Floral innovation through modifications in stem cell peptide signaling

Abstract

Understanding how evolution shapes genetic networks to create new developmental forms is a central question in biology. Flowering shoot (inflorescence) architecture varies significantly across plant families and is a key target of genetic engineering efforts in many crops^1–4. Asteraceae (sunflower family), comprising 10% of flowering plants, all have capitula, a novel inflorescence that mimics a single flower^5,6. Asteraceae capitula are highly diverse but are thought to have evolved once via unknown mechanisms^7,8. During capitulum development, shoot stem cells undergo prolonged proliferation to accommodate the formation of intersecting spirals of flowers (florets) along the disk-shaped head^9,10. Here we show that capitulum evolution paralleled decreases in CLAVATA3 (CLV3) peptide signaling, a conserved repressor of stem cell proliferation. We trace this to novel amino acid changes in the mature CLV3 peptide which decrease receptor binding and downstream transcriptional outputs. Using genetically tractable Asteraceae models, we show that reversion of CLV3 to a more active form impairs Asteraceae stem cell regulation and capitulum development. Additionally, we trace the evolution of CLV3 and its receptors across the Asterales allowing inferences on capitulum evolution within this lineage. Our findings reveal novel mechanisms driving evolutionary innovation in plant reproduction and suggest new approaches for genetic engineering in crop species.

The deep conservation of kingdom specific developmental genes in animals and plants suggests that evolution acts to fine tune developmental networks rather than replace or eliminate them, but few examples of this exist. In plants, inflorescence architecture has been highly modified through domestication in diverse crop species and several artificially selected alleles contributing to this have been identified¹¹. In contrast, little is known about the genetic mechanisms that govern major shifts in the natural evolution of inflorescence form across Angiosperms. As the largest family, Asteraceae (sunflower family) account for about 10% of all flowering plants and are defined by a novel inflorescence form called a capitulum, or head^5,12. The capitulum is composed of multiple individual florets (flowers) that show considerable diversity in morphology and organization across Asteraceae species⁷. The aggregated capitulum mimics a solitary flower (pseudanthium) that is thought to contribute to successful pollination and seed dispersal^12,13, both key traits underlying the global success of Asteraceae. As in other flowering plants, the capitulum is derived from a shoot apical meristem (SAM) that contains a pool of self-maintaining stem cells at the growing apex. In most Angiosperms studied, the size of this pool is relatively constant during inflorescence elaboration, with cells exiting the stem cell zone differentiating into floral primordia or secondary meristems, as in Arabidopsis thaliana¹⁴ (Arabidopsis; Fig. 1a). In contrast, capitulum development in the Asteraceae is characterized by a prolonged period of stem cell proliferation, followed by differentiation of centripetal spirals of florets across the expanded SAM surface, forming a pseudanthium as in Helianthus annuus L. (sunflower; Fig. 1b)^8,10. How the Asteraceae capitulum evolved is unclear, but the conservation of this trait across the family reflects it’s likely monophyletic origin. Closely related families to the Asteraceae in the MGCA (Menyanthaceae, Goodeniaceae, Calyceraceae, Asteraceae) clade, exhibit raceme or thyrse-like inflorescence development: flowering shoots with indeterminate meristems, elongated internodes, and lateral flower or branch formation (in Menyanthaceae and Goodeniaceae). In contrast, the Calyceraceae (sister to the Asteraceae) have head-like inflorescences called cephalioids, which share the compacted disc-like architecture seen in the Asteraceae but maintain terminal flower determinacy and cymose branching patterns of development¹⁵. Currently, there are two models describing the evolution of the Asteraceae capitulum. The inflorescence-origin model posits that the capitulum evolved from an ancestral thryse-like inflorescence via shoot compaction and SAM expansion, followed by a loss of apical determinacy (loss of a terminal flower)¹⁵. In contrast, the floral-unit-meristem model proposes that a capitulum meristem is determinate and similar to a floral meristem in structure, as it evolved through paedomorphosis of a floral meristem¹⁶. While the origins of the capitulum remain unclear, both models require expanded meristem stem cell proliferation as a critical step in capitulum evolution.

Fig 1: Stem cell proliferation occurs prior to floral initiation in Asteraceae capitula.

a-b, Shoot meristem dynamics during inflorescence development in Arabidopsis thaliana (raceme) and Helianthus annuus (sunflower; capitulum) with stem cell niche in blue. c, Confocal micrographs of H. annuus cv. Ha412HO shoot meristems spanning floral transition, meristem expansion and floral initiation; representative of stages used for transcriptomics. d, Volcano plots showing differentially expressed (DE) genes across pairwise comparisons of sunflower inflorescence development, total number of DE genes indicated for each comparison (green = up; red = down). e, Scaled expression of genes from clusters 1 and 2 following clustering analysis of DE genes (solid line = mean; dotted lines = standard deviation) with total number of genes in each cluster and representative CLAVATA pathway genes present indicated. f, Expression of CLV3 and WUSb in Lactuca sativa (lettuce) and sunflower meristems during inflorescence development (TPM = transcripts per million). g, Confocal micrographs of the GhCLV3pro::NLS-eGFP reporter in shoot meristems of Gerbera hybrida across inflorescence development. Upper panel shows a top-down view (xy-axis) and lower panel shows representative longitudinal sections (z-axis) of each shoot meristem with meristems outlined in white. VM = vegetative meristem, TM = transition meristem, IM = inflorescence meristem, FM = floral meristem. Scale bars - 100mm (c), 500mm (g).

To define the genetic requirements of capitulum development, we first used RNA-sequencing (RNAseq) to analyze gene expression in SAMs of cultivated sunflower (H. annuus cv. HA412-HO [PI 642777] – reference genome line¹⁷) at key developmental transitions. Stages were identified using confocal microscopy of longitudinal sections from sunflower shoot apices throughout development (Fig. 1c). Three replicates of sunflower SAMs were sequenced spanning vegetative growth (vegetative meristem – VM), at floral transition (transition meristem – TM), during meristem expansion (inflorescence meristem – IM), and early floret initiation (inflorescence and floral meristems – IM/FM). Differential expression analysis identified 6,657 (VM x TM), 8,265 (TM x IM), and 2,095 (IM x IM/FM) significantly differentially expressed genes (DEGs) during capitulum development (Fig. 1D; Supplementary File 1). We used a clustering analysis to investigate generalized patterns of gene co-expression across the developmental series, using the 12,657 DEGs identified from pairwise comparisons (Supplementary Fig. 1; Supplementary File 1). Thirteen clusters were identified using a hierarchical clustering-derived cutoff to define groups with similar expression patterns (Supplementary Fig. 1b). Clusters 1 (650 genes) and 2 (2,919 genes) exhibited increased gene expression during capitulum meristem expansion and early floral initiation phases (Fig. 1e). Consistent with these developmental stages, both clusters contained orthologs of key meristem regulatory genes including members of the highly conserved CLAVATA (CLV) signaling pathway (Supplementary File 1).

In Angiosperms, shoot stem cell proliferation is repressed by CLAVATA3 peptide signaling (CLV3p), the founding member of the CLAVATA3/EMBRYO-SURROUNDING REGION peptide (CLEp) hormone family^18–20. Mutations in CLV3 lead to uncontrolled stem cell proliferation, enlarged shoot size (fasciation), and increases in floral organ numbers. The mature processed 12 amino acid CLV3p is derived from the CLE domain of the CLV3 apoprotein, secreted from shoot stem cells, and diffuses into the meristem center where it signals through the CLAVATA1 (CLV1) leucine rich repeat (LRR) receptor kinase, and accessory receptors, to repress expression of the stem cell promoting transcription factor WUSCHEL (WUS)^14,21. We leveraged recently published genomic resources and generated new de novo transcriptome assemblies of SAMs in six species from Asteraceae and immediate outgroup taxa to identify CLV pathway orthologs across the Asterales (Moore-Pollard et al 2025). Ortholog analyses consisted of data from major Angiosperm clades (ANA: 2 species, Magnoliid: 3 sp., Monocots: 12 sp., and Eudicots: 39 sp.), including a diverse set of Asteraceae spanning 5 subfamilies and 13 tribes (Supplementary File 2; Supplementary Fig. 2). Through these analyses, we identified Asteraceae-specific duplications of CLV1, the related CLEp receptors BARELY ANY MERISTEM 1/2 (BAM1/2), and WUS (Supplementary Fig. 3). We then used two previously reported Hidden Markov Model (HMM) profiles of CLE motifs to identify and classify Asteraceae CLEs (Supplementary File 3)^22,23. Clustering analysis based on sequence similarity revealed only one CLE1D member (from the 1D1 subclade) in all Asteraceae species examined (16 total), indicating that a single CLV3 ortholog is present in all Asters. This ortholog groups with genetically defined CLV3 genes from Arabidopsis and tomato (Fig. 2a; Supplementary Fig. 4)^22,24. We further validated both CLV3 and CLV1 ortholog identification using a directed BLAST (tblastn) to compare sunflower and Lactuca sativa L. (lettuce; Asteraceae) orthologs against eight genomes (four Asteraceae and four from closely related outgroups) (Moore-Pollard et al 2025). We found that both sunflower and lettuce CLV3 return only one high similarity hit in all genomes surveyed and this analysis also supported the existence of the Asteraceae-specific CLV1 duplication (Supplementary Files 4–5).

