death) (Jia et al., 2018). At the moment, it remains unclear if this cysteine glutathionylation and/or sulfhydration leads to a different ACO activity or whether it is involved in protein interaction and/or protein stability. In general, post-translational modifications of the thiol-groups of cysteine residues by S-glutathionylation or S-sulfhydration are involved in the protection of proteins from irreversible oxidation or redox changes (Gao et al., 2009; Datta et al., 2015) or modulate proteinprotein interactions (Aroca et al., 2015). The exact function of ACO S-glutathionylation and S-sulfhydration remains to be further investigated.

In petunia, Tan et al. (2014) identified GRL2 (Green-like 2) as a novel interacting partner of ACO1 in a yeast two-hybrid screen using GRL2 as bait and a cDNA library of petals and leaves as prey. They also observed that the suppression of GRL2 expression resulted in an increased ethylene production of flowers, leading to an accelerated flower senescence. Therefore, GRL2 is proposed to serve as a negative regulator of ethylene production that can directly influence the activity of ACO1 (Tan et al., 2014).


#### ACO BIOTECHNOLOGY AND APPLICATIONS

Because ethylene plays a crucial role in many plant processes, including climacteric fruit ripening and senescence, excessive ethylene can lead to unwanted decay of plant-based food. Therefore, ethylene biosynthesis or signaling genes have been frequently targeted in biotechnological and transgenic approaches in order to increase the shelf life of plant-based food. Because ACO catalyzes the final step in the ethylene biosynthesis pathway, it is therefore an ideal candidate to target (instead of for example ACS or ethylene signaling components), because there are fewer risk of intervening in other pathways (e.g., ACC metabolism). **Table 3** presents an exhaustive list, although probably not exclusive, of several transgenic applications controlling ethylene production at the level of ACO for important agricultural crops.

One of the most rewarding applications of ACO-directed biotechnology in plants is the reduction of ethylene production during ripening and postharvest storage of climacteric fruit. ACO has been targeted using various antisense RNAi techniques to downregulate its expression and consequently decrease ethylene production, resulting in control over fruit ripening and postharvest storage. **Table 3** shows that this approach was successfully implemented for a wide variety of fruits: apple, lemon, kiwi, melon, papaya, pear, and tomato (Hamilton et al., 1990; Picton et al., 1993; English et al., 1995; Ayub et al., 1996; Guis et al., 1997; Bauchot et al., 1998; Ben-Amor et al., 1999; Dandekar et al., 2004; Silva et al., 2004; Xiong et al., 2005; Nuñez-Palenius et al., 2006; Gao et al., 2007; Lin et al., 2008; López-Gómez et al., 2009; Batra et al., 2010; Atkinson et al., 2011; Sekeli et al., 2014; Zhang et al., 2016). In all these studies, fruit ripening was delayed and shelf-life was prolonged. A similar approach has also been used to prolong the shelf life of vegetative tissue of vegetables, such as for example broccoli (Henzi et al., 1999; Gapper et al., 2005).

Transgenic approaches that silence ACO expression are also beneficial in floriculture. It has been shown that a reduced ACO expression resulted in a delay in flower senescence and flower abscission in petunia, carnation, and torenia (Savin et al., 1995; Aida et al., 1998; Huang et al., 2007; Tan et al., 2014). Besides controlling fruit ripening or flower senescence, ACOs have also been targeted in other ethylene-related processes. For example, tomato plants transformed with an ACO1 antisense construct showed delayed leaf senescence (John et al., 1995) and less

#### REFERENCES


epinasty during soil flooding (English et al., 1995). A mutation in cucumber ACO2 resulted in a mutant that only bears male flowers, suggesting that ACO2 plays an important role in sex determination in cucumber flowers (Chen et al., 2016).

Besides silencing ACO expression and reducing ethylene production, it can sometimes be desirable to boost ethylene production, and then an ACO overexpression construct is most suitable. In safflower, ACO1 overexpression was shown to stimulate the flavonoid biosynthesis pathway, which could be interesting for oilseed production (Tu et al., 2019). Overexpression of ACO1 in poplar (Populus tremula × tremuloides) caused a stimulation of cambial cell division, which in turn resulted in an increased xylem development and an inhibition of elongation growth, which are desirable traits for the wood industry (Love et al., 2009). Altogether, these transgenic examples show the potential of controlling ethylene production levels by targeting the ACO gene family. Perhaps new breeding technologies such as CRISPR/Cas9-mediated mutations in the ACO promotor or coding sequence could also lead to novel strategies to control ethylene production in plants.

### DATA AVAILABILITY

No datasets were generated or analyzed for this study.

### AUTHOR CONTRIBUTIONS

Both authors have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### FUNDING

This work was supported by the University of Leuven (STGBF/16/005 and C14/18/056) and the Research Foundation Flanders (SB/1S18717N, G092419N, and G0G0219N).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.00695/ full#supplementary-material



patterns dominant expression patterns. Science 318, 801–806. doi: 10.1126/ science.1146265



synthase gene expression. Plant Cell 17, 1205–1216. doi: 10.1105/tpc.105. 030775


biosynthesis during postclimacteric ripening. Plant Physiol. 160, 1498–1514. doi: 10.1104/pp.112.206086


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Houben and Van de Poel. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Molecular Analysis of Protein-Protein Interactions in the Ethylene Pathway in the Different Ethylene Receptor Subfamilies

*Mareike Berleth1 , Niklas Berleth2 , Alexander Minges1 , Sebastian Hänsch3 , Rebecca Corinna Burkart4 , Björn Stork2 , Yvonne Stahl4 , Stefanie Weidtkamp-Peters3 , Rüdiger Simon4 and Georg Groth1 \**

*1 Institute of Biochemical Plant Physiology, Heinrich Heine University, Düsseldorf, Germany, 2Institute of Molecular Medicine I, Heinrich Heine University, Düsseldorf, Germany, 3Center for Advanced Imaging, Heinrich Heine University, Düsseldorf, Germany, 4Institute for Developmental Genetics, Heinrich Heine University, Düsseldorf, Germany*

#### *Edited by:*

*Dominique Van Der Straeten, Ghent University, Belgium*

#### *Reviewed by:*

*Derek Gingerich, University of Wisconsin–Eau Claire, United States Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China*

*\*Correspondence:* 

*Georg Groth georg.groth@hhu.de; georg.groth@uni-duesseldorf.de*

#### *Specialty section:*

*This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science*

*Received: 25 March 2019 Accepted: 16 May 2019 Published: 07 June 2019*

#### *Citation:*

*Berleth M, Berleth N, Minges A, Hänsch S, Burkart RC, Stork B, Stahl Y, Weidtkamp-Peters S, Simon R and Groth G (2019) Molecular Analysis of Protein-Protein Interactions in the Ethylene Pathway in the Different Ethylene Receptor Subfamilies. Front. Plant Sci. 10:726. doi: 10.3389/fpls.2019.00726*

Signal perception and transmission of the plant hormone ethylene are mediated by a family of receptor histidine kinases located at the Golgi-ER network. Similar to bacterial and other plant receptor kinases, these receptors work as dimers or higher molecular weight oligomers at the membrane. Sequence analysis and functional studies of different isoforms suggest that the ethylene receptor family is classified into two subfamilies. In *Arabidopsis*, the type-I subfamily has two members (ETR1 and ERS1) and the type-II subfamily has three members (ETR2, ERS2, and EIN4). Whereas subfamily-I of the *Arabidopsis* receptors and their interactions with downstream elements in the ethylene pathway has been extensively studied in the past; related information on subfamily-II is sparse. In order to dissect the role of type-II receptors in the ethylene pathway and to decode processes associated with this receptor subfamily on a quantitative molecular level, we have applied biochemical and spectroscopic studies on purified recombinant receptors and downstream elements of the ethylene pathway. To this end, we have expressed purified ETR2 as a prototype of the type-II subfamily, ETR1 for the type-I subfamily and downstream ethylene pathway proteins CTR1 and EIN2. Functional folding of the purified receptors was demonstrated by CD spectroscopy and autokinase assays. Quantitative analysis of protein-protein interactions (PPIs) by microscale thermophoresis (MST) revealed that ETR2 has similar affinities for CTR1 and EIN2 as previously reported for the subfamily-I prototype ETR1 suggesting similar roles in PPI-mediated signal transfer for both subfamilies. We also used *in planta* fluorescence studies on transiently expressed proteins in *Nicotiana benthamiana* leaf cells to analyze homo- and heteromer formation of receptors. These studies show that type-II receptors as well as the type-I receptors form homo- and heteromeric complexes at these conditions. Notably, type-II receptor homomers and type-II:type-I heteromers are more stable than type-I homomers as indicated by their lower dissociation constants obtained in microscale thermophoresis studies. The enhanced stability of type-II complexes emphasizes the important role of type-II receptors in the ethylene pathway.

Keywords: ethylene receptor subfamilies, signaling, protein-protein interaction, microscale thermophoresis, fluorescence lifetime imaging microscopy

### INTRODUCTION

The gaseous plant hormone ethylene is decisive for many growth and developmental processes in plants, including fruit ripening, senescence, and the control of biotic and abiotic stress responses, such as pathogen defense and wounding (Kieber and Ecker, 1993; O'Donnell et al., 1996; Penninckx et al., 1998; Bleecker and Kende, 2000). Most of the current knowledge about ethylene biosynthesis and signal transduction has been obtained by genetic, physiological, and biochemical studies in the model plant *Arabidopsis thaliana*. Based on these studies, a family of mainly endoplasmic reticulum (ER)-membrane bound receptors was identified to catalyze the first step in all ethylene-regulated phenomena. These receptors in which their functional state form homo- or heterodimers at the ER membrane act as negative regulators of the ethylene-signaling pathway, following an inverse-agonist model in which ethylene-binding switches off the downstream signal transmission (Bleecker et al., 1988; Hua et al., 1995; Hua and Meyerowitz, 1998; Chen et al., 2002).

Sequence analysis and functional studies disclosed that the receptor family is classified into two subfamilies. In *Arabidopsis*, isoforms ETR1 and ERS1 form subfamily-I, whereas subfamily-II contains receptor isoforms ETR2, ERS2, and EIN4 (Hua et al., 1995, 1998; Sakai et al., 1998). Common to all isoforms is a modular structure known from bacterial sensor histidine kinases. In this case, the main elements are a transmembrane (TM) domain with an ethylene binding site at the amino (N)-terminus facing the ER lumen and a large cytosolic domain comprising of a GAF domain followed by a histidine kinase (HK) domain. Genetic and biochemical studies showed that the GAF domain contributes to the formation of the active dimer and that autophosphorylation activity of the kinase domain is inhibited upon ethylene binding. In addition to the GAF and HK domain receptor isoforms, ETR1, ETR2, and EIN4 carry a response regulator domain (RD) at their carboxy (C)-terminus (Chang et al., 1993; Chen et al., 2002; Voet van Vormizeele and Groth, 2008). Despite their similar overall structure, the individual isoforms contain major differences in these modules. For instance, type-II receptors have an additional fourth transmembrane helix. The function of this additional element is still not clear, although it might function as a targeting signal (Chen et al., 2007). In addition, essential residues for histidine kinase activity are missing in this subfamily. However, *in vitro* Ser/Thr kinase activity was demonstrated for these isoforms albeit autokinase activity seems to play only a minor role in ethylene signaling (Wang et al., 2003; Moussatche and Klee, 2004).

Despite these structural dissimilarities, the general function of the different receptor isoforms and response to the plant hormone is highly redundant. Nonetheless, functional specificity among different isoforms has been described, although the underlying molecular mechanisms are still not fully understood (Binder et al., 2006; Wuriyanghan et al., 2009). But even with the exact signal output of the receptors unknown, previous studies have clearly shown that complex formation of receptors with the Raf-like kinase CONSTITUTIVE TRIPLE RESPONSE 1 (CTR1) and the integral membrane protein ETHYLENE INSENSITIVE 2 (EIN2) is an integral part of the ethylene signaling network in the response to the plant hormone. In this process, CTR1 was shown to phosphorylate EIN2 and that interaction of CTR1 with the receptors is critical for CTR1 kinase activity. The presence of ethylene leads to inactivation of the receptors, thereby inactivating CTR1 in turn resulting in dephosphorylation of EIN2. As a consequence, the C-terminal part of EIN2 (EIN2479-1294) is cleaved off by a so far unknown mechanism and translocates to the nucleus. Here, it directly or indirectly stabilizes transcription factors EIN3/EIL1, which activate the transcription of ethylene response genes (Gao et al., 2003; Ju et al., 2012; Qiao et al., 2012; Wen et al., 2012). Remarkably, another mechanism affecting plant ethylene response was shown for the cleaved off EIN2 C-terminus as this digest inhibits the translation of *EBF1/2* mRNA in the cytosol thereby preventing EBF1/2-dependent degradation of the EIN3 master transcription factor (An et al., 2010).

In the past, various approaches analyzing the interaction of type-I receptors and downstream signaling components have been developed. For instance, in our lab, interaction of the type-I receptor ETR1 with EIN2479-1294 was demonstrated by *in vivo* FRET-studies and quantified by *in vitro* tryptophan fluorescence quench analysis (Bisson et al., 2009). Moreover, recent studies in our lab highlighted that the conserved nuclear localization signal (NLS) sequence of EIN2 significantly contributes to the EIN2-receptor interaction and in the form of a synthetic octapeptide (NOP1) delays fruit ripening and flower senescence (Bisson et al., 2016; Kessenbrock et al., 2017; Milić et al., 2018; Hoppen et al., 2019). In contrast, related information on ethylene receptor subfamily-II is still sparse.

In the work presented here, we performed quantitative biochemical and spectroscopic studies on purified receptor preparations and downstream elements in order to elucidate the protein-protein interaction (PPI) landscape of both receptor subfamilies. To this end, we established expression and purification for the *Arabidopsis* ETR2 as a prototype of the type-II subfamily. Our studies indicate similar roles in PPI-mediated signal transduction for both receptor subfamilies. To that, we visualized homo- and heteromer formation of type-I and type-II receptors by *in planta* fluorescence lifetime analysis (FRET-FLIM) and anisotropy and quantified these interactions by microscale thermophoresis (MST) on purified recombinant proteins*.* In the end, our study demonstrates the enhanced stability of type-II receptor complexes compared to complexes consisting of type-I isoforms only stressing an important role of type-II receptors for ethylene signal transduction.

### MATERIALS AND METHODS

#### Production of Recombinant *Arabidopsis* Proteins for *in vitro* Interaction Studies

Codon optimized cDNA encoding full-length AtETR2 (UniProt ID: Q0WPQ2) was purchased from GenScript USA according to the published sequence (NCBI ID: NM\_113216.3). The cDNA sequence was flanked by a 5′ *SmaI* recognition site and 3′ *XhoI* recognition site. Expression vector pGEX-4T-1 (GE Healthcare Life Sciences) and synthetic DNA were digested with *SmaI* and *XhoI* and ligated. In the resulting plasmid pGEX-4T-1\_AtETR2, the region coding for a thrombin cleavage site was changed to a region coding for a tobacco etch virus (TEV) protease cleavage site (ENLYFQG). To this end, a PCR-based approach with 5′-phosphorylated primers was used (Follo and Isidoro, 2008). Furthermore, a 10 × -His-tag was C-terminally fused to ETR2 to facilitate protein purification. The construct was cloned in two successive PCR reactions (Follo and Isidoro, 2008). The resulting plasmid was verified by sequencing and termed pGEX-4T-1\_TEV\_ETR2\_H10. The DNA fragment encoding for the transmembrane domain of ETR1 (aa 1-157) was cloned into expression vector pET16b (Novagen), and base pairs coding for cysteines at positions 4 and 6 in the protein were changed to encode serine. The resulting plasmid was termed ETR1-TMC4SC6S. Full-length cDNA sequence encoding for AtCTR1 (UniProt ID: Q05609) was purchased from GenScript USA according to the published sequence (NCBI ID: NM\_120454.4). The DNA fragment was cloned into expression vector pET30a (Novagen) by using SLIC, and the coding region for CTR1 was extended by a cyan fluorescent protein termed mCerulean (Li and Elledge, 2007; Rizzo et al., 2007). The resulting plasmid was verified by sequencing and termed pET30a\_AtCTR1\_mCerulean. Primer sequences used for cloning are listed in **Supplementary Table S1**.

#### Molecular Cloning of AtERS1 and AtERS2 Fusions for *in planta* Interaction Studies

Expression vectors encoding for receptor proteins AtERS1 (UniProt ID: Q38846) and AtERS2 (UniProt ID: P93825) carrying a C-terminal RFP-label were kindly provided by Klaus Harter (Grefen et al., 2008). TOPO-AtERS1 and TOPO-AtERS2 entry clones were prepared by using pENTR Directional TOPO Cloning Kit (Thermo Fisher Scientific) following the manufacturer's instructions. For transient expression in *Nicotiana benthamiana,* expression vectors encoding ERS1 and ERS2 were prepared *via* Gateway LR-reaction (Thermo Fisher Scientific) using C-terminal mVenus- and mCherry-tagged destination vectors pABindmVenus and pABindmCherry, which are based on vector pMDC7 (Curtis and Grossniklaus, 2003). Expression vector pABindmVmC-ERS2 was cloned using pABindFRET as backbone by Gibson Assembly with amplification of four fragments thereby substituting fluorescent protein (FP) GFP by mVenus (Kremers et al., 2006; Gibson et al., 2009; Bleckmann et al., 2010). The resulting expression vector encodes ERS2 and contains a C-terminal mVenus-mCherry fusion protein. **Supplementary Table S1** gives an overview of primers used for cloning.

#### Transient Expression of AtERS1 and AtERS2 in *N. benthamiana*

*Agrobacterium tumefaciens* strain GV3101 pMP90 was transformed with expression vectors (pABindm Venus-ERS2, pABindmCherry-ERS2, pABindmCherry-ERS1, pABindmCherry-BTI2, pABindmVmC-ERS2) as well as vector pER-Rb encoding for an ER-marker protein tagged to mCherry (Koncz and Schell, 1986; Nelson et al., 2007). To reduce gene silencing *in planta*, expression vectors were co-transfected with silencing suppressor p19 (Qu and Morris, 2002). Cells were cultured in 2YT medium [1.6% (w/v) peptone, 1% (w/v) yeast extract, and 0.5% (w/v) NaCl], precipitated, and resuspended in AS medium [5% (w/v) sucrose, 5 mM MgSO4, 5 mM glucose, 0.01% (v/v) Silwet L77, 450 μM acetosyringone]. *A. tumefaciens* cells containing mVenus- and mCherry-tagged expression vectors were mixed at 1:1 ratio to an optical density at 600 nm (OD600) of 0.2 and infiltrated in 4-week-old *N. benthamiana* leaves. Transient gene expression was induced 72 –96 h after infiltration by 20 μM β-estradiol and 0.1% (v/v) Tween 20, and protein expression was analyzed from 16 to 25 h.

#### Expression of Recombinant *Arabidopsis* Receptor AtETR2 and Protein Kinase AtCTR1 in *Escherichia coli*

The expression vector encoding for recombinant AtETR2 was transformed into chemically competent *E. coli* C43 (DE3) (Lucigen Corporation) cells. Transformants were precultured in 2YT medium containing 100 μg/ml ampicillin at 30°C for 16 h. Preculture was diluted to an OD600 of 0.1 in 500 ml terrific broth (TB) medium (12 g/L peptone, 24 g/L yeast extract, 5 g/L glycerol, 1.8 g/L KH2PO4, 19.8 g/L K2HPO4) containing 100 μg/ml ampicillin and 2% ethanol (v/v). Cells were incubated at 30°C with shaking at 180 rpm. At OD600 = 0.4 temperature was reduced to 16°C. The bacteria were grown to OD600 = 0.6 and heterologous protein expression was induced with 0.5 mM isopropyl β-d-1-thiogalactopyranoside (IPTG). Cells were further grown for 4 h, harvested by centrifugation, flash-frozen in liquid nitrogen, and stored at −20°C. AtCTR1 was expressed in *E. coli* BL21 (DE3) cells. Plasmid pRARE (Novagen) was co-transformed carrying the genes for essential tRNAs encoded by rarely used codons in *E. coli*. Cells were grown at 30°C in 500 ml 2YT medium containing 25 μg/ml kanamycin and 2% ethanol (v/v). Cells were grown to an OD600 = 0.4, and temperature was reduced to 16°C. Protein expression was induced at OD600 = 0.6 by addition of 0.2 mM IPTG. Cells were further incubated for 20 h at 16°C, harvested by centrifugation, flash-frozen in liquid nitrogen, and stored at −20°C. Protein expression of recombinant proteins was analyzed by SDS-PAGE followed by immunoblotting (Laemmli, 1970; Towbin et al., 1979).

### Purification of Recombinant AtETR2 and AtCTR1

Protein purification steps were performed on ice or at 4°C, if not stated otherwise. Resulting AtETR2 cell pellet was resuspended in PBS lysis buffer [PBS pH 8.0, 10% (w/v) glycerol, 1 mM DTT, 0.002% (w/v) phenylmethylsulfonyl fluoride (PMSF) and 10 mg/L DNaseI (PanReac AppliChem), 5 ml PBS lysis buffer per 1 g cells]. Cells were broken with Constants Cell Disruption System (Constant Systems) at 2.4 kbar and 5°C. Cell lysate was centrifuged for 30 min at 14,000 × *g* to remove cell debris and inclusion bodies. The supernatant was centrifuged again for 30 min at 40,000 × *g*. The resulting membrane pellet was resuspended in PBS lysis buffer and isolation of cell membranes were achieved by centrifugation at 34,000 × *g* for 30 min. Membrane pellets were flash-frozen in liquid nitrogen and stored at −80°C. For protein solubilization, membranes were resuspended in buffer S [50 mM Tris/HCl pH 7.8, 200 mM NaCl, 1.2% (w/v) FosCholine-16, 2.5 mM DTT, 0.002% (w/v) PMSF] and stirred at 700 rpm for 1 h. Membrane fragments were removed by ultracentrifugation at 230,000 × *g* for 30 min. The resulting supernatant was loaded to a 5 ml Ni-NTA HisTrap FF column (GE Healthcare Life Sciences), equilibrated in buffer R [50 mM Tris/HCl pH 7.8, 200 mM NaCl, 2.5 mM DTT, 0.015% (w/v) FosCholine-16]. The column was washed with 10 column volumes (CV) of buffer R, followed by 20 CV buffer R-ATP (buffer R with additional 50 mM KCl, 20 mM MgCl2, 10 mM ATP). The column was washed again with 10 CV buffer R, followed by 10 CV buffer R containing 50 mM imidazole. Finally, the receptor AtETR2 was eluted with 250 mM imidazole. To purify His-tagged protein kinase AtCTR1, resulting cell pellet was resuspended in lysis buffer C [50 mM HEPES/NaOH pH 7.6, 300 mM NaCl, 5% (w/v) glycerol, 5 ml lysis buffer C per 1 g cells]. About 10 mg/L DNaseI and 1 × EDTA-free cOmplete Protease Inhibitor Cocktail (Roche) were added to cells prior to cell disruption. Cells were disrupted by passing through a pre-cooled French pressure cell at 12,000 psi (1 psi = 6.9 kPa). The cell lysate was ultracentrifuged at 230,000 × *g* for 60 min. The resulting supernatant was loaded onto a 5 ml Ni-NTA HisTrap HP column (GE Healthcare Life Sciences), equilibrated in buffer C-P [buffer C with 0.002% (w/v) PMSF]. The column was washed with 10 CV of buffer C-P, followed by 20 CV buffer C-ATP [buffer C with additional 50 mM KCl, 20 mM MgCl2, 10 mM ATP, 0.002% (w/v) PMSF]. The column was washed with 10 CV buffer C-P containing 50 mM imidazole and 100 mM imidazole, respectively. AtCTR1 was eluted with 500 mM imidazole in buffer C-P. Purified proteins were concentrated in a 50 kDa Amicon Ultra-15 concentrator (EDM Millipore). Buffer was changed by a desalting step on a PD-10 or PD MiniTrap G-25 column (GE Healthcare Life Sciences) depending on subsequent applications.

#### Circular Dichroism Spectroscopy of Recombinant *Arabidopsis* Receptor Proteins

AtETR1 and AtETR2 were expressed and purified as described in this article (**Figure 1, Supplementary Figure S1A**) and in Milić et al., 2018. Purified receptors were characterized by circular dichroism (CD) spectroscopy. For the far-UV spectra, purified *Arabidopsis* receptors were measured at a final concentration of 0.2 mg/ml in CD buffer (10 mM K2HPO4/KH2PO4 pH 8.0). Therefore, buffer was exchanged to CD buffer using a PD MiniTrap G-25 column (GE Healthcare Life Sciences) and protein samples were ultracentrifuged at 230,000 × *g* for 30 min. Protein and FosCholine-16 concentrations were determined by a Direct Detect infrared spectrometer (EMD Millipore). In blank samples, FosCholine-16 concentrations were adjusted to correspond to the final detergent concentration in the protein samples. For a detailed protocol for protein preparation, see Kessenbrock and Groth, 2017. CD spectra were recorded by a Jasco J715 spectropolarimeter (Jasco GmbH) using a cylindrical quartz cuvette with 1-mm-path-length (Hellma Analytics). Each protein spectrum was measured from 260 to 180 nm at room temperature and represents an average of 10 continuous scans recorded with a bandwidth of 1 nm at 50 nm/min. Secondary structure content of the purified receptors was calculated using reference protein set SMP50 in programs CDSSTR and CONTINLL from the CDpro software package (see **Figure 2, Supplementary Figures S1E**,**F**; Provencher and Glöckner, 1981; Johnson, 1999; Sreerama and Woody, 2000).

#### *In vitro* Autokinase Activity Assay of *Arabidopsis* Receptor Proteins

Autokinase activity of AtETR1 and AtETR2 was assessed by an *in vitro* kinase assay. To this end, full-length receptors were expressed and purified as previously described either with (**Figure 3**) or without the additional ATP purification step (see **Supplementary Figures S1B–D**). Purified proteins (1 mg) were incubated in kinase assay reaction buffer [50 mM Tris/HCl, pH 7.5, 0.1 mM EGTA, 0.1 mM DTT, 10 mM Mg(CH3COO)2] supplemented with 0.1 mM [γ-32P]ATP (Hartmann Analytic) for 30 min at 37°C. Protein denaturation was obtained by the addition of 40 mM DTT and 2% (w/v) SDS for 30 min at 60°C prior to *in vitro* phosphorylation. The kinase reactions were stopped by the addition of SDS sample buffer, and the samples were subjected to SDS-PAGE. After Coomassie staining, the gel was dried and autoradiography was performed for 6 days.

#### Fluorescent Labeling for Microscale Thermophoresis Studies

Protein-protein interactions were analyzed by microscale thermophoresis (MST) (Duhr and Braun, 2006; Jerabek-Willemsen et al., 2011). Therefore, recombinant proteins were labeled with Alexa Fluor 488 succinimidyl-ester (Thermo Fisher Scientific). For this purpose, buffer of purified and concentrated AtETR2 and AtCTR1 samples were exchanged on a desalting PD-10 column. Samples were concentrated again resulting in 500-μl protein sample of AtCTR1 in labeling buffer L (50 mM K2HPO4/KH2PO4 pH 8.0, 300 mM NaCl) and AtETR2 in buffer L-R [buffer L with 0.015% (w/v) FosCholine-16]. Recombinant AtEIN2479-1294 was expressed and purified as previously described (Bisson et al., 2016) and buffer was exchanged to labeling buffer L-E [buffer L with 6% (w/v) glycerol, 10 mM EGTA]. Expression, purification, and labeling of AtETR1 were performed as described in Milić et al., 2018. Alexa Fluor 488 succinimidylester was applied to each protein in 2.5-fold excess and incubated while mixing slightly for 30 min in the dark at ambient temperature. Buffer of labeled proteins was exchanged for AtETR2 to MST buffer 1 [50 mM HEPES/NaOH pH 7.8, 300 mM NaCl, 5% (w/v) glycerol, 0.015% (w/v) FosCholine-16], for AtETR1 to MST buffer 2 [50 mM Tris/HCl pH 7.8, 300 mM

AtETR2 was calculated and adjusted to molar extinction (∆ɛ) considering molecular weight and protein concentration of AtETR2. (B) Secondary structure content was calculated by CONTINLL (solid line) and CDSSTR (dashed line) from the CDpro software package.

NaCl, 5% (w/v) glycerol, 0.015% (w/v) FosCholine-16], and for AtCTR1 and AtEIN2479-1294 to MST buffer 3 [50 mM Tris/ HCl pH 7.8, 300 mM NaCl, 5% (w/v) glycerol]. The protein samples were centrifuged at 230,000 × *g*, for 30 min, and at 4°C. Protein samples solutions were adjusted to a final glycerol concentration of 20% (w/v), flash-frozen in liquid nitrogen, and stored at −80°C.

#### Quantitative Interaction Studies by Microscale Thermophoresis

Protein-protein interactions were analyzed by microscale thermophoresis. Therefore, the receptors AtETR1 and AtETR2 as well as soluble proteins AtEIN2479-1294 and AtCTR1 were purified and labeled as previously described. Experiments were performed on a Monolith NT.115 Blue/Green (NanoTemper Technologies) in three independent replicates, whereas for negative controls, measurements were done in duplicates. If not stated otherwise, measurements were performed in standard glass capillaries (NanoTemper Technologies). For binding studies of AtEIN2479-1294 (in MST buffer 3) to AtETR2 (in MST buffer 1), proteins were used as follows: 20 nM of labeled AtETR2, 2 μM as the highest, and 0.98 nM as the lowest AtEIN2479-1294 concentration. Measurements were performed at 20% MST power. As a negative control, AtEIN2479-1294 was heated to 95°C for 5 min, diluted in MST buffer 3 and mixed with 40 mM DTT and 4% (v/v) SDS. Measurements were carried out as described before. Furthermore, interaction of AtETR2 to AtEIN2479-1294 was analyzed using 75 nM labeled AtEIN2479-1294 and measured with 4 μM as the highest and 1.95 nM as the lowest AtETR2 concentration. Protein samples were incubated

at ambient temperature and measured at 60% MST power. For quantification of receptor AtETR2 and AtETR1 binding to CTR1 (in MST buffer 3), receptors AtETR2 and AtETR1 (in MST buffer 2) were measured with 4 μM as highest and 0.98 nM as lowest receptor concentration. AtETR1 was mixed in a 1:1 volume ratio with labeled AtCTR1 (50 nM final concentration) and measured at 60% MST power. For AtETR2 binding to AtCTR1, labeled AtCTR1 was used in a final concentration of 20 nM. The protein mixture was incubated for 10 min at ambient temperature, transferred into premium glass capillaries, and measured at 60% MST power. As a negative control, the protein mixture was incubated directly after the 10 min incubation step with 4% (v/v) SDS and 80 mM DTT for 5 min in the dark at RT, resulting in the same protein concentrations as before mentioned. For quantification of receptor-receptor interactions, labeled receptors were used at a final concentration of 40 nM. Labeled receptors were mixed in a 1:1 ratio with non-labeled AtETR1, AtETR2 (4 μM as the highest and 0.49 nM as the lowest receptor concentration) or ETR1-TMC4SC6S (MST buffer 2, 16 μM as the highest and 0.98 nM as the lowest concentration), thereby adjusting the detergent concentration to 0.0075% (w/v) FosCholine-16. Sample mixtures containing AtETR2 (labeled as well as non-labeled) were initially incubated for 10 min at RT and transferred into premium glass capillaries. All receptor-receptor binding studies were performed at 60% MST power. As negative controls, denaturation buffer (4% (v/v) SDS, 40 mM DTT in MST buffer 1 or 2, respectively) was added to the receptors. Samples were incubated for 5 min at RT in the dark and MST measurements were carried out as described for the native receptor proteins. All dissociation constants (*K*d) were calculated to a binding model assuming a 1:1 stoichiometry per binding partner.

#### Confocal Fluorescence Microscopy

*N. benthamiana* epidermis cells were imaged for protein expression and protein localization using a LSM 780 laser-scanning confocal microscope (Carl Zeiss GmbH) using a C-Apochromat 40×/1.2 W Corr M27 objective with a zoom factor of 4. The pinhole was set to 1 Airy Unit (AU). The following settings were used: 488/561 beam splitter with 488 nm excitation for mVenus and 561 nm excitation for mCherry. Fluorescence was detected between 508–552 nm for mVenus and 570–624 nm for mCherry by a GaAsP detector. In combination with FM4-64 the detection wavelength of mVenus was between 490 and 552 nm. FM4-64 was detected between 562 and 626 nm. The laser strength was adjusted between 2 and 9% for mVenus, with a gain of 700–900, whereas for mCherry, a laser strength between 1 and 6% with a gain of 750–980 was used. Staining of plasma membrane was carried out using 10 μM FM4-64 which was infiltrated 20 min before image acquisition. Images were recorded and processed using Fiji software (Schindelin et al., 2012).

#### Fluorescence Lifetime Imaging Microscopy in *N. benthamiana* Leaves

Fluorescence lifetime imaging microscopy (FLIM) was performed using a LSM 780 confocal laser-scanning microscope additionally equipped with a single-photon counting device enabling picosecond time resolution (PicoQuant Hydra Harp 400). mVenus fluorescence was excited at 485 nm with a rate of 32 MHz with a linearly polarized pulsed diode laser (LDH-D-C-485, PicoQuant). The pinhole was set to 1 Airy Unit (AU). Excitation power was adjusted to 1 μW at the objective C-Apochromat (40×/1.2 W Corr M27) prior to measurements. mCherry was excited at 561 nm by a continuous wave laser with a laser strength of 0.1%. Emitted light of mVenus was separated into its parallel and perpendicular polarization. mVenus fluorescence was detected by Tau-SPADs (PicoQuant) in a narrow range of its emission spectrum (band-pass filter: 534/30, AHF). mCherry fluorescence detection was set by the band-pass filter (HC 607/70, AHF). Images were acquired with 256 × 256 pixel, zoom factor 4, 12.61 μs pixel dwell time, and a resolution of 210 nm/pixel. A series of 80 frames were merged into one image and further analyzed using SymphoTime64 (PicoQuant).

#### Fluorescence Lifetime and Anisotropy Analysis

The fluorescence lifetime of mVenus was analyzed using the software tool SymPhoTime 64, version 2.3 (PicoQuant, Berlin, Germany). Due to low excitation power to prevent photobleaching during image acquisition and the small pixel size to gain spatial resolution, the number of photons per pixel was still low after merging of frames. An individual ROI for every dataset was generated ensuring that only pixels with a minimum number of 100 photons contributed to the decay histogram. In the donor-only case, a mono-exponential fit model, including background contribution and shifting of the instrument response function, was sufficient to describe the decay histogram of mVenus fluorescence and extract the fluorescence lifetime. In the case of FRET an additional exponent was used to describe the decay and a mean fluorescence lifetime was extracted from the resulting fit.

The steady-state anisotropy r is given by *<sup>r</sup> I GI I GI par per par per* <sup>=</sup> - <sup>+</sup> <sup>2</sup> . *Ipar* and *Iper* are the average fluorescence count rates per pixel with the emission parallel (*Ipar*) and perpendicular (*Iper*) to the excitation polarization direction (Gauthier et al., 2001; Sarkar et al., 2009). Orientation sensitivity differences of the detection system were corrected by determining the *G*-factor by calibration measurements using Rhodamine110. Data were statistically evaluated with GraphPad Prism using Student's *t*-test with Welch correction and Mann-Whitney test, respectively.

#### RESULTS

#### Heterologous Expression and Purification of *A. thaliana* Type-II Receptor ETR2

Previous studies have revealed the direct interaction of *A. thaliana* receptor ETR1 (AtETR1) with the C-terminus of EIN2 (aa 479-1294) *in planta.* These *in vivo* results were supported by tryptophan fluorescence quench studies on purified proteins which disclosed that receptors interact with the EIN2 downstream ethylene signaling protein with high affinity. Along the same lines, *in planta* FRET studies have demonstrated the interaction of type-II receptors and EIN2 (Bisson et al., 2009; Bisson and Groth, 2010). However, these studies did not reveal binding mode and binding affinities of the EIN2 C-terminus with type-II receptors. To clarify whether type-II and type-I receptors interact with EIN2 in a similar way and affinity as the type-I receptor subfamily, we heterologously expressed and purified ethylene receptor ETR2 from *A. thaliana* (AtETR2) as prototype of the type-II subfamily. To this end, AtETR2 expression plasmid pGEX-4T-1\_TEV\_ETR2\_H10 was transformed into *E. coli* strain C43 (DE3) which has been successfully used for expression of AtETR1 (Voet van Vormizeele and Groth, 2008). Bacterial cells harboring the AtETR2 expression plasmid show an increased expression of a protein band with an apparent molecular weight of 120 kDa up to 4 h after induction, whereas 5 h after induction the corresponding protein band appear faded suggesting degradation of the over-expressed receptor by bacterial proteases (**Figure 1A**, line 5). On the basis of the expression studies showing maximum protein production 4 h after induction, cells were harvested 4 h post-induction and solubilized from the host membranes by the detergent FosCholine-16. Solubilized AtETR2 receptors were purified by metal-chelate affinity chromatography on Ni-NTA agarose (GE Healthcare). High purity of the resulting receptor preparations is indicated by the single band observed on the SDS protein gel (**Figure 1B**). Identity of the purified protein band with AtETR2 was confirmed by immunoblotting with an anti-His Tag antibody (**Figure 1C**).

#### Analysis of Secondary Structure and Functional Folding of AtETR2

The folding and secondary structure of AtETR2 was analyzed by circular dichroism (CD) spectroscopy. The CD spectra has two minima at 208 and 221 nm and an isosbestic point at 201 nm (**Figure 2**), indicating a predominately α-helical structure of the purified AtETR2. Secondary structure content was quantified by CONTINLL and CDSSTR—two different algorithms for secondary structure assignment—at 48–50% α-helix and 9–10% β–sheet demonstrating that the recombinant protein adopts a well-folded structure and is likely to reflect a native confirmation of the *Arabidopsis* type-II ethylene receptor ETR2. Additionally, AtETR2 functionality was probed by an *in vitro* radiolabeling autokinase assay. For that, AtETR2 was purified without ATP pre-incubation. Purified protein preparations were then incubated with [γ-32P]ATP. Autoradiography of the SDS gel loaded with samples from the kinase assay revealed incorporation of 32P for those samples incubated in the presence of magnesium—an essential cofactor for histidine kinase activity—but not for those containing chemically denatured AtETR2 (**Figure 3**). These results further confirm functional folding of purified AtETR2. Notably, the purified type-II receptor shows somewhat reduced kinase activity in the presence of manganese (**Supplementary Figure S1D**), which is in line with results of previous studies using the ETR2∆GAF mutant (Moussatche and Klee, 2004). To allow for comparison to the type-I receptor subfamily in terms of protein affinities and protein activities, similar buffer and detergent conditions were applied for purification of AtETR1. The corresponding autokinase assay (**Supplementary Figure S1B**) supports that both receptor subfamilies show similar phosphorylation activities at these conditions and once again emphasizes the purification of the full-length receptors in a functional, catalytically active state.

#### Microscale Interaction Studies of AtETR2 and Downstream Signaling Components AtEIN2479-1294 and CTR1

Protein-protein interactions of purified receptors with downstream ethylene signaling components EIN2 and CTR1 were monitored and quantified by microscale thermophoresis (MST). This biophysical technique is based on the motion of molecules in a temperature gradient and strongly depends on the charge, hydration-shell and size of the moving molecules. At least one of these qualities typically changes upon complex formation. Hence, thermophoresis provides a sensitive and reliable method to analyze and to quantify protein-protein interactions (Duhr and Braun, 2006; Wienken et al., 2010; Jerabek-Willemsen et al., 2011). For a start, we applied this technique to quantify the interaction of the soluble EIN2 C-terminus (AtEIN2479-1294) with the type-II receptor prototype AtETR2. Both recombinant proteins were purified as previously described (see **Figure 1A, Supplementary Figure S2B**; Bisson et al., 2016). Addition of EIN2479-1294 to labeled AtETR2 shows clear changes of thermophoresis and thermophoretic signals obtained with increasing EIN2479-1294 concentrations follow a clear binding curve. From this binding curve, an apparent dissociation constant (*K*d) of 161(30) nM was obtained which is indicative of a tight and highly specific interaction of AtEIN2479-1294 with the AtETR2 receptor. Similarly, clear changes of the thermophoretic signal were observed upon addition of AtETR2 to labeled EIN2479-1294. The corresponding binding curve shows an apparent *K*d of 147(15) nM. Any potential effect of the fluorescence probe Alexa Fluor 488 which is positioned at different sites in the two complementary titration set-ups on the interaction, integrity or stability of any of the two binding partner can be ruled out as both binding studies show almost the same low nanomolar *K*d value (**Figure 4**). The tight and highly specific interaction of AtETR2-AtEIN2479-1294 is further supported in titration studies using denatured AtEIN2479-1294 as negative control. Here, no interaction between both binding partners was detectable. Like EIN2 the Raf-like protein kinase CTR1 is another signaling element downstream of the *Arabidopsis* ethylene receptor family. Recombinant AtCTR1 required for *in vitro* binding studies with receptors AtETR1 and AtETR2 respectively, was purified according to the methods section (see **Supplementary Figure S2A**). As described before for EIN2, protein-protein interactions of receptor proteins and AtCTR1 were analyzed by MST. To this end, AtCTR1 was labeled with the fluorescent dye Alexa 488 and was mixed with increasing concentrations of AtETR1 or AtETR2 until saturation. Binding of receptors to AtCTR1 detected as changes in thermophoresis correspond to dissociation constants (*K*d) of 169(15) nM for the AtCTR1-AtETR1 interaction and of 165(20) nM for the AtCTR1-AtETR2 complex. Again, these numbers indicate a tight and highly specific interaction. Furthermore, it should be noted that both receptor subfamilies show similar affinities for AtCTR1 (**Figure 5**).

#### *In vitro* Quantification of Type-I-Type-II Receptor Interactions

To analyze the mode of receptor-receptor interactions and to quantify binding affinities, we studied protein-protein interactions within receptor subfamilies by microscale thermophoresis. To this end, we analyzed homomeric and heteromeric receptor complexes of AtETR1 and AtETR2. For that, either AtETR1 or AtETR2 was labeled with Alexa Fluor 488, and thermophoresis was recorded after the addition of the corresponding binding partner. Binding affinity and dissociation constants indicate strong binding for both receptor subfamily homomers, AtETR1-AtETR1 and AtETR2-AtETR2 (**Figure 6A**). However, a three-fold higher affinity of the type-II AtETR2-AtETR2 homomer [*K*d of 96(9) nM] was detected in comparison to the type-I AtETR1-AtETR1 homomer [*K*d of 326(18) nM]. In addition, our binding studies revealed that type-I and type-II receptors can also form tight heteromeric complexes with binding constants of 177(18) nM (AtETR1-AtETR2) and 217(14) nM (AtETR2-AtETR1), respectively (**Figure 6B, Supplementary Figures S3A,B**). Notably, type-II:type-I heteromeric complexes seem to be more stable than type-I homomers. No interactions were detected in the related binding studies when chemically denatured receptors were used (**Figure 6A**). Previous studies suggested that dimer and higher complex formations in the ethylene receptor family are substantially mediated by their GAF-domains (Grefen et al., 2008; Mayerhofer et al., 2015). However, the role of the receptor

FIGURE 4 | Interaction studies of *Arabidopsis* ETR2 and EIN2 by MST. Dissociation constants of the interactions were obtained from the related binding curves. Titration of unlabeled AtEIN2479-1294 to AtETR2 (●) is described by a dissociation constant (*K*d) of 161(30) nM. Chemically and thermally denatured AtEIN2479-1294 shows no binding event to AtETR2 (▲). Binding of unlabeled AtETR2 to AtEIN2479-129 is represented by a *K*d value of 147(15) nM (○). All data represent the mean (SD) of three independent measurements (●, ○) and duplicates (▲), respectively.

FIGURE 5 | MST based protein-protein interaction assay between AtCTR1 and receptor proteins AtETR1 and AtETR2. Binding of AtETR1 to fluorescently labeled AtCTR1 measured by MST resulted in a *K*d value of 169(15) nM (○). For AtCTR1-AtETR2 complex formation a *K*d value of 165(20) nM was obtained (●). As negative control, titration of chemically denatured AtCTR1 with AtETR2 is shown. Here, no binding event was observed (▲). Data are given as the mean (SD) of independent triplicates (●, ○) and duplicates (▲), respectively.

transmembrane-domain in complex formation was not resolved in these studies. To address this issue, we analyzed the interaction between the isolated AtETR1 transmembrane domain (ETR1-TM) and full-length receptors AtETR1 and AtETR2 *via* MST. To eliminate any stabilizing effect by disulfide bond formation (Schaller et al., 1995), cysteines in ETR1-TM were substituted to serines (C4SC6S). In the related titration experiments, tight binding of ETR1-TMC4SC6S to both receptor subfamily representatives (AtETR1 and AtETR2) was observed. In line with previously observed heteromeric interactions of full-length receptors, the affinity for AtETR2 was higher than for AtETR1 (**Figure 6B, Supplementary Figures S3C,D**). Taken together, the binding studies with the isolated transmembrane domain (AtETR1-TMC4SC6S) as well as with full-length receptors emphasize that both receptor subtypes interact in homomeric and heteromeric complexes in a selective and specific manner. Moreover, the higher stability of type-II receptor homo- and heteromers highlights the importance of this subfamily to form receptor dimers or higher order oligomers in ethylene signaling.

Supplementary Figure S3. All data represent the mean (SD) of three

independent measurements.

#### *In planta* Detection of Receptor-Receptor Interactions *via* FRET-FLIM Microscopy

To analyze ethylene type-II:type-I receptor complex formations *in vivo*, we performed fluorescence lifetime imaging microscopy (FLIM). FLIM is used in plant cells to study molecular interactions and detects fluorescence resonance energy transfer (FRET) between two fluorescent-tagged proteins in close proximity (Gadella et al., 1993; Valeur, 2001; Stahl et al., 2013; Long et al., 2017). By further analysis of the donor anisotropy, this method allows the discrimination between hetero-FRET (FRET between two different donor and acceptor fluorophores) and homo-FRET (energy migration between identical fluorophores and thus identical receptors). Therefore, monomeric versions of the fluorescent proteins Venus and Cherry fused to the C-terminus of the *Arabidopsis* ethylene receptors were chosen as FRET pair (Shaner et al., 2004; Kremers et al., 2006). Tobacco epidermal leaf cells were (co-)transformed with the relevant receptor(s), and expression was induced by the addition of β-estradiol. The fluorescence lifetime measurements were performed in combination with an inducible expression system which allows the discrimination of autofluorescence and prevents overexpression artifacts (Suhling et al., 2015). To verify localization at the ER membrane in the tobacco leaf cells, we probed the intracellular localization of full-length type-II receptor AtERS2 by confocal microscopy. As a result, we observed a strictly separated localization of the PM-localized dye FM4-64 and AtERS2-mVenus, whereas a colocalization of AtERS2 mVenus with a mCherry-labeled ER-marker protein was detected (**Figures 7A,B**; Nelson et al., 2007). Additionally, we detected colocalization of AtERS2 and other *Arabidopsis* ethylene receptors which are a necessary condition for subsequent analysis of receptor-receptor interactions (**Figures 7C**–**F**). Type-I and type-II receptors AtERS1 and AtERS2, the ER-bound receptor Reticulonlike protein B2 (AtBTI2) as a negative control, and a tandem construct with both FPs fused to AtERS2 as a positive control were used to pinpoint *in vivo* interaction of the two receptor subfamilies. For each receptor combination, fluorescence lifetime (*τ*) and anisotropy (*r*) of the donor fluorophore mVenus were measured. Fluorescence lifetime of a fluorophore represents the average time a fluorophore remains in the fluorescent state after excitation by a pulsed laser. For heteromeric interactions, the mVenus fluorescence is quenched by FRET between mVenus and mCherry fluorophores, thereby shortening the mVenus fluorescence lifetime (Gadella et al., 1993, 1994), whereas fluorescence anisotropy represents the rotational freedom of a fluorophore. This rotation is reduced by fusion of the fluorophore to a protein, increasing the anisotropy value *r*. Occurrence of hetero-FRET additionally increases the anisotropy due to a direct interaction of donor and acceptor fluorophore, again limiting rotation of the donor fluorophore. Whereas in case of homo-FRET, energy can migrate radiation free from one excitated mVenus fluorophore to another mVenus in close proximity but with slightly different dipole orientation. This leads to a reduced overall mVenus anisotropy due to a depolarization of the overall mVenus signal and thus decreases the observed value *r* for the steady state anisotropy. Thus, the measurement of anisotropy enables a direct discrimination between homo- and heteromeric interactions (Gauthier et al., 2001; Bader et al., 2011). In our studies, we measured lifetimes for each AtERS2-mVenus combination by FLIM. As shown in **Figure 8A** cotransfections of AtERS2 with either type-I receptor AtERS1 (2.56 ns) or type-II receptor AtERS2 (2.59 ns) resulted in a significant reduction of the AtERS2-mVenus (2.90 ns) fluorescence lifetime (see **Supplementary Table S2**). In the context of the also observed lifetime reduction of the positive control (AtERS2-mVenus-mCherry) and the unchanged lifetime

FIGURE 7 | Intracellular localization of AtERS1 and AtERS2 transiently expressed in *N. benthamiana* epidermis cells. Confocal laser scanning microscopy images of mVenus (mV) and mCherry (mC)-tagged receptor proteins. (A–A**″**) AtERS2 does not colocalize with the PM dye FM4-64 and is instead found (B–B**″**) at the ER, where colocalization with the ER-mCherry marker protein is detected. AtERS2-mVenus colocalizes with (C–C″) AtERS1 mCherry, (D–D**″**) AtERS2-mCherry and (E–E**″**) BTI2-mCherry at the ER. (F–F**″**) AtERS2 tagged to mVenus and mCherry is also detected at the ER. Bars = 10 μm.

FIGURE 8 | Homo- and heteromeric interaction pattern of ethylene receptors. Depicted are (A) fluorescence lifetime (τ) and (B) anisotropy (*r*) of AtERS2 mVenus coexpressed with mCherry-tagged receptor proteins in *N. benthamiana* leaf epidermal cells. *N. benthamiana* leaf cells were transfected with the indicated proteins, protein expression was induced by β-estradiol and samples were analyzed by confocal microscopy within 16–25 h. Fluorescence lifetime and anisotropy were calculated of free fluorophore mVenus (*n* = 34), donor-only AtERS2-mVenus (*n* = 113) and coexpression of AtERS2-mVenus with either AtERS1-mCherry (*n* = 68) or AtERS2-mCherry (*n* = 18). Coexpression of AtERS2-mVenus with BTI2-mCherry (*n* = 62) was used as negative control, whereas expression of AtERS2-mVenus-mCherry (*n* = 50) was used as positive control. Distribution of mVenus fluorescence lifetime (τ) and anisotropy (*r*) are depicted as box plot: median, first and third quartiles, minimum and maximum, error bars indicate minimum and maximum of distribution. Statistical analysis was performed using Mann-Whitney test (A, AtERS2-mVenus + AtBTI2 mCherry) and Welch's *t*-test (all other), \*\*\*\**p* < 0.0001, mV = mVenus, mC = mCherry. See also Supplementary Tables S2,S3 for data.

of the negative control (AtERS2-mVenus:AtBTI2-mCherry), these data indicate heteromeric interactions of the type-I receptor AtERS1 with type-II AtERS2 as well as homomeric interactions of AtERS2 protomers. Formation of AtERS2 homomers is further supported by the observed changes in fluorophore anisotropy (see **Figure 8B**). In our experiments, we observed a decreased anisotropy (*r =* 0.27) for AtERS2-mVenus compared to free mVenus (*r* = 0.30, see **Supplementary Table S3**). Moreover, formation of type-II homomers can also be inferred from measurements of AtERS2 with AtERS1. Here, the anisotropy of AtERS2 (*r* = 0.29) is increased compared to the mVenus donor only sample, but decreased compared to the free mVenus fluorophore. In accordance to these results, no significant changes were observed for the negative control compared to the AtERS2-mVenus donor. These data are indicative for homo-FRET interactions between AtERS2-mVenus itself only. In summary, our *in planta* fluorescence studies demonstrate homomeric as well as heteromeric interactions of the type-II receptor AtERS2 *in vivo*. Furthermore, we were able to show that at these conditions homo-FRET interactions of AtERS2 take place as well in the presence as in the absence of type-I receptor AtERS1.

### DISCUSSION

Previous studies on *Arabidopsis* revealed that the ethylene signal perception and transduction is mediated by five receptors localized at the ER membrane (Hua and Meyerowitz, 1998; Grefen et al., 2008). Furthermore, genetic studies identified CTR1 and EIN2 as critical regulators mediating ethylene signaling and identified both proteins as direct interaction partner of the ethylene receptor family (Kieber et al., 1993; Alonso et al., 1999; Huang et al., 2003). However, previous protein-protein interaction studies mainly focused on the prototype type-I receptor AtETR1, whereas related information on subfamily-II receptors is still sparse. Although a direct interaction of ethylene type-II receptors with either CTR1 or EIN2 has been demonstrated by yeast two-hybrid assays and *in planta* FRET studies (Cancel and Larsen, 2002; Bisson and Groth, 2010), detailed information about these interactions is still missing. To elucidate type-II receptor interactions with their downstream signaling targets in more detail, we have established a purification protocol for the *Arabidopsis* full-length type-II receptor ETR2. The protocol is based on the expression of AtETR2 in *E. coli* strain C43 (DE3), which has been successfully applied for the expression of type-I receptor AtETR1 and several other membrane proteins in the past (Miroux and Walker, 1996; Voet van Vormizeele and Groth, 2008). In this case, solubilization of AtETR2 was obtained by the zwitterionic detergent FosCholine-16 and a purification protocol similar to AtETR1 (Classen and Groth, 2012; Milić et al., 2018) was applied. Functional folding of the purified recombinant AtETR2 was verified by CD spectroscopy (see **Figure 2**). Secondary structure calculations based on CD measurements on the purified receptor determined an α–helical content of 48–50% and a β–sheet content of 9–10% which correspond well to sequence based secondary structure predictions by SOPMA (α–helical: 48% and a β–sheet: 14%) (Geourjon and Deléage, 1995; Sreerama and Woody, 2000; Kessenbrock and Groth, 2017).

Previous studies revealed that the ethylene receptors function as negative regulators and are in their active state in the absence of ethylene (Hua and Meyerowitz, 1998). *In vitro* phosphorylation assays demonstrated that receptors in their active state show autokinase activity of various degrees depending on the divalent cation used. Along these lines, our studies on purified full-length AtETR2 substantiate that the recombinant type-II receptor has higher autokinase activity in the presence of magnesium than in the presence of manganese, which is in accordance with previous studies on truncated AtETR2 lacking the transmembrane domain (Moussatche and Klee, 2004). On the other hand, almost no measurable phosphorylation activity was found for chemically denatured AtETR2 (see **Figure 3, Supplementary Figures S1C**,**D**). Taken together, our structural and functional studies on the purified AtETR2 attest that the recombinant type-II receptor used in our *in vitro* binding studies was isolated in a functional and active state. Stable complex formation of type-II receptor AtETR2 with AtEIN2 has been previously identified by *in vivo* FRET studies. Nevertheless only the AtETR1-AtEIN2 interaction was further characterized in detail by *in vitro* tryptophan quenching studies due to the lack of a purified functional type-II receptor isoform at that time (Bisson et al., 2009; Bisson and Groth, 2010). Here, we unravel a tight and highly specific complex formation of AtETR2 with the soluble cytosolic domain of EIN2 (AtEIN2479-1294, see **Figure 4**) which reinforces the previously mentioned *in planta* FRET studies (Bisson and Groth, 2010). Noteworthy, our studies disclose similar binding affinities and binding modes for AtETR1 and AtETR2 with AtEIN2479-1294 which substantiate similar roles of the different isoforms in PPI-based signal transfer to their downstream target. Still, the observed similar complex stability is quite surprising considering the low sequence identity of only 39% of both isoforms as well as the pronounced differences of both receptor subfamilies in their kinase activity and transmembrane architecture. However, these differences in the receptor isoforms may only come into effect at specific signaling conditions as previously shown for AtETR1 where receptor phosphorylation or binding of an ethylene agonist were shown to modulate complex stability with EIN2 (Bisson and Groth, 2010).

Genetic and biochemical studies indicate that apart from EIN2 the Raf-like kinase CTR1 also interacts with all members of the ethylene receptor family to mediate ethylene signaling at the ER membrane (Cancel and Larsen, 2002; Gao et al., 2003). Consequently, we also analyzed the interaction of the different receptor subfamilies with CTR1 in this study. As for EIN2, we found similar affinities of the CTR1 kinase with full-length receptors AtETR1 and AtETR2 representing prototypes of both subfamilies. These results differ from previous studies using yeast two-hybrid screening and pull-down assays which propose a preferred interaction of AtCTR1 with AtETR1 (Clark et al., 1998; Cancel and Larsen, 2002), although these studies did not provide clear quantitative analysis. Moreover, pull-down assays were performed with truncated receptor constructs which also affect complex stability with the CTR1 kinase. Taken together, our binding studies determine that both receptor subtypes have similar binding affinities and binding modes with their downstream targets AtEIN2479-1294 and AtCTR1 emphasizing similar roles of both receptor subfamilies in PPI-based signal transduction. Bearing in mind that both, CTR1 and EIN2, have been localized to the ER these results correlate well with the idea of ER-borne signaling complexes consisting of various combinations of receptor subtypes, EIN2 and CTR1 (Gao et al., 2003; Bisson et al., 2009; Bisson and Groth, 2010).

For further analysis of receptor complex formation and the role for ethylene signaling *in planta*, we used fluorescence microscopy of transiently expressed receptor proteins in *N. benthamiana* leaf cells. In their native background *Arabidopsis* receptors are localized at the ER membrane. In order to demonstrate that this localization is not affected at overexpressed conditions in the *N. benthamiana* system, we first determined subcellular localization of transiently expressed full-length type-II receptor AtERS2 by confocal microscopy. In our experiments, AtERS2 shows clear separation from the plasma membrane as evidenced by staining with FM4-64. On the other hand, clear colocalization with an ER-marker protein, with ethylene receptor AtERS1 or ER-bound *Arabidopsis* receptor Reticulon-like protein B2 (AtBTI2, Nziengui et al., 2007) was observed indicative of a proper localization of AtERS2 at the ER membrane in the *N. benthamiana* system. In subsequent *in planta* interaction studies, we analyzed fluorescence lifetime and anisotropy of a type-II receptor construct (AtERS2-mVenus) in combination with the type-I receptor isoform AtETR1. While changes in fluorescence lifetime refer to a direct interaction of different fluorophores (hetero-FRET), changes in anisotropy provide additional information about FRET processes between identical fluorophores (homo-FRET). Here, direct excitation of a fluorophore (mVenus) by another identical fluorophore in close proximity leads to a decreased anisotropy due to a depolarization effect. Hence, by analyzing fluorescence lifetime and anisotropy the homo- and hetero-FRET state of a donor fluorophore can be discriminated and thereby homo- and heteromeric interaction of the tagged protein (Gauthier et al., 2001; Bader et al., 2011). The results obtained in our FLIM experiments clarify that type-II receptor AtERS2 forms uniform homomeric complexes, but also heteromeric complexes with type-I receptor AtERS1 *in planta*. Thereby, the *in planta* interaction studies confirm our *in vitro* analysis which also attests homomeric complex formation of AtETR1 and AtETR2 but also tight heteromeric interaction of both receptor subfamilies. Previous studies proposed that the additional N-terminal α-helix in type-II receptors might serve as signal sequence (Chen et al., 2007). The tight affinity of type-II receptor homomeric and heteromeric complexes observed in our *in vitro* binding studies now suggests that the additional helix of type-II receptor may serve an increased complex stability. In principle, the tight affinity of type-II receptor complexes could also be related to sequence variation in individual helix segments. However, sequence alignment of both subtypes (**Supplementary Figure S4**) reveals that transmembrane helices of AtETR1 and AtETR2 are highly conserved (66% identical residues). In total, only five residues (G59/I88, A67/M96, L73/G102, A90/F119 and M104/T133 with the ETR1 residue and position given first) show low conservation. But even these residues have not experienced drastic changes in terms of physicochemical properties. Moreover, in a regular α-helix they do not align on the same surface. Consequently, a single of these minor changes then would have to account for the observed large differences in complex stability which is highly unlikely.

Heretofore, interactions of the ethylene receptor family were mainly attributed to the receptor GAF-domain (Xie et al., 2006; Grefen et al., 2008). However, we should bear in mind that the TM domain was not expressed in these studies. The quantitative binding studies presented in this work demonstrate for the first time that the isolated TM domain substantially contributes to receptor complex formation and probably plays a major role in mediating receptor-receptor interactions. Furthermore, previous studies demonstrated that receptors are stabilized by covalent cysteine crosslinks, although their impact on complex formation was not addressed (Schaller et al., 1995; Schaller and Bleecker, 1995). To address this point, we mutated the two cysteines near the amino terminus of AtETR1 (positions 4 and 6) to serine in order to prevent disulfide bond formation in the TM domain of the receptor. The tight and highly specific binding of the purified ETR1-TMC4SC6S mutant to either AtETR1 or AtETR2 together with previous genetic studies of Xie et al. (2006) suggest that disulfide linkage is not essential for AtETR1 signaling.

In summary, our studies demonstrate that the type-II receptor AtETR2 binds AtEIN2 and AtCTR1 with similar affinities as type-I receptor AtETR1 indicating a similar role of both receptor subfamilies in PPI-based signal transfer, although they may still differ in other signal output mechanism mediated by post-translational modification such as phosphorylation or ubiquitination (Moussatche and Klee, 2004; O'Malley et al., 2005; Chen et al., 2007; Kevany et al., 2007). Specifically, the similar affinity of both subfamilies for their downstream targets may explain functional redundancy of the ethylene receptor family, i.e., the common function of all members to repress the constitutive ethylene response, although individual receptors can adopt specific functions that are not replaceable by other isoforms (Liu and Wen, 2012). Moreover, the similar affinity of both subfamilies for CTR1 and EIN2 may indicate that both target the same site in the receptors. However, a sequence alignment of the CTR1 N-terminus and the EIN2 C-terminus, which have been identified in previous studies to mediate binding to the receptors, reveals no particular sequence element to support this idea. Still, we have to bear in mind that, in the end, the 3D structure of a protein determines the interaction. Hence, structures of receptor complexes with CTR1 and/or EIN2 will ultimately unravel the binding site and signal transfer mechanism for both downstream signaling proteins.

Noteworthy, both receptor isoforms efficiently and specially bind to representatives of each subfamily, although interaction with type-II receptor protomers is slightly preferred probably due to a stabilizing effect of the additional TM helix in these receptors. These mixed dimers or higher molecular weight oligomers may reflect functional synergism of the different receptor subtypes to control scope, scale, and pace of ethylene responses. The preferred association of ETR1 with type-II receptors determined in our *in vitro* binding studies is fully in line with previous co-purification experiments of ETR1 with tagged versions of ERS1, ETR2, ERS2, and EIN4 from *Arabidopsis* membrane extracts (Gao et al., 2008) underscoring the biological significance of quantitative studies on purified individual components of the signaling pathway. To further dissect the functional role of the receptor heteromeric complexes for ethylene signaling, *in planta* studies on chimeric receptors consisting of different subdomains from the two subfamilies in different ethylene *loss-of-function* backgrounds may prove promising.

#### DATA AVAILABILITY

The datasets generated for this study are available on request to the corresponding author.

#### AUTHOR CONTRIBUTIONS

GG conceived the project, planned, designed, and supervised the research. MB and NB planned and performed experiments with the assistance of SW-P, SH, and RCB. AM assisted with statistical analysis. MB, GG, NB, AM, SW-P, and SH analyzed data. MB, NB, and GG wrote the manuscript. SW-P and AM critical

#### REFERENCES


edited the manuscript. SH, SW-P, BS, YS, and RS provided expertise and feedback.

#### FUNDING

This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—Projektnummer 267205415—SFB 1208 (project B06 to GG, B04 to RS and project Z02 to SW-P) and GR1616/10-1 to GG and STA12/12 1-1 to YS.

#### ACKNOWLEDGMENTS

The authors thank the staff from the botanical garden of the university for cultivation and supply of *N. benthamiana* plants. We gratefully acknowledge technical support in vector cloning of Patricia Bartschies, Kerstin Förster, and Lena Müller.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material file for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019. 00726/full#supplementary-material


*Bioimaging* 2, 139–159. doi: 10.1002/1361-6374(199409)2:3<139::AID-BIO4>3.0.CO;2-T


between cyan and yellow fluorescent proteins. *Microsc. Microanal.* 12, 238–254. doi: 10.1017/s1431927606060235


some applications. *Proc. Natl. Acad. Sci. USA* 76, 4350–4354. doi: 10.1002/ bies.950190612


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Berleth, Berleth, Minges, Hänsch, Burkart, Stork, Stahl, Weidtkamp-Peters, Simon and Groth. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# New Insights in Transcriptional Regulation of the Ethylene Response in Arabidopsis

Likai Wang1,2 and Hong Qiao1,2 \*

1 Institute for Cellular and Molecular Biology, The University of Texas at Austin, Austin, TX, United States, <sup>2</sup> Department of Molecular Biosciences, The University of Texas at Austin, Austin, TX, United States

As any living organisms, plants must respond to a wide variety of environmental stimuli. Plant hormones regulate almost all aspects of plant growth and development. Among all the plant hormones, ethylene is the only gaseous plant hormone that plays pleiotropic roles in plant growth, plant development and stress responses. Transcription regulation is one main mechanism by which a cell orchestrates gene activity. This control allows the cell or organism to respond to a variety of intra- and extracellular signals and thus mount a response. Here we review the progress of transcription regulation in the ethylene response.

Keywords: ethylene, transcription regulation, Arabidopsis, histone, hormone

#### Edited by:

Caren Chang, University of Maryland, United States

#### Reviewed by:

Gyeong Mee Yoon, Purdue University, United States Jose M. Alonso, North Carolina State University, United States

> \*Correspondence: Hong Qiao hqiao@austin.utexas.edu

#### Specialty section:

This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science

Received: 25 March 2019 Accepted: 31 May 2019 Published: 18 June 2019

#### Citation:

Wang L and Qiao H (2019) New Insights in Transcriptional Regulation of the Ethylene Response in Arabidopsis. Front. Plant Sci. 10:790. doi: 10.3389/fpls.2019.00790

#### TRANSCRIPTIONAL REGULATION IN THE ETHYLENE RESPONSE

Like all living organisms, plants must respond to a wide variety of environmental stimuli. Plant hormones, produced in response to environmental stimuli, regulate almost all aspects of plant growth and development. Ethylene is a gaseous plant hormone that plays pleiotropic roles in plant growth, plant development, and stress responses. Histone acetylation, which is modulated through ethylene-mediated signaling, regulates dynamic changes in chromatin structure that result in transcriptional regulation in responses to ethylene.

Ethylene is perceived by a family of receptors bound to the endoplasmic reticulum (ER) membrane (Chang et al., 1993; Bleecker et al., 1998; Hua and Meyerowitz, 1998; Hua et al., 1998; Sakai et al., 1998). Each receptor binds ethylene via a copper cofactor that is provided by the copper transporter RESPONSIVE-TO-ANTAGONIST 1 (RAN1)(Hirayama et al., 1999). In the absence of ethylene, ethylene receptor ETHYLENE RECEPTOR 1 (ETR1) interacts with CONSTITUTIVE TRIPLE RESPONSE 1 (CTR1), a downstream negative regulator of ethylene signaling (Chang et al., 1993; Kieber et al., 1993; Gao et al., 2003; Shakeel et al., 2015). CTR1 is a protein kinase that phosphorylates ETHYLENE INSENSITIVE 2 (EIN2), a key positive regulator of ethylene signaling (Alonso et al., 1999; Ju et al., 2012), preventing the ethylene response. In addition, in the absence of ethylene, EIN2 protein levels are regulated by EIN2 TARGETING PROTEIN 1 and 2 (ETP1/2) via ubiquitin/proteasome-mediated degradation (Qiao et al., 2009).

In the presence of ethylene both ethylene receptors and CTR1 are inactivated, and the C-terminal end of EIN2 is dephosphorylated and cleaved by unknown mechanisms. The cleaved C-terminal end of EIN2 translocates to the nucleus (Qiao et al., 2012; Wen et al., 2012; Ju et al., 2015) where it facilitates the acetylation of histone 3 at K14 and K23 (H3K14 and H3K23, respectively) to regulate ETHYLENE INSENSITIVE 3 (EIN3) and ETHYLENE INSENSITIVE 3

**38**

LIKE1 (EIL1) – dependent transcriptional regulation (Zhang et al., 2016). The cleaved EIN2 C-terminal end also translocates into the P-body through associating with 30UTRs of EIN3 BINDING F-BOX1 (EBF1) and EBF2, further repressing their translation (Guo and Ecker, 2003). EBF1 and EBF2 in turn stabilize EIN3 and EIL1, resulting in activation of EIN3- and EIL1-dependent transcription and the activation of an ethylene response (Li et al., 2015; Merchante et al., 2015).

Both genetic and molecular studies have demonstrated that EIN3 and EIL1 are positive regulators that are necessary and sufficient for the ethylene response (Chao et al., 1997; Guo and Ecker, 2003; Chang et al., 2013). The EIN3 gene encodes a nuclear-localized protein that is essential to the response to ethylene (Chao et al., 1997). In the absence of EIN3, plants are partially insensitive to ethylene both at the morphological and molecular levels (Chao et al., 1997; Guo and Ecker, 2003). The EIN3 binding motif was identified after analysis of the promoters of the genes that are highly up-regulated by ethylene and followed by validation using an electrophoresis mobility shift assay (EMSA) (Ohme-Takagi and Shinshi, 1990; Eyal et al., 1993; Meller et al., 1993; Sessa et al., 1995; Shinshi et al., 1995; Sato et al., 1996; Solano et al., 1998). Using the EMSA assay, EIN3 was shown to form a homodimer in the presence of DNA in vitro (Solano et al., 1998). However, whether the homodimer is formed in vivo and whether the homodimer is required for EIN3 to function in the ethylene signaling are unknown. A number of transcription factors are known to form homodimers or heterodimers, which have different specificities and affinities for certain DNA motifs (Funnell and Crossley, 2012). Finding out whether the dimerization is necessary for EIN3<sup>0</sup> s function in vivo will be an interesting question in the transcriptional regulation of the ethylene response.

To explore the transcription regulation in response to ethylene, Chang et al characterized the dynamic ethylene transcriptional response by identifying targets of EIN3, the master regulator of the ethylene signaling pathway, using chromatin immunoprecipitation sequencing and transcript sequencing during a time course of ethylene treatment (Chang et al., 2013). They found that the number of genes bound by EIN3 does not change significantly in response to ethylene. The amount of EIN3 bound increases upon ethylene treatment, and the expression of most of EIN3-bound genes is up regulated by ethylene, which is consistent with a role of EIN3 as a transcriptional activator. Chang et al. also analyzed the sequences of EIN3-bound regions identified by ChIP-seq (Ohme-Takagi and Shinshi, 1990; Eyal et al., 1993; Meller et al., 1993; Sessa et al., 1995; Shinshi et al., 1995; Sato et al., 1996; Solano et al., 1998). A motif similar to that previously identified in the promoter regions of ethylene up-regulated genes was present in EIN3-bound regions (Ohme-Takagi and Shinshi, 1990; Eyal et al., 1993; Meller et al., 1993; Sessa et al., 1995; Shinshi et al., 1995; Sato et al., 1996; Solano et al., 1998). Intriguingly, Chang et al. also found that ethylene-induced transcription occurs in temporal waves that were regulated by EIN3 with the

potentially distinct layers of transcriptional control (Chang et al., 2013). EIN3 binding was found to modulate a multitude of downstream transcriptional cascades, including a major feedback regulatory circuitry of the ethylene signaling pathway, as well as most of the hormone-mediated growth response pathways, which indicates that network-level feedback regulation results in overall system control and homeostasis (Chang et al., 2013). This type of study can be applied to identify novel components in signaling pathways (Rosenfeld et al., 2002; Amit et al., 2007; Tsang et al., 2007; Avraham and Yarden, 2011; Feng et al., 2011; Yosef and Regev, 2011).

Although the transcriptional activation has been the main focus in the ethylene response, transcriptome analysis in Chang's study clearly showed that nearly 50% of ethylene-altered genes are down regulated and that a subset of ethylene-repressed genes are bound by EIN3 (Chang et al., 2013). Notably, most of the genes are down regulated by ethylene within 1 h of treatment. This result strongly suggests that transcriptional repression plays a critical role in early ethylene response. Interestingly, a recent study from the Qiao lab showed that transcriptional repression dose plays important roles in ethylene response. We identified two histone deacetylases (HDACs) SIRTUIN 1 and 2 (SRT1 and SRT2) that regulate ethylene-repressed genes (Zhang et al., 2018). Notably the study found that SRT2 binds the target promoter regions to inhibit acetylation of histone 3 at K9 (H3K9Ac), repressing gene expression in response to ethylene (**Figure 1**).

Transcriptional repression by chromatin modification is one of the principal mechanisms employed by eukaryotic active repressors (Thiel et al., 2004; Kagale and Rozwadowski, 2011). The importance of HDACs in transcriptional repression during plant growth and development has been well established (Song et al., 2005; Li et al., 2017). For example, in Arabidopsis, the EAR motif containing class II ETHYLENE RESPONSIVE ELEMENT BINDING FACTORS (ERFs), such as ERF3 and ERF4, which are known to function as active repressors in vitro and in vivo, have been shown to physically interact with AtSAP18, which in turn interacts and forms a repression complex with AtHDA19 (Fujimoto et al., 2000; Ohta et al., 2001; McGrath et al., 2005; Song and Galbraith, 2006). AtERF7, another EAR motif-containing class II ERF protein, is also known to recruit AtHDA19 via a physical interaction with AtSIN3 (Song et al., 2005). In planta, coexpression of AtERF3, AtSAP18, and AtHDA19 or AtERF7, AtSIN3, and AtHDA19 results in greater transcriptional repression of reporter genes as compared to when these proteins are expressed alone suggesting a role for AtSAP18, AtSIN3, and AtHDA19 in ERF-mediated transcriptional repression possibly via histone deacetylation (Song et al., 2005; Song and Galbraith, 2006). Yet, whether EAR containing proteins are also required for SRT1 and SRT2 mediated transcriptional repression in response to ethylene is unknown (Zhang et al., 2018). It also remains unclear whether the ethylene response has a molecular mechanism of transcriptional repression similar to that induced by other plant hormones. Exploring the molecular mechanism of transcriptional

repression will provide more insight into ethylene signaling and ethylene response.

#### HISTONE ACETYLATION IN ETHYLENE-MEDIATED TRANSCRIPTIONAL REGULATION

In eukaryotes, the binding of transcription factors is mainly determined by chromatin structure, namely the state of the genome's packaging with specific structural proteins, mainly histones. Chromatin undergoes different dynamics structure changes, further influences transcription factor binding. Among all the regulations, histone acetylation results in a switch between repressive and permissive chromatin. In general, acetylation neutralizes the positive charges of lysine residues and decreases the interaction between histone and DNA, leading to a more relaxed chromatin structure, which is associated with transcriptional activation. In contrast, deacetylation induces a compact chromatin structure, which is associated with transcriptional repression (Fletcher and Hansen, 1996; Steger and Workman, 1996; Luger and Richmond, 1998).

