Di-ACC was synthesized by GenScript Biotech and confirmed by mass spectrometry and HPLC to have a purity of 95.6%. Because the synthesis used Fmoc-ACC (Fluorenylmethoxycarbonyl-ACC) as a building block, which may be converted to ACC in planta, we further purified the di-ACC by HPLC to eliminate any residual ACC contamination. Di-ACC fractions in 0.1% TFA were collected after passing over a Hypercarb^TM column (Thermo Scientific) using acetonitrile as solvent (Supplementary Figure S1A), and purified samples run again on the same column to confirm elimination of contaminant peaks (Supplementary Figure S1B). ACC was run independently to enable calculation of the retention time.

Arabidopsis thaliana ecotype Col-0 was used as wild-type control, the etr1-1 (AT1G66340; N237) EMS mutant and lht1-5 (AT5G40780; SALK_115555C) mutant were obtained from NASC. Seeds were surface sterilized by incubation with 70% ethanol (v/v) for 2 mins, followed incubation in 5% bleach (NaOCl) (v/v) for 5 mins. Seeds were then washed in sterile H₂O and plated in Petri dishes containing 0.5 × Murashige and Skoog (MS) basal salt medium with 1% plant agar (pH 5.7). Media was supplemented with ACC or di-ACC at concentrations indicated. Seeds were imbibed at 4°C for 3 days before transfer to the growth room. For triple response phenotyping and ethylene measurements, seedlings were grown for 4 days in the dark at 21°C before measurements. Root and shoot length was quantified using ImageJ (http://rsbweb.nih.gov/ij/) from photographs. For Nano-PALDI MSI, seedlings were grown for 10 days in a growth chamber (16 h light/8 h dark; 21°C) before treatment with ACC or di-ACC. For triple response assays, plants were treated with 0.5, 1 or 20 μM ACC and increasing concentrations of di-ACC of 0.5, 1, 2, 5, 10, 20, 50 μM. For ethylene measurements, plants were treated with 0.1, 0.5 and 1 μM ACC, and 2 and 10 μM di-ACC.

Marchantia polymorpha Australian ecotype (MEL accession) wild type (“WT1” female) and the Mpein3 mutant used in this study were previously described in Li et al. (2020). Single gemma were plated on solid half-strength Gamborg's B5 salts medium (pH 5.5) supplemented with 0.2% sucrose and 1% plant agar. ACC (100 μM) or di-ACC (10 and 100 μM) was added to the media prior to plating. Plants were grown at 21°C under sunlight-mimicking LED lights at 200 μmol.m⁻².s⁻¹ for a 16/8 h day/night regime. Plants were imaged after 15 days under a stereoscopic microscope (Olympus SZX9) equipped with a camera (Toupcam, Touptek Photonics) or photographed directly with a digital camera (Nikon D3400). The two-dimensional plant area (thallus size) was quantified using ImageJ.

For in planta ethylene measurements, 1 ml of headspace gas from seedlings grown in 4 ml airtight GC vials was injected into a gas chromatograph (Shimadzu GC2014) as described in Houben et al. (2022).

Nano-PALDI MSI was carried out as previously described (Shiono et al., 2017; Ortigosa et al., 2022; Taira and Shiono, 2022), but with a few modifications for small tissues. Briefly, 10 day-old seedlings of Arabidopsis ecotype Col-0 were incubated in 2 μM ACC or di-ACC overnight. The seedlings were then embedded in Super Cryoembedding Medium (SCEM, Leica) and frozen in liquid nitrogen. The specimen block was cut into 10 μm sections using a cryostat (NX-70, Thermo Fisher Scientific, USA) set at −23°C in the chamber and at −25°C on the object holder. The sections were mounted on slides coated with indium tin oxide (ITO) (Bruker Daltonik GmbH). Optical images of sections were obtained by a virtual slide scanner (NanoZoomer-SQ, Hamamatsu Photonics) before analysis by Nano-PALDI-MSI.