Fig 2: Reduced signaling activity of Asteraceae CLV3.

a, Clans-based clustering of angiosperm CLE1D-type CLEs from across flowering plant lineages with genes from A. thaliana (At), Solanum lycopersicum (Sl), Oryza sativa (FCP1 and FON2), sunflower and lettuce labeled. b, LOGO consensus diagrams of mature CLE domains from all CLE1D1-type CLEs showing conservation of CLE motif in all Asteraceae CLV3 orthologs analyzed. Residues 2, 3 and 5 labeled at bottom. c-f, CLE peptide root elongation assays showing responses to 0.1μM AtCLV3p, PgCLV3p (Platycodon grandiflorus – Campanulaceae) and AsterCLV3p in A. thaliana (c-d) and lettuce (e-f) each compared to no peptide (NP) controls. Statistical summary from one-way ANOVA where A and B groups are significantly different (adj. p-value <0.0001). g-h, T1 complementation analysis of carpel number in clv3-9 using a native AtCLV3 promoter::gene (AtCLV3pro::AtCLV3) constructs with AtCLV3, PgCLV3 and AsterCLV3 CLE domains swapped in (related to Supplementary Fig. 8d). In d, n= 13 (NP), 20 (AtCLV3p), 21 (PgCLV3p) and 21 (AsterCLV3p). In f, n= 17 (NP), 18 (AtCLV3p), 20 (PgCLV3p) and 24 (AsterCLV3p).

In all angiosperms studied to date, elevated CLV3 expression represses WUS, and thereby stem cell proliferation, in a CLV1-dependent manner^14,19,21. In contrast, sunflower orthologs of CLV3 (HaCLV3), CLV1 (HaCLV1b), WUS (HaWUSb), and BAM1/2 (HaBAM1/2a) were all present in SAM expression clusters 1 and 2, displaying coincident and increased expression during capitulum meristem expansion (Fig. 1e). We next explored the conservation of CLAVATA pathway regulation during capitulum meristem expansion using a publicly available RNAseq dataset from similar stages of lettuce SAMs (Fig. 1f)²⁵. Again, CLV3 and WUS (WUSb) expression were coincident and peaked in inflorescence meristems during the expansion phase in lettuce (Fig. 1f). In Arabidopsis and tomato, CLV3 is expressed in central zone stem cells (top 1-3 cell layers of the SAM; L1-L3), with WUS and CLV1 expression being more prominent in the subtending organizing center (L3-L5 layers)^14,21. In Arabidopsis clv1 mutants, CLV3 expression shifts laterally, marking expanding stem cell pools, while WUS expression shifts apically into upper layers of the SAM²⁶. To further confirm Asteraceae CLV3 orthology and determine if alterations in the expression zones of CLV/WUS genes underlie capitulum formation, we examined the expression of CLV3, CLV1a/b and WUS genes in select Asteraceae models using in situ hybridization across meristem stages and types. Asteraceae CLV3 is expressed in the stem cell regions of the VM, TM, IM, and IM/FM in representative species from diverged Asteraceae subfamilies; sunflower (HaCLV3, Asteroideae), lettuce (LsCLV3, Cichorioideae), and Gerbera hybrida (Gerbera, GhCLV3, Mutisioideae; Supplementary Figs. 5–7). There is a lateral expansion of CLV3 across the meristem during capitulum expansion in all three species, when stem cell proliferation is maximal (Supplementary Figs. 5–7)¹⁰. The lateral expansion of CLV3 domain was more conspicuous in Gerbera and sunflower, compared to lettuce, tracking with the longer duration of the capitulum expansion phase in these species (and relative final size of their capitula). We further verified this spatial patterning of CLV3 in a developing capitulum using a native GhCLV3pro reporter (GhCLV3pro::NLS-eGFP) line in Gerbera. Clear lateral expansion across the inflorescence meristem was observed at each sequential stage of development (VM, IM, and IM/FM), along with a dip in expression in an adjacent zone of initiating floret primordia followed by re-emergence of GhCLV3 expression in already established floral meristems in the periphery of the IM (Fig. 1g). CLV3 expression was restricted to the upper layers (L1-L3) of meristems in all three species (Fig. 1g; Supplementary Figs. 5–7). Additionally, there was an apical shift in CLV1a/b expression into the L2 and L3 layers compared to the more central expression pattern seen in Arabidopsis. This was also paralleled by an apical expression pattern for WUSb in the upper layers across the meristem in sunflower and Gerbera (Supplementary Figs. 5; 7). As such, CLV3 and WUSb expression patterns and dynamics during Asteraceae capitulum elaboration mimic the behavior of these genes in clv1 mutants in other Angiosperms, suggesting a possible reduction in CLV1-signaling in Asteraceae.

The duration of stem cell proliferation during capitulum development is prolonged relative to other species but not unrestricted as in clv mutants, suggesting a quantitative change in stem cell regulation occurred during Asteraceae evolution. Given the importance of CLV3p in stem cell repression, and the apparent differences in CLAVATA expression and regulatory dynamics in the Asteraceae, we investigated whether capitulum development and evolution could be associated with reductions in CLV3p signaling. To explore if this is due to differences in CLV3p itself, we first compared the CLE motif of Asteraceae and non-Asteraceae CLV3 homologs (CLE1D1-type CLEs), as this represents the mature CLV3p (Fig. 2b). Surprisingly, this revealed that CLV3p is invariant across all 16 Asteraceae species examined. Additionally, the N-terminus of AsterCLV3p diverged significantly from non-Asteraceae CLV3p and appeared unique to Asteraceae (Fig. 2b). In non-Asteraceae, position 2 is variable and there is a bias towards valine and serine at positions 3 and 5, respectively (for CLE1D1s). In contrast these residues in AsterCLV3p were invariably Ala2, Ala3, and Leu5, forming an N-terminal CLEp motif we term the RAAPL CLV3p (Fig. 2b). To test if the AsterCLV3p displayed altered activity, we first performed root growth inhibition assays, a standard measure for CLEp activity²⁷, in response to CLV3p peptides from Arabidopsis AtCLV3p (RTVPS), AsterCLV3p (RAAPL), and from Platycodon grandiflorus PgCLV3p (RKVPS, Campanulaceae), an outgroup to the MGCA clade. While AtCLV3p and PgCLV3p repressed root growth in Arabidopsis seedlings at low peptide concentrations, AsterCLV3p failed to do so (Fig. 2c–d). This was mirrored in lettuce and sunflower seedling root assays, revealing that Asteraceae can respond to CLV3p from other species but fail to respond to their own in this assay (Fig. 2e–f; Supplementary Fig. 8a). We next tested the ability of AsterCLV3p to promote CLV1-dependent stem cell repression in shoots, first using Arabidopsis as a proxy model. Consistent with the root growth inhibition assays, expression of either AtCLV3 (RTVPS) or full-length AtCLV3 containing PgCLV3 CLEp motif substitutions (RKVPS), under the native AtCLV3 promoter (AtCLV3pro:CLV3), fully complemented Arabidopsis clv3 (clv3-9) carpel number increases, while the AsterCLV3 CLEp motif (RAAPL) substitutions failed to do so, indicating that AsterCLV3p also has reduced CLV1-dependent stem cell–restricting activity in Arabidopsis shoots (Fig. 2g–h; Supplementary Fig. 8d–e).

In Arabidopsis, CLV1 and its close homologs BAM1/2/3 mediate developmental responses to diverse CLE peptides beyond the shoot apical meristem, whereas PXY (PHLOEM INTERCALATED WITH XYLEM) receptors specifically recognize TDIF (TRACHEARY ELEMENT DIFFERENTIATION INHIBITORY FACTOR) class CLE peptides (CLE41p, CLE42p, CLE44p) to regulate vascular development^28,29. CLE peptide signaling requires co-receptors^30–32; however, direct and biologically relevant CLEp binding has been only shown for CLV1, BAM, and PXY receptors, with specificity and affinity varying across receptor–ligand pairs^33–38. Although all belong to the same LRR-RLK subfamily, structural predictions and phylogenetic analyses reveal distinct architectures of the ligand-binding pocket between CLV1/BAM and PXY receptors, consistent with their separation into different evolutionary subclades (Supplementary Fig. 9)³⁹. Despite this, genetic, biochemical, and structural studies—based on the PXY-TDIF complex—have identified a conserved mechanism of CLE peptide perception across receptor and peptide subclades^35,38. CLE peptide recognition is primarily mediated by the N-terminal region, which docks the peptide into the receptor pocket through hydrophobic and electrostatic interactions, while the conserved central and C-terminal residues stabilize the complex by facilitating a conformation compatible with the receptor binding groove^35,38. Although no structures of CLV1/CLV3p complexes have been reported in any species, structural modeling and functional analysis supports a similar binding mode for AtCLV3p and AtCLV1, with essential N-terminal and C-terminal residues (Leu10, His11, His12) being required for AtCLV3 function in vivo⁴⁰. In AsterCLV3p, conservation of the key core and C-terminal residues indicates the potential for interaction with CLV1-class receptors (Fig. 2b). Consistent with this, higher doses of AsterCLV3p modestly inhibited Arabidopsis root elongation, suggesting that AsterCLV3 has strongly attenuated, but not eliminated, activity (Supplementary Fig. 8b–c). AsterCLV3p differs from AtCLV3p, and other CLV3p orthologs, at N-terminal positions 2, 3, and 5 (Ala2, Ala3, and Leu5 in AsterCLV3p). Alanine substitution experiments of these positions individually in AtCLV3 previously suggested they play negligible roles in CLV1-dependent stem cell regulation⁴⁰. This implies that the reduced activity of AsterCLV3p arises from the unique combination of these substitutions. Indeed, the reduction in AtCLV3p activity in sunflower, lettuce and Arabidopsis root length peptide sensitivity assays was strongest when Ala2, Ala3, and Leu5 substitutions were combined (Fig. 3a–c). Moreover, these substitutions were additive in attenuating CLV1-dependent stem cell repression in Arabidopsis shoots, as measured by carpel numbers, with AtCLV3^Ser5Leu variants having a negligible effect on activity on their own but synergistic effects with the AtCLV3^Thr2Ala, ^Val3Ala substitutions (thereby recreating the AsterCLV3p RAAPL sequence; Fig. 3c). Despite the failure of AsterCLV3 to complement clv3-9 carpel phenotypes, we observed partial complementation of clv3-9 shoot fasciation and increased rosette leaf numbers, again suggesting that AsterCLV3p signaling is not entirely eliminated (Supplementary Fig. 8f–h). As such, the strongly attenuated activity of AsterCLV3p is due to a unique combination of substitutions outside of the essential conserved CLE peptide residues.