A Number of studies have shown a tight link between histone acetylation and plant hormone responses (Zhu, 2010). In the studies of ethylene signaling, authors found that GENERAL CONTROL NON-REPRESSED PROTEIN 5 (GCN5), which belongs to a family of histone acetyl transferases (HATs), promotes transcriptional activation (Brownell and Allis, 1996; Grant et al., 1997; Bian et al., 2011; Weake and Workman, 2012; Ryu et al., 2014). The Arabidopsis gcn5 mutant shows hypersensitivity to ethylene treatment. In the gcn5 mutant, the histone acetylation at H3K9 and H3K14 in the promoter regions of ethylene response genes is elevated, and the elevation is associated with the up-regulation of gene expression (Poulios and Vlachonasios, 2016). In Arabidopsis, the HAC family, which are the homologs of CREB-binding protein (CBP) and p300, the mammalian family of HAT domain containing transcriptional coactivators, play pleotropic roles in plant growth and development (Pandey et al., 2002; Li et al., 2014). The hac1hac5 double mutant was found to have a constitutive triple response phenotype (Pandey et al., 2002; Li et al., 2014). It was expected that gene expression would be down regulated in the hac1hac5 double mutant due to the reduction of histone acetylation levels; however, similar to that in gcn5 mutant, the downstream ethylene responsive genes are elevated in hac1hac5 double mutant (Li et al., 2014), suggesting an indirect regulation of ethylene responsive genes by HAC1 and HAC5.

Expression of two HDACs, HAD6 and HDA19, are specifically elevated by ethylene treatment. The expression of ethylene responsive gene ERF1 is anti-correlated with the levels of histone H3 acetylation in 35S:HDA19 transgenic plants, showing that HDA19 indirectly influences ERF1 gene expression (Zhou et al., 2005). It is possible that HDA19 induces ERF1 expression by preventing binding of an unknown transcription repressor that regulates ERF1 expression (Zhou et al., 2005). JAZ proteins recruit HDA6 to deacetylate histones and obstruct the chromatin binding of EIN3/EIL1, therefore repressing EIN3/EIL1-dependent transcription and inhibiting jasmonic acid-mediated signaling (Zhu et al., 2011). This provides evidence that histone acetylation regulates EIN3 target genes.

H3K14Ac and H3K23Ac, but not H3K9Ac, H3K18Ac, or H3K27Ac, are elevated by ethylene treatment at the molecular level (Zhang et al., 2016; Yoon et al., 2017). Interestingly, even though the levels of H3K9Ac are not regulated by ethylene, the

levels of H3K9Ac in the promoters of ethylene up-regulated genes are higher than that in those ethylene down-regulated genes both with and without ethylene treatments (Zhang et al., 2016; Yoon et al., 2017). Presumably, H3K9Ac is a pre-existing mark that labels genes regulated by ethylene, whereas the elevation of H3K14Ac and H3K23Ac is positively associated with gene activation. Most importantly, the ethylene-induced change of histone acetylation is EIN2 dependent. However, EIN2 is not a histone or a DNA binding protein, and the biochemical function of EIN2-C remains unknown. Yeast twohybrid screening and ChIP-re-ChIP suggests that EIN2-C is associated with histones at least in part through EIN2 NUCLEAR ASSOCIATED PROTEIN 1 (ENAP1), which has histone binding activity (**Figure 1**; Zhang et al., 2017). It is possible that EIN2- C is a scaffolding protein that is important for the formation of HAT-containing protein complexes in response to ethylene. Identification of HAT or HDAC that functions in cooperation with EIN2-C would validate this assumption. In contrast to transcriptional activation, histone acetylation of H3K9Ac was found to be involved in the transcriptional repression, and the regulation is partially mediated by histone deacetylase SRT1 and SRT2 (**Figure 1**).

Taken together, currently available data suggest that in the absence of ethylene ENAP1 binds to histones to keep chromatin in a relaxed state poised for a rapid response to ethylene (**Figure 1**). In the presence of ethylene, EIN2-C is translocated to the nucleus where it interacts with ENAP1 and potentially HATs resulting in histone acetylation. This causes an uncompacting of chromatin, resulting in more EIN3 binding to target genes and ultimately transcription activation (**Figure 1**). It is not known how the histone acetylation targets are determined in the presence of ethylene. Zhang et al. showed that EIN3 is partially required for the ethylene-induced elevation of H3K14Ac and H3K23Ac (Pandey et al., 2002), suggesting that EIN3 might mark histone acetylation targets. Multi-protein assemblies have been shown to determine the substrate specificities and targeting of integral HAT subunits. The molecular mechanism of how EIN3, ENAP1, and EIN2-C coordinate to integrate the histone acetylation and transcription regulation remains to be elucidated. Beside histone acetylation regulation in transcriptional activation, histone acetylation has been shown to be involved in transcriptional repression in ethylene response. As mentioned above, SRT1 and SRT2 mediate transcriptional repression that requires a low level of H3K9Ac (Yosef and Regev, 2011). How the H3K9Ac levels are determined in the desired targets in the first place is an interesting and important question that needs to be addressed.

#### REFERENCES

Alonso, J. M., Hirayama, T., Roman, G., Nourizadeh, S., and Ecker, J. R. (1999). EIN2, a bifunctional transducer of ethylene and stress responses in Arabidopsis. Science 284, 2148–2152. doi: 10.1126/science.284.5423. 2148

### CONCLUDING REMARKS AND FUTURE PERSPECTIVES

Plants must respond accurately and quickly to hormones, and this necessitates a flexible and rapid way to control gene expression. The acetylation of histone tails by HATs neutralizes positive charges on these proteins that would otherwise interact with negatively charged DNA, facilitating nucleosome unwrapping for rapid transcription activation. How plants utilize a limited number of HATs and HDACs to specifically regulate responses to different hormones is largely unknown. Presumably, the specificity relies on the partners of HATs and HDACs. Identification of the HAT- and HDAC-containing complexes upon ethylene treatment will reveal details of the molecular mechanisms that underlie the ethylene response. Recent studies have clearly shown that different tissues respond to plant hormones differently (Garg et al., 2012; Pattison et al., 2015; Raines et al., 2016). Most available data on histone acetylation induced by plant hormones come from analyses of the whole plant. Studies of histone acetylation in individual tissues and in different cell types will provide more detailed insight into how histone acetylation controls responses to plant hormones. Transcription factor binding in eukaryotes is highly dependent on the context of binding sites on chromatin, but little is known about how EIN3 determines histone acetylation sites in target genes. A more complete understanding of the molecular mechanism of determination of transcriptional activation and transcriptional repression during the ethylene response will facilitate development of methods to improve crop production.

### AUTHOR CONTRIBUTIONS

Both authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### FUNDING

This work was supported by a grant from the National Institutes of Health (NIH) (R01GM115879-01) to HQ.

### ACKNOWLEDGMENTS

We thank the members of the Qiao laboratory for critical reading of this manuscript.



is mediated by the AGCCGCC cis-sequence. Plant Cell Physiol. 37, 249–255. doi: 10.1093/oxfordjournals.pcp.a028939


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Wang and Qiao. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Low Temperature Storage Stimulates Fruit Softening and Sugar Accumulation Without Ethylene and Aroma Volatile Production in Kiwifruit

Oscar W. Mitalo<sup>1</sup> , Sumire Tokiwa<sup>1</sup> , Yuki Kondo<sup>1</sup> , Takumi Otsuki<sup>1</sup> , Ivan Galis1,2 , Katsuhiko Suezawa<sup>3</sup> , Ikuo Kataoka<sup>3</sup> , Anh T. Doan<sup>1</sup> , Ryohei Nakano<sup>1</sup>† , Koichiro Ushijima<sup>1</sup> and Yasutaka Kubo<sup>1</sup> \*

#### Edited by:

Mondher Bouzayen, National Polytechnic Institute of Toulouse, France

#### Reviewed by:

María Serrano, Universidad Miguel Hernández de Elche, Spain M. Teresa Sanchez-Ballesta, Instituto de Ciencia y Tecnología de Alimentos y Nutrición (ICTAN), Spain

#### \*Correspondence:

Yasutaka Kubo ykubo@okayama-u.ac.jp

†Present address: Ryohei Nakano, Graduate School of Agriculture, Kyoto University, Kyoto, Japan

#### Specialty section:

This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science

Received: 15 February 2019 Accepted: 21 June 2019 Published: 05 July 2019

#### Citation:

Mitalo OW, Tokiwa S, Kondo Y, Otsuki T, Galis I, Suezawa K, Kataoka I, Doan AT, Nakano R, Ushijima K and Kubo Y (2019) Low Temperature Storage Stimulates Fruit Softening and Sugar Accumulation Without Ethylene and Aroma Volatile Production in Kiwifruit. Front. Plant Sci. 10:888. doi: 10.3389/fpls.2019.00888 <sup>1</sup> Graduate School of Environmental and Life Science, Okayama University, Okayama, Japan, <sup>2</sup> Institute of Plant Science and Resources, Okayama University, Kurashiki, Japan, <sup>3</sup> Faculty of Agriculture, Kagawa University, Miki, Japan

Fruit ripening in response to propylene (an ethylene analog), 1-methylcyclopropene (1-MCP, an ethylene action inhibitor), and low temperature (5◦C) treatments was characterized in "Kosui" kiwifruit (Actinidia rufa × A. chinensis). Propylene treatment induced ethylene production, with increased expression levels of 1-aminocyclopropane-1-carboxylic acid (ACC) synthase 1 (AcACS1) and ACC oxidase 2 (AcACO2), and rapid fruit softening together with increased expression levels of polygalacturonase (AcPG) and expansin 1 (AcEXP1) within 5 days (d). Fruit soluble solids concentration (SSC) and contents of sucrose, glucose, and fructose together with the expression levels of β-amylase 1 (Acβ-AMY1), Acβ-AMY2, and invertase (AcINV3-1) increased rapidly after 5 d exposure to propylene. Furthermore, propylene exposure for 5 d was sufficient to induce the production of key aroma volatile compounds, ethyland methyl butanoate, accompanied with increased expression levels of alcohol acyl transferase (AcAAT). Application of 1-MCP at the start of the experiment, followed by continuous exposure to propylene, significantly delayed fruit softening, changes in SSC and sugars, and strongly suppressed the production of ethylene, aroma volatiles, and expression of associated genes. During storage, fruit softening, SSC and sugar increase, and increased expression of genes associated with cell wall modification and carbohydrate metabolism were registered without detectable ethylene production; however, these changes occurred faster at 5◦C compared to 22◦C. Interestingly, ethyl and methyl butanoate as well as AcAAT expression were undetectable in kiwifruit during storage, while they were rescued by post-storage propylene exposure, indicating that the production of aroma volatile compounds is strongly ethylene-dependent. Transcript levels of a NAC-related transcription factor (TF), AcNAC3, increased in response to both propylene and low temperature treatments, while AcNAC5 was exclusively up-regulated by propylene. By contrast, transcript levels of a MADS-box TF, AcMADS2, exclusively increased in response to low temperature. The above findings indicate that kiwifruit ripening is inducible by either ethylene or low temperature signals. However, fruit ripened by low temperature were deficient in ethylene-dependent

**44**

aroma volatiles, suggesting that ethylene signaling is non-functional during low temperature-modulated ripening in kiwifruit. These data provide further evidence that ethylene-dependent and low temperature-modulated ripening in kiwifruit involve different regulatory mechanisms.

Keywords: ethylene, ethyl butanoate, low temperature, methyl butanoate, softening, sugars

#### INTRODUCTION

fpls-10-00888 July 4, 2019 Time: 16:11 # 2

The plant hormone ethylene regulates a wide range of plant growth and developmental processes, including fleshy fruit ripening (Lashbrook et al., 1998; Giovannoni, 2004). During ripening, fleshy fruit undergo various physiological, biochemical, and structural changes including softening, starch degradation to sugars, pigment accumulation, and production of aroma volatiles (Klee and Giovannoni, 2011; Osorio et al., 2013). The onset of fruit ripening in climacteric fruit such as tomatoes, apples, and peaches is accompanied by a marked increase in ethylene production (Xu et al., 2012), which is regulated at the transcriptional level via the differential expression of genes encoding two key enzymes: 1-aminocyclopropane-1-carboxylic acid (ACC) synthase (ACS) and ACC oxidase (ACO) (Wang et al., 2002; Cherian et al., 2014).

Kiwifruit (Actinidia spp.) are categorized as climacteric, since fruit ripening is largely driven by ethylene-regulated changes in gene expression (Antunes, 2007; Yin et al., 2008). Exogenous ethylene or propylene treatment initiates rapid fruit softening through the induction of several cell wall modification-associated genes such as polygalacturonase (AcPG) and expansin 1 (AcEXP1) (Wang et al., 2000; Atkinson et al., 2011; Mworia et al., 2012). In addition, kiwifruit respond to exogenous ethylene/propylene by increasing their soluble solids concentration (SSC), which is associated with the induction of various starch degradationrelated genes such as β-amylase (Acβ-AMY) (Nardozza et al., 2013; McAtee et al., 2015; Hu et al., 2016). Ripe kiwifruit produce a composite of aroma volatiles, consisting of mainly aldehydes and esters (Marsh et al., 2006). Characteristic kiwifruit esters have been identified as ethyl butanoate and methyl butanoate (Zhang et al., 2009), and their regulation by ethylene has been previously described (Atkinson et al., 2011; Günther et al., 2011, 2015). Ethylene-induced ripening changes in kiwifruit are usually followed by a sharp increase in ethylene production caused by the up-regulation of key biosynthetic genes AcACS1 and AcACO2 (Pratt and Reid, 1974; Mworia et al., 2010; Atkinson et al., 2011; McAtee et al., 2015).

Kiwifruit exhibit a peculiar ripening behavior, as substantial softening in healthy intact fruit occurs during low temperature storage in the absence of any detectable ethylene (Antunes, 2007; Yin et al., 2009). Arpaia et al. (1987) demonstrated that kiwifruit are sensitive to extremely low ethylene concentrations (as low as 0.01 µLL−<sup>1</sup> ). Consequently, the substantial softening during storage at <1.5◦C in air (presumed to contain 0.001 µLL−<sup>1</sup> ethylene) is believed to be controlled by basal levels of system I ethylene that is present in most fruit and plant tissues (Kim et al., 1999; Pranamornkith et al., 2012; Jabbar and East, 2016). However, to date, there is no direct evidence linking system I ethylene to kiwifruit ripening during low temperature storage, and thus, the regulatory mechanisms associated with this phenomenon remain less well understood.

Over the past few years, this study group has been dedicated to elucidating the molecular mechanisms underpinning low temperature-modulated fruit ripening in kiwifruit. A preliminary study by Mworia et al. (2012) demonstrated that healthy intact "Sanuki Gold" kiwifruit softened; accumulated AcPG, Pectate lyase (AcPL), and AcEXP1 mRNAs; and decreased their titratable acidity (TA) during storage at 4◦C, but not at 25◦C, despite the lack of any detectable increase in ethylene production. These changes were not suppressed by frequent application of 1-methylcyclopropene (1-MCP) to keep fruit insensitive to ethylene, indicating that they occur outside the sphere of ethylene influence. Similar results were reported using different kiwifruit cultivars (Asiche et al., 2017; Mitalo et al., 2018, 2019), confirming that low temperature-modulated ripening is common to all kiwifruit cultivars. Follow-up transcriptomic studies revealed that a distinct set of ripening-associated genes in kiwifruit were uniquely regulated by low temperature, independent of ethylene (Asiche et al., 2018). Despite these findings, low temperaturemodulated ripening in kiwifruit remains a poorly understood phenomenon. There is a growing need to conduct more focused studies to examine the similarities and differences in the molecular regulation of ethylene-induced and low temperaturemodulated ripening in kiwifruit.

Using gas chromatography–mass spectrometry (GC–MS) to analyze the volatiles and soluble sugar contents of "Kosui" kiwifruit, an interspecific hybrid between Actinidia rufa and A. chinensis, this study sought to investigate the impact of low temperature (5◦C) on sugar and aroma volatile profiles relative to ethylene effect. Changes in sugar composition and aroma volatile production in response to propylene, 1-MCP, and during storage at 5 and 22◦C were compared, and their concomitant gene expression patterns are reported. Our results indicate that the production of aroma volatiles is strongly ethylene-dependent and is absent during cold storage, providing evidence that ethylene signaling is non-functional during low temperaturemodulated ripening.

#### MATERIALS AND METHODS

#### Plant Material and Treatments

"Kosui" kiwifruit grown under standard cultural practices were harvested from a commercial orchard in Takamatsu, Japan at a physiological maturity stage [170 days (d) after full bloom (DAFB), firmness: 73.02 ± 3.39 N, SSC: 7.79 ± 0.18%,

TA: 2.20 ± 0.03%]. After harvesting, careful sorting was conducted to exclude fruit with physical injuries, disease symptoms, and those producing ethylene. Fruit were then divided into five sets, corresponding to the various treatments.

#### Ethylene-Dependent Ripening

Three sets of 30 fruit each were used in this experiment. The first set was held in gas-tight plastic containers that were continuously treated with 5000 µLL−<sup>1</sup> propylene, a well-known ethylene analog (McMurchie et al., 1972; Mworia et al., 2010; Asiche et al., 2016). Propylene treatment was done to induce ethylene signaling, and to allow for determination of endogenous ethylene produced by the fruit. The second set was initially exposed to 2 µLL−<sup>1</sup> 1-MCP for 12 h followed by continuous exposure to 5000 µLL−<sup>1</sup> propylene. 1-MCP was released by dissolving SmartFreshTM powder (AgroFresh, PA, United States) in water. The third set contained non-treated fruit as a control. All treatments were carried out at 22◦C for up to 9 d. Soda lime was placed in plastic containers during propylene and 1-MCP treatments to reduce CO<sup>2</sup> accumulation.

#### Low Temperature-Modulated Ripening

Two sets of 300 fruit each were stored either at 5 or 22◦C in air for up to 49 d. During storage, fruit were individually wrapped in perforated polythene bags to reduce water loss, before placing them (∼10 cm apart) in plastic trays. Ethylene production pattern of each fruit was monitored at weekly intervals throughout the storage period. To avoid ethylene accumulation in the storage chambers, fruit that produced detectable ethylene (<10%) were excluded based on previous observations that ethylene production correlated with the appearance of disease symptoms (Asiche et al., 2018). At the end of the storage period, fruit at 5◦C were further divided into two groups before being transferred to 22◦C for up to 14 d; one group was continuously treated with propylene as described above, while the other group was left untreated. Similarly, fruit at 22◦C were divided into two groups; one group was continuously treated with propylene and the other one was left untreated.

#### Evaluation of Changes in Ethylene Production, Firmness, SSC, and TA

To determine ethylene production, individual fruit were incubated in a 440-mL container for up to 1 h. Ethylene production rate was measured by withdrawing 1 mL of headspace gas and injecting it into a gas chromatograph (Model GC4 CMPF, Shimadzu, Kyoto, Japan), equipped with a flame ionization detector (set at 200◦C) and an activated alumina column (set at 80◦C) (Mworia et al., 2012). This procedure has a minimum ethylene detection capacity of 0.01 nLg−<sup>1</sup> h −1 . Fruit firmness was measured at two equatorial regions of peeled fruit using a penetrometer (model SMT-T-50, Toyo Baldwin, Tokyo, Japan) fitted with an 8-mm plunger. Data were recorded as Newtons (N) and firmness was expressed as a mean of five independent biological replications. SSC and TA were determined using fruit juice as described elsewhere (Asiche et al., 2018; Mitalo et al., 2019).

### Collection of Aroma Volatiles

Aroma volatile compounds were collected according to the procedure by Sobhy et al. (2017), with slight modifications. Intact fruit were placed into sealed 1360-mL containers equipped with two ports; one for the air inlet and another port for the outlet. Dried air, purified by passing through a charcoal filter, was introduced to sweep through the headspace at approximately 0.75 L min−<sup>1</sup> . The air was then pulled out through the outlet port fitted with a custom-made 10-cm glass trap (5 mm inner diameter), filled with Porapak Q (200 mg, Supelco Analytical, Bellefonte, PA, United States) held in place by two plugs of silanized glass wool. Porapak Q traps were conditioned before use by flushing with 2 mL methanol (Sigma– Aldrich Co., United States), followed by 2 mL dichloromethane (DCM, Wako Pure Chemical Industries, Japan), dried, and then placed overnight in an oven at 60◦C for complete drying. To exclude the effect of background volatiles in air, empty containers having no fruit were included in the collection setup. Aroma volatiles were collected over 24 h periods. After each collection period, volatile compounds were eluted from Porapak Q traps with 1 mL DCM after adding 400 ng tetralin (1,2,3,4 tetrahydronaphthalene; Nacalai Tesque, Japan) on the top of each column. Tetralin was used as an internal standard. The samples were stored in 1.5 mL glass vials at −20◦C until further analysis.

### Extraction and Derivatization of Soluble Sugars

Soluble sugars were obtained and derivatized as described by Wang et al. (2010). Briefly, the samples (0.1 g) were ground in liquid nitrogen and extracted in 1.4 mL of 100% methanol with ribitol (12 µg) added as an internal standard. After fractionating the non-polar metabolites into chloroform, 150 µL of the polar phase was transferred into a 1.5 mL micro-centrifuge tube to measure the metabolites (soluble sugars). These were dried under vacuum without heating, flushed with nitrogen gas, and then derivatized sequentially with methoxyamine hydrochloride and N-methyl-N-trimethylsilyltrifluoroacetamide (Lisec et al., 2006). The samples were stored at −20◦C until further analysis.

### GC–MS Conditions and Chemical Analysis

Volatile eluates and metabolites were analyzed using an Agilent 240 GC–MS ion trap system coupled with Agilent 7891A GC fitted with the HP-5MS column (5% phenyl methyl silox, 30 m length × 0.25 mm inner diameter × 0.25 µm film thickness) (Agilent Technologies, Santa Clara, CA, United States). One microliter of each eluted sample was injected in split mode (1:30) into the injector port of the GC instrument held at 230◦C via an Agilent 7693A auto-sampler. Helium (1 mL/min) was used as carrier gas, and MS ionization was achieved by electron impact (EI) at emission current 30 µAmps for volatiles (10 µAmps for metabolites) in the ion trap held at 200◦C (transfer line was 260◦C). For headspace volatile analyses, the GC oven temperature was programmed at 40◦C

for 3 min, and then increased at 5◦C min−<sup>1</sup> to 180◦C, followed by 20◦C min−<sup>1</sup> ramp to 300◦C, where it was held for an additional 5 min before returning to initial conditions. For metabolite samples, GC temperature program was 5 min at 60◦C, followed by 5◦C min−<sup>1</sup> to 300◦C, 5 min hold period, and return to initial temperature and equilibration. MS data were collected in full scan mode in mass range m/z 40–300 for volatiles (m/z 40–750 for metabolites), and analyzed by Agilent Workstation Version 7.0.2 software. Aroma volatiles and metabolites were identified by comparing their fragmentation patterns with those from the NIST 2011 Mass Spectral Library and Software (National Institute of Standards and Technology, United States). Coinjection with authentic standards was undertaken to confirm tentative identifications where possible. Quantifications were based on standard curves generated for each target compound and internal standards.

#### RNA Extraction

Total RNA was extracted from ∼3 g of outer pericarp tissue (three biological replications) using a method for polysacchariderich tissues (Ikoma et al., 1996), with slight modifications. DNase I (Nippon Gene, Tokyo) treatment followed by clean-up using FavorPrep after Tri-Reagent RNA Clean-up Kit (Favorgen Biotech Co., Pingtung, Taiwan) were carried out to remove genomic DNA contamination from the extracted RNA.

### Quantitative Real-Time PCR (RT-qPCR)

The method used was similar to that reported in our previous study (Asiche et al., 2018). Briefly, first strand cDNA was synthesized from 2.4 µg of RNA using RevTra Ace reverse transcriptase (Toyobo, Osaka, Japan), and random hexamer primer according to the manufacturer's instructions. Genespecific primers (**Supplementary Table 1**) for RT-qPCR analysis were designed using Primer3 software (version 0.4.0<sup>1</sup> ). Gene expression of three biological replications was examined on a MyiQ Single-Color Reverse Transcriptase-Quantitative PCR Detection System (Bio-Rad, Hercules, CA, United States) using TB GreenTM Premix Ex TaqTM II (Tli RNaseH Plus) (TaKaRa, Shiga, Japan) according to the manufacturer's instructions. AcActin was used as the housekeeping gene. The specificity of all primers was verified by melting curve analysis. Relative gene expression was calculated using the 2−11Ct method with samples at harvest (0 d) calibrated as 1.

#### Statistical Analysis

Data obtained in this study were subjected to statistical analyses using R software (version 3.4.0, R Project). ANOVA followed by post hoc Tukey's tests (p ≤ 0.05) were used to detect differences in fruit ripening characteristics and gene expression levels among the different treatments.

#### <sup>1</sup>http://bioinfo.ut.ee/primer3-0.4.0/

RESULTS

### Induction of Ethylene Biosynthesis in Postharvest Kiwifruit

At harvest, kiwifruit used in this study did not produce any detectable ethylene levels. To understand the molecular mechanisms responsible for ethylene production during postharvest handling, we determined the effects of propylene and 1-MCP treatments, as well as storage temperature (**Figure 1**).

As expected, endogenous climacteric ethylene production was detected in propylene-treated fruit at a level of 0.23 nLg−<sup>1</sup> h −1 after 5 d, increasing to 0.63 nLg−<sup>1</sup> h −1 after 9 d (**Figure 1A**). No endogenous ethylene production was measured in fruit pre-treated with 1-MCP, nor in the control fruit, throughout the experimental period. The expression of ethylene biosynthetic genes, AcACS1 and AcACO2, significantly increased in propylene-treated fruit after 5 and 9 d by 123–176- and 143–179-fold, respectively (**Figures 1B,C**). AcACS1 expression showed no measurable changes in fruit pre-treated with 1-MCP, while AcACO2 expression was significantly reduced to <60-fold. During storage, no endogenous ethylene production was measured in healthy intact fruit either at 5 or 22◦C throughout the experimental period (**Figure 1D**). There were no transcriptional changes observed in AcACS1 (**Figure 1E**), while AcACO2 showed a considerable expression increase (65-fold) in fruit at 5◦C after 49 d (**Figure 1F**). These observations indicate that production of climacteric ethylene in kiwifruit requires ethylene signaling (triggered by propylene), and is largely dependent upon the transcriptional regulation of AcACS1.

### Kiwifruit Softening Is Inducible by Either Ethylene or Low Temperature

Following propylene treatment, kiwifruit firmness rapidly decreased from 73 to 10 N within 2 d, and to <3 N after 5 d (**Figure 2A**). Fruit pre-treated with 1-MCP showed only a slight decrease in firmness, to 50 N after 9 d. The cell wall modification-associated genes, AcPG and AcEXP1, showed a marked increase in expression (>1000- and 71-fold, respectively) during propylene treatment after 5 and 9 d (**Figures 2B,C**). The induction of both genes by propylene was significantly reduced by 1-MCP pre-treatment. Control fruit showed no significant changes in firmness, and the expression of both AcPG and AcEXP1 was maintained at minimal levels throughout the experimental period.

The firmness of fruit stored at 5◦C showed a substantial decrease from 73 to 24 N after 21 d, and to 3 N after 49 d (**Figure 2D**). At 22◦C, fruit firmness showed only a slight decrease to 45 N after 21 d with no further changes thereafter. Both AcPG and AcEXP1 expression showed sustained increases in fruit at 5◦C, whereas their expression in fruit at 22◦C were maintained at low levels throughout the storage period (**Figures 2E,F**).

FIGURE 1 | Ethylene production and expression patterns of ethylene biosynthetic genes in kiwifruit. (A) Ethylene production pattern during ethylene-dependent ripening. Control: non-treated; propylene: continuously treated with propylene (5000 µLL−<sup>1</sup> , 22◦C) for up to 9 d; 1-MCP+Propylene: a single exposure to 1-methylcyclopropene (1-MCP, 2 µLL−<sup>1</sup> ) for 12 h immediately after harvest, followed by continuous propylene treatment for up to 9 d. (D) Ethylene production during low temperature-modulated ripening. Kiwifruit immediately after harvest were separately stored at either 5 or 22◦C in ethylene-free air for up to 49 d. Relative transcript levels of ethylene biosynthetic genes, 1-aminocyclopropane-1-carboxylate (ACC) synthase 1 (AcACS1, Achn364251) and ACC oxidase 2 (AcACO2, Achn326461), were determined against at-harvest (0 d) samples by RT-qPCR using kiwifruit actin (EF063572) as an endogenous control (B,C,E,F). Data are mean (±SE) of at least three independent biological replications. Error bars not shown are smaller than the symbol used. Different letters indicate significant differences in ANOVA (Tukey's test, p < 0.05).

### Changes in the Composition of Major Soluble Sugars

The SSC of propylene-treated fruit increased rapidly from 7.8% to a maximum 16.5% after 5 d (**Figure 3A**). This increase was significantly delayed by 1-MCP pre-treatment, with fruit showing little change for the first 5 d; a substantial increase to 13.2% was observed after 9 d. There was a rapid increase in sucrose content of propylene-treated fruit, from 0.7 to 33.1 mg gFW−<sup>1</sup> after 5 d, also followed by a sudden decrease to 8.9 mg gFW−<sup>1</sup> after 9 d (**Figure 3B**). Both glucose and fructose contents showed a sustained increase in propylene-treated fruit to 40 and 36 mg gFW−<sup>1</sup> , respectively, after 9 d (**Figures 3C,D**). The increase in sucrose, glucose, and fructose contents by propylene treatment was significantly delayed by 1-MCP pre-treatment, although the fruit eventually attained almost similar contents at the end of the experimental period. No measurable changes in

expansin 1 (AcEXP1, Achn336951) were determined against at-harvest (0 d) samples by RT-qPCR using kiwifruit actin (EF063572) as an endogenous control (B,C,E,F). Data are mean (±SE) of at least three independent biological replications. Error bars not shown are smaller than the symbol used. Different letters indicate significant differences in ANOVA (Tukey's test, p < 0.05).

SSC, as well as the contents of sucrose, glucose, and fructose, were observed in non-treated fruit. During storage, SSC and the contents of glucose and fructose exhibited a sustained increase with no significant differences between fruit at 5 and 22◦C (**Figures 3E,G,H**), while the sucrose content rose sharply in fruit at 5◦C to a maximum 20.4 mg gFW−<sup>1</sup> after 21 d followed by a decrease to 7.2 mg gFW−<sup>1</sup> after 49 d (**Figure 3F**). The sucrose content steadily increased in fruit at 22◦C to 33.5 mg gFW−<sup>1</sup> after 49 d.

Since the content of soluble sugars was significantly affected by propylene and storage, we examined the expression patterns of genes associated with starch degradation (Acβ-AMY1 and

1-MCP+Propylene: a single exposure to 1-MCP (2 µLL−<sup>1</sup> ) for 12 h immediately after harvest, followed by continuous propylene treatment for up to 9 d. (E–H) Changes in SSC and concentration of sucrose, glucose, and fructose during low temperature-modulated ripening. Kiwifruit immediately after harvest were stored at either 5 or 22◦C. Data are mean (±SE) of five independent biological replications. Error bars not shown are smaller than the symbol used. Different letters indicate significant differences in ANOVA (Tukey's test, p < 0.05).

Acβ-AMY2) and sucrose metabolism (AcINV3-1) (**Figure 4**). Acβ-AMY1 and Acβ-AMY2 both showed an increase in expression in propylene-treated fruit as the soluble sugars increased, while they were significantly suppressed in fruit pretreated with 1-MCP (**Figures 4A,B**). During storage, there was a small increase in Acβ-AMY1 expression in fruit at 5 ◦C after 49 d, whereas no significant expression changes were observed in fruit at 22◦C (**Figure 4D**). On the other hand, Acβ-AMY2 expression increased in fruit at both 5 and 22◦C, to a maximum (five to sevenfold) after 21 d, and later decreased after 49 d (**Figure 4E**). AcINV3-1 showed only a small expression increase (twofold) in propylene-treated fruit after 5 d, while no significant changes were observed in fruit pre-treated with 1-MCP (**Figure 4C**). However, AcINV3- 1 expression significantly increased (by fivefold after 21 d, and fourfold after 49 d) during storage of fruit at 5◦C; no

measurable changes in expression were recorded in fruit at 22◦C (**Figure 4F**).

#### Aroma Volatile Production Is Strongly Ethylene-Dependent, and Is Undetectable During Low Temperature-Modulated Fruit Ripening

Fruit aroma volatiles were identified and quantified by GC– MS during ethylene-dependent ripening, after 49 d storage at either 5 or 22◦C, and during 14 d shelf life (at 22◦C) (**Figure 5**). The major volatiles detected were esters; particularly ethyl butanoate and methyl butanoate (**Supplementary Figures 1**, **2**), which are considered to form part of the characteristic ripe fruit flavor for kiwifruit (Zhang et al., 2009; Atkinson et al., 2011). Propylene-treated fruit produced large amounts of ethyl butanoate at a rate of 4.2 ngg−<sup>1</sup> h −1 after 5 d, and 20.9 ngg−<sup>1</sup> h −1 after 9 d (**Figure 5A**). Similarly, methyl butanoate was detected at a rate of 0.2 ngg−<sup>1</sup> h −1 in propylene-treated fruit after 5 d; no further increase was observed thereafter (**Figure 5B**). No measurable increase in the production of both esters was observed in fruit pretreated with 1-MCP, as well as in control fruit. During storage, ethyl butanoate and methyl butanoate levels were undetectable in fruit at either 5 or 22◦C, while they only increased during post-storage treatment with propylene (**Figures 5C,D**), consistent with endogenous ethylene production (**Supplementary Figure 3**).

(Tukey's test, p < 0.05).

We also examined the expression of AcAAT, which encodes an alcohol acyl transferase associated with ester production during fruit ripening (Souleyre et al., 2005; Günther et al., 2011). AcAAT showed a dramatic increase in expression in propylene-treated fruit after 5 d (∼2000-fold) and 9 d (∼11,000-fold), while no significant changes were observed in fruit pre-treated with 1- MCP (**Figure 6A**). By contrast, AcAAT showed no significant changes in expression during storage at either 5 or 22◦C, whereas it was massively up-regulated (>10,000-fold) 7 d after poststorage treatment with propylene (**Figure 6B**).

### Expression Analysis of Fruit Ripening-Associated Transcription Factors

As various TFs including MADS-box and NAC domains have been shown to be key regulators of fruit ripening in kiwifruit and other fruit species (Fujisawa et al., 2013; McAtee et al., 2015; Nieuwenhuizen et al., 2015), we determined the expression of three TF-encoding genes (**Figure 7**). AcNAC3 was considerably up-regulated (121- and 309-fold after 5 and 9 d, respectively) in propylene-treated fruit (**Figure 7A**). Its expression was significantly suppressed (64-fold) in fruit pre-treated with 1- MCP, while only a small increase (18-fold) was observed in the control fruit. During storage, AcNAC3 was also remarkably upregulated (>200-fold) in fruit at 5◦C both after 21 and 49 d, as well as after 7 d post-storage propylene treatment; only a small expression increase (<40-fold) was observed in fruit at 22◦C (**Figure 7D**). A second NAC-related gene, AcNAC5, showed an expression increase (53-fold) 9 d after propylene-treatment, while it was significantly inhibited (14-fold) by 1-MCP pre-treatment (**Figure 7B**). During storage, AcNAC5 showed no significant change in expression in fruit at either 5 or 22◦C, except in post-storage propylene-treated fruit where it was up-regulated (45-fold) after 7 d (**Figure 7E**). AcMADS2 showed a completely different expression pattern, as it showed no specific response to

propylene treatment (**Figure 7C**), while it was up-regulated in fruit at 5◦C after 21 d (17-fold) and 49 d (33-fold) (**Figure 7F**). There was no significant change in expression of AcMADS2 in fruit stored at 22◦C. It is also worth noting that AcMADS2 expression in fruit at 5◦C substantially dropped after transfer of the fruit to 22◦C.

### DISCUSSION

The ripening behavior of kiwifruit has elicited great interest, as substantial fruit softening during cold storage essentially occurs in the absence of any detectable ethylene; exogenous or endogenous (Antunes, 2007; Yin et al., 2009). In the present work, our results using "Kosui" kiwifruit are consistent with previous research; fruit softening and the expression of cell wall modification-associated genes were induced during storage at 5 ◦C (**Figure 2**), despite the lack of any measurable increase in ethylene production (**Figure 1**). It has been suggested that fruit softening during low temperature storage is brought about by basal levels of system I ethylene (Kim et al., 1999), based on the high sensitivity of kiwifruit to ethylene (Pranamornkith et al., 2012; Jabbar and East, 2016). However, the present work strongly suggests that ethylene signaling is non-functional during low temperature-modulated ripening in kiwifruit.

In climacteric fruit, the ethylene biosynthetic pathway is well known to be subject to positive feedback regulation (Kende, 1993). In tomato (Barry et al., 2000), pear (Hiwasa et al., 2003), and banana (Inaba et al., 2007), exposure of fruit to exogenous ethylene/propylene triggers a sharp increase in ethylene production with increased expression of ethylene biosynthetic genes. In kiwifruit, the up-regulation of ethylene biosynthetic genes, AcACS1 and AcACO2, by propylene exposure, coupled with their significant inhibition by 1-MCP (**Figures 1B,C**), strongly indicate that ethylene biosynthesis is under the influence of ethylene signaling. These observations agree with previous reports suggesting that ethylene biosynthesis during kiwifruit ripening is regulated by a positive feedback mechanism (Whittaker et al., 1997; Mworia et al., 2010). The expression pattern of AcACS1 suggests that it is strongly ethylenedependent, as 1-MCP pre-treatment absolutely inhibited its induction by propylene (**Figure 1B**). Of great importance is the fact that there was no measurable increase in AcACS1 expression throughout storage of fruit at either 5 or 22◦C (**Figure 1E**), which would account for the undetectable ethylene production (**Figure 1D**). Therefore, lack of ethylene-dependent AcACS1 expression, together with undetectable ethylene production during kiwifruit storage, strongly advocate for the idea that ethylene signaling is non-existent. The increase in AcACO2 expression after 49 d at 5◦C (**Figure 1F**) is insignificant, since it has been shown using tomato that ACS regulates the rate-limiting step in ethylene biosynthesis (Yip et al., 1992; Wang et al., 2002).

It has been shown in previous studies that during fruit ripening, the production of aroma volatiles (especially esters) is regulated by the ethylene signaling pathway. Aroma volatile production was strongly inhibited in ethylene-suppressed kiwifruit (Atkinson et al., 2011), melon (Pech et al., 2008), and apple (Defilippi et al., 2004; Schaffer et al., 2007) lines.

determined against at-harvest (0 d) samples by RT-qPCR using kiwifruit actin (EF063572) as an endogenous control. Data are mean (±SE) of three independent biological replications. Error bars not shown are smaller than the symbol used. Different letters indicate significant differences in ANOVA (Tukey's test, p < 0.05).

Additionally, transgenic lines treated with exogenous ethylene produced increasing concentrations of aroma volatile compounds (Schaffer et al., 2007; Atkinson et al., 2011). Another study by Defilippi et al. (2005) further demonstrated that the expression of MdAAT, together with the activity of the associated enzyme, is a rate-limiting step in ester biosynthesis in apple fruit, and both are regulated by ethylene. In the present work, propylene treatment induced the expression of AcAAT, together with the production of ethyl butanoate and methyl butanoate in kiwifruit (**Figures 5**, **6**), and their complete suppression by 1-MCP confirmed that they are strongly regulated by the ethylene pathway. It is interesting that the production of these aroma volatiles, as well as the induction of AcAAT expression, was not observed in fruit during storage at either 5 or 22◦C, further arguing for the lack of ethylene signaling during low temperature-modulated ripening.

Fruit SSC and the concentrations of sucrose, glucose, and fructose increased in response to propylene, as well as during storage at both 5 and 22◦C (**Figure 3**). These changes coincided with increased expression of Acβ-AMY1 and Acβ-AMY2 (**Figures 4A,B,D,E**), which have been previously linked to starch degradation and sugar accumulation in kiwifruit (Nardozza et al., 2013; McAtee et al., 2015; Hu et al., 2016). The above observations suggest that changes in soluble sugars might involve regulatory mechanisms that are independent of both ethylene and low temperature. This is consistent with previous research using different kiwifruit cultivars (Arpaia et al., 1987; Boquete et al., 2004; Mworia et al., 2012; Asiche et al., 2018). However, Acβ-AMY1 appears to have a stronger response to ethylene since its expression during storage is relatively low (**Figures 4A,D**). Previous studies in tomatoes (Gao et al., 2007), melons (Pech et al., 2008), and apples (Defilippi et al., 2004) have also demonstrated that there is an ethylene-independent component in starch metabolism and sugar accumulation during fruit ripening. Nevertheless, the expression pattern of AcINV3- 1 suggests that it is more aligned to low temperature response than to ethylene, since its expression increased markedly in fruit at 5◦C, but only slightly in response to propylene (**Figures 4C,F**).

The distinction between ethylene-induced and low temperature-modulated ripening has been difficult to accomplish due to the existence of genes that respond to both stimuli, such as AcPG and AcEXP1 (**Figure 2**). The expression of AcNAC3 was also up-regulated by both propylene and low temperature (**Figures 7A,D**), suggesting its potential role in the regulation of both ethylene-induced and low temperature-modulated ripening. However, its induction during low temperature storage is likely to be independent of ethylene, since it was previously shown that inhibiting ethylene signaling by 1-MCP failed to suppress its upregulation at 5◦C (Asiche et al., 2018). NAC-related TFs are involved in regulation of several ripening-associated genes. In tomato, SlNAC4 was shown to regulate fruit ripening and carotenoid accumulation (Zhu et al., 2014), while in banana, several NAC TFs are induced during fruit ripening and are known to physically interact with ethylene insensitive 3 (EIN3) like (EIL), a major component in the ethylene signaling pathway (Shan et al., 2012). In the present study, the observation that AcNAC5 was exclusively induced by propylene and not by low temperature (**Figures 7B,E**) challenges the notion that ethylene

signaling is functional during low temperature-modulated fruit ripening. AcNAC5 expression correlates well with aroma volatile production patterns, suggesting its potential role in regulation of AcAAT and aroma volatile biosynthesis during ethylene-dependent ripening in kiwifruit. By contrast, AcMADS2 expression was substantially induced at 5◦C whereas propylene treatment failed to affect its transcript levels (**Figures 7C,F**), suggesting its potential role in regulation of fruit ripening during cold storage. Our previous studies have also demonstrated the exclusive induction of AcMADS2 by low temperature in different kiwifruit cultivars (Mitalo et al., 2018, 2019), confirming its alignment to regulatory mechanisms associated with low temperature response. However, since aroma volatiles were undetectable in fruit ripened by low temperature (**Figures 5C,D**), it appears that AcMADS2 does not have the capacity to bind or activate AcAAT or other aroma-related genes which essentially belong to the ethylene pathway.

In summary, the present work has demonstrated that kiwifruit ripening is inducible independently by either ethylene or low temperature signals (**Figure 8**). Fruit ripened by either stimulus can attain similar quality characteristics in terms of firmness and soluble sugar levels. However, production of aroma volatiles (especially esters: ethyl butanoate and methyl butanoate) and the expression of AcAAT appear to be strongly dependent on the ethylene signal. These ethylene-dependent components show negligible changes during low temperature-modulated fruit ripening, providing evidence for the absence of ethylene signaling during cold storage. A distinct group of TFs such as those encoded by AcNAC5 are exclusively induced by ethylene, suggesting their involvement in regulating ethylene-induced ripening, while a second group encoded by genes such as AcMADS2 are exclusively aligned to low temperature response. Therefore, it appears that ethylene-induced and low

#### REFERENCES


temperature-modulated ripening in kiwifruit involve distinct regulatory mechanisms.

## DATA AVAILABILITY

All datasets (generated/analyzed) for this study are included in the manuscript and/or the **Supplementary Files**.

## AUTHOR CONTRIBUTIONS

OM, ST, and YKu conceived and designed the study. OM, ST, YKo, TO, and AD performed most of the experiments with close supervision from YKu, IG, RN, and KU. IG did the GC–MS analysis. KS and IK provided technical assistance. OM wrote the first draft of the manuscript. YKu and IG substantially improved the first draft of the manuscript. All the authors have read and approved the submitted version.

#### FUNDING

This work was supported in part by a Grant-in-Aid for Scientific Research (grant nos. 24380023, 25660027, and 16H04873 to YKu) from the Japan Society for the Promotion of Science, Japan and also partially by the Joint Usage/Research Center, Institute of Plant Science and Resources, Okayama University.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.00888/ full#supplementary-material


tomato fruits. Russ. J. Plant Physiol. 54, 80–88. doi: 10.1134/s102144370701 0128


homologs under various conditions. Proc. Nat. Acad. Sci. U.S.A. 89, 2475–2479. doi: 10.1073/pnas.89.6.2475


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Mitalo, Tokiwa, Kondo, Otsuki, Galis, Suezawa, Kataoka, Doan, Nakano, Ushijima and Kubo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Coordination of Ethylene and Other Hormones in Primary Root Development

*Hua Qin1,2 , Lina He3 and Rongfeng Huang1,2 \**

*1 Biotechnology Research Institute, Chinese Academy of Agricultural Sciences, Beijing, China, 2 National Key Facility of Crop Gene Resources and Genetic Improvement, Beijing, China, 3 College of Biological Sciences and Biotechnology, Beijing Forestry University, Beijing, China*

The primary root is the basic component of root systems, initiates during embryogenesis and develops shortly after germination, and plays a key role in early seedling growth and survival. The phytohormone ethylene shows significant inhibition of the growth of primary roots. Recent findings have revealed that the inhibition of ethylene in primary root elongation is mediated *via* interactions with phytohormones, such as auxin, abscisic acid, gibberellin, cytokinins, jasmonic acid, and brassinosteroids. Considering that *Arabidopsis* and rice are the model plants of dicots and monocots, as well as the fact that hormonal crosstalk in primary root growth has been extensively investigated in *Arabidopsis* and rice, a better understanding of the mechanisms in *Arabidopsis* and rice will increase potential applications in other species. Therefore, we focus our interest on the emerging studies in the research of ethylene and hormone crosstalk in primary root development in *Arabidopsis* and rice.

#### *Edited by:*

*Angelos K. Kanellis, Aristotle University of Thessaloniki, Greece*

#### *Reviewed by:*

*Maren Müller, University of Barcelona, Spain Anna N. Stepanova, North Carolina State University, United States*

#### *\*Correspondence:*

*Rongfeng Huang rfhuang@caas.cn*

#### *Specialty section:*

*This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science*

*Received: 08 March 2019 Accepted: 19 June 2019 Published: 10 July 2019*

#### *Citation:*

*Qin H, He L and Huang R (2019) The Coordination of Ethylene and Other Hormones in Primary Root Development. Front. Plant Sci. 10:874. doi: 10.3389/fpls.2019.00874*

Keywords: ethylene, hormones, primary root, elongation, crosstalk

### INTRODUCTION

As an underground organ of plants, the root system plays a vital role in the absorption and translocation of water and nutrients. The root system generally consists of two principal root types: the primary root and secondary roots. Primary root is formed embryonically. It is the basic component of the root system that absorbs mineral nutrients and provides mechanical support for shoot growth in young seedlings (Zheng et al., 2016). The growth of primary roots is maintained by two basal developmental processes: cell proliferation in the root apical meristem (RAM) and cell elongation in the elongation zone. Phytohormones are central regulators of plant root growth and development. Multiple phytohormones, including ethylene, auxin, abscisic acid (ABA), gibberellin (GA), cytokinin (CK), jasmonic acid (JA), strigolactone (SL), and brassinosteroid (BR), have been shown to play vital roles in the regulation of primary root growth (Li et al., 2015; Pacifici et al., 2015; Hu et al., 2017; Qin and Huang, 2018).

Ethylene is a gaseous plant hormone that is synthesized from S-adenosylmethionine (SAM), which is converted to 1-aminocyclopropane-1-carboxylate (ACC) by ACC synthase (ACS). ACC is then converted to ethylene by ACC oxidase (ACO; Yang and Hoffman, 1984). Ethylene is perceived by a family of membrane-bound receptors (Hua et al., 1998; Hua and Meyerowitz, 1998). These receptors then inhibit the function of the Ser/Thr kinase CONSTITUTIVE TRIPLE RESPONSE1 (CTR1; Kieber et al., 1993), leading to the dephosphorylation and C-terminal cleavage of ETHYLENE INSENSITIVE2 (EIN2; Alonso et al., 1999; Ju et al., 2012; Qiao et al., 2012; Wen et al., 2012; Salehin and Estelle, 2015). The split EIN2 C-terminus translocates into the nucleus where it enables activation of the master transcriptional regulators EIN3 and EIN3-LIKE1 (EIL1), resulting in the induction of ethylene-triggered transcriptional response (Chao et al., 1997). In *Arabidopsis*, EIN2 is regulated by proteasomal degradation through EIN2 TARGETING PROTEIN1/2 (ETP1/2), whereas Maohuzi3 (MHZ3) protects OsEIN2 from proteasome-mediated degradation in rice (**Figure 1**; Qiao et al., 2009; Ma et al., 2017). Mutants with enhanced ethylene biosynthesis or signaling exhibit short primary roots (Kieber et al., 1993; Vogel et al., 1998; Woeste et al., 1999). By contrast, loss-of-function mutations of ethylene signaling components such as *ein2*, *ein3*, and *eil1* or treatment of seedlings with ethylene inhibitors such as the biosynthesis inhibitors 2-aminoethoxyvinylglycin (AVG; Ruzicka et al., 2007) and pyrazinamide (PZA; Sun et al., 2017), as well as the perception inhibitor silver nitrate (AgNO3; Ruzicka et al., 2007) and 1-methylcyclopropene (1-MCP; Serek et al., 1995) lead to increased primary root length and reduced ethylene response (Chao et al., 1997; Alonso et al., 1999; Ma et al., 2013; Yang et al., 2015).

An increasing number of investigations have revealed that ethylene affects primary root growth *via* two successive processes: cell proliferation and cell elongation (Ruzicka et al., 2007; Street et al., 2015). Cell proliferation occurs at the RAM, which is composed of a quiescent center (QC), surrounding stem cells, and a group of mitotically active cells. The QC plays important roles in maintaining stem cell populations within the root meristem (Ortega-Martinez et al., 2007; Heyman et al., 2013). Accumulating evidence has revealed that ethylene induces irregular transverse cell divisions in the QC (Ortega-Martinez et al., 2007; Ni et al., 2014). Furthermore, ethylene inhibition of cell proliferation at the RAM is mainly achieved by restricting epidermal cell expansion (Street et al., 2015; Vaseva et al., 2018). Moreover, ethylene also inhibits the elongation of cells in the elongation zone. Ethylene stimulates auxin biosynthesis and basipetal auxin transport toward the elongation zone, where it activates a local auxin response leading to inhibition of cell elongation (Ruzicka et al., 2007). In summary, ethylene inhibits primary root growth by regulating cell proliferation in the RAM and cell elongation in the elongation zone.

In addition to ethylene, other plant hormones are also involved in regulating primary root growth, and accumulating data have shown that plant hormones interact with each other to regulate primary root growth (Thole et al., 2014; Pacifici et al., 2015; Van de Poel et al., 2015; Harkey et al., 2018). As a model plant of dicots and monocots, the results obtained in *Arabidopsis* and rice can be instructive for other species.

FIGURE 1 | Crosstalk of ethylene and other hormones in the primary root growth of *Arabidopsis* and rice. In *Arabidopsis*, ethylene inhibits primary root growth by regulating auxin biosynthesis, transport, and signaling. EIN3/EIL1, ERF1, and HB52 function as crosstalk nodes between ethylene and auxin in this process. ABA promotes ethylene biosynthesis by affecting the posttranscriptional regulation of ACS. GA and ethylene antagonistically regulate the stability of DELLA proteins, which act as growth repressors. CKs induce ethylene biosynthesis by stabilizing ACS stability. Low concentrations of BRs inhibit ethylene biosynthesis by activating BZR1 and BES1 to repress the expression of *ACSs*. High levels of BRs induce ethylene biosynthesis through increasing the stability of ACSs. In rice, ethylene restricts primary root growth by increasing auxin and ABA biosynthesis. Auxin accumulation promotes SOR1-mediated degradation of OsIAA26, thus repressing normal root growth. MHZ3 stabilizes OsEIN2 to facilitate ethylene signal transduction. The solid lines indicate direct interactions, and the dashed lines indicate indirect interactions. The arrows indicate stimulatory effects, whereas the T sharp symbol indicates inhibitory effects.

In this current review, we summarize the current studies on the crosstalk of ethylene with other phytohormones during primary root development in *Arabidopsis* and rice, including auxin, ABA, GA, CKs, BRs, and JA, which contributes to better understanding of the role of ethylene in primary root development.

#### Interaction of Ethylene and Auxin in Primary Root Growth

As an omnipotent regulator of root development, auxin inhibits the primary root elongation, and this effect is most probably mediated in crosstalk with ethylene (Hu et al., 2017; Qin and Huang, 2018). Auxin and ethylene act synergistically in the regulation of primary root elongation (Ruzicka et al., 2007; Swarup et al., 2007; Qin et al., 2017; Zemlyanskaya et al., 2018), and this regulation is mainly through the modulation of cell proliferation in the RAM and cell elongation in the elongation zone (Ruzicka et al., 2007; Street et al., 2015). The transcription of auxin biosynthesis and transport genes is regulated by ethylene, and the treatment of ethylene promotes auxin accumulation in the root tip, as revealed by indole-3-acetic acid (IAA) measurements and *DR5* reporter expression (Stepanova et al., 2005; Ruzicka et al., 2007; Qin et al., 2017; Mendez-Bravo et al., 2019). Moreover, disrupting auxin biosynthesis by L-kynurenine or yucasin [5-(4-chlorophenyl)-4H-1,2,4-triazole-3-thiol] impairs the ethylene effect on primary root growth (He et al., 2011; Qin et al., 2017). *Via* analysis of mutants that showed reduced sensitivity to ethylene in the primary roots, a set of *weak ethylene insensitive* (*wei*) mutants was isolated, such as *wei2/ anthranilate synthase α1* (*asa1*), *wei7/anthranilate synthase β1* (*asb1*), and *wei8/*tryptophan aminotransferase of arabidopsis1 (taa1; Stepanova et al., 2005, 2008). *WEI2/ASA1* and *WEI7/ ASB1* encode α and β subunits of anthranilate synthase, respectively, which is a rate-limiting enzyme involved in the biosynthesis of the auxin precursor tryptophan (Trp; Stepanova et al., 2005). *WEI8/TAA1* encodes key enzymes involved in the indole-3-pyruvic acid (IPyA) pathway, which is the most important pathway for producing auxin in plants (Stepanova et al., 2008; Mashiguchi et al., 2011). In addition to the defects of auxin biosynthesis, mutations affecting auxin transport, perception, or signaling also result in reduced sensitivity to ethylene (Hu et al., 2017; Merchante and Stepanova, 2017; Qin and Huang, 2018). Ethylene is one of the important players that regulate the auxin flow (Luschnig et al., 1998; Ruzicka et al., 2007; Stepanova et al., 2007; Swarup et al., 2007). Abnormal levels of ethylene can result in an imbalance in the auxin gradient, which in turn leads to a higher auxin content, and ultimately result in the inhibition of cell elongation and primary root growth. Mutations in influx carriers of the AUXIN1 (AUX1) family or efflux carriers of the PIN-FORMED (PIN) family lead to ethylene-resistance root phenotype (Luschnig et al., 1998; Ruzicka et al., 2007; Stepanova et al., 2007; Swarup et al., 2007). Loss of function of *AUXIN RESISTANT2* (*AXR2*) and *AXR3*, which are involved in the auxin signaling pathway, results in reduced ethylene sensitivity in primary root (Stepanova et al., 2007; Swarup et al., 2007). Mutants of auxin receptor *transport inhibitor response1* (*tir1*) show an ethylene-insensitive primary root growth (**Figure 1**; Alonso et al., 2003).

With an established method for screening ethylene-response mutants in rice, several mutants with defective ethylene response in primary root were identified (Ma et al., 2013). Among these mutants, some auxin biosynthesis- and signaling-related mutants, such as *maohuzi2* (*mhz2/sor1*) and *rice ethyleneinsensitive7* (*rein7/yuc8*; **Figure 1**), have been identified (Qin et al., 2017; Chen et al., 2018). MHZ2, a RING finger E3 ubiquitin ligase, regulates ethylene response in primary roots by interacting with OsIAA26, an atypical Aux/IAA protein involved in the auxin signaling pathway (Chen et al., 2018). *REIN7* encodes an orthologue of *YUCCA8* (*YUC8*) and loss of *REIN7* function results in reduced auxin biosynthesis and ethylene insensitivity in primary roots (Qin et al., 2017). These studies suggest that ethylene-triggered inhibition of primary root elongation requires auxin biosynthesis, transport, and signaling, and the underlying mechanism is conserved in different species.

Emerging studies have shown that several transcription factors are involved in ethylene signaling or ethylene response, such as EIN3, the canonical transcription factor involved in the ethylene signaling pathway, regulating auxin biosynthesis by activating *YUC* transcripts (Liu et al., 2016; Qin et al., 2017). ETHYLENE RESPONSE FACTOR1 (ERF1), a direct target of EIN3, is responsive to ethylene and directly activates *ASA1*. ASA1 encodes a rate-limiting enzyme in Trp biosynthesis where auxin is derived. The activation of *ASA1* promotes auxin biosynthesis, results in the increased auxin accumulation, and thus inhibits primary root elongation (Solano et al., 1998; Mao et al., 2016). HOMEOBOX PROTEIN52 (HB52), which acts downstream of EIN3, binds to the promoters of *PIN2*, *WAVY ROOT GROWTH1* (*WAG1*), and *WAG2* to increase their expression. WAG1 and WAG2 phosphorylate PIN2, thus modulating auxin transport (**Figure 1**; Miao et al., 2018). These transcription factors act as nodes in the crosstalk between ethylene and auxin in primary root growth, implying that ethylene regulates primary root growth by modulating auxin biosynthesis, transport, and signaling.

### Coordination of Ethylene and Abscisic Acid in Primary Root Growth

ABA regulates many aspects of plant growth and development, including primary root elongation (Luo et al., 2014; Li et al., 2017; Sun et al., 2018). Studies have shown that ABA has biphasic effects on primary root growth, depending on its concentration, environmental conditions, developmental context, genotypes, and plant species. Typically, low concentrations of ABA stimulate primary root growth, whereas high concentrations inhibit it (Rowe et al., 2016; Li et al., 2017).

Increasing numbers of studies in *Arabidopsis* have suggested that ABA inhibits primary root growth mainly acting on cortical cells in the elongation zone, and this process requires the ethylene signaling pathway (Beaudoin et al., 2000; Ghassemian et al., 2000; Dietrich et al., 2017). The ethylene signaling mutant *ethylene response1* (*etr1-1*) and *ein2* showed resistance to ABA-inhibited primary root growth, whereas the ABA-resistant mutant *ABA-insensitive1* (*abi1-1*) and the ABA-deficient mutant *aba2* exhibited a normal ethylene response in the roots (Beaudoin et al., 2000; Ghassemian et al., 2000; Thole et al., 2014). Moreover, disruption of ABA biosynthesis or signaling in the *ein2*, *ein3,* or *ctr1* mutant background by introducing the *aba2* or *abi1* mutation did not alter the ethylene response phenotypes of the respective ethylene mutants (Beaudoin et al., 2000; Cheng et al., 2009). The above results suggest that ABA-inhibited primary root growth requires a functional ethylene signaling pathway but that ethylene-inhibited root growth is ABA-independent. Recent studies have indicated that ABA inhibits primary root growth by increasing ethylene biosynthesis (Luo et al., 2014). ABA activates two calciumdependent protein kinases, CPK4 and CPK11, which phosphorylate the C-terminus of ACS6 and increase the stability of ACS6 in ethylene biosynthesis, thus promoting the biosynthesis of ethylene (**Figure 1**). Disruption of ethylene biosynthesis by AVG relieves the inhibitory effect of ABA on primary root growth (Luo et al., 2014; Li et al., 2017). In addition, a protein phosphatase type 2C, ABI1, a negative regulator of ABA signaling, negatively regulates ethylene biosynthesis by counteracting the phosphorylation of ACS2/ACS6 mediated by MITOGENACTIVATED PROTEIN KINASE 6 (MAPK6; **Figure 1**; Ludwikow et al., 2014). Together, these results suggest that ABA treatment induces ethylene biosynthesis, thus leading to the inhibition of primary root growth, and that ABA responses require normal ethylene signaling.

ABA and ethylene crosstalk in rice have been revealed by two ethylene-response mutants: *mhz4* and *mhz5* (Ma et al., 2014; Yin et al., 2015). MHZ4 is a homologue of *Arabidopsis* ABA4, which is involved in ABA biosynthesis. *MHZ5* encodes a carotenoid isomerase, which is essential for ABA biosynthesis in etiolated rice shoots and roots. Mutations in *MHZ4* and *MHZ5* reduce the ethylene response in roots but enhance the ethylene response in the coleoptiles of etiolated seedlings. Overexpression of *MHZ4/5* results in enhanced ethylene sensitivity in roots and reduced ethylene sensitivity in coleoptiles. Genetic studies have revealed that the MHZ4/5-mediated ABA pathway in rice acts downstream of ethylene signaling to inhibit root growth (**Figure 1**), which is different from that in *Arabidopsis*, where ABA inhibits root growth through promoting ethylene biosynthesis, suggesting that different mechanisms have evolved in these two species. In addition, the MHZ4/5-mediated ABA pathway in rice acts upstream of ethylene signaling to control coleoptile growth (Ma et al., 2014; Yin et al., 2015), suggesting that the interaction between ethylene and ABA is distinctive in different tissues.

#### Integration of Ethylene and Gibberellins in Primary Root Growth

GAs also play an important role in primary root growth. Mutation in GA biosynthesis genes or disruption of GA biosynthesis by paclobutrazol, an inhibitor of GA biosynthesis, substantially reduces the rate of cell proliferation in the *Arabidopsis* root meristem (Achard et al., 2009; Ubeda-Tomas et al., 2009; Lee et al., 2012). GAs have been proposed to promote root growth in *Arabidopsis* by increasing the elongation of both dividing and post-mitotic endodermal cells, thereby indirectly controlling the division and elongation of other types of root cells and the overall root meristem size (Ubeda-Tomas et al., 2009).

GAs are perceived in the cell through a simple pathway that is regulated by a family of proteins known as DELLA proteins, a subfamily of the GRAS family of putative transcriptional regulators (Richards et al., 2001; Daviere and Achard, 2013). In *Arabidopsis*, the DELLA family comprises GA-INSENSITIVE (GAI), REPRESSOR OF GA (RGA), RGA-LIKE1 (RGL1), RGL2, and RGL3 (Peng et al., 1997; Silverstone et al., 1998). GAs promote the degradation of DELLA proteins. However, ethylene treatment delays the GA-induced diminishing of green fluorescent protein (GFP)-RGA fusion constructs from root cell nuclei *via* a CTR1-dependent signaling pathway (**Figure 1**). Moreover, mutations in GAI and RGA reduce the sensitivity to ethylene in primary root growth, and exogenous GA treatment substantially releases the inhibition by ethylene of primary root growth (Achard et al., 2003), suggesting that ethylene and GA act antagonistically in primary root growth.

In rice, GA participates in the modulation of cell elongation and proliferation in the root meristem, GA deficiency leads to short primary roots (Li et al., 2015), and exogenous applications of GA can increase primary root growth and ethylene production (Lee and Yoon, 2018). *Via* an analysis of the ethylene-responsive genes in the roots of *mhz7/osein2* and *mhz6/oseil1* compared with the wild type by RNA sequencing, some genes are involved in GA biosynthesis and the catabolic pathway has been identified (Yang et al., 2015). Considering that ethylene inhibits primary root growth, while GA promotes it, ethylene may reduce endogenous GA contents by the transcriptional regulation of GA biosynthesis and catabolic genes in primary roots.

#### Interactions Between Ethylene, Cytokinins, Jasmonic Acid, and Brassinosteroids in Primary Root Growth

CKs are widely considered to inhibit primary root growth in *Arabidopsis* (Cary et al., 1995; Riefler et al., 2006; Ioio et al., 2012). Accumulating studies showed that CKs regulate primary root growth by interacting with ethylene, and this interaction mainly occurs in cell proliferation in the RAM (Street et al., 2015; Van de Poel et al., 2015; Liu et al., 2017). In *Arabidopsis*, treatment of etiolated seedlings with CKs produces a phenotype similar to ethylene (Cary et al., 1995), suggesting that CKs and ethylene may act synergistically to regulate specific growth and developmental processes. Further research shows that CK treatment increases the stability of the ethylene biosynthesis protein ACS5, leading to increased ethylene production (**Figure 1**; Chae et al., 2003; Hansen et al., 2009). Proteome analysis in *Arabidopsis* roots reveals that CKs upregulate the majority of proteins in the ethylene biosynthesis pathway (Zdarska et al., 2013). These results suggest that CKs inhibit primary root growth through enhancing ethylene biosynthesis. However, Kushwah et al. (2011) reported that CKs can induce primary root elongation, and this response is mediated by the ethylene signaling pathway through the ethylene receptor ETHYLENE RESISTANT1 (ETR1) and its downstream signaling element EIN2, suggesting that the ethylene signaling pathway is required for CK-induced primary root growth response. In rice, recent research suggests that exogenous CKs upregulate the transcription of ethylene biosynthesis genes in primary roots, resulting in increased ethylene biosynthesis and inhibition of primary root growth (Zou et al., 2018).

The function of JA in plant resilience to many environmental challenges has been well studied, and its role in root growth has also been reported (Wasternack and Hause, 2013; Huang et al., 2017; Guo et al., 2018). Exogenous application of JA inhibits various aspects of seedling growth, including primary root growth (Wasternack and Hause, 2013; Huang et al., 2017). In response to the JA signal, the F-box protein CORONATINE INSENSITIVE1 (COI1) recruits JASMONATE ZIM-domain (JAZ) repressors for ubiquitination and degradation, thereby relieving the repression of transcription factors and enabling the expression of JA-responsive genes and JA responses (Thines et al., 2007; Katsir et al., 2008; Fernandez-Calvo et al., 2011; Qi et al., 2015). Complicated modes of interaction between JA and ethylene have been investigated in different processes. JA enhances aluminuminduced primary root growth inhibition, and this process is controlled by ethylene. Aluminum-induced upregulation of pCOI1:COI1-VENUS in the root apex transition zone was significantly repressed in the *ein3eil1* double mutant, suggesting that JA acts downstream of ethylene in aluminum-induced primary root growth inhibition (Yang et al., 2017). In addition, JA and ethylene act synergistically to regulate root hair development, through direct interaction of JASMONATE ZIM-DOMAIN (JAZ) proteins, which are degraded after JA treatment, with EIN3/EIL1 to attenuate its transcriptional activity (Zhu et al., 2006, 2011). The above studies suggest that different mechanisms are present in different tissues, and additional studies are needed to elucidate the involvement of JA-ethylene crosstalk in primary root growth.

BRs, a class of plant-specific steroid hormones, play important roles in regulating primary root growth. Low concentrations of BRs can induce root growth, while high concentrations inhibit root growth (Clouse et al., 1996; Mussig et al., 2003; Lv et al., 2018). Mutations in BR biosynthesis or signaling pathway exhibit reduced cell proliferation in the RAM and decreased cell length in the mature zone, thus leading to shortened roots (Li et al., 1996; Mussig et al., 2003; Lv et al., 2018). BRs, perceived by the BRASSINOSTEROID INSENSITIVE1 (BRI1) receptor, activate the transcription factors BRI1-EMS-SUPPRESSOR1 (BES1) and BRASSINOSTEROID RESISTANT1 (BZR1), which in turn induce the BR response (Yin et al., 2002; Yu et al., 2011; Oh et al., 2012). Several studies have shown that BRs regulate primary root growth through modulating ethylene biosynthesis, namely, low levels of BRs inhibit ethylene biosynthesis by activating BZR1 and BES1 to repress the expression of *ACSs* (Lv et al., 2018). Although high levels of BRs induce ethylene biosynthesis by increasing the stability of ACSs (**Figure 1**), these dual effects on ethylene lead to biphasic effects of BRs on primary root growth (Lim et al., 2002; Hansen et al., 2009; Lv et al., 2018).

### CONCLUSIONS AND PERSPECTIVES

During the past few years, multiple studies have provided molecular connections between ethylene and primary root growth. In this review, we integrated recent data showing the molecular details of the interactions between ethylene and other plant hormones in the regulation of primary root development (**Figure 1**). In particular, ethylene regulates primary root growth through modulating auxin biosynthesis, transport, and signaling, and this regulatory mechanism is conserved in *Arabidopsis* and rice. ABA and CKs inhibit primary root growth in *Arabidopsis* by affecting the posttranscriptional regulation of ACS, which results in increased ethylene synthesis. However, in rice, ethylene inhibition of primary root growth requires ABA function, and this mode of ethylene-ABA interaction is fundamentally different from that in *Arabidopsis*, suggesting that different mechanisms exist in different species. GA and ethylene antagonistically regulate the stability of DELLA proteins, which act as growth repressors. BRs have biphasic effects on primary root growth, namely, low concentrations of BRs inhibit ethylene biosynthesis by activating BZR1 and BES1 to repress the expression of *ACSs*, whereas high levels of BRs induce ethylene biosynthesis through increasing the stability of ACSs. Based on the above results, a general interaction model where ethylene plays a central role is presented (**Figure 1**). Thus, this review provides an overview on the crosstalk of ethylene and other hormones in primary root growth, increasing our understanding of the regulation of ethylene in primary root growth.

Primary root growth begins during embryo development and is easily affected by adverse soil conditions. Plant hormones act as all-encompassing regulators of normal root growth and mediate root morphological responses to abiotic stresses (De Smet et al., 2015). Further investigations should focus on how plants perceive external changes and translate cues into adaptive responses by modulating endogenous hormone crosstalk dynamics. Moreover, there is a complex regulatory network among phytohormones, and different regulatory mechanisms exist in different tissues, at different developmental stages, and in different species (Luo et al., 2014; Ma et al., 2014; Yin et al., 2015; Abts et al., 2017). Thus, researchers should try to understand hormone crosstalk in a multidimensional space, and the development of effective hormone detection methods and computational models will greatly promote research on hormone crosstalk.

With the advancement of technology, high-throughput technology provides an opportunity to track complex regulatory pathways during primary root growth for detecting the molecular mechanisms that govern hormone crosstalk and the nodes of interactions between different hormone pathways (Hung and Weng, 2017). However, caution and further validation are required when using those data. In addition, it is also interesting to clarify the regulatory network of ethylene and other hormones in multiple cell types. Single-cell transcriptome sequencing can provide us with an overview of ethylene and other hormone actions in specific cells (Ziegenhain et al., 2017;

Denyer et al., 2019), and cell type-specific promoters allow us to study the function of a desired protein in particular cell types and to reveal the main sites of ethylene and other hormone actions (Vaseva et al., 2016).

#### AUTHOR CONTRIBUTIONS

All the authors discussed and created the review's outline. HQ wrote the manuscript. RH edited the manuscript.

#### REFERENCES


#### FUNDING

This work was funded by the National Natural Science Foundation of China grant numbers 31801445 and 31871551, the Major Special Foundation of Transgenic Plants in China grant number 2018ZX0800913B, the China Postdoctoral Science Foundation grant numbers 2017M620964 and 2018T110161, the Fundamental Research Funds for Central Non-profit Scientific Institution grant number 1610392019004, and the Innovation Project of Chinese Academy of Agricultural Sciences.


cleaved EIN2 carboxyl terminus. *Cell Res.* 22, 1613–1616. doi: 10.1038/ cr.2012.145


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Qin, He and Huang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Cyanobacteria Respond to Low Levels of Ethylene

Cidney J. Allen<sup>1</sup> , Randy F. Lacey<sup>1</sup>† , Alixandri B. Binder Bickford<sup>2</sup> , C. Payton Beshears<sup>1</sup> , Christopher J. Gilmartin<sup>3</sup> and Brad M. Binder<sup>1</sup> \*

<sup>1</sup> Department of Biochemistry & Cellular and Molecular Biology, The University of Tennessee, Knoxville, Knoxville, TN, United States, <sup>2</sup> West High School, Knoxville, TN, United States, <sup>3</sup> Farragut High School, Knoxville, TN, United States

Ethylene is a gas that has long been known to act as a plant hormone. We recently showed that a cyanobacterium, Synechocystis sp. PCC 6803 (Synechocystis) contains an ethylene receptor (SynEtr1) that regulates cell surface and extracellular components leading to altered phototaxis and biofilm formation. To determine whether other cyanobacteria respond to ethylene, we examined the effects of exogenous ethylene on phototaxis of the filamentous cyanobacterium, Geitlerinema sp. PCC 7105 (Geitlerinema). A search of the Geitlerinema genome suggests that two genes encode proteins that contain an ethylene binding domain and Geitlerinema cells have previously been shown to bind ethylene. We call these genes GeiEtr1 and GeiEtr2 and show that in air both are expressed. Treatment with ethylene decreases the abundance of GeiEtr1 transcripts. Treatment of Geitlerinema with 1000 nL L−<sup>1</sup> ethylene affected the phototaxis response to white light as well as monochromatic red light, but not blue or green light. This is in contrast to Synechocystis where we previously found ethylene affected phototaxis to all three colors. We also demonstrate that application of ethylene down to 8 nL L−<sup>1</sup> stimulates phototaxis of both cyanobacteria as well as biofilm formation of Synechocystis. We formerly demonstrated that the transcript levels of slr1214 and CsiR1 in Synechocystis are reduced by treatment with 1000 nL L−<sup>1</sup> ethylene. Here we show that application of ethylene down to 1 nL L−<sup>1</sup> causes a reduction in CsiR1 abundance. This is below the threshold for most ethylene responses documented in plants. By contrast, slr1214 is unaffected by this low level of ethylene and only shows a reduction in transcript abundance at the highest ethylene level used. Thus, cyanobacteria are very sensitive to ethylene. However, the dose-binding characteristics of ethylene binding to Geitlerinema and Synechocystis cells as well as to the ethylene binding domain of SynEtr1 heterologously expressed in yeast, are similar to what has been reported for plants and exogenously expressed ethylene receptors from plants. These data are consistent with a model where signal amplification is occurring at the level of the receptors.

Keywords: cyanobacteria, ethylene receptor, ethylene binding, Synechocystis, Geitlerinema, phototaxis, biofilm

### INTRODUCTION

Ethylene is an important plant hormone that affects plant growth, development, and responses to many stresses (Mattoo and Suttle, 1991; Abeles et al., 1992). Ethylene receptors in plants have been studied for many decades and much is known about how they bind ethylene and signal to downstream signaling proteins (Merchante et al., 2013; Lacey and Binder, 2014; Bakshi et al., 2015).

