The sequenced genomes of nonflowering land plants reveal the innovative evolutionary history of peptide signaling.

An understanding of land plant evolution is a prerequisite for in-depth knowledge of plant biology. Here we extract and explore information hidden in the increasing number of sequenced plant genomes, from bryophytes to angiosperms, to elucidate a specific biological question-how peptide signaling evolved. To conquer land and cope with changing environmental conditions, plants have gone through transformations that must have required innovations in cell-to-cell communication. We discuss peptides mediating endogenous and exogenous changes by interaction with receptors activating intracellular molecular signaling. Signaling peptides were discovered in angiosperms and operate in tissues and organs such as flowers, seeds, vasculature, and 3D meristems that are not universally conserved across land plants. Nevertheless, orthologs of angiosperm peptides and receptors have been identified in nonangiosperms. These discoveries provoke questions regarding coevolution of ligands and their receptors, and whether de novo interactions in peptide signaling pathways may have contributed to generate novel traits in land plants. The answers to such questions will have profound implications for the understanding of the evolution of cell-to-cell communication and the wealth of diversified terrestrial plants. Under this perspective, we have generated, analyzed, and reviewed phylogenetic, genomic, structural, and functional data to elucidate the evolution of peptide signaling.

Introduction

Embryophytes evolved from freshwater green algae that began to colonize land approximately 470 million years ago (Mya; Delwiche and Cooper, 2015; Ishizaki, 2017). This evolutionary shift required key biological innovations, including the transition from the ancestral algal life cycle of haploid multicellular plants with zygotic meiosis, to alternating multicellular haploid (gametophyte) and diploid (sporophyte) generations. Although the diploid sporophyte and the haploid gametophyte initially might have employed the same genes for growth and development, gene duplications followed by neofunctionalization likely facilitated the development of new specialized organs by extensive changes in gene expression patterns (Ligrone et al., 2012; Bowman et al., 2017; Jill Harrison, 2017; Figure 1). The evolution of complex vasculature in ancestors of seed plants, lycophytes, and ferns, which appeared around 430 Mya (Kenrick and Crane, 1997; Gensel, 2008; Steemans et al., 2009; Morris et al., 2018), provided mechanical support and enabled improved transport of nutrients (Ishizaki, 2017). Roots evolved independently in lycophytes and euphyllophytes (vascular plants excepting lycophytes) providing anchorage for the plants and enabling acquisition of water and nutrients from the soil (Raven and Edwards, 2001; Hetherington and Dolan, 2019). The evolution of novel cell types, tissue types, and organs with increasing multicellular complexity likely required tuning of cell-to-cell communication in early land plants (Grienenberger and Fletcher, 2015).

Figure 1

Phylogenetic distribution of small PTMPs in land plants. Left: Simplified phylogenetic tree of major extant land plant lineages (green lines) based on (One Thousand Plant Transcriptomes Initiative, 2019). The branch lengths are roughly proportional to the estimated divergence dates (Morris et al., 2018). Gains of key morphological innovations are shown as black circles mapped on the tree while some gain events are still under discussion. Right: Green-colored boxes indicate the presence of homologs of the indicated small posttranslationally modified signaling peptides for the model plants Arabidopsis thaliana (Lamesch et al., 2012), Amborella trichopoda (Amborella Genome Project, 2013), Picea abies (Nystedt et al., 2013), Azolla filiculoides and Salvinia cucullata (Li et al., 2018), S. moellendorffii (Banks et al., 2011), Physcomitrium patens (Lang et al., 2018) Marchantia polymorpha (Bowman et al., 2017), and Anthoceros sp. (Li et al., 2020; Zhang et al., 2020a). Triangles indicate putative peptide ligands newly identified in our sequence analyses. Novel peptide-encoding genes were identified in tBLASTn searches against genomic DNA or RNA collections (e.g. IDA in S. moellendorffii), or BLASTp searches adapted for short amino acid queries using the PAM30 matrix (e.g. CLE genes in ferns; Altschul et al., 1997). PHI BLASTp using the aspartic acid (Asp/D) Tyrosine (Tyr/Y) motif present in RGF, CIF, PSK and PSY and/or use of wildcards in positions where the amino acid residues that are less conserved (e.g. DYxxxxxKPPIHN for RGFs), facilitated successful identification of such peptides in P. patens, M. polymorpha, ferns and gymnosperms. Genes encoding CIF and IDA peptides in the genomes of Anthoceros species were identified amongst coding sequences of less than 300 amino acids, using queries DYxxxxPxPPLxxPxPF and PIPxSxPSKRHN with wildcards based on MUSCLE alignments (Edgar, 2004) of previously identified CIF and IDA peptides, and the EMBOSS 6.5.7 tool fuzzpro (http://emboss.open-bio.org/rel/rel6/apps/fuzzpro.html).

Recent advances in whole-genome sequencing of land plants, as well as streptophyte algae, have facilitated large-scale comparative genomic studies. Such studies are helping us to understand the evolutionary processes of plant terrestrialization and diversification (Rensing et al., 2008; Banks et al., 2011; Amborella, 2013; Nystedt et al., 2013; Bowman et al., 2017; Li et al., 2018; Nishiyama et al., 2018; Cheng et al., 2019; Chen et al., 2020; Jiao et al., 2020; Li et al., 2020; Yang et al., 2020; Zhang et al., 2020a). For instance, the ancestry of some components of hormone and stress signaling pathways in land plants has been traced back to algae (Bowman et al., 2017; de Vries et al., 2018). Similarly, in flowering plants (angiosperms), the function of classical hormones as the major mediators of signaling processes is augmented by small, secreted peptides (Matsubayashi, 2014; Olsson et al., 2019). These peptide hormones enable communication between neighboring cells through interactions with leucine-rich repeat (LRR) ectodomains of plasma membrane-bound receptors that have a cytoplasmic serine/threonine kinase domain (KD). The first small signaling peptide in plants was discovered 30 years ago (Pearce et al., 1991). Since then the increase in the number and quality of sequenced genomes, and advances in methods for in silico annotation of small genes, have together allowed the identification of hundreds of potential signaling peptides (Gong et al., 2002; Ghorbani et al., 2015; Tavormina et al., 2015; Goad et al., 2017). The peptides with known receptors comprise two major groups, the cysteine-rich peptides, which attain their 3D structure by disulfide bridges formed between pairs of cysteines, and the posttranslationally modified peptides (PTMPs; Table 1; Supplemental Table S1 and S2), generated from prepropeptides and processed to mature peptides of 5–20 amino acids (Matsubayashi, 2011). The main receptors of both peptide groups are members of a large family of LRR receptor-like kinases (RLKs), which have been assigned to 15 subfamilies according to their unique structure, organization, and number of LRRs (Shiu and Bleecker, 2001; Liu et al., 2017). Here we will focus on the ancestry of the known PTMPs, which signal almost exclusively through a subfamily with more than 20 LRRs identified in Arabidopsis as LRR-RLK XI (hereafter LRXI).

