Glycosylphosphatidylinositol-Anchored Proteins in Arabidopsis and One of Their Common Roles in Signaling Transduction

Diverse proteins are found modified with glycosylphosphatidylinositol (GPI) at their carboxyl terminus in eukaryotes, which allows them to associate with membrane lipid bilayers and anchor on the external surface of the plasma membrane. GPI-anchored proteins (GPI-APs) play crucial roles in various processes, and more and more GPI-APs have been identified and studied. In this review, previous genomic and proteomic predictions of GPI-APs in Arabidopsis have been updated, which reveal their high abundance and complexity. From studies of individual GPI-APs in Arabidopsis, certain GPI-APs have been found associated with partner receptor-like kinases (RLKs), targeting RLKs to their subcellular localization and helping to recognize extracellular signaling polypeptide ligands. Interestingly, the association might also be involved in ligand selection. The analyses suggest that GPI-APs are essential and widely involved in signal transduction through association with RLKs.

makes this association reversible in mammalian cells (Orihashi et al., 2012;Fujihara and Ikawa, 2016). In plants, although similar shedding and release mechanisms are indicated as various GPI-APs were identified in cell walls, thus far, no GPI-specific PLC has been identified yet (Bayer et al., 2006;Yeats et al., 2018). However, a bacterial phosphatidylinositol-specific PLC (PI-PLC) has been used for shedding GPI-APs from lipid bilayers in vitro and identifying them by further proteomic analysis in Arabidopsis (Borner et al., 2003;Elortza et al., 2003;Takahashi et al., 2016;Yeats et al., 2018).

IMPORTANCE OF GPI ANCHORING FOR GPI-APS
GPI-APs and their GPI moieties were demonstrated to be crucial for diverse developmental processes in mammals and in plants, because development was found to be broadly and severely affected if GPI moiety biosynthesis is disrupted (Kawagoe et al., 1996;Gillmor et al., 2005;Kinoshita, 2014b;Bundy et al., 2016).
As the most noticeable feature, GPI anchoring was thought to be essential for the functions of GPI-APs, and their enzymatic activities or subcellular localizations could be altered by the removal of the GPI moiety (Tozeren et al., 1992;Butikofer et al., 2001;Davies et al., 2010). However, there are exceptions: the GPI anchoring of ZERZAUST and FLA4/SOS5 was shown to be dispensable for their functions in Arabidopsis (Vaddepalli et al., 2017;Xue et al., 2017).
GPI moieties also play crucial roles for driving the transient, relatively ordered membrane domains rich in sphingolipids and sterols, which are called lipid rafts or microdomains, to their target regions (Saha et al., 2016;Sezgin et al., 2017;Hellwing et al., 2018;Lebreton et al., 2018). In mammalian and yeast cells, GPI-APs are co-clustered and organized in a mixture of monomers and cholesterol-dependent nanoclusters in the same lipid raft. These exit the ER in vesicles distinct from other secretory proteins and are predominantly sorted to the apical surface to serve in protein trafficking and signaling transduction (Eisenhaber et al., 1998;Morsomme et al., 2003;Legler et al., 2005;Muniz and Zurzolo, 2014;Miyagawa-Yamaguchi et al., 2015;Sezgin et al., 2017). In Arabidopsis, although GPI modification was found essential for protein delivery from the ER to ht eplasmodesmata (Zavaliev et al., 2016), the lipid raft mechanism has not been well revealed yet.

PREDICTION AND IDENTIFICATION OF GPI-APS IN ARABIDOPSIS
To identify GPI-APs, various bioinformatics tools were developed, generally depending on the prediction of a specific The hydrophobic regions at the N and C termini are in black and the spacer region is in light gray. (C) Biosynthesis of GPI-APs. ① Precursor of GPI-AP enters ER, ② C terminus of GPI-AP precursor is recognized when entering ER, ③ Oligosaccharide structure of GPI modification is synthesized separately, ④ Recognized C terminus of GPI-AP is cleaved and covalently linked to the GPI moiety.
In Table 1, 248 genes predicted to encode GPI-APs in 2003 have been listed. Some corrections have been made, as some of them could not be found in databases or turned out to encode non-coding RNA. However, according to more recent experimental data, genes not included in 2003 also turned out to encode GPI-APs, such as At1g09460, At2g30933, At2g03505, and At4g13600 (Simpson et al., 2009), LORELEI (Tsukamoto et al., 2010), and XYP2 (Motose et al., 2004). Interestingly, due to recent achievements on alternative splicing, transcriptional variants of SKS3 (Zhou, 2019a) and CRK10 (Grojean and Downes, 2010) have been found to encode GPI-APs besides their ordinarily reported proteins (Figure 2). Alternative splicing largely enhanced the diversity of transcriptome and proteome, and more and more genes (up to 80% according to recent RNAseq achievements) have been found to be alternatively spliced in Arabidopsis, which could greatly increase the abundance of GPI-APs (Wang et al., 2009;Filichkin et al., 2010;Severing et al., 2011;Reddy et al., 2013;Lee and Rio, 2015;Bush et al., 2017).
In addition, 163 GPI-APs were predicted in 2016, and those not included in Table 1 are listed in Table 2. In this study, a large proportion of possible GPI-APs were discounted as typical GPI-APs in spite of being predicted to possess a GPI signal at the C terminus by various bioinformatics tools. Some of those discounted were transmembrane proteins, such as PIN3 and PIN4 and some receptor-like kinases (RLKs), and the other were cytoplasmic proteins without N-terminal secretory signal peptide, such as SNARE family proteins (listed at the end of Table 2).