Fig 3: Contributions of N-terminal residues to AsterCLV3 signaling.

a-c, CLE peptide root elongation assays showing responses to 0.1μM AtCLV3p, AsterCLV3p and intermediate CLV3p variants in sunflower (a), lettuce (b) and A. thaliana (c) each compared to no peptide (NP) controls. Statistical summary from one-way ANOVA where A-E groups are significantly different (adj. p-value <0.01). c, T1 complementation analysis of carpel number in clv3-9 using a native AtCLV3 promoter::gene (AtCLV3pro::AtCLV3) constructs with AtCLV3 (RTVPS), RTVPL, RAVPL, RAAPS, RTAPL and AsterCLV3 (RAAPL) CLE domains swapped in. d, CLE peptide root elongation assays showing responses to 0.1μM AtCLEp (with 5^th residue Ser) and each AtCLEp S5L variant. Statistics from one-way ANOVA with significance indicated by asterisks. (adj. p-values: * = <0.05, ** = <0.005,**** = <0.0001). In b, n= 30 (NP), 31 (AtCLV3p), 20 (RTVPL), 18 (RAVPL), 38 (RAAPS), 13 (RTAPL), and 32 (AsterCLV3p). In c, n= 21 (NP), 21 (AtCLV3p), 10 (RTVPL), 10 (RAVPL), 10 (RAAPS), 11 (RTAPL), and 20 (AsterCLV3p). In d, n= 252 (NP), 127 and 115 (AtCLV3p and AtCLV3p S5L), 88 and 91 (AtCLE9p and AtCLE9p S5L), 118 and 120 (AtCLE11p and AtCLE11p S5L), 121 and 122 (AtCLE12p and AtCLE12p S5L), 118 and 116 (AtCLE16p and AtCLE16p S5L), 117 and 124 (AtCLE19p and AtCLE19p S5L), 122 and 119 (AtCLE27p and AtCLE27p S5L), 116 and 114 (AtCLE40p and AtCLE40p S5L), 76 and 75 (AtCLE43p and AtCLE43p S5L).

We next inspected Asteraceae CLV1a/b orthologs for potential changes in CLEp binding residues. The N-terminal hydrophobic CLEp binding pocket in CLV1/BAM/PXY receptors is shaped in part by a conserved glycine in LRR6 which accommodates Val3 in CLE peptides (Supplementary Fig. 10a–d)³⁸. This glycine is essential across CLEp receptor classes and is mutated in the strong EMS-derived clv1-4 allele in Arabidopsis (AtCLV1^G201E)⁴¹. CLEp substitutions at position 3 with bulky residues would create steric clashes with the receptor LRR6 glycine and other neighboring residues, while CLEp Val3Ala substitutions, as in AsterCLV3p, are tolerated (Supplementary Fig. 10)^34,38,42. Inspection of CLV1a orthologs revealed that this critical LRR6 glycine is substituted by alanine in the majority of Asteraceae analyzed and is glutamine in lettuce (LsCLV1a^Q209) and other species from its tribe (5 species in the Cichorieae; Supplementary Fig. 11a). For CLV1b orthologs, only 4 out of 34 Asteraceae examined had CLV1b LRR6 glycine substitutions, and those were all from species that had additional CLV1b paralogs which retain the LRR6 glycine. All CLV1a/b orthologs retain other conserved critical residues implicated in CLEp binding (Supplementary Fig. 11a). Only three Asteraceae species analyzed possessed CLV1a/b orthologs where both retained the canonical LRR6 glycine: sunflower (Heliantheae), Smallanthus sonchifolius (Millerieae), and Mikania micrantha (Eupatorieae); all closely related tribes from the Heliantheae alliance. Structure prediction with AlphaFold⁴³ revealed that replacement of glycine with alanine at LRR6 reduces the volume of the hydrophobic peptide-binding pocket, potentially weakening binding, while glutamine substitution would introduce steric clashes that are predicted to block CLEp binding, consistent with the defects observed in Atclv1-4 mutants (Supplemental Fig. 10). Supporting these predictions, native promoter expression of AtCLV1^G201A (AtCLV1pro:CLV^G201A) and AtCLV1^G201Q substitutions failed to complement Arabidopsis clv1-101 null mutants, with AtCLV1^G201A lines displaying milder phenotypes than AtCLV1^G201Q lines. (Supplementary Fig. 11b, Supplementary Fig. 10). Additionally, both AtCLV1^G201A and AtCLV1^G201Q variably enhanced carpel numbers in the clv1-101 background, mirroring the semi-dominant nature of the clv1-4^G201E allele (Supplementary Fig. 11b)¹⁸. To test the conservation of this effect, we generated the same amino acid swaps in the LRR6 glycine of the related AtBAM1 receptor. Native promoter AtBAM1^G199Q (AtBAM1pro:AtBAM1^G199Q) failed to complement the bam1/2 double mutant male sterility and rosette phenotypes⁴⁴ (7/7 lines examined), while AtBAM1^G199A complemented both phenotypes, suggesting AtBAM1^G199A retains affinity for CLE peptides (Supplementary Fig. 11c–d). As such, along with the evolution of attenuated AsterCLV3p activity, diverse Asteraceae evolved CLV1 paralogs with apparent reductions in CLV3p binding potential, while still retaining at least one copy of CLV1 with predicted CLV3p binding capabilities.

We next investigated whether the reduced activity of AsterCLV3p reflected altered ligand affinity for CLV1/BAM receptors in vitro. Despite several attempts and diverse methods, we were unable to purify enough folded AtCLV1 or LsCLV1b proteins. We therefore used AtBAM1 as a proxy receptor for CLEp binding in Isothermal Titration Calorimetry (ITC) assays, due to the conservation of its CLEp-binding pocket with CLV1-type receptors, established CLEp partners, responsiveness to AtCLV3p in vivo, and validated AtCLV3p binding in vitro (Fig. 4a–c, Supplemental Fig. 12)^35,38,45. ITC assays were performed to compare binding of AtBAM1 to AtCLV3p, AsterCLV3p, and the high-affinity BAM1 ligand AtCLE13p⁴⁵. The latter was selected based on its ability to promote BAM1-dependent formative cell divisions in Arabidopsis roots, providing a biologically relevant readout for binding activity (Fig. 4d)⁴⁵. Although AtCLV3p bound BAM1 with ~1000-fold lower affinity than AtCLE13p, it retained robust activity in promoting BAM1-dependent root cell divisions. In contrast, AsterCLV3p showed no detectable binding to BAM1 and failed to induce root cell divisions (Fig. 4c–d, Supplemental Fig. 13).

Fig. 4. The N-terminal region in CLE peptides impacts CLV1/BAM receptor recognition and specificity.

a, Surface representation of CLV1/BAM receptor ectodomains, colored based on sequence conservation across A. thaliana, sunflower and lettuce (related to Supplementary Fig. 12d). b, Magnified view of the AtBAM1 binding pocket (surface representation colored according to conservation) in complex with CLE13 (pink). c, ITC summary table of CLE/CLV peptides binding to AtBAM1 ectodomains. K_d (nM) represents binding affinity, ΔH (kcal/mol) indicates reaction enthalpy, and N denotes stoichiometry (n= 1 for a 1:1 interaction). Data are mean ± SD from at least two independent experiments, with undetected binding shown as n.d. d, Root meristem micrographs (A. thaliana) of an endodermal marker (green; EN7) with no peptide, or grown in the presence of AtCLV3p, AtCLV3p S5L, AsterCLV3p, AsterCLV3p L5S, AtCLE13p, AtCLE13p S5L. Variant residue labeled in red. e-f, CLE13 residues at positions 3 and 5 fit into subpockets on the receptor’s surface, with the Ser5 side chain positioned towards the receptor. Due to the proximity (5.8 Å), Ser5 may facilitate water-mediated interactions between the peptide and receptor. The local environment surrounding CLE13 Ser5 supports a network of potential water-coordinated interactions. f, Close-up view of a computationally predicted water molecule bridging interactions between CLE13 Ser5 and the receptor binding pocket. The models are based on the AlphaFold3 prediction of AtBAM1 in complex with AtCLE13. Scale bar - 25μm (d).