Table 1

Major PTMP families in land plants

Peptide Family	Major Receptor(s)	Amino Acid Sequence of Representative Mature Peptidea,
	(LRR-RLK subfamily)	A. thaliana	M. polymorpha
CEP-C-TERMINALLY ENCODED PEPTIDEs	CEPR-CEP RECEPTOR	CEP1:
	(XI)	DFROTNPGNSOGVGH
CIF-CASPARIAN STRIP INTEGRITY FACTORs	GSO1/SGN3, GSO2, GASSHO/SCENGEN	CIF1:	Putative MpCIF:
	(XI)	DYGNNSOSORLERPPFKLIPN	DYCDCYGFDPSPPLVHAEITF
CLE-CLAVATA3/EMBRYO SURROUNDING REGION-related	CLV1, BAM-CLAVATA1, BARELY ANY MERISTEM	CLV3:	MpCLE2:
(CLV3/R-type)	(XI)	RTVOSGoDPLHH	KEVPNGPNPLHN
CLE	PXY/TDR-PXY, TDIF RECEPTOR	CLE41:	MpCLE1:
(TDIF/H-type)	(XI)	HEVOSGONPISN	HKNPAGPNPIGN
IDA + IDL-INFLORESCENCE DEFICIENT IN ABSCISSION, IDA-LIKE	HAE, HSL-HAESA, HAE-LIKE	IDA:	MpIDA1:
	(XI)	PIPPSAOSKRHN	EVPPQGPSPIHN
Pep-plant elicitor peptides	PEPR - PEP RECEPTOR	Pep1:
	(XI)	ATKVKAKQRGKEKVSSGRPGQHN
PIP + PIPL-PAMP-INDUCED SECRETED PEPTIDE, PIP-LIKE	RLK7-RECEPTOR-LIKE KINASE 7	TOLS2/PIPL3:
	(XI)	ASGOSRRGAGH
PSK-PHYTOSULFOKIN	PSKR-PSK RECEPTOR	PSK1:	Putative MpPSK:
	(X)	YIYTQ	YIYTSQ
PSY1-PLANT PEPTIDE CONTAINING SULFATED TYROSINE1	PSYR-PSY1 RECEPTOR	PSY1:
	(X)	DYGDPSANPKHDPGVoOS
RGF/GLV/CLEL -ROOT GROWTH FACTOR,GOLVEN, CLE-LIKE	RGFR/RGI-RGF RECEPTOR, ROOT GROWTH FACTOR INSENSITIVE	RGF1/GLV11/CLEL8:	Putative MpRGF:
	(XI)	DYSNPGHHPORHN	DYAEPDTHPPESN

O, hydroxylated proline; o, arabinosylated hydroxyproline; and Y, sulfated tyrosine.

LRR-RLKs have from their discovery been studied from a phylogenetic and evolutionary perspective (Shiu and Bleecker, 2001; Lehti-Shiu et al., 2009; Liu et al., 2017; Chakraborty et al., 2019; Hosseini et al., 2020; Man et al., 2020). Signaling peptides, on the other hand, were discovered mainly due to their involvement in developmental processes, such as root growth, organ abscission, or seed development in Arabidopsis, and phylogenetic investigations have focused on their distribution and function within the orders of flowering plants (Delay et al., 2013; Lori et al., 2015; Stø et al., 2015). Subsequent studies have matched LRR-RLKs with endogenous peptides as signaling pairs in angiosperms. A number of papers have reviewed specific peptide or receptor families involved in peptide signaling (Fernandez et al., 2013a; Muschietti and Wengier, 2018; Oh et al., 2018; Taleski et al., 2018; Kaufmann and Sauter, 2019; Segonzac and Monaghan, 2019; Shi et al., 2019) and (Supplemental Table S2). However, comprehensive investigations of how widely these receptors and ligands are conserved are lacking. The increase in available genome sequences of nonflowering plants makes it feasible to elucidate peptide signaling and the likely coevolution of ligands and receptors from an evolutionary perspective.

We consider that changes in peptide signaling systems might be linked to cell-to-cell communication pathways that facilitated the evolution from ancestral algae to land plants and revolutionized terrestrial ecosystems. To address this hypothesis, we started by perusing the literature and data on peptide signaling and LRXI. We identified the need to search for homologs of Arabidopsis LRXI genes in nonflowering plants as well as genes encoding specific peptides to match the presence of their presumptive receptors. Thus, our Perspective reviews published data and is also supplemented with novel findings (Supplemental Data Set S1). We offer cross-comparisons of multiple peptide ligand–receptor pairs and argue that if we are to understand the evolutionary history and innovative importance of signaling peptides in land plant evolution, we need to study cell-to-cell communication through peptide signaling pathways in an integrative manner, using diverse in silico methods that can generate hypotheses to be addressed by experimental approaches.

Evolution of subfamily XI receptors

Evolutionary constraints on LRR and KD domains

Shiu and Bleecker (2001) pioneered the characterization of Arabidopsis RLKs with ectodomains recognizing different types of ligands and highly conserved KD that have been used in previous phylogenetic analyses to study relationships among subfamilies (Zan et al., 2013; Wei et al., 2015; Liu et al., 2016; Magalhães et al., 2016; Zhou et al., 2016; Bowman et al., 2017; Lin et al., 2018; Man et al., 2020). Our focus, however, is different; namely, to better explore the evolutionary history of LRXI receptors known to interact with signaling peptides in angiosperms. Thus, for phylogenetic analysis, 27 Arabidopsis LRXI were used to sample putative LRXI homologs from 11 species including bryophytes, lycophytes, mono-, and dicotyledons, as well as Amborella trichopoda, a sister lineage to all other extant angiosperms. We also identified LRR-RLKs with more than 20 LRRs from the streptophyte algae (the grade of successive sister lineages to land plants) and used them as the outgroup to infer the origin of LRXI. After removal of duplicates and isoforms, and trimming, domains were identified in a final dataset consisting of 249 sequences (Supplemental Data Set S1), which were split on each side of the transmembrane region identified by InterProscan (Jones, 2014). Separate phylogenetic trees were then constructed for the ectodomain and KD segments of the LRXI genes (Supplemental Files S1–S5).

We noted that for most of the clades, the branches were longer in the LRR ectodomain tree compared to the KD tree. This difference suggested that the LRR evolved faster than the KD. To substantiate this, we calculated the pairwise ratio (ω) between nonsynonymous substitutions at nonsynonymous sites (K), and synonymous substitutions at synonymous sites (K), for the KDs and the LRRs encoding DNA sequences (Figure 2). The median ratio for all the angiosperm LRXI genes was less than 1, with a significantly lower value for KD (ωKD= 0.015) than for the LRR (ωLRR= 0.024, p-value < 2.2 × 10−16; Figure 2). This supports the idea that the KD is under a stronger purifying selective pressure than the LRR ectodomain. This was a common trend for the angiosperm LRXI genes, as ωKD of 15 of the 21 subclades was significantly lower than the ωLRR, according to the Wilcoxon test (significance level P < 0.05; Table 2).

Figure 2

The LRR ectodomain evolved faster than the KD. The pairwise ratio (ω) between nonsynonymous substitutions at nonsynonymous sites (Ka) and synonymous substitutions at synonymous sites (Ks) was calculated for phylogenetically identified subclades of the clades PEPR, GSO, Atan, RGI, HSL, TDR/PXY, and CLV/BAM as defined from the phylogenetic tree of the LRR ectodomain (Figure 3; Supplementary File1). All subclades show purifying selection (ω < 1). The boxplots delineate the median and the interquartile range of ω. The error bars extend from a value 1.5 times smaller to a value 1.5 times larger than the interquartile range. p-values were calculated based on a Wilcoxon test. *, Subclade with significant difference between ωLRR and ωKD (P< 0.05); n, number of sequences tested in the subclade; ns, not significant (p > 0.05). Insert: ω ratio between all pairs of sequences, both within and between subclades.Each pairwise comparison is marked with a point. Outliers have been removed from all boxplots for clarity.