FUNCTIONAL DIVERSITY OF GPI-APS IN ARABIDOPSIS
GPI-APs listed in Tables 1 and 2 are from diverse families, such as cell wall structure proteins, proteases, enzymes, receptor-like proteins (RLPs), lipid transfer proteins, and GPI-anchored peptides, which imply a functional diversity of GPI-APs: indeed, they were found functional in most processes, such as cell wall composition, cell wall component synthesis, cell polar expansion, stress responses, hormone signaling responses, pathogen responses, stomatal development, pollen tube elongation, and double fertilization in Arabidopsis.
Among these GPI-APs, the arabinogalactan protein (AGP) family, LORELEI family, COBRA family, and some RLPs, were better characterized. AGP family proteins are ubiquitous cell wall components anchoring on the plasma membrane throughout the Plant Kingdom and abundantly decorated at their Hyp residues by arabinogalactan polysaccharides, which make them be one of the most complex families of macromolecules in plants and play roles in various processes (Schultz et al., 2000;Ellis et al., 2010;Marzec et al., 2015;Showalter and Basu, 2016;Losada and Herrero, 2019;Palacio-Lopez et al., 2019). COBRA families were reported to be involved in various processes by regulating cell wall synthesis in plants (Hochholdinger et al., 2008;Cao et al., 2012;Niu et al., 2015;Niu et al., 2018). LORELEI family proteins associate with cell surface RLK, which is essential not only for ligand recognition but also for RLK transport (Capron et al., 2008;Duan et al., 2010;Tsukamoto et al., 2010;Meng et al., 2012;Yu et al., 2012;Liao et al., 2017;Stegmann et al., 2017;Feng et al., 2018;Guo et al., 2018;Yin et al., 2018).

INVOLVEMENT OF GPI-APS IN SIGNALING TRANSDUCTION IN ARABIDOPSIS
In Arabidopsis, hundreds of RLKs, which possess extracellular ligand recognition domains and intracellular kinase domains, control a wide range of processes, including development, disease resistance, hormone perception, and self-incompatibility (Shiu and Bleecker, 2001;Muschietti and Wengier, 2018;Wei and Li, 2018). Their association with extracellular ligands, including phytohormones, signaling polypeptides, and pathogen molecules, leads to the phosphorylation of the intracellular kinase domain, which consequently activate cytoplasmic signaling components and switch on response mechanisms ( Figure 3A) (Pearce et al., 2001;Asai et al., 2002;Geldner and Robatzek, 2008;Murphy et al., 2012;Breiden and Simon, 2016;Yamaguchi et al., 2016;Chardin et al., 2017).
By summarizing the functional mechanism of those listed GPI-APs in Tables 1 and 2, a group of GPI-APs from various families was found to share a common mechanism of action involving RLK-related signal transduction ( Table 3). The same mechanism has been reported in mammalian cells, for example, that GPI-anchored CD14 possessing leucine-rich repeats (LRR) region associates with not only Toll-like receptor TLR4 to perceive their polypeptide ligand lipopolysaccharide (LPS) leading them to activate mitogen-activated protein kinase (MAPK) cascades (Wright et al., 1990;Schumann, 1992;Zanoni et al., 2011;Li X. et al., 2015) but also TLR3 to perceive viral double-stranded RNA (dsRNA) leading them to activate (Vercammen et al., 2008). This common mechanism 1 | A review of predicted GPI-APs updated from (Borner et al., 2003;Elortza et al., 2003).