#### Edited by:

Caren Chang, University of Maryland, College Park, United States

#### Reviewed by:

Annegret Wilde, University of Freiburg, Germany Shangwei Zhong, Peking University, China

> \*Correspondence: Brad M. Binder bbinder@utk.edu

#### †Present address:

Randy F. Lacey, Department of Microbiology and Biochemistry, School of Biological Sciences, Victoria University of Wellington, Wellington, NZ, United States

#### Specialty section:

This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science

Received: 25 March 2019 Accepted: 08 July 2019 Published: 30 July 2019

#### Citation:

Allen CJ, Lacey RF, Binder Bickford AB, Beshears CP, Gilmartin CJ and Binder BM (2019) Cyanobacteria Respond to Low Levels of Ethylene. Front. Plant Sci. 10:950. doi: 10.3389/fpls.2019.00950

**67**

These receptors, as well as several other plant hormone receptors, have homology to bacterial two-component receptors that signal via a histidine autophosphorylation followed by phosphotransfer to downstream targets (Chang et al., 1993; Schaller et al., 2011; Kabbara et al., 2018). Several research groups have proposed that plants acquired these two-component-like receptors from the cyanobacterium that gave rise to chloroplasts where the freeliving cyanobacterium became an endosymbiont and most of the bacterial genome was acquired by the host cell (Kehoe and Grossman, 1996; Martin et al., 2002; Mount and Chang, 2002; Timmis et al., 2004; Schaller et al., 2011). In support of this, some cyanobacteria contain predicted ethylene binding proteins and several cyanobacteria species have been documented to bind ethylene (Rodriguez et al., 1999; Mount and Chang, 2002; Wang et al., 2006).

Many microorganisms respond to ethylene (Abeles et al., 1992; Bakshi and Binder, 2018). However, it was only recently that a non-plant species, Synechocystis sp. PCC 6803 (hereafter referred to as Synechocystis), was documented to contain a functional ethylene receptor (Lacey and Binder, 2016). In this unicellular cyanobacterium, the receptor is encoded by the slr1212 gene locus and has been variously referred to as Ethylene response 1 (SynEtr1) (Kaneko et al., 1996; Ulijasz et al., 2009; Lacey and Binder, 2016), His-kinase44 (Hik44) (Los et al., 2008), Positive phototaxisA (Narikawa et al., 2011) (PixA), and UV intensity response Sensor (UirS) (Song et al., 2011; Ramakrishnan and Tabor, 2016). It is the only gene in the Synechocystis genome predicted to encode a protein with an ethylene binding domain. The protein contains a functional ethylene binding domain at the N-terminus followed by a phytochrome-like domain known as a cyanobacteriochrome, and a his-kinase domain at the C-terminus (Yoshihara et al., 2004; Ikeuchi and Ishizuka, 2008; Ulijasz et al., 2009; Narikawa et al., 2011; Song et al., 2011; Lacey and Binder, 2016; Ramakrishnan and Tabor, 2016). Thus, this is a receptor for both light and ethylene. Various studies have also delineated components of the signaling pathway downstream of the receptor (Narikawa et al., 2011; Song et al., 2011; Lacey and Binder, 2016; Ramakrishnan and Tabor, 2016). From these studies a model for light signal transduction from SynEtr1 has developed (Ramakrishnan and Tabor, 2016) where UV-A light stimulates histidine autophosphorylation followed by phosphotransfer to a conserved aspartate on the slr1213 response regulator protein. The phosphorylated slr1213 enhances the transcription of a small, non-coding RNA called CsiR1 and slr1214 which encodes a second response regulator protein. In contrast to UV-A light, application of 1000 nL L−<sup>1</sup> ethylene reduces the transcript abundance of CsiR1 and slr1214 (Lacey et al., 2018), but it is not known whether or not this occurs via regulation of SynEtr1 histidine kinase activity. Ethylene also causes changes in the cell surface of Synechocystis cells leading to enhanced biofilm formation, more directed motility of single cells in response to directional light, and faster phototaxis when the cells aggregate (Lacey and Binder, 2016; Kuchmina et al., 2017; Lacey et al., 2018).

Many additional putative ethylene receptors have been identified in a wide array of bacteria and several non-plant eukaryotes (Mount and Chang, 2002; Wang et al., 2006; Lacey and Binder, 2016; Hérivaux et al., 2017; Kabbara et al., 2018), but it remains to be determined whether or not other cyanobacteria respond to ethylene. With this in mind we studied ethylene responses in Geitlerinema sp. PCC 7105. This was originally named Oscillatoria sp. PCC 7105 (Rippka et al., 1979) and is referred to as Geitlerinema in this paper. This is a filamentous cyanobacterium that binds ethylene and is predicted to contain two ethylene receptors (Wang et al., 2006; Lacey and Binder, 2016). Here we document that ethylene alters phototaxis behavior of Geitlerinema and Synechocystis at low concentrations. These concentrations are below the threshold for most, but not all ethylene responses in plants. However, ethylene dosebinding experiments on Geitlerinema, Synechocystis, and the heterologously expressed ethylene binding domain of SynEtr1 indicate that the affinity of ethylene to the cyanobacterial receptors is similar to what has been reported in plants. Thus, we predict that signal amplification occurs at the level of the receptors.

### MATERIALS AND METHODS

### Strains and Growth Conditions

Geitlerinema PCC 7105 cells were from the laboratory of Anthony Bleecker and were originally obtained from the American Type Culture Collection (stock ATCC29120). Synechocystis PCC 6803 cells were obtained from the Pasteur Institute. Liquid cultures of both were maintained in BG-11 medium (Rippka et al., 1979).

#### Phototaxis Assays

Phototaxis assays were conducted at 20–21◦C in flow-through chambers with continuous gas flow with either ethylene-free air or air with ethylene at the concentrations indicated in each figure. All assays were replicated at least three times.

Phototaxis assays for Synechocystis were conducted for 4 days (d) as previously described using directional white light at a fluence rate of 30 µmol m−<sup>2</sup> s −1 (Lacey and Binder, 2016). These assays were quantified by measuring the maximum distance moved by cells from the leading edge of the colony at the start of the assay. For phototaxis assays on Geitlerinema, cells were placed on 0.4% (v/v) agar BG-11 plates and allowed to grow several days under white fluorescent lights. Efforts were made to start with similar quantities of cells. However, the filamentous nature of this species made it difficult to start with identical numbers of cells. A similar problem has been noted by others studying another filamentous cyanobacteria, Nostoc punctiforme (Campbell et al., 2015). The plates were then wrapped with aluminum foil except for a 13 × 13 mm square above the location next to the cells. This area was illuminated from above with white fluorescent lights (42 µmol m−<sup>2</sup> s −1 ) for 5–7 days to allow cells to move into the illuminated area and grow. At this time, we used one of two methods to examine phototaxis. In some assays, we used methods modified from Biddanda et al. (2015) and illustrated

in **Supplementary Figure 1A**. In this method, the first opening above the cells was blocked, a second opening made 25 mm away, and illumination provided from above. Unless otherwise specified, cells were then allowed to respond to light for 2–5 d in response to 42 µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> of white, or 16 µmol m−<sup>2</sup> s −1 of blue (λmax = 462 nm), green (λmax = 528 nm), or red (λmax = 672 nm) light. In other assays, the Geitlerinema cells were exposed to directional white light using methods modified from Campbell et al. (2015). In this method, each petri dish was masked with black paper except for a 5 mm slit at one edge of the plate. The cells were then exposed for 24 h to directional white light through the slit at a fluence rate of 30 µmol m−<sup>2</sup> s −1 . The distance moved toward the light was then quantified by measuring the maximum distance moved by cells from the leading edge of the colony at the start of the assay. For these experiments, white lighting was provided by an LED light panel and monochromatic lighting provided by LED arrays from Quantum Devices Inc. (Barneveld, WI, United States). For both species, images were acquired with a flatbed scanner. In control experiments to examine the effect of these wavelengths of light on growth, we allowed filaments to become established on the BG-11 agar under white light for 1 d, and then exposed the entire plate to either white, red, green, or blue light at the same levels of illumination as used in phototaxis assays. We then scanned the plates 2 d later and used ImageJ to determine the optical density of the colonies.

For polychromatic light experiments, cells were placed on 0.4% (v/v) agar BG-11 plates and exposed to polychromatic light from above the cells for 4 d (**Supplementary Figure 1B**). Polychromatic light was provided by white light from a slide projector focused onto a prism and images were acquired with a Canon EOS Rebel Xsi.

### Biofilm Assays

Biofilm formation by Synechocystis cells was assayed with modifications to the methods of Agostoni et al. (2016). For this, cells were grown in BG-11 liquid culture to a density of OD<sup>750</sup> = 0.5 and 15 mL placed into a 250 mL flask. Samples were then incubated in flow through chambers with ethylenefree air or various dosages of ethylene for 5 d under white light (30 µmol m−<sup>2</sup> s −1 ) provided by an LED panel. Non-adhered cells were removed by aspiration and 0.5% (w/v) crystal violet added for 2 min to stain the cells that remain attached. The stain was removed and the cells washed three times with 15 mL of phosphate-buffered saline. The cells were then resuspended in 10 mL of 95% (v/v) ethanol for 30 min, and the OD<sup>588</sup> measured. All assays were replicated at least three times.

### RNA Isolation, Complementary DNA Synthesis, and Quantitative Real-Time Reverse Transcriptase (qRT)-PCR

For Synechocystis, cells were exposed to phototaxis conditions for 1 d in ethylene-free air at which time they were either kept in ethylene-free air or treated with ethylene at concentrations ranging from 1 to 1000 nL L−<sup>1</sup> for 4 h using methods previously described (Lacey et al., 2018). Briefly, ethylene was injected into sealed chambers to yield the designated concentration. Cells were harvested off the agar 4 h later for RNA isolation and further processing. For Geitlerinema, cells were maintained in non-directional light and exposed to ethylene-free air or 1000 nL L−<sup>1</sup> ethylene for 4 h after which the cells were harvested with forceps. To make harvesting of cells easier, cells were kept in Petri dishes filled with 30 mL BG-11 media. For both species, RNA isolation, complementary DNA synthesis, and qRT-PCR were carried out as previously described (Lacey and Binder, 2016). Primers used for SynEtr1, CsiR1, slr1213, and slr1214 from Synechocystis have previously been described (Lacey et al., 2018). Synechocystis data were normalized to the tryptophan synthase gene (TrpA) gene (Zhang et al., 2008) and then to levels of each gene transcript in airtreated controls.

In Geitlerinema, we first determined which housekeeping gene to use. For this, we analyzed the RNA abundance of candidate genes in air- and ethylene-treated cells using qRT-PCR. This was normalized to total RNA (**Supplementary Figure 2**). From this we determined that the abundance of these transcripts was not significantly altered by application of 1000 nL L−<sup>1</sup> ethylene. We chose a gene (gene locus WP\_026097408) annotated as a tRNA pseudouridine synthase (TruB) as our housekeeping gene because its levels were very similar in air versus ethylene. Thus, we normalized the RNA abundance of GeiEtr1 and GeiEtr2 to the abundance of TruB and then to air-treated controls. Primers for qPCR were 5<sup>0</sup> - ATGTGGGAAACTGTCAAAACTTTATTTT-3<sup>0</sup> (forward) and 5 0 -CCGAAGCCTGCTGGGTAA-3<sup>0</sup> (reverse) for GeiEtr1, 5<sup>0</sup> -AT GTGGACCGCTCTCGAATCGCTCC-3<sup>0</sup> (forward) and 5<sup>0</sup> -CCC GAACGAAATCCATGACTGCTGA-3<sup>0</sup> (reverse) for GeiEtr2, and 5<sup>0</sup> -ATGGCGGGCTTTCTGAACCTGG-3<sup>0</sup> (forward) and 5<sup>0</sup> - CCGAAAATGGTGTTTGATCGC-3<sup>0</sup> for TruB.

Quantitative PCR was performed as described in Lacey and Binder (2016). All data represent the average ± SEM from three technical replicates done on three biological replicates.

## Ethylene Binding Assays

Ethylene binding assays were conducted on Geitlerinema and Synechocystis cells as previously described for bacteria (Wang et al., 2006) and on the ethylene binding domain of SynEtr1 fused to glutathione-S-transferase (SynEtr1[1-130]GST) expressed in Pichia pastoris as described by Lacey and Binder (2016). In control samples, specific ethylene binding at 1000 nL L−<sup>1</sup> ethylene to Synechocystis cells lacking SynEtr1 (1SynEtr1) and P. pastoris with empty vector (pPICZ) was determined. The 1SynEtr1 Synechocystis have been previously described (Lacey and Binder, 2016). Assays were conducted using <sup>14</sup>C2H<sup>4</sup> custom synthesized by ViTrax (Placentia, CA, United States). Briefly, experiments were conducted on 0.8 (wet weight) of bacteria cells placed on Whatman No. 1 paper filters or 1 g (wet weight) P. pastoris cells placed on glass filters. Samples were then treated with either <sup>14</sup>C2H<sup>4</sup> at the indicated concentrations to determine total binding or <sup>14</sup>C2H<sup>4</sup> plus 1000-fold excess <sup>12</sup>C2H<sup>4</sup> to determine non-specific background

binding. Specific binding was calculated by subtracting nonspecific binding from total binding. All experiments were done in triplicate.

### RESULTS

### Geitlerinema Has Two Putative Ethylene Receptors

The sequenced Geitlerinema genome contains two genes predicted to encode proteins with an ethylene-binding domain (see the **Supplementary Data** for full DNA and amino acid sequences). We are calling them Geitlerinema Ethylene response1 (GeiEtr1) and GeiEtr2 following the nomenclature for the first ethylene receptor discovered, AtETR1 from Arabidopsis thaliana. GeiEtr1 is at DNA coordinates 3597293–3599260 and is predicted to encode a protein 655 amino acids long, whereas GeiEtr2 is at DNA coordinates 3147733–3150810 and is predicted to encode a 1025 amino acid long protein. An examination of the genome neighborhoods of these two genes (**Supplementary Figure 3**) reveals one gene grouped with GeiEtr1 predicted to encode a lycopene cyclase and two genes grouped with GeiEtr2 where one is predicted to encode a protein with a diguanylate cyclase domain and the other annotated as a starch phosphorylase.

The predicted ethylene binding domain of GeiEtr1 shares 43% homology and GeiEtr2 41% homology with the binding domain of the canonical ethylene receptor, AtETR1. An alignment of the ethylene domains from GeiEtr1 and GeiEtr2 with the ethylene receptor from Synechocystis and the five receptors from A. thaliana (AtETR1, AtETR2), Ethylene Response Sensor1 (AtERS1), AtERS2, and Ethylene Insensitive4 (AtEIN4) reveals that both proteins from Geitlerinema have retained many amino acids in common with functional ethylene receptors. Seven amino acid residues are required for ethylene binding to AtETR1 (Rodriguez et al., 1999; Wang et al., 2006). All seven of these amino acids are conserved in GeiEtr1 and GeiEtr2 suggesting that they too can bind ethylene (**Figure 1A**). Consistent with previous research (Hua et al., 1998), this alignment also shows that the subfamily 1 receptors from A. thaliana, AtETR1 and AtERS1, have a short N-terminus extension (approximately 19 amino acids) ahead of the ethylene binding domain, whereas there is a longer hydrophobic stretch of amino acids (approximately 50 amino acids) in the subfamily 2 receptors, AtERS2, AtETR2, and AtEIN4. The proteins from cyanobacteria have an N-terminal stretch intermediate in length (approximately 30 amino acids) between the subfamily 1 and 2 receptors. It is also interesting to note that whereas the plant receptors have two cysteines near the N-terminus that form disulfide bonds to form stable homodimers (Schaller et al., 1995; Hall et al., 2000), the cyanobacteria proteins only have one cysteine in this region of the protein that may fulfill the same function. An unrooted cladogram based on the amino acid sequences of the ethylene binding domains of these seven proteins shows that the ethylene receptors from plants fall into two subfamilies with the bacterial receptors forming a distinct third subfamily (**Figure 1B**), consistent with a previous analysis (Wang et al., 2006). The predicted domain structure of GeiEtr1 is similar to AtERS1 and AtERS2 where there is a N-terminal ethylene binding domain followed by a GAF (for cGMP-specific phosphodiesterase, adenyl cyclases, and FhlA) domain and a C-terminal kinase domain, but no receiver domain (**Figure 1C**). By contrast, GeiEtr2 has additional domains with a PAS (for Per-Arnt-Sim) and PAC domain between the ethylene binding and GAF domains as well as a C-terminal receiver domain. The arrangement of PAS, PAC, and GAF domains in GeiEtr2 is reminiscent of the domain arrangement of SynEtr1 that functions as a photoreceptor.

We were curious to know if either receptor is expressed and whether or not ethylene affected the transcript abundance of either gene. To answer these questions we extracted RNA from samples kept in air versus 1000 nL L−<sup>1</sup> ethylene for 4 h. We chose this dosage of ethylene because it is commonly used in plant research and has been shown to affect the physiology and growth of Synechocystis (Lacey and Binder, 2016; Henry et al., 2017) and causes wide-spread changes in the transcriptome of Synechocystis (Lacey et al., 2018). Also, we have previously found that 1000 nL L−<sup>1</sup> ethylene alters transcript abundance of various genes in Synechocystis cells within 4 h (Lacey et al., 2018). From this analysis we observed that both GeiEtr1 and GeiEtr2 are expressed in air. Upon application of ethylene, GeiEtr1 abundance decreased, but the abundance of GeiEtr2 showed no statistically significant change (**Figure 1D**).

## Ethylene Alters Phototaxis of Geitlerinema

We have previously shown that ethylene accelerates phototaxis of Synechocystis toward white light (Lacey and Binder, 2016). We therefore examined the effect of ethylene on phototaxis of Geitlerinema. Because this is a filamentous cyanobacterium, we adapted the methods of Biddanda et al. (2015) to conduct these assays where the bacteria were exposed to an area of illumination at a distance from their starting location (**Supplementary Figure 1A**). In air, cells displayed phototaxis movement in response to white light where most, but not all cells, moved to the new position of illumination after 5 d (**Supplementary Figure 4**). Given this result we conducted phototaxis assays toward white light for 5 d to determine whether application of ethylene increased or decreased movement. Interestingly, the application of 1000 nL L−<sup>1</sup> ethylene altered the response to white light so that the cells formed a ring outside the area of illumination (**Figure 2A**). To determine if this ring of cells is due to higher light sensitivity causing the cells to avoid the area with the highest light levels, we conducted these assays at a 10 fold dimmer light intensity. At this dimmer light level, the cells move into the entire area of illumination so that no ring of cells is present (**Supplementary Figure 5**) suggesting that ethylene sensitizes Geitlerinema to higher light intensities.

We have also previously demonstrated that ethylene accelerates phototaxis of Synechocystis toward monochromatic light including red, green, and blue light (Lacey and Binder, 2016). Therefore, we examined phototaxis of Geitlerinema cells in response to these colors of light. In

the average ± SEM. P-values were determined using Student's t-test.

air, phototaxis occurred faster in response to red and green light compared to white light (**Supplementary Figure 4**). By contrast, the cells did not phototaxis in response to blue light and appeared to show unbiased movement with few cells accumulating in the area of illumination. This is consistent with results using Synechocystis where blue light normally does not cause phototaxis and has been found to promote growth and inhibit motility (Wilde et al., 2002; Chau et al., 2017). We also examined the effects of these different light qualities on cell growth and found that growth occurred in white, red, and green light at similar rates but growth did not occur in blue light (**Supplementary Figure 6**). This suggests that the larger number of cells in the new area of illumination in red and green light is not simply caused by more growth compared to white light.

Because red and green light cause faster phototaxis compared to white light, we shortened the assay time-frame from 5 to 2 d in order to determine whether ethylene stimulates or inhibits phototaxis. Under these conditions, 1000 nL L−<sup>1</sup> ethylene altered the phototaxis pattern in response to red light where the cells aggregated at the edge of the area of illumination (**Figure 2A**). By contrast, ethylene caused no measurable change in phototaxis in response to either green or blue light. This is in contrast to our observations with Synechocystis where 1000 nL L−<sup>1</sup> ethylene caused a measurable increase in phototaxis toward both (Lacey and Binder, 2016).

Cyanobacteria integrate various light inputs including wavelength information (Chau et al., 2017). We were therefore curious to examine the response of Geitlerinema cells to polychromatic light. In air, the cells moved into the light spectrum between orange and blue (**Figure 2B**). By contrast, in the presence of 1000 nL L−<sup>1</sup> ethylene the cells moved into a wider range of wavelengths of light that included red light. Together,

these results indicate that ethylene affects the phototaxis response of Geitlerinema cells to different wavelengths of light.

#### Low Levels of Exogenous Ethylene Affect Geitlerinema

Plants respond to a wide range of ethylene concentrations and even show transient growth inhibition at very low concentrations (0.2 nL L−<sup>1</sup> ) of exogenous ethylene (Chen and Bleecker, 1995; Binder et al., 2004). We therefore wished to know the threshold concentration of ethylene that affects phototaxis of Geitlerinema. To examine this we conducted phototaxis assays in response to white light for 5 d in the presence of different concentrations of ethylene or ethylene-free air (**Figure 3A** and **Supplementary Figure 7**). Under these conditions, even the lowest concentration used, 8 nL L−<sup>1</sup> , resulted in faster phototaxis since more filaments of cells moved into the illuminated area of the plate than were observed in ethylene-free air. Increasing the ethylene levels to 70 nL L−<sup>1</sup> caused more phototaxis. By contrast, treatment with 700 nL L−<sup>1</sup> ethylene caused the cells to form a distinct band

FIGURE 3 | Geitlerinema responds to low levels of exogenous ethylene. (A) Cells were placed on soft agar and exposed to white light from above using a protocol modified from Biddanda et al. (2015) as described in the section "Materials and Methods." The position where the colony of cells started and end position after 5 d of lighting are indicated. Scale bar = 1 cm. (B) Phototaxis assays were conducted by exposing cells to directional light for 24 h using a protocol modified from Campbell et al. (2015) as described in the section "Materials and Methods." The maximum distance moved from the initial colony position was measured and normalized to distance moved in air. Data represent the average ± SEM from five replicates. Statistical analysis was done with ANOVA and the different letters indicate significant differences (P < 0.05). In both panels, cells were kept in flow-through chambers maintained at the indicated concentration of exogenous ethylene.

around the edge of the illuminated area reminiscent to the ring that formed farther from the light with 1000 nL L−<sup>1</sup> (**Figure 2**).

To gain a better understanding about what appears to be enhanced phototaxis at low levels of ethylene, we also conducted phototaxis experiments where the cells were exposed to directional white light. In air, cells moved from the original colony in all directions relative to the light, but the largest distance moved was toward the light source (**Supplementary Figure 8**). Application of ethylene at the lowest dose used (8 nL L−<sup>1</sup> ) significantly increased the distance moved by cells toward directional white light and the response seems to be saturated by this level of ethylene (**Figure 3B** and **Supplementary Figure 8**). These data support the idea that low ethylene levels increase phototaxis in this cyanobacterium.

#### Low Levels of Exogenous Ethylene Affect Synechocystis

Because low levels of exogenous ethylene affect Geitlerinema, we tested whether or not low levels of ethylene also affect the physiology of Synechocystis. For this we measured phototaxis and biofilm formation, both of which are increased by application

of 1000 nL L−<sup>1</sup> ethylene (Lacey and Binder, 2016). Application of ethylene at the lowest dose used (8 nL L−<sup>1</sup> ) significantly increased the distance moved by cells in response to directional white light (**Figure 4A**). There is a slight increase in phototaxis at 70 nL L−<sup>1</sup> but the response seems to be largely saturated with 8 nL L−<sup>1</sup> applied ethylene. A similar dose–response curve was observed with biofilm formation except that saturation of the response at 8 nL L−<sup>1</sup> is more clearly seen (**Figure 4B**). As a comparison, the threshold ethylene concentration typically observed for responses in plants is above 10 nL L−<sup>1</sup> with saturation of responses occurring in a concentration range of 1–100 µL L−<sup>1</sup> ethylene depending on the response being measured and the species studied (Chadwick and Burg, 1970; Lyon, 1970; Burdett, 1972; Goeschl and Kays, 1975; De Munk and Duineveld, 1986; Beaudry and Kays, 1988; Chen and Bleecker, 1995).

These assays were conducted using flow-through chambers to maintain constant O<sup>2</sup> and CO<sup>2</sup> concentrations over the long time period (4 d) of the assays. A limitation of this method is that it is difficult to reliably deliver lower ethylene concentrations. Previously, we showed that application of 1000 nL L−<sup>1</sup> ethylene causes a rapid (within 30 min) decrease in the transcript levels of CsiR1 and slr1214 (Lacey et al., 2018). This allowed us to conduct shorter term experiments where ethylene was simply injected into a sealed chamber with phototaxing cells. We compared the transcript levels of SynEtr1, slr1213, CsiR1, and slr1214 after 4 h treatments with varying dosages of exogenous ethylene. Results from this showed that ethylene at dosages between 1 and 1000 nL L−<sup>1</sup> had no significant effect on the transcript abundance of either SynEtr1 or slr1213 (**Figure 4C**). By contrast, CsiR1 transcript abundance was altered by application of as low as 1 nL L−<sup>1</sup> ethylene and increasing levels of ethylene resulted in a concomitant decrease in the levels of CsiR1 transcript. By contrast, lower dosages of exogenous ethylene had no effect on slr1214 transcript. However, treatment with 1000 nL L−<sup>1</sup> ethylene resulted in a decrease in slr1214. These results with 1000 nL L−<sup>1</sup> ethylene are consistent with our prior results examining these four transcripts (Lacey et al., 2018). Together, these data indicate that Synechocystis responds to ethylene at levels as low as 1 nL L−<sup>1</sup> .

### Ethylene Dose-Binding to Geitlerinema, Synechocystis, and SynEtr1[1-130]GST

The ethylene dose-dependency of binding to AtETR1 parallels the dose-dependency for long-term growth inhibition of darkgrown A. thaliana seedlings in response to ethylene (Chen and Bleecker, 1995; Schaller and Bleecker, 1995). This suggests that this response is related to the binding affinity to the receptors. We were curious to know if a similar relationship between ethylene binding and responses in Synechocystis and Geitlerinema existed. If true, we predicted that ethylene dose-binding characteristics should saturate at around 10–100 nL L−<sup>1</sup> ethylene. To test this, we conducted ethylene dose-binding experiments across a range of ethylene concentrations on Geitlerinema and Synechocystis cells, and SynEtr1[1-130]GST expressed in P. pastoris (**Figure 5**). In all three cases, ethylene binding continued to increase as ethylene levels were increased up to 1000 nL L−<sup>1</sup> , demonstrating that ethylene-binding activity does not saturate at 100 nL L−<sup>1</sup> . At 1000 nL L−<sup>1</sup> , there was no specific ethylene-binding activity detected in 1SynEtr1 Synechocystis cells with SynEtr1 deleted or in P. pastoris cells that were expressing empty vector, consistent with what has previously been reported (Rodriguez et al., 1999; McDaniel and Binder, 2012). Ethylene binding did not reach obvious saturation in the range of concentrations we tested so we estimated the Kd-values with curve fitting. We fitted the data with GraphPad Prism (Ver. 7.04) using default settings for one binding site and measuring specific binding. This yielded sigmoidal curves with estimated Kd-values of 335 ± 227 nL L−<sup>1</sup> for Geitlerinema, 248 ± 16 nL L−<sup>1</sup> for Synechocystis, and 127 ± 12 nL L−<sup>1</sup> for SynEtr1[1-130]GST. These values are in the same range as what has been observed in many plants (Sisler, 1979; Goren and Sisler, 1986; Sisler et al., 1986; Smith et al., 1987; Blankenship and Sisler, 1989, 1993; Sanders et al., 1990) and somewhat higher than what has been reported for A. thaliana plants (Sanders et al., 1991) and exogenously expressed AtETR1 (Schaller and Bleecker, 1995). This suggests that the sensitivity of ethylene responses in these cells is not explained simply by the binding affinity of ethylene to the receptors.

## DISCUSSION

We have previously shown that the cyanobacterium Synechocystis contains a functional ethylene receptor that regulates cell physiology including phototaxis and biofilm formation and that many other bacteria may contain functional ethylene receptors (Rodriguez et al., 1999; Wang et al., 2006; Lacey and Binder, 2016). In this study we provide evidence supporting the idea that another cyanobacterium, Geitlerinema, responds to ethylene and contains two ethylene receptor isoforms. Application of ethylene to Geitlerinema alters phototaxis indicating a conservation of function for ethylene signaling in both Geitlerinema and Synechocystis. At low ethylene concentrations, ethylene stimulates phototaxis toward white light for both species. It is unclear if this is occurring because of faster movement of individual cells/filaments or movement that is more directed toward the light or both. Using single-cell tracking assays, Kuchmina et al. (2017) demonstrated that higher concentrations of endogenously produced ethylene do not affect the speed of cells, but rather cause movement that is more directed toward the illumination (Kuchmina et al., 2017). It is unknown if this also applies to Synechocystis cells once they aggregate as used in our assays or to Geitlerinema filaments. At higher concentrations, ethylene appeared to sensitize Geitlerinema to the light so that the cells tended to avoid higher levels of illumination. No such sensitization caused by ethylene has been observed in Synechocystis. Like Synechocystis (Lacey and Binder, 2016), ethylene alters phototaxis of Geitlerinema in response to white light and monochromatic red light. However, unlike our prior results studying Synechocystis, ethylene does not have a measurable effect on phototaxis of Geitlerinema in response to monochromatic blue or green light. Interestingly, ethylene affected responses to polychromatic light where Geitlerinema cells phototaxed into a wider range

flow-through chambers maintained at 0, 8, 70, 290, or 700 nL L−<sup>1</sup> exogenous ethylene for 4 d and data are the average ± SEM. (C) The gene transcript abundance of SynEtr1, slr1213, CsiR1, and slr1214 was measured using qRT-PCR from RNA extracted from cells maintained at 0, 1, 10, 100, or 1000 nL L−<sup>1</sup> exogenous ethylene for 4 h in sealed chambers under phototaxis conditions. Data were normalized to the transcript levels of the TrpA reference gene and normalized to cells kept in ethylene-free air. Data represent the average ± SEM from three biological replicates with three technical replicates each. In all panels, statistical analyses were done with ANOVA and the different letters indicate significant differences (P < 0.05).

of wavelengths of light in the presence of ethylene versus in ethylene-free air. This shows that ethylene can affect wavelength integration by these cells.

A model for signal transduction from SynEtr1 in Synechocystis has developed (Ramakrishnan and Tabor, 2016) where UV-A light stimulates histidine autophosphorylation of SynEtr1 followed by phosphotransfer to a conserved aspartate on slr1213. The phosphorylated slr1213 enhances the transcription of CsiR1 and slr1214 which contain a common transcription start site and appears to be co-transcribed in response to UV-A light (Ramakrishnan and Tabor, 2016). In contrast to UV-A light, application of 1000 nL L−<sup>1</sup> ethylene reduces the transcript abundance of CsiR1 and slr1214 (Lacey et al., 2018). It is currently unknown whether or not ethylene signaling from SynEtr1 also affects histidine kinase activity to regulate CsiR1 and slr1214 levels. However, application of ethylene reveals that the regulation of CsiR1 and slr1214 is not entirely overlapping. First, we previously showed that application of 1000 nL L−<sup>1</sup> ethylene caused a rapid and prolonged down regulation of CsiR1, whereas this treatment caused a rapid and transient down-regulation of slr1214 (Lacey et al., 2018). In this study we provide further evidence for more complex regulation where low levels of ethylene decrease CsiR1 levels, but it requires a much higher level of ethylene to reduce the abundance of slr1214. We currently do not know the basis for these differences in regulation but it is possible that the transcript stabilities of these genes are regulated differently.

In Synechocystis, the ethylene dose–responses for phototaxis and biofilm formation show an inverse correlation with the dose–response for CsiR1 abundance suggesting that CsiR1 plays an important inhibitory role for these two responses. However, its function in Synechocystis and the mechanisms by which it affects Synechocystis physiology are poorly studied. One possibility is that CsiR1 is functioning in transcriptional or posttranscriptional regulation in Synechocystis (Georg et al., 2009; Hernández-Prieto et al., 2012). In Synechocystis, both biofilm formation and motility are dependent on type IV pili (Bhaya et al., 1999, 2000; Yoshihara et al., 2001; Burriesci and Bhaya, 2008; Schuergers et al., 2015). The RNA chaperone Hfq affects Synechocystis motility and type IV pilus function (Dienst et al., 2008; Schuergers et al., 2014) suggesting that CsiR1 may also affect motility and biofilm formation by altering type IV pili, perhaps by altering expression of certain pilin proteins. However, it is likely that regulation by ethylene is more complex than simply regulating CsiR1 since removing slr1214 eliminates physiological responses to ethylene (Lacey and Binder, 2016).

We have previously studied the effects of ethylene on Synechocystis using 300 and 1000 nL L−<sup>1</sup> exogenous ethylene (Lacey and Binder, 2016; Henry et al., 2017; Lacey et al., 2018). This is a common concentration range to use in plants, but in aqueous environments, ethylene levels are often reported to be lower than this (Abeles et al., 1992). We therefore examined the physiological responses of Synechocystis and Geitlerinema to a range of ethylene concentrations and found that both species physiologically respond to levels of ethylene as low as 8 nL L−<sup>1</sup> . We quantified the effects of ethylene on increasing biofilm formation and phototaxis in Synechocystis and discovered that both responses saturate at between 8 and 70 nL L−<sup>1</sup> . This suggests that both responses occur over a narrow dynamic range of ethylene concentrations, although we cannot rule out that these cells have physiological responses that occur at ethylene concentrations lower than used in this study. By comparison, plants have a wide dynamic range over several orders of magnitude of ethylene concentration. For instance, the growth of dark-grown A. thaliana seedlings is inhibited in a dose-dependent manner over a range of ethylene concentrations between 10 and 1000 nL L−<sup>1</sup> (Chen and Bleecker, 1995). It is interesting to note that induction of chitinase-B also has dose-dependent induction over two orders of magnitude ethylene concentration. However, whereas growth inhibition saturates at 1000 nL L−<sup>1</sup> , chitinase-B induction saturates at a 10-fold higher concentration (Chen and Bleecker, 1995). It is therefore of note that the reduction of CsiR1 transcript abundance in Synechocystis occurs in a dose-dependent manner over a wide range of ethylene concentrations between 1 and 1000 nL L−<sup>1</sup> . This raises the possibility that there may be other physiological responses in Synechocystis occurring in different ranges of ethylene concentration from biofilm formation and phototaxis. It is also interesting that CsiR1 transcript abundance was reduced by application of as low as 1 nL L−<sup>1</sup> ethylene. This low level of ethylene is below the threshold for most responses reported in plants (Abeles et al., 1992). Exceptions are that stimulation of Ricinodendron rautanenii seed germination has been observed with 1 nL L−<sup>1</sup> treatments (Keegan et al., 1989) and transient growth inhibition responses occur in dark grown A. thaliana seedlings down to 0.2 nL L−<sup>1</sup> ethylene (Binder et al., 2004).

The long-term growth inhibition response of dark-grown A. thaliana to ethylene has a dose-dependency that parallels the dose-dependency of ethylene binding to AtETR1 (Chen and Bleecker, 1995; Schaller and Bleecker, 1995). This led us to predict that ethylene binding to Geitlerinema and Synechocystis cells, as well as heterologously expressed SynEtr1, would saturate at between 8 and 100 nL L−<sup>1</sup> ethylene. However, our results indicate that this is not the case. Thus, unlike the long-term growth inhibition response of dark-grown A. thaliana seedlings to ethylene, there is no obvious correlation between the amount of ethylene bound and physiological responses in these two species of cyanobacteria. It is interesting to note that the dosedependency of ethylene binding to Synechocystis and SynEtr1[1- 130]GST parallels the decrease in CsiR1 transcript caused by ethylene. The K<sup>d</sup> for ethylene binding to AtETR1 heterologously expressed in yeast is reported at 36 nL L−<sup>1</sup> (Schaller and Bleecker, 1995) yet A. thaliana seedlings are able to respond to ethylene concentrations approximately 300-fold below this level with a transient growth inhibition response (Binder et al., 2004). Our results indicate there may be a similar discrepancy between ethylene binding and ethylene responses where Synechocystis responds to 1 nL L−<sup>1</sup> ethylene yet the estimated binding affinity is over 100-fold higher. These observations suggest that signal amplification is occurring. In bacterial chemotaxis receptors, such amplification occurs because of physical clustering of the receptors. In this model, binding of ligand to one receptor causes conformational changes in surrounding, ligand-free receptors

through physical interactions (Briegel and Jensen, 2017; Bi and Sourjik, 2018). Thus, it is possible that amplification of ethylene signaling in these cyanobacteria, as well as in plants, is occurring at the level of the receptors. However, it remains to be determined whether a similar clustering of ethylene receptors and amplification of signal is occurring in either plants or cyanobacteria.

The ecophysiological role of ethylene for cyanobacteria remains unanswered but it is likely that cyanobacteria encounter ethylene in the environment produced by other organisms or produced abiotically from sunlight photochemically converting dissolved organics to ethylene (Swinnerton and Linnenborn, 1967; Wilson et al., 1970; Swinnerton and Lamontagne, 1974; Ratte et al., 1993, 1998). It is known that ethylene diffuses in water and ethylene levels vary in aqueous environments depending on environmental conditions resulting in ethylene concentrations in the range where we observe responses (Swinnerton and Lamontagne, 1974; Plass et al., 1992; Ratte et al., 1993, 1998). Thus, ethylene may be acting as a signal for cells to move into light conditions that optimize photosynthesis or reduce light stress or it may be a signal to establish a symbiotic relationship with another organisms. Our results showing that ethylene causes Geitlerinema cells to avoid high light supports the idea that it is a stress signal. However, these are not mutually exclusive ideas and ethylene may be regulating a variety of functions.

In summary, we have shown that two cyanobacteria species respond to low levels of ethylene. Our results support the hypothesis that ethylene perception evolved prior to green plants and is likely to be wide-spread in cyanobacteria species. Given the diversity of putative receptors in various microbes and the presence of multiple receptor isoforms in certain species (Mount and Chang, 2002; Wang et al., 2006; Lacey and Binder, 2016; Hérivaux et al., 2017; Kabbara et al., 2018; Papon and Binder, 2019), it is likely that the mechanisms of ethylene signaling and responses controlled by ethylene are diverse.

#### REFERENCES


### DATA AVAILABILITY

All datasets for this study are included in the manuscript and/or the **Supplementary Files**.

#### AUTHOR CONTRIBUTIONS

RL, CA, and BB designed the experiments. BB wrote the manuscript with the help of RL. All authors performed the experiments.

### FUNDING

This work was supported by funds from the University of Tennessee Program for Excellence & Equity in Research (PEER) (to CA). Research funding was provided by grants from the US National Science Foundation (IOS-1254423 and MCB-1517032 to BB).

### ACKNOWLEDGMENTS

The authors thank M. B. Stalans, J. B. Case, A. J. Schafer, J. Shreck, T. Payne, A. Patterson, J. Mondal, E. E. Helmbrecht, and B. D. Bruce for technical assistance and advice.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.00950/ full#supplementary-material


phototaxis in hormogonium filaments of < span class = "named-content genusspecies" id = "named-content-1" > Nostoc punctiforme < /span >. J. Bact. 197, 782–791. doi: 10.1128/jb.02374-14



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Allen, Lacey, Binder Bickford, Beshears, Gilmartin and Binder. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Ethylene Induces a Rapid Degradation of Petal Anthocyanins in Cut *Vanda* 'Sansai Blue' Orchid Flowers

*Sudarat Khunmuang1,2 , Sirichai Kanlayanarat1,2 , Chalermchai Wongs-Aree1,2 , Shimon Meir3 , Sonia Philosoph-Hadas3 \*, Michal Oren-Shamir4 , Rinat Ovadia4 and Mantana Buanong1,2*

*1 Division of Postharvest Technology, School of Bioresources and Technology, King Mongkut's University of Technology Thonburi (Bangkhuntien) Thakam, Bangkok, Thailand, 2 Postharvest Technology Innovation Center, Office of the Higher Education Commission, Bangkok, Thailand, 3 Department of Postharvest Science, Agricultural Research Organization (ARO), The Volcani Center, Rishon LeZion, Israel, 4 Department of Ornamental Plants and Agricultural Biotechnology, Agricultural Research Organization (ARO), The Volcani Center, Rishon LeZion, Israel*

#### *Edited by:*

*Autar Krishen Mattoo, United States Department of Agriculture, United States*

#### *Reviewed by:*

*Antonio Ferrante, University of Milan, Italy Kazuo Ichimura, Institute of Vegetable and Floriculture Science, NARO, Japan*

> *\*Correspondence: Sonia Philosoph-Hadas vtsoniap@volcani.agri.gov.il*

#### *Specialty section:*

*This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science*

*Received: 09 April 2019 Accepted: 18 July 2019 Published: 09 August 2019*

#### *Citation:*

*Khunmuang S, Kanlayanarat S, Wongs-Aree C, Meir S, Philosoph-Hadas S, Oren-Shamir M, Ovadia R and Buanong M (2019) Ethylene Induces a Rapid Degradation of Petal Anthocyanins in Cut Vanda 'Sansai Blue' Orchid Flowers. Front. Plant Sci. 10:1004. doi: 10.3389/fpls.2019.01004*

Ethylene plays a major role in the regulation of flower senescence, including in the ethylenesensitive *Vanda* 'Sansai Blue' orchid flowers. This cut flower is popular in Thailand due to its light blue big size florets possessing a beautiful shape pattern. In the present study, we further examined the rapid ethylene-induced process of active anthocyanin degradation in cut *Vanda* 'Sansai Blue' flowers, which occurred much before detection of other typical senescence-related symptoms. For this purpose, the cut inflorescences were exposed to air (control), 1 or 10 μl L−1 ethylene for 24 h, or to 0.2 μl L−1 1-methylcyclopropene (1-MCP) for 6 h followed by 10 μl L−1 ethylene for 24 h at 21°C, and the effects of these treatments on various parameters were assayed. While the fading-induced effect of ethylene was not concentration-dependent in this range, the ethylene treatment significantly reduced the flower vase life in a concentration-dependent manner, further confirming the separation of the bleaching process from senescence. Exposure of the inflorescences to 1-MCP pre-treatment followed by 10 μl L−1 ethylene, recovered both inflorescence color and anthocyanin content to control levels. Quantification of total anthocyanin content, performed by HPLC analysis on the basis of cyanidin-3-glocuside equivalents, showed that ethylene reduced and 1-MCP recovered the anthocyanins profile in non-hydrolyzed anthocyanin samples of *Vanda* 'Sansai Blue' florets, assayed at half bloom and bloom developmental stages. The results showed that the ethylene-induced color fading, observed immediately after treatment, resulted from a significant reduction in the levels of the two main anthocyanidins, cyanidin and delphinidin, as well as of other anthocyanidins present in low abundance, but not from changes in the levels of flavonols, such as kaempferol. This anthocyanin degradation process seems to operate *via* ethyleneincreased peroxidase activity, detected at the bud stage. Taken together, our results suggest that the ethylene-induced rapid color bleaching in petals of cut *Vanda* 'Sansai Blue' flowers is an outcome of *in-planta* anthocyanin degradation, partially mediated by increased peroxidase activity, and proceeds independently of the flower senescence process.

Keywords: anthocyanidins, color fading, cyaniding, delphinidin, ethylene sensitivity, peroxidase activity, senescence symptoms, cut *Vanda* orchid flowers

### INTRODUCTION

The plant hormone ethylene (Bleecker and Kende, 2000) plays a vital role in the regulation of flower senescence, manifested in a range of symptoms including wilting, discoloration, bud degeneration, and abscission (Reid and Wu, 1992), which also occur in senescing orchid flower species (Goh et al., 1985; Woltering and Van Doorn, 1988). The responses to ethylene vary widely between species (Reid and Wu, 1992), although they are often consistent within families or subfamilies (Van Doorn, 2001). The *Orchidaceae* is classified as one of the ethylene-sensitive flower families, with a variable sensitivity to ethylene among the species and cultivars (Akamine, 1963; Burg and Dijkman, 1967; Goh et al., 1985; Woltering and Van Doorn, 1988). *Cattleya, Paphiopedilum*, *Dendrobium*, *Phalaenopsis*, and *Cymbidium* orchids were found to be highly sensitive to ethylene, which caused color fading and wilting of sepal tips, as well as bud and flower abscission (Woltering and Van Doorn, 1988; Porat et al., 1995; Ketsa and Rugkong, 2000). Additionally, *Cymbidium* orchids showed a dramatic response to exogenous ethylene, including induction of anthocyanin formation in female reproductive parts (Goh et al., 1985). Thus, the variation in postharvest life can partly be ascribed to differences in endogenous ethylene biosynthesis, as well as to differences in sensitivity to endogenous and exogenous ethylene (Goh et al., 1985; Woltering and Van Doorn, 1988).

1-Methylcyclopropene (1-MCP), an effective blocker of ethylene perception, is considered to bind to the ethylene receptor irreversibly, resulting in the inhibition of ethylene action (Serek et al., 1994; Sisler et al., 1996). 1-MCP prevents damage from exogenous ethylene in numerous potted plants and cut flower species (Serek et al., 1994, 2006; Serek and Sisler, 2001). Application of 1-MCP suppressed 1-aminocyclopropane-1-carboxylic acid oxidase (ACO) activity and ethylene production in *Cattleya alliance* orchids (Yamane et al., 2004). 1-MCP also delayed senescence in *Cymbidium* flowers with damaged pollinia, thereby extending their vase life, protected the flowers from the deleterious effects of exogenous ethylene, and prevented premature flower senescence generated by the damaged pollinia (Heyes and Johnston, 1998).

The *Vanda* 'Sansai Blue' orchid is a hybrid of *V. Crimson Glory* × *V. coerulea*, with big beautiful light blue florets, and a vase life of about 11–12 days (Khunmuang et al., 2016, 2018). The main senescence symptoms were flower wilting, epinasty, petal discoloration, and abscission. Our previous study showed that exposure of three *Vanda* cultivars, 'Patchara Delight', 'Pure Wax', and 'Sansai Blue' to 10 μl L−1 exogenous ethylene for 24 h significantly reduced by about 50% their vase life (Khunmuang et al., 2019). Ethylene treatment resulted in partial reduction of the anthocyanin content of 'Patchara Delight' after 2 days of vase life, mainly in the full bloom developmental stage, but had no effect on the coloration of 'Pure Wax' except in the bud stage. The flowers of 'Sansai Blue' showed a fast discoloration that occurred much before the wilting and other senescence symptoms (Khunmuang et al., 2019).

While regulation of anthocyanin biosynthesis at the physiological and molecular levels in flowers has been well studied and documented (Bradley et al., 1998; Weiss, 2000; Quattrocchio et al., 2006; Albert et al., 2009, 2010; Davies et al., 2012; Schwinn et al., 2014), the process of anthocyanin degradation was hardly investigated (Vaknin et al., 2005; Oren-Shamir, 2009). Recent studies reported that vacuolar peroxidases, belonging to the class III peroxidase, were responsible for the *in-planta* degradation of anthocyanins in *Brunfelsia calycina* flowers and in ripening grape berries grown in high temperatures (Zipor et al., 2014; Movahed et al., 2016; Lecourieux et al., 2017; Pastore et al., 2017). Therefore, it was of interest to further investigate in the present study, the rapid color fading in response of cut *Vanda* 'Sansai Blue' flowers to exogenous ethylene, focusing on the mechanism of anthocyanin breakdown process.

#### MATERIALS AND METHODS

#### Plant Materials and Treatments

Inflorescences of *Vanda* 'Sansai Blue' orchid were obtained from a commercial farm in Kanchanaburi province, Thailand. Orchid inflorescences bearing 5–8 open florets and 2–4 buds were selected for the experiment, and transported to King Mongkut's University of Technology Thonburi (KMUTT), Bangkhuntien campus, Bangkok within 1.5 h. Upon arrival to the laboratory, flower stems were re-cut under water to a 20-cm length from the stem end to the lower first flower.

For application of ethylene, inflorescences in vases with distilled water were exposed either to air as control, or to 1 or 10 μl L−1 ethylene for 24 h, in a 43-L glass chamber. The 1-MCP and ethylene treatment was applied by exposing the inflorescences to 0.2 μl L−1 1-MCP (0.14% w:w; Ethylbloc®, FloraLife, Walterboro, SC, USA) for 6 h, followed by exposure to 10 μl L−1 ethylene for 24 h, in a 43-L glass chamber. All treatments were performed in a controlled environment room, maintained at 21 ± 2°C, 70–80% RH, under cool-white fluorescence light for 12 h day−1. After treatments, flowers in vases with distilled water were incubated in the observation room throughout the experimental period.

Individual florets from five different developmental stages (tight bud; colored bud; half bloom; bloom; and full bloom) were detached from the inflorescences immediately or 2 days after treatments, photographed for their visual appearance, and assayed for their anthocyanin content as indicated below.

#### Evaluation of Inflorescence Vase Life Longevity

The number of senescing florets in the orchid inflorescences was recorded during the experiment. The vase life was terminated when more than 30% of the florets in an inflorescence lost quality due to petal necrosis, wilting (expressed in loss of turgidity – "sleepiness"), and/or abscission.

#### Determination of Anthocyanin Content by Spectrophotometer

Total anthocyanin was extracted and quantified as previously described (Rodriguez-Saona and Wrolstad, 2005). Samples (about 0.1 g) of fresh petals of florets at different developmental stages were grounded under liquid nitrogen to a fine powder, mixed with 10 ml of 0.01% HCl in methanol, and the extracts were incubated overnight in darkness at 4°C. Absorbance of these extracts was monitored at 530 nm using a spectrophotometer (UV-1800 Shimadzu). Anthocyanin content in the samples was expressed as OD530 mg FW−1.

#### Extraction and Purification of Anthocyanin Samples for HPLC

*Vanda* floret petal samples (approximately 5 g FW) were homogenized in 20 ml of acidified methanol (containing 0.01% HCl), and incubated overnight in darkness at 4°C. The samples were filtered through Whatman no. 1 filter papers. The supernatant was removed, the pellet was re-extracted in 20 ml of acidified methanol, and the extracts were combined. The combined supernatants were dried at 40°C under vacuum (Buchi Evaporator; Vacuum controller V-800: Rotavapor R-205: Heating Bath B-490: Vac®V-500: Recirculating chiller B-740). The samples were evaporated until droplet, and the volume of the anthocyanin extract was adjusted to 0.5 ml with acidified water (containing 0.01% HCl).

The anthocyanin extraction-purification method was performed according to Rodriguez-Saona and Wrolstad (2005), with slight modifications. The anthocyanins were purified using a C18 cartridge (Water Sep-Pak®). The C18 cartridge was activated by flushing twice with methanol, followed by flushing three times with acidified water. The anthocyanins extract (dissolved in acidified water) was loaded into the activated cartridge, and then washed twice with acidified water to remove sugars, acids, and water soluble compounds, followed by washing with ethyl acetate for elimination of procyanidins. Finally, anthocyanins were eluted by acidified methanol. The purified anthocyanins were concentrated by evaporation at 40°C under vacuum, using a rotary evaporator.

#### HPLC Analysis and Composition of Total Anthocyanin

The non-hydrolyzed anthocyanin extracts were analyzed by HPLC (Shimadzu; DGU-20AS Degasser: LC-20AT Liquid Chromatograph: SPD-M20A Photo Diode Array Detector: SIL-20A Auto Sampler: C18 Inertsil® ODS-3; 4.6 mm × 250 mm, 5 μm column). The mobile phase consisted of solvent A (4% phosphoric acid in water) and solvent B (100% acetonitrile). Elution was performed in a linear gradient at the following ratios of solvent A and solvent B – 95:5, 77:23, 77:23, and 95:5 during 1, 25, 29, and 29.01 min, respectively, at a flow rate of 0.7 ml min−1 for 40 min. Chromatograms were obtained by LC solution program. Purified anthocyanins were dissolved in 4% phosphoric acid prior to HPLC purification. Anthocyanins were identified by comparing the retention time (RT) and spectral patterns of standard compounds. The content of the main three non-hydrolyzed anthocyanins at RT of 27.6, 33.8, and 36.0 min, were calculated and expressed as cyanidin-3-glucoside equivalents, after running a standard curve in HPLC under the same conditions. The contents were calculated according to the following equation: *Y* = 5586.6*X* − 34,050 (*Y* = peak area; *X* = ng cyanidin-3-glucoside).

For determination of anthocyanidins composition, another sample of purified anthocyanins (1.5–10 mg DW, lyophilized) was hydrolyzed by boiling in 2 N HCl for 1 h, and separated as described by Dela et al. (2003). Hydrolyzed anthocyanin samples were analyzed by HPLC (Shimadzu, Japan) equipped with an LC-10AT, an SCL-10A, and an SPD-M10AVP photodiode array detector. Separation was performed on a RP-C18 column (201TP54, Grace Vydac) at 27°C with the following solutions: (A) H2O, pH 2.3 and (B) H2O:MeCN:HOAc (107:50:40), pH 2.3. The solutions were applied as a linear gradient from a ratio of 4:1 (A:B) to 3:7 over 45 min, and held at a ratio of 3:7 for an additional 10 min at a flow rate of 0.5 ml min−1. Anthocyanidins and flavonols were identified by comparing both the RT and the absorption spectrum from 250 to 650 nm to those of standard purified anthocyanidins and flavonols (obtained from Apin Chemicals, UK; Polyphenols, Norway; Sigma Aldrich, USA).

#### Extraction and Determination of Peroxidase Activity

Peroxidase (POD) activity of *Vanda* florets was determined as previously described (Gerailoo and Ghasemnezhad, 2011), with some modifications. Enzyme extraction for POD activity was prepared by homogenization of 2 g of floret petal samples in 20 ml extraction buffer composed of 50 mM phosphate buffer, pH 7, and 1% PVPP (w/v). The samples were centrifuged for 30 min at 12,000× *g*, and the supernatant was used to determine enzyme activity. POD activity was assayed by measuring spectrophotometrically the formation of guaiacol in l ml reaction mixture composed of 450 μl guaiacol 25 mM, 450 μl H2O2 225 mM, and 1 ml crude enzyme. The formation of tetraguaiacol was measured at 470 nm, and the activity was expressed as units per mg protein. Protein concentrations were measured as described by Bradford (1976).

### Statistical Analysis

Experiments were arranged in a completely randomized design (CRD), with 6–8 replicate stems for each treatment. Data were analyzed using ANOVA, and differences among means were compared using Tukey Test.

## RESULTS

Exposure of cut *Vanda* 'Sansai Blue' inflorescences to two ethylene concentrations for 24 h resulted in a dramatic bleaching of the florets at development stages B (colored bud) and C (half bloom) already during the 24 h of ethylene treatment (**Figure 1A**). On the other hand, in the more developed stages, D (bloom) and E (full bloom), the decrease in petal pigmentation

was less noticeable at the end of the ethylene exposure (**Figure 1A**). Indeed, the appearance of a bunch of inflorescences held in the vase after their removal from the atmosphere of both ethylene concentrations, showed a faded color compared to control, but they were still blueish because most florets were in stages D and E (**Figure 1B**). It should be noted that the fast bleaching of the florets at developmental stages B and C (**Figure 1A**), and the visual color appearance of the whole inflorescences at day 0 (**Figure 1B**) were similar for the two ethylene concentrations. By contrast, the significant reduction in their vase life longevity following exposure to ethylene was concentration-dependent (**Figure 1C**). Thus, exposure of the inflorescences to 1 or 10 μl L−1 ethylene significantly shortened their vase life to 7.2 and 5.2 days, respectively, as compared to control flowers which lasted for 12 days of vase life (**Figure 1C**).

In order to further examine the effect of ethylene on flower pigmentation during vase life, the inflorescences were exposed to 10 μl L−1 ethylene or to 0.2 μl L−1 1-MCP followed by 10 μl L−1 ethylene. The results depicted in **Figure 2** demonstrate that the color intensity of the ethylene-treated flowers dramatically decreased on days 4 and 8, and this effect was completely inhibited by the 1-MCP pretreatment, which recovered the color appearance to that of control flowers at these time points (**Figure 2**).

A similar pattern of changes in response to these treatments was obtained in the anthocyanin content of florets, analyzed during 2 days of vase life after treatment. Thus, ethylene treatment significantly reduced the anthocyanin content of florets at the developmental stages of colored bud (**Figure 3A**) and half bloom (**Figure 3B**) on day 0, while the ethyleneinduced reduction was less significant for florets at bloom stage (**Figure 3C**). On the other hand, on day 2, the reduction in anthocyanin content was significant for all three developmental stages (**Figure 3**). 1-MCP pretreatment abolished completely the ethylene effect on the anthocyanin content in all floret development stages at both time points, and recovered the anthocyanin levels to those of control florets (**Figure 3**).

More than eight anthocyanin peaks were observed in the chromatogram of *Vanda* 'Sansai Blue' florets at bloom stage following the HPLC analysis of non-hydrolyzed samples (**Figure 4A**). Two major anthocyanins were separated at RT of 27.6 and 33.8 min, and six less abundant anthocyanins appeared at RT of 19.5, 22.5, 25.0, 30.4, 36.0, and 38.9 min. The content of all these anthocyanins (peak height and area) remained similar in control florets at bloom stage after 2 days of vase life (**Figure 4D**). A drastic decrease of all anthocyanins could be observed in the ethylene-treated flowers, when flowers were removed from the ethylene atmosphere (day 0) (**Figure 4B**), and the small remaining residues of anthocyanin peaks further decreased to almost nullified levels on day 2 (**Figure 4E**). 1-MCP pretreatment prevented completely and very efficiently the ethylene-enhanced effect on the anthocyanin degradation

both on day 0 (**Figure 4C**), and on day 2 (**Figure 4F**). Thus, the 1-MCP pretreatment recovered the anthocyanin levels to those of control untreated flowers (**Figures 4A,D**).

Similar results were obtained also for florets analyzed at the half bloom stage, in which the anthocyanins were nullified already on day 0 (data not shown). A quantitative data for the HPLC analysis described above is presented in **Figure 5** for total anthocyanins (the sum of peak area of all anthocyanins), expressed as cyanidin-3-glucoside equivalents. The data show that the ethylene treatment nullified the content of total anthocyanins during the ethylene exposure of florets at half bloom stage (**Figure 5A**), while at the bloom stage there was a continuous degradation of anthocyanins in response to ethylene between day 0 and day 2 (**Figure 5B**). 1-MCP pretreatment prevented the ethylene-induced anthocyanin degradation, and the levels of anthocyanins at half bloom stage even increased over control levels (**Figure 5A**), suggesting that anthocyanins were also synthesized during this period.

Analysis of hydrolyzed anthocyanin samples revealed that most of the anthocyanin pigments in *Vanda* 'Sansai Blue' flowers were based on delphinidin and cyanidin backbones (**Figure 6**). The results confirmed the previous findings, and show that ethylene reduced all anthocyanidins during the treatment at bloom stage florets (**Figure 6B**). Kaempferol was found to be the dominant flavonol in *Vanda* 'Sansai Blue' flowers (**Figure 7**), but unlike the anthocyanidins, the flavonols were not affected by the ethylene treatment (**Figure 7B**).

Anthocyanin degradation was reported to be mediated by enzymatic activity of the class III peroxidase (POX) (Zipor et al., 2014; Movahed et al., 2016; Lecourieux et al., 2017; Pastore et al., 2017). Therefore, in an attempt to investigate the mechanism of anthocyanin degradation in cut *Vanda* 'Sansai Blue' flowers, we have examined the effect of ethylene treatment on total peroxidase (POD) activity at different developmental stages (**Figure 8**). The results show a significant ethylene-induced increase in POD activity on day 2 in the colored bud stage (**Figure 8A**). 1-MCP pretreatment resulted in the lower POD activity, generally at all developmental stages, and it inhibited completely the ethylene-induced increase in POD activity (**Figure 8**). These results suggest that the ethylene-induced anthocyanin degradation in the cut *Vanda* 'Sansai Blue' flowers seems to be mediated by increased POD activity.

### DISCUSSION

#### Ethylene Induces Color Fading in *Vanda* Flowers Independently of Senescence

The effect of exogenous ethylene treatment on cut *Vanda* 'Sansai Blue' flowers was very dramatic and rapid, and was already pronounced during the 24 h of exposure to the ethylene atmosphere (**Figures 1A,B**). In the younger flower developmental stages of colored buds and half bloom, the florets became almost completely white, while in the more advanced developmental stages, the bleaching was only partial during the ethylene treatment (**Figure 1B**), and the pigment degradation continued to take place throughout vase life (**Figure 2**). Both ethylene concentrations of 1 and 10 μl L−1 induced the same degree of color fading (**Figures 1A,B, 2**), indicating that the fading-induced effect of ethylene was not concentration-dependent in this range. On the other hand, the effect of ethylene treatment on flower senescence, expressed in a significant reduction of vase life due to floret wilting, was dependent on ethylene concentrations (**Figure 1C**). This suggests that the two processes of the rapid ethylene-induced anthocyanin degradation and the ethylene-induced senescence,

FIGURE 5 | Effect of ethylene and 1-MCP pre-treatments on changes, detected 2 days after treatments, in the content of total anthocyanins extracted from florets at half bloom (A) or bloom (B) developmental stages, based on the chromatogram peak areas of non-hydrolyzed samples of *Vanda* 'Sansai Blue' florets. Ethylene and 1-MCP pre-treatments were applied as detailed in Figure 2, and the anthocyanin level was calculated as cyanidin-3-glucoside equivalents. The results represent means ± SE of three floret replicates per treatment, and different letters indicate significant differences among treatments at the different time points, at *p* < 0.01.

proceed as separate processes at a different timing in cut *Vanda* 'Sansai Blue' flowers.

This conclusion is further supported by the following additional observations: (1) the wilting of the florets occurred first in the florets at the more advanced developmental stages, which are located at the bottom of the inflorescence, and proceeded upward. However, the upper florets that became completely bleached were still turgid and continued to bloom and grow until the end of the experiment (**Figure 2**). The decrease in water uptake in the ethylene-treated flowers

was the reason for the enhanced decrease of the inflorescences FW, but the first wilted florets at the bottom of the inflorescence could be observed only when their FW decreased to about 93% of the initial value (Khunmuang et al., 2016, 2018). (2) The ethylene-induced bleaching of cut *Vanda* 'Sansai Blue' flowers was not accompanied by wilting symptoms or any other well-documented senescence parameters, such as ion leakage, protein degradation, and increased amino acid content (Mayak, 1987; Van Doorn, 2001; Van Doorn and Woltering, 2008; Rogers, 2013; Dar et al., 2014), during more than 7 days after treatment (Khunmuang et al., 2019). The presented results further suggest that the rapid ethyleneinduced anthocyanin degradation of cut *Vanda* 'Sansai Blue' flowers proceeds as a separate process, independently from the well-characterized senescence-associated processes.

A similar fast ethylene-induced color fading was reported long ago in *Vanda* flowers, but in an indirect manner in response to pollination or emasculation, which induced high levels of endogenous ethylene production (Akamine, 1963; Burg and Dijkman, 1967; Goh et al., 1985). Thus, emasculation of *Vanda* 'Rose Marie' resulted in increased ethylene evolution, which started after a 10-h lag period, and fading became evident after additional 8–12 h (Burg and Dijkman, 1967). A similar time course of ethylene evolution and fading was reported for *Vanda* 'Miss Agnes Joaquim' flowers after their emasculation, in which ethylene production was correlated with the degree of color fading (Akamine, 1963). In *Vanda* 'Petamboerant' flowers, endogenous ethylene production rates increased in control flowers after 75 h, while pollination or emasculation enhanced the process by inducing ethylene production within 1 or 28 h, respectively (Burg and Dijkman, 1967). Consequently, the lower petals of pollinated or emasculated *Vanda* 'Petamboeran' flowers started to fade after 8–10 or 35 h, respectively, as compared to petals of control flowers which faded after about 80 h (Burg and Dijkman, 1967). Also in other orchid flowers of *Phalaenopsis* (Porat et al., 1995) and *Dendrobium* 'Pompadour' (Ketsa and Rugkong, 2000), pollination induced endogenous ethylene production, which enhanced their senescence.

chromatographed. The results in graph (B) represent means ± SE of three floret replicates per treatment. NS, not significant.

A more direct effect of exogenous ethylene was reported recently, describing pulsing of cut *Vanda* 'Sansai Blue' flowers with the ethylene-releasing compound, ethephon. This treatment resulted in reduced water uptake and vase life longevity, and this effect was prevented by 1-MCP pretreatment, but the effect on color bleaching was less visible (Khunmuang et al., 2016). This discrepancy between the treatments could be ascribed to the fact that the ethylene released by ethephon pulsing affected more the stems, and less directly the florets at the younger developmental stages. In other orchid flowers, application of 10 μl L−1 ethephon by dipping for 5 min decreased significantly the water uptake of cut *Dendrobium* 'Planty Fushia' flowers, and reduced their vase life. Termination of vase life was due to senescence and abscission of the open florets at the bottom of the inflorescences, and prevention of opening of the florets at the bud stages that were abscised (Mohammadpour et al., 2015). These ethephon effects on cut *Dendrobium* "Planty Fushia" flowers were completely inhibited by 1-MCP pretreatment. Application of 0.25–2 μl L−1 1-MCP to cut *Dendrobium* 'Burana Jade' flowers was even more effective in maintaining the FW as compared to control flowers (Yoodee and Obsuwan, 2013). This indicates that 1-MCP is effective also in inhibition of endogenous ethylene.

#### Anthocyanin Degradation Is Responsible for the Rapid Color Fading

The visible bleaching of the cut *Vanda* 'Sansai Blue' flowers (**Figures 1A,B, 2**) was due to anthocyanin pigment degradation induced by the ethylene treatments (**Figure 3**). The reduction in anthocyanin content in florets at half bloom and bloom stages in response to ethylene was quantitatively quite similar during the ethylene treatment (**Figure 3**). The light blueish color (**Figure 1B**) on day 0 and during vase life (**Figure 2**) observed in the ethylene-treated flowers reflected the pigment residues in the florets at more advanced developmental stages. The initial content of the anthocyanins in the more advanced

(*Continued* )

developmental stages (**Figure 3C**) was much higher than in the early developmental stages (**Figure 3A**), and the anthocyanins continued to decrease in the ethylene-treated flowers (**Figures 2, 3**). The higher anthocyanin contents in the more developed floret stages (**Figure 3C**) indicate that anthocyanins continued to be synthesized in the flowers until reaching the full bloom stage. 1-MCP pretreatment completely inhibited the ethylene-induced floret bleaching (**Figure 2**), and anthocyanin degradation (**Figures 3–5**). Indeed, the visual appearance on day 8 of inflorescences pretreated with 1-MCP shows that their color was more pronounced than that of the control untreated flowers (**Figure 2**), but on the other hand, more florets with "sleepiness" symptoms at the bottom were detected (Khunmuang et al., 2019).

The results presented in **Figure 3** clearly demonstrate that anthocyanin content increases with development (zero time in each stage), but this process takes time (about 10 days from bud to a fully open flower). On the other hand, during the initial 2 days of vase life, anthocyanin levels did not change in the control samples of each developmental stage, but they were reduced significantly in the ethylene-treated samples, and this reduction was inhibited by 1-MCP. These results clearly show that the decreased anthocyanin levels resulted from the ethylene-induced degradation rather than from inhibition of anthocyanin biosynthesis.

Our previous study showed that while the effect of ethylene treatment in reducing the vase life longevity was similar in three cut *Vanda* cultivars, the effect of ethylene on flower color bleaching and anthocyanin content varied among the cultivars and floret stages (Khunmuang et al., 2019). The anthocyanin content of *Vanda* 'Pure Wax' flowers was almost unaffected by ethylene, except at the bud stage, while in cut *Vanda* 'Patchara Delight' flowers it was partially reduced after 2 days of vase life, mainly in the full bloom developmental stage. This suggests that the three *Vanda* cultivars differ in their sensitivity to ethylene, which is expressed at different developmental stages, and the *Vanda* 'Sansai Blue' flowers are unique among the three cultivars in their high sensitivity to ethylene manifested in the rapid color fading.

Variation in ethylene sensitivity may be related to differences in the concentration and affinity of the ethylene receptors and/ or to the activity of downstream components in the signal transduction pathway, which activates gene transcription and translation (Bleecker and Kende, 2000). It was previously suggested that *Vanda* 'Sansai Blue' flowers are sensitive to ethylene but the flower itself produced very low amount of ethylene (Goh et al., 1985; Khunmuang et al., 2016, 2018).

### Ethylene Induces Degradation of Anthocyanins *in planta*

More than eight anthocyanins were observed in *Vanda* 'Sansai Blue' flowers analyzed at the bloom stage by HPLC of unhydrolyzed samples, and all of them were fast degraded by ethylene treatment (**Figure 4**). Pigment accumulation, anthocyanin structure and the expression of floral anthocyanin genes were analyzed in anthocyanin-based colored florets of a pale-mauve *Vanda* hybrid (*V. teres* × *V. hookeriana*) (Junka et al., 2011). The anthocyanins gradually accumulated during all the developmental stages of the florets. Based on HPLC and LC-ESI-MSn analyses, the anthocyanins in the pale-mauve hybrid were composed of only five types of cyanidin derivatives, which were diversely conjugated with some hexose sugars and organic acids, such as ferrulic, sinapic, and malonic acids (Junka et al., 2011). On the other hand, more than 11 anthocyanins were observed in the violet-blue and red-purple flowers of the *Vanda* hybrid cultivars, from which eight major acylated anthocyanins were isolated (Tatsuzawa et al., 2004). Four of those pigments were based on cyanidin 3,7,3′-triglucoside, and the other four pigments were based on delphinidin 3,7,3′-triglucoside as their deacylanthocyanins. The distribution of these pigments was investigated in the flowers of four species and 13 hybrids by the analytical process of HPLC. Unfortunately, the *Vanda* 'Sansai Blue' was not included in this survey, but our anthocyanins hydrolyzed extract analysis revealed that most anthocyanidins in this cultivar are based on both delphinidin and cyanidin as well (**Figure 6**). The acylated anthocyanins of cyanidin and delphinidin contribute to make the blue flower color in the *Vanda* cultivars, as well as the presence of delphinidin glycosides (Tatsuzawa et al., 2004). This is generally true according to previous studies of orchids, which indicated that the bluing effect was dependent on the numbers of hydroxycinnamic acids (Lu et al., 1992; Honda and Saito, 2002).

The main flavonols that were found in *Vanda* 'Sansai Blue' florets were kaempherol and another unknown flavonol, which were not affected by the ethylene treatment (**Figure 7**). It seems, therefore, that they are not involved in the ethyleneinduced color fading. Based on these results, it is clear that the ethylene-induced bleaching of *Vanda* 'Sansai Blue' flowers occurring *in planta* is ascribed only to the degradation of anthocyanins and not of flavonols.

### Ethylene-Induced Anthocyanin Degradation Is Mediated by Peroxidase

Anthocyanin degradation has been detected *in vivo* in some systems such as the loss of red pigmentation in maturing leaves of *Photinia* spp. (Oren-Shamir and Nissim-Levi, 1999). The enzymatic degradation hypothesis was strongly supported by the investigation on *B. calycina* Benth., in which active anthocyanin degradation by oxidation was reported *in planta* (Vaknin et al., 2005; Oren-Shamir, 2009; Zipor et al., 2014). This loss of color from dark purple to white was dependent on anthocyanin degradation, and *de novo* synthesis of mRNAs and proteins during the different stages of development, well before flower senescence has started (Vaknin et al., 2005), similar to the results reported in the present study for *Vanda* 'Sansai Blue' flowers. A candidate peroxidase was partially purified and characterized, its intracellular localization was determined for *B. calycina* flowers (Zipor et al., 2014), and the transcript sequence of this peroxidase was fully identified. A basic peroxidase, BcPrx01, was responsible for the *in planta* degradation of anthocyanins in *B. calycina* flowers. BcPrx01 had the ability to degrade complex anthocyanins, it co-localized with these pigments in the vacuoles of petals, and both the mRNA and protein levels of BcPrx01 were greatly induced in parallel to the degradation of anthocyanins (Zipor et al., 2014). Recent studies confirmed the degradation of anthocyanins *in planta* by peroxidases, which exhibited higher activity at elevated temperatures (Movahed et al., 2016; Lecourieux et al., 2017; Pastore et al., 2017). Overexpression of the grapevine peroxidase gene (*Vvi-Prx31*) decreased anthocyanin contents in *Petunia hybrida* petals under heat stress condition, suggesting that a high temperature can stimulate peroxidase activity and anthocyanin degradation in ripening grape berries (Movahed et al., 2016). BcPrx01 and Vvi-Prx3 are vacuolar peroxidases, belonging to the class III peroxidase family (POX), and are able to catalyze the reduction of toxic H2O2 that reaches the vacuoles by oxidizing a variety of secondary metabolites (Hiraga et al., 2001).

Class III plant peroxidases (POXs) are plant-specific oxidoreductase, which participate in lignification, suberization, auxin catabolism, wound healing, and defense against pathogen infection (Hiraga et al., 2001). Studies have provided information on the regulatory mechanisms of wound- and pathogen-induced expression of some *POX* genes. These studies suggest that *POX* genes are induced *via* different signal transduction pathways from those of other known defense-related genes (Ishige et al., 1993; Ito et al., 1994). Furthermore, high temperature (35°C) increased ethylene production and concentration of H2O2, and the activity of POX in plum fruit (Niu et al., 2017). It seems, therefore, that ethylene is an enhancer of peroxidase activity. It is important to emphasize that in the present study, total peroxidase (POD) was assayed, rather than the specific class III peroxidase (POX). This may explain the insignificant effects of ethylene on peroxidase activity obtained on day 0 at all assayed developmental stages (**Figure 8**). Nevertheless, we could demonstrate that ethylene treatment significantly increased total POD activity on day 2 at the bud developmental stage, and this effect was inhibited completely by 1-MCP pretreatment (**Figure 8A**). These results are consistent with the results of anthocyanin content at the bud stage (**Figure 3A**). This may indicate that POD activity is highly affected by ethylene at the bud stage, relative to the other developmental stages. Since the anthocyanins accumulate in the vacuole, it is necessary that the vacuolar peroxidase (POX) will degrade them. Our results indicate that at the bud stage, the activity of POX in the POD extract is relatively high, and therefore the results for the ethylene effects at this stage were significant (**Figure 8A**). Alternatively, since the anthocyanins are synthesized in the cytoplasm at the bud stage, it is still possible that they are degraded by the cytoplasmic peroxidase in response to ethylene, before reaching the vacuole. Additionally, the insignificant effects of ethylene on POD activity obtained at the other developmental stages (**Figures 8B,C**), may indicate that another unknown mechanism is possibly involved in this process, which needs a further study.

Taken together, our results suggest that the ethylene-induced anthocyanins degradation in cut *Vanda* 'Sansai Blue' flowers seems to be mediated by increased POD activity *in-planta*. This effect, which is fast and independent of the flower senescence process, is mainly expressed at the bud developmental stage, in which the anthocyanin degradation was most prominent.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the supplementary files.

#### AUTHOR CONTRIBUTIONS

SKh, MB, SM, SP-H, CW-A, and SKa were responsible for the conception, design of the experiments, and interpretation of data.

#### REFERENCES


SKh performed the laboratory experiments and the HPLC analyses. MO-S and RO were responsible for the anthocyanin analyses and determination. SKh, MB, SM, and SP-H were involved in drafting the work. SP-H and SM were responsible for the writing, editing, and final approval of the version to be published. All authors revised and approved the final version.

#### FUNDING

The research was supported by the Royal Golden Jubilee (RGJ) Ph.D. Program (Grant No. PHD/0120/2553), the Thailand Research Fund (TRF).