Table 2

Median values of pairwise K/K (ω) for subclades of angiosperm LRR-RLKs subfamily XI genes with pvalues from Wilcoxcon rank-sum test

Subcladea	ωKD	ωLRR	P-Valueb
BAM3	0.0176	0.0523	0.066
BAM0	0.0047	0.0492	4.7 × 10−05
BAM1-BAM2	0.0053	0.0169	2.9 × 10−07
CLV1	0.0083	0.0110	0.029
AMPS-BAM-CLV	0.0029	0.0174	0.001
PXL1-PXL2	0.0068	0.0313	1.5 × 10−09
PXY	0.0207	0.0245	0.186
AMpS-PXY	0.0351	0.0173	0.589
HSL3	0.0100	0.0099	0.002
CEPR1	0.0054	0.0187	0.013
IKU2-RKL7	0.0070	0.0114	0.153
CEPR2	0.0135	0.0171	0.064
HSL1	0.0060	0.0109	0.001
HSL2	0.0298	0.0119	0.699
AMPS-HSL-CEPR	0.0090	0.0093	0.517
RGI1-RGI2	0.0106	0.0192	0.051
RGI0	0.0231	0.1142	0.7
RGI3-RGI4	0.0055	0.0111	3.4 × 10−14
RGI5	0.0166	0.0096	0.27
aMpS-RGI	0.0033	0.0346	0.004
At2g33170-At563930	0.0323	0.0386	0.027
At1g17230	0.0490	0.0409	0.672
amPS-unnamed	0.0230	0.0081	0.7
GSO1-GSO2	0.0056	0.0112	0.004
PEPR1-PEPR2	0.0531	0.0436	0.51
AMPS-GSO-PEPR	0.0066	0.0203	0.001

For convenience the three bryophytes A. agrestis, M. polymorpha and P. patens and the lycophyte S. moellendorffii are collectively named AMPS after their first letter, and a capital letter denotes presence of the relevant ortholog in the given subclade.

Significantly lower ω than ω in bold (P < 0.05); significantly higher ω than ω underlined (P < 0.05).

Figure 3

Phylogeny of the LRR ectodomains encoded by LRXI genes. LRR: Maximum likelihood phylogeny for the LRR domain encoded by 249 LRXI genes from 12 species constructed with IQ-tree (Nguyen et al., 2015) and based on a 755 amino acids long alignment, aligned with MAFFT l-INS-I and trimmed with trimal with the gappyout setting (See Supplemental File S1; Supplemental Data sets S1–S5 for details). A–G, The seven colored boxes represent monophyletic clades of the indicated angiosperm LRXIs with a common origin and at least one common bryophyte and lycophyte ancestor. These include (see blue branches) the hornwort Anthoceros agrestis (Frangedakis et al., 2021; Li et al., 2020; Zhang et al., 2020b), the liverwort Marchantia polymorpha, the moss Physcomitrium patens, and the lycophyte S. moellendorffii, collectively referred to as AMPS (the first letter for each species, capital letter when present, see also Table 2). Branches leading to A. thaliana receptors (in dark green) are labeled with the protein names. Other branches correspond to the basal flowering plant Amborella trichopoda (purple); the three monocots banana, rice, and maize (Musa acuminata, Oryza sativa, and Zea mays; yellow), and three dicots pink Shepherd's-purse, poplar, and tomato (Capsella rubella, Populus trichocarpa, and Solanum lycopersicum; cyan). LRR-RLKs with more than 20 LRRs from the recently sequenced algae Penium margaritaceum (the Penium Genome Database: http://bioinfo.bti.cornell. edu/cgi-bin/Penium/home.cgi; Jiao et al., 2020), Spirogloea muscicola, and Mesotaenium endlicherianum (Cheng et al., 2019) were used as outgroup (gray branches). Monophyletic groups of the same kind (i.e. either monocots, dicots, A. trichopoda, AMPS, or the outgroup) have been collapsed for clarity. Branches drawn with solid lines are well supported (SH-aLRT >0.95%), while dashed lines represent branches with lower support (see values for all branches in Supplemental File S1). KD: The lower part of the corresponding phylogenetic tree based on the KD illustrates ambiguities between the LRR and the KD trees and coincides with lower support (cf. eclipses; aLRT <0.85%). Branch lengths represent the average number of changes in amino acids per site. See Supplemental File S1 for the full KD tree and all branch values.

These differences in selection pressure can be explained by the differential structural and functional requirements facing the respective domains. An overall KD configuration must be preserved to secure the kinase activity by maintaining its compact structure with 13 tightly positioned conserved motifs (Liu et al., 2017). On the other hand, the ectodomain requires a structural scaffold making up nearly half of each LRR, and in addition, amino acid residues involved in peptide recognition. The land plant LRR motif consists of 24 residues with the consensus sequence of PxxLxxLxxLxxLxLxxNxLxGxI (P—Pro; L—Leu; N—Asn: G—Gly; I—Ile; x—any amino acid). The LRR units that are most important in peptide binding often deviate from the strict consensus of the scaffold. Conserved nonscaffold amino acids are on the other hand likely to represent functional sites, including residues interacting with ligands or coreceptors.

Seven ancient clades of subfamily XI receptors

The phylogenetic tree based on the LRR domain, rooted by the algal outgroup LRRs, comprises seven major clades (Figure 3, A–G in LRR; Supplemental File S1 for all support values). The clades, named after the relevant Arabidopsis receptors (PEP RECEPTORs - PEPR, GASSHO - GSO, At1g17230/At5G63930/At2G33170, ROOT GROWTH FACTOR INSENSITIVE - RGI, HAESA-LIKE - HSL/CEP RECEPTOR - CEPR, TDIF RECEPTOR - TDR, and BARLEY ANY MERISTEM - BAM/CLAVATA1 - CLV1), are defined by an early branch consisting of a combination of one or more orthologs from bryophytes, and from the lycophyte S. moellendorffii, together with a number of receptor orthologs from A. trichopoda, monocots, and dicots including A. thaliana. All seven major clades have high support (Shimodaira–Hasegawa approximate likelihood ratio test [SH-alrt] >95; Shimodaira and Hasegawa, 1999; Supplemental File S1). However, a comparison of the LRR tree with the corresponding KD tree (Supplemental File S1) revealed differences in the phylogenetic relationship between PEPR/GSO and other groups (Figure 3, A and B in LRR and KD). The phylogenetic trees could not unambiguously resolve the relationship between the hornwort Antheceros agrestis, the liverwort Marchantia polymorpha, the moss Physcomitrium patens, the vascular lycophyte Selaginella moellendorffii (for convenience collectively named AMPS based on the first letter in their names), and the angiosperm homologs as the support values for the deep nodes were low in both phylogenies, reflecting a common problem of inferring ancient relationships for LRR-RLKs (Liu et al., 2017). The PEPR (Figure 3A) and the GSO (Figure 3B) form a paraphyletic group basally to the remaining five land plant clades. This suggests that PEPR and GSO sequences are more closely related to the algal sequences (Figure 3, A and B).

One of the most plausible hypotheses of the land plant phylogeny proposes an early separation of both monophyletic bryophytes and vascular plant lineages (Figure 1; One Thousand Plant Transcriptomes Initiative 2019; Li et al., 2020; Zhang et al., 2020b). In line with this scenario, our phylogenetic analyses indicate that the 27 LRXI Arabidopsis genes originate from six or seven ancestral genes present in a common ancestor of bryophytes and vascular plants (Figure 3, Supplemental File 1and Table 3; Bowman et al., 2017; Liu et al., 2017; Li et al., 2020; Yang et al., 2020; Zhang et al., 2020a). After the two lineages split, these clades followed different trajectories. The bryophytes evolved into three distinct lineages: hornworts, mosses, and liverworts. This is reflected in their genomes, which have differentially gained and lost gene families. While M. polymorpha lack PEPRs orthologs and the clade with three A. thaliana genes with unknown function (hereafter named Atanonymous, Atan), the newly sequenced Anthoceros species also lack Atan, and in addition RGIs, while P. patens also lack RGIs and in addition TDR (Table 3). How these losses may have restricted the evolution of these three bryophytes remains to be elucidated.