Group
Sub-group Total Gene No. Name Descriptions

AGP
Classical AGP 17 At1g68725 AGP19 AGP17-19 encode a subclass of lysine-rich AGPs, among which AGP18 was reported to be essential for the initiation of female gametogenesis both at the sporophytic and gametophytic levels, and AGP19 functions in cell division and expansion (Acosta-Garcia and Vielle-Calzada, 2004;Sun et al., 2005;Yang et al., 2011;Zhang et al., 2011a;Zhang et al., 2011b).
At5g14380 AGP6 AGP6 and AGP10 are co-expressed and co-localized in pollen grains and pollen tubes and essential for pollen grain development and pollen early germination, possibly because they are essential components of the nexine layer in pollen cell wall (Levitin et al., 2008;Coimbra et al., 2009;Coimbra et al., 2010;Costa et al., 2013;Palareti et al., 2016).
At4g09030 AGP10 At3g01700 AGP11 Its GPI modification has been experimentally confirmed, but its function has not been characterized yet (Schultz et al., 2000).
Shown as At4g40091 in Borner et al. (2003). At5g10430 AGP4/ JAGGER Essential for the degeneration of synergid cells, which guide the pollen tube attraction after acceptance of the unique pollen tube, and for prohibition of polytubey (Pereira et al., 2016a;Pereira et al., 2016b).
At5g55730 FLA1 Involved in lateral root initiation and shoot regeneration potentially by regulating cell-type specification (Johnson et al., 2011).
FLA3 Specifically expressed in pollen grains and tubes and involved in microspore development potentially through the regulation of cellulose deposition . At3g46550 FLA4/SOS5 Directly associates with cell wall RLKs FEI1/2 to perceive environmental stimuli in apoplast by altering its conformation and association with FEI1/2. This complex could regulate cell wall synthesis and composition by collaborating with CESA5. Interestingly, this regulation could also be controlled by ethylene and ABA with unclear mechanism. Surprisingly, the absence of GPI anchors only affected their PM localization but not their function (Harpaz-Saad et al., 2012;Seifert et al., 2014;Basu et al., 2016;Griffiths et al., 2016;Xue et al., 2017;Turupcu et al., 2018).

17
At5g53870 ENODL1 At4g27520 ENODL2 Catalyzes the formation of pyroglutamic acid at the N-terminus of several peptides and proteins (Schilling et al., 2007).
COBL2 Plays a role in the deposition of crystalline cellulose in secondary cell wall structures during seed coat epidermal cell differentiation, and the regulation is independent of the FEI-SOS pathway (Ben-Tov et al., 2015;Ben-Tov et al., 2018).
COBL10 Crucial for pollen tube directional growth by affecting the deposition of the apical pectin cap and cellulose microfibrils of pollen tubes and might also be involved in male-female communications .
AT3-MMP This subgroup of proteases contribute to the MAMP-triggered callose deposition in response to the bacterial flagellin peptide flg22, which suggests their involvement in the pattern-triggered immunity in interactions with necrotrophic and biotrophic pathogen .