#### ACKNOWLEDGMENTS

We would like to thank the Postharvest Technology Innovation Center, Office of the Higher Education Commission, Bangkok, Thailand, for the laboratory facilities.

*New Zealand J. Crop Hortic. Sci.* 26, 319–324. doi: 10.1080/01140671. 1998.9514070


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Khunmuang, Kanlayanarat, Wongs-Aree, Meir, Philosoph-Hadas, Oren-Shamir, Ovadia and Buanong. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Shaping Ethylene Response: The Role of EIN3/EIL1 Transcription Factors

*Vladislav A. Dolgikh1,2, Evgeniya M. Pukhovaya1,2 and Elena V. Zemlyanskaya1,2\**

*1 Institute of Cytology and Genetics, Siberian Branch of Russian Academy of Sciences, Novosibirsk, Russia, 2 Department of Natural Sciences, Novosibirsk State University, Novosibirsk, Russia*

#### *Edited by:*

*Jin-Song Zhang, Institute of Genetics and Developmental Biology (CAS), China*

#### *Reviewed by:*

*Xing Wen, Southern University of Science and Technology, China Hong Qiao, University of Texas at Austin, United States Shangwei Zhong, Peking University, China*

#### *\*Correspondence:*

*Elena V. Zemlyanskaya ezemlyanskaya@bionet.nsc.ru*

#### *Specialty section:*

*This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science*

*Received: 30 April 2019 Accepted: 23 July 2019 Published: 26 August 2019*

#### *Citation:*

*Dolgikh VA, Pukhovaya EM and Zemlyanskaya EV (2019) Shaping Ethylene Response: The Role of EIN3/EIL1 Transcription Factors. Front. Plant Sci. 10:1030. doi: 10.3389/fpls.2019.01030*

EIN3/EIL1 transcription factors are the key regulators of ethylene signaling that sustain a variety of plant responses to ethylene. Since ethylene regulates multiple aspects of plant development and stress responses, its signaling outcome needs proper modulation depending on the spatiotemporal and environmental conditions. In this review, we summarize recent advances on the molecular mechanisms that underlie EIN3/EIL1 directed ethylene signaling in Arabidopsis. We focus on the role of EIN3/EIL1 in tuning transcriptional regulation of ethylene response in time and space. Besides, we consider the role of EIN3/EIL1-independent regulation of ethylene signaling.

Keywords: ETHYLENE-INSENSITIVE3, ETHYLENE-INSENSITIVE3-LIKE, epigenetic regulation, protein–protein interactions, cross-talk

#### KEY COMPONENTS OF ETHYLENE SIGNALING PATHWAY

Plant hormone ethylene coordinates numerous developmental processes (including germination, soil emergence, seedling growth, fruit ripening, senescence, abscission, etc.), as well as diverse biotic and abiotic stress responses (Abeles et al., 2012). Ethylene has also been shown to induce typical morphological changes in dark-grown seedlings (inhibition of hypocotyl and root elongation, radial swelling of hypocotyl, and exaggeration of apical hook) known as "the triple response" (Ecker, 1995). Ethylene is produced from *L*-methionine, which is consequently converted to *S*-adenosyl-*L*-methionine (by SAM-synthetases), 1-aminocyclopropane-1-carboxylic acid (ACC) (by ACC synthases), and ethylene (by ACC oxidases) (reviewed in Booker and DeLong, 2015). Ethylene is perceived by a family of receptors (ETHYLENE RESPONSE 1, ETR1; ETHYLENE RESPONSE SENSOR 1, ERS1; ETR2, ETHYLENE INSENSITIVE 4, EIN4; and ERS2 in Arabidopsis) localized in the endoplasmic reticulum (ER) membrane (reviewed in Lacey and Binder, 2014). Upon binding, ethylene inactivates them and thereby blocks the serine–threonine protein kinase CONSTITUTIVE TRIPLE RESPONSE 1 (CTR1) activity promoting the cleavage of ER-anchored EIN2 protein (reviewed in Chang, 2016; Hu et al., 2017). EIN2 C-terminal domain (EIN2-C) released upon cleavage indirectly triggers EIN3 and EIN3-Like (EIL) transcription factors (TFs) that are considered the key transcriptional regulators of ethylene response (**Figure 1**). Noteworthy, these TFs function as a hub that integrates and processes different cues to "shape" ethylene response in accordance with spatiotemporal and environmental conditions. Below, we will focus on the nuclear events that conduct EIN3/EIL activation and set their functional output.

### ACTIVATION OF EIN3 AND ITS HOMOLOGS IN RESPONSE TO ETHYLENE

EIL is a small family of plant-specific proteins. There are six genes encoding the members of this family in *Arabidopsis thaliana* genome (*EIN3*, *EIL1*-*5*) (Chao et al., 1997; Guo and Ecker, 2004). They harbor a conservative N-terminal DNA-binding domain with a unique fold structure (Song et al., 2015). EIN3, EIL1, and EIL2 represent functionally homologous proteins involved in the regulation of ethylene-responsive genes (Chao et al., 1997; Solano et al., 1998; Alonso et al., 2003; An et al., 2010). The most closely related EIN3 and EIL1 are considered the major regulators since *ein3 eil1* double mutants show complete ethylene insensitivity in terms of the triple response, pathogen resistance, and the ability to fully suppress *ctr1* mutation (reviewed in Guo and Ecker, 2004; Cho and Yoo, 2015). Two paralogs differentially regulate ethylene response in the seedlings (EIN3) and in adult leaves and stems (EIL1) (An et al., 2010). Yet, a minor, *EIL2* role in the regulation of ethylene response is supported by its capability to complement *ein3* mutation when overexpressed (Chao et al., 1997). In **Figure 2**, we visualized tissue-specific expression levels of *EIL* genes based on publicly available data on transcriptome profiling in different Arabidopsis tissues retrieved from ThaleMine v1.10.4 (https://apps.araport.org/thalemine/; Krishnakumar et al., 2017). Unlike *EIN3* and *EIL1*, *EIL2* transcripts level is low throughout plant tissues; moderate *EIL2* expression is restricted to root apical meristem and pollen (**Figure 2**). Therefore, EIL2 function could be limited to specific spatiotemporal conditions. EIL3/SLIM1 does not function in ethylene pathway but regulates sulfur deficiency response; no defined roles of EIL4 and EIL5 have been reported to date (reviewed in Guo and Ecker, 2004; Wawrzyńska and Sirko, 2014).

EIN3 and EIL1 activation in response to ethylene is the target for complex regulation. EIN3 and EIL1 are short-living proteins that undergo ubiquitination and proteasomal degradation driven by ubiquitin-ligases EIN3 BINDING F-BOX1 (EBF1) and EBF2 (**Figures 1** and **3A**) (Gagne et al., 2004; An et al., 2010). Stabilization of EIN3/EIL1 upon ethylene release plays a pivotal role in triggering ethylene-directed gene expression. Ethylene dampens EBF1/2 levels *via* i) translational repression of *EBF1*/*2* mRNA in the cytosol promoted by EIN2-C (Li et al., 2015; Merchante et al., 2015), and ii) EIN2-dependent proteasomal degradation of EBF1/2 proteins (An et al., 2010) (**Figure 1**). Stabilized EIN3/EIL1 accumulate in the nucleus.

EIN3/EIL1 are predominantly transcriptional activators (Chang et al., 2013; reviewed in Cho and Yoo, 2015). In Arabidopsis, EIN3, EIL1, and EIL2 specifically bind a short DNA sequence referred to as EIN3 binding site (EBS) in gene promoters (**Figure 1**) (Solano et al., 1998; Chang et al., 2013; Song et al., 2015; O'Malley et al., 2016). EIN3 binds its target loci as a homodimer, and the dimerization is DNA independent (Solano et al., 1998; Song et al., 2015). Accordingly, EIN3 demonstrates higher binding affinity to the inverted repeats of EBS compared to the monomeric site in the *in vitro* experiments (Song et al., 2015). EIN3 binding to the targets is facilitated by elevated levels of H3K14 and non-canonical H3K23 histone acetylation both promoted by a EIN2-C-scaffolded histone acetylation complex, which is triggered upon EIN2-C interaction with a histone binding protein EIN2 NUCLEAR ASSOCIATED PROTEIN 1 (ENAP1) (Zhang et al., 2016; Zhang et al., 2017; reviewed in Wang and Qiao, 2019) (**Figure 3**). Since neither EIN2-C nor ENAP1 possess histone acetyltransferase domains, they might recruit other proteins to promote histone modifications. EIN3 is capable of interacting with ENAP1, too, and it is thought to contribute to ethylene-induced elevation of H3K14 and H3K23 acetylation as well (Zhang et al., 2016).

FIGURE 3 | Nuclear events that promote ethylene response. (A) Without ethylene, EIN3 undergoes EBF1/2-driven degradation. (B) Upon ethylene treatment, EIN3 is stabilized. On one hand, EIN2 C-terminal domain interacts with ENAP1, which results in elevation of H3K14Ac and H3K23Ac levels, facilitated EIN3 binding to the target promoters and activation of gene expression. On the other hand, SRT1 and SRT2 histone deacetylases mediate ethylene-directed transcriptional repression by downregulating the levels of H3K9 acetylation. The models are based on the findings reported previously (Gagne et al., 2004; Li et al., 2015; Merchante et al., 2015; Zhang et al., 2016; Zhang et al., 2017; Zhang et al., 2018a). Gray and white solid circles depict negative and positive regulators of ethylene signaling, correspondingly. EIN3 is depicted in orange, H3K9Ac—in red, H3K14Ac and H3K23Ac—in blue. Dashed circles denote putative regulators (with a question mark inside) and putative regulations (with a question mark outside). HAT, histone acetyltransferase; 26S, 26S proteasome.

Along with well-known EIN3/EIL1-promoted gene transcriptional activation, ethylene downregulates a considerable set of genes (Chang et al., 2013; Harkey et al., 2018). In a recent work, Zhang et al. (2018a) demonstrated that histone deacetylases SRT1 and SRT2 mediate transcriptional repression in response to ethylene by downregulating the levels of H3K9 acetylation (at least for a particular set of ethylene-repressed genes) (**Figure 3**). Both deacetylases interact with ENAP1, and the function of *SRT2* is *EIN2*- and *EIN3/EIL1*-dependent. The mechanism used to distinguish between the activator and repressor pathways as well as the role of EIN3/EIL1 in SRT1/2-mediated gene repression are still unclear and need further investigations.

#### EIN3/EIL1-REGULATED TRANSCRIPTIONAL NETWORKS

Upon DNA binding, EIN3/EIL1 modulate multiple transcriptional cascades. Ethylene-sensitive EIN3 target genes encoding TFs include *ERF1*, involved in a range of ethylene responses (Solano et al., 1998), *PIF3*, *RSL4*, *ESE1*, and *CBF1/2/3*, the regulators of de-etiolation, root hair development, salt and cold stress responses, correspondingly (Zhang et al., 2011; Shi et al., 2012; Zhong et al., 2012; Feng et al., 2017). To supplement this list, numerous TF-encoding genes comprising representatives of AP2/ERF, WRKY, NAC, and other families were retrieved from whole-genome data on EIN3 binding and ethylene-induced transcriptomes (Chang et al., 2013). Besides, EIN3 directly regulates expression of chlorophyll biosynthesis genes *PORA/B* (Zhong et al., 2009), the pigment-binding proteins *LHC* essential for photosynthesis initiation (Liu et al., 2017), the immune receptor *FLS2* (Boutrot et al., 2010), and the apical hook formation regulator *HLS1* (Lehman et al., 1996; Shen et al., 2016). EIN3/ EIL1 affect the pathways of many hormones (Chang et al., 2013), including direct regulation of hormones biosynthesis (e.g., salicylic acid biosynthesis gene *SID2*, Chen et al., 2009), and signaling (e.g., type-A negative regulators of cytokinin signaling *ARR5/7/15*, Shi et al., 2012). To maintain a homeostasis, EIN3 activates a feedback regulatory circuit by inducing transcription of *EBF2* (Konishi and Yanagisawa, 2008) and probably some other negative regulators of ethylene signaling (Chang et al., 2013).

To provide a proper phenotypic outcome upon ethylene release, these transcriptional cascades and the downstream growth control pathways should be tightly coordinated, which is supported by data on the dynamic changes of ethyleneinduced transcriptomes in etiolated Arabidopsis seedlings where four distinct transcriptional waves are segregated (Chang et al., 2013). The observed transcription kinetics may be due to distinct mechanisms of transcriptional control, or the heterogeneity of the ethylene response in different tissues (Chang et al., 2013). Transcriptome profiling of Arabidopsis mutants identified large groups of EIN3/EIL1-regulated genes that were co-regulated by the other TFs such as RHD6 (root hair development) and PIFs (light signaling) (Feng et al., 2017; Shi et al., 2018), which implies co-regulation of EIN3/ EIL1-triggered transcription by certain developmental and environmental cues. In the following sections, we illustrate that EIN3/EIL1 proteins represent crucial targets for tuning the downstream transcriptional cascades in time and space.

#### TUNING TRANSCRIPTIONAL REGULATION OF ETHYLENE RESPONSE

#### Epigenetic Regulation of Spatiotemporal Expression of EIN3/EIL1 Target Genes

Climacteric fleshy fruits (the ones that demonstrate a respiratory burst at the beginning of ripening) use ethylene as a ripening signal (McMurchie et al., 1972). Mature fruit produces ethylene in an autocatalytic manner (system II) unlike immature fruit and vegetative tissues where self-inhibitory ethylene production (system I) is implemented. Autocatalytic regulation suggests a positive feedback loop controlling ethylene synthesis. Presumably, the corresponding regulatory circuit includes EIN3 triggered transcriptional cascade that finally activates ethylene biosynthesis genes (*ACSes* and *ACOs*) (Vandenbussche et al., 2012; Lü et al., 2018). To prevent uncontrolled ethylene production, this circuit should be under a tight spatiotemporal regulation.

Epigenetic modifications often promote spatiotemporal regulation of plant hormone responses (reviewed in Yamamuro et al., 2016). In Arabidopsis, a repressive mark H3K27me3 regulates expression of a large number of genes (Lafos et al., 2011). A systematic analysis of epigenome and transcriptome data suggests that climacteric fruits use removal of H3K27me3 to trigger autocatalytic system II ethylene production specifically in the mature fruit (Lü et al., 2018). Accordingly, EIN3 targeted promoters—a part of transcriptional feedback circuit controlling climacteric fruit ripening (*RIN* in tomato, *NAC* in peach and banana)—are associated with the repressive histone mark H3K27me3 in leaf and immature fruit. They become demethylated and therefore accessible only in the ripening fruit tissues. Presumably, this epigenetic mechanism prevents autocatalytic ethylene production in vegetative and immature fruit tissues.

Recently, using a systematic analysis of publicly available ChIP-Seq data on EIN3 binding in Arabidopsis, we have demonstrated that EIN3 direct targets are enriched in a chromatin state 4 according to the classification of Sequeira-Mendes et al. (2014), which is associated with H3K27me3 repressive mark (Zemlyanskaya et al., 2017b). Therefore, H3K27me3-associated epigenetic silencing might be a more general mechanism providing spatiotemporal specificity of ethylene response *via* restriction of EIN3 function.

#### Modulation of EIN3/EIL1 Protein Stability

Regulation of EIN3/EIL1 levels *via* the control of the protein stability by EBF1/2 is a pivotal mechanism of EIN3/EIL1 adjustment in ethylene signaling. Simultaneously, it can be affected by environmental factors resulting in a modulation of transcriptional response to ethylene. Plants germinating in the darkness assume a light-regulated developmental program known as skotomorphogenesis, which phenotypically results in rapid hypocotyl elongation, small closed chlorotic cotyledons, and apical hook formation (McNellis and Deng, 1995). EIN3/EIL1 and their target genes (e.g., *HLS1*, *ERF1*, *PIF3*, *PORA/B*) play essential roles in these processes. They contribute in chlorosis and increased apical hook curvature of buried seedlings, induce shortening and thickening of hypocotyl to enhance lifting capacity of the seedling, and finally promote seedlings greening upon light irradiation (Zhong et al., 2009; Zhong et al., 2012; Zhong et al., 2014; Shen et al., 2016).

In the seedlings growing through the soil, EIN3/EIL1 are stabilized by both light signaling and ethylene, which accumulates in response to mechanical pressure. In the former case, E3 ubiquitin ligase CONSTITUTIVE PHOTOMORPHOGENIC 1 (COP1), a central repressor of light signaling, directly targets EBF1/2 for ubiquitination and degradation (Shi et al., 2016a). As seedlings grow toward the surface, light intensity gradually increases. As a result, COP1 activity, which is negatively regulated by photoreceptors (Podolec and Ulm, 2018), gradually decreases attenuating ethylene response.

When the seedling reaches the soil surface, light triggers a dramatic developmental transition known as de-etiolation that leads to immediate termination of ethylene responses. Light-activated photoreceptor phytochrome B (phyB) directly interacts with both EIN3 and EBF1/2 proteins, thereby stimulating robust EIN3 degradation, rapidly turning off ethylene signaling (Shi et al., 2016b; Luo and Shi, 2019).

#### Repression of EIN3/EIL1 Transcriptional Activity

In this section, we consider the cross-talk of ethylene signaling pathway with jasmonic acid (JA) and gibberellins (GA) based on an inhibition of EIN3/EIL1 transcriptional activity due to their physical interactions with repressor proteins (**Table 1**). These protein– protein interactions (PPI) rather prevent EIN3/EIL1 binding to DNA than cause changes in protein stability (Zhu et al., 2011; An et al., 2012; Zhang et al., 2014). JA and ethylene synergistically regulate certain aspects of plant development (such as root hair development and inhibition of root growth) and tolerance to necrotrophic fungi. The transcriptional repressors JASMONATE ZIM-DOMAIN (JAZ) are the master regulators that interact with MYC2, MYC3, and MYC4 TFs and negatively control JA signaling (reviewed in Wasternack and Song, 2017). JAZ proteins interact with EIN3/EIL1 and enhance EIN3/EIL1 binding to HDA6, an RPD-type histone deacetylase (Zhu et al., 2011; Zhu and Lee, 2015). The resulting complex inhibits EIN3/EIL1-mediated transcription. Upon JA treatment, JAZ degrades, attenuating HDA6-EIN3/ EIL1 association and therefore activating EIN3/EIL1. Therefore, pathogenesis-related genes *ERF1* and *ORA59*, directly regulated by EIN3/EIL1, as well as their downstream target *PDF1.2*, are upregulated in response to JA.

At the same time, MYC2, MYC3, and MYC4 transcriptional regulators of JA signaling interact with EIN3/EIL1, inhibiting their function (Song et al., 2014; Zhang et al., 2014). Thus, *ERF1*, *ORA59*, and *PDF1.2* genes are upregulated in *myc2* mutants. This inhibitory mechanism underlies ethylene-JA antagonism. Particularly, JA-directed abolishment of ethylene-promoted apical hook formation proceeds *via* MYC2-mediated attenuation of *HOOKLESS1* (*HLS1*) expression, which is the key regulator of hook development and a direct EIN3/EIL1 target (Lehman et al., 1996; An et al., 2012; Song et al., 2014; Zhang et al., 2014). Additionally, MYC2 targets *EBF1*, inducing its expression and therefore promoting EIN3/EIL1 degradation (Zhang et al., 2014). Noteworthy, the inhibitory effect in the EIN3–MYC2 complex is reciprocal: the interaction suppresses MYC2 activity as well and thereby ethylene attenuates JA-regulated plant defense response against insect attack (Song et al., 2014). Similarly, EIN3 plays an inhibitory role in sulfur deficiency response, forming heterodimers with EIL3/SLIM1 TF and preventing its target gene recognition by EIL3/SLIM1 (Wawrzyńska and Sirko, 2016).

Just as in the case of JA-ethylene synergy, GA enhances apical hook curvature at least partially *via* a release of EIN3/EIL1 from repressor proteins. DELLA proteins are the main transcriptional repressors of GA responses (Sun and Gubler, 2004). Two members of this family (RGA and GAI) are capable of associating with EIN3/EIL1 DNA-binding domain and inhibiting EIN3/EIL1 function (An et al., 2012). In response to GAs, DELLA proteins rapidly degrade, thereby de-repressing EIN3/EIL1-mediated transcription of at least the *HLS1* gene.

#### EIN3/EIL1 Cooperate With Other TFs in an Interdependent Manner

EIN3/EIL1's capability to function cooperatively with the transcriptional regulators of the other signaling pathways provides another possibility to shape spatiotemporal patterns of ethylene response. This cooperation implies the cross-talk of TFs bound to DNA that goes along with the physical interaction of these TFs (**Table 1**). In buried seedlings, the chloroplasts' development is arrested at the etioplast stage, characterized by an immature arrangement of the inner membranes and pigment molecules (Solymosi and Schoefs, 2010; Jarvis and López-Juez, 2013). EIN3 and PHYTOCHROME INTERACTING FACTOR3 (PIF3), a darkness-stabilized transcriptional regulator of light signaling, form an interdependent module that represses chloroplast development in buried seedlings (Liu et al., 2017). Namely, EIN3 and PIF3 directly interact and bind the promoters of *LIGHT HARVESTING COMPLEX* (*LHC*) genes in a cooperative manner to synergistically suppress their expression. Upon light exposure, the levels of EIN3 and PIF3 decrease, and activation of *LHC* expression triggers chloroplast differentiation.

Interestingly, another TF from PIF family, PIF4, interacts with EIN3 as well (Yazaki et al., 2016), and both TFs target *HLS1*, the key regulator of apical hook development (An et al., 2012; Zhang et al., 2018b). However, EIN3 and PIF4 activate *HLS1* transcription independently (Zhang et al., 2018b).

Cooperative regulation also guides ethylene functioning in root hair development. EIN3 promotes root hair elongation by directly activating *RHD6-LIKE4* (*RSL4*) gene (Feng et al., 2017). Besides, EIN3 physically interacts with ROOT HAIR DEFECTIVE6 (RHD6), a major regulator of root hair development that targets *RSL4* as well (Yi et al., 2010; Feng et al., 2017). Both EIN3 and RHD6 co-activate *RSL4* more efficiently than either of them alone (Feng et al., 2017). The role of EIN3–RHD6 cooperative action is most likely not limited to *RSL4* regulation, but rather covers a quite extensive set of genes and contributes to ethylene-promoted root hair initiation as well (Feng et al., 2017). Similarly, in papaya, EIN3 homolog CpEIN3a interacts with CpNAC2, and both TFs directly activate the transcription of carotenoid biosynthesisrelated genes *CpPDS4* and *CpCHY-b* expressed during fruit ripening (Fu et al., 2017). Both TFs possess a combinatory effect on the regulation of their targets.

Besides, EIN3/EIL1 are capable of binding gene promoters and affecting gene expression indirectly *via* physical interactions with other TFs and modulation of their activity (**Table 1**). Increase of auxin biosynthesis in the root tip epidermis *via* upregulation of *TRYPTOPHAN AMINOTRANSFERASE OF ARABIDOPSIS 1* (*TAA1*) plays a pivotal role in ethylene-induced inhibition of root growth (Vaseva et al., 2018). EIN3 targets *TAA1* promoter through a "piggyback" interaction with RESPONSE REGULATOR 1 (ARR1), a transcriptional regulator of cytokinin signaling, thereby enhancing ARR1 transcriptional activity (Yan et al., 2017). Similarly, EIN3/EIL1 interact with FER-LIKE FE DEFICIENCY-INDUCED TRANSCRIPTION FACTOR (FIT), a central regulator of Fe acquisition in roots, activating FIT abundance (Lingam et al., 2011). Moreover, EIN3/EIL1 bridges FIT to the transcriptional Mediator complex to recruit RNA-pol and promote the regulation of iron homeostasis (Yang et al., 2014).

#### EIN3/EIL1-INDEPENDENT ETHYLENE SIGNALING

There is growing evidence that despite their essential role, EIN3/EIL1 TFs are not indispensable components of ethylene response. Thus, kinetic studies distinguish two phases of ethylene-induced growth inhibition of the hypocotyl in etiolated Arabidopsis seedlings: a transient phase I (up to 2 h) and a sustained phase II (Binder et al., 2004; Chang et al., 2013). Both phases require EIN2 function, while only the second requires EIN3/EIL1 (Binder et al., 2004). Intriguingly, unlike etiolated seedlings, light-grown *ein3 eil1* double mutants do not demonstrate the total loss of longterm ethylene responses (Harkey et al., 2018). Moreover, osmotic stress-induced cell cycle arrest in leaf primordia that coincides with enhanced activation of the ethylene signal is EIN3 independent (Skirycz et al., 2011). These observations favor the existence of an alternative pathway. One possible candidate to promote such regulation is a PAM domaincontaining protein EER5. It negatively regulates ethylene signaling during hypocotyl elongation in etiolated seedlings regardless of EIN3 by promoting downregulation of a gene subset upon ethylene treatment. In addition, it physically interacts with EIN2-C (Christians et al., 2008). EER5 regulates magnitude of ethylene response *via* perception of ERS1 signal (Deslauriers et al., 2015).

### CONCLUDING REMARKS AND PERSPECTIVES

Ethylene response is a target for a complex regulation, in which EIN3/EIL1 TFs play a crucial role. Recent studies shed light on multiple layers of complexity in tuning EIN3/EIL1 function (including epigenetic gene silencing and modulation of EIN3/EIL1 stability and activity *via* PPIs) that facilitate the

TABLE 1 | Protein–protein interactions involved in modulation of EIN3/EIL1 function.


*PPI, protein–protein interaction; JA, jasmonic acid; GA, gibberellins; TF, transcription factor; RP, repressor protein; PK, protein kinase.*

"shaping" of ethylene response according to spatiotemporal and environmental conditions. At the same time, these findings open up new perspectives for further research. Growing evidence of the important role that epigenetic landscape plays in EIN3/ EIL1 functioning requires its more detailed characterization. Particularly, the contribution of distinct epigenetic modifications as well as ENAP1 patterns in modulation of EIN3/EIL1 function is of interest. In view of interdependent cooperation of EIN3/ EIL1 with some TFs described recently, the detailed analysis of nucleotide context surrounding EIN3 binding sites requires more attention, and genome-wide research appears helpful both to generalize resent findings and to predict new connections. Moreover, it would be interesting to clarify the role of epigenetic regulation and PPIs in suppression of gene expression upon ethylene treatment.

Yet, despite their essential role, EIN3/EIL1 are not indispensable regulators of ethylene response. To couple the molecular events and phenotypic responses more precisely, EIL2 function in ethylene signaling and EIN3/EIL independent pathways are to be elucidated.

#### REFERENCES


### AUTHOR CONTRIBUTIONS

VAD and EMP performed the literature search and drafted the paper. VAD performed the analysis of the transcriptome datasets. EVZ revised and edited the manuscript. All authors read and approved the final manuscript.

#### FUNDING

The work was funded by the Russian Foundation for Basic Research through grant № 18-29-13040. Meta-analysis of the transcriptome datasets was done in the frame of the project supported by the Russian Foundation for Basic Research and the government of Novosibirsk region through grant № 18-44-540039.

#### ACKNOWLEDGMENTS

We thank Nadya Omelyanchuk for the critical reading of the manuscript and for fruitful discussions.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Dolgikh, Pukhovaya and Zemlyanskaya. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Targeted Proteomics Allows Quantification of Ethylene Receptors and Reveals SlETR3 Accumulation in Never-Ripe Tomatoes

#### *Edited by:*

*Dominique Van Der Straeten, Ghent University, Belgium*

#### *Reviewed by:*

*Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China Maria Manuela Rigano, University of Naples Federico II, Italy*

#### *\*Correspondence:*

*Véronique Santoni veronique.santoni@inra.fr Christian Chervin christian.chervin@ensat.fr*

*†These author share first authorship*

#### *Specialty section:*

*This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science*

*Received: 22 March 2019 Accepted: 29 July 2019 Published: 29 August 2019*

#### *Citation:*

*Chen Y, Rofidal V, Hem S, Gil J, Nosarzewska J, Berger N, Demolombe V, Bouzayen M, Azhar BJ, Shakeel SN, Schaller GE, Binder BM, Santoni V and Chervin C (2019) Targeted Proteomics Allows Quantification of Ethylene Receptors and Reveals SlETR3 Accumulation in Never-Ripe Tomatoes. Front. Plant Sci. 10:1054. doi: 10.3389/fpls.2019.01054*

*Yi Chen1†, Valérie Rofidal2†, Sonia Hem2, Julie Gil1,2, Joanna Nosarzewska1, Nathalie Berger2, Vincent Demolombe2, Mondher Bouzayen1, Beenish J. Azhar3,4, Samina N. Shakeel3,4, G. Eric Schaller4, Brad M. Binder5, Véronique Santoni2\* and Christian Chervin1\**

*1 GBF, Université de Toulouse, INRA, Toulouse, France, 2 BPMP, CNRS, INRA, Montpellier SupAgro, Université de Montpellier, Montpellier, France, 3 Department of Biochemistry, Quaid-i-azam University, Islamabad, Pakistan, 4 Department of Biological Sciences, Dartmouth College, Hanover, NH, United States, 5 Department of Biochemistry, Cellular, and Molecular Biology, University of Tennessee, Knoxville, TN, United States*

Ethylene regulates fruit ripening and several plant functions (germination, plant growth, plant-microbe interactions). Protein quantification of ethylene receptors (ETRs) is essential to study their functions, but is impaired by low resolution tools such as antibodies that are mostly nonspecific, or the lack of sensitivity of shotgun proteomic approaches. We developed a targeted proteomic method, to quantify low-abundance proteins such as ETRs, and coupled this to mRNAs analyses, in two tomato lines: Wild Type (WT) and Never-Ripe (NR) which is insensitive to ethylene because of a gain-of-function mutation in ETR3. We obtained mRNA and protein abundance profiles for each ETR over the fruit development period. Despite a limiting number of replicates, we propose Pearson correlations between mRNA and protein profiles as interesting indicators to discriminate the two genotypes: such correlations are mostly positive in the WT and are affected by the NR mutation. The influence of putative post-transcriptional and post-translational changes are discussed. In NR fruits, the observed accumulation of the mutated ETR3 protein between ripening stages (Mature Green and Breaker + 8 days) may be a cause of NR tomatoes to stay orange. The label-free quantitative proteomics analysis of membrane proteins, concomitant to Parallel Reaction Monitoring analysis, may be a resource to study changes over tomato fruit development. These results could lead to studies about ETR subfunctions and interconnections over fruit development. Variations of RNA-protein correlations may open new fields of research in ETR regulation. Finally, similar approaches may be developed to study ETRs in whole plant development and plant-microorganism interactions.

Keywords: ethylene, receptor, hormone, signaling, tomato

## INTRODUCTION

Ethylene is a plant hormone involved in many developmental processes such as seed germination, root initiation, root hair development, flower development, sex determination, fruit ripening, senescence, and responses to biotic and abiotic stresses (Merchante et al., 2013). Recent research has shown that ethylene sensing is also found in cyanobacteria, such *Synechocystis* (Lacey and Binder, 2016) and possibly in early diverging fungi, such as *Rhizophagus* (Hérivaux et al., 2017).

Ethylene gas is perceived by specific receptors (EThylene Receptors, ETRs) localized at the endoplasmic reticulum (Chen et al., 2002). Since the initial description of the first ethylene receptor, AtETR1 from *Arabidopsis thaliana* (Chang et al., 1993), several studies combining genetics, molecular biology, and biochemistry have led to a model whereby the receptors function as negative regulators and ethylene releases this inhibition (Shakeel et al., 2013; Lacey and Binder, 2014; Ju and Chang, 2015). Thus, ETR abundance may be a critical determinant of ethylene signaling. This is supported in tomato where a study showed that the level of insensitivity to ethylene is related to the expression level of an ETR1 gain-of-function (GOF) mutant (Gallie, 2010). Additionally, other authors observed that ethylene insensitivity, due to a receptor GOF mutant, can be partially overcome with increased gene dosage of WT gene (Hall et al., 1999). In other words, the ethylene signaling may be governed by the relative amount of WT ETRs versus mutant ETRs.

A major bottleneck in understanding ETR roles is the absence of a method to quantify the protein levels of all receptor isoforms in the same sample mainly due to the absence of specific antibodies against ETRs (Chen et al., 2002; Kevany et al., 2007; Mata et al., 2018). Hence, two studies correlating receptor protein abundance using antibodies to transcript levels of each ETR isoform made conflicting observations (Kevany et al., 2007; Kamiyoshihara et al., 2012) raising the need for a better method of ETR protein detection. To reach this objective, a targeted mass spectrometry proteomic method, called parallel reaction monitoring (PRM) was recently described to study ETR receptor abundance in tomato fruit (Mata et al., 2018). We adapted this strategy, focusing on single peptides of rare proteins, to compare the abundance of ETRs in WT and in the NR mutant. In this mutant, *ETR3* harbors a Pro36Leu mutation in the ethylene-binding domain, which renders the plant ethylene insensitive to block fruit ripening as well as downregulating the mRNA levels of *ETR1* and *ETR4* at Breaker stage (Hackett et al., 2000). Additionally, these authors showed that the NR fruit changes from green to orange, but never completes ripening by turning red, due to a lack of lycopene accumulation at the end of the ripening period.

#### RESULTS AND DISCUSSION

#### Development of the PRM Analyses for the Seven Tomato ETRs

To better understand the ETRs roles in the control of important traits such as tomato fruit ripening, it is critical to have a method to quantify the levels of all receptor isoforms at different developmental stages. The tomato (*Solanum lycopersicum)* genome encodes seven ETR isoforms (SlETR1 through SlETR7). Recent advances in large-scale shotgun proteomics have led to identify a large set of proteins including SlETR3 and SlETR4 in green to red ripe tomato fruits using the ITAG 2.3 database (Feb 2013) (Szymanski et al., 2017) and SlETR1, 3 and 4, using the UniProt FASTA database (Dec 2015) in red ripe tomatoes (Mata et al., 2017). In a large-scale labelfree proteomic study, we identified SlETR1, 4, 6, and 7 using the most recent ITAG 3.2 (June 2017), in pooled skin and flesh tissues of both the WT and NR genotypes of the MicroTom cultivar, in four development stages from immature green to Breaker + 8 days (**Table S2a**; **Methods S1b**). These four ETRs were identified but not quantified in all fruit development stages (**Table S2b**). Such large-scale shotgun studies can identify thousands of proteins in biological samples but may result in an under representation of low-abundance proteins such as ETRs. In contrast, targeted approaches such as the PRM performed on quadrupole-Orbitrap mass spectrometers, offers clear advantage in targeting and quantifying lowabundance analytes (Bourmaud et al., 2016).

A PRM strategy was thus developed to identify ETRs in tomato fruit over the ripening period (**Figure 1A**). Microsomal proteins were extracted from tomato fruits at four developmental stages. Proteins were fractionated through SDS-PAGE gel electrophoresis and subsequently digested by trypsin (**Figure 1A**). The success of a PRM-based targeted assay depends on choosing the most appropriate proteotypic peptides for use as specific tracers of each of the proteins of interest (Bourmaud et al., 2016). An *in silico* analysis was performed in order to discriminate between the 7 ETRs and 16 labeled proteotypic peptides (at least 2 proteotypic peptides/ETR) were synthesized (**Table S1a**) and used in a PRM approach to identify the corresponding endogenous ETRs (**Figure 1B**, **Figure S1**).

Among the 16 proteotypic ETR peptides, 15 were identified with high confidence (rdopt > 0.95), except the peptide GLHVLLTDDDDVNR that belongs to ETR5 (rdopt = 0.94) (**Table S1a**, **Figure S1**). Thus, the seven ETRs encoded by the tomato genome were identified in the two genotypes whatever the developmental stage (**Figure 2B**, **Figure S1**). To quantify the ETRs over fruit maturation, the labeled peptides were spiked into a biological matrix using seven adapted peptide concentrations to obtain calibration curves used to determine their quantification limit (**Figure S2**). All identified peptides showed linear regressions with regression coefficients above 0.90 allowing their relative quantification (**Figure S2**). The accumulation profiles of the different peptides for each ETR revealed high correlation coefficients (**Figure 2C**) except in the case of ETR5, likely due to a low protein accumulation during fruit ripening and a limited dynamic range (**Figure S2**). However, the power that reflects the reproducibility of the significance (Zhang and Wen, 2019) appears low with either ETR1 or ETR2 or ETR7 pep3, suggesting that more replicates would be necessary to make better predictions.

FIGURE 1 | PRM workflow for identification and relative quantification of ethylene receptors. (A) Tomato fruits (*Solanum lycopersicum*) from wild-type (WT) plants and NR (never ripe) mutants were collected at four developmental stages: IMG, ImMature Green; MG, Mature Green; Br, Breaker; and Br8, Breaker + 8 days. Membrane proteins were extracted and fractionated through SDS-PAGE electrophoresis. Proteins above 50 kDa were digested with trypsin, and peptides were injected in LC-MS/MS [(nano-HPLC coupled to a quadrupole Orbitrap Qexactive + (Thermo)]. A PRM analysis was optimized through i) the design of ETR peptides, ii) the optimization of PRM parameters, and iii) calibration curves with labeled peptides to identify and relatively quantify ETRs. (B) LC−PRM data validating the identification of ETR1. Heavy peptide (ISPNSPVAR) was spiked into IMG WT biological sample. Selected transitions were extracted for the heavy and endogenous peptides, and rdopt value was calculated using Skyline software (see *Materials and Methods*).

#### Changes in the Seven ETR Proteins Over Tomato Development in WT and NR Backgrounds

Using PRM, we successfully measured the relative amount of the seven ETRs in a series of ripening tomatoes (**Figure 2B**). This showed that the protein levels of ETR1, 2, 5, 6, and 7 dropped from Br to Br8 stages in WT, but this was not the case with ETR3 and ETR4 (**Figure 2B**) indicating that there is a differential regulation of ETRs. In addition, one interesting result is an accumulation of ETR3 in the NR fruit between the mature green stage (MG), and the Breaker + 8 days stage (Br8), which delimits the ripening phase (**Figure 2B**) (Hoeberichts et al., 2002). ETR3 is mutated in the NR background rendering the plant insensitive to ethylene (Wilkinson et al., 1995), and tomato fruit ripening has previously been shown to be blocked by GOF mutations in ETR1 (Okabe et al., 2012). However, since protein content was not determined in earlier studies (Wilkinson et al., 1995; Okabe et al., 2012), our study brings further understanding on how ripening may be blocked in NR fruits at the ETR protein level. Various studies indicate that ethylene acts as a negative regulator. In this model, in air without ethylene, the receptors output leads to inhibition of the ethylene signaling pathway. When ethylene is present, it alleviates this inhibition (Shakeel et al., 2013). Receptors that cannot bind ethylene, such as the mutant ETR3 receptor in the NR background, are thus incapable of turning off. Based on this, we propose that the low levels of the mutant ETR3 in the NR at the early stages of fruit ripening only leads to partial ethylene insensitivity because there is not enough mutant receptor to mask ethylene perception when the other receptor isoforms bind ethylene. In contrast, when mutant ETR3 levels increase at later stages during ripening, the increased signaling from the mutant receptor masks the perception of ethylene by the other receptors.

To evaluate whether such dynamic regulation is possible, we examined the ethylene growth inhibition kinetics of hypocotyls of two *Arabidopsis* ethylene receptor mutants, *etr1-1* and *etr2-1*. The *etr1-1* plants are ethylene insensitive, and *etr2-1* has a large reduction in ethylene sensitivity (Bleecker et al., 1988; Chang et al., 1993; Sakai et al., 1998). *ETR1* is constitutively expressed, whereas *ETR2* occurs at low levels in air and is induced by ethylene within 2 h (Binder et al., 2004a; Hua and Meyerowitz, 1998). Similarly, we observed an induction of *etr2-1* by ethylene (**Figure 3A**). We predicted that if this model of regulation is correct, *etr1-1* seedlings should show no response to ethylene. In contrast, the *etr2-1* seedlings should have a transient growth inhibition response because initially the levels of *etr2-1* are predicted to be too low to block ethylene perception, but upon induction by ethylene, the higher *etr2-1* levels should block ethylene signaling. As shown in **Figure 3B**, WT seedlings had ethylene response kinetics similar to previous studies where growth was inhibited for as long as ethylene was present (Binder et al., 2004b; Binder et al., 2006). In contrast, the *etr1-1* seedlings had no

between the profiles of the various specific peptides tested in this study. *P* is the probability of the correlation; the power values were calculated at the 0.05 risk.

measurable response to ethylene, but did have a slow decline in growth rate over time similar to what has been observed in WT seedlings in air (Binder et al., 2006). Interestingly, *etr2-1* seedlings responded transiently to the application of ethylene with an acceleration in growth rate starting at approximately 2 h after the initial application of ethylene. These results are consistent with our model that proposes that increased levels of a mutant ethylene receptor can cause ethylene insensitivity *in planta*.

The receptors form higher-order complexes much like bacterial chemoreceptors (Shakeel et al., 2013). Thus, in this model, it is possible that the increased levels of mutant receptors are blocking perception by direct interactions between mutant and non-mutant receptors. Alternatively, the increase in mutant receptor levels might be blocking access of WT receptors to downstream effectors such as CTR1. In either case, this model explains why NR fruits start to ripen, but then stop at later stages. This model is consistent with observations in *Arabidopsis* where the ethylene insensitivity of several receptor gain-of-function mutants are overcome by increasing levels of WT receptors (Hall et al., 1999).

The NR mutant fruit fails to turn red (**Figure 4A**), and this is due to a limited accumulation of lycopene (red pigment) as previously shown (Liu et al., 2012). Support for this is that using a large-scale label-free quantitative proteomic approach on the same microsomal extracts, with three biological replicates (see **Table S2**), we observed a decreased accumulation of two key enzymes for lycopene synthesis, zeta-carotene desaturase and phytoene desaturase, in the NR samples compared to WT (**Figure 4B**); ratios around 2.5 show enzymes that were 2.5 fold more present in WT than in NR. Lower accumulation of lycopene in fruits has also been observed with a GOF mutation in ETR1, but the abundance of receptor protein was not determined (Okabe et al., 2012). The label-free data available

difference was found between the *ETR2* allele expression levels in WT and etr2-1 at any time point (t-test, *P* > 0.05). Error bars show SE. (B) Effect of *etr1-1* and *etr2-1* on the short-term ethylene response in etiolated *Arabidopsis* seedlings. A kinetic analysis of hypocotyl growth was carried out on *etr1-1* and *etr2-1* mutants, by time lapse imaging. For comparison, Columbia (wt) seedling responses are included. The seedlings were grown in air for 2 *h*, at which time, 10 µL L**−**1 ethylene was added. The average ± SEM from at least six seedlings is shown.

through the ProteomeXchange database are interesting resources to mine for additional changes occurring at the membrane in the tomato fruit development.

Additionally, the NR mutation led to higher C2H4 production than in WT in ripening tomatoes (**Figure 4C**), which was already observed in ETR1 GOF mutant tomato (Mubarok et al., 2015). Finally, the NR mutation also resulted in higher levels of ETR5 and ETR7 at Br8 (**Figure 2B**). However, because these ETRs are WT proteins and do not harbor mutations altering their sensitivity to ethylene, their higher accumulation is predicted to cause a milder change in ethylene sensitivity, as observed by Hall et al. (1999), as opposed to accumulation of NR, which leads to ethylene insensitivity. Moreover, care should be taken as we observed relative protein values, as discussed below.

Indeed, additional experiments will be necessary to switch from peptide quantification to protein quantification, mainly because of post-translational modifications that can alter the true protein quantification. For instance, the increased abundance of the peptide GNIWIESEGPGK in NR at Br8 stage could be the consequence of *in vivo* serine dephosphorylation, inducing an apparent increase in the quantity of the non-modified peptide. This latter hypothesis would suggest that ETR3 is phosphorylated at this site and less phosphorylated in NR than in WT. It remains an open question whether or not this site is phosphorylated. Other reasons for such discrepancies between peptides are different digestion efficiencies along the protein sequence and partial adsorption of labeled peptides into vials. We obtained similar accumulation profiles for ETR3, ETR4, ETR6, and ETR7 to those reported by Mata et al. (2018) in the WT plant and small differences for ETR1, 2 and 5. Mata et al. (2018) did not test NR. However, there are important differences to note between our study and Mata's study as they used a different cultivar, very different growth conditions, and different methods for protein extraction. Despite these differences, we think both data sets will shed new light on ethylene signaling during fruit development.

#### Are There Positive or Negative Correlations Between the Seven ETR mRNAs and Protein Levels?

Another critical question for ETRs is to understand the relationship between the abundance of mRNA and of the corresponding proteins because prior studies revealed conflicting results about such correlations in the tomato fruit (Kevany et al., 2007; Kamiyoshihara et al., 2012). Therefore, we examined the transcript levels of each *ETR* using qRT-PCR and correlated this information with protein quantification results (**Table S3**, and **Figures 2A**, **B**). With only four points per correlation, the powers, which represent the reproducibility of the significance, are too weak to make any solid conclusion (**Table S3**). Mata et al. (2018) found positive correlations between RNAs and proteins, but the correlation was only significant for ETR3. In WT, a positive correlation was generally observed between RNA and protein levels (**Figures 2A**, **B**), and this can be verified by averaging results of pep1 and pep2, then piling up all RNA data and all protein data to generate the Pearson correlation coefficient, with a total of 28 values. The Pearson correlation is then 0.754, the *P* value is 3.5810−6, and the power is 0.998. This is very global, and suppresses all possible analyses between the different stages and different ETRs, but at least, it validates the positive correlation proposed by Kamiyoshihara et al., (2012) and invalidates the negative correlation proposed by Kevany et al. (2007). In NR, when comparing RNA and protein levels globally, as described above, the Pearson correlation coefficient with a total of 28 values is then 0.586, the *P* value is 1.0410−3, and the power is 0.919. Thus, NR modifies the correlation

FIGURE 4 | Phenotypes and biochemical changes in WT and NR tomato cultivars. (A) Four development fruit stages, used in this study, in two MicroTom tomato lines, WT stands for wild type, and NR stands for never ripe. IMG stands for IMmature Green, MG stands for Mature Green, Br stands for Breaker, and Br + 8 stands for Breaker + 8 days. Both cultivars originate from LE Pereira Peres' Laboratory (Universidade de São Paulo, Brazil) and have been described previously (see *Materials and Methods*). (B) Ratios WT/NR of two enzymes involved in lycopene accumulation as a function of fruit development stages. Data obtained by label-free analysis on the same extracts as for PRM. The protein analysis was performed as described in Methods S1b, using the ITAG 3.2 annotation. The ratios were performed by means of three biological replicates (data available in Supp. Table 2b). \* and \*\*\* stands for P < 0.05 and P < 0.001, respectively, resulting of t-test comparisons between WT and NR means at a similar stage. Carotene desaturase: Solyc01g097810, Phytoene desaturase: Solyc03g123760. (C) Ethylene production by developing fruits (two lines and four stages as described above). The results are the means of three independent biological replicates, error bars show SE, and the small letters show significant differences at 0.05 level (Fisher's LSD test).

compared to WT. When comparing WT to NR, at the Br8 stage, the NR mutation caused a decreased accumulation of mRNA of *ETR1*, *ETR3*, *ETR4*, *ETR6*, and *ETR7* (**Figures 2A**, **B**) suggesting some as yet unknown transcriptional controls. In *Arabidopsis*, the etr1-1 gain-of-function mutation did not cause changes in the transcript levels of the other four receptor isoforms (O'Malley et al., 2005). However, this mutation did result in higher levels of mutant etr1-1 protein compared to ETR1 levels in wild-type plants, even though transcript levels for this receptor were unchanged (Zhao et al., 2002). The mRNA variations observed of the seven ETRs matched previous observations analyzed from various RNAseq in tomatoes (Chen et al., 2018).

#### The NR Mutation seems to Affect the Correlation Between mRNA and Protein Levels

The NR mutation also causes several changes in the correlations between mRNA and protein abundance (**Figure 2B**), in particular, in the case of ETR1 where the Pearson coefficient changed from negative in WT (−0.17 and 0.59 for PEP1 and PEP2, respectively) to positive in NR (0.93 and 0.96 for PEP1 and PEP2, respectively) (**Table S3**). However, the weakness of power values would require a higher number of points to strengthen the correlation analysis; thus, **Table S3** is used to give trends at a glance. For the ETR4 and ETR7 receptors, mRNA levels decreased at BR8 in NR with either little to no change in protein levels (**Figures 2A**, **B**) suggesting that breakdown of these receptors is reduced in the NR mutant background. Further analysis will be required to determine the mechanism by which this occurs. For other ETRs such as ETR4 and ETR6, mRNA/protein correlation coefficients were very high and minimally affected by the NR mutation (**Table S3**); however, the power values were still too low for validating the trends.

#### CONCLUSIONS

We developed a PRM strategy that allowed the comparison of the abundance of ETRs in NR and WT tomato plants. Because ethylene has important roles in regulating plant development and responses to stresses, this method will be of wide use to study the roles of this phytohormone in diverse responses and plant species. However, calibration will be necessary for each peptide in each plant species. The observation that the GOF mutant ETR3 protein accumulates in orange mature fruit of the NR mutant is an example of regulation that would have remained unknown without the development of this new method. mRNA/protein correlations could also bring information about the regulation that occurs in ethylene signaling in fruit tissues, but more replicates are necessary. Given that ETRs in *Arabidopsis* show patterns of subfunctionalization (Shakeel et al., 2013), the use of PRM in tomato and other plant species will provide critical information about ETR subfunctionalization across the plant kingdom.

### MATERIALS AND METHODS

#### Plant Materials and Growth Conditions

Two tomato lines (*Solanum lycopersicum*) cv. Micro-tom were used, WT and NR mutant (Pro36Leu), both previously described (Carvalho et al., 2011). In addition to fruit color difference, these authors showed that NR seedlings are less sensitive to exogenous ethylene than WT seedlings, a classical response of ETR GOF mutants. Plants were grown in culture rooms with the following conditions: day/night (26°C for 18 h, 18°C for 8 h), light intensity 250 µmol.m−2.s−1, relative humidity at 80%. Four fruit stages were studied: IMmature Green (IMG), Mature Green (MG) fruit were harvested 20 and 38 days after flower anthesis, respectively; Breaker (Br) fruit was harvested once fruit color changed from green to yellow and red fruit (Br + 8) was harvested 8 days later (**Figure 4A**). Ethylene was analyzed using gas chromatography as previously described (Trapet et al., 2016) by incubating the fruit for 3 h. *Arabidopsis* lines (*Arabidopsis thaliana*) cv. Colombia were used, WT and *etr1-1 and etr2-1*, using growing conditions and growth monitoring described previously (Shakeel et al., 2013, and refs herein).

#### mRNA Purification and qPCR Analysis

For each fruit stage, the skin, together with pericarp tissues were collected and divided in three biological replicates of five fruits each, originated from different fruits, then ground to a fine powder in liquid nitrogen using a ball grinder. Total RNAs were purified from 100 mg of frozen sample with ReliaPrepTM RNA Tissue Miniprep System (Promega), according to the manufacturer's instructions. RNA was treated with DNase I (Invitrogen-AM1906), then 2 μg of RNA was treated with GoScript Reverse Transcriptase (Promega-A5003). Quantitative real-time PCR (qPCR) reactions were performed using 5 ng of cDNA per well as described before (Chervin and Deluc, 2010). EF1α, GAPDH, and actin were selected as house-keeping genes. All primers (**Table S1c**) were designed with primer-blast (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). SlETR1 expression in WT at IMG stage was used as control for all genes at all stages.

#### Microsomal Protein Extraction

The fruit samples used for mRNA extraction were also used for protein extraction, performed at 4°C according to previous studies (Bono et al., 1995; Kamiyoshihara et al., 2012) with some modifications. Briefly, 3 g of frozen ground powder was mixed with 25 ml of extraction buffer (50 mM Tris-HCl pH 7.0, 10 mM EDTA, 0.5 M sucrose, 3% PVPP w/v, 10 mM DTT, 100 μM PMSF, cOmplete™ Protease Inhibitor Cocktail (one tablet/100 ml), 1 mM phenantrolin, 1 mM Na-orthovanadate). The slurry was filtered through glass cotton at 300*g*, for 5 min and 900*g* for10 min). Then, left-over tissue bits were removed at 3,000*g* for 15 min. The resulting supernatant was centrifuged at 48,000*g* for 60 min. The pellet was resuspended in 25 mM TrisHCl buffer pH 7.0, 250 mM sucrose, 1 mM CaCl2, 1 mM MgCl2, and cOmpleteTM Protease Inhibitor Cocktail (one tablet/10 ml). Proteins were quantified with DCTM Protein Assay (Bio-Rad). Proteins (80 µg/lane) were fractionated using 10% precast SDS-PAGE gel electrophoresis (Biorad) after incubation at 37°C for 30 min in a loading buffer (50 mM Tris-HCl pH7.0, 10% glycerol, 4% SDS, 100 mM DTT, stained with bromophenol blue). The gels were then stained with Coomassie blue (R250, BioRad), then rinsed with acetic acid/methanol (Destain, BioRad). Each lane was cut in two bands, and bands containing proteins with a molecular weight above 50 kDa (with ETR dimers and monomers) were further analyzed by mass spectrometry.

#### Targeted LC-Parallel Reaction Monitoring Analyses

*Protein digestion:* Gel band treatments and trypsin digestion were performed as described in **Methods S1b**. Briefly, proteins in gel slices were reduced, alkylated, and digested overnight at 37°C with modified trypsin at a 1:100 enzyme/protein ratio (Promega, Madison, WI, USA). Peptides were extracted twice by the addition of 200 µL of 80% acetonitrile (ACN) and 2% formic acid (FA), and then dried in a vacuum centrifuge. Peptides were then resuspended in 20 µl FA 2%.

*ETR peptide selection:* To select ETR peptides to be studied in the PRM experiment, ETRs were digested *in silico* using MS digest (ProteinProspector tool, v. 5.19.1, University of California). Search criteria included digestion by trypsin, peptide mass from 500 to 4,000 Da, a minimum peptide length of six amino acids, and a uniqueness in the ITAG 3.2 database digested *in silico*. The peptides should also contain a minimal number of methionine residues because of their putative oxidation, of asparagine and glutamine residues because of their putative deamidation, of glutamic acid or glutamine as first amino acid because of the pyro-glutamination, of serine, threonine, or tyrosine residues because they can be phosphorylated. Then, the presence of proline was privileged because of its property to facilitate the MS/MS fragmentation. In addition, the proteotypic peptides previously identified in shotgun analyses (Mata et al., 2017; Szymanski et al., 2017) were preferentially selected. For the seven selected ETRs, 16 proteotypic peptides were selected. Labeled (or heavy) crude synthetic peptides were synthetized (PEPotec, ThermoFisher Scientific) with carbamidomethylation of cysteins and isotopic labeling of the last sequence amino acid (R: +10 Da (13C6, 15N4) or K: +8 Da (13C6, 15N2) (**Table S1a**).

*Parallel Reaction Monitoring (PRM):* Labeled peptides were mixed together in a hand-adjusted concentration-balanced mixture to equilibrate individual peptides signals and spiked in a biological matrix made of IMG WT sample in a similar quantity to the one used in all samples further analysed (**Figure 2** and **Figure S1**). The peptide mixture was analyzed using an UltiMateTM NCS-3500RS Ultra High Performance Liquid Chromatography system interfaced online with a nano easy ion source and a Q Exactive Plus Orbitrap mass spectrometer (ThermoFisher Scientific Inc, Waltham, MA, USA). Peptides were first loaded onto a pre-column (Thermo Scientific PepMap 100 C18, 5 μm particle size, 100 Å pore size, 300 μm i.d. x 5 mm length) from the Ultimate 3000 autosampler with 0.05% trifluoroacetic acid for 3 min at a flow rate of 10 μL/min. Then, the column valve was switched to allow elution of peptides from the pre-column onto the analytical column (Thermo Fisher Scientific Inc, Waltham, MA, USA, C18, 2 μm particle size, 100 Å pore size, 75 μm i.d. x 50 cm length). Loading buffer (solvent A) was 0.1% formic acid (FA) and elution buffer (solvent B) was 80% ACN + 0.1% FA. The three step gradients were 4–25% of solvent B for 103 min, then 25–40% of solvent B up to 123 min, and 40–90% of solvent B from 123 to 125 min, at a flow rate of 300 nL/min. The total chromatographic run time was 150 min including a high organic wash and re-equilibration steps. Peptides were transferred to the gaseous phase with positive ion electrospray ionization at 1.7 kV. Labeled peptides were checked by High-energy Collisional Dissociation MS/MS with regard to their retention time, charge, and m/z (**Table S1a**). A schedule PRM method was developed to simultaneously target all peptides (16 light peptides and 16 heavy peptides) in the protein sample (analytical details provided in **Methods S1a**). The Q-Exactive Plus Orbitrap instrument was operated as follows: a full MS scan spectra considering a mass range of 350–2,000 m/z was acquired with a resolution of 17.500 with an automatic gain control (AGC) fixed at 3e6 ions and a maximum injection time set at 100 ms. Targeted MS/MS spectra were acquired with a resolution of 140.000 with an AGC fixed at 2e5 and with the maximum injection time set at 1,000 ms. An MS/MS spectral library was acquired using a mixture of 16 heavy labeled synthetic peptides (**Methods S1a**). After manual checking of effective co-elution of endogenous and isotopically labeled peptides and after elimination of transitions showing interference, the Rdot-product (rdotp) values were calculated with Skyline (MacLean et al., 2010) (**Figure S1**), and peptides were relatively quantified with at least four transitions (**Figure 1**, **Table S1a**, and **Figure S1**).

Calibration curve was established using stable isotopelabeled peptides spiked into WT IMG samples prior to LC– MS/MS analysis using seven different peptide concentrations adapted for each peptide (**Figure S2**). Provided that the regression coefficient was above 0.90 and the rdopt was above 0.95, the peptide was qualified to be further quantified. For each peptide, the ratios of the endogenous to labeled peak areas were compared to obtain a relative quantification according to the genotypes and the development stages, as follows: relative level of endogenous peptide = sum of all transition intensities of the endogenous/sum of all transition intensities of the labeled.

The Pearson correlations have been calculated using R code, *via* the Wessa online tool (https://www.wessa.net/rwasp\_ correlation.wasp). The powers of the Pearson correlations were calculated using Sigmaplot (Systat Sotware, Inc.) at the 0.05 level.

#### Quantitative Real-Time PCR

For qPCR of *Arabidopsis* seedlings, 4-day-old dark-grown seedlings were grown in hydrocarbon-free air as described (Hall et al., 2012) and treated with 10 µL of L-1 ethylene for the indicated times at the end of their growth cycle. Total RNA was extracted from seedlings using the E.Z.N.A. Plant RNA Kit (Omega Bio-Tek), DNase treatment was performed using TURBO DNA free kit (Invitrogen), and cDNA was synthesized using the SuperScript III First Strand cDNA Synthesis Kit (Invitrogen, USA) according to the manufacturer's instructions. Real-time PCR was performed using iTaq Universal SYBR Green Supermix (Bio-Rad) and primer sets specific for ETR2 (5′-AGAGAAACTCGGGTGCGATGT-3′ and 5′-TCACTGTCGTCGCCACCATC-3′) and b-tubulin (At5g62700) control (5′-TGGTGGAGCCTTACAACGCTACTT-3′ and 5′-TT CACAGCAAGCTTACGGAGGTCA-3t).

#### Time Lapse Imaging

Ethylene growth response kinetics of etiolated *Arabidopsis* seedlings were determined according to methods previously described (Binder et al., 2004a and Binder et al., 2004b) on 2-dayold, dark-grown *Arabidopsis* seedlings grown on 0.8% (w/v) agar plates with half-strength Murashige and Skoog medium at pH 5.7 (Murashige and Skoog, 1962).

### DATA AVAILABILITY

The PRM data are deposited to PeptideAtlas, accessible *via*  ftp://PASS01274:DB4724xpa@ftp.peptideatlas.org/, Username: PASS01274, Password: DB4724xpa; and the label-free data are deposited to ProteomeXchange with the dataset identifier PXD011412, Username: reviewer72717@ebi.ac.uk, Password: g7EGcQI4.

#### AUTHOR CONTRIBUTIONS

CC and VS conceived the study. YC performed tomato culture, fruit sampling, protein extraction and purification, and mRNA extraction and analyses. VR performed protein digestion and MS/MS analyses. VR and SH designed specific peptides with help by JG, JN, and NB, who performed preliminary studies. VR, SH, and VD analyzed the MS/MS data. BA performed Arabidopsis gene expression analysis under the supervision of SS and GS. BB performed Arabidopsis kinetic analysis. BA, SS, BB, and GS analyzed and interpreted Arabidopsis data. YC, BB, VR, SH, MB, VS, and CC interpreted the data and wrote the manuscript. All authors read and approved the final manuscript.

### FUNDING

The authors acknowledge CSC for the PhD studentship to Y Chen, and INRA for a Starter Research grant to V Santoni and C Chervin. Part of this work was supported by the National Science Foundation Grants IOS-1456487 to GS and MCB-1517032 to GS and BB, and the International Research Support Initiative Program of Higher Education Commission of Pakistan to BA.

#### ACKNOWLEDGMENTS

Thanks to B Cooper (USDA, Beltsville, USA) for the initial advice about protein extraction and, later, for the critique of the results. Thanks to JJ Bono (CNRS, Toulouse, France) and HJ Klee (University of Florida, Gainesville, USA) for help to optimize the microsomal protein extraction, and to LE Pereira Peres (Universidade de São Paulo, Brazil) for the NR MicroTom cultivar. The authors acknowledge the Mass Spectrometry Proteomic Platform (MSPP)

#### REFERENCES


for PRM and quantitative label-free experiments. Finally, thanks to the two reviewers for their time and fruitful comments.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01054/ full#supplementary-material.


Zhao, X., Qu, X., Mattews, D. E., and Schaller, G. E. (2002). Effect of ethylene pathway mutations upon expression of the ethylene receptor ETR1 from Arabidopsis. *Plant Phys.* 130, 1983–1991. doi: 10.1104/pp.011635.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling editor is currently organizing a Research Topic with one of the authors MB, and confirms the absence of any other collaboration at the time of review

*Copyright © 2019 Chen, Rofidal, Hem, Gil, Nosarzewska, Berger, Demolombe, Bouzayen, Azhar, Shakeel, Schaller, Binder, Santoni and Chervin. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Biochemical Characterization of the *Fusarium graminearum* Candidate ACC-Deaminases and Virulence Testing of Knockout Mutant Strains

#### *Edited by:*

*Panagiotis Kalaitzis, Mediterranean Agronomic Institute of Chania, Greece*

#### *Reviewed by:*

*Luca Sella, University of Padova, Italy Guotian Li, Huazhong Agricultural University, China*

*\*Correspondence: Thomas Svoboda thomas.svoboda@boku.ac.at*

#### *Specialty section:*

*This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science*

*Received: 16 April 2019 Accepted: 07 August 2019 Published: 10 September 2019*

#### *Citation:*

*Svoboda T, Parich A, Güldener U, Schöfbeck D, Twaruschek K, Václavíková M, Hellinger R, Wiesenberger G, Schuhmacher R and Adam G (2019) Biochemical Characterization of the Fusarium graminearum Candidate ACC-Deaminases and Virulence Testing of Knockout Mutant Strains. Front. Plant Sci. 10:1072. doi: 10.3389/fpls.2019.01072*

*Thomas Svoboda1\*, Alexandra Parich2, Ulrich Güldener3, Denise Schöfbeck2, Krisztian Twaruschek1, Marta Václavíková2, Roland Hellinger2, Gerlinde Wiesenberger1, Rainer Schuhmacher2 and Gerhard Adam1*

*1 Department of Applied Genetics and Cell Biology, University of Natural Resources and Life Sciences, Vienna (BOKU), Tulln, Austria, 2 BOKU, Department for Agrobiotechnology (IFA-Tulln), Institute of Bioanalytics and Agro-Metabolomics, Tulln, Austria, 3 Department of Bioinformatics, TUM School of Life Sciences Weihenstephan, Technical University of Munich, Freising, Germany*

*Fusarium graminearum* is a plant pathogenic fungus which is able to infect wheat and other economically important cereal crop species. The role of ethylene in the interaction with host plants is unclear and controversial. We have analyzed the inventory of genes with a putative function in ethylene production or degradation of the ethylene precursor 1-aminocyclopropane carboxylic acid (ACC). *F. graminearum*, in contrast to other species, does not contain a candidate gene encoding ethylene-forming enzyme. Three genes with similarity to ACC synthases exist; heterologous expression of these did not reveal enzymatic activity. The *F. graminearum* genome contains in addition two ACC deaminase candidate genes. We have expressed both genes in *E. coli* and characterized the enzymatic properties of the affinity-purified products. One of the proteins had indeed ACC deaminase activity, with kinetic properties similar to ethylene-stress reducing enzymes of plant growth promoting bacteria. The other candidate was inactive with ACC but turned out to be a d-cysteine desulfhydrase. Since it had been reported that ethylene insensitivity in transgenic wheat increased *Fusarium* resistance and reduced the content of the mycotoxin deoxynivalenol (DON) in infected wheat, we generated single and double knockout mutants of both genes in the *F. graminearum* strain PH-1. No statistically significant effect of the gene disruptions on fungal spread or mycotoxin content was detected, indicating that the ability of the fungus to manipulate the production of the gaseous plant hormones ethylene and H2S is dispensable for full virulence.

Keywords: *Fusarium graminearum*, ACC (1-aminocyclopropane-1-carboxylic acid), ACC synthase, ACC deaminase, ketobutyrate

## INTRODUCTION

Ethylene is a gaseous plant hormone mediating developmental processes, such as fruit ripening, flower senescence, leaf abscission, as well as root elongation and has a strong influence on growth and yield of crop plants (Dubois et al., 2018). Furthermore, in a complex interplay with other plant hormones, ethylene has an important role in plant immunity (Broekaert et al., 2006; Ent and Pieterse, 2018; Li et al., 2019).

In plants, ethylene is derived from methionine *via* S-adenosyl-L-methionine. In a first dedicated and typically rate limiting step, 1-aminocyclopropane carboxylic acid (ACC) synthase converts S-adenosyl-L-methionine into the precursor ACC, from which ethylene is released by ACC oxidase. Ethylene perception and signalling were elucidated by genetic analysis of the model system *Arabidopsis thaliana*. The hormone is perceived by the copper containing receptor proteins ETR1 and related proteins (Gallie, 2015) that are present in the endoplasmic reticulum membrane. The Raf-kinase like protein CTR1 (Constitutive Triple Response 1) is associated with the ethylene receptor. In the absence of ethylene, CTR1 phosphorylates the ER-localized EIN2 protein at its cytosolic C-terminal part, which is then in the inactive state. When ethylene binds to the receptor, CTR is inhibited and EIN2 becomes dephosphorylated by phosphatases. The C-terminal domain of dephosphorylated EIN2 is then cleaved off and enters the nucleus, which ultimately leads to stabilization of the EIN3 transcription factor against proteasomal degradation (Qiao et al., 2012). EIN3 and the related EIL1 protein subsequently bind to promoters and activate the expression of multiple ethylene-responsive transcription factors (Huang et al., 2016), most importantly of the Ethylene Response Factor 1 (ERF1), which in turn induce the expression of a large number of ethylene responsive genes. With respect to plant defense, genes encoding pathogenesis-related proteins and biosynthetic genes for defense metabolites are induced by ethylene, often in a complex interplay with jasmonic acid. Ethylene and jasmonic acid signaling are considered to mediate defense against necrotrophic pathogens (Glazebrook, 2005) and hemibiotrophic pathogens switching to a necrotrophic mode after an initial biotrophic phase.

Studies with mutants that are deficient in ethylene biosynthesis or are ethylene-insensitive revealed the relevance of ethylene signaling in plant defense, but also the complexity of the interaction (Broekaert et al., 2006). For instance, Arabidopsis *ein2* mutants lost the ability to induce several pathogenesis-related (PR) genes in response to *Alternaria brassicicola* but nevertheless showed unchanged resistance, while the same mutants showed increased susceptibility to *Botrytis cinerea* (Thomma et al., 1999). In transgenic tobacco, which was engineered to become ethylene insensitive by overexpression of a dominant interfering ethylene receptor allele from Arabidopsis (*etr1-1*), breakdown of non-host resistance to several soil fungi was observed (Knoester et al., 1998), and also increased susceptibility to the pathogens *Botrytis cinerea* and *Fusarium oxysporum* (Geraats et al., 2003). Not all aspects of ethylene signaling are conserved between dicotyledonous and monocotyledonous plants (Yang et al., 2015). Nevertheless, ethylene clearly has a role, for instance in defense of rice against the fungal pathogen *Magnaporthe grisea*. Yet, in general, the outcome of virulence tests may be strongly dependent on the combination of host plant and pathogen species (Ent and Pieterse, 2018).

In contrast to the observed trend indicating a role for ethylene in mediating resistance, there are reports that ethylene signaling may also be utilized by fungal pathogens to increase host susceptibility. In the rice interaction with *Cochliobolus miyabeanus* (causing brown spot disease) treatment of plants with the ethylene releasing compound ethephon enhanced susceptibility, while plants with a silenced *OsEIN2* homolog (showing decreased expression of ethylene responsive defense genes) were more resistant (De Vleesschauwer et al., 2010). Similarly, in the interaction between *Fusarium oxysporum* f. sp. *raphani* and *Arabidopsis thaliana*, the ethylene insensitive *etr1-1*  mutant was more resistant (Pantelides et al., 2013).

In line with this, the agriculturally relevant pathogen and mycotoxin producer *Fusarium graminearum*, which is in the focus of our interest, was also reported to be less virulent on transgenic wheat with silenced *EIN2* homologs, and importantly, much lower amounts of the mycotoxin DON accumulated in the infected plants. Thus, the fungus may exploit ethylene signaling to promote virulence (Chen et al., 2009). Yet, this is highly controversial. Most gene expression results comparing wheat cultivars of either high or low *Fusarium* resistance found correlations with gene expression consistent with a stronger resistance-associated ethylene-response. A stronger and more rapid response of genes involved in ethylene biosynthesis, ethylene signaling and ethylene response (Ding et al., 2011; Gottwald et al., 2012) and also higher amounts of ACC early in the infection were found in highly resistant cultivars (e.g. Wang et al., 2018). Furthermore, chemical manipulation of ethylene signaling revealed that treatment of highly resistant wheat cultivars with ethylene biosynthesis inhibitors (1-methylcyclopropene and cyclopropane-1,1-dicarboxylic acid) negatively affected resistance to initial infection and disease spread, while treatment with compounds enhancing ethylene production (ethephon and ACC) increased resistance of wheat to *F. graminearum* in highly susceptible cultivars (Foroud et al., 2018).

Several microbes are capable to synthesize ethylene and can consequently trigger ethylene dependent gene expression. In contrast, some microbes can downregulate ethylene production (reviewed in Ravanbakhsh et al., 2018). Fungi can form ethylene at unphysiologically high methionine concentrations in the medium by first transaminating methionine to 4-(methylsulfanyl)-2 oxobutanoate, better known by the synonym 2-keto-methyl-thiobutyrate (KMBA). In a second non-enzymatic reaction, KMBA is then cleaved by hydroxyl radicals (e.g. generated from H2O2 and Fe2+) and ethylene is released [see *MetaCyc* Pathway: ethylene biosynthesis III (microbes)]. This unspecific pathway is seemingly the predominant one in the fungus *Botrytis cinerea* (Chagué et al.,

**Abbreviations:** DON, deoxynivalenol; D3G, deoxynivalenol-3-O-glucoside; ACC, 1-aminocyclopropane carboxylic acid; KBA, ketobutyric acid; ACS, ACC synthase; ACD, ACC deaminase; ACO, ACC oxidase; EFE, ethylene forming enzyme; KMBA, 4-methylthio-2-oxobutanoic acid; CTR, constitutive triple response; EIN, ethylene insensitive; EIL, ethylene insensitive like; PR, pathogen related; FMM, fusarium minimal media.

2002). In contrast, many bacteria, particularly plant pathogenic *Pseudomonas* species, can synthesize ethylene from oxoglutarate and arginine *via* the Ethylene Forming Enzyme (EFE). This enzyme is also present in some fungi [*MetaCyc* Pathway: ethylene biosynthesis II (microbes)], for instance in *Penicillium digitatum* (Johansson et al., 2014) which causes fruit rot in citrus. In addition, some fungi, e.g., *Penicillium citrinum* (Kakuta et al., 2001) contain ACC synthase and can synthesize ethylene using the plant pathway [*MetaCyc* Pathway: ethylene biosynthesis I (plants)]. Microbes can also have the ability to downregulate ethylene signalling by interfering with ethylene production. For instance, a bacterial type III effector protein (HopAF1 from *Pseudomonas syringae*) reduces ethylene production by targeting methionine recycling in the Yang cycle (Washington et al., 2016). Also, small molecules can be used as effectors, such as the bacterial rhizobitoxin, which directly inhibits ACC synthase (Yasuta et al., 1999; Sugawara et al., 2006). The ethylene precursor ACC, a signal molecule in plants, is transported *via* the xylem and phloem and causes ethylene production in tissue distant to its site of production (Van de Poel and Van Der Straeten, 2014). The most widely employed mechanism of plant associated bacteria (Gamalero and Glick, 2015; Nascimento et al., 2018) to downregulate ethylene production is the production of ACC deaminase enzymes, which degrade ACC into α-ketobutyric acid (KBA) and ammonia. Reduction of "stress-ethylene" *via*  ACC deaminase has been found to be the mode of action of plant growth promoting rhizobacteria that improve the ability of plants to cope with different abiotic stresses, like salt stress (Singh et al., 2015; Qin et al., 2016) or heavy metal stress (Grichko et al., 2000). Expression of ACC deaminase in transgenic tomato was sufficient to enhance growth despite increased heavy metal accumulation (Grichko et al., 2000). ACC deaminases were found and characterized also in several plant associated fungi, for instance in the pathogen *Penicillium citrinum* (Jia et al., 2000), and the biocontrol strain *Trichoderma asperellum* (Viterbo et al., 2010). Such genes have been detected in many fungal genomes, and phylogenetic analysis has provided evidence for multiple independent acquisitions of bacterial ACC genes by horizontal gene transfer (Bruto et al., 2014).

Our research group is interested in virulence mechanisms of *Fusarium graminearum*. Due to the reported ability of this fungus to exploit ethylene signaling for virulence (Chen et al., 2009) and the controversial proposed roles of ethylene in Fusarium head blight resistance, we started to search for genes in *Fusarium graminearum* potentially involved in either ethylene production or downregulation of ethylene signaling, and to experimentally test candidate genes by heterologous expression and gene disruption.

### MATERIALS AND METHODS

#### Bioinformatical Analysis

For the bioinformatic analysis of the genome annotation of *F. graminearum* (PH-1) the PEDANT genome database was used (Walter et al., 2009). The FGSG gene models are still available at http://gbrowse.boku.ac.at/cgi-bin/gb2/gbrowse/

Fusarium\_graminearum\_PH1/and in FungiDB (https://fungidb. org/fungidb/). BLAST searches were performed at NCBI (https:// blast.ncbi.nlm.nih.gov/Blast.cgi).