In contrast, in the vascular plant lineage, S. moellendorffii has members in all seven clades, each with one to three paralogs, and altogether 14 LRXIs, indicating lineage-specific gene duplications for lycophytes (Table 3). In A. trichopoda, representing the sister lineage to all other flowering plants, the number of LRXI receptors has increased to 19, implicating duplications that may have contributed to the origin and early evolution of angiosperms. Nine of the present-day Arabidopsis LRXIs are likely results of genome duplication during the evolution of the eudicot species (Table 3; Vanneste et al., 2014). Independent gene duplications have occurred in the monocots (Figure 3).

Table 3

Number of orthologues of 27 Arabidopsis LRR-RLKs subfamily XI in A. trichopoda and AMPS.

Vascular plants			Bryophytes			Common ancestor of vascular plants and bryophytes
Angiosperm		Lycophyte	Hornwort	Moss	Liverwort
Arabidopsis thaliana a	Amborella trichopoda	Selaginella moellendorffii	Anthoceros agrestis	Physcomitrium patens	Marchantia polymorpha
PEPR1	1	2b	2b	1b	0	PEPR
PEPR2
GSO1	1				1	GSO
GSO2
At1g17230	1	1	0	2	0	At1g17230/ At2g33170/ At5g63930
At2g33170	1
At5g63930
RGI5	1	3	0	0	2	RGI
RGI4	1
RGI3
RGI2	1
RGI1
HSL2	1	3	3	2	1	HSL
HSL1	1
HAE
HSL3	1
CEPR1	1
CEPR2	1
RLK7	1
IKU2
TDR/PXY	1	2	1	0	1	PXY
PXL1	1
PXL2
BAM1	1	3	1	2	1	CLV1/BAM3
BAM2
CLV1	1
BAM3	1
Total	18	14	7	7	6	6 or 7

aBold represents novel genes in the eudicot lineage.

bUnsolved phylogeny

Success stories—the HSL clade and the BAM/CLV clade

In the vascular plant genomes analyzed in this study, none of the ancestral LRXI genes were lost, suggesting that they all are essential for the evolution of vascular plants. In contrast, each bryophyte lineage experienced losses of different ancestral genes (Table 3). Consequently, only the two related clades, HSL (Figure 3E) and CLV1/BAM3 (Figure 3G), are universally present in all lineages in our analyses, with the former showing a notable increase in gene copies in angiosperms. Gene duplications and subsequent neofunctionalization in the HSL clade may have refined cell-to-cell communication and facilitated the evolution of flowering plants. The HSL clade forms three major branches. In one branch that includes A. thaliana HAESA (HAE) and HSL2, HAE appeared during the evolution of dicots (Stø et al., 2015). These two receptors redundantly control cell separation events (Cho et al., 2008; Stenvik et al., 2008a; Shi et al., 2019). In another branch, the ancestral HSL3 gene has proliferated both before and after the rise of flowering plants, but especially in monocots (Figure 3E; Supplemental File S1). The HSL3 clade was the only subclade for which ωKD > ωLRR (p= 0.002), suggesting that with the expanded number of paralogs the selective pressure was relaxed on the KD relative to LRR domain. HSL3 was very recently shown to be involved in the regulation of stomata closure and drought stress responses by modulation of H2O2 levels in guard cells (Liu et al., 2020).

The last subclade includes CEPRs mediating nitrogen starvation signaling and controling root system architecture, (Tabata et al., 2014), and RLK7 involved in innate immunity and lateral root formation (Hou et al., 2014; Toyokura et al., 2019), and finally, there is a recent duplication in the eudicot linage leading to HAIKU2 (IKU2, Figure 3E). IKU2 is specifically expressed during seed development and controls seed size in Arabidopsis (Luo et al., 2005) and may have been recruited to regulate angiosperm- or eudicot-specific characters (Friedman and Williams, 2004).

Interestingly, a comparison of the phylogenetic trees based on the LRR and the KD shows different branching patterns for the two CEPR receptors; RLK7/IKU2 is monophyletic with CEPR1 in the KD tree but forms a clade with CEPR2 in the LRR tree (Figure 4A;Supplemental File S1). Closer inspection of the alignment of these four receptors revealed that CEPR1 has lost a repeat compared to the other three (Figure 4B). In the related HAE receptor, this repeat is involved in the interaction with the SERK1 coreceptor (Santiago et al., 2016). In the KD, RLK7 and IKU2 are more similar to CEPR1, and the CEPR2 is deviating, for instance, with a truncated kinase motif M3 (Figure 4C). Thus, RLK7 and IKU2 may be more like CEPR2 with regards to perceiving signals, but more similar to CEPR1 with regards to output from the KD. Altogether, the HSL clade suggests that gene duplications in the subclades, differential specificity in gene expression and, additionally, changes in the KD and the LRR influencing receptor output and ligand recognition, may contribute to evolutionary changes in biological function. Thus, the number and diversity of LRXI may have facilitated the rapid radiation of angiosperms.

Figure 4

Differential sequence similarities in the ectodomains and the KDs of CEPR1, CEPR2, RLK7, and IKU2. A, Schematic representation of the phylogenetic difference between the LRR (left) and the KD (right) domain of CEPR1, CEPR2, RLK7, and IKU2 based on the phylogenies in Supplemental File S1. B, Alignment of three repeats of the ectodomain of the indicated receptors from A. thaliana and Capsella rubella corresponding to the Arabidopsis HAE repeats LRR18, LRR19, and LRR20. HAE residues interacting with the coreceptor SERK1 (Santiago et al. 2016) are encircled. C, Alignment of the N-terminal part the KD of the indicated receptors from A. thaliana and C. rubella. The first four conserved motives of the KD (Liu et al. 2016) are indicated below the alignment. The deviating M3 motif of CEPR2 is indicated by a red rectangle.

Another cosmopolitan clade, BAM/CLV1, originated from an ancient duplication in the common ancestor of vascular plants and bryophytes, which produced its sister clade, TDR/PXY. Subsequently, TDR/PXY was lost in the moss lineage that includes P. patens (Figure 3F;Supplemental File S1; Table 3), illustrating differential evolutionary pressures acting on two sister clades. Within the monophyletic clade of vascular plant TDR/PXY sequences, two S. moellendorffii sequences are sister to the other angiosperm PXL1/2 genes in the LRR tree but to the PXY genes in the KD tree, which perhaps reflects stronger purifying selection on the PXL1/2 KD (Figure 3; Supplemental File S1 and Supplemental Table S2). While PXY remains as a single copy, the PXL1/2 family expanded independently in eudicots and monocots, suggestive of their potential functional diversification in these lineages.

The BAM/CLV1 gene family expanded markedly in the angiosperms with all the AMPS BAM/CLV1 sequences in the sister grade. This phylogeny based on sequence information precludes a simple deduction of the ancestral state of vascular plant BAM/CLV1 genes. Molecular genetic studies, however, started to shed light on the functional conservation between the vascular plant and bryophyte BAM/CLV1 genes. Before further exploring the evolution of BAM/CLV1 and other LRXI receptors, we now turn to their partner, PTMPs to make a case for the receptor–ligand coevolution.

PTMPs in land plants

Members of 10 families of PTMPs with characteristic length and amino acid composition have been matched with LRR-RLKs (Table 1). Each family has been previously reviewed (Supplemental Table S2). Under the assumption that LRXI receptors are activated by PTMP ligands, conservation of LRXI orthologs in bryophytes and lycophytes triggers questions regarding the presence and the roles of signaling peptides in different lineages of land plants (Bowman et al., 2017; Liu et al., 2017). We, therefore, initiated our study by reviewing the PTMPs known to be present in nonflowering plants. PTMPs are generated from prepropeptides by posttranslational processing and amino acid modifications (Matsubayashi, 2014; Olsson et al., 2019). Gene and genome duplications have led to evolutionary changes, but the amino acid sequences of C-terminal bioactive peptides have been conserved. All known PTMPs were originally identified in angiosperms. Four families, all recognized by LRXI receptors, were reported to be conserved in nonseed plants: the H- and R-type of CLE peptides, each mainly recognized by TDR/PXY and BAM/CLV1; IDA interacting with HAE and HSL2 receptors; and ROOT GROWTH FACTOR (RGF)/GOLVEN (GLV)/CLE-LIKE (CLEL) peptides recognized by RGFR/RGI) receptors. Growing availability of nonseed plant genomes, however, demands reexamination of their phylogenetic distribution.