Among the residues that diverge from AtCLV3p, position 5 has not previously been implicated in receptor binding or activity. In our refined BAM1-CLE13p structural model, the hydroxyl group of CLE13p Ser5 is oriented towards the receptor interface, where it may engage in water-mediated interactions that stabilize the peptide-receptor complex (Fig. 4e–f, Supplementary Fig. 10e; Supplementary Fig. 12e). Supporting this model, substitution of CLE13p Ser5 with Ala or Leu—as found in AsterCLV3p—markedly reduced AtCLE13p binding affinity to BAM1 and CLE13p^Ser5Leu and AtCLV3p^Ser5Leu both failed to promote BAM1-dependent root cell divisions (Fig. 4c–d; Supplementary Fig. 13). Notably, in the PXY/TDR-CLE41p complex, the equivalent Ser5 side-chain is solvent-exposed and does not contribute to the receptor-peptide binding, suggesting a clade-specific role³⁸. Consistent with this, Leu5 substitutions impaired activity in 8 of 9 tested Arabidopsis CLE peptides in root growth inhibition assays (Fig. 3d), supporting a general requirement for position 5 in effective peptide–receptor interactions. Replacing Leu5 with Ser in AsterCLV3p restored partial BAM1 binding, but to a level considerably lower than AtCLV3p or AtCLE13p, again suggesting that additional residues—particularly at positions 2 and 3—contribute to full receptor engagement (Fig. 3a–c. 4c, and Supplementary Fig. 13). Consistent with this AsterCLV3p^Leu5Ser displayed variable activity across bioassays, with significant root growth inhibition activity in sunflower, lettuce, and Arabidopsis (Fig. 3a–c, 4c, Supplementary Fig.13), a partial ability to complement clv3 carpel numbers in Arabidopsis (Fig. 3d), and no ability to promote BAM1-dependent root cell divisions. Together, these results identify position 5 as a context-dependent regulatory residue that modulates CLEp activity and receptor specificity across CLE/TDIF receptor clades. Consistent with our structural model, we observed no binding of any peptide to AtBAM1^G199Q, while AtBAM1^G199A exhibited reduced binding affinities for most CLE peptides and variants tested, mirroring the predicted role of CLEp position 3 in BAM1 LRR6 G199 interactions. Notably, neither AtBAM1^G199A nor AtBAM1^G199Q restored binding to AsterCLV3p (RAAPL), indicating that LRR6 glycine substitutions in Asteraceae species cannot compensate for AsterCLV3p-specific changes (Fig. 4a and Supplementary Figs 10 and 13). Collectively, these findings indicate that the unique and conserved substitutions in AsterCLV3p impair its signaling capacity by reducing receptor binding affinity.

While we were unable to detect AsterCLV3p binding in our ITC assays, several observations suggest AsterCLV3p retains low levels of signaling capacity in vivo. First, AsterCLV3p can restrict Arabidopsis root growth at micromolar concentrations, but not at nanomolar concentrations like AtCLV3p or PgCLV3p (Supplementary Fig. 8b–c). Second, while AsterCLV3 failed to complement Arabidopsis clv3 carpel number defects, it did partially suppress other clv3 phenotypes including increased leaf numbers and stem fasciation (Supplementary Fig. 8e–h). Third, stem cell proliferation is prolonged during capitulum development, but not unrestricted as in clv3 mutants in other Angiosperms¹⁹. We therefore hypothesized that Asteraceae may have evolved a CLV3p with attenuated but not eliminated signaling capacity in order to permit prolonged stem cell division during capitulum development but still buffer against run-away stem cell proliferation. If so, this would also explain the invariance of the AsterCLV3p across the family and the retention of essential AsterCLV1b CLEp binding residues. To test this and determine the function of AsterCLV3 in vivo, we generated CRISPR induced LsCLV3 knockout mutants in lettuce (lsclv3-cr) and GhCLV3 RNAi knockdown lines in Gerbera. lsclv3-cr mutants displayed enhanced capitulum and capitulescence (capitula bearing shoot axis) meristem size, stem fasciation and increased floral organ numbers, which we quantified in both petals and stigmata (a proxy for carpel number; Fig. 5a–b; e–f and Supplementary Fig. 14a, c–e). lsclv3-cr also developed clusters of fused florets in the capitulum center, which we called terminal florets, a phenotype of unknown origin not seen in clv mutants in other species (Fig. 5d; Supplementary Fig. 14f). Similar phenotypes were seen in Gerbera GhCLV3 RNAi knockdown lines, disrupting inflorescence architecture along with increases in meristem size and floral organ numbers (Fig. 5c–d; Supplementary Fig. 14b). Together, these data demonstrate that the canonical role for AsterCLV3p in repressing stem cell proliferation is conserved in Asteraceae.

Fig. 5: AsterCLV3 has attenuated stem cell restrictive function.

a-b, Capitula and scanning electron micrographs (SEM) of developing florets in WT and Lsclv3-cr. c-d, WT and GhCLV3 RNAi knockdown capitula and SEM images of developing florets. e-f, petal and stigma counts of WT vs Lsclv3-cr florets. Statistics from Mann-Whitney U test (adj. p-values <0.0001). g-h, Complementation of increased stigma number phenotype comparing LsCLV3pro::LsCLV3 and LsCLV3pro::AtCLV3 expressed in Lsclv3-cr. i, Floret counts (per capitulum) comparing LsCLV3pro::LsCLV3 and LsCLV3pro::AtCLV3 expressed in Lsclv3-cr. Statistical summaries from Kruskal-Wallis and Dunn’s multiple comparison tests where A-D groups are significantly different (adj. p-value <0.05). Related to Supplementary Fig. 14g–i. j-k, differential gene expression of sunflower inflorescence meristem cores 5 and 24 hours after treatment with CLV3 and CLV3-variant peptides vs no peptide mock treatment. j, total differentially expressed genes. k, log₂ fold change of WUSb expression after peptide treatment. Blue represents amino acids encoded by AtCLV3 while red represents LsCLV3. Scale bar - 100μm (a-b), 1cm (c), and 1mm (d, f).

We next asked if the reduction in signaling capacity of AsterCLV3p relative to other species is important for shoot stem cell homeostasis in the Asteraceae. Using a synthesized LsCLV3 native promoter construct (~3kb upstream of and ~1kb downstream of LsCLV3), we attempted to complement lsclv3-cr with LsCLV3pro:LsCLV3 (RAAPL), LsCLV3pro:LsCLV3 containing the AtCLV3 CLEp motif (RTVPS) or containing the Scaevola aemula CLV3 (SaCLV3 – Goodeniaceae) CLEp motif (RTAPL, differing only from Asteraceae at residue 2). We quantified complementation using stigma number and number of florets produced per capitulum across three (LsCLV3) or four (AtCLV3 and SaCLV3) independent stable transgenic lines. Both LsCLV3 and SaCLV3 complemented stigma and floret number at similar levels (Fig. 5g–i; Supplementary Fig. 14g–h). In contrast, lettuce expressing AtCLV3 displayed gain-of-function effects, with reductions in the number of stigma per floret to 1 or even 0 in ~25-50% of plants across lines (compared to the invariant 2 stigmata in wildtype plants), and a reduction in the number of florets per capitulum compared to wild-type lettuce in 3 out of 4 lines (Fig. 5g–i). As such, normal stem cell homeostasis in Asteraceae requires attenuated CLV3p receptor affinity. We next wanted to determine if this correlates with altered downstream CLV-signaling outputs in Asteraceae capitula. To determine this, we used RNAseq to assay transcriptional responses of sunflower capitula to treatment with either AsterCLV3p, AtCLV3p, and corresponding CLEp 5^th residue swaps (AsterCLV3p^Leu5Ser and AtCLV3p^Ser5Leu), as our biological and genetic analysis demonstrates that the conserved AsterCLV3p Leu5 impairs signaling capacity and receptor binding. Sunflower was used because the larger inflorescence meristems of sunflowers are readily accessible (easily dissected) and both HaCLV1a and HaCLV1b paralogs retain the canonical LRR6 glycine necessary for CLEp binding, allowing the assessment of AsterCLV3p signaling without confounding impacts from CLV1 alleles with potentially reduced CLEp binding. As the outer waxy cuticle of inflorescence meristems blocks CLEp diffusion⁴⁶, we devised a strategy in which we isolated tissue cores from the center of expansion phase sunflower capitula prior to floret differentiation (IM – 30d old sunflowers), incubated these IM tissue cores with CLV3p containing solutions and then compared transcriptional response to cores incubated with control solutions (Supplementary Fig. 15a–b). Tissues were collected for RNAseq 5 and 24 hrs post-treatment, a period that did not impact capitulum cell viability. Sunflower capitula responded most strongly to AtCLV3p (RTVPS), with 161 and 93 DEGs at 5 hrs and 24 hrs, respectively (Fig. 5j). Ser5Leu substitutions in AtCLV3p (RTVPL) reduced the number of DEGs, with 21 DEGs at 5 hrs and 0 at 24 hrs, consistent with a reduction in signaling capacity and the critical role for Ser5 in receptor binding (Fig. 5j). In contrast to AtCLV3p, sunflower capitula responses to their own AsterCLV3p (RAAPL) were strongly attenuated, with just 11 DEGs at 5 hrs, and only 1 DEG at 24 hrs (Fig. 5j). Capitulum transcriptional responses to AsterCLV3p were enhanced in AsterCLV3p^Leu5Ser (RAAPS) treatments, with 104 (5 hrs) and 96 (24 hrs) DEGs, again highlighting the importance of Ser5 for CLV3p signaling (Fig. 5j). Remarkably, the 1 DEG identified in AsterCLV3p 24hr treatments was HaWUSb. In other Angiosperms studied, WUS is rapidly downregulated in response to CLV3p^14,21. HaWUSb was significantly downregulated at 5 hrs and 24 hrs by both AtCLV3p and AsterCLV3p^Leu5Ser (Fig. 5k). In contrast, AsterCLV3p treatments delayed HaWUSb downregulation as HaWUSb is only significantly repressed 24 hrs post-treatment. As such, the reduced capacity of AsterCLV3p to repress stem cell proliferation in Asteraceae capitula is associated with dampened downstream CLV transcriptional outputs. Interestingly, Ser5 also conveyed specificity in signaling outputs as there was considerable overlap in downstream regulation in treatments with AtCLV3p (RTVPS) and the AsterCLV3p var (RAAPS) at both 5 hrs (53 shared DEGs) and 24 hrs (54 shared DEGs) (Supplementary Fig. 15c–d, Supplementary Files 6–7). Collectively, our data support a model wherein the downstream responses of CLV signaling in Asteraceae are conserved but dampened due to a quantitative reduction in CLV3p potency, thereby permitting the relaxed stem cell homeostasis critical for normal capitulum development.