#### Cloning of *Fusarium graminearum* ACS and ACD Candidate Genes in *E. coli* DH10B

Both *ACD* candidate genes were amplified from gDNA of *F. graminearum* PH-1 using fusion PCR. All PCRs were done as follows: 2 min/95°C initial denaturation followed by 25 cycles of 30 s/95°C, 30 s at primer-dependent calculated annealing temperatures, 1 min/kb at 72°C and a single 72°C/5 min step at the end. All primers and fragment lengths are listed in **Supplementary Table 1**. The two *ACD* candidate genes were cloned in pETDuet-1. The coding region of *ACD1* (FGSG\_02678) was cloned *via*  BamHI/NotI, and *ACD2* (FGSG\_12669) as EcoRI/NotI fragment. Plasmid pET30-UW4-651, containing an ACC deaminase from *Pseudomonas putida* was kindly provided by Prof. Glick, Waterloo, Ontario (Hontzeas et al., 2004). The BglII/HindIII fragment from this plasmid containing the ACD coding sequence was cloned into pETDuet-1 and into pACYCDuet-1 (BamHI/HindIII), a plasmid containing P15A ori (compatible with pETDuet-1) and a chloramphenicol resistance marker.

*ACS1* (FGSG\_05184) was amplified from genomic DNA due to the lack of introns, while *ACS2* (FGSG\_07606) was amplified from cDNA which was prepared from mycelium harvested from PDA medium according to the protocol of the "RevertAid H Minus First Strand cDNA Synthesis Kit" (Thermo Scientific, Vienna, Austria). *ACS3* (FGSG\_13587) was cloned from genomic DNA by fusion PCR. First the exons were amplified with the respective primers shown in **Supplementary Table 1**. Due to the overhangs which are homologous with the primer for amplification of the adjacent exon, amplification of two fused exons was achieved by fusion PCR. *ACS1* was cloned in pET-Duet1 into MCS1 *via* EcoRI/NotI while ACS2 and ACS3 were cloned using BamHI/NotI. The plasmids are listed in **Supplementary Table 4**.

#### Expression of *ACS* Candidate Genes in *E. coli*

After induction of transformed *E. coli* strain BL21/DE3 with 1 mM IPTG cells were harvested by centrifugation and the pellets were resuspended in 0.1 M Na-phosphate buffer pH 7.6. For cell disruption a Branson Sonifier W-250 D (Branson Ultrasonics Corporation, Danbury, CT, USA) was used with the following settings: 12 x 5 s pulse and 1 min pause. After centrifugation at 18,000 g the proteins were purified from the supernatants by affinity chromatography using a HisTrap© FF crude 1 ml column using an ÄKTA purifier (GE Healthcare, Austria). For rapid preparations His-select® spin columns (H7787, Sigma-Aldrich, Vienna, Austria) were used.

#### Tests of ACC Synthase Activity

The *ACS* candidate genes were expressed in *E. coli* and ACC synthase activity was tested in living cells by measuring the conversion of methionine into ACC. For this IPTGinduced cultures were supplemented with methionine to a final concentration of 100 mg/l. Samples were drawn after 0, 0.5, 1, 3 and 24 h. The cells were harvested by centrifugation, resuspended in 0.1 M HEPES-KOH buffer pH 8.5 and disrupted by sonication as described above. The concentrations of methionine and ACC in the culture supernatant and in the soluble cell extract were measured by GC-MS.

For *in vitro* tests the proteins were purified using a rapid procedure (His-Select spin columns) according to the manufacturer's instructions. The activity of the ACC synthases was tested in presence of 50 mM HEPES-KOH buffer (pH 8.5) with 200 µM S-adenosyl-L-methionine (A7007, Sigma-Aldrich, Vienna, Austria) and 10 µM pyridoxal phosphate (P9255, Sigma-Aldrich, Vienna, Austria). Samples were taken every 30 min for 3 h and then after 16, 20, and 24 h and measured by GC-MS.

#### GC-MS Measurements of ACC and KBA

The standards for 1-aminocyclopropane-carboxylic acid (ACC; EMD 149101-1G) and 2-ketobutyric acid (KBA; K401, Sigma-Aldrich, Vienna, Austria) were purchased from Sigma-Aldrich. The solvents methanol LC-MS grade (Honeywell 34966) and pyridine p.A. (Merck 1.09728.0500) were obtained from Merck. The derivatisation chemicals methoxyamine hydrochloride (MOX; 226904, Sigma-Aldrich, Vienna, Austria) and N-methyl-Ntrimethylsilyl trifluoroacetamide (MSTFA; 701270.201, Macherey-Nagel) were purchased from Sigma Aldrich and Macherey-Nagel, respectively. The stock solutions of ACC and KBA were prepared in 50% aqueous methanol.

The unpurified protein extracts were centrifuged for 10 min at 14,000 rpm at 4°C, 10 μL of the supernatant were transferred into micro-inserts in GC/HPLC vials and dried overnight using a centrifugal evaporator (Labconco, Kansas City, MO) at 15°C. Subsequently an automated two step derivatisation was carried out using the GC auto sampler (PAL LHX-xt, CTC Analytics, Carrboro, NC). To this end the residue was resuspended in 50 µL MOX (20 mg/ml pyridine) and agitated for 90 min at 90°C. After addition of 50 µL MSTFA the mixture was again agitated at 90°C for 60 min.

For GC-MS analysis, 1 µL of the derivate was injected into the split-/splitless injector of an Agilent 7890A gas chromatograph coupled to a 5975C inert XL MSD detector (Agilent, Waldbronn, Germany), equipped with ChemStation software (version E.02.01.1177) and operated in pulsed splitless mode at 250°C. Chromatographic separation was carried out on an HP5-ms column (30 m x 0.25 mm x 0.25 µm; Agilent Technologies, Waldbronn, Germany) at a constant flow of 1 ml/min helium. Temperature program: 50°C, 2 min hold, to 120°C (10°C/min, 5 min hold), to 150°C (5°C/min), to 325°C (70°C/min, 10 min hold). The MSD was operated with interface set to 335°C, the EI source to 230°C and the MSD quadrupole to 150°C, solvent delay 6 min, scan range of *m/z* 50–300, dwell time 100 ms. SIM mode was used for quantification of the ACC 1TMS derivate *m/z* 83 (quant ion), 130, 173; ACC 2TMS derivate *m/z* 147 (quant ion), 202, 230 and KBA 1TMS derivate *m/z* 89, 172, 188 (quant ion).

#### Ethylene Measurements

Ethylene (Ethylen Ecocyl 2.5) was purchased from Linde. Ethylene concentrations were determined by GC-FID (Hewlett Packard, Series 2, 5890 Series 2 plus, Agilent, Santa Clara, USA) coupled to a head space autosampler (Hewlett Packard, HP 7694, Agilent, Santa Clara, USA) using the following parameters. Headspace sampler: Oven, sample loop and transferline temperature 45°C, loop 3 ml, loop fill time 0.1 min and injection time 0.2 min; vial parameters: equilibration time 0.5 min, pressurizing time 0.2 min, GC cycle time 11 min. Chromatographic separation was carried out on a Restek Rt-QS-Bond column (30 m × 0.53 mm × 20 µm), using Helium as carrier gas (3.3 ml/min constant flow). Temperature program: 30°C, 1 min hold, to 60°C with 30°C/min, 3.5 min hold. Split injection (30:1) was used and the injector and the flame ionization detector (FID) were set to 250°C. Quantification was based on peak heights and external calibration (10 concentration levels between 0.17 µg/L and 10.65 µg/L). Calibration results were used to estimate the limit of detection (LOD = 0.67 µg/L) and limit of quantification (LOQ = 2.3 µg/L) with the ValiData software Version 3.02.48. Data were processed using MassHunter (Agilent Technologies, Qualitative Analysis B.06.00). Peak areas of the quantification ions were used for comparative quantification and the compound identity was confirmed by comparison with MS spectra of reference standards.

#### *In Vitro* Test of ACC Deaminase Candidate Genes

To follow the fate of ACC and KBA in liquid cultures of the *E. coli* BL21/DE3 strain transformed with empty vector or ACD expression vectors, a feeding experiment was performed. The empty vector control allows tracing of possible metabolization products of the target substances by *E. coli*. Expression of the target genes was induced by adding 1 mM IPTG and incubation overnight at 20°C, 140 rpm. ACC and KBA were added (0.5 mg/ ml final concentration) to the IPTG-induced cultures. Samples were taken after 0, 1, 3 and 24 h and analyzed for ACC and KBA by GC-MS.

For characterization of the kinetic properties of the ACC deaminase from *Fusarium graminearum* a colorimetric assay relying on the reaction of phenylhydrazine with a ketone to phenylhydrazone was used (Penrose and Glick, 2003). The assay contained purified ACD2 (0.5 mg/ml), ACC concentrations ranging from 0.1 to 100 mM as well as 100 mM Na-phosphate buffer pH 7.6. A calibration curve was generated using KBA.

#### d-Cysteine Desulfhydrase Assay of ACD1 With d-Cysteine and l-Cysteine

To test ACD1 (later renamed DCS1 since the protein has d-cysteine desulfhydrase instead of ACC deaminase activity) for activity, 5 µl of the raw protein extract, which was obtained by sonification of IPTG induced cells, were tested in a total volume of 60 µl containing 0.1 M Tris–HCl pH 7.6 and 8 mM d-cysteine and incubation at 37°C for 15 min. The colorimetric assay was performed according to Penrose and Glick (2003), using pyruvic acid for calibration. ACD1 was purified by liquid chromatography *via* the His-tag using HisTrap™ columns (GE29-0510-21, Sigma Aldrich, Vienna, Austria). Affinity purified ACD1 was tested for substrate specificity in 50 mM phosphate buffer pH 7.6 using 8 mM of the respective substrate, d-cysteine, l-cysteine, 2-aminoethyl-l-cysteine, or water as negative control. For determination of kinetic properties 0.1–30 mM D-cysteine as substrate were used. All reactions were incubated for 5 min at 30°C and stopped by the addition of 900 µl 0.56 M HCl.

#### *ilvA* Knock-Out in *E. coli* T7-Express

In order to be compatible with the T7 polymerase based expression system we knocked out *ilvA* in the expression host strain. The resulting strain is auxotrophic for isoleucine due to the loss of the ability to synthesize KBA, an intermediate of isoleucine biosynthesis. We obtained the *ilvA* mutant *E. coli* strain JW3745-2 from "The Coli Genetic Stock Center." The kanamycin resistance gene replacing the *ilvA* gene was amplified with primers 5´CGGAGATGTGGTAGTAATTC-3´ and 5´-GCCGTTTATTATGGCCGATC-3´, which bind in the neighbouring genes *ilvD* and *ilvY*. The 2085 bp fragment was purified and adjusted to a final concentration of 100 ng/µl. The strain to be transformed, *E. coli* T7-express (NEB, Frankfurt am Main, Germany), was first endowed with plasmid pKD46, which contains an arabinose inducible phage lambda *red* recombinase (Datsenko and Wanner, 2000). The PCR product (100 ng) was used for electroporation followed by selection on LB+kanamycin plates at 37°C, leading also to the loss of the temperature sensitive pKD46. Kanamycin resistant mutants unable to grow on M9 plates were tested by PCR. The resulting strain—T7 express *ilvA*::*KanR*—is available upon request. This strain was further transformed with the empty pET-DUET-1 vector or with pET-DUET-1 containing either ACD1, ACD2, or the positive control of *Pseudomonas putida* in MCSI. Expression was induced by IPTG (1 mM final concentration).

The positive control from *P. putida* (*ACDP*) was released from pET30a by BglII/HindIII cleavage and cloned into BamHI/ HindIII digested pACYCDUET-1. This vector was transformed together with one of the *ACS* genes in pETDUET-1 into the generated *E. coli* T7-express lacking *ilvA*. *ACS4* of *Arabidopsis thaliana* (AT2G22810.1) in pETDUET-1 served as a positive control. Transformants were selected on LB plates containing ampicillin (100 mg/l) and chloramphenicol (25 mg/l). For the spottings, M9 was supplemented either with 3 mM methionine or 3 mM ACC.

#### Preparation of Knock-Out Mutants in *F. graminearum*

PEG-mediated protoplast transformation was performed as described by Twaruschek et al. (2018). For gene disruptions the split marker strategy was applied and fragments for transformation were produced as described below. The 3´ and 5´ flanking region of the candidate genes were cloned in pASB42 (see Twaruschek et al., 2018) adjacent to *hph*-*amdS* cassette flanked by two loxP sites (Steiger et al., 2011) using the primers listed in **Supplementary Table 2**. Screening of transformants was performed as described in (Twaruschek et al., 2018) using the primers listed in **Supplementary Table 3**. For multiplex PCR, three primers which are either located outside of the flanking region which was used for homologous recombination, or in the resistance cassette, or in the target gene to be disrupted were used. The corresponding primers in the resistance cassette were 5´AGAAGTACTCGCCGATAGTG-3´ for the 5´-flanking region, and 5´-ACACCTGCCGTGTCAGCC-3´ for the 3´-flanking region. The primers were designed in a way that the band of the wild type and the one of the knock-out strain can be easily differentiated.

For generating *acd1 acd2* double mutants, a knockout construct with HSVtk-nptII as selection marker was generated: *ACS* flanking region containing plasmids pTS24, pTS35 and pTS45 were cut with HindIII/SalI. pTS14 and pTS58 were cut with BcuI/SfiI and pTS61 was digested BglII/BcuI (**Supplementary Table 4**). The HSVtk-nptII cassette from pKT245 was digested the same way to replace amdS-hyg (Twaruschek et al., 2018). Two independent, genotypically identical mutants (strains TS\_ACD1Δ\_2 and TS\_ACD1Δ\_13, listed in **Supplementary Table 5**) lacking *ACD1* were transformed using the split marker strategy. For screening of the double knock-out candidates, the same outer primer and the one within the gene as for the single knock-out screening were used. Due to a different resistance cassette the primers in the cassette were adapted to 5´-GTAGACCGCAAATGAGCAAC-3´ for the screening of the 5´ region, and 5´-GCCACAGCAGCCACGACA-3´ for the 3´ region. The resulting strains have the genotype *dcs1*∆::*loxP-hyg-amdS-loxP acd2*∆::*loxP-nptII-HSVtk-loxP* (**Supplementary Table 6**). Six PCR confirmed doubleknockout strains were chosen for the virulence test with 10 replicates per strain. The progress of infection was observed over a time period of 16 days followed by analysis of DON and D3G after harvesting.

#### Test of Utilization of ACC as Sole Nitrogen Source

Fusarium Minimal Medium (FMM, for recipe see Twaruschek et al., 2018) was modified by replacing NaNO3 with 3 mM ACC. Suspensions containing 105 spores were spotted onto the plates, which were incubated for 10 days at 20°C in the dark.

### Wheat Infections

For the infection assays the cultivar Apogee (Mackintosh et al., 2006) was used. Ten microliters of a spore suspension with 4\*104 spores/ml was pipetted into each of four florets of two spikelets in the middle of the ear. Moistened plastic bags were used to cover the infected ears for 24 h to maintain high humidity. Plants were kept in a growth chamber (20°C with 16 h/8 h light/dark periods) and the disease progression was observed over a time period of 16 days by counting the infected spikelets every other day. For toxin analysis the ears at the endpoint of 16 days were harvested, frozen in liquid nitrogen and ground in a Retsch mill (Retsch MM400) using steel balls. For extraction 400 µl solvent (acetonitrile/water/acetic acid = 79: 20: 1) per 100 mg sample was added. The samples were extracted for 1 h at 20°C, 180 rpm, and centrifuged for 10 min at 14,000 g, and a 1:10 dilution was prepared for analysis by HPLC-MS.

#### Quantitative Analysis of DON and D3G

Deoxynivalenol (DON) and DON-3-O-glucoside (D3G) were determined by HPLC-MS/MS. A 1290 UHPLC system from Agilent Technologies (Waldbronn, Germany) equipped with Gemini C18 column (150 × 4.6 mm, 5 µm; Phenomenex, Aschaffenburg, Germany) was used for separation of analytes. The mobile phase eluents were composed of water and methanol (A: 80:20, v/v; B: 3:97, v/v) and contained both 5 mM ammonium acetate. The applied gradient was as follows: 0–1 min (0% B); 1–6 min (linear gradient from 0% B to 50% B); 6.1–8 min (flushing of the column with 100% B), 8–10 min (column equilibration with 0% B). The chromatographic system was maintained at 25°C and the flow rate of mobile phase was set to 800 µl/min.

Detection and quantification of both target analytes were performed on mass spectrometer QTrap 4000 (Sciex, Foster City, CA, USA), equipped with a TurboV electrospray ionization source. The system was operating in negative electrospray ionization mode (ESI-). The following source parameters were applied: curtain gas 35 psi (240 kPa), ion spray voltage (4 kV), temperature 550°C, ion source gas 1 and 2 both 50 psi (344 kPa), collision gas (nitrogen) high, and the interface heater on. The following selected reaction monitoring (SRM) transitions with a dwell time of 25 ms were used: DON (retention time 5.75 min) m/z 355.1 (declustering potential, DP, 65 V), product ions m/z 265.2 (collision energy, CE, 28 V) and m/z 59.2 (CE, 21 V); D3G (retention time 5.45 min) m/z 517.3 (DP, 56 V), product ions m/z 427.1 (CE, 25 V) and m/z 59.1 (CE, 18 V). Data were processed with Analyst 1.6.3 software from Sciex and further data processing and calculation of concentrations were performed using Microsoft Excel 2010. The limits of detection and quantification for DON and D3G were 10 and 25 ng/ml, respectively.

#### Cysteine Racemase Test

To check whether *F. graminearum* is able to convert L-cysteine into D-cysteine or vice versa, protein extracts were tested for racemase activity. Pre-cultures were prepared by inoculation of 105 spores of strain PH-1 in 100 ml FMM medium. After 3 days of incubation the mycelia were harvested using a sterile filter funnel, washed and inoculated in modified FMM (MgSO4 replaced by MgCl2 and supplemented with 1 mM Dor L-cysteine). The cultures were incubated at 20°C overnight followed by filtration of the mycelium. The mycelia were blotted dry, frozen in liquid nitrogen, pulverized using Retsch mill and 2 µl 50 mM Na-phosphate buffer pH 6.6 was added per milligram mycelium followed by vortexing 3 × 30 s. The protein extracts were centrifuged and the supernatant was used.

The assay contained 20 µl of the protein extract, 75 µl 50 mM phosphate buffer pH 6.6, and 25 µl d- or L-cysteine (10 mM final). The negative controls were supplemented with buffer instead of cysteine. Samples were taken after 0 and 24 h by transferring 50 µl of the samples into a tube containing 150 µl MeOH. One volume of ice-cold chloroform and four volumes of ice-cold methanol were added. The mixtures were shaken at 4°C for 15 min and precipitated proteins were pelleted by centrifugation at 12,000 g at 4°C for 10 min. The clear aqueous phase was transferred into another 1.5 ml reaction vial. The process was repeated once and the collected methanol phase was evaporated in a centrifugal vacuum evaporator. The dried residue was kept at −20°C until derivatization. Marfey's reagent (Pierce TS-48895) was used for derivatization of enantiomers to receive diastereomers. This derivatization allows chromatographic separation and detection of isobaric compounds derived from d-/ l-cysteine.

#### Quantification of d-, l-Cysteine Racemization Using LC-HR-MS Measurements

For quantification of D-cysteine desulfhydrase activity and to measure racemase activity an ultra-high performance liquid chromatography (UHPLC) Vanquish system coupled to a QExactive HF Orbitrap (both ThermoFisher) was applied to measure D-/L-cysteine levels. Five microliters of sample extract was injected for chromatographic separation on a Gemini NX-C18 column (150 × 3 mm ID, 5 µm, 110 Å, Phenomenex), which was protected by a pre-column and maintained at 25°C at a constant flow rate of 400 µl/min. To achieve chromatographic separation of the cysteine derivatives a gradient program with linear gradient segments was applied. Buffer A was water containing 0.1% formic acid and buffer B was acetonitrile containing 0.1% formic acid. The gradient program was as follows (% acetonitrile/ time in minutes): 10/0, 10/3, 45/56.1, 99/57, 99/62, 10/63, 10/70. UV traces were recorded at 220 and 340 nm.

The MS system was equipped with a heated electrospray ionization (HESI) source. The ESI interface was operated in fast polarity-switching mode, using 55 L/min sheath gas flow, 5 L/ min auxiliary gas flow, a spray voltage of 3.5/3.0 kV respectively for the positive and negative ionization mode, S-lens level 55, capillary temperature 320°C and auxiliary gas temperature 350°C. FT-Orbitrap was operated in full scan mode acquiring profile spectra for the scan range m/z 100–1000 with a resolving power of 120,000 FWHM (at m/z 200) and automatic gain control setting of 3x106 with a maximum injection time of 200 ms. The mass analyzer's mass accuracy was calibrated on a daily basis using Pierce LTQ Velos ESI ion calibration solution (ThermoFisher). A 1:1 mixture of the solution for positive and for negative ionization was prepared for calibration. The system was operated with TUNE 2.8 SP1 software and the direct control plugin for Chromeleon 7.2 SR4. Data processing was performed with XCalibur 4.0.27.19 (ThermoFisher). Quantification of area under the curve for the detected analytes was obtained with the XCalibur plug-in QuanBrowser. The retention times and m/z species used for quantification of the analytes were as follows: Marfey's reagent 23.80 min, l-cysteine 46.75 min, d-cysteine 49.90 min, D-valine 38.70 min, d8-D-valine 38.60 min, LL-cystine 47.20 min, DD-cystine 53.90 min, and DL-cystine 44.7 min. Note that Marfey's reagent reacts with both the primary amines and the free sulfhydryl groups in cysteine. The retention time of the diastereomeric amino acid derivatives was determined using pure enantiomeric reference standards (Sigma-Aldrich, Vienna, Austria). Reference materials were prepared in 10 mM HCl and working solution for calibration were diluted in water. An internal standard calibration was applied to account for matrix effects as well as for variation of chemical derivatization with the Marfey's reagent. Due to the lack of sufficient isotopically labelled D- and L-cysteine and DD-, DL-, and LL-cystine reference standards, the representative amino acid D-valine was chosen as internal standard. Accordingly, quotient of analyte/ISTD area under the curve was calculated for all analytes. The ISTD concentration of 6.94 µM after spiking the analytes into the sample and derivatization was kept constant for all calibration standards and the biological samples. The applied concentrations for cysteine to calculate calibration function and for analyte quantification were 16, 80, 400, 2,000, and 10,000 µM. The obtained calibration function was linear for the chosen concentration range. The limit of detection was estimated in a serial dilution experiment for the analytes with greater than 2.89 µM.

### RESULTS

#### Bioinformatical Analysis of the *F. graminearum* Genome—Presence of EFE?

The first question we addressed was whether the genome sequence of *F. graminearum* contains candidate genes that might allow the fungus to synthesize ethylene and thereby exploit ethylene signaling to increase virulence. The EFE pathway is the most efficient mechanism to produce ethylene. Therefore, we started the search for candidate genes using an EFE with biochemically confirmed activity (GenBank: EKV19239.1) as a query in a BLASTP search. As shown by Johansson et al. (2014), the associated mitochondrial import sequence was missed in the genome annotation of *Penicillium digitatum* (Marcet-Houben et al., 2012). The protein sequence in the database starts at the second in-frame ATG and is missing a number of N-terminal amino acids present in the protein sequence of the purified enzymes (see **Supplementary Figure 1**). We therefore used the modified gene model proposed by Johansson et al. (2014), which is highly similar to gene model (XP\_002562422.1) from *Penicillium chrysogenum*/*P. rubens* (van den Berg et al., 2008). The BLASTP search using the N-terminally extended sequence (EKV19239.1) as query against the *F. graminearum* PH-1 genome yielded five hypothetical proteins (FGSG\_00893, FGSG\_00048, FGSG\_03213, FGSG\_08081, FGSG\_02301) with sequence identities lower than 24% (low E-values below 9e-15, query covers below 75%). These low-similarity genes most likely encode members of the 2OG-Fe(II) oxygenase superfamily (pfam03171) with functions other than ethylene formation. The whole *F. graminearum* species complex (taxid:569360) contained no good candidate *EFE* gene, with the notable exception of *F. venenatum* (XP\_025583657.1), having 66.39% identity with a query cover of 99%. The sequences differ to the highest extent in the putative mitochondrial import sequence, which is shorter in *Fusarium*. Interestingly, many other *Fusarium* species, particularly from the *F. oxysporum* and *F. fujikuroi* species complex had also highly similar genes, indicating that *EFE* could be present and conserved in the genus *Fusarium* (taxid:5506) after all (see **Supplementary Figure 1**). Yet, also in other *Penicillium* species, non-functional homologs highly similar to the enzymatically active *P. digitatum EFE* exist. Site-specific mutagenesis had revealed that introduction of several amino acids, which are different in the inactive enzyme from *P. chrysogenum,* into the active *Pseudomonas syringae* pv. *phaseolicola* enzyme (P32021.1, Fukuda et al., 1992), leads to loss of activity of the latter (Johansson et al., 2014). The candidate gene from *F. oxysporum* f. sp. *cepae* (RKK90967.1, as published by Armitage et al., 2018), has the highest similarity with the N-terminally extended *P. digitatum* gene (65% sequence identity over 88% of the query sequence). Using this protein sequence as query, the BLASTP results indicate that many gene models of other *Fusarium oxysporum* isolates are probably missing the C-terminal exon and therefore run into a premature stop. Closer analysis of one of these putatively truncated EFE candidates (Accession PCD42226.1) has shown that this is due to a GT → GA mutation in a donor splice site, which is then no longer recognized by automated annotation algorithms. However, studies of *Fusarium*  splicing by (Zhao et al., 2013) have shown that the organism does actually use GA as a donor site, therefore the algorithm used to detect these "premature stops" might have overlooked an intron, and the truncated gene models might, in fact, not be truncated at all. **Supplementary Figure 1** shows the respective alignment, indicating that the hypothetical protein AU210\_004756 from *Fusarium oxysporum* f. sp. *radicis-cucumerinum* can easily be modified to resemble the non-truncated candidates of other species. A potential exception is the unordered region (blue letters in **Supplementary Figure 1**), which was not included in the structural model of Johansson and coworkers (2014). In summary, we conclude that the *EFE* homologs seem to be widespread in different *Fusarium* species, but functional testing is necessary to find out whether the candidates are indeed encoding active enzymes or not. In contrast, *F. graminearum* and close relatives (*F. culmorum*, *F. pseudograminearum*) do not possess *EFE*.

#### Identification of ACC Synthase, Oxidase, and Deaminase Candidate Genes in the *F. graminearum* Genome

Using headspace GC-MS we found increased ethylene production of *F. graminearum* PH-1 on medium with 20 mM methionine added (**Supplementary Figure 2**), which suggests that the ACC synthase pathway might be active. Yet, we cannot rule out that the observed ethylene production is due to the unspecific pathway III found in a broad range of microbes.

We investigated whether *F. graminearum* possesses ACC synthase and ACC oxidase candidate genes. A BLASTP search with the sequence of the enzymatically active enzyme from *Penicillium citrinum* (BAA92149.1, described by Kakuta et al. in 2001) revealed three candidate genes: *FGSG\_05184*/*FGRAMPH1\_01G17303*, (named *ACS1* in this study), *FGSG\_07606/FGRAMPH1\_01T25199* (*ACS2*), and *FGSG\_13587* (*ACS3*). These genes were already previously annotated as ACC-synthases in the Fusarium Genome Database (FGDB) by Wong et al. (2011). The alignment score between the ACC synthase of *P. citrinum* and *Fusarium* genes is 52.3 for ACS1, 37.1 for ACS2 and 29.8 for ACS3. An alignment of these three predicted proteins with the ACC synthase of *P. citrinum* and an ACC synthase from *Triticum aestivum* (AAB18416.1), which were both confirmed to be active when expressed in *E. coli* (Subramaniam et al., 1996; Kakuta et al., 2001), is shown in **Supplementary Figure 3**. The phylogenetic tree shows that ACS3 is closely related to ACC synthase of *T. aestivum* and ACS1 is closely related to the confirmed gene of *P. citrinum*. The candidate genes have highly conserved residues involved in the catalytic activity and were therefore selected for experimental testing by heterologous expression in *E. coli*.

A BLAST search with a confirmed ACC oxidase from the basidiomycete *Agaricus bisporus* (Swiss-Prot H9ZYN5) showed that *F. graminearum* genome contains two predicted genes, *FGSG\_09103* and *FGSG\_11522*, with 79% and 88% query cover and 38% and 34% identical amino acids, respectively, while three other hits showed lower similarity. Both of these candidate genes are annotated as "isopenicillin N synthase and related dioxygenases," and are rather unlikely to have ACO activity, however, confirming this would require experimental testing.

While the results of Chen et al. (2009) suggest that *F. graminearum* exploits signaling for virulence by having the ability to produce ethylene, also, the opposite scenario cannot be excluded. We therefore searched for ACC deaminase candidate genes. A BLASTP search with a biochemically characterized ACC deaminase of *Trichoderma asperellum* (ACX94231.1, Viterbo et al., 2010) showed two hits in which the query cover was 99% and 98%, respectively, with 84% and 42% identical amino acids. These two genes (*FGSG\_02678*/*FGRAMPH1\_01T06417*, named *ACD1* and *FGSG\_12669*/*FGRAMPH1\_01T16927*, named *ACD2*) are also annotated as ACC deaminases also in FungiDB. The alignment of the confirmed ACC deaminase of *T. asperellum* with both candidate genes of *F. graminearum* is shown in **Supplementary Figure 4**. The phylogenetic analysis showed that *ACD2* is closer related to the confirmed ACC deaminase of *T. asperellum*.

#### Activity Test of *ACS* Candidate Genes

To test for ACC formation activity, we cloned all three *ACS* candidate genes into pETDuet-1. This vector allows co-expression with an ACC deaminase in an *in vivo* test (see below). *E. coli* transformants were induced to express the *Fusarium ACS* candidate genes. The protein extracts were analysed by SDS-PAGE and bands of the expected sizes were detected (see **Supplementary Figure 5**). Feeding assays were performed with induced cells, and the consumption of methionine added to the medium, as well as the potential formation of ACC in the cell pellets were measured by GC-MS. Yet, none of the transformants showed activity. Similarly, protein extracts generated from induced ACS-expressing *E. coli* were tested using S-adenosyl-L-methionine as substrate, but again, ACC formation could not be detected. Some ACC synthases have been reported to be quite unstable proteins and difficult to handle. We therefore purified the proteins with spin columns to speed up the purification process, again without success. Consequently, we tried an *in vivo* approach by complementation of an isoleucine deficiency in *E. coli via* coupled ACS/ACD activity*.* The test was based on the ability of an *ilvA* mutant

(1:10, 1:50, 1:100, 1:200 from left to right) onto the following media: M9 (left panel), M9 + isoleucine (middle) and M9 + 0.05 mM ACC + 1 mM IPTG + 100 mg/L ampicillin (right). 1, T7 express; 2, T7 express ∆*ilvA*; 3, ∆*ilvA* + pETduet1; 4, ∆*ilvA* + *ACD1* from *P. putida*; 5, ∆*ilvA* + *ACD1* from *F. graminearum*; 6, ∆*ilv5* + *ACD2* from *F. graminearum*.

to grow on minimal medium (M9) without added isoleucine (Tarun et al., 1998), first due to the conversion of S-adenosylmethionine into ACC by an active ACS, and further conversion of ACC to the isoleucine precursor α-ketobutyrate by an ACC deaminase (see **Figure 1A**). For these experiments, the *Pseudomonas* ACD was cloned into pACYCDuet-1, which is compatible with pETDuet-1 and allows both proteins to be produced in one cell. We also combined the two genes in one plasmid (pETDuet-1), but no growth could be detected after co-expression, indicating a lack of ACS activity. Tests with ACC added to the medium and expression of ACD showed the functionality of the ACD part of the assay (see below).

#### Testing ACD Candidate Genes in the *E. coli ilvA* Knock-out Strain

The product of the ACC deaminase reaction, α-ketobutyrate, is an intermediate of isoleucine biosynthesis in *E. coli*. If ACC is supplemented and ACD activity is present, an *ilvA* knockout strain is able to grow using the KBA formed from ACC to produce isoleucine. To allow testing of expression vectors based on the T7 RNA polymerase expression system (pET vectors), the *ilvA* mutation was introduced into the *E. coli* strain T7 Express by transferring the *ilvA*::KanR mutation from a strain from the *E. coli* knockout collection into the expression host (see *Materials and Methods* section). The ACD expression plasmids were introduced into the strain generated (T7 Express *ilvA*::KanR). Both the pETDuet-1 vector containing the *Pseudomonas putida* ACD gene serving as positive control and the *Fusarium ACD2* expression vector allowed growth of the *ilvA* mutant on minimal media supplemented with ACC, demonstrating the presence of active enzymes. In contrast, no evidence for activity of the *ACD1* encoded protein was obtained with this assay (**Figure 1B**).

#### Characterization of the ACC Deaminase Activity of the *ACD2* Gene Product

The pilot experiment with the modified *E. coli ilvA* strain already indicated that Acd2 has the predicted ACC deaminase activity. This was also confirmed by a colorimetric assay with permeabilized cells (see **Supplementary Figure 6**). Furthermore, a test of the crude protein extract revealed that Acd2 shows ACC deaminase activity similar to the positive control of *Pseudomonas putida,* while Acd1 did not show measurable activity. Likewise, GC-MS analysis did not reveal activity of Acd1, but clearly KBA formation was detected with Acd2. A feeding assay using ACD-expressing *E. coli* revealed that independently of the expressed gene, supplemented KBA was consumed by *E. coli* and also partly transaminated to 2-aminobutyric acid released to the medium. The activity test was repeated with the affinity purified protein (**Figure 2A**). The Acd2 protein remained constantly active over the whole assay period observed (60 min) and was used to determine the enzymatic properties (**Figure 2B**). The Km value of Acd2 for ACC was determined to be 3.3 ± 0.7 mM and vmax to 1.3 ± 0.06 µmol mg−1 min−1.

#### Acd1 Shows d-Cysteine Desulfhydrase Activity

It had previously been shown in an elegant work, that ACC deaminases and D-cysteine desulfhydrases from *P. putida* not only share a high sequence similarity, but can be interconverted into each other by site directed mutagenesis (Todorovic and Glick, 2008).

An alignment of both *Fusarium* candidate proteins with these *Pseudomonas* proteins is shown in **Figure 3**. The highlighted amino acids at positions 302 and 322 are specific for ACC deaminases and d-cysteine desulfhydrases. While Acd1 and the cysteine desulfhydrase from *P. putida* contain leucine and threonine, respectively, at these positions, Acd2 and the *Pseudomonas* ACC deaminase have methionine and leucine. Acd1 was therefore tested with D-cysteine as substrate. d-cysteine desulfhydrases convert d-cysteine into pyruvate, H2S and NH3. The formed pyruvate reacts with 2,4-dinitrophenylhydrazine, and this activity could be confirmed using the colorimetric assay. Neither L-cysteine nor S-(3-aminoethyl)-L*-*cysteine, which were used as alternative substrates, were consumed confirming the substrate specificity. The Km value of Acd1 for d-cysteine was determined to be about 18 mM and vmax was 5.5 µmol mg−1 min−1. Since the ACD1/FGSG\_02678 gene product shows d-cysteine desulfhydrase activity but not ACC deaminase activity we renamed the gene *DCS1*.

Due to the high specificity of the d-cysteine desulfhydrase and the presumably very low concentrations of d-cysteine *in planta*, we assumed that the fungus itself might be able to racemize L-cysteine to d-cysteine instead of obtaining the latter from the plant. To test this, a protein extract from a *Fusarium* culture, which had been pre-treated with d-cysteine, was incubated with L-cysteine for 24 h. Indeed, an increase in the d-cysteine concentration from a level below the limit of detection to around 2.5 mM was determined (data not shown, see Materials and Methods section) indicating the presence of a cysteine racemase. Yet, protein extracts of *Fusarium* cultures, which had been pre-exposed to L-cysteine first, did not show any detectable d-cysteine formation.


#### or for an ACC deaminase (yellow), respectively.

### *ACD2* and *DCS1* Knock-Out Strains Show Unaltered Virulence

We constructed *acd2*∆ and *dcs1*∆ knock-out strains using a *hphamdS* cassette flanked by two *loxP* sites (Steiger et al., 2011). The virulence of three independent mutants lacking *DCS1* and four independent *ACD2* deletion strains (**Supplementary Figures 7A** and **B**) was tested with 15 replicates per strain. On average, a marginally lower virulence of both, the *acd2*∆ and *dcs1*∆ knock-out strains was observed, but the results did not reach the threshold of statistical significance (**Figures 4A, B**). Virulence tests were also carried out with the double-knockout strains. Neither the single (*acd2*∆ or *dcs1*∆), nor the double knock-out (*acd2*∆ *dcs1*∆) strains (**Supplementary Figure 7C**) showed a significantly reduced virulence, indicating that these genes are dispensable during infection (**Figure 4C**).

The main problem with *F. graminearum* is contamination of grain with the mycotoxin deoxynivalenol (DON) and the plant detoxification product DON-3-glucoside (D3G). The *Arabidopsis*  DON-detoxifying glucosyltransferase UGT73C5 is rapidly and transiently induced by ACC and jasmonic acid (Poppenberger et al., 2003). We therefore determined the concentrations of DON and D3G in infected ears at the endpoint by grinding and extracting 15 infected ears individually. The average DON values of the heads infected with *acd1*∆ were lower compared to wild type treated ears. Yet, these values were not significantly different due to high fluctuations in DON content between individual samples (**Supplementary Figure 8**).

#### *Fusarium graminearum* Is Able to Use ACC as Sole Nitrogen Source

Plant growth promoting rhizobacteria (PGPR), which provide increased stress tolerance to plants, had been shown to be able to grow on media with ACC as sole nitrogen source (Penrose and Glick, 2003). We therefore tested whether also *F. graminearum* can grow on ACC. Spores of four independent *acd2*∆ knockout strains were spotted on modified FMM with ACC as sole nitrogen source together with one ectopic mutant and the wild type control (PH-1). After 1 week of incubation, the wild-type and the ectopic mutant grew well on ACC as sole nitrogen source, while

FIGURE 4 | Virulence tests of the single and double knock-out mutants. Ears of the wheat cultivar Apogee were infected with PH-1 and the indicated mutant strains. (A) and (B): PH-1 vs. ∆*acd2* and ∆*acd1*, respectively; the mean number of infected spikelets from 15 inoculated ears are shown for each strain and time point. (C) PH-1 (mean number of infected spikelets from 10 inoculated ears) compared to six independent double knock-out strains (∆∆*acd1,2*) derived from two independent acd1 single knock-out strains. For each graph, the mean number of infected spikelets from 10 inoculated ears for each of the three independent double knock-out strains were used at each time point. t-test was performed but no statistically significant difference was evidenced.

growth of the *ACD2* knockout strains was strongly retarded but not completely blocked (**Figure 5**). In contrast, all strains were equally well growing on standard FMM plates which were used as control (not shown). A possible explanation for continued growth of the mutant strains might be a partial non-enzymatic degradation of ACC in the medium.

#### DISCUSSION

Our bioinformatical investigation revealed that various *Fusarium*  species (particularly from the *F. oxysporum* and *F. fujikuroi* complex) might possess EFE as a means to synthesize ethylene. Yet, it remains to be tested whether the candidate genes are indeed encoding enzymatically active proteins. Existing gene models have to be used with caution. For instance, in the case

of *Fusarium mangiferae*, where fungal ethylene production has been implicated in malformation of mango fruits (Ansari et al., 2013), the gene model is questionable (see **Supplementary Figure 1**). A closer look at the DNA sequence of FMAN\_06358 showed that the annotated intron might not be real. However, intron retention leads to a model with a frameshift, as indicated in **Supplementary Figure 1**. Questionable non-functional splice sites might be a cause for truncated *F. oxysporum* models lacking the last exon, and hence, for these species, experimental testing is warranted. Nevertheless, the focus of this study was on the cereal pathogen *F. graminearum*, and for this species, ethylene synthesis *via* the EFE pathway can be excluded.

Three candidates for ACC synthases were found in *F. graminearum*. Orthologs of these genes also exist in other *Fusarium* species. Therefore, certain *Fusarium* species apart from *F. graminearum* might potentially possess both the EFE and the ACS pathway, a phenomenon so far only described for *Penicillium digitatum* (Chalutz and Lieberman, 1977). Activity of heterologously expressed ACC synthases in *E. coli*, e.g. of the wheat ACC gene (Subramaniam et al., 1996, shown in the alignment in **Supplementary Figure 3**), has been described. Yet, many plant ACC synthases are notoriously unstable and difficult to handle *in vitro*. Our efforts to demonstrate synthesis of ACC with S-adenosyl-methionine *in vitro* with the *E. coli*-expressed affinity-purified protein were unsuccessful with all three candidates. Likewise, upon feeding of IPTG induced *E. coli* cells with methionine, no ACC formation in cell extracts could be detected. We therefore adopted the growth-based assay originally described by Tarun et al. (1998) for use with the T7 expression system. With the functional *Pseudomonas putida* ACC deaminase as a bridge (see **Figure 1**), no growth upon expression of any of the *F. graminearum* ACS candidate genes was observed on M9 medium (data not shown). However, in an experiment using headspace GC-MS, ethylene production by *F. graminearum* on minimal media supplemented with 20 mM methionine was observed, which means the fungus either has ACC synthase activity or uses the transamination pathway *via* KMBA.

Potentially, the *Fusarium* ACC synthases may be active only in a complex with other proteins and could play a role in the synthesis of secondary metabolites containing ACC or ACC-like substances. Recently, a first bacterial ACC synthase was described (Xu et al., 2018), which is involved in synthesis of guangnanmycin. Coronatine, a well-known *Pseudomonas* metabolite interfering with defense signaling (Geng et al., 2014), contains coronamic acid, which has strong structural similarity to ACC and induces the synthesis of ethylene from methionine (Kenyon and Turner, 1992) by activating JA signaling (Uppalapati et al., 2005). An isolate of the fungus *Trichothecium crotocinigenum*, belonging to the hypocreales like *Fusarium*, is able to produce a diketopiperazine type cyclic "dipeptide" containing ACC (cyclo-(L-pipecolinylaminocyclopropane-carboxylic acid) (Long, 2014)). Also in the cotoxin II nonribosomal peptide from the plant pathogen *Bipolaris zeicola* (Ueda et al., 1992) ACC is a building block. Yet, none of the ACC synthase candidate genes is located in or near a predicted gene cluster suggestive of a role in secondary metabolite biosynthesis. Both ACC deaminase genes and two of the three ACC synthase candidates are located on the slowly evolving, conserved sub-genome (Wang et al., 2017). Only *ACS2* (FGSG\_07606) is located on the fast evolving subgenome. Analysis of published transcriptomics data from Zhang et al. (2012) and Wang et al. (2017) indicates that both *ACD1* (renamed *DCS1*) and *ACD2* are preferentially expressed during infection as compared to vegetative mycelial growth (5.3 fold and 6.8-fold higher, respectively), while of the ACC synthase genes, only the first candidate *ACS1* (FGSG\_05184) exhibits increased expression in planta (3.8-fold). We plan in further work to perform metabolomic analyses using the generated ACS knockout strains. While this research was going on, an alternative gene model for ACS3 has been deposited in FUNGI DB. The former ACS3, FGSG\_13587 gene model was modified upon availability of more robust RNA sequencing data. The current model, FGRAMPH1\_01G27057, incorporates some changes due to an updated splicing pattern and extended N- and C-termini (**Supplementary Figure 3**). Furthermore, extensive adenosine to inosine editing (Bian et al., 2019) during sexual development was found in all three ACS candidate genes (Liu et al., 2019).

Regarding ACC oxidases, we are not aware of any biochemically characterized functional enzyme from an ascomycete. There are reports that ethylene synthesis and signalling play a role in sexual development of the slime mold *Dictyostelium discoideum* and also in the basidiomycete *Agaricus bisporus* (Wood and Hammond, 1977; Zhang et al., 2016). Silencing and overexpression of a putative *ACO* gene (http://dictybase.org/gene/DDB\_G0277497) from the slime mold affected ethylene production (Amagai et al., 2007). Using the *D. discoideum ACO* gene as a BLAST query against the *F. graminearum* genome yields a list of genes that is similar to the results of a query using the *A. bisporus ACO*, but with a different order. Since we could not demonstrate ACC synthase activity, we did not further investigate whether these multiple candidates have indeed ACO activity, but this is clearly a topic for further research.

The search for potential ethylene production genes was inspired by the claim by Chen et al. (2009) suggesting *F. graminearum* might exploit ethylene production to trigger increased plant susceptibility. Yet, many transcriptome studies suggest the opposite (Kazan and Gardiner, 2018). This is true also for other *Fusarium* species. For instance, upregulation of banana ethylene biosynthesis and of ethylene-responsive transcription factors were implicated in the resistance response against *Fusarium oxysporum* f. sp. *cubense* (Li et al., 2013). Consequently, we also searched for ACC deaminase candidate genes having the potential to counteract ethylene defense signaling. Only one of the two purified ACD candidate gene products, Acd2, showed activity with ACC in the classical colorimetric assay with toluene-permeabilized cells (Penrose and Glick, 2003). The *F. graminearum* Acd2 enzyme has kinetic properties very similar to previously described ACDs from plantbeneficial bacteria or pathogenic fungi. For instance, the ACC deaminases from *Pseudomonas putida* and *Penicillium citrinum* showed Km values of 3.40 and 4.80 mM, respectively, compared to 3.31 mM for *F. graminearum* Acd2. The enzymatic activity of Acd2 did allow growth of isoleucine-deficient *E. coli* on M9-medium supplemented with ACC. Yet, the *Fusarium* Acd2 protein seems to work better at lower temperatures, e.g., 30°C, than the routine *E. coli* growth temperature of 37°C.

With an increasing number of genome sequences available, it became evident that gene products from bacteria (Ekimova et al., 2018) and from plants with high similarity to ACC deaminases exist. However, they are inactive with ACC, but have D-cysteine desulfhydrase activity (DCS). It was shown by site-directed mutagenesis that exchanging only two amino acids was sufficient to convert *Pseudomonas* ACC deaminase into an enzyme with Dcysteine desulfhydrase activity, and *vice versa* (Todorovic and Glick, 2008). We noticed that the inactive ACD1 (DCS1) candidate had the amino acids suggesting it might be a D-cysteine desulfhydrase (**Figure 3**). This was confirmed by enzymatic assays with the purified protein. The *F. graminearum* enzyme is inactive with L-cysteine. D-cysteine desulfhydrase catalyses the following reaction: D-cysteine + H2O → H2S + NH3 + pyruvate. The product H2S is, like ethylene, a gaseous signaling molecule with an increasingly recognized importance in animals and plants (Li et al., 2016; Filipovic and Jovanovic, 2017; Ziogas et al., 2018; Corpas et al., 2019). In crosstalk with other hormones, H2S is mediating increased resistance to abiotic stresses (Shi et al., 2015), for instance drought resistance in wheat (Li et al., 2017). In banana, hydrogen sulfide slows down senescence *via* affecting ethylene signaling (Ge et al., 2017) by downregulating the expression of both ACC synthase and ACC oxidase genes. Hydrogen sulfide can modify many proteins and change their activity by persulfidation or S-sulfhydration, where the cysteine-SH is converted into a persulfide group (-SSH) (Aroca et al., 2015). It has been recently shown that H2S in tomato negatively regulates ethylene biosynthesis by persulfidation of ACC oxidase (Jia et al., 2018). The relevance of H2S in plantpathogen interaction is still largely unknown with the exception of some storage diseases (Huo et al., 2018).

*F. graminearum* and other species are able to produce auxin, and auxin levels are increased in *Fusarium* infected plants (Kidd et al., 2011; Wang et al., 2018). Elevated auxin levels trigger induction of ACC synthase and oxidase in various plants (Grossmann, 2000). We have unpublished evidence that, in an early stage of the interaction, auxin production is indeed a virulence factor of *F. graminearum* (Svoboda et al., in preparation). We hypothesized that ethylene might increase resistance of the plant when formed later in a necrotrophic interaction, and that the ACC deaminase candidate genes might be used by *Fusarium* to counteract ethylene formation. In this case, inactivation of ACC deaminase should reduce the virulence of mutants. To test the relevance of the ACC deaminase *ACD2* and *DCS1* (formerly *ACD1*) in plant disease development we generated single-gene knockout strains, and additionally the *acd2*∆ *dcs1*∆ double mutants. These strains were used for infection of the rapid cycling and highly *Fusarium*-susceptible wheat cultivar Apogee. No significant difference was detectable at the end-point or in the area under the disease progression curve. Furthermore, the levels of the mycotoxin DON and the plant detoxification product DON-3-O-glucoside determined at the endpoint of the infections after 16 days were highly variable. No significant difference was observed between wild type, the single *acd2*∆ and *dcs1*∆ strains, the ectopic insertion mutants or the double mutant. ACC deaminase and D-cysteine desulfhydrase are obviously dispensable for full virulence of *F*. *graminearum* on wheat, at least on the highly susceptible cultivar Apogee. This is in contrast to the interaction of tomato with the root-infecting pathogen *Verticillium dahliae*. In this pathosystem, host resistance to the wilt pathogen could be increased *via* expression of a bacterial ACC deaminase in tomato, which was under control of promoters limiting the expression to the site of infection (Robison et al., 2001). In line with this, it was recently reported that ACC deaminase knockout mutants of *Verticillium* were less virulent, while overexpression led to increased virulence (Tsolakidou et al., 2018). The only other evidence for a role of a fungal ACD we are aware of is the case of the plant growth promoting fungus *Trichoderma asperellum* T203. In this case, ACD was not deleted, but silenced using RNAi, the effects of which are often ephemeral. Nevertheless, the results suggested that the strain with reduced ACD activity has a decreased ability to promote root elongation of *Brassica napus* (Viterbo et al., 2010).

We tested whether *F. graminearum* can utilize ACC as a nitrogen source, which is clearly the case (**Figure 5**). Thus, instead of preventing the mobile signaling molecule ACC, generated at the infection front, from moving to other plant parts and triggering resistance, the main but marginal function of ACC deaminase could be utilization of a low-abundance nitrogen source. The enzyme has a Km in the mM range. The highest reported (Wang et al., 2018) ACC concentration in infected ears of the highly *Fusarium* resistant cultivar Sumai3 of around 45 mg/kg could be relevant, especially assuming the local concentrations might be higher than this average. On the other hand, these <0.5 mM amounts of ACC seem to be of minor relevance, considering that reported levels of free amino acids in developing grain are in the range of 300 to 50 mM from 7 days post-anthesis to ripening (Howarth et al., 2008).

Likewise, the role of D-cysteine desulfhydrase in *Fusarium* is enigmatic. The expression of both *ACD2* and *DCS1* is higher during plant infection than in axenic cultures. A cysteine desulfhydrase gene (dcyD/yedO) also exists in *E. coli*. It is upregulated during sulfur starvation and may allow utilization of D-cysteine (if available in the environment). The only other known function of dcyD is protection against toxicity of very high levels of D-cysteine when present in minimal medium, since D-cysteine blocks threonine deaminase, the gene product of *ilvA* (Soutourina et al., 2001). Toxicity of D-cysteine is low in rich medium or when isoleucine is supplemented. Nevertheless, there are D-cysteine containing secondary metabolites that are orders of magnitude more toxic. If *Fusarium* is able to degrade them, *DCS1* might play a role to utilize the released D-cysteine. For instance, the *Aspergillus* metabolite malformin A contains two D-cysteine residues in the cyclic pentapeptide. Recently, malformin E was described to have a minimal inhibitory concentration of about 7 µM for *Fusarium solani* (Ma et al., 2016) and below 1 µM for *E. coli*. Many microbes can also produce highly toxic aminovinyl-cysteine-containing peptides that contain D-cysteine (Sit et al., 2011). This suggests a role in microbe-microbe interaction rather than in plant-pathogen interaction. Yet, the advantage seems minimal, since a *dcs1∆* mutant is still able to grow on D-cysteine (data not shown), which is expected since the *F. graminearum* genome contains D-amino-acid oxidases.

We did not obtain evidence supporting a role of *DCS1* in plant disease. Still, utilization of a (marginal) sulfur source could be a function. The major sulfur transport forms during grain filling are glutathione and S-methyl-methionine (Bourgis et al., 1999), and both are highly abundant. While we are not aware of any reports of relevant concentrations of D-cysteine in plants, it has been described that peas possess transaminases producing D-alanine—and possibly other D-amino acids—in seedlings (Ogawa et al., 1973). It is, however, difficult to envision how a low-affinity enzyme with a Km of ~18 mM could make a relevant contribution to the utilization of sulfur sources. On the other hand, plants also have D-cysteine desulfhydrases (e.g. Arabidopsis AT1G48420), which may be involved in H2S signaling. Potentially, the fungus itself may be able to synthesize D-cysteine from L-cysteine. Our preliminary experiments indicated that *F. graminearum* has racemase activity that can potentially convert L-cysteine into D-cysteine. Cysteine racemase is a rarely reported activity (Espaillat et al., 2014). Formed D-cysteine might then be used by Dcs1 to produce a volatile H2S signal. Potentially, *F. graminearum* might be able to take up cysteine from the host and affect levels of glutathione, which is crucial for maintenance of the redox balance during infection. Glutathione (GSH) levels clearly have an impact on plant defense in general, most likely *via* affecting NPR1 and salicylic acid production (Ghanta et al., 2011; Kovacs et al., 2015) but also direct effects on ethylene production were reported (Datta et al., 2015). *Arabidopsis thaliana* transgenic plants with increased GSH content upregulated several ACS transcripts, and in addition, ACO protein levels were also increased, while in a *pad2-1* mutant with reduced GSH levels, the opposite effect was observed (Datta et al., 2015). Yet, our virulence tests do not support such a scenario. Inactivation of both ACC deaminase and D-cysteine desulfhydrase did not have a significant impact on the virulence of *F. graminearum*, at least not in the highly susceptible wheat cultivar Apogee. This outcome is not unprecedented. Other genes with a highly suggestive function were also found to be dispensable, such as genes encoding enzymes for degradation of salicylic acid (Hao et al., 2018; Qi et al., 2019; Rocheleau et al., 2019), despite clear evidence that salicylic acid plays an important role in the defense of wheat against *F. graminearum* (Makandar et al., 2006; Makandar et al., 2012). In future work we intend to

#### REFERENCES


investigate the role *Fusarium* ACC synthase candidate genes by generating triple knock-out mutants, to clarify the role of fungal ethylene production in the interaction with the host plant.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and the supplementary files.

### AUTHOR CONTRIBUTIONS

TS constructed plasmids. TS performed experiments. GW, RS, and GA conceived the concept. GW supervised experimental work. KT and UG did the bioinformatical analyses. DS, RH, MV, and AP performed the analytical measurements. TS, KT, GW, and GA wrote the paper and all authors amended and corrected the paper.

### FUNDING

This work was funded by the Austria Science Fund (FWF), SFB Fusarium (F3702 and F3715), and the DFG (LAP3714).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01072/ full#supplementary-material


N and S deficiency during wheat grain filling. *J. Exp. Bot.* 59, 3675–3689. doi: 10.1093/jxb/ern218


*Fusarium* head blight. *Plant Cell Rep.* 25, 313–319. doi: 10.1007/s00299-005- 0059-4


favors homologous integration and employs reusable bidirectionally selectable markers. *Appl. Environ. Microbiol.* 77, 114–121. doi: 10.1128/AEM.02100-10


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Svoboda, Parich, Güldener, Schöfbeck, Twaruschek, Václavíková, Hellinger, Wiesenberger, Schuhmacher and Adam. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Light Modulates Ethylene Synthesis, Signaling, and Downstream Transcriptional Networks to Control Plant Development

*Alexandria F. Harkey1, Gyeong Mee Yoon2, Dong Hye Seo2, Alison DeLong3 and Gloria K. Muday1\**

*1 Department of Biology and Center for Molecular Signaling, Wake Forest University, Winston-Salem, NC, United States, 2 Department of Botany and Plant Pathology, Purdue University, West Lafayette, IN, United States, 3 Department of Molecular Biology, Cell Biology and Biochemistry, Brown University, Providence, RI, United States*

#### *Edited by:*

*Dominique Van Der Straeten, Ghent University, Belgium*

#### *Reviewed by:*

*Jan Smalle, University of Kentucky, United States Anna N. Stepanova, North Carolina State University, United States*

> *\*Correspondence: Gloria K. Muday muday@wfu.edu*

#### *Specialty section:*

*This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science*

*Received: 02 June 2019 Accepted: 09 August 2019 Published: 12 September 2019*

#### *Citation:*

*Harkey AF, Yoon GM, Seo DH, DeLong A and Muday GK (2019) Light Modulates Ethylene Synthesis, Signaling, and Downstream Transcriptional Networks to Control Plant Development. Front. Plant Sci. 10:1094. doi: 10.3389/fpls.2019.01094*

The inhibition of hypocotyl elongation by ethylene in dark-grown seedlings was the basis of elegant screens that identified ethylene-insensitive Arabidopsis mutants, which remained tall even when treated with high concentrations of ethylene. This simple approach proved invaluable for identification and molecular characterization of major players in the ethylene signaling and response pathway, including receptors and downstream signaling proteins, as well as transcription factors that mediate the extensive transcriptional remodeling observed in response to elevated ethylene. However, the dark-adapted early developmental stage used in these experiments represents only a small segment of a plant's life cycle. After a seedling's emergence from the soil, light signaling pathways elicit a switch in developmental programming and the hormonal circuitry that controls it. Accordingly, ethylene levels and responses diverge under these different environmental conditions. In this review, we compare and contrast ethylene synthesis, perception, and response in light and dark contexts, including the molecular mechanisms linking light responses to ethylene biology. One powerful method to identify similarities and differences in these important regulatory processes is through comparison of transcriptomic datasets resulting from manipulation of ethylene levels or signaling under varying light conditions. We performed a meta-analysis of multiple transcriptomic datasets to uncover transcriptional responses to ethylene that are both light-dependent and light-independent. We identified a core set of 139 transcripts with robust and consistent responses to elevated ethylene across three root-specific datasets. This "gold standard" group of ethylene-regulated transcripts includes mRNAs encoding numerous proteins that function in ethylene signaling and synthesis, but also reveals a number of previously uncharacterized gene products that may contribute to ethylene response phenotypes. Understanding these light-dependent differences in ethylene signaling and synthesis will provide greater insight into the roles of ethylene in growth and development across the entire plant life cycle.

Keywords: ethylene, light, transcriptomic meta-analysis, ethylene response, ethylene biosynthesis, hypocotyl, root

## INTRODUCTION

Plant responses to the gaseous hormone ethylene are an excellent model for studying the relationships between hormone synthesis, signaling, transcriptional changes, and development. The identification of ethylene-insensitive mutants in Arabidopsis using molecular genetics opened a new era in dissecting plant hormone signaling (Bleecker et al., 1988; Guzman and Ecker, 1990). Ethylene-insensitive mutants were identified as lacking the ethylene "triple response" in dark-grown seedlings (short, thick hypocotyl and exaggerated apical hook), remaining tall in the presence of excess ethylene (Alonso et al., 2003; Guo and Ecker, 2003; Yanagisawa et al., 2003). This approach enabled the isolation of mutations affecting the activities of core ethylene response machinery, including receptors, signaling proteins, and transcription factors. The functions of these signaling components, as well as the pathways for ethylene synthesis, have subsequently been assayed in additional tissues beyond dark-grown hypocotyls, demonstrating that many of these proteins function in all tissues and growth conditions, but also revealing branches of the ethylene signaling and synthesis pathways that have distinct roles in light-grown plants and in other developmental stages. In particular, ethylene-responsive transcriptional networks and regulatory controls of ethylene biosynthesis show profound differences between light- and dark-grown tissues. Although some of these differences have been reviewed previously (Rodrigues et al., 2014; Booker and DeLong, 2015; Yoon, 2015; Yu and Huang, 2017), recent studies have identified new mechanisms and yielded insight into lightdependent differences. This review highlights the similarities and differences in light-dependent regulation of ethylene synthesis and response in seedlings grown at a range of light levels, focusing on recent publications establishing that the genetic redundancy in ethylene biosynthetic machinery, ethylene receptors, and transcriptional machinery may allow a complex suite of lightdependent developmental responses to this important hormone.

#### Basics of the Ethylene Signaling Pathway

The triple response of dark-grown seedlings was exploited in elegant genetic screens that identified mutants exhibiting either ethylene-insensitivity (*ein* or *etr* mutants) (Bleecker et al., 1988; Guzman and Ecker, 1990; Chang et al., 1993), enhanced ethylene signaling in the constitutive triple response (*ctr*) (Kieber et al., 1993; Huang et al., 2003), or synthesis in the ethylene overproducer (*eto*) mutants (Guzman and Ecker, 1990). The genes responsible for these phenotypes have been cloned and mapped to the ethylene signaling and biosynthetic pathways. The signaling pathway begins with ethylene binding to ER-localized receptor proteins (Kendrick and Chang, 2008), which act as negative regulators of the pathway (Hua and Meyerowitz, 1998). In Arabidopsis, these receptors are ETR1, ETR2, EIN4, ERS1, and ERS2 (Chang et al., 1993; Schaller and Bleecker, 1995; Hua and Meyerowitz, 1998; Sakai et al., 1998), which fall into two subfamilies based on sequence similarity of the ethylene binding domains and the presence of conserved histidine kinase domains (Kendrick and Chang, 2008; Stepanova and Alonso, 2009; Shakeel et al., 2013). When ethylene binds, the receptors are turned off, resulting in decreased activity of the inhibitory CTR1 protein kinase and increased EIN2 output (Kieber et al., 1993; Alonso et al., 1999; Huang et al., 2003; Qiao et al., 2009). C-terminal proteolytic cleavage of EIN2 promotes the nuclear localization of the EIN2 C-terminal proteolytic fragment (EIN2-CEND) (Ju et al., 2012; Qiao et al., 2012; Wen et al., 2012). EIN2-CEND-mediated targeting of *EBF1/2* mRNA to the processing body further enhances signaling output (Li et al., 2015; Merchante et al., 2015). Nuclear EIN2-CEND alters transcription *via* activation of the EIN3 and EIN3-LIKE (EIL1 and EIL2) transcription factors (TFs), which then turn on expression of genes encoding other TFs, such as *ERF1* and *EDF1-EDF4* (Chao et al., 1997; Solano et al., 1998; Alonso et al., 2003; Chang et al., 2013). These core TFs likely work with other TFs as part of a gene regulatory network leading to a diversity of transcriptional responses, which have been characterized in multiple genome-wide transcriptional studies (Stepanova et al., 2007; Chang et al., 2013; Feng et al., 2017; Harkey et al., 2018). Ethylene signaling is also modulated by EIN2-mediated translational regulation (Merchante et al., 2015), as well as F-box dependent proteolysis of EIN2 and EIN3 *via* ETP1/2 and EBF1/2, respectively (Guo and Ecker, 2003; Potuschak et al., 2003; Qiao et al., 2009). EBF1/2 are also destabilized by ethylene in an EIN2-dependent manner, allowing increased accumulation of EIN3 (An et al., 2010).

Ethylene signaling proteins have roles that extend beyond their functions in dark-grown Arabidopsis hypocotyls. Genes encoding these proteins have been found across the plant kingdom (Wang et al., 2015), and the proteins have been shown to function in a diversity of tissues and under a range of light conditions (Lanahan et al., 1994; Binder et al., 2006; Plett et al., 2009; Wilson et al., 2014a). Both CTR1 and EIN2 are required for normal ethylene responsiveness in all light conditions in Arabidopsis, indicating that each of these gene products plays a central and non-redundant role in ethylene signaling, regardless of light conditions. Mutants lacking *CTR1*  show constitutive ethylene responses in roots and shoots grown in light or dark (Kieber et al., 1993). Mutations in *EIN2* confer insensitivity to added ethylene in dark-grown hypocotyls (Alonso et al., 1999), light-grown rosettes (Kieber et al., 1993), light-grown hypocotyls (Smalle et al., 1997), and roots of darkgrown (Stepanova et al., 2005) and light-grown seedlings (Negi et al., 2008; Harkey et al., 2018).

Ethylene receptors are members of a conserved multigene family (Shakeel et al., 2013). As these receptors function as negative regulators, dominant gain-of-function (GOF) mutations, such as *etr1-1* and *etr1-3* in Arabidopsis (Bleecker et al., 1988; Guzman and Ecker, 1990; Chang et al., 1993) and *Neverripe* in tomato (Wilkinson et al., 1995), yield ethyleneinsensitive plants. In contrast, null or loss-of-function (LOF) alleles can confer constitutive ethylene response phenotypes (Hua and Meyerowitz, 1998; Shakeel et al., 2013). In Arabidopsis, the five ethylene receptors have been shown to have distinct roles that are tied to specific developmental responses (Shakeel et al., 2013), some of which can be studied only in older plants, which are necessarily grown in light. Similarly, the tomato *Neverripe* gene belongs to a seven-member ethylene receptor gene family and the *Neverripe* mutant carries a GOF mutation that confers ethylene insensitivity in phenotypes observed in both light and dark conditions (e.g., fruit ripening, hypocotyl triple response, and root development) (Wilkinson et al., 1995; Negi et al., 2010; Klee and Giovannoni, 2011). Tomato plants with knockdown of mRNA encoding receptors have also revealed distinct functions for two tomato ethylene receptors (Kevany et al., 2007). In the sections below, we highlight studies that have revealed differences in ethylene responses that are influenced by light and developmental stage, and which require distinct ethylene signaling or synthesis machinery.

#### Basics of the Ethylene Biosynthesis Pathway

The enzymatic steps of the ethylene biosynthetic pathway were uncovered in fruit; subsequent work in fruit and in dark-grown Arabidopsis seedlings identified a conserved biosynthetic pathway and revealed important regulatory mechanisms that control pathway activity (Adams and Yang, 1979; Yang and Hoffman, 1984; Booker and DeLong, 2015; Yoon, 2015). The simple and highly conserved pathway has only two committed steps: conversion of S-adenosyl-l-methionine (SAM) to 1-aminocyclopropane-1-carboxylic acid (ACC) by ACC synthase (ACS), followed by conversion of ACC to ethylene by ACC oxidase (ACO) (Houben and Van de Poel, 2019). ACS has been a primary target for researchers interested in understanding regulation of ethylene biosynthesis, as this enzyme catalyzes the first biosynthetic step, which is frequently described as the ratelimiting step (Adams and Yang, 1979; Yang and Hoffman, 1984). *ACS* gene families in land plants encode isozymes belonging to three classes, type-1, type-2, and type-3 (El-Sharkawy et al., 2008; Lin et al., 2009; Booker and DeLong, 2015; Zhu et al., 2015; Lee et al., 2019). The evolution and regulation of ACO, including consideration of conditions under which ACO activity is limiting for ethylene production, have been recently reviewed (Houben and Van de Poel, 2019). There are both transcriptional and post-translational mechanisms that control which ACS and ACO isozymes are expressed and active, leading to distinct enzyme populations in tissue- and developmental stage-specific contexts (Booker and DeLong, 2015; Houben and Van de Poel, 2019). Positive feedback loops, largely driven by transcriptional controls of these biosynthetic enzymes, drive dramatic increases in ethylene production to accelerate fruit ripening (Klee and Giovannoni, 2011). This review will examine new insight into the molecular mechanisms by which ethylene synthesis is modulated by light levels at both transcriptional and post-translational levels.

### LIGHT-DEPENDENT AND -INDEPENDENT ETHYLENE RESPONSES

#### Ethylene Effects in Hypocotyls Are Opposite in Light and Dark

The ethylene response in the hypocotyls of young seedlings is highly dependent on light level. The triple response of etiolated seedlings, including inhibited hypocotyl elongation, is the basis of much of the current molecular insight into ethylene signaling (Bleecker et al., 1988; Guzman and Ecker, 1990). Ethylene treatment under shade covering, rather than complete darkness, also leads to decreased hypocotyl growth (Das et al., 2016). The hypocotyl response to ethylene is coordinated with light-dependent hypocotyl elongation changes during photomorphogenesis (Yu and Huang, 2017). Light inhibits hypocotyl elongation, which is important as plants growing in soil transition to light (Montgomery, 2016). In opposition to the effect of ethylene in the dark, light-grown Arabidopsis seedlings show increased hypocotyl elongation in response to ethylene (Smalle et al., 1997; Le et al., 2005; Das et al., 2016; Seo and Yoon, 2019), as illustrated in **Figure 1**. In both light and dark, the ACC or ethylene response is tied to differences in cell expansion (Smalle et al., 1997; Seo and Yoon, 2019). These light-dependent differences have more frequently been reported in response to treatment with the ethylene precursor, ACC (Smalle et al., 1997; Le et al., 2005), but ethylene yields the same light-dependent increases in elongation (**Figure 1**), and ethylene-insensitive mutants are shorter than wild-type in the light (Le et al., 2005). Intriguingly, the nutrient content of the growth media affects the ethylene response in light-grown, but not dark-grown, seedlings (Smalle et al., 1997; Collett et al., 2000).

Another striking feature of the ethylene triple response in etiolated seedlings is the accentuation of the apical hook. As part of photomorphogenesis, the apical hook opens and cotyledons expand, so it is important to ask whether this ethylene response, like hypocotyl elongation, is also light dependent (Bleecker et al., 1988; Raz and Ecker, 1999; Mazzella et al., 2014; Van de Poel et al., 2015). The formation of apical hooks in etiolated seedlings protects the shoot apical meristem during growth through soil, and ethylene build-up in denser soil exaggerates this hook to assist in emergence (Zhong et al., 2014; Shi et al., 2016a). Ethylene insensitive mutants with receptor and signaling defects show impaired hook formation, while the *ctr1-1* null mutant has an exaggerated hook (Abbas et al., 2013). Localized accumulation of ACO across the hook may also contribute to hook maintenance in dark-grown seedlings (Peck et al., 1998; Raz and Ecker, 1999). Mutants with elevated ethylene synthesis show enhanced hook formation (Guzman and Ecker, 1990). A central feature of ethylene-accentuated hook formation is crosstalk with auxin. Asymmetries in auxin synthesis and auxin transport, which lead to accumulation of growth-inhibiting auxin levels on the inside of the hook, are enhanced by ethylene treatment (Vandenbussche et al., 2010; Zádníková et al., 2010). The process of hypocotyl hook opening in response to light is also ethylene regulated (Vandenbussche et al., 2010; Zádníková et al., 2010; Van de Poel et al., 2015). In dark-grown seedlings, the *ein3-1 eil1-1* double mutant has enhanced hook opening, while an EIN3 overexpression line has a tightly closed, exaggerated hook like *ctr1-1*, and shows delayed hook opening in the light (Zhang et al., 2018), consistent with ethylene negatively regulating hook opening in both light and dark.

#### Ethylene Modulates Light-Dependent and Light-Independent Root Development

In seedling roots, ethylene and ACC inhibit elongation in both light and dark conditions (Rahman et al., 2001; Ruzicka

et al., 2007; Stepanova et al., 2007; Swarup et al., 2007; Negi et al., 2008; Negi et al., 2010; Strader et al., 2010) while enhancing root hair initiation (Cutter, 1978; Tanimoto et al., 1995; Pitts et al., 1998; Dolan, 2001; Rahman et al., 2002; Strader et al., 2010). In both light- and dark-grown seedlings, these root responses to ethylene are lost in ethylene-insensitive *etr1-3*, a dominant gain of function (GOF) receptor mutant, and in the *ein2-5* signaling mutant (Ruzicka et al., 2007; Swarup et al., 2007; Negi et al., 2008; Lewis et al., 2011a). These effects on root elongation are tied to auxin and ethylene cross-talk in a light-independent fashion. Ethylene enhances auxin synthesis, transport, and signaling to control root development (Stepanova et al., 2005; Ruzicka et al., 2007; Stepanova et al., 2007; Swarup et al., 2007; Negi et al., 2008; Stepanova et al., 2008; Lewis et al., 2011a; Muday et al., 2012).

In contrast, the inhibitory effect of ethylene and ACC on lateral root (LR) formation in Arabidopsis and tomato has been examined only in light-grown seedlings, as LRs do not form in roots of dark-grown seedlings (Ivanchenko et al., 2008; Negi et al., 2008; Negi et al., 2010; Lewis et al., 2011b; Lewis et al., 2011a). Ethylene and ACC block early stages of LR initiation (Ivanchenko et al., 2008). As with the inhibition of root elongation, ethylene inhibits LR formation by modulating auxin synthesis, signaling, and transport, which control this process (Stepanova et al., 2007; Muday et al., 2012). Similarly, the effects of ethylene and ACC on root gravitropism and root waving, which have been assayed only in light-grown seedlings, also are blocked in the ethylene signaling mutants *ein2-5* and the GOF *etr1-3* receptor mutant (Buer et al., 2003; Buer et al., 2006). Overall, published data support a light-independent function of the EIN2 protein in ethylene signaling in roots (Ruzicka et al., 2007; Stepanova et al., 2007; Swarup et al., 2007; Negi et al., 2008; Lewis et al., 2011a). However, these data do not reveal which specific receptors function in the roots, because the use of GOF mutants (like *etr1-1* and *etr1-3*) can perturb the functions of the entire receptor family (Chang et al., 1993; Shakeel et al., 2013). Using LOF alleles in each receptor subtype is a powerful strategy to resolve the specific function of the family of ethylene receptors; this approach has been used to understand ethyleneregulated growth and development in a light-dependent context, as discussed below.

### MECHANISTIC CONNECTIONS BETWEEN LIGHT RESPONSE AND ETHYLENE BIOSYNTHESIS

Changes in ethylene synthesis in response to changing light levels have been reported in many different species and under many different growth conditions, with dramatically varying results. The ability of light to modulate ethylene synthesis was reported half a century ago, when a single dose of red light was shown to decrease ethylene levels in etiolated pea seedlings in a farred reversible manner, suggesting that phytochrome negatively controls ethylene biosynthesis (Goeschl et al., 1967). Conversely, high-intensity illumination of green seedlings induced an increase in ethylene synthesis, demonstrating a positive effect of light on ethylene production (Weckx and Van Poucke, 1989). Subsequent studies have confirmed that the effect of light on ethylene synthesis is complex and context-dependent (Foo et al., 2006; Khanna et al., 2007; Jeong et al., 2016; Song et al., 2018), and is also affected by crosstalk with other plant hormone response pathways (Vandenbussche et al., 2003; Arteca and Arteca, 2008; Muday et al., 2012; Lee et al., 2017). For instance, etiolated Arabidopsis seedlings show age- and light-dependent increases in ethylene biosynthesis with higher levels in the light; increased ethylene production is detectable as rapidly as 4 h after transfer to light, but becomes more dramatic with increasing time in light (Seo and Yoon, 2019). As discussed below, these effects are mediated at both the transcriptional and post-translational levels, and although much work has focused on regulation of ACS expression and activity, additional data reveal light-dependent effects on regulation of ACO function.

#### Light-Mediated Transcriptional Regulation of ACS and ACO

Regulation of ethylene synthesis *via* alteration of *ACS* and/or *ACO* gene expression is a primary mechanism through which differences in the quality, quantity, or periodicity of light modulate ethylene production and signaling outputs to coordinate plant growth and development (Yamagami et al., 2003; Tsuchisaka and Theologis, 2004; Wang et al., 2005). The combinatorial effects of light with phytohormones and biotic or abiotic stresses add further complexity to light-mediated control of ethylene biosynthesis. For example, IAA treatment induces expression of Arabidopsis *ACS* genes in seedlings grown in darkness or in constant light, but this induction is less dramatic in plants grown with a light/dark cycle (Rashotte et al., 2005). Furthermore, light differentially influences the transcript levels of various *ACS* genes, depending on the developmental stage and the length of light treatment (Seo and Yoon, 2019). The mRNA levels of a subset of type-1 and type-2 *ACSs* (*ACS6* and *ACS5*, *8,* and *9*, respectively) declined rapidly and steeply after etiolated seedlings were transferred to light, and these transcript levels remained low for 5 days. Meanwhile, *ACS2* (type-1) and *ACS4* (type-2) showed gradual increases in their transcript levels after light exposure (Seo and Yoon, 2019). Together, these data suggest distinct roles for ACS isozymes depending on the light conditions, with *ACS5*, *6*, *8*, and *9* playing the primary roles in dark-grown seedlings, while expression of *ACS2* and *ACS4* is implicated in controlling ethylene production in the light.