Furthermore, when we initiated our study, the remaining peptide families of CIFs, CEPs, PIP/PIP-LIKE, Peps, phytosulfokine [PSK], and PSY1) with members interacting with GSO, CEPR, RLK7, PEPR, PSKR, and PSYR receptors, respectively (for names see Table 1), had not (yet) been identified in nonseed plants. Hence, during this study, we have seen the need of searching for “missing” peptide ligand. However, this is not a trivial task since small open reading frames often are overlooked during the annotation of newly sequenced genomes, and prepropeptide sequences are highly variable except for the short region encoding the mature peptide.

It is generally difficult to find peptide homologs from distantly related species by Basic Local Alignment Search Tool for proteins (BLASTp; Altschul et al., 1990; https://blast.ncbi.nlm.nih.gov/Blast.cgi). Consequential false-negative problems may be circumvented by using exhaustive tBLASTn searches against whole-genome sequences or RNA collections translated in all six frames (Altschul et al., 1997). Limitation of the query to the most conserved part of the prepropeptide and limitation of BLASTp targets to coding sequences of less than 300 amino acids reduce the chance of getting false positive results. The search can be optimized for short query sequences, and the query further optimized using Pattern Hit Initiated (PHI) BLASTp and/or wildcards in positions where the amino acid residues are less conserved. Further options may be to test Machine Learning or Deep Learning methods that recently have been developed for peptide discovery (Plisson et al., 2020; Serrano, 2020; Zhang et al., 2020b).

The confidence of retrieved hits is usually low, and hits must therefore in all cases be analyzed for an appropriate prepropeptide length (<200 amino acids), the position of the assumed mature peptide sequence near or at the C-terminal end and the presence of a hydrophobic N-terminal secretion signal. When interpreting the sequences, it should be kept in mind that sequences from extant species may have diversified from ancestral sequences.

Using these strategies, we have identified putative peptide gene homologs in hornworts, lycophytes, ferns, and gymnosperms (Figure 1, indicated by dark triangles, see Supplemental Data Set S1). These data are presented below together with data on the presence of LRXI.

Coevolution of peptide ligands and their receptors

PEPR and GSO and the conservation of ligand-binding residues

PEPR and GSO were the receptors most closely related to the algal receptors used as outgroup in the phylogenetic analyses (Figure 3), although their functions in Arabidopsis may be angiosperm-specific. GSO is involved in the regulation of stem cell identity, control of seedling root growth, and establishment of the Casparian strip, which works as a diffusion barrier in the root vasculature (Tsuwamoto et al., 2008; Racolta, 2014; Doblas et al., 2017; Nakayama et al., 2017; Figure 3, A and B), while A. thaliana PEPR1 is involved in amplification of biotic and abiotic stress responses and interact with endogenous stress-induced Pep peptides (Figure 1; Safaeizadeh and Boller, 2019). The phylogenetic distribution of the Pep ligands shown to interact with the AtPEPR is, however, limited to angiosperms, and even within flowering plants, Pep peptides show interfamily incompatibility. Peps of heterologous origin cannot be recognized as ligands due to the rapid coevolution of PEPR LRRs, and the PEPR1–PEPR2 subclade has a higher ω value than most of the LRXI subclades (Table 2).

To elucidate why the PEPR ectodomain seems to be so closely related to the outgroup we used Repeat Conservation mapping which is a computational method that first removes the scaffold residues from the alignment of orthologous LRRs and thereafter calculates the conservation of the remaining residues (Koller and Bent, 2014). The result is presented as a heat map reflecting the degree of amino acid identity and similarity in a given position in a given repeat and can be interpreted with the help of 3D structure models. For PEPR, this analysis confirmed that the scaffold residues as well as putative coreceptor-interacting residues are conserved, but the majority of the Arabidopsis PEPR1 residues interacting with the peptide ligand Pep according to the solved crystal structure (Tang et al., 2015), are not conserved among PEPR orthologs from the species used in our studies (Figure 5, A, C, and E). This situation has likewise been found for another LRR-RLK receptor involved in defense, FLAGELLIN-SENSITIVE2 of subfamily XII, which, like PEPR, interacts with the coreceptor BAK1/SERK3 (Koller and Bent, 2014). This is in contrast to the conserved ligand-interacting residues of receptors involved in developmental processes, exemplified here by GSO1 interacting with the confirmed ligand CIF2 (Figure 5B, D and F; Okuda et al., 2020).

Figure 5

Conservation of LRR scaffold residues and ligand-interacting amino acids. A and B, Crystal structure of PEPR1 (PDBid:5GR8) with the ligand Pep1 (Tang et al., 2015) and GSO1 (PDBid: 6S6Q) with the ligand CIF2 (Okuda et al., 2020), respectively, interacting along the inner side of the LRR structures. The conserved Pro (P) and Asn (N) residues of the scaffolds are highlighted in mustard and green colors. C and D, Alignment of the 24 amino acids long leucine-rich repeat units of PEPR1 and GSO1 with coloration based on amino acid properties according to the default Clustal X color scheme. Above a WebLogo consensus sequence (http://weblogo.berkeley.edu/) visualizing the conservation of the scaffold residues. Residues of the LRR that according to the crystal structures shown in (A) and (D) interact with residues of the respective peptides, are marked with red rectangle, and amino acids of PEPR1 interacting with the co-receptor BAK1/SERK3 are marked with orange squares. E and F, Heat maps generated using Repeat Conservation Mapping (http://www.bentlab.russell.wisc.edu/main/main.php) reflecting the degree of amino acid identify and similarity in a given position (X-axis) and a given repeat (Y-axis) for the nonscaffold residues of PEPR and GSO orthologs, respectively. The most conserved residues are in red and least conserved in blue. The position of the conserved Ps and Ns of the scaffolds are indicated by mustard and green colored vertical lines, respectively. Note that the conserved amino acid residues of PEPR orthologs (LRR 13 and onward) are not coinciding with the majority of the amino acids of PEPR1 interacting with the Pep1 ligand (LRRs 2–13), but rather with the coreceptor-interacting residues. In contrast, ligand-interacting residues overlap with a substantial number of conserved residues in GSO orthologs (LRRs 14–23).

The solved crystal structures of LRXIs with ligands show that ligands interact along the inner surface of the ectodomain (Figure 5, A and B; Chakraborty et al., 2019). Hence, the length of the entire ectodomain must fit the length of the peptide ligand. PEPRs and GSOs illustrate this: they have 26 and 31 LRRs, respectively, while most LRXI has 21–23 repeats. Correspondingly, the peptides Pep and CIF interacting with PEPR and GSO are longer (17 and 20 amino acids, respectively) than the common 12–14 amino acids of the ligands of other LRXI receptors (Okuda et al., 2020; Tang et al., 2015).

RGIs and GSO and sulfated peptide ligands

In a search for ligands, it is important to take into account the number of LRRs and the length of the peptide ligand. Another factor to keep in mind when investigating potential peptide–receptor interactions is the presence of modified amino acid residues. Both CIFs and RGFs (Fernandez et al., 2013b) are known to have Asp (Asp, D) and a sulfated tyrosine (Tyr, Y) at the N-terminus (Table 1) and may be difficult to tell apart since both also are enriched in Pro residues, and often end with a C-terminal Asn (N). They do, however, differ in length (Figure 6, A and B; Table 1). Members of the RGF family serve diverse roles in root development, such as meristem homeostasis, gravitropism, and lateral root development (Fernandez et al., 2013b). The solved crystal structure of RGF1 and its receptor RGI3 (At4g26540; Song et al., 2016), also called RGFR3 (Shinohara et al., 2016), revealed that the DYsulf residues are recognized by the amino acid motif RxGG of the receptor. Importantly, the RGI orthologs from different land plant lineages contain this motif, which distinguishes RGIs from other LRXI members (Supplemental Figure S1). RGFs have been suggested to be present in the lycophyte S. moellendorffii (Ghorbani et al., 2015), and signal through the RGI receptors (Shinohara et al., 2016; Song et al., 2016; Qian et al., 2018).