We next traced the evolution of CLV3 across the Asterales, specifically focusing on Asteraceae outgroups within MGCA clade, using newly generated chromosome-level assemblies and transcriptomic datasets for species in the Campanulaceae (P. grandiflorus), Menyanthaceae (Nymphoides indica), Goodeniaceae (S. aemula), and Calyceraceae (Acicarpha procumbens) (Moore-Pollard et al 2025). Strikingly, A. procumbens (Calyceraceae), which has a cephalioid inflorescence that also requires prolonged stem cell proliferation as in Asteraceae capitula¹⁵, contains a single CLE1D1 CLV3 homolog (ApCLV3) whose CLEp motif was identical to the Asteraceae CLV3p (RAAPL; Supplementary File 3). Thus, the evolution of AsterCLV3p RAAPL predates a Calyceraceae/Asteraceae split, suggesting that it evolved only once across the MGCA clade. In contrast Menyathaceae CLV3p (NiCLV3p, N. indica) contained an RRVPA N-terminal peptide sequence, while the Goodeniaceae (SaCLV3p, S. aemula) N-terminal CLV3p residues contained an RTAPL. Notably, in our Arabidopsis clv3 complementation assays RTAPL variants of AtCLV3p display intermediate signaling capabilities compared to AtCLV3p and AsterCLV3p; while SaCLV3p and LsCLV3p both complement lsclv3 to similar levels (Figs. 3c, 5g and Supplementary Fig. 14g–h). Therefore, the evolution of head-like inflorescences in the MGCA clade appears to have been paralleled by a stepwise reduction in CLV3p activity and receptor affinity, from high activity in the Campanulaceae, to intermediate activity in the Goodeniaceae, to low affinity in the Calyceraceae and Asteraceae.

Discussion

The re-wiring of conserved developmental networks is thought to underly evolutionary novelty in body plans across kingdoms, but few mechanistic examples exist. Here we show that evolutionary innovation in Asteraceae inflorescence form was accompanied by quantitative reductions in CLV3p signaling due to a novel combination of substitutions in the AsterCLV3p CLE domain which lower signaling capacity and receptor affinity. At the same time, AsterCLV3p retains essential receptor binding CLEp residues. This balanced signaling capacity correlates with reduced repression of WUS, a conserved target of CLV signaling in Angiosperms. The reduced signaling threshold promotes stem cell proliferation sufficient for normal capitulum development, while avoiding the adverse effects of unchecked proliferation. This is likely to explain the conservation of the peptide and the maintenance of at least one CLV1 paralog with potential peptide-binding capability across the Asteraceae. Our findings provide a molecular mechanism for the meristem expansion central to both the inflorescence-origin and floral-unit-meristem models of capitulum evolution. Beyond shoot development, CLE peptides have diverse roles in plant growth. Our structural and biochemical data reveal distinct peptide-binding mechanisms that contribute to specificity in CLE-receptor interactions. Notably, Ser5 residues serve a context-dependent regulatory role, modulating CLE peptide activity and receptor specificity across different CLE-receptor clades. The evolution of the RAAPL CLV3p predates the Calyceraceae/Asteraceae divergence, with both families producing head-like inflorescences. Despite this, the Calyceraceae are a small and geographically isolated family, while the Asteraceae spread globally from their ancestral origin to become the largest flowering plant family on earth. As such, other evolutionary innovations likely contributed to the success of the Asteraceae, possibly through diversification of pollinator recruitment, seed dispersal traits, and the evolution of a rich arsenal of secondary metabolites^5,12,47. Moreover, recent findings suggest that ancient genome duplications, along with gene family expansions and subsequent selection, played a pivotal role in enabling the Asteraceae to diversify and successfully disperse across the globe (Moore- Pollard et al., 2025). Several data points suggest that the Goodeniaceae CLV3p ortholog is likely intermediate in signaling capacity between AsterCLV3p and outgroup CLV3p orthologs. Unlike the Calyceraceae and Asteraceae, species in the Goodeniaceae exhibit variable inflorescence forms, with some possessing head-like inflorescences⁴⁸. This raises the possibility that variation in inflorescence architecture may be linked to intermediate CLV3p activity. Finally, our work suggests that engineering receptor-ligand affinity could provide a novel avenue for the quantitative re-shaping of crop shoot development.

Methods

Plant material and growth conditions

Sunflower, lettuce, Gerbera and additional Asteraceae species accessions

Helianthus annuus cv Ha412HO (sunflower; PI 642777), Lactuca sativa (lettuce; PI 251246), Calendula suffruticosa (PI 607419), Gaillardia pinnatifida (W6 55897), Packera multilobata (W6 30293), Stokesia laevis (PI 346981), and Symphyotrichum novae-angliae (PI 667514) were all sourced from the USDA National Plant Germplasm System. Bidens ferulifolia cv. Compact Yellow was acquired as cuttings from North Carolina Farms Inc. Micropropagated Gerbera hybrida cv. Terra Regina plantlets were provided by Dummen Orange, The Netherlands. H. annuus cv. Pacino cola seeds were purchased from Siemenliike Siren Oy, Finland.

Asteraceae growth conditions

Sunflower achenes were sown on pre-wet vermiculite (A-3 coarse) and germinated under a dome at the UNC greenhouse in long-day conditions; LD, 16hr light/8hr dark (light supplemented to 400W/m²) with temperature kept between 20-25°C. At 7 days (7d) post germination, seedlings were transplanted to 4.5-inch geranium pots (Greenhouse Megastore CN-STD-0450) in soil (Metro-Mix 360/sand/perlite supplemented with Marathon pesticide) and continuously watered with Peter’s 20:20:20 [N:P:K] at 150ppm (parts per million). At 20d post germination, plants were then up-potted to 1-gallon pots in fresh soil and Osmocote [14:14:14] was added for top watering. Lettuce achenes were sown on 0.8% phytoagar with half-strength Murashige-Skoog (½ MS) media buffered with MES at pH 5.7 and germinated in a growth chamber (Percival CU-36L4) in LD conditions at 22°C for 5 d. Seedlings were transplanted to soil (2.8 cf PRO-MIX BX, 40g Marathon 1% G Insecticide, and 1 gal 200ppm Peter’s 20-10-20 fertilizer) in 6 cell seedling inserts (BWI FG12067) and grown in growth chambers (either Percival AR-66L2_or AR-75L2) in LD conditions at 22-25°C. 2-week-old seedlings were transplanted to 4.5 inch geranium pots (Greenhouse Megastore CN-STD-0450). 4-week-old plants were supported with bamboo stakes and thereafter fertilized weekly with 200 ppm Peter’s 20-10-20 fertilizer. Gerbera plants were all grown in a greenhouse as described previously⁴⁹. All other Asteraceae species were grown similar to sunflower methods above but in a custom-built grow room at UNC with environmental control (temperature maintained at 21-25°C) under red and blue LED lights in LD conditions.