ASSOCIATION BETWEEN GPI-AP AND RLK
Interestingly, the association between GPI-AP and RLK could be involved in not only ligand recognition but also RLK transport and subcellular localization. One of the best characterized GPI-APs, LORELEI, not only participates in ligand recognition by associating with FERONIA but also plays a crucial role in chaperoning the transport of FERONIA from the ER to the plasma membrane (Capron et al., 2008;Duan et al., 2010;Tsukamoto et al., 2010;Meng et al., 2012;Yu et al., 2012;Liao et al., 2017;Stegmann et al., 2017;Feng et al., 2018;Guo et al., 2018;Yin et al., 2018) (Figure 3C). This special chaperone and transport mechanism might be due to the GPI-APs becoming involved with lipid rafts, which determine distinct protein sorting and protein traffic (Eisenhaber et al., 1998;Legler et al., 2005;Miyagawa-Yamaguchi et al., 2015;Sezgin et al., 2017).
GPI-APs appear to be important not only for ligand recognition but also essential for ligand selection. For example, RLK FERONIA recognizes ligands RALF1 or RALF22/23 when associated with GPI-anchored LORELEI or LRX5, respectively Zhao et al., 2018). This potential GPI-AP-dependent selection mechanism could greatly enhance the ligand recognition abundance of RLK but could also mediate the cross-talk between various signaling perception and transduction ( Figure 3D).
The associations between GPI-AP and RLK could be structure independent, such as SKU5-TMK1, LRE/LLGs-FERONIA, FLA4-FEI1/FEI2, ENODL14-FERONIA, and LRX5-FERONIA, or structure dependent, such as TMM and ERECTA both possessing LRR structure at the extracellular domain and LYM1/LYM3 and CERK1 both possessing LyM structure at extracellular domain in Arabidopsis. Interestingly, the same structure dependence is also present in one of the best characterized GPI-APs in mammalian cells, CD-14, and together with its partner receptor kinases TLR3 and TLR4 all possess an LRR structure.
The structure-dependent associations between GPI-APs and RLKs largely increased the curiosities of the group of GPIanchored RLPs, which shared the same structures or sequence similarities with RLKs but lack intracellular kinase domains. They might recognize specific RLKs depending on sequence and structure similarities and form heterodimers with various Hydroxyproline-rich glycoprotein family protein At5g26290 RAMCAP Involved in both male and female gametophytic development (Singh et al., 2017). At5g26300 TRAF-like protein At3g24518** Natural antisense transcript overlaps with AT3G24520 At5g35890 β-galactosidase-related protein At1g21090 Cupredoxin superfamily protein At1g56320 At1g61900 At2g28410 At2g29660 Zinc finger (C2H2-type) family protein At3g26110 Anther-specific protein agp1-like protein At3g44100 MD-2-related lipid recognition domain-containing protein At3g58890 RNI-like superfamily protein At3g61980 KPI-1 Putative Kazal-type serine proteinase inhibitor, which is supposed to limit and control the spread of serine proteinase activity, and function during defense mechanism (Pariani et al., 2016). At4g14746 Neurogenic locus notch-like protein At4g28085 At4g38140 RING/U-box superfamily protein At5g08210** MIR834A Encoded a microRNA of unknown function At5g14190** Does not exist Shown in Borner et al. (2003) but could not be found in genomic or proteomic database actually At5g16670** Does not exist Shown in Borner et al. (2003) but could not be found in genomic or proteomic database actually At5g22430 Pollen Olee 1 allergen and extensin family protein *Shown incorrectly in Borner et al. (2003). **Shown in Borner et al. (2003) but does not exist in genomic or proteomic database or encodes non-coding RNA. ***Not shown in Borner et al. (2003) but could be predicted or studied as GPI-APs. August 2019 | Volume 10 | Article 1022 Frontiers in Plant Science | www.frontiersin.org 3 AT3g22620 AT2g13820 XYP2 Functions as a mediator of inductive cell-cell interaction in vascular development (Motose et al., 2004). AT4g22505 Bifunctional inhibitor/lipid-transfer protein/ seed storage 2S albumin superfamily protein β-1,3 Glucanases 3 AT1g11820 O-Glycosyl hydrolase family 17 protein AT4g34480 O-Glycosyl hydrolase family 17 protein AT3g57240 PLC-like phosphodiesterases 1 AT4g36945 Wall-associated kinase family protein AT4g18760 RLP51/SNC2 Functions upstream of ankyrin-repeat transmembrane protein BDA1 to regulate plant immunity through transcriptional factor WRKY70 (Zhang et al., 2010;Yang et al., 2012). Oligogalacturonide oxidase 2 AT5g66680 DGL1 Subunit of the oligosaccharyltransferase complex, which catalyzes N-glycosylation of nascent secretory polypeptides in the lumen of the ER (Lerouxel et al., 2005;Qin et al., 2013;Jeong et al., 2018). AT4g20830 ATBBE20/OGOX1 Required in plant immunity (Benedetti et al., 2018).
Whether the GPI-anchored RLKs encoded by transcriptional variants, such as GPI-CRK10 and its variant of CRK10, can form homodimers based on the same extracellular domains and play a role in RLK regulation, is a very interesting question.  (Leshem et al., 2010;Ichikawa et al., 2015;Xue et al., 2018

CONCLUSION AND PERSPECTIVES
Previous genomic and proteomic assays that predicted and identified GPI-APs from Arabidopsis have been listed. Due to recent experimental data and knowledge of alternative splicing, more and more GPI-APs have been identified, suggesting that GPI-APs in Arabidopsis might be more abundant than we expected.
Previous studies on those listed GPI-APs from diverse families were discussed, and they were found to be involved in diverse biological processes, including cell wall composition, cell wall component synthesis, cell polar expansion, hormone signaling response, stress response, pathogen response, stomata development, pollen tube elongation, and double fertilization. Those reports demonstrated the functional diversity and indispensability of GPI-APs in Arabidopsis.
Among these reports, direct associations were found between various GPI-APs and their partner cell surface RLKs, demonstrating not only participation in their ligand recognition but also essential roles in RLK transport and localization. Localization might due to specific protein sorting and protein traffic driven by GPI-AP-related lipid rafts.
Surprisingly, GPI-APs have also been shown to participate in ligand selection, which made one RLK and its downstream intracellular target activated by various ligands. Such protein cross-reactivity greatly enhanced the ligand recognition abundance of RLKs, which can also be considered as a common mechanism of cross-talk between various ligands or various signaling pathways.
In this review, the most predicted or identified GPI-APs in Arabidopsis were listed and discussed, and a common involvement of them in signing transduction was summarized. This involvement could be very helpful for understanding the ligand-RLK signaling transduction in plants, especially for understanding the polar localization of RLKs, and the crosstalk between various ligand-RLK signaling transduction. It would be interesting to identify more associations between various GPI-APs and RLKs and study their recognition and selection of ligands and downstream intracellular signaling components in Arabidopsis.