Analysis of light signaling mutants and transgenic lines expressing light signaling components has also provided insight into the light-mediated regulation of ethylene biosynthesis. Mutations in the phytochrome genes *PHYA* and *PHYB* increased ethylene biosynthesis in pea, consistent with a negative effect of light on ethylene synthesis, with a more profound effect observed in the *phyA* mutant (Foo et al., 2006). Intriguingly, in Arabidopsis and sorghum, *phyA* mutants show less profound increases in ethylene biosynthesis than do *phyB* mutants, indicating species-specific functions of these photoreceptors in controlling ethylene levels. Similarly, transgenic lines overexpressing Arabidopsis PHYTOCHROME-INTERACTING FACTOR5 (PIF5), a basic helix-loop-helix transcription factor that specifically interacts with the photoactivated form of PhyB, showed a marked increase in ethylene production in the dark that is correlated with increased abundance of *ACS4*, *ACS8*, and other *ACS* transcripts (Khanna et al., 2007). Although the *pif1 pif3 pif4 pif5* (*pifq*) mutant initially produced less ethylene than wild-type seedlings, consistent with the higher ethylene levels in *PIF5* overexpression lines, at later time-points the *pifq*  mutant showed higher ethylene production (Jeong et al., 2016), indicating a developmental stage-dependent role of PIFs in controlling ethylene biosynthesis.

The regulation of *ACO* gene expression has received much less study than that of *ACS* (Houben and Van de Poel, 2019), yet the levels of *ACO* transcripts are also regulated by light and other factors that control pathway activity (Argueso et al., 2007; Rodrigues et al., 2014). In tomato fruits, *ACO1* is upregulated by pulses of white light (Scott et al., 2018). Classic work demonstrated that *ACO*  expression is both a driver of ethylene production and a reporter for ethylene response in etiolated tissues (Peck and Kende, 1995; Kim et al., 1997), creating a positive feedback loop. *ACO* transcript increases have also been reported after ACC treatment of aerial tissues of light-grown seedlings (Zhong and Burns, 2003). The meta-analysis discussed below provides strong support for this feed-forward mechanism. Furthermore, when ACS activity is elevated during climacteric ripening in tomato or banana fruits (and during flooding stress), ACO activity becomes rate limiting, and ACO expression is up-regulated (Ruduś et al., 2013; Xiao et al., 2013; Houben and Van de Poel, 2019). This suggests that one role of the feed-forward mechanism is to "clear" excess ACC when ACO activity limits ethylene production.

#### Light-Mediated Post-Translational Control of ACS and ACO Activity

An early study suggested that light regulates ethylene biosynthesis by altering stability/activity of ACS isozymes (Rohwer and Schierle, 1982). More recent work has confirmed that light modulates ethylene biosynthesis *via* post-translational mechanisms including reversible phosphorylation and protein turnover (Steed et al., 2004; Chae and Kieber, 2005; Yoon and Kieber, 2013b; Zdarska et al., 2015; Seo and Yoon, 2019). Post-translational regulation of ACS is largely dependent on the regulatory motifs located in the C-terminus of ACS proteins (Chae and Kieber, 2005). All three ACS types contain a well-conserved N-terminal catalytic domain, whereas the C-termini vary among ACS isoforms. Type-1 ACSs (ACS1, 2, and 6 in Arabidopsis) possess phosphorylation target sites for mitogen-activated protein kinases (MAPKs) and calcium-dependent protein kinases (CDPKs) (Tatsuki and Mori, 2001; Hernández Sebastià et al., 2004; Liu and Zhang, 2004). Type-2 ACSs (ACS4, 5, 8, 9, and 11 in Arabidopsis) contain a phosphorylation site for CDPKs and a unique regulatory motif called Target of ETO1 (TOE) in the C-terminus. The TOE motif is the binding site for ETHYLENE OVERPRODUCER1 (ETO1) and its two paralogs, ETO1-LIKE1 and 2 (EOL1 and EOL2). ETO1/EOL1/EOL2 are BTB/TRP-containing E3 ligases that control the degradation of type-2 ACS proteins *via* the 26S proteasome (Yoshida et al., 2005). In contrast to both type-1 and type-2 ACSs, the single type-3 ACSs does not contain known regulatory motifs in the C-terminus, but as discussed below, an N-terminal motif may control the stability of Arabidopsis ACS7 (Xiong et al., 2014), a sole type 3 isozyme in Arabidopsis.

The protein stability of all three ACS isozyme types is regulated by 14-3-3 proteins (Yoon and Kieber, 2013a). 14-3-3 proteins are an evolutionarily well-conserved family of regulatory proteins involved in numerous cellular processes such as cell cycle regulation, cell division, cell metabolism, proliferation, and protein oligomerization and localization (Dougherty and Morrison, 2004; Darling et al., 2005; Oecking and Jaspert, 2009; Freeman and Morrison, 2011). 14-3-3 activity influences ethylene biosynthesis by destabilizing ETO/EOL proteins and by stabilizing ACS proteins in an ETO/EOL-independent manner (Yoon and Kieber, 2013a). The range of light-dependent developmental phenotypes observed in 14-3-3 LOF mutants (Pnueli et al., 2001; Mayfield et al., 2007; Tseng et al., 2012; Adams et al., 2014) suggests interaction with multiple light signaling components. Although there is no direct evidence that light regulates interactions between 14-3-3 proteins, ACS isozymes, and ETO/EOLs, the 14-3-3s proteins are logical candidates to mediate crosstalk between light signaling and ethylene biosynthesis pathways.

Light-dependent post-translational control of ACS5 (and perhaps other type-2 ACSs) and the associated increase in ethylene production are critical for regulating hypocotyl elongation during the dark-to-light transition. Intriguingly, PIF3 may be involved in this process (Seo and Yoon, 2019). As described above, PIF3 is required for ethylene-induced stimulation of hypocotyl elongation in the light, and ethylene treatment specifically antagonizes light-induced degradation of PIF3 (Zhong et al., 2012). Light-induced stabilization of type-2 ACS enzymes should lead to increased ethylene production, which may play a role in PIF3 stabilization, thereby driving ethyleneinduced hypocotyl elongation in the light (Seo and Yoon, 2019). PP2A is another regulatory component that contributes to post-translational regulation of ACS stability. Genetic analysis indicated that PP2A-mediated dephosphorylation negatively controls the protein stability of ACS6 in the dark, but has a much weaker effect on ethylene production in the light (Skottke et al., 2011). Paradoxically, the stability of ACS5, a type-2 isozyme, is positively regulated by PP2A; differential effects on the two isozyme types likely accounts for the lesser effect of PP2A inhibition in light-grown plants (Muday et al., 2006; Skottke et al., 2011).

Compared to type-1 and type-2 ACS isozymes, the sole Arabidopsis type-3 isozyme, ACS7, has unique protein stability characteristics; regulation of ACS7 turnover remains somewhat controversial (Lyzenga et al., 2012; Xiong et al., 2014; Lee et al., 2017). Because of the lack of C-terminal regulatory motifs in type-3 ACS, it was thought that these isozymes might be generally stable compared to other ACS isozymes. However, recent work showed that the stability of type-3 ACS is negatively regulated by ubiquitin-dependent turnover mediated by XBAT32, a RING-type E3 ligase (Lyzenga et al., 2012). Moreover, a putative N-terminal degron of ACS7 is active only in light-grown plants (Xiong et al., 2014) and is poorly conserved (Booker and DeLong, 2015). This lightdependent regulation of ACS7 stability may be similar to the turnover regulation of type-2 ACS, allowing the fine-tuning of ethylene production to impose transient growth control under changing conditions. Considering the regulatory role of the N-terminal domain in ACS7, it may be important to revisit the question of N-terminal motifs that could be involved in regulating the stability of other ACS proteins in response to various stimuli, including light.

The post-translational modifications of ACO have been examined in less detail than those that regulate ACS activity. However, recent work has identified several posttranslational mechanisms for controlling ACO activity, including glutathionylation (Dixon et al., 2005) and sulfhydration of cysteine residues on ACO (Friso and van Wijk, 2015). While the effect of glutathionylation on ACO activity has not been reported, S-sulfhydration of LeACO1 and LeACO2 results in a decrease in ACO activity and ethylene production (Jia et al., 2018), establishing an *in vivo* role for post-translational control of ACO. Determining whether these modifications contribute to light-dependent regulation of ethylene production is an open question for future research.

## MECHANISTIC CONNECTIONS BETWEEN LIGHT RESPONSE AND THE ETHYLENE SIGNALING PATHWAY

#### Ethylene Receptor Function Is Dependent on Light and Developmental Context

The five ethylene receptors in Arabidopsis are not functionally equivalent, with sub-functionalization observed for responses in different tissues and developmental stages (as reviewed by Shakeel et al., 2013). This subfunctionalization was revealed though detailed phenotypic analysis of LOF receptor mutants (Wang et al., 2003; Binder et al., 2004; Binder et al., 2006; Qu et al., 2007; Liu et al., 2010; McDaniel and Binder, 2012; Wilson et al., 2014b; Bakshi et al., 2015; Harkey et al., 2018). This subfunctionalization is likely due to diversity in receptor structure and signaling capabilities (O'Malley et al., 2005; Wang et al., 2006; Shakeel et al., 2013; Bakshi et al., 2015). Like the central signaling mutant *ein2-1*, a GOF *etr1-3* mutant was insensitive to ethylene or ACC in seedlings growth in light or dark (Guzman and Ecker, 1990; Roman et al., 1994; Negi et al., 2008). In an examination of nutation of etiolated hypocotyls, ethylene-dependent nutations were lost in the *etr1-7* LOF mutant no other single receptor LOF mutations affected this process (Binder et al., 2004; Binder et al., 2006). In contrast, the function of EIN4 was light-dependent. In dark-grown seedlings the *ein4-1* receptor GOF mutant showed no ethylene response (Roman et al., 1994). When grown in the light, however, *ein4-1* seedlings show a partial response to ACC (Smalle et al., 1997), suggesting differences in this receptor's role in dark vs. light.

The functional role of the five ethylene receptors has been explored in roots of light-grown Arabidopsis seedlings (Harkey et al., 2018). Transcripts encoding all five ethylene receptors are expressed in roots, and the abundance of transcripts encoding three receptors, *ETR2, ERS1*, and *ERS2*, is increased by treatments that elevated ethylene (Hua et al., 1998; Harkey et al., 2018). The GOF *ETR1* mutant (*etr1-3*) is insensitive to the effects of ethylene on root elongation, LR development, and root hair initiation (Negi et al., 2008; Lewis et al., 2011a). Using null mutants in each of the five receptors, the major role of ETR1 in controlling root responses to ACC was reported, with subtle changes in development in null mutants in any of the other receptors (Harkey et al., 2018). Using multiple LOF mutants in two or three receptor genes, minor and redundant roles for *ETR2* and *EIN4* were identified, especially in root hair formation. A triple mutant carrying *etr1-6, etr2-3*, and *ein4-4* LOF mutations has short roots, with no LRs and with extreme proliferation of root hairs. All three phenotypes are largely complemented with a genomic copy of ETR1 (Harkey et al., 2018). These results argue that the ETR1 receptor has a predominant role in controlling ethylene-inhibited LR formation, and ethylene-stimulated root hair initiation in light-grown roots, similar to the major role of this receptor in controlling nutations and responses to silver ions (Shakeel et al., 2013). Two specific receptors regulate the size of the root apical meristem, however (Street et al., 2015). In contrast with findings in LRs and root hairs, LOF *etr1-9* or *ers1-3* single mutants showed wild-type meristem size, but the LOF *etr1-9*  *ers1-3* double mutant exhibited a substantially reduced root apical meristem size, similar to that found in the *ctr1-2* mutant, consistent with multiple receptors controlling this aspect of root development (Street et al., 2015).

The role of specific ethylene receptors in root elongation in dark-grown seedlings has also been reported. Images of dominant GOF mutants in *ETR1*, *ERS1*, *ERS2*, and *EIN4* show roots that appear to be ethylene-insensitive (Hua et al., 1995; Hua et al., 1998). Responses to added ACC were quantified for several *etr1* and *ers1* mutant alleles, which showed reduced sensitivity (Hua et al., 1995). In comparison, the GOF *etr2-1* mutant appears to have an intermediate phenotype, with roots shorter in ethylene than in air, but not as short as wild-type roots in ethylene (Sakai et al., 1998). One study observed that subfamily 2 receptors (*ETR2*, *ERS2*, and *EIN4*) are not required for ethylene root response, as the *etr1-9 ers1-3* double mutant which carries strong LOF alleles has constitutive ethylene signaling, suggesting that the remaining receptors were not sufficient to repress ethylene signaling (Hall et al., 2012). Additionally, complementation with a wild-type copy of *ETR1* was adequate to restore ethylene sensitivity (Hall et al., 2012). Another group assayed phenotypes of receptor mutants in both light and dark (Adams and Turner, 2010), but in the absence of sucrose, which is also known to influence ethylene response (Zhou et al., 1998; Gibson et al., 2001; Yanagisawa et al., 2003; Haydon et al., 2017). Root length in the GOF *etr1-1* mutant was unchanged in response to ethylene under conditions of continuous darkness, but not continuous light. Some differences in the responses of other receptor LOF mutants were observed in dark- versus light-grown seedlings, but all receptors were at least partially required under both conditions (Adams and Turner, 2010). Together, these results demonstrate that ethylene receptors in Arabidopsis have distinct functions, dependent on tissue and light context.

#### The EIN3 and EBFs Mediate Light-Dependent Transcriptional Responses to Ethylene

The EIN3 TF is an essential mediator of ethylene response in hypocotyls of dark-grown seedlings, but its role is more complex in light-grown seedlings. The *ein3-1* mutant has ethylene-insensitive hypocotyl elongation in either light- or dark-grown hypocotyls (Chao et al., 1997; Smalle et al., 1997), suggesting that elongation responses to ethylene in the hypocotyl require EIN3 in a light-independent manner. EIN3 also regulates chlorophyll biosynthesis during the dark-to-light transition (Liu et al., 2017). However, the function of EIN3 in roots is light-dependent. In roots of dark-grown seedlings, double mutants between *ein3-1* and either *eil1-1* or *eil1-2* show no response to added ACC, while single mutants in *ein3* and *eil1* show partial response to this treatment (Alonso et al., 2003). In contrast, in roots of light-grown seedlings, *ein3-1*, *eil1-1*, and the double mutant all exhibit ACC-inhibition of root elongation and LR formation, and ACC stimulation of root hair formation (Harkey et al., 2018). A subset of ethylene-responsive transcripts from light-grown roots were identified as binding targets of EIN3 (Harkey et al., 2018) as reported by a DAP-Seq dataset (O'Malley et al., 2016), but many other transcripts were not direct EIN3 targets. These results are consistent with EIN3 and EIL1 controlling only a subset of ethylene responses in roots of light-grown seedlings. One example where there is lightdependent function of EIN3 is in regulation of *ACO2* transcript abundance. Upregulation of *ACO2* after ethylene treatment was lost in dark-grown *ein3-1* mutant seedlings (and EIN3 has been shown to bind to *ACO2 via* ChIP-Seq) (Chang et al., 2013). In contrast, in light-grown plants, that upregulation was present in the *ein3-1* single mutant, but was lost in both *ein3-1 eil1-1* and *ein3-1 eil1-2* double mutants (Lee et al., 2006), suggesting that EIL1 can compensate for EIN3 in regulating *ACO2* only in lightgrown plants.

Recent results have suggested that differences in EIN3 function in the light and dark may be controlled at the level of turnover of this protein. Although *EIN3* transcript accumulation is not regulated by ethylene (Chao et al., 1997; Harkey et al., 2018), EIN3 and EIL1 protein accumulation is tightly controlled *via* ethyleneregulated turnover. In the absence of ethylene, EIN3 and EIL1 are ubiquitinated by EIN3-BINDING F-BOX PROTEIN1 and 2 (EBF1 and 2), two F-box proteins that act in SCF complexes, leading to EIN3 degradation. When ethylene levels rise, EBF1 and 2 are targeted for degradation in an EIN2-dependent manner, stabilizing EIN3 (Guo and Ecker, 2003; Gagne et al., 2004; Binder et al., 2007; An et al., 2010). EIN3 and EIL1 protein turnover is also regulated by crosstalk with light signaling *via*  cryptochromes and HY5. The stimulation of hypocotyl elongation by ethylene in light-grown plants requires CRY1 or CRY2 (Vandenbussche et al., 2007), as well as HY5 (Yu et al., 2013). In darkness, CONSTITUTIVE PHOTOMORPHOGENESIS 1 (COP1), an integrator of light signaling, targets EBF1/2 and HY5 for ubiquitination and degradation, allowing EIN3 accumulation (Shi et al., 2016a), and preventing HY5-mediated inhibition of hypocotyl elongation. Movement of COP1 from the nucleus to the cytoplasm in light conditions allows HY5 to accumulate and inhibit growth. If ethylene signaling is activated in light conditions, EIN3 antagonizes HY5 and stimulates elongation by promoting nuclear localization of COP1, leading to HY5 degradation (Yu et al., 2013). The red light receptor PhyB also directly interacts with EIN3 and EBF1/2 after exposure to red light and enhances degradation of EIN3 (Shi et al., 2016b).

EIN3 regulation of *PIF3* and *ERF1*, which have antagonistic roles in regulating growth, constitutes one of the primary mechanisms driving the inverse hypocotyl responses to ethylene in light *versus* dark (Zhong et al., 2012). Both *PIF3* and *ERF1* are direct transcriptional targets of EIN3 (Chang et al., 2013). ERFs are stabilized by light, and they generally inhibit growth. EIN3 upregulates *ERF1* both in darkness and in light, but ERF1 effects on hypocotyl growth are only measurable under darkness, where other ERFs are absent. Conversely, *pif3* mutants are insensitive to ethylene-induced hypocotyl elongation in light, but not to hypocotyl inhibition in the dark (Zhong et al., 2012). PIFs generally promote elongation, and are destabilized in light, contributing to reduced elongation in light-grown seedlings. Transcriptional regulation of *PIF3* by ethylene *via* EIN3 is inconsequential in darkness, where many other PIFs are also active, but becomes significant under light, where other PIFs are degraded, and *PIF3* activation leads to increased hypocotyl growth. EBF1/2 also mediate red light-dependent degradation of PIF3 (Dong et al., 2017). EBFs can synergistically reduce PIF3 levels both directly, by promoting PIF3 degradation, and indirectly, by targeting EIN3 for degradation and thus reducing *PIF3* mRNA. This modulation of EIN3 and its targets by light enables complex responses to ethylene under different light contexts, such as opposite response in hypocotyl elongation. As discussed above light-dependent ethylene synthesis may also contribute to PIF3 stabilization and amplification of ethylene responses.

Downstream transcriptional effects of EIN3 and light signaling pathways cannot be completely disentangled. Recent work revealed that an *ein3 eil1* double mutant retains shade response, although ethylene-stimulated hypocotyl elongation is abolished (Das et al., 2016), suggesting that shade does not induce hypocotyl elongation by acting directly through the EIN3/EIL1 response pathway. The similar growth effects of ethylene and light are accompanied by many common transcriptional responses (Das et al., 2016). The COP1 effects on EIN3 targets are also complex. COP1 has been shown to increase EIN3 protein levels by targeting EBF1/2 for degradation in the dark (Shi et al., 2016a). In the light, ACC treatment and EIN3 overexpression lead to increased transcript levels of growth-promoting genes such as *YUCCA1* and *5*. This effect is lost in the dark but is restored in the *cop1-4* null mutant (Liang et al., 2012). This suggests that COP1 works by some mechanism downstream of EIN3 to fine tune expression of these particular genes so that they promote elongation in the light, but not in the dark. EIN3 and PIF1 transcriptionally regulate many of the same gene targets independently from one another, but mostly in the same direction (Jeong et al., 2016), and EIN3 and PIF1 pathways are each sufficient to maintain skotomorphogenesis (Shi et al., 2018). Overlapping transcriptional responses are also involved in EIN3/EIL1- and PIF3-mediated regulation of hypocotyl hook opening (Zhang et al., 2018). Downstream transcription factors, such as ERF72, may also have activity that is modulated by light to influence developmental responses (Liu et al., 2018). As described above, differential regulation of specific proteins, such as HY5, contributes to the opposing ethylene effects observed in light and dark (Smalle et al., 1997).

### DOWNSTREAM ETHYLENE TRANSCRIPTIONAL EFFECTS ARE INFLUENCED BY LIGHT

A number of ethylene transcriptome studies have been performed with plants grown under a range of light conditions, revealing distinct transcriptional networks downstream of ethylene perception. We previously compared a dataset from dark-grown seedlings treated with ethylene (Chang et al., 2013) with another dataset from lightgrown roots treated with ACC (Harkey et al., 2018). Both datasets used similar time points across a 24-h period after treatment, and we used the same statistical analysis of both datasets. However, we found limited overlap in differentially expressed (DE) genes (71 common genes out of 449 in the light-grown root dataset and out of 971 in the dark-grown seedling dataset). In principle, these changes could be explained by differences in light condition, tissue type, and/or method of elevating ethylene levels (ACC treatment vs. ethylene gas). This last possibility seems unlikely because all ACC responses were lost in the ethylene-insensitive *etr1-3* and *ein2-5* mutants (Harkey et al., 2018). Comparing a larger number of transcriptomic data sets is essential for more complete understanding of the light-dependent effects of ethylene on transcript accumulation.

To identify transcriptional responses to ethylene that are lightand tissue-specific, we looked for datasets that were suitable for a meta-analysis that could resolve differences and similarities in ethylene-responsive transcriptomes in the light and dark. We searched the Gene Expression Omnibus (GEO) for the term "ethylene." Twenty-five datasets were identified in the original search based on treatment with ACC, ethylene, or with compounds that block ethylene synthesis (such as AVG), and/ or mutations or transgenes that alter ethylene production or response. Many of these datasets were not usable because of dissimilar approaches or incomplete information. Five datasets were excluded due to insufficient information on experimental methods; another five used specific mutants or transgenic lines that were not found in any other dataset and did not include wild-type seedlings treated with ACC or ethylene. Although there were many datasets utilizing Col-0 and/or *ein2*, *ein3*, and *eil1* mutants in light and dark conditions, they used experimental methods, tissue types, or plants that were not developmentally matched. Seven additional datasets used 3- or 4-day-old whole dark-grown seedlings, while the remaining five datasets came from light-grown material using a variety of ages and tissue types. This highlights the need for future work that directly compares ethylene effects in light versus dark.

Ultimately, we identified three datasets with highly similar experimental methods and plant age in which transcript abundance was quantified after 4 h of ethylene or ACC treatment in roots (Stepanova et al., 2007; Feng et al., 2017; Harkey et al., 2018), and a fourth that provided an interesting comparison between ethylene treatment and shade treatment in hypocotyls or in cotyledons (Das et al., 2016). The most relevant differences between the three root datasets can be found in **Figure 3**, and further details on the process of identifying these datasets can be found in **Supplemental Datasheet 1**, along with a description of the experimental conditions used in each study. The fourth dataset was of particular interest because the authors compared the transcriptional effects of shade and ethylene in experimental conditions that were otherwise identical (Das et al., 2016). The authors noted that the effect of combined shade and ethylene on hypocotyl elongation was intermediate between the two individual treatments, consistent with ethylene and light signaling pathways sharing downstream signaling and/or effector components. However, samples treated with both ethylene and shade were not included in the transcriptomic analysis. Among genes that responded to ethylene and shade consistently and with the same direction of change, the authors found enrichments for annotations including hormone signaling, cell wall, and photomorphogenesis, among others, as well as two TFs, *AtHB28* and *IBL1*. Analysis of mutant and overexpression lines showed that *AtHB28* and *IBL1* are important for both shade and ethylene response (Das et al., 2016).

We developed a statistical pipeline to apply to all datasets used in our analysis to avoid discrepancies that might arise from differences in data analysis methods. We generated lists of DE genes that could more properly be compared to one another. (Note that this re-analysis results in DE lists that differ from those derived in the original publications.) For the three root datasets, we combined expression data from all three experiments into one master dataframe; both this dataframe and the Das et al. dataset were analyzed for differential expression using *limma* and other packages in R (Davis and Meltzer, 2007; R Core Team, 2014; Ritchie et al., 2015; Gu et al., 2016). Additional details of these analyses can be found in **Supplemental Datasheet 1**.

To identify the entire overlap between ethylene and shade transcriptional responses in the Das et al. (2016) dataset, we used this data analysis pipeline. First, we identified the complete set of ethylene-responsive genes, and then queried their expression responses in the shade dataset. Compared to cotyledons, hypocotyls showed a greater response to ethylene, which is expected given the changes in hypocotyl growth that occur in etiolated seedlings treated with ethylene, described above, so we focused on that tissue type. Not surprisingly, of the 7,248 hypocotyl transcripts that showed a significant response to ethylene, more than half of those genes also showed a shade response (4,239; **Figure 2A**). The majority of these gene expression changes occurred in the same direction and with similar kinetics. Full results for all ethylene-responsive transcripts can be found in **Supplemental Datasheet 2**.

To better illustrate the relationship between ethylene and shade response, we plotted the log2 fold-changes in transcripts in response to ethylene against the fold-change in response to transition to shade (using the 25.5-h time point, which showed the most striking changes from the control) using the previously published transcript abundance values from Das et al. (2016). This graph highlights the strong correlation between ethylene response and shade response (**Figure 2B**). The correlation between the magnitude of change in response to ethylene and shade is statistically significant both for genes with the same direction of response (Pearson's correlation, r = 0.89, p < 0.001) and in genes with the opposite direction of response (Pearson's correlation, r = −0.87, p < 0.001). Dark- or shade-grown plants exhibit a different transcriptional landscape than their lightgrown counterparts. Our analysis illustrates that many transcripts show similar responses to ethylene and shade; thus, studies that use dark-grown tissues to examine ethylene response will likely miss changes that occur only in light-grown plants.

We performed a meta-analysis using the three root-specific ethylene-response datasets identified as sufficiently matched for comparison (Stepanova et al., 2007; Feng et al., 2017; Harkey et al., 2018) to screen for light-dependent and light-independent changes in ethylene-regulated transcript abundance. We used our new pipeline to reanalyze the root-specific transcriptomes to identify differences that are linked to the light environment of seedling growth. This analysis yielded interesting patterns of light-dependent and light-independent changes in transcript abundance that are summarized in a Venn Diagram in **Figure 3**. A list of all transcripts that showed significant responses to ethylene or ACC in at least one dataset and their magnitude of change

can be found in **Supplemental Datasheet 2**. As expected, many more DE genes were identified in the RNA-seq dataset (Feng et al., 2017) than in the microarray-based datasets (Stepanova et al., 2007; Harkey et al., 2018), because RNA-Seq has a greater dynamic range. Although only 3% of the DE genes identified responded to ethylene in all three datasets, nearly a third (32%) were DE in two datasets. A number of genes were DE in the two datasets from light-grown seedlings (Feng et al., 2017; Harkey et al., 2018), but

not in the dark (Stepanova et al., 2007), suggesting light-dependent regulation by ethylene. There was also substantial overlap (433 transcripts) between the two datasets that used ethylene treatment but differed in the presence of light during growth. We identified 169 transcripts in the overlap between the dark-grown ethylene dataset (Stepanova et al., 2007) and light-grown ACC dataset (Harkey et al., 2018). This number is greater than in our previously reported comparison of these two datasets (80, transcripts; Harkey et al., 2018), due to the common filtering used for both datasets in this meta-analysis. A surprising number of genes, however, were specifically regulated in one dataset, and not in the other two, despite the similarity of experimental techniques. These differences may be related to other conditions such as plant age (3, 5, or 6 days), light cycle (continuous light vs. 16 h light 8 h dark), or differences in media (e.g., sucrose concentration, which is also known to influence ethylene response; Gibson et al., 2001; Haydon et al., 2017; Yanagisawa et al., 2003). These results demonstrate the need for direct comparisons of ethylene effects under experimental conditions that vary only by light level.

In addition to the light-specific transcripts described above, this analysis identified a core set of 143 transcripts that responded to ethylene or ACC in all three datasets, regardless of light. Of these transcripts, 139 (97%) changed in the same direction in all treatments (**Figure 3**). This set of 139 genes with consistent direction of change should be considered the "gold standard," for root ethylene response, much like a previously identified set of cytokinin-responsive genes from another meta-analysis (Bhargava et al., 2013). The full list of ethylene- or ACCresponsive genes from any dataset can be found in **Supplemental Datasheet 2** , with "gold standard" genes indicated.

A subset of the "gold standard" genes is summarized in **Table 1**. This group of 44 genes was chosen based on three criteria: the largest logFC values (in the positive or negative direction), known roles in ethylene synthesis or signaling (highlighted in red in Table 1), and/or known EIN3 targets based on DAP-Seq (O'Malley et al., 2016) and/or CHiP-Seq (Chang et al., 2013) analysis. Interestingly, most of the upregulated "gold" genes were identified as EIN3 targets by at least one method (72.4%), but very few downregulated "gold" genes were bound by EIN3 (6.3%). "Gold standard" genes also included a number of auxin-related genes (e.g., *SAUR76*, *SAUR8*, *IAA2*, and *IAA4/AUX2-11*), and genes involved in cell wall regulation (e.g., a pectin methylesterase inhibitor). Not surprisingly, the 139 transcripts were also enriched in gene annotations for cellular response to ethylene stimulus and negative regulation of the ethylene pathway.

Within this group of 139 transcripts, we identified 13 core genes in ethylene signaling or synthesis whose levels increased in all three datasets (and in Das et al., 2016). This core gene set includes genes encoding TFs that participate in ethylene signaling (for example, EDF1, EDF3*,* EDF4, and several ERFs), negative regulators of the signaling pathway CTR1, EBF2 and ARGOS, and the ethylene receptors ETR2, ERS1, and ERS2. Thus, a core output of the ethylene response is upregulation of its own signaling pathway TABLE 1 | Selected gold standard transcripts regulated in all three datasets. The transcripts in red are all implicated in ethylene signaling or synthesis.


components including both positive and negative regulators of ethylene responses. The core set also includes transcripts encoding ethylene biosynthetic proteins. There is consistent upregulation of transcripts encoding the ACO enzymes, with *ACO1* and *ACO2*  upregulated in all three datasets and *ACO3*, *ACO4*, and *ACO5* upregulated in two of the three datasets. *ACO2* was also upregulated by ethylene, although down-regulated in shade in Das et al. (2016). Interestingly, *ACS* transcript levels show less consistent positive regulation, showing no changes for any *ACS* gene in two datasets (Stepanova et al., 2007; Harkey et al., 2018) and changes in only two to four *ACS* transcripts (out of 11 family members) in two other data sets (Das et al., 2016; Feng et al., 2017). These results indicate that a positive feedback loop drives ethylene synthesis *via*  upregulation of *ACO* expression, while *ACS* mRNA levels appear to be subject to a more complex control network, as discussed above (see *Light-Mediated Transcriptional Regulation of ACS and ACO*).

Finally, included in this comparison is an annotation of genes that are regulated by ethylene in dark-grown whole seedlings as detected by RNA-Seq (Chang et al., 2013) (as found in a separate column in the **Supplemental Datasheet 2**). Of the 77 up-regulated genes in the gold-standard list, 40 were also found to be sites of EIN3 binding while only 2 of the 62 down-regulated genes showed ethylene-regulated expression. Therefore, one can further refine these genes into root-specific and tissue-independent transcripts, using the detailed annotations in **Supplemental Datasheet 2**. Together, this meta-analysis reveals many candidate genes for conserved ethylene responses that are also induced by the ethylene precursor, ACC, and transcripts whose responses depend on light or tissue type. This information can allow formulation of a wealth of hypotheses that can be tested to further refine our understanding of ethylene signaling across plant development.

### CONCLUSIONS

As seedlings germinate, elongate through soil, and then emerge into light, they undergo profound changes in development. The importance of ethylene levels in controlling development is best understood in the early dark phases, but new studies that examine the role of ethylene during developmental transitions from dark to light or in light-dependent development are providing new insight into the functions of ethylene during seedling development. Recent studies have revealed novel mechanisms that modulate ethylene biosynthesis, including important transcriptional and posttranslational regulatory strategies that control production of this hormone. The pathways that control ethylene response include central signaling proteins that function in ethylene response under all conditions, but also receptors and transcription factors with light- and developmental stage-specific functions. Comparison of genome-wide transcriptional datasets allows identification of candidate genes that contribute to all ethylene responses and other genes that may contribute to developmental outputs that are specific to the light environment. Together, light regulation of ethylene biosynthesis, signaling, and developmental response have far-reaching effects on a plant's ability to adapt to the environment in early stages of development and throughout the life cycle. Understanding the mechanisms by which light and ethylene interact at the molecular and organismal levels is an important goal of future research.

### REFERENCES


### AUTHOR CONTRIBUTIONS

AH performed the meta-analysis, drafted text, prepared figures, and edited the manuscript; GY drafted text, prepared figures, and edited the manuscript; DS prepared figures; AD and GM drafted text and edited the manuscript.

### FUNDING

This work was supported by grants from the US National Science Foundation to GM (MCB-1716279) and GY (MCB-1817286).

### ACKNOWLEDGMENTS

We appreciate the assistance of Joëlle Mühlemann with the metaanalysis and the helpful comments on the manuscript from Emily Martin.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01094/ full#supplementary-material

SUPPLEMENTAL DATASHEET 1 | Extended methods for meta-analysis of ethylene-regulated transcripts and ethylene- and shade-regulated transcripts

SUPPLEMENTAL DATASHEET 2 | Summary of differentially expressed transcripts from roots in response to ethylene or ACC and overlap of shade-and ethylene-regulated transcriptomes


encodes a member of the Raf family of protein kinases. *Cell* 72, 427–441. doi: 10.1016/0092-8674(93)90119-B


*of Ethylene Production in Lower and Higher Plants: proceedings of a Conference held at the Limburgs Universitair Centrum, Diepenbeek, Belgium, 22–27 August 1988 Advances in Agricultural Biotechnology*. Eds. H. Clijsters, M. De Proft, R. Marcelle, and M. Van Poucke (Dordrecht: Springer Netherlands), 279–290. doi: 10.1007/978-94-009-1271-7\_32


1-aminocyclopropane-1-carboxylate synthases. *BMC Plant Biol.* 5, 14. doi: 10.1186/1471-2229-5-14


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Harkey, Yoon, Seo, DeLong and Muday. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Ethylene Signaling Is Required for Fully Functional Tension Wood in Hybrid Aspen

*Carolin Seyfferth1, Bernard A. Wessels 1, András Gorzsás 2, Jonathan W. Love3, Markus Rüggeberg4,5, Nicolas Delhomme6, Thomas Vain7, Kamil Antos 8, Hannele Tuominen1, Björn Sundberg6,9 and Judith Felten6\**

*1 Umeå Plant Science Centre, Department of Plant Physiology, Umeå University, Umeå, Sweden, 2 Department of Chemistry, Umeå University, Umeå, Sweden, 3 Arevo AB, Umeå, Sweden, 4 Institute for Building Materials, Swiss Federal Institute of Technology Zurich (ETH Zurich), Zurich, Switzerland, 5 Laboratory of Wood Materials, Swiss Federal Laboratories of Materials Science and Technology, Dubendorf, Switzerland, 6 Umeå Plant Science Centre, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, Umeå, Sweden, 7 DIADE, Univ Montpellier, IRD, Montpellier, France, 8 Department of Integrative Medical Biology, Umeå University, Umeå, Sweden, 9 Stora Enso AB, Nacka, Sweden*

#### *Edited by:*

*Dominique Van Der Straeten, Ghent University, Belgium*

#### *Reviewed by:*

*Deqiang Zhang, Beijing Forestry University, China Antonio Ferrante, University of Milan, Italy*

> *\*Correspondence: Judith Felten Judith.felten@slu.se*

*Specialty section:*

*This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science*

*Received: 24 April 2019 Accepted: 12 August 2019 Published: 26 September 2019*

#### *Citation:*

*Seyfferth C, Wessels BA, Gorzsás A, Love JW, Rüggeberg M, Delhomme N, Vain T, Antos K, Tuominen H, Sundberg B and Felten J (2019) Ethylene Signaling Is Required for Fully Functional Tension Wood in Hybrid Aspen. Front. Plant Sci. 10:1101. doi: 10.3389/fpls.2019.01101*

Tension wood (TW) in hybrid aspen trees forms on the upper side of displaced stems to generate a strain that leads to uplifting of the stem. TW is characterized by increased cambial growth, reduced vessel frequency and diameter, and the presence of gelatinous, cellulose-rich (G-)fibers with its microfibrils oriented parallel to the fiber cell axis. Knowledge remains limited about the molecular regulators required for the development of this special xylem tissue with its characteristic morphological, anatomical, and chemical features. In this study, we use transgenic, ethylene-insensitive (ETI) hybrid aspen trees together with time-lapse imaging to show that functional ethylene signaling is required for full uplifting of inclined stems. X-ray diffraction and Raman microspectroscopy of TW in ETI trees indicate that, although G-fibers form, the cellulose microfibril angle in the G-fiber S-layer is decreased, and the chemical composition of S- and G-layers is altered than in wild-type TW. The characteristic asymmetric growth and reduction of vessel density is suppressed during TW formation in ETI trees. A genome-wide transcriptome profiling reveals ethylene-dependent genes in TW, related to cell division, cell wall composition, vessel differentiation, microtubule orientation, and hormone crosstalk. Our results demonstrate that ethylene regulates transcriptional responses related to the amount of G-fiber formation and their properties (chemistry and cellulose microfibril angle) during TW formation. The quantitative and qualitative changes in G-fibers are likely to contribute to uplifting of stems that are displaced from their original position.

Keywords: xylem, wood, ethylene, tension wood, lignin, microfibril angle, Raman microspectroscopy, transcriptomics

### INTRODUCTION

When angiosperm tree stems are displaced from their original growing position, due to external factors including wind, snow load, and/or growth on uneven terrain, they form tension wood (TW) in order to reorient/uplift the stem towards its original growth position (Felten and Sundberg, 2013). TW formation is characterized by asymmetric xylem growth in the stem, originating from enhanced activity of the secondary cambium at the upper (TW) side of the stem as compared with the lower [opposite wood (OW)] side (Timell, 1986). TW is characterized by an increased fiber-to-vessel ratio compared with OW (Esau, 1977). Another typical feature of TW is the presence of an altered cell wall layer structure in fiber cells. In many tree species, TW fibers [called gelatinous (G)-fibers] are characterized by the presence of a tertiary cell wall layer, mainly composed of a gelatinous, celluloserich layer (G-layer) with microfibrils having high crystallinity (Müller et al., 2006). Typically, a bimodal distribution of the cellulose microfibril angle (MFA) is observed in TW fibers using X-ray diffraction. The S2-layer exhibits an MFA of 20–40°, which is considerably larger than the values reported for the S2 of normal wood (NW) (5–15°), whereas the G-layer shows a very small MFA of 0–5° (Müller et al., 2006; Goswami et al., 2008; Rüggeberg et al., 2013). The highly porous structure of the G-layer (Norberg and Meier, 1966; Gierlinger and Schwanninger, 2006; Clair et al., 2008; Chang et al., 2009) allows for water incorporation resulting in a gelatinous aspect, from which the name gelatinous- or G-layer originates. The G-layer with its low MFA together with the S2 layer with its high MFA is thought of as a crucial factor to establish the tensile force that leads to the uplifting of the tree in response to the initial displacement (Goswami et al., 2008; Mellerowicz and Gorshkova, 2012). Even though the G-layer formation is a common phenomenon observed in the TW of many tree species, numerous tree species form TW without typical G-fibers, but rather with other cell wall layer modifications that all have small MFAs in common, which similarly exerts a tensile strain that leads to re-orientation of the stem (summarized in Ruelle, 2014).

In aspen, the plant hormone ethylene (ET) has been demonstrated to influence many of the features that characterize TW formation (Love et al., 2009; Felten et al., 2018). Indeed, after stem displacement, ET biosynthesis increases on the TW side in parallel with an asymmetric induction of 1-aminocyclopropane-1-carboxylate oxidase (ACO), which converts the ET precursor 1-aminocyclopropane-1-carboxylic acid (ACC) to ET (Andersson-Gunnerås et al., 2003). Furthermore, exogenous application of ET or ACC could mimic all typical TW characteristics such as enhanced cambial growth, increased fiberto-vessel ratio and induction of the G-layer formation (Felten et al., 2018). Finally, the role of endogenous ET signaling in the unilateral growth response induced by leaning was investigated by constructing ET-insensitive (ETI) hybrid aspen trees. This study demonstrated an overall reduction in the TW/OW ratio in the ETI trees compared with wild-type trees (Love et al., 2009). Whether this reduced asymmetry affected the functionality of the TW was, however, not shown. Also, whether endogenous ET induced by leaning mediates the secondary wall modification in fibers observed in TW remains to be evaluated.

TW-forming tissues exhibit a strongly modified transcriptome (Paux et al., 2005; Andersson-Gunnerås et al., 2006; Chen et al., 2015; Zinkgraf et al., 2018), and it seems plausible that this is at least partly regulated through ET signaling. ET is perceived by the ET receptors localized at the endoplasmic reticulum (Bleecker et al., 1988; Cancel and Larsen, 2002). Ethylene binding derepresses the downstream ET signaling cascade (Merchante et al., 2013; Xu and Zhang, 2015). This leads to phosphorylation of the downstream component ETHYLENE-INSENSITIVE 2 (EIN2) and the cleavage of the EIN2 C-terminus (EIN2Cend) that translocates into the nucleus. The translocation triggers the expression of the ETHYLENE-INSENSITIVE 3 (EIN3)/EIN3- LIKE 1 (EIL1) transcription factors (TFs) and stabilizes EIN3/ EIL1 proteins by preventing their proteasomal degradation (Qiao et al., 2009; Wen et al., 2012; Li et al., 2015; Merchante et al., 2015; Zhang et al., 2016). It has long been thought that EIN3/EIL1 activates the expression of ETHYLENE RESPONSE FACTORS (ERFs), which then trigger expression of their target genes through binding to GCC-boxes in the promoter region (Solano et al., 1998; Zhang et al., 2011; Shi et al., 2012). Recently, however, EIN3/EIL1 has been shown to directly bind to the promoter of many targets (beyond ERFs) and to modulate their expression (Chang et al., 2013). Furthermore, the ET downstream genes that were activated in hybrid aspen by exogenously applied ACC more frequently contained the EIN3/EIL1-binding motif (TEIL motif) in their promoter rather than the GCC-box (ERF-binding motif), and only a few ACC-activated ERFs contained the TEIL motif in their promoter (Felten et al., 2018). Therefore, it is likely that both ERFs and EIN3/EIL1 connect the ET signaling pathway to the molecular program of (tension) wood formation. In NW, we have identified *in silico* that ERFs and EIN3/EIL1 members act as hubs for regulating genes involved in wood formation (Seyfferth et al., 2018). Also, certain ERFs are transcriptionally induced in TW (Vahala et al., 2013).

We hypothesize here i) that ET signaling is required for all aspects of TW formation (cambial activity, vessel differentiation, G-layer formation, and MFA orientation) and that TW in ETI trees lacks these typical TW features and consequently the capacity to upright an inclined stem, and ii) that there is a molecular connection on transcriptional level between ET signaling and the molecular pathways regulating the typical developmental modifications in TW. To test these hypotheses, we conducted leaning experiments with wild-type and ETI hybrid aspen (*Populus tremula* L. × *Populus tremuloides* Michx) trees (Love et al., 2009), and we investigated how ET insensitivity affected TW formation, the uplifting response, and transcriptome changes in TW-forming tissues.

## MATERIALS AND METHODS

### Biological Material and Growth Conditions

The genetic background of the trees used in this study was hybrid aspen (*P. tremula* L. × *P. tremuloides* Michx.) clone T89. The strongest ETI genotype, according to previous experiments (Love et al., 2009; Felten et al., 2018), *pLMX5::etr1-1*, was chosen (line 6E) for all experiments. Transgenic and wild-type trees were transferred to soil after 4 weeks of *in vitro* propagation (Nilsson et al., 1992). For TW experiments (note: conditions specific for uplifting experiments are different and stated below), trees were propagated *in vitro* (for details see Felten et al., 2018) and at about 8- to 10-cm height transferred to the greenhouse and grown in a mixture of commercial soil–sand–fertilizer mixture (Yrkes Plantjord, Kronmull, Weibulls Horto, Sweden) under 18-h photoperiod at 20/15°C (light/dark). In total, 25 trees per genotype were grown in the greenhouse. Trees were fertilized with about 150 ml of 1% Rika-S (N/P/K 7:1:5, Weibulls Horto) once a week. Trees were grown to about 0.8- to 1-m height with weekly rotation to minimize positional effects before 10 trees per genotype were inclined to an angle of 45° against a table and bound to that table approximately 10 cm below the tip of the stem for TW formation or kept upright for NW controls. After 22 days of inclination, stems were harvested between 10 and 50 cm above soil level; split into three parts for RNA extraction and transcriptomics, MFA measurements, and Fourier transform infrared (FT-IR) and Raman microspectroscopy; frozen in liquid nitrogen; and stored at −80°C until processing.

#### Uplifting Experiments

After being outplanted into the greenhouse, wild-type and ETI trees were grown upright for 6 weeks. Then trees were either kept upright to continue forming NW, or the pots were horizontally inclined (90°) on the surface of a table to produce TW (see Supplementary Video S1). The uplifting response of the inclined trees was assayed over 4 weeks. All trees were grown in controlled greenhouse conditions with an 18/6-h photoperiod under light-emitting diode (LED lamps) (Fiona Lighting FL300 Sunlight), with an average day temperature ranging between 24°C and 25°C and average night temperature between 15°C and 17°C. Relative humidity ranged between 50% and 60%. Trees were fertilized once per week with Rika-S (Weibulls Horto, Hammenhög, Sweden), starting at the third week after planting and ending at the week before the experiments began in the case of the horizontally inclined trees, in order to reduce the additive effects on TW formation due to high nitrogen fertilization (Pitre et al., 2010). Measurements of height and diameter (twice, perpendicular at 10 cm above soil level) were taken weekly for upright grown trees, but only before inclination and after the end of the uplifting phase for the horizontally inclined trees. The stem spanning 5–12 cm above soil level was sampled in two parts. The lower half was frozen in liquid nitrogen and stored at −80°C, and the upper half was used directly for sectioning. Experiments were carried out on six biological replicates for each treatment and genotype. Time-lapse movies of the uplifting response were produced from photographs taken using a custom Raspberry Pi (Rpi) setup. Three cameras were connected to each Rpi, forming one unit. Units were clamped to the sides of opposing tables, and each camera was positioned to photograph two plants undergoing the gravibending response on the opposite tables. Using this setup, we captured 12 plants placed horizontally in each experiment. The Rpi was programmed to take photographs every 10 min the first day and thereafter once per hour for the remainder of the experiment. Quantification of stem lift and curvature was analyzed as described in Gerttula et al. (2015). In brief, we marked the point of primary bending with pink elastics (see in **Supplementary Video S1**). This point is defined as "the basal point of tropism of the primary, herbaceous portion of the stem" (Gerttula et al., 2015) and is clearly visible after a few hours of inclination. Thereafter, using the first picture taken each day (*n* = 28), we traced the stem from the base to the marked position (see in **Supplementary Video S1**). These stem traces were digitized in the form of *XY* coordinates in ImageJ (ImageJ, version 1.51j8 USA) and used to quantify the uplifting response. Normalized degree of lift was determined as the difference in height between the primary bending point and the base of the stem (in terms of their "*Y*" positions), divided by the respective length of the entire stem for each day [(*Y*apex − *Y*base)/(Total Stem Length)]. We calculated the curvature (the summed differences in angle between subsequent pairs of *XY* coordinates or vectors), which indicates the deviation from a straight line in degrees. Scripts used in this experiment are available under https://github. com/UPSCb/UPSCb/tree/master/manuscripts/Seyfferth2019

#### Histology, Microscopy, and Vessel Quantification

For both the upright and horizontally inclined trees, the stem at 10 cm above soil level was used to prepare fresh 70-µm-thick sections (vibrating blade microtome Leica VT1000S, Wetzlar, Germany) stained with Safranin : Alcian Blue (1:2) for 30 s, washed and mounted in 50% glycerol, and imaged with a Zeiss Axioplan2 microscope, AxioCam HRc camera, and AxioVision V 4.8.2 software (Carl Zeiss Light Microscopy, Göttingen, Germany). For TW : OW ratio determination, stereomicroscope images were obtained with a Canon PowerShot G7 digital camera. The distance from the center of the pith to the cambium of both the upper (TW) and lower (OW) sides was used to calculate the TW/OW ratio. Images for vessel quantification were taken with 10× magnification on a Zeiss Axioplan2 microscope. Images just inwards of the cambium, taken for NW, OW, and TW, were analyzed using ImageJ (version 1.51j8 USA).

#### Vibrational Microspectroscopy and Data Analysis

FT-IR microspectroscopy with a single-element detector was performed using existing protocols (Gorzsás and Sundberg, 2014). Transverse sections with thickness of 20 µm were prepared from frozen stem samples using a Microm cryotome HM505E and dried in a desiccator between two polished rectangular BaF2 windows (38.5 mm × 19.5 mm × 4 mm, International Crystal Laboratories, Garfield, NJ, USA) for at least 48 h. One section from each of five biological replicates per genotype and treatment was analyzed. Single-element detector images were recorded over four positions, approximately 90° from one another around the entire cross section. A single spectrum was recorded from each position covering an area of approximately 430 × 430 µm, using a Bruker Equinox 55 spectrometer equipped with a Hyperion 3,000 microscopy accessory (Bruker Optics GmbH, Ettlingen, Germany). Spectra were recorded in transmission mode over the range of 700–3,800 cm−1, with a spectral resolution of 4 cm−1. Six hundred interferograms were co-added to improve signal-tonoise ratio, and a zero-filling factor of 2 was employed. Spectra were converted to data point tables using OPUS (v7.0.122, Bruker Optik GmbH, Ettlingen, Germany), baseline corrected (2-point linear baseline at 812 and 1,809 cm−1), and total sum (area-) normalized in the region of 812–1,809 cm−1, using in-house scripts (https://www.umu.se/en/research/infrastructure/visp/ downloads/) written in Matlab (version 7.0, MathWorks, Natick, MA, USA, http://www.mathworks.com). Spectra were extracted, and the area measuring 816–1,803 cm−1 was used in subsequent multivariate analyses. Principal component analysis (PCA) and orthogonal projections to latent structures discriminant analysis (OPLS-DA) (Trygg and Wold, 2002) were performed using SIMCA-P+ (versions 12-14, Umetrics AB, Umeå, Sweden).

For Raman microspectroscopic analysis, 20-µm-thick transverse sections were prepared from frozen stem material (ca. 30 cm above soil level) using a Microm cryotome HM505E and kept between a standard glass microscope slide and cover slip in deuterated water (D2O). Spectral maps were recorded using a Renishaw inVia Raman spectrometer and microscope with a 100× oil immersion lens and a 514-nm Ar+ laser. Spectral maps comprising of 60–80 voxels with step sizes of 1 µm in the *X* and *Y* directions were recorded on two positions from each one to two cross sections of TW and NW, of each of three wild-type and three *pLMX5::etr1-1* trees that were grown upright or had been leaned at 45° and fixed in this position for 22 days. The maps were recorded in radial orientation over three fiber cells from lumen of cell 1 to lumen of cell 3. The following settings were used for spectral recording: static scans centered at 1,190 cm−1 (resulting in a spectral range of ca. 510 to 1,802 and ca. 1 cm−1 resolution with a 2,400 lines grating), standard confocality, laser power set at 100% in the software, and exposure time of 30 s per spectrum. Spectra were noise filtered and cleaned from cosmic rays using the chemometrics package of WiRE (version 3.4, Renishaw Plc, Wotton-under-Edge, UK). Data pre-processing was carried out in Matlab (version 2018a, MathWorks Inc., Natick, Massachusetts, USA), using an open source graphical user interface adapted from Felten et al. (2015) (https://www.umu.se/en/research/ infrastructure/visp/downloads/) with the following parameters: spectra were cut to a range of 510 to 1,800 cm−1, asymmetric least squares baseline corrected (λ = 5,000, pVal = 0.001, Eilers, 2004), total area normalized, and smoothed with Savitzky–Golay filtering (first-order polynomial, with a frame size of 3). Multivariate curve resolution-alternating least squares (MCR-ALS) analysis was first conducted on TW samples. Based on singular value decomposition, the data were fitted using four components and only non-negativity constraints for both spectral and concentration directions. Two of the resolved components showed spectral features corresponding to bands previously assigned to cellulose and lignin. The third resolved spectral profile showed features characteristic of aromatic extractives, while the fourth had residual, unidentified contributions (**Supplementary Figure S1A**). These resolved spectral profiles were used as initial estimates in the MCR-ALS modeling of NW spectra (**Supplementary Figure S1B**). The resolved components were used to classify voxels into chemically distinct categories, using hard *k*-means clustering. Three clusters were used, corresponding to distinct zones: lumen (in all samples), G-layer and S-layer (in TW and NW samples, respectively), and S-layer plus middle lamella (in all samples) (**Figure 3**, **Supplementary Figure S2**). Spectral maps where no zones could be distinguished were removed from further analyses. Spectra originating from the distinct chemical zones of interest (cell wall layers) of the respective genotypes/conditions were used for PCA and, when appropriate, for OPLS-DA analysis (Trygg and Wold, 2002) in SIMCA-P+ (version 14.0, Umetrics AB, Umeå, Sweden) to reveal chemical changes in cell wall layers between the genotypes and conditions.

#### Microfibril Angle Measurements

Radial-longitudinal stem sections of 100-µm thickness and 0.5 to 2-mm width were generated from six stems per genotype and condition (ca. 35 cm above soil level) using a Microm cryotome HM505E and dried between two microscopy glass slides. Xylem strips were transferred to metal holders for X-ray diffraction analysis. The X-ray diffraction experiments were carried out at the µ-spot beamline at the synchrotron facility BESSY II, Berlin, Germany. The radiation energy was set to 15 keV, corresponding to a wavelength of 0.8265 A. The (200)-Bragg peaks of cellulose, which were taken for orientation analysis, occurred at a scattering angle 2θ of 11.8. The samples were measured in ambient air, with the long axis of the fiber cells being perpendicular to the incident X-ray beam. The sample-detector distance was set to 245 mm, and the exposure time to 60 s. The beam diameter was set to 100 µm, so several measurements could be performed on one sample. Azimuthal intensity profiles (azimuthal angle φ vs. intensity) of the diffraction patterns were obtained by radially integrating the intensity of the (200)-peaks within 2θ ± 0.2° with an azimuthal step size of 1°. Microfibril orientation distributions and mean MFA were calculated by the simulation and fitting routine as described in detail previously (Rüggeberg et al., 2013).

#### Transcriptomics and Data Analysis

From greenhouse grown upright or inclined trees (22 days fixed at 45°) a 10- to 30-cm stem piece above soil level was harvested and frozen in liquid N2. We chose to analyze transcriptomes at a late stage of TW formation (22 days after inclination) since at this stage all processes that lead to uprighting are active, which is suggested by reduced performance in stem correction of ETI trees compared with WT (**Figure 1**, **Supplementary Video S1**). From trees forming TW, the bark was peeled off from the TW side of the stem, and developing wood and phloem/cambium (inner surface of the bark) were scraped with a scalpel into liquid nitrogen (Gray-Mitsumune et al., 2004). For NW, phloem/ cambium and developing wood were scraped all around the stem after peeling off the bark. Scrapings from three trees per genotype and condition were pooled equally by weight and ground to a fine powder using a mortar and pestle in liquid nitrogen. One hundred milligram of powder was used to extract RNA using the cetyl trimethylammonium bromide (CTAB) method (Chang et al., 1993). DNA removal was carried out using DNA-free™ DNase (Ambion). After DNA removal, RNA was cleaned with the Qiagen MinElute kit. RNA concentration was measured with a Nanodrop ND-1000 (Nano-Drop Technologies, Delaware, USA), and its quality was assessed using an Agilent 2100 Bioanalyzer with Agilent RNA 6000 Nano Chips according to manufacturer's instructions. Sequencing library generation and sequencing using Illumina HiSeq 2000 were carried out at SciLifeLab (Science for Life Laboratory, Stockholm, Sweden). Data sequences are available from the European Nucleotide Archive under the accession number PRJEB32252. Quality assessment and pre-processing of the obtained RNA-Seq reads were done

following the guidelines described in Delhomme et al. (2014). Briefly, FastQC (v0.11.5; Andrews, 2010) was used to assess the quality of the raw data. Then, ribosomal RNAs and sequencing adapters were removed using SortMeRNA (Kopylova et al., 2012) and Trimmomatic (Bolger et al., 2014), respectively. Read quality was assessed using FastQC after both these steps and reads that passed the quality assessment were aligned to the *P. tremula* genome (v1; available at http://plantgenie.org/, Sundell et al., 2015). HTSeq was used for summarizing read counts (Anders et al., 2015). Read counts were normalized in R (v3.5.3, R Core Team, 2019) using DESeq2 (v1.30.0, Love et al., 2014) and used for PCA and differential gene expression analysis. Gene expression was normalized considering the effect from the genotype (wild type or *pLMX5::etr1-1*) and tissue (TW or NW) using the following model: gene counts ~ genotype + tissue. Variancestabilizing transformation was applied using the DESeq2 package. A PCA was conducted (using the plotPCA command in DESEq2) to show general differences between all analyzed samples. Differentially regulated genes (DRGs) were selected using an absolute log2 fold change (log2FC) larger or equal to 0.5 (pAdj < 0.05) as cutoff (suggested in Schurch et al., 2016). The script used for the transcriptome analysis is available under https://github. com/UPSCb/UPSCb/tree/master/manuscripts/Seyfferth2019. Gene ontology (GO) enrichment analysis was performed with the GO and PFAM enrichment tool at www.aspwood.org (Sundell et al., 2015). Visualization of comparative expression analyses was performed using the drawing tool available at http:// bioinformatics.psb.ugent.be/webtools/Venn and gplots (v3.1.0). Promoter sequences of *Populus trichocarpa* of up to 2,000 bp in length were obtained and screened for the presence of the TEIL (tobaccoEIN3like) motif (AYGWAYCT; Kosugi and Ohashi, 2000), the GCC-box (AGCCGCC; Ohme-Takagi and Shinshi, 1995), and the ET-responsive element (ERE) (ATTTCAAA; Itzhaki et al., 1994). For precise promoter and cis-element prediction, we chose the *P. trichocarpa* genome (v3) as reference for this analysis, because it has a higher coverage in non-coding genome regions than does the available *P. tremula* genome sequence (v1). The "best hit" (*P. trichocarpa* sequence with the highest sequence similarity to *P. tremula* sequence) was defined as *P. trichocarpa* homolog, and the respective promoter sequences were obtained using the "Sequence Search" webtool from Popgenie (http:// popgenie.org/sequence\_search). Sequence match analyses are described in Felten et al. (2018).

#### Quantitative Polymerase Chain Reaction (qPCR)

Total RNA extraction for qPCR analysis was carried out from 100 mg wood powder with Qiagen RNeasy Plant Mini kit according to the manufacturer's instructions (using RLT buffer for lysis). Each of the three biological replicate samples per condition and genotype was a pool of xylem scrapings from three individual trees similar as described for material used for RNA-Seq experiments. An on-column DNA removal was carried out using Qiagen RNase-Free DNAse. Samples were subjected to a second DNA-removal step using the Invitrogen DNA-free kit according to the manufacturer's manual. RNA quality was assessed by gel electrophoresis and quantitated by a Qubit 2.0 fluorometer using the Invitrogen Qubit RNA BR Assay. cDNA was synthesized using Biorad iScript from 300-ng RNA per sample according to the manual. qPCR was carried out on a Biorad CFX96 realtime PCR machine with a C1000 Touch Thermal Cycler in duplicate reactions measuring 15 µl consisting of 7.5 µl 2× Roche LightCycler 480 SYBR Green I Master solution, 0.3 µM of each primer, and 2 µl of cDNA. For evaluating primer efficiency, equal aliquots of every cDNA sample were taken, mixed, and diluted five times with a dilution factor of 5. For expression analysis, a 1/10 dilution of cDNA was used. The primer list is found in **Supplementary Table S8**. The qPCR program consisted of an initial denaturation step (95°C, 5 min); followed by 40 cycles of denaturation (95°C, 10 s), annealing (55–58°C, 10 s), and elongation (72°C, 20 s); and followed by a melting curve (65– 95°C with a 0.5°C interval for fluorescence reading). Biorad CFX maestro software was used to determine the most stable reference genes among the five tested ones and yielded ideal values (*M* = 0.339 and stability Ln(1/Avg*M* = 1.082) for two genes, *UBC21* (*Potri.006G240900*) and *DAT* (*Potri.002G127700*), which were hence used for normalization of target gene expression data. Experimentally assessed primer efficiencies were taken into account for delta Cq calculations.

#### Statistical Analysis

Statistical analysis of OW/TW ratios (**Figure 1C**) between wild-type and ETI trees was calculated using a Student *T*-test with Welch correction. Differences were significant with pVal <0.05 (\*), pVal < 0.01 (\*\*), and pVal < 0.001 (\*\*\*). Statistical differences between the number of vessels in TW and OW (**Figure 1D**) in wild-type and ETI trees were calculated by a two-way analysis of variance (ANOVA), and a post-ANOVA Tukey test was used for pairwise comparisons (pVal < 0.05). An ANOVA and a Tukey test were also used to assigned statistical differences in MFAs obtained in the G- or S2-layer and in gene expression of ET-dependent genes in wild-type and ETI trees (**Figure 2**). Details about PCA and OPLS-DA (Trygg and Wold, 2002) models generated in SIMCA-P+ (version 14.0, Umetrics AB, Umeå, Sweden) on spectroscopic data are given in the

obtained from six individual trees per genotype. Letters indicate statistical significant differences between angles obtained in both genotypes and the G- or S2-layer calculated using a two-way analysis of variance (ANOVA) and a Tukey test (using a pVal cutoff = 0.05). ETI, ethylene insensitive; TW, tension wood.

respective figure legends (**Figures 3** and **4**), and the choice of these methods is described in detail in Felten et al. (2018). Statistical analysis for differential gene expression analysis (**Figure 5**) was performed using pre-set statistical settings in the DESeq2 package (Love et al., 2014). Briefly, pValues were calculated using the Wald test, and multiple testing correction was performed using the Benjamini–Hochberg approach. Statistical tests for GO term enrichment were performed on the basis of Fisher's exact test and included multiple testing (for detailed description, see Sjödin et al., 2009). Hypergeometric distribution was used to determine statistical significant enrichment of cis-elements (Felten et al., 2018; **Figure 5**).

#### RESULTS

#### Ethylene Signaling Is Required for Fully Functional Tension Wood

We compared the uplifting response of wild-type and ETI hybrid aspen after horizontal inclination through time-lapse photography over a period of 28 days. The ETI trees overexpress the dominant negative *Atetr1-1* receptor under the control of the wood-specific *pLMX5* promoter. We have previously shown that transformation of hybrid aspen trees with the *pLMX5::etr1-1* construct conveys strong ET insensitivity to woody tissues (Love et al., 2009; Felten et al., 2018). Representative videos tracking the response of one out of the six tested wild-type trees and one out of the six ETI trees over the first 21 days of leaning were compiled using the same number of pictures per day and per genotype, enabling direct comparison (**Supplementary Video S1**). The videos illustrate qualitatively a delayed bending and reduced

FIGURE 3 | Identification and chemical comparison of cell wall layers in NW and TW of wild-type and ETI (*pLMX5::etr1-1*) hybrid aspen trees. (A) Overlay of white light images on one representative section out of six scanned sections per genotype and scanned maps after *k*-means clustering based on four MCR-ALS resolved components and three clusters. The sizes of the spectral maps are as follows: (A) 2 × 29 spectra, (B) 2 × 25 spectra, (C) 2 × 30 spectra, and (D) 2 × 34 spectra. Lateral step size is 1 µm2, and scale bar in (D) is 10 µm. Clusters in NW correspond to lumen (yellow), middle lamella + S-layer (red), and S-layer (blue); in TW, clusters correspond to lumen (yellow), middle lamella + S-layer (orange), and G-layer (violet). (E) Score plots (PC1 and PC2) of a PCA of spectra from all cluster maps shown in Supplementary Figure S2. Colors correspond to cell wall layers as described above. Triangles represent wild type, and circles *pLMX5::etr1-1.* Model details: autofitted (70 components, 893 observations), PC1 explained 79% of variation, PC2 explained 10% of variation. R2X(cum) = 0.996, Q2(cum) = 0.99. The ellipse in (E) indicates a 95% confidence interval (Hotelling's *T*2). ETI, ethylene insensitive; MCR-ALS, multivariate curve resolution-alternating least squares; NW, normal wood; PCA, principal component analysis; TW, tension wood.

uplifting of the ETI trees as compared with the wild-type tree. To quantify this response, we selected the first time-lapse image of each day over the entire experimental period (*n* = 28) and calculated both the normalized lift and the curvature from stem traces captured in ImageJ (**Figures 1A, B**). While these two traits are interdependent, they also reveal slightly different aspects of the TW response since the same normalized lift may be supported by different curvatures. The curvature also gives an indication of the shape of the stem (e.g., 180° is a perfect half circle or "C" shape). The ETI trees showed a delayed bending response with a dramatic 52% reduction in lift relative to wild type on

the final day of the experiment (**Figure 1A**). Stem curvature was also negatively affected in ETI trees, but to a lesser extent (**Figure 1B**). This illustrated that functional ethylene signaling is required in xylem tissues for the uplifting response.

#### Asymmetric Growth and Vessel Reduction, but Not G-Fiber Formation, Require Functional Ethylene Signaling

To assess potential anatomical causes of the reduced uplifting response in ETI trees, we inspected the xylem anatomy

FIGURE 4 | Chemical differences among G-layer (A and C) and S-layer (B and D) in TW of wild-type and ETI (*pLMX5::etr1-1*) hybrid aspen revealed by Raman microspectroscopy. (A and B) OPLS-DA score plots of 1 + 2 (predictive + orthogonal) component models based on the respective cluster extracted from all cluster maps shown in Supplementary Figure S2. Specification for the models: (A) 296 observations, R2X(cum) = 0.723, R2Y(cum) = 0.924, and Q2(cum) = 0.921; (B) 149 observations, R2X(cum) = 0.804, R2Y(cum) = 0.946, and Q2(cum) = 0.941. The ellipses in (A) and (B) correspond to 95% confidence intervals (Hotelling's *T*2). (C) and (D) The corresponding correlation-scaled loadings plots with a randomly selected spectrum underneath each representing the spectra used in the models. Wavenumbers with high correlation to separation are labeled. Noticeably, these do not always correspond to distinct peaks in (D) but can indicate shoulders (1,616 cm−1) or shifts towards different energy levels (lower, e.g., 1,075 cm−1; higher e.g., 1,390 cm−1), the latter being indicative of structural rather than proportional changes (qualitative rather than quantitative differences). ETI, ethylene insensitive; OPLS-DA, orthogonal projections to latent structures discriminant analysis; TW, tension wood.

in cross sections of leaned trees. The TW : OW ratio was significantly reduced in ETI trees (**Figure 1C**), confirming previous observations of reduced asymmetric growth in this genotype (Love et al., 2009). Horizontally inclined wild-type trees displayed a dramatic and expected decrease in vessel density on the TW side compared with the OW side after 4 weeks (**Figure 1E**). Compared with wild-type trees, the ETI trees had a higher vessel density in both TW (increase of

FIGURE 5 | Transcriptome analysis between wild-type (WT) and ETI (*pLMX5::etr1-1*) hybrid aspen trees identifies ethylene-downstream genes with potential roles in TW formation. (A) PCA of transcriptome data. PC1 explains 41% of the variation and refers to the difference in NW and TW data. PC2 (16%) separates the TW transcriptomes of the WT and *pLMX5::etr1-1*. Each of the three replicate samples per genotype and condition originates from a pool of xylem/phloem scrapings from three trees. ERE, ethylene-responsive element; ETI, ethylene insensitive; NW, normal wood; PCA, principal component analysis; TW, tension wood. (B) Venn diagram highlighting the number of differentially regulated genes (DRGs; |log2FC| ≥ 0.5 and pAdj <0.05) in NW-to-TW comparisons. DRGs that only showed a significant change in expression in WT, but not in *pLMX5::etr1-1*, are further called "ET-dependent" (Supplementary Tables S1–S3). (C) Heat map showing expression changes (log2FC) of ET-dependent DRGs in TW compared with NW (Supplementary Table S5). Red indicates induction and blue suppression of gene expression in TW as compared with NW. White indicates absence of significant change in gene expression between TW and NW. Genes depicted in the heat map function in transcriptional regulation of cell division, xylem differentiation, and secondary cell wall biosynthesis; phytohormone signaling; microtubule and actin organization; and cell expansion and cell wall chemistry. Black boxes next to each gene indicate the presence of the ERE, TEIL-motif, and/or the GCC-box in its promoter region (Supplementary Table S7). ABA, abscisic acid; BR, brassinosteroid; CK, cytokinin; ET, ethylene; GA, gibberellins; JA, jasmonic acid.

133%) and OW (increase of 30%) but not in NW (**Figures 1D, E**, **Supplementary Figure S3**). Strikingly, ETI trees presented G-fibers with no obvious anatomical difference compared with wild-type G-fibers, despite their reduced capacity to lift and the commonly held view of G-fiber involvement in generating the mechanical strain needed to uplift the stem (Mellerowicz and Gorshkova, 2012). Taken together, ETI trees showed decreased asymmetric growth, higher vessel density in both TW and OW, and a reduced uplifting response after inclination than did wild type, indicating that ET signaling is required for many characteristic features of the TW response, but not necessarily G-fiber formation.

#### S-Layers in G-Fibers of Ethylene-Insensitive Trees Have an Altered Cellulose Microfibril Angle and Chemical Composition

With the aim to reveal potential causes of the attenuated uplifting response despite the presence of G-fibers in ETI TW, we inspected the cellulose MFA and chemical composition of the secondary cell wall more closely. When measuring MFA in TW of six wild-type and ETI trees that were each inclined to 45° and fixed in this position for 22 days, we observed in both genotypes the bimodal MFA distribution typical for TW and originating from small (0–5°) G-layer MFAs and larger (20–40°) S-layer MFAs (Müller et al., 2006; Goswami et al., 2008; Rüggeberg et al., 2013). This was expected as both genotypes formed S2- and G-layers (**Figure 1E**). There was no significant difference in average G-layer MFAs between the genotypes (**Figure 2**). However, the average S2-layer MFAs in ETI trees were significantly lower than those in wild-type trees (36.7 ± 2.2° for ETI trees and 41.2 ± 4° for wild-type trees).

We furthermore assessed whether G-fiber-rich TW areas were chemically altered, in cross sections of ETI and wildtype trees using FT-IR microspectroscopy with a singleelement detector. PCA performed on these spectra did not show clear grouping (**Supplementary Figure S4**), indicating that the overall chemical composition of the sampled areas is not altered. As unaltered cells/cell wall layers can mask subtle changes confined to small zones/specific layers, we used Raman microspectroscopy with high lateral resolution to specifically focus on the chemical information from G- and S-layers in wild-type and ETI trees. While analyzing data from spectral maps scanned over three adjacent fiber cells with 1-µm step size in TW and NW of both genotypes, four chemically different zones were assigned based on spectral information that overlapped with cell layers visible in the sections: (1) cellulose rich G-layer, (2) cellulose- and lignin-containing S-layer, (3) lignin-rich zone including both the S-layer and the middle lamella (hereafter called S + ML), and (4) lumen (**Figures 3A**–**D**, **Supplementary Figures S1, S2**). Comparison of NW S-layer spectra between wild-type and ETI trees or NW S + ML layer spectra between wild-type and ETI trees by PCA did not reveal any clear grouping and thereby ruled out any genotype-related chemical differences at the start of the leaning experiment (**Supplementary Figure S5**). PCA of non-lumen spectra from TW and NW from both genotypes revealed clear groupings of TW and NW samples (**Figure 3E**) as well as between the respective cell wall layers. This suggests that TW is chemically different from NW not only by the addition of the G-layer to fibers but also by a chemical alteration of the S-layer itself. The respective OPLS-DA models for TW G-layer and TW S-layer spectra of both genotypes indicated a chemical difference in both layers between wild-type and ETI trees (**Figures 4A, B**). The corresponding correlationscaled loadings revealed a limited number of distinct bands that contribute to the difference in the G-layer (**Figure 4C**), a higher proportion of –C═O double-bond vibrations in the ETI G-layer (1,730 cm−1, not related to cellulose), together with a lower proportion of –C–H vibrations (1,376 cm−1, typically assigned to cellulose based on Gierlinger and Schwanninger, 2006). The correlation-scaled loadings for S-layers in G-fibers revealed more variables contributing to the differences between wild-type and ETI trees (**Figure 4D**). The S-layer spectra in ETI G-fibers had, in addition to bands already marked in the G-layer loadings in **Figure 4C**, a lower proportion of lignin-associated vibrations (1,616 and 1,660 cm−1) and a higher contribution of bands associated with polysaccharidic compounds (carbohydrate ring-associated vibrations around 1,100 cm−1). Taken together, Raman microspectroscopy was able to identify chemical differences among S- and G-layers of wild-type compared with ETI TW, with a clearer difference in the S-layers of G-fibers, as compared with their G-layers. This is in line with results from X-ray diffraction that also revealed a stronger effect on the S-layer ultrastructure of the G-fibers as compared with the G-layers themselves in ETI trees.