Arabidopsis lines and growth conditions

Mutant alleles used in this study were all in the Col-0 (Columbia-0) background with specific allele information as follows: clv3 (clv3-9)⁵⁰, clv1 (clv1-101)⁵¹, bam1 (bam1-4)⁵⁰, and bam2 (bam2-4)⁵⁰. The published endodermal reporter EN7pro::YFP-H2B⁵² was used in root meristem analyses. All Arabidopsis were grown as follows: seeds were sterilized and plated on ½ MS media (with MES buffer at pH 5.7) and stratified in the dark at 4°C for 2 days. Plates were then either moved to a custom-built grow room at UNC with environmental control (temperature maintained at 21-25°C) or two Percival growth chambers (AR-75L3 at UNC or Auburn). By 10d, seedlings were transplanted to soil (Metro-Mix 360/sand/perlite supplemented with Marathon pesticide and Peter’s 20:20:20 at recommended levels) then grown in same conditions they were germinated in until mature and flowering for phenotypic analyses.

Confocal microscopy

For sunflower, shoot apices from 10d, 20d, 30d, 35d, and 37d (days post germination) plants were removed, cut longitudinally and fixed overnight in a 1:20 (tissue:fixative) volume of FAA (2% formaldehyde, 5% acetic acid, 60% ethanol v/v) at 4°C with constant agitation on a rocker. Meristems were cleared and autofluorescence was imaged on a confocal laser scanning microscope as described previously for the 10d, 20d and 30d stages on a Zeiss 710 using a Plan-APOCHROMAT 10x objective (NA=0.45)^19,53. For 35d and 37d stages, meristems were imaged on a Zeiss 880 using an EC Plan-Neofluar 10x objective (NA=0.30) in tiling mode. For both stages, 24 images were stitched together using Nikon NIS-Elements to produce larger spatially resolved micrographs of entire developing capitula. For Gerbera, images of proGhCLV3:NLSeGFP:GhCLV3ter were taken with a Leica SP8 upright laser scanning confocal microscope equipped with a HC Fluotar L 25x/0.95 VisIR water dipping objective. The GFP was excited with a 488nm laser, and the emission was collected from 495nm to 545nm. Given the large size of the IM/FM stage, 9 tile scans were taken with a single sample and merged by the LAS X software. The images were then exported into tiff image stacks and further visualized with the MorphoGraphX software⁵⁴. For Arabidopsis, dissected root meristems of EN7pro::YFP-H2B were imaged on a Zeiss 710 using a C-Apochromat 40x water immersion objective (NA=1.20). Seedlings were fixed in room temperature 4% PFA for 4 minutes to preserve fluorophores. Fixed seedlings were stained with 0.1% Calcofluor white (Fluorescent Brightener 28; Sigma-Aldritch) in Clearsee (10% xylitol, 15% sodium deoxycholate, 25% urea)⁵⁵ and following dissection of the root tips mounted in Clearsee. Root meristems were imaged with following excitation/emission settings: YFP (514nm; 519 to 564nm) and Calcofluor white (405nm; 410 to 551nm). Unless specified, all images were processed as separate channels and then merged when necessary in Fiji (v.2.14.0/1.54f)⁵⁶.

Scanning electron microscopy

Lettuce inflorescences (WT and Lsclv3-cr) were dissected and fixed in Formaldehyde Acetic Acid overnight, before serial dehydration in increasing concentrations of ethanol (70%, 80%, 90%,100%; 2hrs each step). The dehydrated samples were then critical point dried in Tousimis Samdri-PVT-3D and gold sputter coated using a Quorum EMS 150R ES. The samples were then imaged using a Zeiss EVO50. Gerbera inflorescences (WT and GhCLV3 RNAi lines) were dissected, fixed and imaged as described previously⁵⁷ using a Quanta 250 SEM (FEI Corporation).

Photography and stereoscope imaging

Images of lettuce stigmata, florets, and capitulescence meristems were captured on a stereo microscope (Leica M60, Flexacam C3). Photographs of inflorescences from Arabidopsis, lettuce and Gerbera were taken using a DSLR Camera (Cannon EOS Rebel T5 with a Tokina 100mm f/2.8 AT-X M100 AF Pro D macro lens or similar). Images were edited in Gimp (v2.10) or Adobe Photoshop. Arabidopsis bam1/2 rosette phenotypes and carpel pictures were made using a Samsung Galaxy S23 phone. For bam1/2 inflorescence pictures (sterility phenotype), source images were photographed using a Canon Eos Rebel T6i DSLR camera on a tripod equipped with a Canon EF-S 35mm f/2.8 Macro IS STM lens. A focus stack of varying focal planes through the sample were acquired, and composite inflorescence images were made using Affinity Photo (version 1.10.6.1665) Focus Merge tool.

RNA sequencing and analysis

Staging, CLEp treatments and RNA extraction

For sunflower developmental RNAseq, shoot meristems from 10d (vegetative meristems), 20d (floral transition meristems), 30d (inflorescence meristems; IMs) and 35d (inflorescence and floral meristems) post-germination sunflower plants were dissected and flash frozen in 1.5mL microcentrifuge tubes in liquid nitrogen. Meristems were collected in three biological replicates with 10-20 meristems pooled per sample and stored at −80°C until further processing. For de novo transcriptome assemblies of Asteraceae species, shoot meristems were dissected spanning different stages of development including vegetative and floral (when available). Approximately 8-10 meristems were pooled into one sample per species, flash frozen in liquid nitrogen and stored at −80°C until processing. For both the sunflower developmental series and Aster shoot meristem transcriptomes, samples were hand ground into fine powder using small pestles fitted to 1.5mL microcentrifuge tubes in liquid nitrogen. For CLV3p variant IM response experiment, 30d sunflower shoot meristems were dissected and stored on inflorescence meristem media (1/2 MS minimal, 4% sucrose (w/v), 0.05% casein hydrolysate (w/v), pH 5.5, 100mg/L myo-inositol, 1.2mg/L kinetin, 0.1% plant preservation mixture (PPM), 1% agarose) while all ~100 meristems for a single rep were dissected (~1-2hrs). A single tissue core was taken from the center of each 30d IM using a 1.5mm Biopsy Punch (Miltex) to facilitate CLEp delivery to meristematic cells upon treatment. Approximately 8-10 30d IM cores were used for each treatment and replicate. Cores were incubated in inflorescence meristem media (without agarose) containing 0.1uM CLEp vars (no peptide control, AtCLV3p, AsterCLV3p, AtCLV3p S5L, and AsterCLV3p L5S) for 5hrs or 24hrs at room temp with constant agitation on a table-top plate rotator. Samples were then collected in 2mL microcentrifuge tubes and flash frozen in liquid nitrogen. Three biological replicates at 5hrs and 24hrs (30 total samples) were then ground using a liquid nitrogen cooled genogrinder (Spex). RNA was extracted from each sample using the EZNA Plant RNA kit (Omega Bio-tek) and treated with RNase-free DNase (Omega Bio-tek). Approximately 1.5μg of RNA was used in library preparation, using the stranded mRNA-seq kit (Kapa Biosystems) at the High-Throughput Sequencing facility at UNC. 50bp paired-end reads were generated on either an Illumina HiSeq 4000 or 6000 with read depths that varied across experiments. Samples for de novo transcriptomes were sequenced at ~80-100 million (M) reads per sample, while samples for differential expression analyses were sequenced at ~50-100M reads (developmental series) and ~30-40M reads (CLEp treated IMs) per replicate.

Gene expression analyses

For sunflower, Illumina paired-end RNAseq reads were quality assessed (FastQC v0.11.9⁵⁸, MultiQC v1.14⁵⁹) and genome indexed (STAR v2.7.10b⁶⁰). Reads were mapped to the HA412-HOv2 sunflower reference genome¹⁷ and raw read counts per gene were quantified (STAR v2.7.10b). Combat-Seq was used to reduce unwanted variation between biological replicates (sva v3.42.0⁶¹). Differential gene expression analysis was done with DESeq2⁶² (v1.34.0) with the model ~0 + development stage using a Wald test for significance and multiple test correction with Benjamini & Hochberg. For clustering, per the DESeq Workflow, the count data was transformed using variance stabilizing transformation without blind dispersion estimation. Hierarchical clustering was done on the VST-transformed data in R⁶³ (v4.1.2) using hclust. All genes that were differentially expressed between the pairwise comparisons are included (12,657 genes total). The distance matrix was calculated with Euclidean distance and the complete linkage method for clustering was used. The tree was cut at a height of 2.7 (Supplementary Fig. 1b). For lettuce, publicly available datasets were downloaded using sratoolkit (data: SRR6388382- SRR6388389²⁵) and paired-end reads were mapped to the lettuce reference genome (v8⁶⁴) using hisat2⁶⁵ (v2.2.1). Stringtie⁶⁶ (v2.1.1) was used to produce a shoot meristem transcriptome-based annotation (LsCLV3 is unannotated in the available reference genome annotation) which was then leveraged to quantify gene-level expression of the stem cell regulators LsCLV3 and LsWUSb using featureCounts in subread⁶⁷.

De novo transcriptome assembly

Illumina paired-end RNAseq reads from shoot meristems (SAMs) of six Asteraceae species (Supplementary Fig. 2c) were quality assessed (FastQC v0.11.9⁵⁸) then assembled into de novo transcriptome assemblies using the Trinity pipeline^68,69 (v.2.11.0) containing SAM-expressed genes. Coding regions of transcripts were identified and protein sequences extracted using the TransDecoder⁷⁰ (v5.5.0) LongOrf and Predict functions with -m flag set at 60 to capture smaller proteins of potential interest (i.e CLE peptides). The protein fasta file output from each Asteraceae SAM transcriptome were used for downstream identification of stem cell signaling orthologs.