Figure 6

Conservation of sulfated peptides. A, Alignment of the CIF and TWS1-related peptides. Above the alignment shown is the sequence logo of the top seven sequences (boxed), which represent canonical CIFs. TWS1 and its putative orthologs are also boxed. B, Alignment of the presumptive RGF peptides, with amino acids colored according to chemical properties. RGFs are classified into two distinct groups (boxed), represented by the sequence logos above and below the alignment. C, Phylogenetic tree based on the ectodomain amino acid sequences of RGIs in Arabidopsis, and their orthologues in A. trichopoda, the gymnosperm P. abies, the ferns A. filiculoides and S. cucullata, the lycophyte S. moellendorffii and the bryophyte M. polymorpha constructed by the maximum likelihood method. The GSO homolog of M. polymorpha (Mp1g04830.1) from the neighboring clade was used as outgroup. Bootstrap values were obtained from 500 bootstrap replicates and indicated at the nodes as percentages. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. D, Alignment of the presumptive PSY peptides, with a sequence logo of the first 13 amino acids above the alignment. E, Alignment of the PSK1 mature peptide (YIYTQ) and adjacent conserved motif. Sequence logos were generated using WebLogo 2.8.2 (https://weblogo.berkeley.edu). Conserved amino acids are colored based on amino acid properties according to the default Clustal X color scheme. Amtr, Amborella trichopoda; Anag, Anthoceros agrestis; Anpu, Anthoceros punctatus; Arth, Arabidopsis thaliana; Azfi, Azolla filiculoides; Brdi, Brachypodium distachyon; Elgu, Elaeis guineensis, Mapo, Marchantia polymorpha; Orsa, Oryza sativa; Phpa, Physcomitrium patens; Piab, Picea abies; Sacu, Salvinia cucullata; Semo, Selaginella moellendorffii, Zema, Zea maize.

Whole-genome sequencing of nonflowering plants has facilitated the detection of GSO and RGI receptors (Bowman et al., 2017; Liu et al., 2017; Man et al., 2020), and we indeed retrieved ligand candidates from several species using Pattern-initiated BLASTp (PHI-BLASTp) with wildcards (Figures 1 and 6). The candidate CIF peptides show limited sequence identity (Figure 6A); however, this is also the case for the four Arabidopsis CIF peptides and the recently identified TWISTED SEED1 (TWS1) peptide which show differential affinities to GSO1 and GSO2 (Fiume et al., 2016; Barbosa et al., 2019; Doll et al., 2020; Okuda et al., 2020). TWS1 functions with the GSOs to form a functional cuticle around the developing embryo of Arabidopsis (Tsuwamoto et al., 2008).

To further clarify phylogenetic relationships among RGI homologs from different lineages, we bridged the evolutionary gap between Arabidopsis and S. moellendorffii by finding genes encoding highly similar receptors in the water ferns Azolla filiculoide and Salvinia cucullata and gymnosperm Picea abies (Norway spruce). In each of these species, two or three RGI genes were identified, one or two belonging to a clade that includes Arabidopsis RGI5 and the other to a branch with three A. trichopoda and four Arabidopsis RGIs genes (Figure 6B), suggesting one gene duplication in the common ancestor of ferns and gymnosperms generated the RGI1–4 and RGI5 clades. In the RGI1–4 clade, at least three gene duplications are inferred: one in the seed plant lineage, generating two subgroups, both of which experienced duplication in the eudicot lineage. A previous study also indicated that gene copies increased through lineage-specific gene duplications (Man et al., 2020).

In Arabidopsis, Tyr sulfation regulates protein–protein interactions and affects receptor binding (Matsubayashi and Sakagami, 1996; Shinohara et al., 2016; Kaufmann and Sauter, 2019). The gene encoding the responsible enzyme, TYROSYLPROTEIN SULFOTRANSFERASE (AtTPST), was discovered in Arabidopsis (Komori et al., 2009). TPST-encoding genes are conserved across land plants and found in S. moellendorffii, P. patens, and M. polymorpha, as well in ferns and gymnosperms (Supplemental Figure S2) supporting the likely presence of sulfated signaling peptides and the plausible contribution of PTMs to the ligand diversity in early land plant evolution.

PSY1 and PSK are two sulfated peptides of 18 and 5 amino acids, respectively. PSY1 has an N-terminal DYsulf, and promotes cellular proliferation and expansion (Amano et al., 2007). The mature PSK peptide is only five amino acids long (YIYIQ) and promotes cell growth, acts in the root apical meristem, contributes to pollen tube guidance, and integrates growth and defense (Sauter, 2015; Kaufmann and Sauter, 2019). We identified putative orthologs both for PSY1 and PSK in P. patens, M. polymorpha as well as P. abies (Figure 6, D and E). PSK-like genes and bioactivities have previously been detected in gymnosperms (Igasaki et al., 2003; Wu et al., 2019), but surprisingly, we did not find any ortholog of the Arabidopsis PSY1-receptor in the bryophytes. In contrast to all the subfamily XI receptors discussed here, the two PSK-RECEPTORs (PSKRs) of Arabidopsis belong to the subfamily X, of which binding to small ligands is facilitated by a so-called island domain in the LRR structure that does not fit in with the rest of the repeats. In the PSK-like sequences found in P. patens and M. polymorpha, the Asp (D) residue preceding the YIYIQ sequence is missing, but a conserved N-terminal region can still be recognized (Figure 6E). Such conserved sequences outside the assumed mature peptide might be involved in precursor processing and peptide maturation steps, which have been less characterized in plant peptidic signals.

The HSL receptors and their ligands evolved through successive changes

The angiosperm HAE-HSL receptors showcase dynamic processes of their coevolution with peptide ligands. In Arabidopsis the receptors HAE and HAE-LIKE2 (HSL2) control cell separation events through the IDA ligand which is a 12 amino acid proline-rich peptide with a concerved histidine–asparagine (His–Asn, HN) or HH at the C-terminus, residues that have been shown to interact with two closely positioned arginine (Arg, R) residues found in many LRXI receptors (Table 1; Hou et al., 2014; Vie et al., 2015; Stø et al., 2015; Song et al., 2016; Shi et al., 2019). As stated previously, the four AMPS have LRXI in the HSL clade. This and the recent identification of orthologs of IDA in M. polymorpha (Bowman et al., 2017), encouraged us to search for IDA in the other nonflowering plants. We identified potential IDA orthologs in S. moellendorffii and Anthoceros species, however, not in P. patens (Figures 1 and 7).