For nano-PALDI-MSI, iron oxide-based nanoparticles (Fe-NPs) were prepared by stirring aqueous solutions of FeCl₂·4H₂O (5 ml, 100 mM), and 3-aminopropyltriethoxysilane (5 ml; γ-APTES) at room temperature for 1 h. The resulting precipitate was washed several times with ultrapure water and then suspended in methanol. One mg Fe-NPs was suspended in 1 ml methanol and sprayed on tissue sections on ITO-coated glass slides with an airbrush (nozzle caliber, 0.2 mm). To obtain MS images, each data point on the section were irradiated with 200 laser shots in the positive ion detection mode of the mass spectrometer. Only signals between 80 and 500 m/z were analyzed to detect the correlated ACC (m/z 102.1) and ACC dipeptide (m/z 185.0) as protonated ions, respectively. For each section, ~100,000 data points were obtained, 5 μm apart. The MS image was reconstructed from the obtained MS spectra with a mass bin width of m/z ± 0.05 from the exact mass using flexImaging 5.0 (Bruker Daltonik GmbH). The accurate mass of the ions was used for image generation, and mass accuracy and root-mean-square error (RMSE) were automatically calculated by the imaging software to avoid false-positive signals.

The coding sequence of AtACO2 (At1g62380) was amplified and cloned into the pET28a bacterial expression vector, and transformed into E. coli strain BL21 (DE3). Starter cultures were grown in 5 ml LB media with 50 μg/ml kanamycin and chloramphenicol overnight at 37°C shaking at 200 rpm. Starter cultures were used to inoculate 2 L cultures, which were grown at 37°C shaking at 100 rpm until an OD600 of 0.8 was reached, measured using an Ultrospec 10 Cell Density meter (Amersham Biosciences). Cultures were then cooled on ice for 30 mins, protein expression induced by adding IPTG to a final concentration of 0.5 mM, and then grown overnight at 18°C shaking at 100 rpm. Cells were pelleted by centrifugation at 5,500 rpm for 30 mins at 4°C, resuspended in lysis buffer, rotated at room temperature for 30 mins, and sonicated with a Branson 450 sonifier (VWR) with a power of 3 and duty cycle of 80, for six cycles of 1 min sonication and 2 mins rest on ice. Cell debris was removed by centrifugation for 40 min at 13,000 rpm at 4°C, and filtration of the supernatant through a 0.45 μm filter. AtACO2 was first purified by IMAC chromatography (Qiagen Ni-NTA, 5–250 mM Imidiazole, 200 mM NaCl, 50 mM NaH₂PO₄, 1 mM DTT, pH 8) followed by anion exchange chromatography on an Äkta^TM Prime Plus chromatography system (GE Healthcare) fitted with a 5 ml HiTrap^TM Q HP column (Cytiva) (25–1,000 mM NaCl, 50 mM NaH₂PO₄, 1 mM DTT, pH 8). SDS PAGE confirmed protein purity at every step. Protein concentration was calculated using a NanoDrop^TM 2000 spectrophotometer (Thermo Scientific).

For in vitro ACO activity measurements (Van de Poel et al., 2014), recombinant purified AtACO2 (5 μg) was incubated in freshly prepared activity buffer containing 50 mM MOPS, 5 mM ascorbic acid, 20 mM sodium bicarbonate, 10% glycerol, 0.1 mM DTT and the mentioned ACC or di-ACC concentration. The reaction was incubated in 4 ml airtight GC vials for 60 min at 30°C while shaking. One ml of headspace was sampled and analyzed for ethylene content using gas chromatography as described above.

The di-ACC-ACO2 complex was modeled in Molecular Operating Environment Software (MOE) (Chemical Computing Group ULC), using crystal structures of Arabidopsis ACO2 (PDB 5GJ9) and the ACC-bound ACO of Petunia hybrida (PDB 5TCV) (Sun et al., 2017). Amber10 was used for complex simulation and calculation of partial charges (Case et al., 2008).

Data was analyzed in Graphpad Prism version 8.0.2. The Shapiro-Wilk normality test was performed, and normally distributed data tested by ordinary one-way ANOVA and a Tukey post-hoc test. When the assumption of normality was not met, a non-parametric Kruskal–Wallis one-way ANOVA test was used. For ethylene production, biological variation between data sets was high, so a Welsh ANOVA with Holm-Sidak's multiple comparisons test was used.