#### Ethylene Signaling Suppresses Lignin Biosynthesis Genes and Induces Expression of Microtubule and Actin-Related Genes in Tension Wood

To understand the impact of ET signaling on transcriptional reprogramming during TW formation, we compared transcriptomes from TW (trees inclined to and fixed at 45° for 22 days) and NW (upright trees) forming tissues in wild-type and ETI trees (**Figure 5**). PCA suggested that the predominant difference between the obtained transcriptomes (describing 41% of variation in the data set) is explained by the tissue type (NW or TW) (**Figure 5A**). The second component (describing approximately 16% of the overall variation in the data set) separated transcriptomes based on the genotypic background. The separation between wild-type and ETI trees was seen for both TW and NW transcriptomes. We next selected genes with significantly changed expression (DRGs, |log2FC| ≥ 0.5 and pAdj < 0.05) in TW compared with NW in both the wild-type and ETI trees (**Supplementary Tables S1, S2**). We detected 6,193 DRGs in wild type, while only 3,747 DRGs were identified in ETI trees (**Figure 5B**). We observed an almost equal distribution of induced and repressed genes in both genotypes (**Supplementary Figure S6A**). Expression of 3,010 DRGs, with close to equal numbers of induced and repressed ones, was significantly changed only in TW of wild-type trees (**Figure 5B**, **Supplementary Table S3**). Loss of transcriptional regulation of these genes in the ETI trees implicates that their expression requires functional ethylene signaling. We therefore defined these genes as "ET-dependent genes." GO term analysis of the ET-dependent genes revealed an enrichment of genes related to ribosomes and translation (**Supplementary Table S4**). Functional characterization of the ET-dependent genes identified several TFs with known functions in xylem differentiation and gene networks related to hormone pathways, cell expansion, cytoskeleton organization, and cell wall chemistry (**Figure 5C**, **Supplementary Table S5**). To investigate a potential direct transcriptional regulation of these ET-dependent genes by TF families that function downstream of the ET receptors, we screened promoter regions of ET-dependent genes (2,000-bp upstream of the start codon) for the presence of the ERE, EIN3 (TEIL)- and ERF (GCCbox)-binding motifs (**Figure 5C**, **Supplementary Table S6**). The vast majority of the selected ET-dependent genes with a function in wood development, anatomy, or chemistry contained the ERE motif (90 genes), the TEIL motif (66 genes), and/or a GCC-box (six genes) in their promoter region. Approximately 37% (101 genes) of the 234 ET-dependent genes that were selected based on their association with hormone pathways, cell expansion, cytoskeleton organization, cell wall chemistry, or transcriptional regulation during wood formation did not contain any motif.

We detected 35 hormone-related genes among the ET-dependent DRGs, with the largest group of them (15 genes) being related to auxin transport and signaling (**Figure 5C**). The ET-dependent genes suggested that ET signaling stimulates induction of abscisic acid (e.g., *9-CIS-EPOXYCAROTENOID DIOXYGENASE* (*PtNCED*)) and jasmonic acid (e.g., *DELAYED DEHISCENCE 2* (*DDE2*) biosynthesizes genes and suppresses genes involved in ET biosynthesis (e.g., *PtACO6*) and degradation of cytokinins (e.g., *CYTOKININ OXIDASE*s (*CKX*s)). We also identified genes involved in signaling of auxin (e.g., *AUXIN RESPONSE FACTOR*s (*ARF*s)), brassinosteroids (e.g., *BRI1- EMS-SUPPRESSOR 1* (*BES1*)), and gibberellins (GAs) (e.g., *A-STIMULATED IN ARABIDOPSIS 14* (*GASA14*)). We evaluated the effect of overexpression of *Atetr1-1* on the expression of genes related to the ET pathway (**Supplementary Figure S6B**). Whereas TW formation resulted in differential expression of ET biosynthesis genes (*PtSAM*s, *PtACO*s, and *PtACS*) regardless of ET insensitivity, differential expression of many ET downstream genes (mainly TFs like *PtEIN3* and various *PtERF*s) was absent or attenuated in ETI trees during TW formation. We further found that wild type, but not ETI trees, showed induced expression of genes involved in cell division (*WOX4* (*WUSCHEL RELATED HOMEOBOX 4*), *E2F3* (*E2F TRANSCRIPTION FACTOR 3*), and *CYCLIN*s), actin and microtubule organization (e.g., *MICROTUBULE-ASSOCIATED PROTEINS* (*MAP*s)), and cell expansion (e.g., *EXPANSIN*s) in TW (**Figure 5C**; **Supplementary Figure S6D**). The ET-dependent genes also included two key TFs for xylem fiber differentiation and vascular patterning, *SND2* (*SECONDARY WALL-ASSOCIATED NAC DOMAIN PROTEIN 2*) and *SHR* (*SHORT ROOT*). These TFs were suppressed in TW of wild-type trees, while no change in expression was observed in ETI trees. In line with the observed lower amount of vessels in TW of wild-type trees (**Figure 1D**), we observed during TW formation in wild-type trees a slight repression of two *Populus*

homologs (*PtVND7.1* and *PtVND7.2*) of the master regulator of vessel formation *VASCULAR-RELATED NAC-domain 7*  (*VND7*) (although statistically not significant, **Supplementary Figure S6C**). During TW formation in ETI trees, no transcript change was observed for *PtVND7.2* but a slight, but not statistically significant, induction of *PtVND7.1*. Statistically significant ET-mediated suppression of genes involved also genes in the lignin pathway (e.g., *MY52*, *MYB83*, *CAFFEATE O-METHYLTRANSFERASE2* (*PtCOMT2*), and *CINNAMYL ALCOHOL DEHYDROGENASE* (*PtCAD2*)) (**Figure 5C**), whereas no phenylalanine ammonia lyase (PAL)-encoding genes that act upstream during lignin biosynthesis were among our target genes. qPCR results validated that ETI trees lost regulation of genes involved in lignin biosynthesis such as *PtCOMT2* during TW formation (**Supplementary Figure S6D**). The ET-dependent genes also comprised genes encoding cellulose synthases (e.g., *PtCSLD5*), pectin lyases/acetyltransferase/methylesterases (e.g., *PME44*), CAZymes (e.g., *CELLULASE1*), and sugar transporters (e.g., *SUGAR TRANSPORTER1* (*STP1*)). Despite the large impact of ET on gene regulation in TW, only two G-layerassociated fasciclin-like arabinogalactan (*FLA*) genes (Lafarguette et al., 2004; Kaku et al., 2009) were found among the ET-dependent genes (*FLA1* and *AGP14*, **Supplementary Table S3**), which is in line with the presence of a G-layer even in ETI trees (**Figure 1E**).

This analysis of ET-dependent genes suggests that ET signaling triggers expression of genes underlying cell division and expansion and suppresses genes involved in vessel differentiation and biosynthesis and polymerization of lignin. A part of these genes may be directly regulated by TFs of the ET signaling cascade as indicated by the presence of GCC-box, TEIL, or ERE motifs in the promoters of these genes.

#### DISCUSSION

Understanding the mechanisms by which trees sense displacement and react to correct their stem position through TW formation is a long-standing question. A role for phytohormones, such as GAs, auxin, and ET, has been discussed in this context (Andersson-Gunnerås et al., 2003; Love et al., 2009; Nugroho et al., 2012; Gerttula et al., 2015), but the significance of ET signaling remains unclear. In this study, we report on the importance of signaling by the phytohormone ET for both TW formation and functionality. ET signaling is often induced in response to environmental stimuli such as abiotic and biotic stresses (Dubois et al., 2018). In the absence of any stress signals, such as during upright growth in greenhouse conditions, ETI trees are similar to wild-type trees and do not show any obvious growth defects (Love et al., 2009). Therefore, these trees are ideally suited to study the role of ET signaling during an induced stress like displacement. Our data indicate that loss of ET signaling in ETI trees restricts the lifting response of inclined stems (**Figure 1**) and interferes with most developmental responses that are characteristic for TW, but not the formation of G-layers. Stem eccentricity observed in TW is the result of enhanced cambial growth at the upper side of the leaning stem (Ruelle et al., 2006; Ruelle, 2014). Stimulation of xylem growth on the TW side was attenuated in ETI trees during tilting resulting in a lower TW : OW ratio than in wild-type trees. The concept that the reduced TW : OW ratio results from reduced cambial cell division in ETI TW (**Figure 1C**) is supported on a molecular level by the attenuated reprogramming of cell cycle regulators, such as *CYCD3* and *E2F3* in ETI trees during TW formation (Kosugi and Ohashi, 2002), as well as regulators of cambial cell division activity like *WOX4* (Hirakawa et al., 2010; Ji et al., 2010; Suer et al., 2011; Kucukoglu et al., 2017; Shi et al., 2019; **Figure 5C**). *WOX4* expression is controlled by auxin signaling. Auxin forms a gradient over the cambium in the stem and is known to function as a positional signal and to positively influence cambial growth (Tuominen et al., 1997; Uggla et al., 1998; Smetana et al., 2019). Conversely, inhibiting auxin signaling in stems of transgenic trees negatively impacts (in addition to primary also) secondary growth (Nilsson et al., 1992). We identified several genes related to auxin transport and signaling (e.g., *PIN6* and *ARF*s) among the ET-dependent DRGs in our study (**Figure 5C**), which is in line with previous reports that showed that ET influences both auxin biosynthesis and transport (Swarup et al., 2007; Muday et al., 2012; Vaseva et al., 2018). Based on these results, it is possible that the reduced stimulation of cambial growth observed in ETI TW is explained through ET–auxin crosstalk. ETI TW showed a reduced amount of G-fibers than did wild-type TW, due to the mentioned reduced cambial stimulation and as a result of lower fiber-to-vessel ratio because of reduced suppression of vessel formation (**Figure 1C**; Love et al., 2009). Reduced vessel frequency is observed in many other species during TW formation (Yoshizawa et al., 2000; Jourez et al., 2001; Ruelle et al., 2006; Sultana et al., 2010). Results from yeast one-hybrid assays identified the cell-cycle regulator E2Fc as a regulator of transcriptional gene networks for secondary cell wall formation in *Arabidopsis*. E2Fc can directly bind to promoter regions of NAC TFs involved in vessel differentiation, namely, *VND6* and *VND7* (Taylor-Teeples et al., 2015). Only TW of wild-type trees showed induction of a *Populus E2F* homolog (*E2F3*) and a slight, yet not statistically significant, suppression of *PtVND7.1* and *PtVND7.2*, while such expression changes were not observed in ETI trees (**Figure 5C**, **Supplementary Figure S6C**, **Supplementary Tables S3, S7**). Furthermore, *PtLBD18* (*LATERAL ORGAN BOUNDARIES DOMAIN 18*), a homolog of *AtLBD18*, which has been shown to regulate expression of *AtVND7* (Soyano et al., 2008), was among the ET-dependent genes. The EIN3 binding motif is present in promoters of *Populus E2F3*, *LBD18*, and *VND7*, suggesting that expression of these genes might be under the direct control of EIN3 during TW formation (**Figure 6**). Furthermore, ET–auxin interaction might also contribute to the changes in vessel formation. Auxin, when exogenously applied, suppresses expression of *NAC*s involved in fiber differentiation and stimulates expression of *NAC*s involved in vessel formation in hybrid aspen stems (Johnsson et al., 2018). Expression of several genes involved in auxin signaling was regulated by ET in TW (**Figure 5C**), some of them harboring TEIL or ERE motifs in their promoters (e.g., *PtPIN6*, *PtARF1*, and *PtARF10*). Therefore, in addition to a direct effect of ET on *NAC* expression and vessel density, a potential regulation of vessel differentiation through *NAC* regulation *via* ET–auxin crosstalk is also plausible. The absence of repression of *E2F3* and *VND7* in ETI TW could also contribute to the reduced suppression of vessel development during TW formation in ETI trees. The reduction of vessel elements in TW of wild-type trees as compared with NW likely causes an absolute decrease of lignin levels on tissue level, because lignin content in vessel cell walls is generally higher as compared with the one in fiber S2 walls (Saka and Goring, 2009; Xu et al., 2005). As less of the highly lignified vessels are formed in wild-type TW compared with NW, the lignin biosynthesis machinery is expected to be downregulated. During formation of TW in ETI trees, the absence of vessel suppression goes in line with an absence of downregulation of genes involved in lignin biosynthesis as observed in our data (**Figure 5C**). An effect of ET signaling and regulation of lignification has previously been demonstrated in experiments where lignification of tracheary elements (TEs) in cell culture was blocked by inhibiting ET signaling (Pesquet et al., 2013). For a few of the ET-dependent lignin biosynthesis genes from our data set, we identified the EIN3, GCC-box, or ERE-motif in their promoters, suggesting a direct regulation of their expression by TFs in the ET pathway (**Figures 5C** and **6**).

We previously reported that exogenously enhancing ET signaling was sufficient to induce the G-layer formation in hybrid aspen in an ET signaling-dependent manner (Felten et al., 2018). Unexpectedly, in the present study, we observed that G-fibers with G-layers could form during TW formation in ETI trees (**Figure 1E**), despite inhibited ET signaling, with no altered ultrastructure in the G-layer (**Figure 2**) and no dramatic changes in gene expression typically associated with G-layer presence, such as induction of *FLA*s (**Supplementary Table S2**). This is, to a large extent, similar to what has been found with GAs. Although it remains to be shown that GA biosynthesis is altered during tilting, exogenously applied GAs can induce G-fiber formation and xylem growth (Jiang et al., 1998; Jiang et al., 2008; Nugroho et al., 2012). Conversely, inhibition of GA signaling counteracts stem bending and lifting and lowers eccentric growth in displaced stems, despite G-fibers being formed (Jiang et al., 1998; Jiang et al., 2008; Nugroho et al., 2012). This suggests that both GA and ET signaling can induce G-fiber formation but that inhibiting one of these pathways does not abolish G-fiber formation. We found very few GA-associated genes among the ET-dependent DRGs, suggesting that GA signaling is not impacted in ETI trees. It will be interesting to investigate whether a combination of both ET and GA signaling inhibition can fully inhibit the G-layer formation during stem tilting and whether these phytohormone pathways can compensate for each other's absence. The present study also confirms what was observed with exogenous ACC/ET application experiments (Felten et al., 2018), that the effect of ET signaling on vessel and the G-layer formation is uncoupled and must therefore occur as independent routes downstream of ET signaling.

Even though G-fiber formation as such was not inhibited during TW formation in ETI trees, we observed ultrastructural and chemical alteration of G-fiber S-layers in ETI TW

microfibril angles. Direct regulation through EIN3 and ERFs can be assumed for *EXPANSIN*s harboring TEIL and GCC-boxes in their promoter as well as indirect effects through ethylene–auxin interaction. TW, tension wood.

(**Figures 2**–**4**). The significantly lower MFA of the S2 layer in ETI TW suggests a role of ET in modification of secondary cell wall MFA in TW. G-layers had, however, no alternated MFA in ETI TW. Using Raman microspectroscopy, we observed chemical alterations in both the G- and S-layers of G-fibers as a result of ET insensitivity (**Figure 4**). Our spectroscopic data suggest differences both in the relative proportions of major cell wall polymers and in the structure of the polymers such as changes in cellulose crystallinity or a difference in crosslinking of polymers. It is possible that the chemical differences identified originate either from the layers itself or from a shift in clustering related to the composition of layer boundaries (e.g., the boundary of S- and G-layers). Gradients across the G-layer in both the polymer composition and the cellulose ultrastructure have been observed (Prodhan et al., 1995; Joseleau et al., 2004; Gierlinger and Schwanninger, 2006; Arend, 2008; Sandquist et al., 2010; Ruelle, 2014). At the S-G-layer interface, an accumulation of xyloglucan and rhamnogalacturonan has been detected (Sandquist et al., 2010) and linked to the mechanisms required for stress generation in G-fibers (Nishikubo et al., 2007; Mellerowicz and Gorshkova, 2012). As such boundaries and gradients are narrower than what can be resolved in a separate class by the Raman microspectroscopy, their chemical composition will influence the cluster that they are included in by *k*-means clustering. If the composition of this boundary was altered in ETI TW, the pixels comprising them may chemically become more similar to another neighboring layer (S instead of G) and therefore be included in that latter cluster. In conclusion, the chemical differences detected in the ETI G-fiber S-layer (lower lignin and higher polysaccharide contribution) could result either from a chemical alteration within the layer or from a higher proportion of S–G-boundary pixels, low in lignin and high in hemicellulose, included in the cluster of this layer. This is, to our knowledge, the first report that shows that ET can interfere with both the cellulose ultrastructure and the chemical composition of the secondary cell wall.

In summary, we have detected both ET-dependent quantitative (G-fiber number) and qualitative (G-fiber properties) aspects that could contribute to the observed attenuated uplifting in ETI plants. The G-layer itself does not seem to be a prerequisite for high tensile stress generation in TW, since other cell wall layer arrangements leading to a low cellulose MFA in the main cell wall layer can also induce the tensile stress required for stem lifting (Okuyama et al., 1994; Yoshizawa et al., 2000; Ruelle et al., 2007a; Ruelle et al., 2007b). However, in species containing a G-layer, there is a relationship between its thickness, its cellulose MFA, and the potential growth strain exerted (Fang et al., 2008; Goswami et al., 2008). Typically, in TW of *Populus*, the MFA in the S-layer is much larger than in NW, while the G-layer inherits an MFA close to 0° (Timell, 1986; Lautner et al., 2012). It has been hypothesized that the interplay between the G-layer and the S2-layer with its high MFA facilitates generation of the tensile stress upon swelling, required to correct the stem position through uplifting (Goswami et al., 2008). G-layer swelling leads to a pressure on the surrounding cell walls that is translated into axial stress. According to a model by Goswami et al. (2008), this stress enhancement factor depends on the orientation of the MFAs in the S2 layer. The MFA alteration observed in our study for the S2-layer of ETI TW is within the range of stress enhancement factors leading to a sufficient axial stress for uplifting and lies in a range of MFAs also observed in TW of other wild-type trees with G-fibers. Taken together, as the G-layer MFA is not altered in ETI TW and the S-layer MFA is not altered to an extent that would limit the uplifting response according to the model proposed by Goswami et al. (2008), the origin of inhibited lifting observed in ETI trees (**Figure 1**) can not only be attributed to altered MFA angles. Other authors have observed a correlation between the stem bending response and the amount of G-fibers formed (Nugroho et al., 2012). This indicates that there is both qualitative and quantitative effects of G-layers/G-fibers on the potential of stem reorientation. Given the rather low changes in qualitative G-fiber properties in ETI TW in comparison with the large effect on G-fiber quantity, we propose that it is rather the latter that is decisive for the reduced uplifting response, even though we cannot exclude that G-fiber amounts and S-layer chemistry and ultrastructure in ETI G-fibers have additive effects.

Taken together, the results presented here demonstrate the importance of ET signaling to define TW characteristics and to correct stem position after tilting. We also revealed key genes that potentially can coordinate these responses downstream of ET signaling during TW formation (**Figure 6**).

#### DATA AVAILABILITY

The raw sequencing data for this study can be found in the European Nucleotide Archive (ENA) under accession PRJEB32252.

#### AUTHOR CONTRIBUTIONS

CS performed transcriptomic analysis. BW performed and analyzed uplifting experiments. AG performed FT-IR and Raman microspectroscopic measurements and data analysis. JL supplied biological material for FT-IR measurements. MR carried out cellulose MFA measurements. ND gave support for transcriptomics analysis. TV and BW performed vessel density quantification. KA developed software for data analysis of uplifting experiments. BS and JF conceived the study. JF, BS, and HT supervised the experiments. JF generated biological material for the experiments. CS and JF wrote the manuscript with contribution of all authors.

#### FUNDING

The experiments as well as BW, JWL, and JF were funded by the Swedish Research Council Formas (grant nos. 213-2011-1148 and 239-2011-1915) and the Swedish Foundation for Strategic Research (grant no. RBP14-0011). CS was funded by The Kempe Foundation (SMK-1649 and SMK-1533) and the Lily och Sven Lawskis fund. Experimentation was financially supported also by the Swedish Governmental Agency for Innovation Systems Vinnova (grant no. 2016-00504) and the Knut and Alice Wallenberg Foundations (grant no. 2016-0341).

#### ACKNOWLEDGMENTS

We thank the students of the Design Build Test course 2016 at Umeå University (Carolina Eriksson, Karolina Eriksson, Oskar Hallberg, Alexander Lindström, and Johan Lundberg) for developing the imaging tools to record and measure tree uplifting. The support of the Chemical Biological Centre and the Department of Chemistry, Umeå University, Sweden, for the Vibrational Spectroscopy Core Facility is gratefully acknowledged. We would also like to acknowledge support from Science for Life Laboratory, the National Genomics Infrastructure funded by the Swedish Research Council, and Uppsala Multidisciplinary Center for Advanced Computational Science for assistance with massively parallel sequencing and access to the UPPMAX computational infrastructure. MR would like to thank Stephan Siegel and Chenghao Li for the support with the X-ray powder diffraction (XRD) measurements at BESSY II.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01101/ full#supplementary-material

SUPPLEMENTARY VIDEO S1 | Time-lapse video of the uplifting response of one representative pLMX5::etr1-1 and one representative wild type hybrid aspen tree imaged over the first 21 days of the experiment.

### REFERENCES


of woody stems in Populus. *Plant Cell* 27, 2800–2813. doi: 10.1105/ tpc.15.00531


**Conflict of Interest Statement:** Björn Sundberg is employed by Stora Enso AB, Sweden. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Seyfferth, Wessels, Gorzsás, Love, Rüggeberg, Delhomme, Vain, Antos, Tuominen, Sundberg and Felten. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Super-*Agrobacterium* ver. 4: Improving the Transformation Frequencies and Genetic Engineering Possibilities for Crop Plants

*Satoko Nonaka1,2\*, Tatsuhiko Someya2, Yasuhiro Kadota3, Kouji Nakamura2 and Hiroshi Ezura1,2\**

#### *Edited by:*

*Dominique Van Der Straeten, Ghent University, Belgium*

#### *Reviewed by:*

*Benoit Lacroix, Stony Brook University, United States Kan Wang, Iowa State University, United States*

#### *\*Correspondence:*

*Satoko Nonaka nonaka.satoko.gt@u.tsukuba.ac.jp Hiroshi Ezura ezura.hiroshi.fa@u.tsukuba.ac.jp*

#### *Specialty section:*

*This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science*

*Received: 24 April 2019 Accepted: 02 September 2019 Published: 07 October 2019*

#### *Citation:*

*Nonaka S, Someya T, Kadota Y, Nakamura K and Ezura H (2019) Super-Agrobacterium ver. 4: Improving the Transformation Frequencies and Genetic Engineering Possibilities for Crop Plants. Front. Plant Sci. 10:1204. doi: 10.3389/fpls.2019.01204*

*1 Tsukuba Plant Innovation Research Center, Gene Research Center, University of Tsukuba, Tsukuba, Japan, 2 Faculty of Life and Environmental Sciences, University of Tsukuba, Tsukuba, Japan, 3RIKEN Center for Sustainable Resource Science, Plant Immunity Group, Yokohama, Japan*

*Agrobacterium tumefaciens* has been utilized for both transient and stable transformations of plants. These transformation methods have been used in fields such as breeding GM crops, protein production in plant cells, and the functional analysis of genes. However, some plants have significantly lower transient gene transfer and stable transformation rates, creating a technical barrier that needs to be resolved. In this study, Super-*Agrobacterium* was updated to ver. 4 by introducing both the ACC deaminase (*acdS*) and GABA transaminase (*gabT*) genes, whose resultant enzymes degrade ACC, the ethylene precursor, and GABA, respectively. *A. tumefaciens* strain GV2260, which is similar to other major strains (EHA105, GV3101, LBA4404, and MP90), was used in this study. The abilities of the Super-*Agrobacterium* ver. 4 were evaluated in *Erianthus ravennae*, *Solanum lycopersicum* "Micro-Tom," *Nicotiana benthamiana*, and *S. torvum*. Super-*Agrobacterium* ver. 4 showed the highest T-DNA transfer (transient transformation) frequencies in *E. ravennae* and *S. lycopersicum*, but not in *N. benthamiana* and *S. torvum*. In tomato, Super-*Agrobacterium* ver. 4 increased the stable transformation rate by 3.6-fold compared to the original GV2260 strain. Super-*Agrobacterium* ver. 4 enables reduction of the amount of time and labor required for transformations by approximately 72%, and is therefore a more effective and powerful tool for plant genetic engineering and functional analysis, than the previously developed strains. As our system has a plasmid containing the *acdS* and *gabT* genes, it could be used in combination with other major strains such as EHA105, EHA101, LBA4404, MP90, and AGL1. Super-*Agrobacterium* ver. 4, could thus possibly be a breakthrough application for improving basic plant science research methods.

Keywords: *Agrobacterium tumefaciens*, Super-*Agrobacterium*, genetic engineering, plant transformation, gamma-aminobutyric acid, ethylene, GABA transaminase, ACC deaminase

### INTRODUCTION

*Agrobacterium tumefaciens* is an α-proteobacteria that causes crown gall disease in many agriculturally and economically important species, such as those from the families *Rosaceae* (rose, apple, cherry, and pear), *Vitaceae* (grape), and the genus Juglans (walnut) (Kado, 2014). *A. tumefaciens* has the ability to transfer T-DNA from bacterial cells to plant cells (T-DNA transfer). Transferred T-DNA is integrated into the plant genome *via* complicated plant cell systems (Guo et al., 2019), and results in crown gall disease. To utilize this unique ability of *A. tumefaciens* for research purposes, there has been a great deal of effort to remove its oncogenesis characteristics, and to develop a binary vector system (Zambryski et al., 1983; Hoekema et al., 1983; Bevan, 1984; Komari et al., 2006). There has been further effort to increase the T-DNA transfer frequency of *A. tumefaciens*; one effective strategy was to upregulate its *vir* gene expression levels. The application of *vir* gene inducers (Stachel et al., 1985; Stachel et al., 1986; Cangelosi et al., 1990; He et al., 2009; Hu et al., 2013), using ternary system (van der Fits et al., 2000), utilization of Super-binary vectors (Komari, 1990; Hiei et al., 1994; Ishida et al., 1996), and a modification of the Ori of the binary vector (Ye et al., 2011; Vaghchhipawala et al., 2018), have subsequently improved its transformation frequencies.

Another strategy to increase T-DNA transfer frequencies was the removal of the negative factors of the *Agrobacterium*plant interactions, such as ethylene, the gaseous phytohormone. Applications, such as aminoethoxyvinylglycine (AVG), an ethylene biosynthesis inhibitor, and AgNO3 or silver thiosulfate (STS), ethylene perception inhibitors, were effective at improving the T-DNA transfer frequencies in tomato, melon, and bottle gourd (Davis et al., 1992; Ezura et al., 2000; Han et al., 2005; Nonaka and Ezura, 2014). An alternative strategy was the utilization of ACC deaminase (AcdS) activity, which cleaves ACC, the ethylene precursor, to ammonia and α keto-butyrate. This enzyme was found in some plant growth promoting bacteria (PGPBs), such as *Pseudomonas* sp., which were found on the plant surface (Sheehy et al., 1991), and these bacteria utilize ACC as a nitrogen source. However, *A. tumefaciens* C58 strain, which was the original strain for *Agrobacterium*-mediated transformation, does not have the *acdS* gene or its activities (Wood et al., 2001; Nonaka et al., 2008a). Therefore, the utilization AcdS activity seemed to be reasonable. Indeed, *A. tumefaciens* GV2260 that had AcdS activity introduced into it, was efficacious in the suppression of ethylene evolution from plant tissues during co-cultivation and increasing T-DNA transfer [Nonaka et al., 2008a (Super-*Agrobacterium* ver.1); Ntui et al., 2010; Hao et al., 2010]. Moreover, Super-*Agrobacterium* ver.1 showed stronger inhibition of ethylene evolution and higher T-DNA transfer frequencies than chemical treatments in melon and wild water melon (Nonaka et al., 2008a; Malambane et al., 2018). For the further improvement of Super-*Agrobacterium* ver.1, a stronger promoter was used to drive *acdS*. In Super-*Agrobacterium* ver.1, the expression of *acdS* gene was under the control of the *lac* promoter, which shows constitutive expression in *A. tumefaciens*. Instead of the *lac* promoter, the *virD* promoter was cloned from *A. tumefaciens* and was used in Super-*Agrobacterium* ver. 2 (Someya et al., 2013). *virD* genes are induced by acetosyringone at pH 5.2, which is the co-cultivation condition. This promoter showed higher gene expression levels than the *lac* promoter during co-cultivation (Someya et al., 2013), resulting in higher T-DNA transfer frequencies in Super-*Agrobacterium* ver. 2 than in ver.1 (Someya et al., 2013).

Gamma-aminobutyric acid (GABA), an amino acid, was determined to be another negative factor in *Agrobacterium*plant interactions (Chevrot et al., 2006; Haudecoeur et al., 2009; Nonaka et al., 2017). GABA is taken up into *A. tumefaciens* and triggers the degradation of the quorum-sensing (QS) signal, resulting in the reduced horizontal gene transfer of the Ti plasmid and the aggressiveness of the plant host (Chevrot et al., 2006; Haudecoeur et al., 2009). GABA is a biologically active agent in animals, plants, and bacteria. In animals, GABA is particularly well known as an effector, lowering blood pressure (Elliott and Hobbiger, 1959; Takahashi et al., 1961; Takahashi et al., 1995), and its mechanism of action has been well studied. It was found that some chemical compounds control GABA effect in animals. While contrarily, in plants, GABA was know as modulator cell elongation, abiotic stress and pathogen attack (Park et al., 2010; Renault et al., 2011; Shelp et al., 2012; Forlani et al., 2014), but the action mechanisms of GABA in-plants are still to be clarified, and the chemical compounds related with GABA perception or signal transduction in plants have not been identified. Some bacteria are known to harbor GABA transaminase (*gabT*), a GABA degradation enzyme. Utilization of GabT activities increased the transient and stable transformation frequencies in tomato and grass plants [Nonaka et al., 2017 (Super-*Agrobacterium* ver.3)]. Super-*Agrobacterium* ver. 3 was also effective in the agroinfiltration method (Hoshikawa et al., 2019; Knoch et al., 2019).

Stable transformation techniques are important as they are used for breeding GM crops. Transient transformations are also widely used in plant science; for example, protein production by excessive gene overexpression and gene function analysis by the virus-induced gene silencing (VIGS) system (Velásquez et al., 2009), are based on transient gene transfers. However, some plants have significantly lower transient gene transfer rates, creating a limitation in plant science research that should be resolved by increasing the transient transformation (T-DNA transfer) frequency in a wide variety of plant hosts. Therefore, the host range of *A. tumefaciens* must be enlarged, and its transformation efficiency increased. In this study, to further increase the transient and/or stable transformation frequencies, Super-*Agrobacterium* was updated to ver. 4 by introducing both AcdS and GabT activity to the GV2260 strain, which has similar abilities, compared with other strains such as EHA105, EHA101, LBA4404, and MP90 (Sun et al., 2006 and Chetty et al., 2013). The abilities of the Super-*Agrobacterium* ver. 4 were evaluated in *Erianthus ravennae*, *Solanum lycopersicum* "Micro-Tom," *Nicotiana benthamiana*, and *S. torvum* for both transient and stable transformations.

#### MATERIALS AND METHODS

#### Bacterial Strains, Vectors, and Culture Conditions

All bacterial strains and vectors, which were used in this study, were listed up in **Table 1**. The vector maps were described in

#### TABLE 1 | List of *A. tumefaciens* strains and plasmids that are used in this study.


**Supplemental Figure 1**. *A. tumefaciens* strains were grown at 28°C in Luria Broth (LB) medium (1% bacto-tryptone, 0.5% yeast extract, and 0.5% NaCl). Antibiotics were added at the following final concentrations: ampicillin at 100 µg/ ml, gentamicin at 50 µg/ml, spectinomycin at 50 µg/ml, and kanamycin at 50 µg/ml. *A. tumefaciens* strains were then cultured on solid LB medium at 28°C for 2 days. A single colony was picked and cultured in 2 ml of LB medium at 28°C and 200 rpm for 2 days until the pre-culture reached the stationary phase. From this, 15 µl of culture was harvested and added to 15 ml of LB medium and cultured at 28°C and 200 rpm. When the O.D.600 of the culture reached 0.7 to 0.9, the cells were then centrifuged, collected, and checked for enzymatic activity. For transformations, the pelleted bacterial cells were resuspended in liquid Murashige and Skoog (1962) (MS) containing 30 g/l glucose, and 500 µM acetosyringone at pH 5.2. The cell density was then adjusted to 0.4–0.5 at O.D.600.

#### Construction of *acdS* and *gabT* Expression Plasmids

The *gabT* gene was cloned from pBBRgabT in a previous study (Nonaka et al., 2017) by polymerase chain reaction (PCR) with the primers acdSF (5′- TCTGCGCGTAATCT GCTGCTTGAGCGCAACGCAATTAATG -3′) and gabTR (5′- CGATTCTGGACTACTGCTTCGCCTCATCAAAAC-3′). The transcription terminal sequence of the ampicillin resistance gene was cloned from the pUC18 vector using PCR with the primer's amp\_ter-for2 (5′-GCTAGAATTCCTGTCAGACCA

AGTTTACTC-3′) and amp\_ter-rev2 (5′-CATTAATTG CGTTGCGCTCAAGCAGCAGATTACGCGCAGA-3′). Then, the two fragments were combined by fusion-PCR with the primer's amp\_ter-for2 and gabTR. The ligated fragment was inserted into pBBRacdS (Nonaka et al., 2008a) and digested with *Eco*RI and *Xba*I (New England Biolabs, Hirchin, UK). The expression of both genes was under the control of the *lacZ* gene promoters (pBBRacdSgabT, **Figure 1A**). pBBR1MCS5, pBBRacdS, pBBRgabT, and pBBRacdSgabT were introduced into *A. tumefaciens* GV2260 (pEKH2-nosPNPTII-ubiPGUS-35SPHPT; pEKH2), *A. tumefaciens* GV2260 (pIG121-Hm), or *A. tumefaciens* GV2260 (pEAQ-GFP-HT) *via* electroporation.

#### ACC Deaminase Activity Assay

Cells were collected and washed twice with 100 mM Tris–HCl (pH 8.5) and resuspended in 1.5 ml of lysate buffer. The cells were lysed on ice by sonication and centrifuged at 5,000 × *g* at 4ºC for 15 min. The AcdS activity was measured according to a modified protocol based on that of Honma and Shimomura (1978). The AcdS activity was measured spectrophotometrically at 340 nm. The protein content of the extracts was determined using the Bradford method (Bradford, 1976).

#### GABA Transaminase Activity in *A. tumefaciens*

The pellet of *A. tumefaciens* was re-suspended in 100 µl of BugBuster Master mix (Novagen, MA, USA) for lysate

FIGURE 1 | Effect of ACC deaminase and GABA transaminase activity on the transfer of T-DNA. (A) Map of a plasmid for the expression of ACC deaminase (*acdS*) and GABA transaminase (*gabT*) in *A. tumefaciens*. *lacP*, lac gene promoter from *E. coli; acdS*, ACC deaminase gene from *Psedomonas* sp (Sheehy et al., 1991, Accession No. M73488); *gabT*, GABA transaminase gene from *E. coli* (Accession No. CP040667); pBBR1 Rep, replication protein for the broad-host-range plasmid pBBR1 from *Bordetella bronchiseptica*; pBBR1 oriV, replication origin of the broad-host-range plasmid pBBR1 from *B. bronchiseptica*; pBBR1 Rep, protein for replication required by pBBR1 oriV, *GmR*, Gentamicin resistance gene. (B) Growth curve of *A. tumefaciens*. Open and solid circles represent *A. tumefaciens* C and V1, respectively. Open and solid squares represent V3 and V4, respectively. (C) ACC deaminase activity in *A. tumefaciens*. (D) Detection of GABA transaminase activity in *A. tumefaciens*. C, GV2260 (pBBRMCS1-5); V1, *A. tumefaciens* GV2260 (pBBRacdS); V3, *A. tumefaciens* GV2260 (pBBRgabT); V4, *A. tumefaciens* GV2260 (pBBRacdSgabT). Bars represent the standard deviation (n = 3). Different characters indicate values that were statistically different in the one-way ANOVA and Tukey-Kramer method, multiple comparison method (*P* < 0.01).

preparation. The protein concentration of the lysate was measured with a BCA Protein Assay Kit (Novagen, MA, USA). The amount of protein was adjusted to 100 µg *per* reaction mixture. The reaction mixture contained 0.1 M bicine–NaOH, 0.1 M pyridoxal phosphate, 10 mM 2-ketoglutarate, 10 mM GABA, and a proteinase inhibitor cocktail. The reaction mixture was incubated at 37°C. GabT metabolizes GABA to glutamate; therefore, to estimate the GabT activity, we detected the glutamate concentration in the reaction mixture using a Yamaki glutamate assay kit (Yamaki, Tokyo, Japan) (Akihiro et al., 2008).

#### T-DNA Transfer Assay in *E. ravennae* and *S. lycopersicum*

Calli of *E. ravennae* were kindly provided by Prof. Masahiro Mii from Chiba University, Japan. The calli were induced directly from the seeds on MS medium, containing 1 g/l casamino acids, 2 mg/l 2,4-dichlorophenoxyacetic acid (2,4-D), 0.2 mg/l 6-benzylaminopurine (BAP), 30 g/l 4-O-α-D-glycopyranosyl-Dglycopyranose (maltose H) (Wako, Tokyo, Japan), and 3% Gelrite (Wako, Tokyo, Japan), were subcultured for 2 weeks before inoculation. After co-cultivation, the β-glucuronidase (GUS) activity of the *E. ravennae* calli were assayed histochemical with X-Gluc buffer containing 100 mM phosphate buffer, 10 mM EDTA, 2.5 mM potassium ferricyanide, 2.5 mM potassium ferrocyanide, 0.1% Triton X-100, and 0.5 mg/l X-glucuronide. The GUS-stained calli were observed using a stereoscopic microscope (Leica: MX FLIII, DFC300 FX, Application Suite, Leica, Germany), the number of GUS-stained spots per 1 g of calli calculated, and the T-DNA transfer efficiency was estimated, based on the relative number of GUS spots.

Tomato seeds were washed with 70% ethanol for 10 s, sterilized with 5% hypochlorous acid containing 10% Triton X-100 for 45 min, and washed three times with sterilized water. After the third wash, the seeds were kept in water for 2 days. The sterilized tomato seeds were sown on MS medium, containing 15 g/L sucrose (Wako, Tokyo, Japan) and 0.3% Gelrite (Wako, Tokyo, Japan). Cotyledons from 7-day-old tomato seedlings were cut into four pieces and used to generate two locations for inoculations with *A. tumefaciens*. Thirty explants were subjected to each treatment. The inoculated explants were cultured on co-cultivation medium (pH 5.2) containing MS salts, 30 g/L glucose, 500 µM acetosyringone, and 0.3% Gelrite (Wako, Tokyo, Japan) at 25°C, for 3 days in the dark. After 3 days of co-cultivation, the tomato explants were assayed histochemically for GUS activity with X-Gluc buffer, described above. GUS stained tomato cotyledon explants were observed and images were taken using a stereoscopic microscope system (Leica: MX FLIII, DFC300 FX, Application Suite, Leica, Wetzlar, Germany). The GUS stained areas were converted into numerical values by Image J (National Institutes of Health: http://rsbweb.nih.gov/ ij/) and the percentage of GUS stained area for each explant was calculated. According to the results, GUS stained tomato explants were categorized into 4 classes: (Class 1) less than 5%, (Class 2) 5–10%, (Class 3) 10–20%, and (Class 4) more than 20%. To estimate the T-DNA transfer, the frequency of more than 20% was calculated.

#### Tomato Stable Transformation

Tomato transformations followed the protocol by Sun et al. (2006). In brief, after 3 days of co-cultivation, tomato cotyledon segments were placed on a callus-induction medium [MS medium containing 0.3% Gelrite (Wako, Tokyo, Japan), 1.5 mg/l zeatin, 100 mg/l kanamycin, and 375 mg/l Augmentin (GlaxoSmithKline, London, UK)] for 4 weeks. Calli that formed segments were cultured on shoot-induction medium [MS medium containing 0.3% Gelrite (Wako, Tokyo, Japan), 1.0 mg/l zeatin, 100 mg/l kanamycin, and 375 mg/l Augmentin (GlaxoSmithKline, London, UK)] for 4 weeks. The shoots were then placed on rooting medium, which consisted of half-strength MS medium, 0.3% Gelrite (Wako, Tokyo, Japan), 100 mg/L kanamycin, and 375 mg/l Augmentin, for 2 weeks. Tissues were each subcultured for 10–14 days.

### Agro Infiltration Method

*A. tumefaciens* GV2260 carrying pEAQ-GFP-HT (Sainsbury et al., 2009) was grown in LB media, resuspended in 10 mM MgCl2, 10 mM MES, pH 5.6, 150 µM acetosyringone, and incubated for 3 h at room temperature. The leaves were then syringe infiltrated with the *A. tumefaciens*. Concentrations of *A. tumefaciens* were 0.3 at O.D.600 for *N. benthamiana* and 1 at O.D.600 for *S. torvum*. GFP fluorescence was detected 3 and 5 days after infiltration for *N. benthamiana* and *S. torvum*, respectively. Each experiment was repeated three times.

### Ploidy Analysis

The ploidy of the rooting shoots was checked with flow cytometry. One square centimeter of leaf was cut from the rooting shoots, chopped, and added to 250 µl of nucleus-extraction solution (CyStain UV Precise P, Sysmex, Hyogo, Japan). To purify the nucleus-extraction solution, 1 mm2 mesh was used. After purification, 1 ml of staining solution (CyStain UV Precise P, Sysmex, Hyogo, Japan) was added and incubated for 1 min. This solution was applied to an Attune focusing analyzer (ABI, CA, USA), and 2n plants were selected. The 2n plants were then planted on solid medium and acclimatized.

### Southern Blot Analysis

Genomic DNA was extracted from young tomato leaves using Maxwell 16 System DNA Purification kits (Promega, WI, USA). The purified DNA was digested with *Hin*dIII, electrophoretically separated in 0.8% agarose gel, and transferred onto Gene Screen Plus nylon membranes (Roche Diagnostics, Basel, Swiss) with 20× saline–sodium citrate (SSC) buffer. After ultraviolet (UV) crosslinking, the membranes were hybridized in a solution containing 7% sodium dodecyl sulfate (SDS), 50% deionized formamide, 50 mM sodium phosphate (pH 7.0), 2% blocking solution, 0.1% N-lauroylsarcosine, 0.75 M NaCl, and 75 mM sodium citrate at 42ºC overnight. For hybridization, a digoxigenin (DIG)-labeled DNA probe, specific for *nptII* (0.8 Kb), was used. A DIG-labeled probe was generated by DIG-High Prime, and the DIG signal was detected according to the manufacturer's protocol (Roche Diagnostics, Basel, Swiss).

#### Statistical Analysis

The average values were obtained from three biological replicates. One-way analysis of variance (ANOVA) and Tukey Kramer's multiple range test, with *P* < 0.01 or *P* < 0.05, were carried out to determine the significant differences. Statistical analyses were carried out using the SAS statistics program (version 8.0, SAS Institute Cary, NC, USA).

#### RESULTS

#### Introduction of AcdS and GabT Activity Did Not Affect Bacterial Growth

Since ethylene and GABA suppress the transfer of T-DNA in different ways, we predicted that the introduction of AcdS and GabT activity into *A. tumefaciens* would be effective at increasing the T-DNA transfer. These two genes were introduced by pBBR1MCS-5, the broad host range plasmid (Kovach et al., 1995) (**Figure 1A**, pBBRacdSgabT), and expressed under the control of the *lac* promoter (Nonaka et al., 2008a; Nonaka et al., 2017). To estimate whether these two genes affect bacterial growth or not, growth curves were compared for the four strains [(C), (V1), (V3), and (V4)]. In all strains, the accelerated growth period began 10 h after culturing, and after 18 to 26 h, the logarithmic growth phases were observed (**Figure 1B**). These results indicate that introducing the *acdS* and *gabT* at the same time in *A. tumefaciens* did not affect its bacterial growth. To measure the AcdS and GabT activity, cells were collected by centrifugation, and the lysate was prepared. Then, the AcdS and GabT activity were measured, as described in previous studies (Nonaka et al., 2008a; Nonaka et al., 2017). Both activities were detected in the V4 strain, but the AcdS and GabT activities in V4 were one-third of the V1 and V3, respectively (**Figures 1C**, **D**).

#### Evaluation of the Super-*Agrobacterium* for T-DNA Transfer in Plants

To examine whether the AcdS and GabT activities were enough to increase the transfer of T-DNA, the T-DNA transfers in *E. ravennae*  and *S. lycopersicum* "Micro-Tom" were observed. *E. ravennae* is known for its high bio-mass production and is relevant for practical agriculture. After 3 days of co-cultivation, the number of blue spots were counted to evaluate the T-DNA transfer in *E. ravennae*. Four strains, (C-E), (V1-E), (V3-E), and (V4-E) were used for the transformation. The V1-E, V3-E, and V4-E showed higher T-DNA transfer frequencies than the control, C-E. The inoculation of V4-E increased the T-DNA transfer frequency by 7.2, 2.4, and 1.7 times, compared to the C-E, V1-E, and V3-E, respectively (**Figure 2A**). Next, we evaluated the abilities of Super-*Agrobacterium* V4 using *S. lycopersicum* "Micro-Tom." Almost 100 explants of "Micro-Tom" were inoculated for each bacterial strain [(C-G), (V1-G), (V3-G), and (V4-G)]. The *uidA* gene was used as an indicator of T-DNA transfer, and the blue area indicated transformed cells (**Figure 2B**). The GUS-stained area was determined in each of the explants with Image J, as described in the *Materials and Methods* section (in "2.5 T-DNA transfer assay in *E. ravennae* and *S. lycopersicum*"). The degree of staining was categorized into 4 classes (**Figure 2B**). To evaluate the ability of the T-DNA transfer in C-G, V1-G, V3-G, and V4-G, the frequency of class 4 was compared. V4-G showed the highest frequency of class 4; the frequencies were 3.9, 1.4, and 1.5 times higher than the C-G, V1-G, and V3-G, respectively. V1-G

FIGURE 2 | Transient transformations in tomato *via* tissue culture and co-cultivation method. (A) Occurrence of T-DNA transformations in *E. ravennae*. The number of GUS-stained spots *per* 1 g of *E. ravennae* calli were counted for each treatment. The bars indicate the standard deviation (n = 3). Different characters indicate values that were significantly different according to the one-way analysis of variance, multiple comparison method (*P* < 0.01). C-E: *A. tumefaciens* GV2260 (pBBR1MCS-5, pEKH2), V1-E: *A. tumefaciens* GV2260 (pBBRacdS. pEKH2), V3-E: *A. tumefaciens* GV2260 (pBBRgabT, pEKH2), V4-E: *A. tumefaciens* GV2260 (pBBRacdSgabT, pEKH2). (B) GUS stained explants of *S. lycopersicum* "Micro-Tom." Explants were prepared from 7 days old seedlings. After 3 days of co-cultivation, explants were stained. Classification of GUS-stained cotyledon explants. GUS stained tomato cotyledons were categorized depending on the stained area. Categorized into 4 classes: (Class 1) less than 5%, (Class 2) 5–10%, (Class 3) 10–20%, and (Class 4) more than 20%. (C) Appearance ratio Class 4 in tomato cotyledons. C-G: *A. tumefaciens* GV2260 (pBBRMCS1-5, pIG121-Hm); V1-G: *A. tumefaciens* GV2260 (pBBRacdS, pIG121-Hm); V3-G: *A. tumefaciens* GV2260 (pBBRgabT, pIG121-Hm); V4-G: *A. tumefaciens* GV2260 (pBBRacdSgabT, pIG121-Hm). Bars represent the standard deviation (n = 3). Different characters indicate values that were statistically different in a one-way ANOVA and the Tukey-Kramer method, multiple comparison method (*P* < 0.01). (D) Transient transformation *via* agroinfiltration methods on *N. benthamiana.*  (E) Transient transformation *via* Agroinfiltration methods on *S. torvum.* GFP signals were used as indicators of transformation. C-Q: *A. tumefaciens* GV2260 (pBBRMCS1-5, pEAQ-GFP-HT); V1-Q: *A. tumefaciens* GV2260 (pBBRacdS, pEAQ-GFP-HT); V3-Q: *A. tumefaciens* GV2260 (pBBRgabT, pEAQ-GFP-HT); V4-Q: *A. tumefaciens* GV2260 (pBBRacdSgabT, pEAQ-GFP-HT).

and V3-G showed almost the same levels (**Figure 2C**). Therefore, the activities of AcdS and GabT in V4-E and V4-G, were enough to increase the T-DNA transfer frequencies in *E. ravennae* and *S. lycopersicum* "Micro-Tom." Additionally, we evaluated the ability of T-DNA transformation with V4 in the Agroinfiltration method. Plasmid pEAQ-GFP-HT was used as a binary vector in Agroinfiltration method. *N. benthamiana* and *S. torvum* were inoculated with four strains [(C-Q), (V1-Q), (V3-Q), and (V4- Q)]. The strength of the GFP signal was used as an indicator of the frequency of the T-DNA transfer. In *N. benthamiana*, V4-Q induced higher GFP expression than the C-Q strain and V1-Q (**Figure 2D**), but it was the same level as the V3-Q. In *S. torvum,* the success of the V4-Q strain with the Agroinfiltration treatment was greater than that of the C-Q strain, but the same as that of the V1-Q and V3-Q strains (**Figure 2E**).

#### *A. tumefaciens* With Both AcdS and GabT Activities Resulted in the Enhanced Stable Transformation of Tomato

V4 was effective at the T-DNA transfer in *E. ravennae* and *S. lycopersicum* "Micro-Tom" (**Figures 2A**, **C**), however this is one step of the stable transformation process. The entire process for *Agrobacterium*-mediated stable transformation is divided into four steps: i) T-DNA transfer and integration into the plant genome, ii) calli induction, iii) the regeneration of the shoots, and iv) rooting. It was not clear if the V4 affected these other steps. To ascertain this information, the frequency of each process was observed in *S. lycopersicum* "Micro-Tom," which has a well-established regeneration system for processes (ii) to (iv) (Sun et al., 2006). To characterize each strain, C-G, V1-G, V3-G, and V4-G were used. One-month after inoculation, the calli inductions were observed. All Super-*Agrobacterium* strains increased the callus inductions compared with the C-G (**Figures 3A**–**D**). V1-G and V4-G showed slightly higher calli induction ratios (calli number / segments number) than the V3-G. The C-G, V1-G, V3-G, and V4-G showed calli induction frequencies of 51.5 ± 0.6, 85.2 ± 8.8, 73.8 ± 2.03, and 91.8 ± 3.7%, respectively (**Figure 3E**, **Table 2**). Shoot regeneration ratios (shooting number / calli number) were increased with the inoculation of the V3-G and V4-G. V1-G was slightly higher than that of the C-G. The frequency of the shoot regeneration ratios with the inoculations of C-G, V1-G, V3-G, and V4-G were 49.7 ± 10.9, 80.4 ± 29.2, 181.8 ± 23.4, and 176.3 ± 58.9%, respectively (**Figure 3F**, **Table 2**). The frequencies for rooting from the shoots (rooting number/ shoots number) were similar for all strains (**Figure 3G**, **Table 2**). These results suggest that V1-G had positive effects on step (ii) calli induction, V3-G increased step (iii) shooting, and V4-G accelerated both steps (ii) and (iii) in the *Agrobacterium*mediated stable transformation process. After regenerated diploid shoots (2n) were selected, the exogenous T-DNA was detected by PCR (data not shown) and Southern hybridization analysis (**Supplemental Figure 2**). The stable transformation frequencies were evaluated with single-copy-number plants, and identification by Southern hybridization analysis. These results imply that all of the lines we obtained were independent and did not contain a cloned plant. The C-G, V1-G, V3-G, and V4-G showed the stable transformation efficiencies of 4.3 ± 1.9, 9.7 ± 0.4, 9.8 ± 1.6, and 15.2

± 1.1%, respectively (mean ± SD of three repetitions) (**Figure 3H** and **Table 2**). Thus, V4 exhibited approximately 3.6, 1.6, and 1.6 times the stable transformation frequency of C-G, V1-G, and V3-G, respectively. The frequency of regenerated rooting shoot with single copy was same level in all *A. tumefaciens* strains (**Figure 3I**).

#### DISCUSSION

*A. tumefaciens* with AcdS and GabT was expected to cause reduced ethylene and GABA content in plants. Indeed, the ethylene levels in the plant tissues during the transformation were reduced by the *A. tumefaciens* with AcdS activity (Nonaka et al., 2008a; Malambane et al., 2018). On the other hand, significant differences in GABA content during the co-cultivation were not observed between *A. tumefaciens* with GabT activity and the control (data not shown). As *A. tumefaciens* takes the GABA from the plant into the bacterial cell through a kind of ABC transporter (Planamente et al., 2010; Planamente et al., 2013), if the GabT activity introduced *A. tumefaciens*, GABA taken into bacterial cell would be degraded. The degradation of GABA occurred only in bacterial cells, and at very localized areas. Therefore, it was difficult to detect the differences of GABA content. ACC deaminase activity and GabT activity in *A. tumefaciens* were effective at increasing the T-DNA transfer frequency (Nonaka et al., 2008a; Nonaka et al., 2017; Ntui et al., 2010; Hao et al., 2010), but it was not clear which was more effective in *Agrobacterium*-plant interactions. To ascertain this, the T-DNA transfer abilities of Super-*Agrobacterium* ver. 1 and ver. 3 were compared in *E. ravennae* and tomato with the tissue culture and co-cultivation methods. No differences were observed between the strains in *E. ravennae* and *S. lycopersicum* "Micro-Tom" (**Figures 2A**, **C**). With the Agroinfiltration method, the same tendency was observed in *S. torvum* (**Figure 2E**). These results mean that ethylene and GABA influence the T-DNA transfer frequencies at almost the same level in these plant species. On the other hand, in *N. benthamiana*, Super-*Agrobacterium* ver. 3 and ver. 4 showed higher level of T-DNA transfer than GV2260 and Super-*Agrobacterium* ver. 1, but the level of T-DNA transfer was same in ver. 3 and ver. 4. This showed that in the *N. benthaminana*, AcdS activity did not improve the T-DNA transfer, but GabT activity was effective at increasing the T-DNA transfer. Therefore, in *N. benthamiana*, GABA is a stronger negative factor than ethylene. From these results, the effect of the Super-*Agrobacterium* was found to be different, dependent on the plant species, thus the selection of the most suitable strain is important for the successful application of the technology.

Even under conditions where the *vir* gene is sufficiently expressed, our Super-*Agrobacterium* strains could further improve T-DNA transfer. In this study, with the tissue culture and co-cultivation methods, 500 µM of acetosyringone, which was enough to induce *vir* gene expression (Nonaka et al., 2008b), was used during co-cultivation. Super-*Agrobacterium* ver. 1, ver. 3, and ver. 4 further increased the T-DNA transfer frequency, despite the existence of enough *vir* gene inducers (**Figures 2A**, **C**). The additional effects of AcdS and GabT under the acetosyringone indicate that in the T-DNA transfer, the ethylene and GABA affect


TABLE 2 | Effect of the Super-Agrobacterium ver.1, ver.3, and ver.4 on plant regeneration and transformation of the 'Micro-Tom' cotyledons.

*Each column "Segments," "Calli," "Shoots," "Rooting," "Diploid," and "Single copy" indicated the number of occurrences. "Segments" means the number of segments used for inoculation of A. tumefaciens. "Calli," "Shoots," and "Rooting" showed the number of regenerate calli, shoot, and rooting. "Diploid" was the number rooting shoots that were diploid, detected by the ploidy analyzer. "Single copy" means the number of diploid rooting shoots with single copy of T-DNA, identified by Southern blot analysis. Transformation frequency was calculated as follows: the total number of transgenic plants with diploid and single copy was divided by the number of explants inoculated and then multiplied by 100. Only one plant regenerated per cotyledon explant was considered to calculate transformation efficiency. Three replications were done for this experiment.*

different from the level of *vir* gene inducer. Previous studies have demonstrated that GABA was independent of *vir* gene expression (Chevrot et al., 2006; Haudecoeur et al., 2009). Thus, the inhibition of GABA further increased the transformation. On the other hand, research has shown that the ethylene target points are involved with *vir* gene expression, and the ethylene perceiving plant would reduce *vir* gene inducers or release antagonists of the *vir* gene inducers (Nonaka et al., 2008b). If the target point of the ethylene was the reduction of the *vir* gene inducer, the effect of the Super-*Agrobacterium* ver. 1 would be masked by acetosyringone. In fact, our results showed that the Super-*Agrobacterium* ver. 1 increased the transformation frequency up to 3.2 and 2.8 times in *E. ravennae* and *S. lycopersicum*, respectively, even with the application of the *vir* gene inducers. Therefore, these results suggest that the target point of ethylene is not the reduction of *vir* gene inducers, but the suppression of the antagonists.

In Super-*Agrobacterium* ver. 4, the enzymatic activity was one third of the Super-*Agrobacterium* ver. 1 and ver. 3, but it was effective in *E. ravennae* and *S. lycopersicum* "Micro-Tom" (**Figures 2A**, **C**). The amount of AcdS protein in the Super-*Agrobacterium* ver. 4 was one third of that found in Super-*Agrobacterium* ver. 1 (**Supplemental Figure 3**). Expressing multiple genes using the same promoter may reduce the expression levels of each gene. In this study, we used the lac promoter to drive both the *acdS* and *gabT* genes. If stronger promoters were used, the expression levels would be increased. Previous research compared the *vir* gene promoters (*virB*, *virC,* and *virD*) with the *lac* promoter activities; the *vir* gene promoters were found to show higher promoter activities than the *lac* promoter (Someya et al., 2013). Therefore, using these promoters would be effective to increase *acdS* and *gabT* expression in the Super-*Agrobacterium* ver. 4. Indeed, replacement of the promoter increased the *acdS* gene expression and the activity in *A. tumefaciens*, resulting in increased T-DNA transformation frequencies (Super-*Agrobacterium* ver. 2) (Someya et al., 2013). Therefore, replacing promoters would increase the transient transformations in *S. torvum via* the agroinfiltration method.

In this study, the stable transformation frequency was 15.2 ± 1.1%. This value was calculated as the ratio between independently transformed plants with diploid and single copy number in soil (checked by a ploidy analyzer and Southern blot analysis) and the total number of explants infected with *A. tumefaciens*. Previous studies have reported transformation frequencies that differ from ours (Sun et al., 2006; Khoudi et al., 2009; Khuong et al., 2013; Chetty et al., 2013). The transformation frequency might depend on the bacterial strain, binary vector and the selection method. In our study, the transformation frequency was calculated using regenerated rooting shoot with diploid and a single copy *per* inoculated segment. In contrast, most previous studies calculated this frequency from the PCR-positive tomatoes (Sun et al., 2006; Khoudi et al., 2009; Khuong et al., 2013; Chetty et al., 2013). Therefore, it would be difficult to compare between our and previous results. To create a transgenic plant, it is important to avoid somaclonal variation and multiple copies, which our method did.

The process of *Agrobacterium*-mediated stable transformation contains four steps: i) the T-DNA transfer into plant cells and integration into the host genome, ii) callus induction, iii) the regeneration of shoots, and iv) rooting. The activity of AcdS and GabT increased step (i) (**Figures 2A**, **C**). In the rooting step, there were no significant differences detected between them. Other steps showed different responses to AcdS and/or GabT activity (**Figure 3** and **Table 2**). The AcdS activity showed higher callus induction frequencies than the GabT activity, whereas the GabT activity induced higher shoot regeneration ratios than the AcdS. The browning callus appearance was suppressed by *A. tumefaciens* with AcdS (**Figures 3A**–**D**), as the ethylene induced hypersensitive responses and programmed cell death (Wang et al., 2017); infection of *A. tumefaciens* with AcdS (Super-*Agrobacterium* ver. 1 or ver. 4) with the ability to remove ethylene, suppressed the browning phenomena. Although the function of GABA in plants needs to be clarified, there have been several studies regarding the functions of GABA as a signaling compound in plant growth and development. The increased endogenous concentrations of GABA seem to be the reason for impaired cell elongation in the *Arabidopsis thaliana* mutants, *pop2,* and *her1*, and the corresponding phenotypes (Renault et al., 2011). Infection of *A. tumefaciens* with GabT activity locally decreased GABA content in the plant calli and maintained higher shoot regeneration frequencies. Since both activities have different effective points in the transformation process, Super-*Agrobacterium* ver. 4. with AcdS and GabT activity at the same time, enhanced the stable transformation frequency approximately 3.6 times, compared with that of the original GV2260 strain.

We succeeded in producing an *A. tumefaciens* strain with improved potential for transformation by imbuing it with the ability to remove ethylene and GABA, which are negative factors in the *Agrobacterium*-plant interactions. *A. tumefaciens* with AcdS and GabT increased the T-DNA tranfer and stable transformation frequency. Especially in tomato, this newly bred bacterium (Super-*Agrobacterium* ver. 4) enables us to decrease the number of cotyledons used for transformation and allows us to reduce 72% of the time and labor required for transformation. Moreover, because our system was the plasmid with *acdS* and *gabT* gene, it is used in combination with other strains, such as the EHA105, EHA101, LBA4404, MP90, and AGL1. Based on this, we conclude that this new system is a useful tool for plant genetic engineering. On the other hand, the frequency is still not enough depending on the plant species and cultivars (**Figures 3E**, **F**). Therefore, the additional effort should have been required to adapt *Agrobacterium*-mediated transformation for a wide variety of plants. Other negative factors in *Agrobacterium*plant interactions, aside from ethylene and GABA, have been reported by previous studies (Liu and Nester, 2006; Yuan et al., 2007; Yuan et al., 2008; Anand et al., 2008). Therefore, to expand plant spices and cultivars adapting *Agrobacterium*-mediated transformation, multiply suppress of these negative factors would be also effective.

#### DATA AVAILABILITY STATEMENT

All datasets GENERATED for this study are included in the manuscript/**Supplementary Files**.

### AUTHOR CONTRIBUTIONS

SN designed the experiments, analyzed the data, and wrote the manuscript. TS constructed the plasmid pBBRacdSgabT and did the western blot analysis. YK performed the experiments about Agroinfiltration. HE and KN critically revised and approved the manuscript for publication.

#### FUNDING

This research was supported by grants from the New Energy and Industrial Technology Development Organization (NEDO) to HE and from JSPS KAKENHI (Grant Numbers JP24780001 and JP19K05964) to SN. Cooperative Research Grant from Plant Transgenic Design Initiative (PTraD), Gene Research Center in Tsukuba Innovation Plant Research Center (T-PIRC), University of Tsukuba, Japan supported this research.

#### ACKNOWLEDGMENTS

We appreciate the help of Prof. Mii (Chiba University, Japan) for kindly providing the *E. ravennae* calli. We also thank Prof. Nakamura (Chiba University, Japan) and Prof. Mitsui (Tohoku University, Japan) for the gift of the pEKH2 plasmid and pBBR1MCS-5, respectively. We would like to thank Editage (www.editage.jp) for English language editing.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01204/ full#supplementary-material

SUPPLEMENTAL FIGURE 1 | Map of binary vectors. (A) Map of pEKH2 nosPNPTII-ubiPGUS-35SPHPT. NosP; Nopalin synthesis gene promoter, UbiP; ubiquitin gene promoter from rice, NosT; Nopalin synthesis gene terminator, *nptII*; neomycin phosphotransferase gene, *uidA*; beta-glucuronidase gene, *hptII*; hygromycin phosphotransferase gene, OriV; replication origin V (IncPα, plasmid RK2 from *E. coli*, GeneBank accession #J01780), *KanR*; Kanamycin resistance gene. RB; Right border sequence, LB; Left border sequence. (B) Map of pIG121-Hm. NosP; Nopalin synthesis gene promoter, CaMV 35S P; Cauliflower mosaic virus 35S promoter, NosT; Nopalin synthesis gene terminator, *nptII*; neomycin phosphotransferase gene, *uidA*; beta-glucuronidase gene, *hptII*; hygromycin phosphotransferase gene, OriV; replication origin V (IncPα, plasmid RK2 from *E. coli*, GeneBank accession #J01780), *KanR*; Kanamycin resistance gene, RB; Right border sequence, LB; Left border sequence. (C) Map of pEAQ-GFP-HT. CaMV 35S P; Cauliflower mosaic virus 35S promoter, NosT; Nopalin synthesis gene terminator, GFP; green fluorescence gene, RB; Right border sequence, LB; Left border sequence.

SUPPLEMENTAL FIGURE 2 | Map of the pIG121-Hm vector and the southern blot analysis of the T0 generation. (A) Present the maps of the T-DNA region in the pIG121-Hm expression vectors used for the stable transformation. Red bars represent the position of probes used in the southern blot analysis. *Hin*dIII indicates the restriction enzyme sites that were used in the southern hybridization. NPTII probes were used in. (B) Southern blot analysis of the T0 generation. Red numbers indicate the transgenic lines with single copy. C-G: *A. tumefaciens* GV2260 (pBBRMCS1-5, pIG121-Hm); V1-G: *A. tumefaciens* GV2260 (pBBRacdS, pIG121-Hm); V3-G: *A. tumefaciens* GV2260 (pBBRgabT, pIG121-Hm); V4-G: *A. tumefaciens* GV2260 (pBBRacdSgabT, pIG121-Hm).

SUPPLEMENTAL FIGURE 3 | Western blot analysis of ACC deaminase expression in *A. tumefaciens* strains using the cell lysate at the 'Early stage' (O.D.600 0.7). ACC deaminase was probed with an anti-ACC deaminase antibody. Coomassie Brilliant Blue staining (bottom panel) is shown as an internal control. C-G: *A. tumefaciens* GV2260 (pBBRMCS1-5, pIG121-Hm); V1-G: *A. tumefaciens* GV2260 (pBBRacdS, pIG121-Hm); V3-G: *A. tumefaciens* GV2260 (pBBRgabT, pIG121-Hm); V4-G: *A. tumefaciens* GV2260 (pBBRacdSgabT, pIG121-Hm).

### REFERENCES


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Nonaka, Someya, Kadota, Nakamura and Ezura. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

## Red to Brown: An Elevated Anthocyanic Response in Apple Drives Ethylene to Advance Maturity and Fruit Flesh Browning

*Richard V. Espley1, Davin Leif1, Blue Plunkett1, Tony McGhie2, Rebecca Henry-Kirk1, Miriam Hall1, Jason W. Johnston3, Matthew P. Punter3, Helen Boldingh4, Simona Nardozza1, Richard K. Volz3, Samuel O'Donnell1 and Andrew C. Allan1,5\**

*1 Plant & Food Research, Auckland, New Zealand, 2 Plant & Food Research, Palmerston North, New Zealand, 3 Hawke's Bay Research Centre, Plant & Food Research, Havelock North, New Zealand, 4 Plant & Food Research, Hamilton, New Zealand, 5 School of Biological Sciences, University of Auckland, Auckland, New Zealand*

#### *Edited by:*

*Autar Krishen Mattoo, Agricultural Research Service (USDA), United States*

#### *Reviewed by:*

*Prabodh Kumar Trivedi, National Botanical Research Institute (CSIR), India Ai-Sheng Xiong, Nanjing Agricultural University, China*

*\*Correspondence:*

*Andrew C. Allan andrew.allan@plantandfood.co.nz*

#### *Specialty section:*

*This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science*

*Received: 16 May 2019 Accepted: 06 September 2019 Published: 09 October 2019*

#### *Citation:*

*Espley RV, Leif D, Plunkett B, McGhie T, Henry-Kirk R, Hall M, Johnston JW, Punter MP, Boldingh H, Nardozza S, Volz RK, O'Donnell S and Allan AC (2019) Red to Brown: An Elevated Anthocyanic Response in Apple Drives Ethylene to Advance Maturity and Fruit Flesh Browning. Front. Plant Sci. 10:1248. doi: 10.3389/fpls.2019.01248*

The elevation of anthocyanin contents in fruits and vegetables is a breeding target for many crops. In some fruit, such as tomato, higher anthocyanin concentrations enhance storage and shelf life. In contrast, highly anthocyanic red-fleshed apples (*Malus* x *domestica*) have an increased incidence of internal browning flesh disorder (IBFD). To determine the mechanisms underlying this, 'Royal Gala' cultivar apples over-expressing the anthocyanin-related transcription factor (TF) MYB10 (35S:MYB10), which produces fruit with highly pigmented flesh, were compared with standard 'Royal Gala' Wild Type (WT) grown under the same conditions. We saw no incidence of IBFD in WT 'Royal Gala' but the over-expression of MYB10 in the same genetic background resulted in a high rate of IBDF. We assessed concentrations of potential substrates for IBDF and a comparison of metabolites in these apples showed that anthocyanins, chlorogenic acid, pro-cyanidins, flavon-3-ols, and quercetin were all higher in the MYB10 lines. For the flavol-3-ols subgroup, epicatechin rather than catechin was elevated in MYB10 lines compared with the control fruit. Internal ethylene concentrations were measured throughout fruit development and were significantly higher in 35S:MYB10 lines, and ethylene was detected at an earlier developmental stage pre-harvest. Expression analysis of key genes associated with ethylene biosynthesis (aminocyclopropane-1-carboxylic acid synthase and oxidase; *ACS* and *ACO*) and polyphenol oxidase (*PPO*) showed the potential for increased ethylene production and the mechanism for enhanced PPO-mediated browning. The expression of a transcription factor of the ethylene response factor (ERF) class, *ERF106*, was elevated in red flesh. Analysis of transcriptional activation by MYB10 showed that this transcription factor could activate the expression of apple *ACS*, *ACO,* and *ERF106* genes. Our data show a link between the elevation of anthocyanin-related transcription factors and an undesirable fruit disorder. The accelerated advancement of maturity *via*  premature ethylene induction has implications for the breeding and storage of these more highly pigmented plant products.

Keywords: apple, anthocyanin, flavonoids, ethylene, peroxidase, transcription factors, enzymatic browning, ripening

### INTRODUCTION

Apples have a long history of association with human civilisation and over 8,000 years of domestication. The current domesticated apple (*Malus* x *domestica*) has become a fruit with high economic value with over 83 million tonnes harvested worldwide (FAO, 2018). Novel cultivars continue to be bred, such as redfleshed apples created by crossing wild-red fleshed apples with domesticated varieties (Espley et al., 2007; Volz et al., 2009). While large red-fleshed apples are now available, they suffer from a variety of fruit quality issues. Most notably, red-fleshed apples suffer from an increased incidence in enzymatic browning (Volz et al., 2013).

In fruit development, ripening represents the final stage. During this period, important structural, biochemical, and physiological changes such as net starch degradation, softening of the flesh, changes in aroma and flavour profiles occur (Giovannoni et al., 2017). Additionally, two types of fruit ripening behaviour are present amongst plants, climacteric and non-climacteric (Lelievre et al., 1997). Climacteric fruits usually undergo a large burst of ethylene just before, during, or after a respiratory peak also known as a climacteric rise (Wills et al., 2001). Apples fall under this category. Non-climacteric fruits do not exhibit the large burst of ethylene nor any changes in respiration. Fruits that fall under this category include citrus (Alonso et al., 1995), grape, and strawberry (Giovannoni, 2004).

Apple fruit, such as 'Royal Gala', reach the stage of commercial harvest maturity at around 130 days after full bloom (Janssen et al., 2008), although maturity ranges from around 100 to 190 days after full bloom (DAFB), depending on cultivar and climate. In this phase, cell wall modifying enzymes are produced, causing changes in texture which makes the fruit palatable (Schaffer et al., 2013). This causes cell wall polysaccharides, including pectin and cellulose, to be broken down by cell wall degrading enzymes such as polygalacturonase. Changes in flavour are mainly due to a change in sugar-acid balance, a breakdown of bitter compounds (tannins and flavonoids) and an increase in production of volatiles such as methyl esters (ocimene and myrcene) to a total of at least 34 esters upon ripening in 'Royal Gala' (Young et al., 2004; Defilippi et al., 2009). Apple also undergoes a colour change as a result of the reduced production of chlorophyll and its degradation causing the unmasking of pigments that were already previously formed and synthesis of new pigments (carotenoids or anthocyanins) (Ferrer et al., 2005).

Browning (enzymatic browning) is a reaction that often occurs in fruit at the end of ripening and the beginning of over-ripening (Murata et al., 1995). Apple fruit, because of their high phenolic content, are highly susceptible to browning (Holderbaum et al., 2010). Internal browning can be triggered when the fruit is wounded (Queiroz et al., 2008). Browning is also related to abiotic stress in storage, such as low temperature, low oxygen, and/or high carbon dioxide (Mellidou et al., 2014). Browning occurs when phenolic compounds are oxidised by the enzyme polyphenol oxidase (PPO) which causes brown pigments to be generated (Queiroz et al., 2008). PPOs belong to a large gene family and of these, ten PPO genes have been mapped to the 'Golden Delicious' apple genome (Di Guardo et al., 2013). Polyphenol compounds are stored in vacuoles whilst PPOs are found in plastids (Holderbaum et al., 2010). Wounding, such as cutting or dropping the fruit, can weaken the cells, which allows both PPO and phenolic compounds to interact, causing polyphenols to revert to their corresponding quinones when oxidised by PPO. These then polymerise with other quinones or phenols to form brown pigments (Murata et al., 1995). Similarly, ripening also causes weakening effects in cells because enzymes, such as polygalacturonases, break down cell walls (Kramer and Redenbaugh, 1994). In apple cultivars, such as 'Aori27' and 'Mellow', chlorogenic acid is the major phenolic compound that is oxidised. In 'Fuji', chlorogenic acid and epicatechin are the major phenolic compounds, whilst in 'Elstar', epicatechin and procyanidin predominate (Holderbaum et al., 2010). The activity of PPO has also been shown to decrease as the apple ripens. This is due to a denaturing of the protein and not reduced PPO production (Murata et al., 1995). The ability to control enzymatic browning is important because it negatively impacts on colour, taste, flavour, and nutritional value in many fruits and vegetables (Holderbaum et al., 2010). Recently one apple has been engineered to reduce enzymatic browning. Named the 'Artic' apple, it possesses a non-browning trait conferred to it by silencing the PPO through RNA interference (RNAi), therefore reducing its expression (Waltz, 2015). This results in an apple which retains its colour, taste, flavour, and nutritional value even when damaged or cut.

Anthocyanins are a class of flavonoids commonly found in fruits and vegetables responsible for the vivid red, blue, and purple colours commonly found in nature (He and Giusti, 2010). In apple, the anthocyanins are primarily composed of cyanidin-3-galactoside with cyanidin-3-arabinoside, cyanidin-7 arabinoside, and cyanidin-3-xyloside usually in minor amounts (Vrhovsek et al., 2004; Meng et al., 2016). Overall, anthocyanins compose around 1–3% of total polyphenols in apples (Vrhovsek et al., 2004).

Anthocyanins are generated by enzymes of the phenylpropanoid pathway, including chalcone synthase (CHS) and chalcone isomerase (CHI), flavanone 3-hydroxylase (F3H), flavonoid 3'-hydroxylase (F3'H), flavonoid 3',5'-hydroxylase (F3'5'H) dihydroflavonol reductase (DFR), anthocyanidin synthase (ANS), and leucoanthocyanidin dioxygenase (LDOX) flavonoid-3-O-glycosyltransferase (UFGT/UF3GT) (Holton and Cornish, 1995; Winkel-Shirley, 2001; Tanaka et al., 2008). These enzymes are regulated at the transcriptional level by a well-studied complex termed the MBW complex (proteins of the transcription factor [TF] classes MYB, basic helix-loop-helix bHLH and a WD40 repeat protein). In apple, MYB10, which is allelic to MYB1/MYBA, regulates anthocyanin biosynthesis, having 58% overall protein identity to PRODUCTION OF ANTHOCYANIN PIGMENT 1 (PAP1/MYB75) in Arabidopsis. The apple bHLHs MdbHLH3 and MdbHLH33 are both members of the IIIf subfamily (Heim et al., 2003) and interact with MdMYB10. Both share high homology with TT8 in Arabidopsis and Delila in snapdragon (Espley et al., 2007).