Ortholog Identification

Proteome analysis for CLV3 and CLV1 across Angiosperms

To identify CLV1 and CLV3 sequences in Asteraceae, we analyzed a comprehensive set of complete proteomes from 13 Asteraceae species and 37 outgroup species, representing major orders from Monocots, Magnoliids, and the ANA grade (taxonomic information and accession numbers are available in Supplementary Fig. 2 and Supplementary File 2). Among these, eight represent newly assembled genomes from the Asterales (Moore-Pollard et al. 2025), and our analysis of CLV1-related genes was extended to include the six unpublished de novo transcriptomes (derived from shoot meristems; see above) from diverse Asteraceae species.

CLE-clustering and identification of AsterCLV3

For the CLV3 analysis, we conducted two independent HMMER3⁷¹ screenings using profiles of CLE motifs derived from Oelkers et al. (2008)⁷² and Zhang et al. (2020)²³, applying a domain e-value threshold of 40 to reduce the chance of false positives. Since CLE peptides are unsuitable for standard phylogenetic methods due to their small size and high sequence variability outside the conserved domain, we employed a phenetic clustering approach using CLANS⁷³ to identify putative Asteraceae orthologs of CLV3. During the initial run, we tested seven e-value thresholds for high-scoring segment BLASTP⁷⁴ pairs, ranging from 1e-04 (default) to 1e-10. The screening based on the Zhang et al. profile was slightly more conservative (retrieving 1052 sequences vs. 1200), resulting in fewer singletons; thus, we used this dataset for the final analyses. An e-value threshold of 1e-7 produced clustering patterns consistent with previous studies^22,72, including a clearly defined 1D1 cluster with canonical CLV3 from A. thaliana.

CLV1 Ortholog Identification

The identification of CLV1 genes involved BLASTP screening, followed by InterProScan⁷⁵ scanning to confirm the presence of the signature protein-kinase domain (PF00069) and leucine-rich repeat (LRR) regions (PF00560) at an e-value of 1e-10. In transcriptomes where multiple splicing variants were represented, we retained the longest variant for subsequent analyses. The initial search retrieved 1065 genes, of which 302 were confirmed as belonging to the CLV1/BAM family based on preliminary phylogenetic inference. For the final reconstruction, sequences were re-aligned using the G-INS-i algorithm implemented in MAFFT⁷⁶ (v7.271). After manually correcting misaligned positions, we collapsed duplicated sequences using ElimDupes and performed quality trimming with trimAl⁷⁷ (v1.2rev59). Lastly, phylogeny was reconstructed using RaxML⁷⁸ (v8.2.4), with automatic protein model selection (-m PROTGAMMAAUTO flag), and branch support estimated with 200 rapid bootstrap replicates.

We further validated the identity of CLV1 and CLV3 orthologs by blasting the protein sequences of CLV1a, CLV1b, and CLV3 from Helianthus annuus L. and Lactuca sativa L. against coding sequences of eight Asterales species (with reference-level assemblies and new gene annotations) (Moore-Pollard et al. 2025) using TBLASTN⁷⁹ (v2.13.0). CLV genes of each taxon were confirmed by having the lowest e-value and highest score from BLAST (Supplementary Files 4–5). GhCLV1a and GhCLV1b were identified from Gerbera transcriptome. The sequences were amplified and validated by sequencing.

In situ hybridization

The cultivars used for in situ hybridization are Gerbera cv. Terra Regina, sunflower cv. Pacino cola, and lettuce (PI 251246). GhWUSb, HaCLV1a and LsCLV1a in situ hybridization was done with gene-specific probes, and GhCLV1a, GhCLV1b, HaCLV1b, HaWUSb, LsCLV1b, and LsWUSb with full length hydrolyzed probes. CLV3 genes were done with full length probes.. The probes were amplified with PCR. After purification, the probes were synthesized as previously described⁸⁰ and labeled with DIG RNA Labeling Kit (Roche) according to the instructions. When appropriate, the labeled probes were further hydrolyzed to approximately 150bp⁸¹. The preparation of plant samples, sectioning, and hybridization were done as described by Elomaa et al. 2003⁸². Oligo sequences used for in situ hybridization are found in Supplementary Table 1.

Root CLE peptide treatments

All CLE peptides and variants used in this study were synthesized by Biomatik (>90% purity) as in previous studies⁴⁵.

CLE peptide variant root length assay

Arabidopsis (Col-0) seeds were sterilized then plated on a fine mesh placed on ½ strength Murashige and Skoog medium supplemented with 0.8% Phytoagar. The seeds were stratified for 48 hours at 4°C, then germinated vertically under 24-hour light at 22°C for four days. The fine mesh that the seedlings were germinated on was then transferred to ½ strength Murashige and Skoog medium supplemented with 0.8% Phytoagar and 0.1μM mock treatment, CLEp or CLEp variants (with S5L substituions) under sterile conditions. Seedlings grew for 4 days vertically following the transfer, then plates were scanned and root lengths were measured using Fiji⁵⁶.

CLV3p root length assay

For Arabidopsis, seeds were sterilized then plated on ½ strength Murashige and Skoog medium supplemented with 0.8% Phytoagar, and either 0.1μM of CLV3p (from Arabidopsis, Asteraceae, outgroup taxa or amino acid substitutions as indicated) or varying concentrations of Arabidopsis and Asteraceae CLV3p (0μM, 0.01μM, 0.1μM, 1μM, and 10μM). The seeds were stratified for 48 hours at 4°C, then germinated vertically under 24-hour light at 22°C for ten days. Plates were then scanned, and root lengths were measured using Fiji⁵⁶. For lettuce and sunflower, achenes were sterilized and plated on ½ MS buffered with MES at pH 5.7 containing 0.1μM of CLV3p (species and amino acid substitutions as indicated). For sunflower, achenes were grown in sterile cups containing ~200mL of media for ten days in long-day conditions. Only seedlings with fully expanded cotyledons and true leaves initiated were used for root length comparisons. For lettuce, achenes were grown vertically on plates in 24-hour light at 22°C for seven days. Plates were scanned and primary root length measured using Fiji⁵⁶.

Transgenic lines

For Arabidopsis, new transgenic lines were generated through floral-dip transformation⁸³ of binary vectors into indicated backgrounds using Agrobacterium tumefaciens (GV3101). Transgenic plants were selected for on ½ MS plates using appropriate antibiotic selection (Kanamycin or Hygromycin). Unless specified differently, T1 plant lines were used to assess transgene function across many (typically >10) independent transgene insertion events to reduce selection bias.

For lettuce, achenes were germinated as described above. 5-day-old cotyledons were removed and their apical tips were dissected off with a razor blade. Cotyledon explants were added to a suspension of A. tumefaciens (GV3101) and nutated for 5 min. A. tumefaciens inoculation suspension was created by pelleting 2 mL of overnight culture and resuspending in 20 mL inoculation solution (½ MS salts and vitamins, 1% sucrose, 0.5 g/L casamino acids, 200 mM acetosyringone, pH 5.33). Explants were lightly patted dry to remove excess A. tumefaciens and placed abaxial side down on co-cultivation media (½ MS salts, full strength MS vitamins, 3% sucrose, pH 5.7, 0.8% phytoagar, 0.1 mg/L NAA, 0.5 mg/L BAP, and 200 mM acetosyringone). A piece of sterile filter paper was placed between explants and co-cultivation media to prevent A. tumefaciens overgrowth. Co-cultivation plates were then kept in the dark for 3 days at 22°C. Explants were then transferred to callus induction media containing Timentin (RPI T36000-5.0) to kill A. tumefaciens and Kanamycin for antibiotic selection (½ MS salts, full strength MS vitamins, 3% sucrose, pH 5.7, 0.8% phytoagar, 150 mg/L Timentin, 100 mg/L kanamycin, 0.1 mg/L NAA, and 0.1 mg/L BAP) and kept at 22°C for 2 weeks in a tissue culture chamber (Percival CU-36L4). The resulting callus and nascent shoots were then transferred to shoot elongation media (½ MS salts, full strength MS vitamins, 3% sucrose, pH 5.7, 0.8% phytoagar, 150 mg/L Timentin, 100 mg/L kanamycin, and 0.01 mg/L BAP) for 1-3 weeks. Once shoots elongated, they were excised from surrounding callus and placed in rooting media (½ MS salts, full strength MS vitamins, pH 5.7, 0.8% phytoagar, 150 mg/L Timentin, 100 mg/L kanamycin, and 0.1 mg/L NAA). Once roots developed, plants were transferred to soil, covered with humidity domes for 2 d before acclimated to chamber conditions fully. For Gerbera, wild-type Gerbera hybrida cultivar ’Terra Regina’ transformation was performed as previously described⁸⁴.

Molecular cloning

Primer sequences used for cloning can be found in Supplementary Table 1.