In analogy to the existence of other non-proteinogenic dipeptides in plants, we hypothesized that ACC may also exist as a dipeptide (Figure 1A), and as such may represent a novel ACC derivative which could influence ethylene biosynthesis or ACC signaling. To investigate this possibility, we had di-ACC custom synthesized, followed by HPLC purification to remove residual trace amounts of ACC from the synthesis process (Supplementary Figure S1). First, we examined whether di-ACC is able to evoke an ethylene response in plants, deploying the characteristic triple response assays of 4-day-old dark-grown Arabidopsis seedlings. We observed that di-ACC induces shorter roots and hypocotyls, and an exaggerated apical hook, reminiscent of ethylene responses (Figures 1B–D). However, these effects were only observable at higher di-ACC concentrations as compared with ACC treatment, with hypocotyl shortening only seen with di-ACC treatment of 5 μM or higher.

Since the di-ACC is a newly synthesized molecule, we have no data on its stability. It is therefore possible that the ACC dipeptide was broken down during the course of experiments (for example by thermal instability) and release ACC, leading to the observed triple response phenotype (Figures 1B–D). To rule out di-ACC instability, we performed a triple response assay including di-ACC which had been boiled for 1 h prior to making media plates (Figures 1E,F). Hypocotyl and root length of seedlings treated with boiled di-ACC were not significantly different from those treated with the same concentration of fresh di-ACC, indicating that phenotypes seen with higher di-ACC concentrations are not due to release of free ACC from degradation of di-ACC. We also included boiled ACC as a control, which also shows similar activity as fresh ACC (Supplementary Figure S2), indicating that both ACC and di-ACC are thermostable molecules.

To investigate the observation that di-ACC evokes an ethylene response through the synthesis of ethylene (similar as ACC), we measured ethylene production of 4-day-old dark-grown seedlings treated with ACC and di-ACC (Figure 1G). This revealed that a di-ACC treatment is able to elevate ethylene production levels, albeit a ten-fold higher concentrations of di-ACC is needed to get similar levels of ethylene compared to ACC. This is similar to the requirement for higher concentrations of di-ACC to generate triple response phenotypes (Figures 1B–D), as compared to ACC. Overall, this proves that di-ACC is converted into ethylene in planta.

Figure 1 showed that di-ACC is able to augment ethylene production and evoke ethylene responses in Arabidopsis seedlings (reviewed by Li et al., 2022). To validate if the triple response phenotype of di-ACC treated plants (Figures 1C,D) is ethylene-dependent, we phenotyped the ethylene insensitive mutants etr1-1. The etr1-1 mutant shows complete insensitivity toward both ACC and di-ACC for both the root and the hypocotyl (Figures 2A,B), suggesting that the triple response evoked by di-ACC occurs via ethylene signaling by the ethylene receptor. In order to determine if the ACC dipeptide is taken up and transported in a similar way as ACC, we evaluated the triple response phenotype of the lht1-5 mutant, harboring a reduced ACC uptake due to a dysfunctional Lysine Histidine Transporter1 (LHT1) (Shin et al., 2015). In the lht1-5 mutant, the di-ACC treatment causes a partial triple response phenotype, with a reduced root and hypocotyl length, but not as severe as compared to the Col-0 background (Figures 2B,C). This indicates that di-ACC, similar as ACC, is taken up in part via LHT1, but also that di-ACC can still be taken up by other transporters.

To visualize ACC and di-ACC in planta we made use of Nano Particle-Assisted Laser Desorption/Ionization mass spectrometry imaging (Nano-PALDI MSI), which has recently been used to identify the spatial distribution of hormones and metabolites in plant tissue sections (Shiono et al., 2017; Ortigosa et al., 2022; Taira and Shiono, 2022). Ten day-old Arabidopsis seedlings were treated with 2 μM ACC, di-ACC or both overnight, before hypocotyl sections were analyzed using Nano-PALDI MS imaging (Figure 2C). Untreated control samples show a weak signal for ACC in tissue mainly surrounding of the vasculature, whereas there is no detection of the di-ACC, suggesting di-ACC is naturally not present in Arabidopsis hypocotyls (or not detectable). When ACC was fed, a strong ACC mass signal (m/z 102.1) was seen in all tissues, while no di-ACC mass signal (m/z 185.0) was observed in the sample. A di-ACC treatment showed that di-ACC was taken up by the sample, and largely confined to the vascular tissue. When both ACC and di-ACC were fed, ACC was observed much more broadly in the surrounding tissues, while di-ACC seemed more restricted to the vasculature. This tissue-specific accumulation suggests that di-ACC is less efficiently transported from the vasculature to the surrounding tissues compared to ACC. Furthermore, we observed a relatively lower signal intensity for di-ACC compared to ACC, suggesting that di-ACC has a lower affinity for ACC transporters, and consequently ACC can outcompete di-ACC for uptake.