Figure 7

Modeling of the interaction between putative IDA ligands and HSL2 receptors of M. polymorpha and S. moellendorffii. A, The AtHAE-AtIDA crystal structure (PDBid:5ixq; Santiago et al. 2016) with the IDA peptide lined along the inner face of the LRR structure. B, Overall models of the interaction between M. polymorpha and S. moellendorffii putative HSL2 receptors (MpHSL2 in magenta and SmHSL2 in green) with, respectively, two and three M. polymorpha and S. moellendorffii superimposed putative IDA peptides (gray backbones) built on the AtHAE-AtIDA structure using SWISS-MODEL (Arnold et al., 2006). Note in particular receptor interaction with the Asn (N) at the C-terminal end of the peptides. C, Close-up view of the central parts of the respective receptor models and the surrounding hydrogen bonding network, with a central Pro (P) in the ligands. D, Close-up view as in (C), however, with hydroxylation of the central Pro (P*), which facilitates formation of additional hydrogen bonds. Central amino acids of the receptors, as well as the peptides, are shown as sticks and colored by atom type. Water molecules are shown as red spheres, and modeled based on coordinates from the AtHAE-AtIDA crystal structure. Hydrogen bonds are depicted as dotted lines (yellow). Residues involved in hydrogen bonding to the peptides are depicted with three-letter symbols in colors according to the respective structures. The peptide residues are shown in one-letter symbols. All structure figures were prepared using PyMOL (Schrödinger, LLC).

To substantiate that the MpIDA and SmIDA peptides could be the ligands of MpHSL2 and SmHSL2, respectively, we took advantage of the solved crystal structure of synthetic AtIDA peptide bound to its receptor AtHAE (Santiago et al., 2016). Using this structure, we have recently successfully generated a 3D model of IDA-HSL2 orthologs from oil palm (Shi et al., 2019)). Using the same modeling strategy for M. polymorpha and S. moellendorffii, we demonstrate that a substantial fraction of the AtHAE amino acids generating hydrogen bonds with IDA residues, are identical in MpHSL2 and SmHSL2 (Figure 7, A and B; Supplemental Figure S3). The potential overall hydrogen bonding interactions between the modeled MpHSL2 and SmHSL2 receptors and their corresponding IDA peptides (MpIDA1-2 and SmIDA1-3) support receptor–peptide binding (Figure 7, B and C). Hydroxylation of the centrically positioned Pro residues has been shown to increase the binding of IDA and HAE/HSL2 in A. thaliana (Santiago et al., 2016), and accordingly, hydroxylation of this P in the in silico model generated more hydrogen bonds between the receptor and the IDA ligand (Figure 7, C and D). Crucial residues, in particular, the central Pro and the C-terminal His–Asn, are also conserved in the amino acid sequence of the MpIDA1-2 and SmIDA peptides (Figure 7B). Thus, the generated models are consistent with a function as ligand–receptor pairs in early land plant lineages. It remains to be demonstrated whether such a pair plays a role in cell wall remodeling and/or cell separation in nonflowering plants.

Interestingly, other receptors in the HSL clade, namely, CEPRs and RLK7, do not bind IDA peptides, but instead have the closely related CEPs and PIP–PIPLs as their ligands. They are similar to IDA peptides with respect to the size and amino acid composition, possibly reflecting structural constraints to stabilize homologous receptor–ligand interactions. A notable difference is that most CEPs and PIP–PIPLs lack the C-terminal HN or HH. Instead, CEPs and PIP–PIPLs share the C-terminal GxGH motif (Vie et al., 2015; Table 1). Such deviation could be explained by recent origins of both receptors and ligands. The origin of CEPs in gymnosperms (Delay et al., 2013; Roberts et al., 2013) predates the emergence of PIP–PIPLs in angiosperms. Elucidating whether gymnosperms have CEPR orthologs and deorphanizing HSL3 and IKU2 would help us trace coevolutionary processes of receptor homologs and their potentially homologous ligands. Considering what we now know about the HSL clade, gene duplications seem to have resulted in novel functions.

CLE peptides and evolution of peptide signaling

The CLE family peptides illustrate contrasting themes. The 12–13 amino acid CLE peptides are proline (Pro, P)-rich and characteristically have the above-mentioned C-terminal HN or HH (Table 1). CLE peptides are classified into two groups by the N-terminal residue: the R (arginine, Arg) and H groups. Many R-CLEs interact with CLAVATA1 (CLV1) or BARELY ANY MERISTEM (BAM) receptors involved in meristem maintenance in Arabidopsis (DeYoung and Clark, 2008; Ogawa et al., 2008), while the three H-CLE peptides interact with PXY/TDR involved in vascular differentiation in Arabidopsis (Hirakawa et al., 2010). The CLE peptides are the first PTMPs from bryophytes that have been assigned with biological functions. These studies revealed that the CLE peptide signaling has evolutionarily conserved roles from P. patens and M. polymorpha to angiosperms (Hata and Kyozuka, 2021). One of the known key roles of the CLE signaling in angiosperms is to regulate cell proliferation in the sporophytic (2n) meristems, which grow indeterminately. In the bryophytes, the sporophyte shows determinate growth while indeterminate meristems are present in the gametophyte body (1n). In M. polymorpha an R-type peptide (MpCLE2) functions in a paracrine manner as a haploid stem cell-promoting signal in dichotomous branching of the thallus and is genetically dependent on the MpCLV receptor (Hirakawa et al., 2020). In P. patens, the R-type PpCLE peptide with its PpCLV receptor is involved 3D growth by determination of the orientation of stem cell division plane (Whitewoods et al., 2018).

To our surprise, several Arabidopsis CLE peptides have been found to interact with LRRs of different clades. The AtCLE9/CLE10 peptides interact with both the HSL1 and BAM1 receptors, depending on tissue type (Qian et al., 2018). AtCLE14 peptide signals through AtPEPR2 (Gutierrez-Alanis et al., 2017). While PEPs and CLEs differ in length, the C-terminal 12 amino acids of mature PEPs share sequence similarity to CLEs (Table 1), offering a possible explanation for the recognition mechanism. Another CLE peptide, AtCLE45, which differs from other CLE peptides by a number of arginine (Arg, R) residues near its N-terminus, is perceived by canonical CLE receptors as well as an RGF receptor homolog, AtRGI4/SKM2 (Endo et al., 2013; Hazak et al., 2017).

An MpTDR receptor, is on the other hand confined to the proliferative activity of gametophytic meristem and affects the overall size of reproductive organs (gametangiophore; Hirakawa et al., 2019). Functions of the H- and R-type CLE peptides in Anthoceros remain to be revealed (Li et al., 2020; Zhang et al., 2020a; Figure 1). The studies in the moss and liverwort support the idea that peptide–receptor modules are conserved and employed in a parallel manner, although the evolutionary relationship between the bryophyte gametophytic meristems and the vascular plant sporophytic meristems is still debated (Bowman, 2013). A key question is whether conserved mechanisms at cellular levels (e.g. downstream signaling events and target pathways) underpin functional conservation between analogous systems of distantly related species.

Toward decoding peptide signals

The “unconventional” ligand–receptor pairs discussed above raise the question of how the specificity of ligand–receptor pairs evolves. It is of note that genes encoding CLE peptides highly similar to AtCLE9/CLE10 are present in both M. polymorpha and P. patens (PpCLE5/CLE6; Whitewoods et al., 2018; Figure 8). Our phylogenetic investigations suggest that founder receptors, representing the seven clades, by the onset of the angiosperm era had increased about three times, and in each of the clades, a new gene has evolved during dicot evolution. We therefore speculate that during the early evolution of peptide signaling, there were fewer peptides, and receptors were less specific. Accordingly, large-scale clustering analyses of land plant CLE peptides based on their entire prepropeptide sequences found a smaller number of clusters in AMPS, indicative of the diversification of CLE peptide sequences, which could have resulted in changes in the specificity of ligand–receptor interactions during land plant evolution (Goad et al., 2017). Hirakawa et al. (2017) found that a synthetic CLE peptide, KIN named after the K (2nd), I (10th), and N (12th) residues crucial for activity, exerts both R-type and H-type CLE activities and interact directly with both the CLV1 and the TDR receptors (Figure 8; Hirakawa et al., 2017), illustrating the potential for a broader specificity in ligand–receptor interactions. It is of critical importance to reveal what underlies common features and family-specific characteristics in order to understand both causes (e.g. homology or convergence) and consequences (e.g. family expansion, birth of new ligands, or signaling crosstalk) of evolutionary changes in signaling peptides.