For the generation of CLV1 and BAM1 mutant vectors, recombinant PCR was used to create LRR DNA fragments with point mutations in BAM1 or CLV1 as described before⁵⁰. Fragments were cloned into the pBLUNTII Topo vector (Invitrogen) and sequenced. Fragments were then subcloned into pENTR vectors containing the corresponding wild-type full length genomic-2XGFP receptor clones, sequenced, and then recombined into previously created gateway compatible native promoter binary vectors⁵⁰. For CLV3 variants, CLV3 variants were synthesized by Twist Bioscience into manufacturer provided entry plasmids and then recombined into a previously created gateway compatible native promoter CLV3 binary vector.

Entry plasmids (pENTR) containing the LsCLV3 (lettuce) genomic sequence with CLE domain encoding different variants (AtCLV3, AsterCLV3, and SaCLV3) and an empty vector containing both 5’ upstream and 3’ downstream LsCLV3 cis-promoter regions were synthesized by Twist Bioscience. LsCLV3 variants were amplified by PCR using cloning primers with PacI sites, then digested and ligated into the LsCLV3pro binary to form expression vectors. All binary vectors were introduced into chemically competent Agrobacterium tumefaciens (GV3101).

GhCLV3 full length coding sequence was cloned into the Gateway compatible RNAi vector pK7GWIWG2(II)⁸⁵. For GhCLV3 transcriptional reporter construct, 1652bp upstream and 236bp downstream of GhCLV3 were cloned into entry vectors pDONR_Amp_P4P1R and pGEMtP2rP3 as promoter and 3’ sequences (both entry vectors were kindly provided by Prof. Ari Pekka Mähönen, University of Helsinki, Finland). The sequences were verified by plasmid sequencing. An LR reaction was performed to clone the promoter, NLSeGFP and 3’ sequences into destination vector pKm43GW⁸⁵. The constructs were transferred into the A. tumefaciens strain C58C1.

Lettuce mutant analyses and Gerbera RNAi

For lettuce, two Lsclv3 mutant alleles were generated, a +1 insertion (Lsclv3-cr) and a −6 deletion (Lsclv3-cr2) (Supplementary Fig. 14a). Lsclv3-cr was detected by dCAPS PCR and BslI digestion, and Lsclv3-cr2 was detected by PCR amplification and MwoI digestion with listed primers (Supplementary Table 1). As Lsclv3-cr plants rarely produced viable seed, and transgenic LsCLV3 variants weren’t easily distinguished from WT LsCLV3 by PCR, transgenes were introduced into an Lsclv3-cr/Lsclv3-cr2 biallelic mutant background (as the Lsclv3-cr2 −6 deletion is in frame and there was no detectable phenotype). Transgenic achenes were sown on ½ MS media with kanamycin for antibiotic selection and confirmed using a 35S::tdTomato selectable marker. All complementation lines were confirmed as both transgene positive and Lsclv3-cr −/− mutants prior to phenotypic analysis.

For Gerbera, six independent transgenic GhCLV3 RNAi lines showed fasciated inflorescence phenotypes. The transgenic line TR5 was selected for gene expression analysis. The undifferentiated meristem was dissected from inflorescences of 2 to 3 mm in diameter from both GhCLV3 RNAi and wild-type plants. Three independent biological replicates, each with at least three meristem samples, were collected for transgenic and wild-type plants. Total RNA preparation, cDNA synthesis (500 ng total RNA was used), quantitative RT-PCR, and data analysis were done as previously described⁸⁶. Gene expression levels were normalized to the reference gene GhACTIN. All primers used are listed in Supplementary Table 1.

Phenotypic analyses

All phenotypic data was collected and analyzed using Prism 10 (GraphPad). Details of statistical analyses can be found in Supplementary File 8.

Leaf numbers for the Arabidopsis clv3 complementation experiment were made on 3-week-old soil grown plants using a tabletop dissecting scope. Stem width measurements were performed in the same experiment as described before⁸⁷. bam1 bam2 rosette phenotypes were scored in 3-week-old soil grown plants, and fertility was confirmed by the production of siliques containing seeds. All Arabidopsis carpel number quantifications were made as described previously⁸⁷, with 15 flowers counted per T1 plant (clv3 experiments), and 20 flowers counted per T1 (clvl experiments). bam1 bam2 inflorescences were photographed using a Canon Eos Rebel T6i DSLR camera on a tripod equipped with a Canon EF-S 35mm f/2.8 Macro IS STM lens. A focus stack of varying focal planes through the sample were acquired, and composite inflorescence images were made using Affinity Photo (version 1.10.6.1665) Focus Merge tool.

Lettuce (PI 251246) forms a paniculate capitulescence with a terminal capitulum that reaches anthesis first, followed by capitula terminating the secondary branches in basipetal order. To avoid variation due to developmental age, only capitula terminating secondary branches were analyzed in petal, stigma, and total floret counts for lettuce mutant and complementation experiments. Two capitula were quantified per plant in the initial WT vs Lsclv3-cr analyses, then four of the first six capitula were quantified in mutant complementation analyses. Stem width was measured at the time of floral organ counts using digital calipers around stem above the second true leaf.

Protein expression and purification

The codon optimized ectodomains of AtBAM1, AtBAM1^G199A, AtBAM1^G199Q(residues 20–629) was cloned into a modified pFastBac vector (Geneva Biotech) under a polyhedrin promoter and fused to Drosophila BIP signal peptide⁸⁸. The BAM1 ectodomains were followed by a TEV (tobacco etch virus protease) cleavage site, and a tandem affinity tag Strep-9xHis at its carboxy terminus. Baculovirus generation was performed in DH10 cells and virus amplification was done in Spodoptera frugiperda Sf9 cells. The proteins were produced in Trichoplusia ni Tnao38 cells, infected with a multiplicity of infection (MOI) of 3. Cells were grown 1 day at 28 °C and two days at 22 °C at 110 rpm. The secreted proteins were purified separately by sequential Ni²⁺ (HisTrap excel, GE Healthcare), equilibrated in 25 mM KPi pH7.8 and 500 mM NaCl and Strep (Strep-Tactin Superflow high-capacity, (IBA, Germany), equilibrated in 25 mM Tris pH 8.0, 250 mM NaCl, 1 mM EDTA) affinity chromatography. Affinity tags were removed by incubating the purified proteins with recombinant His-tagged TEV protease, in a 1:50 ratio overnight at 4° C. The proteins were further purified and the cleaved tag removed by size-exclusion chromatography on a HiLoad 16/600 Superdex 200 column (Cytiva, 28-9893-35), equilibrated in 20 mM citric acid, 150 mM NaCl, pH 5.0. Peak fractions containing the proteins were concentrated using Amicon Ultra 30kDa concentrators (UFC9030). The wild-type BAM1 ectodomain was analyzed for purity and integrity by analytical SEC (size exclusion chromatography) with and SDS-PAGE³⁵.The BAM1 pocket variants were analyzed for purity and integrity by analytical SEC using a BIO-RAD ENrich SEC 650 (7801650) and SDS-PAGE (Supplemental Fig. 13b).

Isothermal titration calorimetry (ITC)

Binding experiments were performed at 25 °C using a MicroCal PEAQ-ITC (Malvern Instruments) with a 200 μL standard cell and a 40 μL titration syringe. AtBAM1 and pocket variants were gel filtrated into the ITC buffer (20 mM sodium citrate, 150 mM NaCl, pH 5.0). Protein concentrations were determined based on their molar extinction coefficients and calculated molecular weights after removal of the signal peptide and affinity tags. A typical experiment consisted of injecting 2 μL of a range between 100 and 800 μM solution of CLE peptides and variants, into a 10 to 15 μM AtBAM1 or AtBAM1 variant solution in the cell. Experiments were run at 150 s intervals and 500 rpm stirring speed. Experiments were done in duplicates and data were analyzed using the MicroCal PEAQ-ITC Analysis Software provided by the manufacturer. Control experiments of ligand vs buffer were performed independently, and no dilution heat was observed. All ITC runs used for data analysis have an N ranging from 0.9 to 1.1. The N values were fitted to 1 in the analysis.

Structural modeling and prediction of water molecules within the binding pocket

Structural models of BAM1/CLV1 receptors in complex with peptides were generated using the Alphafold3 server (https://alphafoldserver.com/)⁸⁹. For AtCLE13 in complex with AtBAM1, final peptide-receptor refinement was performed using high-resolution modeling through the Rosetta FlexPepDock tool, with total energy score of −430.181 and a RMSD of 0.997. The prediction of water molecules within the binding pocket was done using Waterdock 2.0⁹⁰. Generation of molecular images was performed using ChimeraX⁹¹ and PyMOL (The PyMOL Molecular Graphics System, Version 2.6.1 Schrödinger, LLC.).

Conservation analysis of receptors surface

Conservation analysis within CLV1/BAM1/TDR receptors was made using ESPript 3.0 server (https://espript.ibcp.fr/ESPript/ESPript/)⁹². A multiple sequence alignment produced with Clustal Omega⁹³ was used as input. Images were generated using PyMOL⁹⁴.

Data Availability

RNA-seq data from all Asteraceae species sequenced are available on the National Center for Biotechnology Information in BioProject PRJNA1261082. Sunflower data is in SAMN48412296, and all other species can be found in SAMN48412297- SAMN48412302. RNA-seq data from lettuce was obtained from Chen et al²⁵.