In di-ACC treated plants, the ACC signal may look more intense compared to control plants, suggesting di-ACC may be catabolized to ACC. However, it should be noted that comparisons of Nano-PALDI MS images are semi-quantitative, because each signal value was derived from the relative intensity normalized by the highest intensity spot per slide. Therefore we cannot be certain that catabolism of di-ACC to ACC takes place in planta. Nonetheless, the restricted spatial distribution of the di-ACC is a clear indicator of its reduced mobility in planta compared to ACC. Additionally, the requirement for higher concentrations of di-ACC to evoke the triple response (see Figures 1B–D), could partially be explained by a reduced uptake and a reduced mobility of the ACC dimer in planta.

To rule out that di-ACC is also able to function via an ethylene-independent ACC signaling pathway, we tested the response of Marchantia polymorpha to the ACC dimer (Figure 3). The Marchantia genome does not encode any ACOs, and therefore ACC is not used as an efficient precursor for ethylene production. However, an ACC treatment results in an ethylene-independent reduction of the thallus area (Li et al., 2020), a phenotype which is amplified in the Mpein3 mutant. Treatment of Marchantia polymorpha with di-ACC did not reduce thallus area in either background, while an ACC treatment resulted in a much smaller thallus area (Figure 3). This result indicates that di-ACC is not able to function in the ethylene-independent ACC signaling pathway in Marchantia.

Having ruled out the possibility that di-ACC associated phenotypes are caused by di-ACC breakdown into ACC during treatment (Figures 1E,F), and ethylene independent ACC signaling (Figure 3), it remains unknown how di-ACC is able to elevate ethylene biosynthesis (Figure 1G) and subsequently activate the ethylene signaling pathway (Figure 2) leading to ethylene responses in Arabidopsis (Figures 1A–C). It is possible that di-ACC is catabolized to ACC in planta (by an unknown peptidase) which can subsequently be processed to form ethylene. Alternatively, it is possible that di-ACC is used directly by ACO as a substrate for ethylene synthesis. To address this question, we measured in vitro ethylene production, using recombinant purified AtACO2. In this system, we were able to measure low levels of ethylene production when di-ACC was fed (Figure 4A), but could not detect any ethylene in control samples without any substrate (not shown). Interestingly, in vitro ethylene production levels for di-ACC were around 4 times lower compared to feeding the same concentration of ACC. This result indicates that ACO2 is able to use di-ACC directly as a substrate for ethylene synthesis, albeit with a lower efficiency compared to ACC.

To explore how di-ACC can act as a substrate for ACO, we examined the molecular docking prediction between di-ACC and ACO. The complex was modeled using the Arabidopsis ACO2 structure, and the ACC-bound complex of an ACO from Petunia hybrida (Sun et al., 2017). Superimposition of the di-ACC on the mono-ACC followed by a short local energy optimization revealed no steric hindrance, while the free carboxyl end of di-ACC is compensated by the hydrogen bonds formed by Ser163, Tyr165 and Ser249. The second cyclopropyl group sits sandwiched between Val117 and Val239. The Fe-bound water molecule required for catalytic activity is also not influenced by the di-ACC placement. The lower catalytic efficiency of di-ACC vs. ACC in our in vitro assay, could be explained by the lower affinity of the metal ion for the coordination oxygen, which has a calculated partial charge of −0.9 in the carboxyl free (ACC) compared to −0.57 in case of the amide-form (di-ACC), as calculated by the Amber10-EHT forcefield (Case et al., 2008).

This suggests that di-ACC is able to bind to the catalytic site of AtACO2, albeit with a lower efficiency than ACC. The higher concentrations of di-ACC needed to evoke a similar triple response phenotype compared to ACC, is likely due to reduced affinity of di-ACC for ACO, in combination with a reduced uptake and in planta mobility of the ACC dimer.