Figure 8

Conservation of CLE and IDA peptides. Alignment of mature CLE and IDA peptides in land plants and a synthetic peptide, KIN. Note shared core residues (Pro-Ser-Gly-Pro (PSGP) and C-terminal end (HN) between IDA and some of the CLE peptides, and the presence of almost identical R-CLE, H-CLE, and IDA peptides in AMPS, ferns, and Arabidopsis. Conserved amino acids are colored according to chemical properties using the default Clustal X color scheme. Amtr, Amborella trichopoda; Anag, Anthoceros agrestis; Anpu, Anthoceros punctatus; Arth, Arabidopsis thaliana; Azfi, Azolla filiculoides; Brdi, Brachypodium distachyon; Mapo, Marchantia polymorpha; Orsa, Oryza sativa; Phpa, Physcomitrium patens; PAB, Picea abies; Sacu, Salvinia cucullata; Semo, Selaginella moellendorffii.

Perspectives

Genetics and nonflowering model plants

In the 20 years since Shiu and Bleecker categorized the LRR-RLKs of Arabidopsis (Shiu and Bleecker, 2001), our knowledge of their functions and their confirmed ligands has gradually increased. Principally it is expected that disruption of the genes encoding a receptor or its ligand display the same mutant phenotype. The CLAVATA1 (CLV1) receptor and the CLV3 peptide ligand were discovered at the end of the last century by their spectacular meristem phenotypes (Clark et al., 1995, 1997). A double mutant of the closely related receptors HAE and HSL2 was needed to disclose their involvement in floral organ abscission, like their peptide ligand IDA (Butenko et al., 2003; Cho et al., 2008; Stenvik et al., 2008b). Furthermore, triple to quintuple mutants were necessary to disclose the involvement of the RGI receptors in root development, with RGFs peptides as ligands (Shinohara et al., 2016; Song et al., 2016). Thus, functional redundancy on the receptor side as well as the peptide side has made it complicated to identify ligand–receptor pairs through genetics. However, in nonseed plants that often have fewer duplicated genes, for example, M. polymorpha, and with the establishment of CRISPR-Cas mutagenesis, a genetic approach for the identification of peptide ligand receptor-pairs and their biochemical and biological functions, is becoming feasible. There is thus a need for developing more nonseed, genetically tractable model plants (Rensing, 2017). For instance, ligand(s) or function(s) have not been reported for the LRXI clade including the three Arabidopsis Atan genes (Figure 3C); studying its moss orthologs could provide first functional insight into this clade.

Families and orphans

A number of receptors and members of peptide families are orphans, in the sense that their signaling partners are unknown. This may result from the accumulation of mutations in duplicated genes that weakened ligand–receptor interaction. An alternative (and likely) explanation is that more specific genetic, molecular, and biochemical factors and conditions for peptide-mediated cell-to-cell communication remain to be discovered. In line with this suggestion, CLE9/10 of Arabidopsis bind HSL1 efficiently in the presence of the small multifaceted coreceptors SERK1, -2 or -3, but prefer BAM1 in the absence of SERK (Qian et al., 2018). To dig deeper into the evolution of peptide signaling, we recommend matching phylogenetic, functional, and structural data to reveal additional interactions with other XI receptors (Smakowska-Luzan et al., 2018) and coreceptors as key partners in receptor complexes in diverse peptide signaling pathways (Ma et al., 2016; Hohmann et al., 2017; He et al., 2018; Liang and Zhou, 2018).

Considering that a fourth of the present-day XI receptors of Arabidopsis, with seemingly angiosperm-specific functions, has a history all the way back to the ancestor of the bryophytes and vascular plants, expanding our knowledge of nonflowering plants will give a new perspective on the roles and evolution of cell-to-cell signaling. In addition to CLE peptide ligands and their receptors (BAM/CLV and TDR; Whitewoods et al., 2018; Hirakawa et al., 2019; Hirakawa et al., 2020), our approach has disclosed the early occurrence of orthologs of GSO and RGI receptors and their CIF and RGF ligands. In addition, direct interactions between HSL2 and IDA orthologs of M. polymorpha and S. moellendorffii have been substantiated, taking advantage of the solved 3D structure of the Arabidopsis peptide–receptor pair (Figure 7). We have also bridged the gap between AMPS and Arabidopsis by identifying likely orthologs of LRXI in ferns and gymnosperms (Figure 6C). These may be starting points for investigating the roles of peptide signaling and LRXI evolution prior to the emergence of angiosperms. An approach similar to what we have shown here with the integration of phylogenetic, functional, and structural information, should facilitate the disentanglement of distinct evolutionary histories of peptide ligands and their receptors. One study along these lines has recently been published on the BAM/CLV1/TDR receptors and CLE peptides (Cammarata and Scanlon, 2020).

Two ancient types of signaling PTMPs

Interestingly, the GSO/RGI clades and the HSL/TDR/BAM/CLV clades seem to have ligands representing two different groups—those with sulfated Tyr and the CLE/IDA-like peptides, respectively. Founder genes for both were possibly present in the most recent common ancestor of vascular plants and bryophytes. Alternatively, convergent evolution cannot be dismissed and might account for the highly variable central part of peptide precursor sequences. Signaling peptides can evolve not only from changes in the primary sequences, but also through changes in proteolytic processing of prepropeptides, or by PTM enzymes conferring Tyr sulfation or Pro hydroxylation. PTM processes tend to be irreversible and hence require precise regulation, pointing to the possible contribution of PTMs in the diversification of peptide ligands and in defining the specificity of ligand–receptor interaction since the early stages of land plant evolution. Our structure-based modeling of IDA-HSL2 is consistent with an ancient origin of peptide hydroxylation. Structural modeling and analyses of bioactive forms of PTMPs in AMPS will be critical for confirming predictions of molecular interactions.

New functions and new perspectives

Land plant evolution was influenced by a broadening spectrum of receptors as more complex organs evolved and interspecies interactions increased. However, except for the bryophyte CLE signaling pathways, biological roles have only been assigned for peptide LRXI signaling in angiosperms, and most often have been studied in angiosperm-specific organs. Where and when a signaling system triggers a given outcome (e.g. cell division, cuticle formation, or cell separation) may be the crux of the matter.

Not all the known signaling peptide families are conserved in nonseed plants, including bryophytes (Table 2). We know very little to nothing about peptide signaling pathways that uniquely evolved in these lineages. As such, much remains to be discovered to truly appreciate the diversity of land plant peptide signaling and to postulate testable hypotheses regarding its origin and evolution. In this Perspective, using diverse in silico methods, we broadened our knowledge on the molecular inventory of peptide-signaling components and presented hypotheses to be addressed. A very important next step in tracing the evolutionary origin and history of land plant peptide-signaling pathways will be to investigate thoroughly the temporal and spatial expression patterns of LRXI receptors and their putative peptide ligands, and to study downstream molecular events in diverse species.

Supplemental data

The following materials are available in the online version of this article.

. Conservation of ligand-binding residues in RGI receptors. Support for Figure 6.

. Phylogenetic analysis of TYROSYLPROTEIN SULFOTRANSFERASE (TPST). Support for Figure 6.

. Conservation of amino acids residues involved in ligand binding. Support for Figure 7.

. Major cysteine-rich peptide families in land plants. Support for Table 1.

. References for the signaling peptides. Support for Table 1.

. Phylogeny for subfamily XI LRR-RLKs based on the KD and the LRR ectodomain. Support for Figure 2.

. Alignment of LRRs for phylogenetic analyses.

. Alignment of KDs for phylogenetic analyses.

. Phylogenetic tree file for LRRs.

. Phylogenetic tree file for KDs.

. Names and IDs for proteins and genes presented in Figure 2 and Tables 1 and 2. Excel format.

Click here for additional data file.