In addition to the MBW complex, other TFs affect anthocyanin biosynthesis *via* protein-protein interaction with the MBW complex, or regulation of expression levels of genes encoding the MBW. A SQUAMOSA MADS-box gene regulates anthocyanin accumulation in bilberry (Jaakola et al., 2010). More recently, ethylene response factors (ERFs) have been shown to affect fruit colour at this level. In apple, MdERF1B has been shown to bind to the promoters of anthocyanin and proanthocyanidin regulating MYBs to alter anthocyanin and proanthocyanidin concentration (Zhang et al., 2018). In pear, PyERF3 interacts with PyMYB114 and its partner PybHLH3 to co-regulate anthocyanin biosynthesis (Yao et al., 2017) while in apple MYB1/10 was shown to activate expression of apple ERF3 (a close homologue to pear ERF3) and this increase in ERF expression increases ethylene emission from apple callus (An et al., 2018). Furthermore, it was shown that EIL1 directly bound to the promoter of MYB1 to induce anthocyanin production, while MYB1 was able to interact with ERF3, providing a positive feedback loop for ethylene biosynthesis. The apple bHLH3, a known partner to MYB1/10 in the MBW complex, has also been implicated in activating ACO and ACS1 and ACS5 genes to drive ethylene production (Hu et al., 2019). In a recent study of the transcriptome of red-fleshed apple (Wang et al., 2018), WRKY11 and ERF106 were identified as being differentially regulated and both were more expressed in red flesh.

Here, we use isogenic lines of 'Royal Gala' that differ genetically by only the over-expression of *MYB10* (Espley et al., 2007). These lines show that MYB10 can activate the expression of apple *ACS*  and *ACO* genes, possibly *via* up-regulation of *ERF* genes. This activation links anthocyanin concentrations and ethylene, as well as enhanced amounts of PPO and substrates for the browning reaction. This linkage has implications for the quality of highly pigmented apples.

#### MATERIALS AND METHODS

#### Plant Materials

To generate highly anthocyanic apple fruit, the *Malus* x *domestica* 'Royal Gala' apple cultivar was transformed with an overexpression construct containing *MYB10* cDNA under the control of the CaMV35S promoter as previously described (Espley et al., 2007). Fruit from multiple trees for three independent transgenic lines (A1, A3, and A4) over two seasons (2013–2014 and 2014–2015) were used in this study and compared with nontransformed wild type (WT) 'Royal Gala' fruit grown under the same conditions in a containment glasshouse. Samples from each season were collected at seven different time points: T1 = 35 Days After Full Bloom (DAFB); T2 = 65DAFB; T3 = 85 DAFB; T4 = 110 DAFB = T5, 120 DAFB; T6 = 130 DAFB (WT 'Royal Gala' commercial maturity); T7, 140 DAFB, using fruit from two or three transgenic lines (A1, A3, A4; dependent on fruit number) and WT control fruit. For gene expression analysis T2 to T6 were assayed, to capture the most relevant developmental stages.

#### Fruit Maturity and Ripening Assessments

Ethylene concentrations were determined both during fruit development and at harvest on three transgenic lines (A1, A3, and A4) and 'Royal Gala' WT fruit. For fruit *on planta*, a needle was inserted into the core cavity and the insertion site was sealed with wax to prevent wounding-related damage or gaseous loss; this remained in place during fruit development. Internal ethylene concentration was determined by extracting a 1 ml core cavity gas sample, using a compatible syringe to the inserted needle, and injecting it into a gas chromatograph (Hewlett Packard, 5890 series II) as previously described (Johnston et al., 2009). For harvested fruit, ethylene samples were extracted at one of the developmental time points using the same analysis method with five biological replicates. Fruit firmness was assessed by Texture Analyser TAXT plus (Stable Microsystems, United Kingdom) fitted with a 7.9-mm Effegi penetrometer probe (Johnston et al., 2009). Soluble solids content (SSC) was assessed by hand-held refractometer. Fruit were also maintained in industry-standard storage conditions for ten weeks after harvest at 0.5°C and assessed for extent of internal browning.

#### Peroxidase Enzyme Assay

Enzyme activity was assessed spectrophotometrically at 25°C. All assays were performed on a SpectroMax plus 384 UV-vis spectrophotometer and analysed using the SoftMax Pro v5.4.5 software. Peroxidase (POX) was extracted and assayed based on a modified version previously described (Lester et al., 2004). Apple cortical tissue was frozen and ground into a fine powder in liquid nitrogen; 250 mg of tissue was then homogenised in 1 ml of 100 mM phosphate buffer pH 7.0 containing 0.5 mM cysteine and 4% w/v polyvinylpolypyrrolidone (PVPP). POX activity was determined by adding 25 µL of the extract to 225µL of assay reagent (50 mM potassium phosphate, pH 7.0, 10 mM guaiacol, and 10 mM H2O2) in a 96-well plate. The reaction was initiated by adding 10 mM guaiacol, and the activity determined by the rate of formation of tetraguaiacol at 470 nM. Assays were performed in triplicate for each of the biological replicates. Results are presented as the maximum rate of change per mg of protein.

#### Non-Structural Carbohydrates and Anthocyanin Analysis

Fruit flesh tissue was ground under liquid nitrogen using an IKA (IKA, Staufen, Germany) grinder and stored at -80ºC. Soluble sugars were extracted from 200 mg aliquots of ground flesh in 80% ethanol. The soluble sugars (glucose, fructose, sucrose, and sorbitol) were analysed by ion chromatography as described in Nardozza et al. (2013). Starch was analysed by a colorimetric method following enzymatic digestion of the pellet obtained from the sugar extraction, as previously described by Smith et al. (1992). Anthocyanin and flavonoids were extracted in methanol (75% in water) and quantified by High Performance Liquid Chromatography (HPLC), using a method previously described (Espley et al., 2013). Starch accumulation rate/degradation was calculated using the Relative Growth Rate (RGR) where the primary data were logarithmically transformed to homogenise variability (Opara, 2000).

#### Gene Expression Analysis

RNA from apple flesh was isolated using the Spectrum™ Plant Total RNA kit (Sigma) and cDNA was synthesised using the Qantitect® Reverse Transcription kit (Qiagen), both according to the manufacturer's recommendations. Primers used for qPCR analysis was designed using Geneious 8.1.9 (Kearse et al., 2012) and synthesised by Macrogen (Republic of Korea). Primers are listed in **Table S1**. qPCR analysis was performed using the LightCycler® 480 using LightCycler FastStart SYBR Green Mix (Roche Diagnostics), according to the manufacturer's guideline. Reactions contained 2.5 μl Master Mix, 0.25 μl of each primer (10 mM), 1.25 ml diluted cDNA (1:25), and nuclease-free water (Roche Diagnostics) to a total volume of 5 μl, using reaction conditions previously reported (Espley et al., 2007). Expression was normalised against *Malus* x *domestica Elongation factor 1* (*MdEF1α*), a 'housekeeping' gene known for its consistent transcript level in apple fruits and leaves (Chagné et al., 2013).

### Isolation and Cloning of Apple Gene Promoter Sequences

For the cloning of apple promoters used in the Dual Luciferase assays, nested primers were designed to the target sequence. The inner primers for use in promoter isolation were designed to isolate approximately 2 kb upstream of the transcription start site of each gene. PCR fragments were cloned using Platinum® Taq DNA Polymerase High Fidelity (Thermofisher Scientific) as per manufacturer's protocols. For all candidate gene promoters, outer PCR amplification was first performed and checked on agarose gels. The PCR product was diluted 100 x and 1.5 μl of this was used as the template for the inner PCR amplification. Fragments were cleaned using DNA Clean and Concentrator™ (Zymo Research) as per manufacturer's protocol and cloned into the pGreenII 0800- Luc using the In-Fusion® HD Cloning Kit. Promoter fragments were verified by sequencing. Analysis of putative binding TF sites in the MdERF106 promoter was carried out using PlantPAN 2.0 (Chien et al., 2015). The MYB TF over-expression constructs used in the luciferase assays were as previously described (Lin-Wang et al., 2011; Brendolise et al., 2017).

#### Dual Luciferase Assays

Glasshouse-grown *Nicotiana benthamiana* plants were used for the dual luciferase assays to determine transcription factor activation of promoter sequences, as previously described (Hellens et al., 2005).

### Phylogenic and Statistical Analysis

Arabidopsis and other plant protein sequences were obtained from NCBI. Full-length sequences were aligned using MEGA 6.06 (Tamura et al., 2013) MUSCLE (open = -2.9, gap = 0) (Edgar, 2004). Phylogenetic analysis of the proteins were performed using MEGA 6.06 using a maximum likelihood method based on the JTT matrix-based model (Jones et al., 1992) and 1,000 bootstrap replicates. Analysis of the ERF genes was performed using MEGA 6.06 (6140226) *via* the Neighbour-Joining method (Saitou and Nei, 1987) with evolutionary distances calculated using the JTT matrix-based model and 1,000 bootstrap replicates.

The effects of the transgenic modification and fruit developmental stage (factors) on secondary metabolites, ethylene, peroxidase enzyme activity, and non-structural carbohydrates were analysed by ANOVA (type 3 sums of squares Kenward-Roger's method) using a linear mixed effects model in R (version 3.5.1). The effect of transcription factors (which is the factor in the analysis) on activating a selected promoter was analysed with the same method. The biological replicates were treated as random effects. The means were separated on the base of all pairwise comparisons of least-squares means (letters assigned).

### RESULTS

#### Elevated Anthocyanin Content Causes a Higher Incidence of Internal Browning Flesh Disorder (IBFD)

'Royal Gala' is a cultivar with little reported incidence of IBFD. There are reports of flesh browning but these tend to be after prolonged storage (Lee et al., 2016). We used three independent transgenic lines with high concentrations of flesh anthocyanin to test any link with flesh browning (**Figure 1A**). At harvest (T7), some of the transgenic fruit showed high incidence of IBDF (**Figure 1B**). In WT 'Royal Gala' and two of the transgenic lines (A1, A3) there was no IBFD at harvest. However, in the line with the highest anthocyanin content in the flesh, A4, IBDF was apparent in 16% of fruit at harvest (**Figure 1B**). After 10 weeks in standard storage conditions at 0.5°C, the incidence of IBFD was 67, 23, and 60% for the transgenic lines A1, A3, and A4, respectively. No IBFD was detected in any of the 'Royal Gala' fruit assessed. All the fruit tested at harvest, including WT, were relatively small compared with fieldgrown fruit and line A4, in particular, appears smaller than WT (**Table 1**). Both A1 and A3 showed lower fruit firmness than WT. Line A4 also showed a difference in SSC (*P <*0.001).

### Metabolic Comparison of MYB10 Fruit During Development

Previously it has been shown that the most dramatic effect of over-expression of MYB10 is a greater than a 20-fold increase in total anthocyanins in whole fruit (Espley et al., 2013). This is particularly evident in the fruit flesh, which goes from nondetectable amounts of anthocyanin in 'Royal Gala' to more than 700 μg/gFW in the transgenic lines. However, there are also significant changes in other polyphenols and these, together with the anthocyanin, may contribute to a substrate pool. We analysed this in the highly anthocyanic lines under-going IBFD (**Figure 1C**). This confirmed a trend to have higher concentrations of epi-catechin, chlorogenic acid (CGA), and quercetin (**Figures 1D**–**G**). In contrast, by maturity catechin was not significantly different between control and MYB10 lines (**Figure 1D**). As has been commonly reported, many of the major polyphenols in apple are highly abundant at the earlier stages in fruit development (Henry-Kirk et al., 2012) but the concentrations reduce as the fruit expands and matures. While there is some evidence for this in the anthocyanic transgenic lines, the reduction in these key metabolites is at a lower rate than in 'Royal Gala' and relatively high concentrations of epicatechin, CGA and quercetin are maintained in the ripe fruit. Since these polyphenols are a potential source

TABLE 1 | Fruit assessment for average weight, firmness and soluble solids content (SSC) of WT and three independent transgenic lines of 'Royal Gala' at harvest and after 10 weeks in storage. Incidence of browning by percentage of flesh showing symptoms by visual observation.


*Values are averages ± SEM.*

of enzymatic oxidation substrate, we measured the peroxidative potential of fruit throughout the development series.

Enzymatic browning in fruit can occur when phenolic compounds are oxidised, so we measured total peroxidase activity using the rate of conversion of guaiacol to tetraguaiacol. Total peroxidative activity in control 'Royal Gala' fruit flesh steadily declined during fruit development to very low rates by maturity (**Figure S1**). In contrast, peroxidative activity was seen to increase in the MYB10 lines.

#### Ethylene Induction in MYB10 Lines

Internal ethylene concentration (IEC) measurements were performed on attached fruit through fruit development using an *in planta* method. A sample of apple fruit core cavity gas was extracted while fruit remained on the tree, by insertion of a needle into the core cavity space. These measurements, made on three individual apples from the three independent lines of MYB10 'Royal Gala', and four individual control 'Royal Gala' fruit, were carried out over a 90-day period. Ethylene production for the transgenic lines increased at a much earlier point in fruit development (**Figure 2A**). Low rates of ethylene production occurred in these lines before T4, after which ethylene increased. By T5, IEC in two of the transgenic lines had reached greater than 2 ml/L while in WT it was less than 0.5 ml/L. As expected, ethylene was detected in the WT control fruit commensurate with the onset of on-tree ripening between T5 and T6.

IEC was also assessed destructively with off-tree fruit measurement at six developmental time points equivalent to 65, 85, 95, 100, and 108 DAFB on three separate biological replicates for each line and controls (**Figure 2B**). These standard assays confirmed results for the on-tree assessments with an early ethylene production detected in all transgenic lines compared with the control fruit.

#### Effects on Starch Concentration

Flesh starch concentration in transgenic MYB10 apple fruit was significantly lower than in WT 'Royal Gala' fruit (ANOVA main effect of wild type over transgenic lines; *P* < 0.0001) (**Figure S2**). Differences were evident from developmental stage T2 and were maintained throughout fruit growth. Starch degradation is an indicator of apple fruit maturity (Brookfield et al., 1997). Minimal starch concentrations at T6 and T7 indicated that fruit maturation was near completion in transgenic fruit at T6, whilst in 'Royal Gala' WT fruit starch degradation was still occurring by T7 stage. Furthermore, the rate of starch degradation was faster in transgenic MYB10 apple fruit compared to the WT (**Figure S3A**). Flesh sorbitol concentration in transgenic MYB10 apple fruit was significantly higher than in WT 'Royal Gala' fruit (ANOVA main effect of transgenic lines over wild type; *P* < 0.001) (**Figure S2B**). Fructose concentration was higher in WT 'Royal Gala' fruit than that in the transgenic fruit (ANOVA main effect of wild type over transgenic lines; *P* < 0.001) (**Figure S3B**). Fruit sucrose concentration was significantly higher in WT 'Royal Gala' than the transgenic lines at T7 (*P* < 0.05; **Figure S3C**). Glucose concentration was overall higher in transgenic lines, although the differences were not statistically significant (**Figure S3D**).

#### Expression Analysis of PPO and Ethylene-Related Genes

We tested a number of PPO genes (PPOa, PPOb, PPOc, and PPOd) for transcript abundance in both red and white apple flesh tissue. There were high rates of expression in both transgenic lines (A1 and A4) compared with WT control for all tested genes (no transcript was detected for PPOc) (**Figure 3**). This expression was evident at early time points in fruit development, T2 for PPOb and PPOd and T3 for PPOa in the transgenic lines. In WT 'Royal Gala' there was no detectable expression in PPOa and PPOb and a gradual increase in expression of PPOd, albeit at a lower rate than the red-fleshed lines. In most cases line A4, the most highly pigmented and prone to IBFD, showed the highest expression for all the genes tested.

With evidence for an early peak in ethylene production in the anthocyanic fruit, the key genes in the ethylene biosynthesis pathway, ACO and ACS, were tested for transcript abundance in the fruit flesh of two transgenic lines (A1 and A4) and WT (**Figure 3**). For ACO1, transcript was detectable in both A1 and A4 at T4, while in 'Royal Gala' WT transcript was not detected until T5. The highest ACO transcript abundance was detected in the A1 and A4 transgenic lines at the final time point, T6. For ACS1, transcript was at low levels with an anomalous peak of expression for RG at T5. For the previous year tested, an early peak of ACS1 expression was evident for both transgenic lines tested at T4 but at the final time point the highest expression was seen in control tissue (**Figure S4**). For ACS3, the most abundant ASC gene, transcript was detected at an even earlier stage (T3) for A1 and A4 but was evident in all samples, including WT control, at T4. By the final time point (T6), expression was higher in the transgenic lines. ACS5 was never detected in control samples but there was low expression in A1 and greater expression, albeit at a low level, in A4.

Based on data from a previous study using RNA-sequencing on red and white fleshed apples (Wang et al., 2018), we selected ERF106 as a possible candidate gene for involvement in the regulation of anthocyanin-related early ethylene production. We found low levels of transcript abundance throughout fruit development in 'Royal Gala' but consistently higher abundance in both of the transgenic lines tested. MdERF106 is similar to Arabidopsis AtDEWAX and AtDEWAX2 (AtERF107/ERF106) (**Figure 4**, **Figure S5**) which negatively regulate cuticular wax biosynthesis (Go et al., 2014; Kim et al., 2018), and other ERFs such as AtERF5 and AtERF6 involved in biotic and abiotic stress response (Moffat et al., 2012;

(B) PPOb, (C) PPOd, (D) ACS1, (E) ACS3, (F) ACS5, (G) ACO1, and (H) ERF106. T1, 35 DAFB; T2, 65DAFB; T3, 85 DAFB; T4, 110 DAFB; T5, 120 DAFB; T6, 130 DAFB. Expression was normalised to EF1a and error bars are SE of four technical replicates of at least 5 pooled fruit per time point. Expression is shown for three lines; dark grey—A1, medium grey—A4, and light grey—RG.

Dubois et al., 2013). Previously it has been shown that MYB1/10 activates expression of apple ERF3 (a close homologue to pear ERF3) to increase ethylene emission from apple callus (An et al., 2018), while MdERF1B has been shown to alter anthocyanin and proanthocyanidin concentration (Zhang et al., 2018). In pear PyERF3 interacts with PyMYB114 to co-regulate anthocyanin biosynthesis (Yao et al., 2017). These ERFs are less similar to MdERF106.

#### Activation of Non-Anthocyanin Related Genes by MYB10

To test for the possibility of direct activation of ethylene-related gene by MYB10, we cloned the promoters of ACO, ACS1, ACS3, ACS5, and ERF106, which were fused with the luciferase reporter gene sequence. These were then tested for MYB10 activation using the dual luciferase assay. Since MYB10 is known to partner with bHLH3 in the MBW complex and is required for full activation of the anthocyanin structural genes, the apple bHLH3 TF was also included. A non-anthocyanin related MYB, MYB8, was used as a control MYB to test for specificity of MYB10. The promoter fragment of apple DFR was used as a positive control for MYB10 activation. As expected, MYB8 had little activation potential on DFR, while MYB10 did induce promoter activity and luciferase production (**Figure 5A**). This was further enhanced by the co-infiltration of bHLH3.

For the ACS3 and ACS5 promoters there was no apparent activation by MYB10, with or without bHLH3, or for the control MYB8. However, both MYB8 and MYB10 appeared to activate ACS1, suggesting a non-specific MYB activation. This was somewhat reduced for MYB10 when co-infiltrated with bHLH3. For ACO1, there was also activation by MYB8 and MYB10, with higher activation for the latter that was then somewhat diminished by the inclusion of bHLH3.

We tested the MYBs for activity on the ERF106 promoter fragment. Here we found no detectable activation by the control MYB8 but a strong activation by MYB10. Again, this was reduced with the co-infiltration of bHLH3.

To further test the specificity of MYB10 on the ERF106 promoter, a number of known apple flavonoid activating and repressing MYB TFs were assayed against the ERF106 promoter. The strongest activation was seen with the anthocyanin-related activators MYB10 and MYB110 (**Figure 5B**). A lower level of activation was seen with infiltration of the flavonoid-related MYB12, while the anthocyanin repressors showed mixed results, with no activation by MYBs 15 and 16 but some activation by both MYB27 and MYB111.

## DISCUSSION

Of the widely cultivated varieties of apple, 'Royal Gala' does not normally exhibit IBFD. However, the MYB10-induced ectopic accumulation of anthocyanin in fruit flesh leads to a severe browning phenotype, suggestive of a link between flesh colour and browning. Alternatively, an increase in other polyphenols may be the cause, providing increased substrate for downstream oxidation. The concentration of polyphenols in apple fruit usually reduces over development but in MYB10 lines a relatively high concentration of epi-catechin, CGA, and quercetin is still present at maturity, in addition to the high concentration of cyanidingalactoside. The incidence of IBFD in these highly anthocyanic transgenic lines is resonant with data from a traditionally bred red-fleshed apple population (Volz et al., 2013). In the redfleshed breeding populations tested, quantitative trait loci (QTL) for IBFD were detected at two genetic locations on Linkage Group (LG) 9 and LG6, consistent with a QTL for red flesh, and suggesting a strong genetic link between MYB10 and IBFD.

Adding to these possibilities, there is also an increase in the ripening-related gaseous hormone, ethylene. Since the developmental accumulation of anthocyanin colour in apple is part of the ripening process, it is likely that these processes are connected. Here we show that MYB10 may play a role in ethylene biosynthesis, *via* ethylene-related transcription factors and that this premature ethylene production is closely associated with a reduction in fruit quality, ultimately leading to IBDF. A previous study in a F1 population of red and white fleshed apples showed a higher concentration in flavonoid content as well as anthocyanin in red fleshed apples (Wang et al., 2015). In this study a comparative transcriptome showed a higher incidence of abiotic and biotic stress-related genes in red flesh, and that this was associated with an increased production of flavonoids. Whether these stress-related genes are up-regulated in direct response to flavonoid production or the potential detrimental effects of an increase in flavonoids leading to premature ripening is yet to be fully determined. This study was conducted on ripe fruit although some ethylene-related genes such as ACO (gene ID 103404960, equivalent to MDP0000195885) showed greater transcript abundance in red flesh compared to white.

#### Anthocyanin Is Associated With Major Changes in Fruit Physiology

The ectopic accumulation of anthocyanin does not necessarily lead to reduced fruit quality. This is convincingly demonstrated in tomato where elevation of anthocyanin in transgenic purple tomato fruit leads to an extended shelf life and a reduction in susceptibility to grey mould (Zhang et al., 2013). In this study, it appears that the increased anthocyanin boosts antioxidant capacity in the fruit which is likely to slow the over-ripening process. This is in direct contrast to results presented here for apple. We have previously shown an elevated antioxidant capacity for these highly anthocyanic apples (Espley et al., 2014) despite a reduction in ascorbic acid, suggesting that the increase in anthocyanins and flavonoids are the reason for this increase. However, this increased capacity did not improve shelf life and the red-flesh apples over-ripened more quickly than the whitefleshed controls. Like apple, tomato is a climacteric fruit and so reliant on ethylene for ripening. In the purple tomatoes, ethylene production was 2-fold greater than control fruit, although this was just after the breaker stage. This differs for apple, where we see an ethylene burst earlier in fruit development.

There are reports of flesh browning (flesh breakdown) in 'Royal Gala' but these tend to be after prolonged storage of more than three months and may be more prevalent in larger fruit (Lee et al., 2013; Lee et al., 2016). In these transgenic lines, a high incidence of IBFD is visible at (and even before) harvest (Line

numbers for other species as follows: LePti4 (NM\_001347076.1), LePti5 (U89256.1), ORA59 (NM\_100497.3), AaERF1 (JN162091.1), PyERF3 (MF489220), NtERF4 (NM\_001325253.1), and EjAP2-1 (KM506584.1).

A4) and is severe by 10 weeks in storage, while WT 'Royal Gala' showed no visible symptoms.

Apple has an unusual carbohydrate metabolism compared to the more commonly studied model plants (e.g. Arabidopsis) as sucrose is not the major carbohydrate metabolite. Indeed sorbitol is the major photosynthetic product in apple leaves (60–80% of sugars), and it is translocated in the phloem to sink tissues where it is quickly converted into fructose by sorbitol dehydrogenase (Loescher, 1987; Klages et al., 2001; Wu et al., 2015), so that sorbitol concentrations in fruit are usually low (Yamaki and Ishikawa, 1986). Apple fruit are mainly fructose accumulators, although early in development they store both fructose and transitory starch in similar amounts (Li et al., 2018). Fruit sorbitol dehydrogenase genes respond to variable levels of source sorbitol, decreasing in expression when leaf sorbitol levels decrease. In sorbitol dehydrogenase antisense lines, translocated sorbitol is reduced, lowering the levels of sorbitol in fruit (Teo et al., 2006; Li et al., 2018). In our experiment, transgenic MYB10 lines had fruit with higher a concentration of sorbitol in the fruit flesh than the wild type 'Royal Gala'. This phenotype could be due to the up-regulation of leaf sorbitol metabolism in transgenic MYB10 lines or by a down-regulation of the sorbitol dehydrogenase genes in the fruit so that sorbitol is converted more slowly into fructose. However, it could be also linked to the maturation stage the fruit have reached. Onset of net transitory starch degradation in apple occurs whilst fruit are still on the tree (Zhang et al., 2017), and it is an indication of fruit maturation (Brookfield et al., 1997). In apple, starch degradation is a process with low dependency but high sensitivity to ethylene (Johnston et al., 2009) which means an earlier onset of endogenous production of the hormone (from T3 rather than T5 in WT) could trigger faster starch degradation in the fruit, as observed in this study with the transgenic MYB10 lines. Sorbitol is likely to be linked to the ripening stage (Aprea et al., 2017). The accumulation of sorbitol in the MYB10 lines could be a result of the fruit switching to a ripening stage at T4–T5, and sorbitol translocated to the fruit is converted to fructose more slowly.

In the highly anthocyanic MYB10 fruit, there is an increase in a number of polyphenols as previously shown (Espley et al., 2013). These polyphenols could provide added substrate to fruit flesh for

enzymatic browning to occur. The peroxidative activity is also higher in all the MYB10 lines, therefore, both the potential substrates for browning and the peroxidative enzymes are at greater concentration in red-fleshed lines than in WT control. This is further shown in the PPO expression analysis, where an increase in transcript in both transgenic lines tested was evident for PPOa, PPOb, and PPOd. Results from the published transcriptomic analysis of red versus white apple flesh at harvest (Wang et al., 2015), shows that PPOa, and particularly PPOb, are more highly expressed in the red flesh while PPOc and PPOd are more highly expressed in the white fleshed fruit. It would appear that the key finding presented in our data is the demonstration of very early PPO expression.

#### Early Production of Ethylene *via* the Up-Regulation of Ethylene Genes in Red-Fleshed Fruit

The rate of ethylene production can vary widely between apple cultivars (Harada et al., 2000). It has also been previously shown that mature red-fleshed apples start to produce more ethylene than those with white flesh just two days after harvest (An et al., 2018). In our study, apples from the same genetic background were used, so from a genetic level it might be expected that ethylene production would be similar. However, we detected a very different ethylene production profile for all three transgenic lines compared with WT. Production was at levels normally consistent with the onset of the ripeningrelated ethylene burst but at four to six weeks earlier than expected. The collection of ethylene samples while fruit were still attached to the tree was chosen as the most representative method of assessing the actual increase in ethylene production in developing fruit. As this is method is novel, we also used a standard practice for ethylene assessment for harvested fruit and found very similar patterns.

Studies have shown that the biosynthesis of ethylene is transcriptionally regulated (Schaffer et al., 2013). ACC synthase (ACS) converts *S*-adenosyl-L-methionine to 1-aminocyclopropane-1-carboxylate (ACC) and is thought to be the rate limiting step

(Yang and Hoffman, 1984). We tested the transcript abundance of three versions of ACS: ACS1, ACS3, and ACS5. For all three versions we found an increase in transcription at a considerably advanced stage in the anthocyanic fruit compared to WT 'Royal Gala'. ACS1 is associated with the ripening-specific climacteric burst (Harada et al., 2000), while ACS3 is transcriptionally active before ACS1 and is likely to be involved in the transition from system-1 to system-2 ethylene biosynthesis (Wang et al., 2009). ACS3 transcript was just detectable in the transgenic lines at T3, and clearly detectable at T4, as was ACS1. However, there did not appear to be negative feedback for ACS3, as previously reported (Wang et al., 2009) and transcription increased as fruit developed. ACS5 is possibly more associated with wounding rather than ethylene-related ripening (Costa et al., 2005). The transcript abundance of ACS5 was detectable at low levels in the MYB10 lines, particularly A4, but was at very low rates.

The oxidation step performed by ACC oxidase (ACO) to form ethylene follows the ACS-mediated conversion to ACC. ACO transcript abundance was measured and the first major peak in expression was seen at T4 but only in the transgenic lines A1 and A4. At T5 transcript was also detected in control fruit. There are other ACO gene family members (not tested here) that may also contribute to the rise in ethylene and ACO has been linked to internal browning in white-fleshed fruit (Mellidou et al., 2014). The pattern for the highly anthocyanic fruit shows that the presence of ACS3 transcript at T3 precedes both ACS1 and ACO transcript at T4 while the later expression of ACS3 in control fruit at T4 ACO precedes ACO1 at T5 and ACS1 at T6. Since ACS1 has been shown to be responsible for system-2 type autocatalytic ethylene production (Tan et al., 2013) the results here are entirely consistent with premature ripening in MYB10 fruit.

#### A Proposed Model for the Association Between Anthocyanin and IBFD

The Ethylene Signalling Cascade Ends With the ERFs and These Can Either Activate or Repress Ethylene production as well as regulating the expression of ripening-related genes (Liu et al., 2015). In apple, the ERF2 TF has been shown to negatively regulate ethylene biosynthesis by suppressing the transcription of ACS1 (Li et al., 2016). Previous data has shown that one ethylenerelated TF, ERF106, was differentially expressed in red compared with white apple flesh (Wang et al., 2018).

A study by Zhang et al. (2018) showed that ERF1B was capable of interacting with the promotors of anthocyanin and proanthocyanidin MYB TFs. This interaction leads to increases in the related metabolites and suggests that ethylene regulation and anthocyanin regulation might be linked in either direction.

It appears that both MYB apple flesh anthocyanin regulators, the Type 1 MYB10 and Type 2 MYB110 (Chagné et al., 2013) were capable of activating the promoter of ERF106. Some of the other MYBs tested did demonstrate some level of activation, but to a lesser extent. The MYB binding sites may confer some nonspecific binding and hence, activation. Interestingly, two known apple anthocyanin repressors, MYBs 15 and 16 (Lin-Wang et al., 2011), showed no activation with reporter levels below that of the control. Further work is required to test if these MYB repressors can compete with activators or repress promoter activity.

We propose a model (**Figure 6**) whereby the up-regulation of MYB10 directly increases anthocyanin concentration by regulating the anthocyanin biosynthetic pathway and indirectly causes the increased production of flavonoids by enhanced pathway flux. MYB10 is also associated with an increase in early ethylene production, possibly *via* interaction with ERF106. This ethylene production advances mechanisms that drive fruit maturity, including the elevation of PPO. These direct and indirect events produce a scenario where early ripening is triggered and where the production of peroxidise enzymes and large pool of additional substrate come together to culminate in IBFD.

#### Summary

The data presented here describes a mechanism for elevated IBFD in anthocyanin tissue in apple fruit. This has implications for the breeding of high quality novel red-fleshed cultivars. The genetic background may be key to IBDF evasion, such as the use of low ethylene producing parents in a breeding population. Alternative strategies to avoid IBDF could include genetic manipulation such as demonstrated by knocking out PPO in Arctic® apple (Waltz, 2015) which could also be achieved using CRISPR/Cas9 gene editing approach.

### DATA AVAILABILITY STATEMENT

All datasets for this study are included in the manuscript/ **Supplementary Files**.

### AUTHOR CONTRIBUTIONS

RE, DL, BP, RH-K, MH, and SO'D contributed to the harvest and postharvest assessments, qPCR, transactivation experiments. MP and BP conducted peroxidase assays. HB and SN performed carbohydrate assessment. TM performed chemical analysis. RE, AA, RV, and JJ developed the experimental design. RE, AA, SN, and BP wrote the manuscript and all authors contributed to editing.

## ACKNOWLEDGMENTS

This work was funded by the New Zealand Government, Ministry of Business, Innovation and Employment. We thank Robert Schaffer and Adam Friend for helpful advice on the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01248/ full#supplementary-material

### REFERENCES


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Espley, Leif, Plunkett, McGhie, Henry-Kirk, Hall, Johnston, Punter, Boldingh, Nardozza, Volz, O'Donnell and Allan. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# The Ethylene Precursor ACC Affects Early Vegetative Development Independently of Ethylene Signaling

*Lisa Vanderstraeten†, Thomas Depaepe†, Sophie Bertrand and Dominique Van Der Straeten\**

Laboratory of Functional Plant Biology, Department of Biology, Ghent University, Ghent, Belgium

The plant hormone ethylene plays a pivotal role in virtually every aspect of plant development, including vegetative growth, fruit ripening, senescence, and abscission. Moreover, it acts as a primary defense signal during plant stress. Being a volatile, its immediate biosynthetic precursor, 1-aminocyclopropane-1-carboxylic acid, ACC, is generally employed as a tool to provoke ethylene responses. However, several reports propose a role for ACC in parallel or independently of ethylene signaling. In this study, pharmacological experiments with ethylene biosynthesis and signaling inhibitors, 2-aminoisobutyric acid and 1-methylcyclopropene, as well as mutant analyses demonstrate ACC-specific but ethylene-independent growth responses in both dark- and light-grown Arabidopsis seedlings. Detection of ethylene emanation in ethylene-deficient seedlings by means of laser-based photoacoustic spectroscopy further supports a signaling role for ACC. In view of these results, future studies employing ACC as a proxy for ethylene should consider ethylene-independent effects as well. The use of multiple knockout lines of ethylene biosynthesis genes will aid in the elucidation of the physiological roles of ACC as a signaling molecule in addition to its function as an ethylene precursor.

Keywords: 1-aminocyclopropane-1-carboxylic acid, ethylene signaling, root growth, rosette growth, triple response, vegetative growth

## INTRODUCTION

The gaseous plant hormone ethylene has been shown to regulate a myriad of physiological and developmental processes including germination, root growth and root hair formation, leaf expansion, leaf and flower senescence, abscission, fruit ripening, nodulation, and the response to numerous stresses (Burg and Burg, 1962; Abeles et al., 1992; Vandenbussche et al., 2012). Ethylene is formed from the amino acid methionine in three subsequent steps with S-adenosylmethionine (SAM) and 1-aminocyclopropane-1-carboxylic acid (ACC) as intermediates. The rate-limiting step in the biosynthesis of ethylene is the conversion of SAM to ACC, catalyzed by the enzyme ACC synthase (ACS) (Boller et al., 1979; Yang and Hoffman, 1984). In *Arabidopsis*, ACS proteins are encoded by a multigene family, eight of which are functional ACC synthases. The transcription of *ACS* genes is highly regulated during plant development and in response to a wide variety of developmental, hormonal, and environmental stimuli (Liang et al., 1992; Van Der Straeten et al., 1992; Tsuchisaka and Theologis, 2004). The final step of ethylene biosynthesis, the oxidation of ACC to ethylene, is catalyzed by the enzyme ACC oxidase (ACO) (Ververidis and John, 1991). In *Arabidopsis*, the *ACO* gene family consists of five members that are also differentially regulated (Barry et al., 1996; Nakatsuka et al., 1998). Although ACS is the major rate-limiting enzyme in

#### Edited by:

Gerrit T. S. Beemster, University of Antwerp, Belgium

#### Reviewed by:

Jie Le, Institute of Botany (CAS), China Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China

#### \*Correspondence:

Dominique Van Der Straeten Dominique.VanDerStraeten@UGent.be

†These authors have contributed equally to this work and are named in reverse alphabetical order

#### Specialty section:

This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science

Received: 17 July 2019 Accepted: 13 November 2019 Published: 06 December 2019

#### Citation:

Vanderstraeten L, Depaepe T, Bertrand S and Van Der Straeten D (2019) The Ethylene Precursor ACC Affects Early Vegetative Development Independently of Ethylene Signaling. Front. Plant Sci. 10:1591. doi: 10.3389/fpls.2019.01591

1 **189** ethylene biosynthesis, under certain conditions, for example, during fruit ripening, ACO can also become rate-limiting (Barry et al., 1996; Van De Poel et al., 2012). Moreover, the levels of ACC are not only regulated at the level of ACS and ACO activity, but are also dependent on conjugation and deamination of ACC (Amrhein et al., 1981; Martin et al., 1995; Glick et al., 1998; Mcdonnell et al., 2009).

As the immediate and water-soluble precursor of ethylene, the main role of ACC is to act as a mobile signal for shortand long-distance communication within the plant. Transport of ACC throughout the plant has been observed in numerous cases (Bradford and Yang, 1980; Lurssen, 1981; Zarembinski and Theologis, 1993; Morris and Larcombe, 1995; Jackson, 2002; Almeida et al., 2003; Jackson, 2008; Vanderstraeten and Van Der Straeten, 2017). Recently, the amino acid transporter LYSINE HISTIDINE TRANSPORTER1 (LHT1) has been demonstrated to transport ACC in etiolated *Arabidopsis* seedlings (Shin et al., 2015). While it is clear that a major role of ACC is to act as the precursor of ethylene, several studies suggest that ACC itself can act as a signal independent of its oxidation to ethylene. Exogenous ACC is widely applied as a tool to study ethylene responses in plants. Both the triple response phenotype in etiolated seedlings and the reduced rosette size in light-grown plantlets, typical ethylene-related phenotypes, are triggered by ACC as well (Guzman and Ecker, 1990; Van Der Straeten et al., 1993; Roman et al., 1995; Smalle et al., 1997). The comparison of null mutations in key ethylene signaling components and the octuple *ACS* (*acs8x*) ethylene biosynthesis mutant revealed that not ethylene but ACC is crucial for viability, since only the latter resulted in embryo lethality (Tsuchisaka et al., 2009). Moreover, Xu et al. (2008) suggest that ACC might act as a signaling molecule to regulate cell expansion in the FEI/SOS5 pathway. Investigating the *fei1fei2* mutant they found that the cell expansion phenotypes in roots could be reversed by blocking ethylene biosynthesis [using AOA (2-aminooxyacetic acid, an ACS inhibitor) or AIB (2-aminoisobutyric acid, an ACO inhibitor)] but could not be reversed by chemical [using 1-MCP (1-methylcyclopropene) or silver thiosulfate] or genetic (using *etr1* or *ein2* ethylene insensitive mutants) disruption of ethylene perception. A couple of years later, Tsang et al. (2011) observed that the short-term response to cell wall damage or PAMPs resulting in rapid reduction of primary root elongation depends on the biosynthesis of ACC but is independent of the perception of ethylene. They were able to show that AIB is capable of fully restoring the LEH (length of the first epidermal cell with a visible root hair bulge) in isoxaben-treated (inhibitor of cellulose biosynthesis) roots but did not affect the ACC response. Recently, a signaling role for ACC in stomatal development has been demonstrated (Yin et al., 2019). The symmetric division of the guard mother cell (GMC) into two guard cells represents the last step in stomatal development, a process depending on ACC. Pharmacological manipulation of ACC levels showed that ACC acts as a positive regulator in GMC division. Reduced levels of ACC, in the multiple *acs* knockout lines increased the occurrence of single guard cells (SGC). This phenotype could be rescued by addition of ACC but not by treating SGCs with the ethylene-releasing chemical ethephon. Altogether, these reports demand for a reassessment of the physiological role of ACC as a signaling molecule. In this study, the ethylene-independent signaling role of ACC has been investigated during early vegetative growth. Specifically, ACC negatively affected both rosette development and hypocotyl growth, and inhibited primary root elongation independently of ethylene perception. However, similar to ethylene dosedependent growth inhibitory effects, roots were more sensitive to ACC compared to shoots.

## MATERIALS AND METHODS

### Plant Material and Growth Conditions

*Arabidopsis thaliana* (L.) Heynh. Columbia (Col-0) was used as wild-type (WT) in this study. Col-0, *ein2-1* (Roman et al., 1995) and *acs1-1acs2-1acs4-1acs5-2acs6-1acs7-1acs9-1acs11* (*acs8x*; Tsuchisaka et al., 2009), both in Col-0 background, were obtained from the Nottingham Arabidopsis Stock Center (NASC; arabidopsis.info/). Seeds were surface-sterilized for 12 min in a bleach solution containing 5% NaOCl and subsequently washed at least 3 times with sterile distilled water. Seeds were plated on halfstrength Murashige and Skoog (1/2 × MS) medium containing 0% (**Figure S1**) or 1% sucrose (all other assays) and 0.8% agar and supplemented with the indicated chemicals. 1-aminocyclopropane-1-carboxylic acid (ACC) and 2-aminoisobutyric acid (AIB) were purchased from Sigma-Aldrich and stock solutions were prepared in distilled water. After a stratification of 3 days at 4°C, plates were transferred to a tissue culture chamber (21°C; 16/8-h photoperiod; 70 µmol s−1 m−2) for the desired time. For assays with dark-grown seedlings, seeds were exposed to light for 6 h after stratification to induce germination and were subsequently grown for 4 days in complete darkness (21°C).

#### Gas Treatments 1-MCP

At the start of the experiment, plates were transferred to a dedicated gassing chamber (Van Cleven, Belgium) containing both a treatment and a control cell. Plants were treated with 1-methylcyclopropene (1-MCP) in the treatment cell for 20 h, followed by 4 h of flushing with 1-MCP-free air. The required amount of 1-MCP (EthylBloc) was dissolved in a buffer containing 0.9% KOH and 0.9% NaOH in a 200-ml beaker, which was immediately transferred to the chamber, to give a final concentration of 50, 100, or 250 ppm inside the treatment cell. In parallel, control plants were placed in the control cell and flushed continuously with 1-MCP-free air. The treatment was repeated on a daily basis until the end of the experiment.

#### Ethylene

Ethylene treatment on 2-week-old plants (**Figure S3**) was carried out in the gassing chambers described above. For combined treatments, plants were first treated with ethylene (Air Liquide, Belgium) for 4 h at a final concentration of 100 ppm, and subsequently treated with 1-MCP for 20 h. Both treatments were repeated on a daily basis until the end of the experiment.

## Phenotypic Analysis

Plants were photographed with a Canon EOS 550D camera (Canon, Tokyo, Japan)) after 14 days of growth on horizontally standing plates. To evaluate effects on shoot growth, rosette area was measured with Rosette Tracker (De Vylder et al., 2012), an open-source plug-in in ImageJ (National Institutes of Health). In addition, root length was measured in ImageJ. The triple response assay was carried out, as described previously (Hu et al., 2017), to evaluate the growth of etiolated seedlings. The length of 4-day-old hypocotyls and roots were measured in ImageJ. Average values were obtained from three independent replicates.

### Measurement of Ethylene Emanation

Ethylene emanation was monitored essentially as described in Van de Poel and Van Der Straeten (2017). For the detection of ethylene levels emitted by etiolated seedlings, approximately 30 seeds were sown in 10-ml chromatography vials (Chromacol, VWR) containing 5 ml ½ × MS medium (four independent biological repeats), transferred to a sterile box and grown in darkness at 21°C. Average values were obtained from four independent replicates. For the detection of ethylene levels in 14-day-old plants, three seeds were sown in 10-ml vials, containing 8 ml ½ × MS medium, and grown in a tissue culture chamber (21°C; 16/8-h photoperiod; 70 µmol s−1 m−2). Average values were obtained from eight independent replicates 24 h before the start of the measurement, vials were sealed off with a rubber septum and a snap-cap (Chromacol, VWR) to allow ethylene accumulation. Ethylene levels were analyzed by means of laser-based photoacoustic spectroscopy (ETD-300, Sensor Sense, The Netherlands).

### Ethylene Complementation Assay

To monitor residual ethylene biosynthesis in the presence of ACC and the ACO blocker AIB, ethylene emission was examined, as described above, in etiolated Col-0 seedlings on a daily basis. The effect of this concentration of ethylene was assessed as follows. Col-0 seeds were sown in 10-ml chromatography vials on media containing 0, 10, or 50 µM ACC in the absence or presence of 2 mM AIB. After germination, ethylene was injected with a gastight syringe (Hamilton) to a final concentration equivalent to the levels measured, over a 24-h period from day 3 to day 4, upon treating with 10 [= ETH(10)] or 50 µM ACC [= ETH(50)] in the presence of 2 mM AIB. The final concentrations were 116 and 585 ppb, respectively. Seedlings were allowed to grow for 3 days in complete darkness, after which the phenotypic effects were evaluated at the level of hypocotyl and root growth. Seedlings were grown in the same vials on media containing ACC and/ or AIB with ethylene-free air as a control. Average values were obtained from four independent replicates.

### Statistical Analysis

All statistical analyses were carried out in the free software environment for statistical computing and graphics R 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria, www.Rproject.org). Data are presented as means, error bars are standard deviations. Statistical analysis comparing two means was performed using the Wilcoxon rank sum test/T-test (P < 0.01). Statistical analysis comparing multiple means was performed using One-Way ANOVA/Kruskal-Wallis (one independent variable) or Scheirer-Ray-Hare (two independent variables) tests (*P* < 0.01) followed by *post hoc* Tukey's HSD/Dunn tests (*P* < 0.01) with Benjamini and Hochberg correction for multiple pairwise comparisons. In addition, effect sizes for multifactorial analyses are presented and interpreted with partial η² according to Richardson (2011). Small, medium, and large effects correspond with effect sizes of 0.01, 0.06, and 0.15, respectively. Partial η² is calculated as the sum of squares (SS) divided by the sum of SS and SS of the residuals. Effect size for comparisons between two groups is presented as r (Rosenthal, 1994). R was calculated as *Z N* / (Wilcoxon rank sum tests, Tukey's HSD and Dunn tests) or as √(t^2/(t^2+df)) (T-tests). Small, medium, and large effects correspond with effect sizes of 0.1, 0.3, and >0.5, respectively. Relevant output of effect sizes can be found in **Table S1**.

## RESULTS

#### Dose-Dependent Effects of ACC on Shoot and Root Development Upon Ethylene Insensitivity

To explore the role of ACC in rosette development, possibly independent of ethylene, dose-response assays were conducted in 2-week-old WT Col-0 and in ethylene insensitive *ein2-1* plants (**Figures 1A**, **B**). In parallel, 1-methylcyclopropene (1-MCP) gas treatments, specifically blocking ethylene from binding to its receptor, were carried out to further investigate ACC-specific effects on rosette growth (**Figures 1C**, **D**). In WT Col-0, ACC reduced overall shoot growth in a dose-dependent manner, reflected by a decrease in rosette area (**Figures 1A**, **B**). Compared to the mock treatment, 10 µM ACC already reduced rosette area severely. A saturated response was visible as of 100 µM ACC. As expected, the phenotype of the rosettes was reminiscent of that of plants treated with exogenous ethylene or *ctr1* mutants, namely severe dwarfism caused by smaller leaf blades and petioles, as a result of inhibited cell expansion (Kieber et al., 1993; Rodrigues-Pousada et al., 1993). With a defective ethylene signaling pathway in *ein2-1*, rosettes responded differently to ACC compared to Col-0 (**Figures 1A**, **B**). At 10 µM ACC, *ein2-1* rosette size was slightly larger compared to a treatment with 0 µM ACC. Contrarily, at 100 µM ACC, the mean rosette area was decreased, as seen in Col-0 plants. The observation that ACC inhibited rosette development at higher doses regardless of the genotype, is indicative for an ethylene-independent effect of ACC. In addition, a small-scale experiment was carried out to determine whether the inhibitory effect of ACC on rosette growth is influenced by the presence of sucrose (**Figure S1**). In general, the omission of sucrose supplementation in the growth medium resulted in a decrease in rosette area in both Col-0 and *ein2-1* plants and irrespective of ACC concentration (**Figure S1B**). In the absence of ACC, a lack of sucrose resulted in a small inhibitory effect on rosette area in Col-0 and large decrease in *ein2-1*. However, at high concentrations of ACC (e.g., 100 µM) rosette area of Col-0 and *ein2-1* was reduced severely, irrespective of the presence of sucrose. However, the magnitude of the effect was

larger upon sucrose supplementation. For instance, in *ein2-1*, 100 µM ACC decreased rosette area with 5.88 mm² and 12.30 mm² in the absence or presence of sucrose, respectively. Hence, although sucrose affects rosette growth in both WT and ethylene-insensitive plants, ACC is capable of inhibiting growth independently of ethylene and sucrose signaling.

When ethylene perception was blocked with 250 ppm 1-MCP, Col-0 rosettes were slightly larger compared to mocktreated rosettes, though this increase was negligible (**Figure 1C**). Furthermore, in the presence of 1-MCP, ACC was still able to reduce growth in a dose-dependent manner, indicating that ACC can affect shoot growth independently of ethylene perception. On 10 µM ACC, MCP-treated rosettes reached 30.01 mm² compared to 0 µM ACC, while 100 µM ACC further decreased rosette area to 8.72 mm² (**Figure 1C**). Nevertheless, in the presence of 1-MCP Col-0 exhibited a reduced sensitivity to ACC as compared to the absence of 1-MCP, consistent with the ACC dose-response in *ein2-1* (**Figures 1B**, **C**). Furthermore, 1-MCP did not substantially change the response of *ein2-1* to increasing concentrations of ACC (**Figures 1A**, **D**). An additional experiment using 50 ppm 1-MCP was conducted to verify that the observed inhibitory effects were not due to an excess of 1-MCP (**Figure S2**). The effects of ACC on rosette area were comparable to those in the presence of 250 ppm 1-MCP.

ACC can also negatively affect root growth independently of ethylene signaling (**Figure 2A**). Col-0 seedlings were grown on increasing concentrations of ACC in the absence or presence of 250 ppm 1-MCP. Both ACC and 1-MCP dramatically altered root growth. In the absence of 1-MCP, a reduction in root length was already apparent at 10 µM ACC (**Figure 2B**), consistent with previously reported effects of ethylene on root growth inhibition (Le et al., 2001). In contrast, at the same concentration of ACC in the presence of 1-MCP, a much smaller inhibition was observed (**Figure 2B**). A five-fold higher dose was required for an effective inhibition of root elongation in Col-0 plants subjected to 1-MCP treatment (**Figures 2A**, **B**). It is conceivable that the observed effects of ACC on both rosette and root growth, in the presence of 1-MCP, are due to an ethylene signal remaining present under an insufficient dose of 1-MCP. However, the binding affinity of 1-MCP to the ethylene receptors was demonstrated to be at least 10 times greater compared to ethylene (Hall et al., 2000; Binder et al., 2004). Additionally, previous reports demonstrated that 500

ppb of ethylene is sufficient to mimic the phenotypic effects of 50 µM ACC at the level of rosette growth inhibition (Vaseva et al., 2018). Moreover, a treatment of etiolated seedlings with 1 mM ACC was shown to give rise to ethylene levels ranging from 1 to 10 ppm (Woeste et al., 1999). Altogether, these findings support the interpretation that 250 ppm of 1-MCP is more than sufficient to block the effects of 500 µM ACC. To further corroborate this assumption, plants were treated for 2 weeks with ethylene in the presence of 1-MCP to assess ethylene insensitivity (**Figure S3**). A dose of 100 ppm ethylene was chosen to effectively supersede the effects of 500 µM ACC. WT plants supplemented with ethylene displayed a typical dwarfed phenotype and reduced root growth, in the absence of 1-MCP (**Figure S3**). However, these phenotypic effects disappeared when ethylene perception was blocked with 100 ppm 1-MCP, or similarly in *ein2-1*, which has a defective ethylene signaling pathway and is fully unresponsive to ethylene. Thus, ACC, at the concentrations tested, is capable of reducing overall growth during early plant development, independently of ethylene.

#### Effects of ACC on Vegetative Growth in the Presence of ACO Inhibitor AIB

To further investigate the role of ACC in vegetative development, experiments using the ACO inhibitor AIB were conducted. Ethylene emanation was measured in 2-week-old plants grown on increasing doses of ACC in the absence and presence of 2 mM AIB (**Figure 3A** and **Table S2**). A dose-dependent increase in ethylene levels was observed when plants were grown on ACC-containing media. In addition, AIB effectively blocked ACO-mediated conversion of ACC to ethylene, though a small increase in ethylene levels, could be observed at higher concentrations (50 µM ACC; **Table S2**). Nevertheless, the ethylene levels in plants treated with 50 µM ACC + AIB were more than two-fold lower than those in plants treated with 1 µM ACC alone (**Figure 3A** and **Table S2**). Concentrations of ACC exceeding 50 µM were henceforth omitted, since in such conditions 2 mM AIB was unable to outcompete

ACC for binding to the catalytic site of ACO, resulting in notable ethylene emanation. Moreover, if the phenotypic effect upon treatment with 50 µM ACC + AIB was stronger compared to that of 1 µM ACC alone, it was considered a *bona fide* ACC effect.

In subsequent experiments, the phenotypic responses upon AIB with increasing doses of ACC were investigated both at the level of rosette and root growth (**Figure 3B**). Fifty micromolar of ACC was capable of reducing rosette expansion of 2-week-old WT Col-0 plants treated with 1-MCP (**Figures 1A**, **B**). The addition of 2 mM AIB resulted in a slight decrease in rosette area upon 50 µM ACC, compared to 0 µM ACC (**Figures 3B**, **C**). In *ein2-1* plantlets, 50 µM ACC decreased rosette size substantially in the absence of AIB. However, in the presence of 2 mM AIB, rosette area reduced only slightly upon treatment with 50 µM ACC (**Figures 3B–D**). Contrarily, when roots were evaluated after 2 weeks of growth on vertically standing plates, a dose-dependent reduction in root length was discovered in Col-0 and *ein2-1* even in the presence of 2 mM AIB (**Figures 4A–C**). In Col-0, 50 µM ACC decreased the average root length from 39.17 mm to 5.57 mm in the absence of AIB. In the presence of AIB, root length was decreased from 22.37 mm to 5.35 mm (**Figure 4B**). Likewise, 50 µM ACC reduced root elongation substantially in *ein2-1* both without and with AIB supplementation (**Figure 4C**). Since the application of 50 µM ACC + AIB led to ethylene levels lower than those observed upon 1 µM ACC alone (**Figure 3A**), and the inhibitory effect of the latter dose was relatively small for both rosettes and roots, the stronger inhibitory effect observed on 50 µM ACC + AIB is assumed to be a true ACC effect (**Figures 3B**, **C** and **4B**, **C**). In conclusion, AIB hinders the growth-inhibiting effect of ACC in light-grown plantlets, in an organ-dependent manner.

#### ACC Effects During Skotomorphogenic Development

As ACC/ethylene affect skotomorphogenic development as well, an ethylene-independent role for ACC was investigated

in etiolated seedlings (**Figure 5**). First, ethylene emanation by seedlings treated with ACC, AIB or a combination of both, was measured using laser-based photoacoustic spectroscopy (**Figure 5A**). AIB was able to effectively reduce ethylene synthesis to negligible levels up to doses of 10 µM ACC. At 10 µM ACC, AIBtreated seedlings emitted ethylene levels (µ = 1.29 pl seedling−1 h−1) comparable to seedlings subjected to 0.1 µM ACC without AIB (µ = 1.00 pl seedling−1 h−1) (**Table S3** and **Figure 5A**). A significant rise in ethylene levels could be observed at 50 µM ACC + AIB (µ = 6.50 pl seedling−1 h−1), which was similar to ethylene levels released upon treatment with 0.75 µM ACC alone (µ = 5.96 pl seedling−1 h−1) (**Table S3** and **Figure 5A**). Therefore, differences in phenotypic effects between the aforementioned treatments are indicative for true ACC effects.

Next, an AIB dose-response assay employing Col-0 and *ein2-1* was carried out on 4-day-old etiolated seedlings (**Figures 5B–F**). In Col-0, a dose-dependent inhibition of hypocotyl (**Figure 5C**; **Figure S4A**) and root (**Figures 5D** and **S4B**) elongation was observed, with roots being slightly more responsive to ACC at lower concentrations (as of 0.1 µM). Upon the addition of the ACO inhibitor AIB, hypocotyls demonstrated significant resistance to lower concentrations of ACC (0.5–10 µM), while reacting comparable to the wild type upon larger doses (>50 µM) (**Figure 5C**). For instance, 1 µM ACC reduced the average hypocotyl length from 10.88 mm to 4.08 mm, while in the presence of AIB, it only decreased from 9.81 mm to 8.11 mm (**Figure 5C**). In contrast, 50 µM ACC strongly inhibited hypocotyl length, irrespective of AIB treatment. A similar response was observed in AIB-treated roots (**Figure 5D**). In the ethylene-insensitive mutant *ein2-1*, ACC was able to significantly reduce hypocotyl and root growth as well, though to a much smaller extent than in Col-0 treated with AIB (**Figures 5B**, **E–F** and **S4C, D**). At 50 µM ACC and in the absence of AIB, hypocotyl length was merely reduced from 11.46 mm to 8.57 mm (**Figure 5E**).

Given that in Col-0, 10 and 50 µM ACC in the presence of AIB resulted in stronger inhibitory effects compared to 0.1 and 0.75 µM ACC, respectively, in the absence of AIB (**Figures 5B**, **C**), we hypothesized that these represent *bona fide* ACC responses. An ethylene complementation assay was conducted to rule out the possibility that residual ethylene biosynthesis upon AIB treatment is the cause of the observed phenotype in darkgrown seedlings (**Figure S5**). Specifically, the phenotypic effects

of the residual levels of ethylene emanated by seedlings treated with either 10 µM ACC + AIB [designated ETH (10); 116 ppb] and 50 µM ACC + AIB [designated ETH (50); 585 ppb] were investigated (**Table S3**). When seedlings were treated with ETH (10), hypocotyls and roots were almost indistinguishable from the mock treatment (**Figures S5A–C**). In the presence of 2 mM AIB, the effect of ETH (10) was slightly larger in both organs.

However, the effect of 10 µM ACC + AIB on hypocotyl and root elongation was stronger than that of ETH (10) [**Figures S5A–C**; compare ETH (10) with 10 µM ACC; gray bars]. Similarly, the effect of ETH (50) was less pronounced compared to 50 µM ACC + AIB [**Figures S5A–C**; compare ETH (50) with 50 µM ACC; gray bars]. Nevertheless, it is clear that the inhibitory effect on elongation of ETH (50) is stronger than that of ETH (10).

FIGURE 5 | Triple response development of Arabidopsis wild-type Col-0 and ethylene insensitive ein2-1 on increasing concentrations of ACC with and without AIB treatment. Arabidopsis wild-type Col-0 and ethylene insensitive mutant ein2-1 were sown in darkness on 0, 0.1, 0.25, 0.5, 0.75, 1, 5, 10, and 50 µM ACC with or without treatment with 2 mM AIB. (A) Ethylene production rates of 4-day-old etiolated Col-0 seedlings (n = 30; 4 independent replicates). (B) Pictures of representative 4-day-old etiolated Col-0 and ein2-1 seedlings. (C) Hypocotyl length and (D) root length of Col-0 seedlings (38 ≤ n ≤ 55; 3 independent replicates). (E) Hypocotyl length and (F) root length of ein2-1 seedlings (22 ≤ n ≤ 42; 3 independent replicates). Different letters indicate statistically significant differences between the different groups (Kruskal-Wallis, Dunn's Multiple Comparison Test, P < 0.01). Error bars are SD. The relative values of panels (B–E) are depicted in Figure S4. Effect sizes are presented in Table S1. P < 0.01 is indicated with \*.

Interestingly, AIB reduced root length in both Col-0 and *ein2- 1*, even in the absence of exogenous ACC or in the presence of ethylene (**Figures 4B**, **C**, **5D**, **F**, and **S5C**). In darkness, Col-0 and *ein2-1* roots were approximately 25% shorter when treated with 2 mM AIB (**Figures 5D**, **F**). To assess whether this inhibitory effect is caused by an accumulation of endogenous ACC or is due to a side effect of AIB, we investigated the response of the *acs8x* mutant, which is almost completely devoid of endogenous ACC, to AIB treatment (**Figure 6**). Etiolated *acs8x* seedlings exhibited significantly longer hypocotyls and shorter roots compared to the WT (**Figures 6A–C**), consistent with previous reports (Tsuchisaka et al., 2009). Upon addition of 2 mM AIB, both WT and *acs8x* roots were 30% shorter compared to roots in absence of AIB (**Figures 6C**, **E**). Furthermore, *acs8x* hypocotyls were slightly more sensitive to AIB compared to the WT (**Figures 6B**, **D**). These results strongly indicate that the response of seedlings to AIB is not related to an accumulation of ACC, as *acs8x* seedlings were not resistant to the treatment.

Altogether, these data corroborate an ethylene-independent role for ACC during skotomorphogenic shoot and root development (**Figure 5**), in addition to its negative effect on rosette and root development in light conditions (**Figures 1** and **4**). Furthermore, ACC evokes distinct responses in ethylenesensitive compared to -insensitive backgrounds.

### DISCUSSION

Since the discovery of ACC as a biosynthetic precursor of the plant hormone ethylene about 4 decades ago (Adams and Yang, 1979), most studies were focused on identifying the mechanisms controlling ethylene synthesis and its subsequent *in vivo* responses. Few studies, however, investigated whether ACC has a non-canonical function, independent of ethylene signaling or bypassing ethylene perception. Here, we provide evidence for a role of ACC in the regulation of early vegetative development in *Arabidopsis thaliana.* Using chemical inhibitors of ethylene biosynthesis and perception, in WT and in the ethyleneinsensitive mutant background *ein2-1*, our study demonstrates that ACC can act as a negative regulator of rosette development (**Figures 1**, **3** and **S1**, **S2**), hypocotyl elongation in darkness (**Figures 5** and **S4**), and root growth in both light (**Figures 2** and **4**) and etiolated conditions (**Figure 5**), in parallel of ethylene perception. Based on these findings, we propose to revisit the current model on ACC/ethylene biosynthesis and signaling, including ACC-specific responses (**Figure 7**).

In roots, ACC or ethylene treatment is known to cause growth inhibition through a reduction in LEH, in a concentration- and time-dependent manner (Le et al., 2001). Specifically, ethylene perception in the epidermal layer of the transition zone is

acs8x seedlings. (B) Hypocotyl length and (C) root length of Col-0 and acs8x seedlings (13 ≤ n ≤ 26; 3 independent replicates). (D) Relative hypocotyl length and (E) relative root length of acs8x seedlings. In panels (D) and (E), lengths are expressed relative to the corresponding genotype on medium without AIB. In panels (B) and (C), different letters indicate statistically significant differences between the different groups (Kruskal-Wallis, Dunn's Multiple Comparison Test, P < 0.01). T-tests were carried out in panels (D) and (E) to compare differences within a genotype (P < 0.01 is indicated with \*; P < 0.0001 is indicated with \*\*\*). Error bars are SD. Effect sizes are presented in Table S1.

FIGURE 7 | Revisited model for ACC/ethylene biosynthesis and signaling. (A) ACC/ethylene signaling pathway in normal conditions, where the responses are mainly mediated by ethylene. A major portion of ACC is converted to ethylene by ACO. The signal is then transferred via key signaling components ultimately leading to ethylene responses. The remaining ACC is transported to other tissues, conjugated to storage forms or perceived by (a) putative ACC receptor(s). Subsequent ACC responses are assumed to be minimal. (B) ACC/ethylene signaling pathway in conditions where ACC accumulates and leads to ACC-specific responses. ACC accumulation can result from an inhibition of ACO by inhibitors (e.g., AIB; shown in red), saturation of ACO activity (leading to overflow), or any other inhibition of ACO caused by internal or external signals. Elevated levels of ACC bind to its receptor(s), activating unidentified downstream components, ultimately leading to ACC-specific responses. Alternatively, ACC could, upon interaction with its receptor(s) act via EIN2, activating ethylene-mediated gene expression, while bypassing the need for ethylene perception. Additionally, the ACO inhibitor AIB potentially affects other ACC binding proteins such as putative receptors, transporters or conjugating enzymes. Arrow-headed lines represent stimulatory interactions. Bar-headed lines represent inhibitory interactions. Thick lines depict the predominantly active pathway. Dashed lines indicated relations that have not been demonstrated experimentally. 1-MCP, 1-methylcyclopropene; ACC, 1-aminocyclopropane-1 carboxylic acid; ACO, ACC oxidase; ACS, ACC synthase; AIB, 2-aminoisobutyric acid; CTR, constitutive triple response; ETR, ethylene response; ERS, ethylene response sensor; EIN, ethylene insensitive; EIL, EIN3-like.

sufficient to inhibit root growth (Vaseva et al., 2018). Moreover, this response is mediated by changes in intracellular auxin levels concomitant with an apoplastic alkalinization, which in turn regulates the activity of cell-wall loosening agents and peroxidases (Staal et al., 2011; Barbez et al., 2017; Vaseva et al., 2018). Though ACC and ethylene root responses are certainly overlapping, our data demonstrate that ACC can affect primary root growth independently of ethylene perception and/or signaling (**Figures 2**, **4**, and **5**). We propose that this role might be either specific to ACC or related to a feedback regulatory mechanism involving auxin, mimicking ethylene-mediated root inhibition (Vaseva et al., 2018). An ethylene-independent role for ACC as a signal driving root growth has been proposed previously (Xu et al., 2008; Tsang et al., 2011). They suggest that ACC is involved in the sensing of cell wall integrity, a feature crucial in the control of root elongation. It is conceivable that crosstalk between ACC and ethylene exists, since the latter also acts on factors related to cell wall integrity. Until now, no role for ACC in the regulation of hypocotyl or shoot growth has been thoroughly characterized. Higher order mutants of *ACS* genes, however, do exhibit reduced branching, a phenotype clearly distinct from single ethylene signaling mutants (Tsuchisaka et al., 2009). In addition, embryo lethality in a homozygous octuple *acs* mutant points to a primary role of ACC in early vegetative growth (Tsuchisaka et al., 2009). Contrarily, a non-canonical ethylene pathway, such as the controversial CTR1-MKK9– MPK3/MPK6-EIN3 signaling cascade, could regulate embryo development as well (Yoo et al., 2008). Analysis of an *ACO* null mutant or a heptuple *etr1ers1etr2ers2ein4ctr1ein2* will shed light on this issue. Our work supports a role for ACC as a regulator of hypocotyl and shoot growth, independently of ethylene (**Figures 1**, **3**, and **5C**). Ethylene is known to regulate hypocotyl elongation and leaf expansion *via* changes in cell wall integrity and microtubule orientation (Collett et al., 2000; Le et al., 2005; Pierik et al., 2007). In addition, downstream effects on auxin and gibberellin homeostasis are linked to ethylene-mediated shoot growth as well (Vriezen et al., 2004; Stepanova et al., 2007). It is unclear how ACC controls expansion in shoot tissues, but it could involve ACC-specific effects or depend on crosstalk with the abovementioned growth hormones too.

Sensitivity of plants to ACC or ethylene, as for other hormones (e.g., auxin) depends on the concentration, tissue, organ, developmental stage, species, and growth conditions (Abeles et al., 1992; Le et al., 2001; Vandenbussche et al., 2012; Vaseva et al., 2018). *Arabidopsis* roots are generally more sensitive to ACC/ethylene [> 0.1 µM in darkness (**Figure 5D**); > 1 µM in light (**Figure 4B**)] compared to hypocotyls (> 0.25 µM; **Figure 5C**) or rosettes (> 10 µM; **Figure 3C**). When ethylene effects were excluded using pharmacological or genetic approaches, a lower sensitivity to ACC could be observed in all tissues studied (**Figures 1**–**5**). The need for the application of relatively high doses of exogenous ACC is consistent with an "overflow model". Low doses of ACC are converted to ethylene by ACO (**Figure 7A**), whereas increasing concentrations of ACC are supposed to accumulate intracellularly due to ACO limitation, ultimately leading to ACC-specific responses (**Figure 7B**). Nevertheless, the possibility that phenotypic changes are due to toxicity rather than *bona fide* ACC effects when applied at higher doses cannot be excluded. However, this notion holds true for all pharmacological experiments with compounds that are not 100% pure. In addition, the activity of ACC most likely depends on the tissue, developmental stage, and environmental conditions. During flooding, for instance, ethylene evolution in tissues subjected to hypoxia is limited due to the lack of oxygen, a key factor for ACO activity (**Figure 7B**; Adams and Yang, 1979; Vanderstraeten and Van Der Straeten, 2017). On the one hand, ACC can act as a mobile signal, which is transported from root-to-shoot, where it can be converted to ethylene and induce the appropriate phenotypic response to stress (e.g., hyponasty, petiole and shoot elongation, etc.), if oxygen is present. On the other hand, upon complete submergence, ACC itself might act as a growth-limiting factor in roots and shoots, contributing to a quiescence strategy, eventually leading to enhanced survival. The precise role of ACC in these conditions should be further evaluated.

Whether ACC acts as a signal completely independent of ethylene signaling or whether it can interact with ethylene signaling components downstream of perception remains a key question to be resolved in future studies (**Figure 7B**). Given that a depletion of ACC confers embryo lethality and reduces branching, in contrast to ethylene insensitivity, distinct roles for ACC are probable (Tsuchisaka et al., 2009). Moreover, the stimulation of GMC division by ACC does not require the major ethylene signaling components, further hinting at an ethylene-independent ACC pathway (Yin et al., 2019). Tsang et al. (2011) have postulated the second option. ACC could act as a shortcut, bypassing the need for ethylene perception. In the latter scenario, ACC would have the capacity to elicit certain responses before the required threshold levels of ethylene are reached or when the synthesis of ethylene is hampered (e.g., upon hypoxia). Our results support the latter hypothesis, since at least with respect to skotomorphogenic development—*ein2-1* is less sensitive to ACC compared to the WT supplemented with AIB (**Figures 5C–F**). Contrarily, a negative effect of ACC on rosette development or primary root development (in light conditions) appears to occur independently of ethylene (**Figures 1**–**4**), downstream or in parallel with EIN2. A thorough analysis employing different ethylene signaling mutants in combination with pharmacological treatments will shed light on the events occurring downstream of ACC. On the other hand, the differences in root growth inhibition by ACC could be stageor light-dependent. For instance, a recent report demonstrated that root growth defects upon phosphate deficiency are related to blue light illumination, and are suppressed by growing roots in darkness, as usually the case in nature (Zhang et al., 2019). Furthermore, though we observe the same effect in the absence of sucrose, we cannot rule out the possibility that ACC effects are modulated by sucrose. Sucrose is known to affect various vegetative growth processes (Gibson, 2005). It was shown that 1% sucrose decreased hypocotyl length and enhanced root length in etiolated seedlings (Lu and Wen, 2019). Therefore, the degree of ethylene response diminished in hypocotyls, while being enhanced in roots. In addition, increasing concentrations of sugar reduces the stability of EIN3, the major transcription factor of the ethylene signaling pathway (Yanagisawa et al., 2003). Less is known about the effects of sucrose—and its interaction with ethylene—on rosette development. High levels of sucrose or glucose are inhibitory for growth (Zhou et al., 1995). Here, we demonstrated that 1% sucrose increased rosette size, though this effect disappeared upon increasing doses of ACC, irrespective of the genetic background (**Figure S1**). Hence, the growth inhibitory properties of ACC are not dependent on the presence of sucrose.

Interestingly, differences in ACC responsiveness could be observed between AIB and 1-MCP application. While hypocotyl or root growth was equally inhibited by ACC in the presence of AIB or 1-MCP (**Figures 2**, **4**, and **5**), rosettes were less responsive to ACC upon combination with AIB (**Figures 1** and **3**). These differences did not arise due to ethylene-related effects, as the concentrations used were considered sufficiently effective. Two millimolar of AIB effectively blocked ACOmediated conversion of ACC to ethylene in plants treated with 50 µM ACC (**Figure 3B**). Furthermore, residual ethylene emanation was even lower compared to a treatment with 1 µM ACC without AIB (**Table S2**). For 1-MCP treatments, a dose of 100 ppm completely blocked the phenotypic effects induced by 100 ppm ethylene (**Figure S3A**). Therefore, notwithstanding the fact that a strict comparison remains difficult, the observed discrepancy is most likely caused by another factor. To further corroborate the results obtained with AIB, parallel analyses including higher order *aco* mutants as well as their crosses with *ein2-1* will be instrumental. Blocking the conversion of ACC to ethylene in a pentuple *aco* mutant would also lead to an excess of endogenous ACC similar to a treatment with AIB. However, such a mutant has not been constructed to date, and it remains to be seen whether it is even viable.

AIB was discovered as an ACO inhibitor, based on its structural similarities with ACC (Satoh and Esashi, 1982; Pirrung et al., 1998). Hence, it is plausible that AIB also has the capacity to interact with other ACC binding proteins, be it (a) putative ACC receptor(s), conjugating enzymes or ACC transporters (**Figure 7B**). The decrease in ACC sensitivity at the level of rosette expansion when AIB is present (**Figure 3**), hints at possible competition for binding to the putative ACC receptor. On the other hand, in the presence of AIB, one might expect an increase in endogenous ACC levels resulting from a feedback effect at the level of ACS, leading to ACC effects at lower concentrations of exogenous ACC. However, this is not supported by our analysis. Moreover, AIB still inhibited root growth in the *acs8x* mutant, suggesting that AIB exhibits side effects unrelated to the accumulation of ACC (**Figure 5**). No additive effects were observed upon a combined application of AIB and low concentrations of ACC (**Figure 5D**), indicative for the absence of a genuine side effect. Therefore, it is more conceivable that AIB and ACC act on the same target, such as a putative ACC receptor. Different ACC receptors could have different sensitivities to ACC or structural analogues (e.g., AIB) and can be expressed in a tissue-specific manner or are developmentally regulated, explaining the variation in AIB sensitivity among tissues and conditions. Furthermore, our data do not determine whether AIB exhibits agonist or antagonist properties for the putative ACC receptor(s). Lastly, it remains to be clarified whether ACC itself, a derivative or a downstream element is the *bona fide* signal. The function of all three ACC conjugates is largely unknown (Amrhein et al. 1981; Kionka and Amrhein 1984; Martin et al. 1995). They are proposed to act as storage forms, depleting the pool of free ACC, but could also have signaling functions.

With this work, we demonstrated that ACC could function as a regulator of growth during early vegetative development, apart from its role as an ethylene precursor. Hence, researchers employing ACC as an ethylene precursor should be mindful of putative ACC effects confounding ethylene responses. The exact mechanism underlying the ACC response, however, remains to be identified. The discovery of a putative receptor will shed light on the molecular players involved in the ACC response, either directly or downstream. Additionally, phenotypic, genetic and transcriptomic analyses of higher order *aco* mutants and the identification of ACC transporters and conjugating enzymes *in vivo*, are merely few of the studies required to elucidate ACC metabolism and signaling. Though ACC research is still relatively uncharted territory, novel

#### REFERENCES


findings related to the non-canonical role of ACC and the related molecular mechanisms will open up exciting new avenues in plant hormone physiology, shedding light on the complex signaling networks shaping plant growth and development.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/ **Supplementary Material**.

### AUTHOR CONTRIBUTIONS

All authors conceived and designed the research. LV, TD, and SB conducted the experiments. LV, TD, and DS analyzed the data. LV, TD, and DS wrote the first draft of the manuscript. All authors read and commented on the manuscript. DS coordinated the project.

### FUNDING

DS gratefully acknowledges financial support from the Research Foundation Flanders (project G032717N) and Ghent University (Bijzonder Onderzoeksfonds, BOF-BAS 2018). LV is indebted to the Research Foundation Flanders for an SB-PhD fellowship (1S17917N).

#### ACKNOWLEDGMENTS

The authors thank Mira Haegman, Els Fostier, and Piet Vankeirsbilck for technical assistance with the ethylene and 1-MCP treatments.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01591/ full#supplementary-material


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Vanderstraeten, Depaepe, Bertrand and Van Der Straeten. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*