Cantil: a previously unreported organ in wild-type Arabidopsis regulated by FT, ERECTA and heterotrimeric G proteins

Despite the phenotypic variation of individuals, organisms customarily reach a final, generalized body plan that is characteristic of the species. In plants, this plan is plastic, as the indeterminate plant body allows recurrent addition or suppression of distinct organs (e.g. leaves, branches and flowers), thereby fine-tuning the developing plant architecture to the environment. For example, as a facultative long-day plant, short days (SDs) promote Arabidopsis thaliana vegetative growth and delay flowering, while long days (LDs) stimulate flowering (Giakountis et al., 2010). The archetypal architecture of Arabidopsis plants is well established (Provart et al., 2016) and, at maturity, Arabidopsis plants exhibit three distinct metamers: a basal rosette bearing leaves (M1), an upright co-florescence-producing proximal primary stem (M2) and a distal stem segment bearing nodes that directly produce single flowers borne on a pedicel (M3) (Fig. 1A). Archetypal pedicel attachment is exemplified by upward-slanting pedicels that emerge directly from the stem (Fig. 1B) (Denay et al., 2017).

Pedicel emergence directly from archetypal M3 nodes, without a lateral intermediary structure, is a robust feature that previous publications suggest is controlled by specific genetic determinants; overexpression of multiple proteins, in both Arabidopsis and Brassica napus (canola), causes atypical development of a supporting structure basal to the pedicel (Kirik et al., 1998; Wen and Walker, 2006; Dumonceaux et al., 2009; Ikeuchi et al., 2011; Colling et al., 2015). These gain-of-function studies point towards the existence of a signaling cascade that restricts pedicel attachment directly to the stem, yet do not conclusively identify any of the components, particularly as overexpression can result in neomorphic effects.

In our work, we noticed the infrequent occurrence of similar M3 node morphological deviations in diverse Arabidopsis knockout lines, particularly in null mutants of the heterotrimeric G protein complex. Heterotrimeric G protein complexes, comprising Gα, Gβ and Gγ subunits, function as conserved eukaryotic signaling hubs that play fundamental roles in development and environmental responses (Urano et al., 2016b). Plant Gα and Gβ proteins are known to modulate the size of the shoot apical meristem (SAM) (Bommert et al., 2013; Ishida et al., 2014; Urano et al., 2016a), which is the progenitor of emergent aboveground organs, and localization of the sole canonical Arabidopsis Gα (GPA1) and sole Gβ (AGB1) subunits in the SAM is well documented (Weiss et al., 1993; Huang et al., 1994; Anderson and Botella, 2007; Ishida et al., 2014). G protein signaling specificity is achieved via 12 distinct G protein complexes in Arabidopsis via the association of GPA1 (Gookin and Assmann, 2014), or of one of three atypical extra-large Gα subunits (XLG1, XLG2 and XLG3) (Chakravorty et al., 2015), with one of three obligate Gβγ dimers (AGB1-AGG1, AGB1-AGG2 or AGB1-AGG3). In each case, dissociation of the complex leads to free Gα and Gβγ dimers, which can signal to additional effectors (Urano et al., 2013). Through observation, we linked the infrequent morphological deviations we observed in LDs to wild-type and mutant Arabidopsis plants that were initially grown for an extended time period in SDs. Here, we identify the wild-type expressivity of these M2-M3 transition morphologies, the underlying requirements for their development, and specific positive and negative genetic determinants that govern their formation.

Wild-type Arabidopsis accession Columbia (Col-0) plants flowering under SD (8-10 h) morphologically deviate from the archetypal pedicel arrangement by developing a macroscopic anatomical structure we call a ‘cantil’, for cantilever, as it holds the pedicel at a distance from the stem (Fig. 1C-E). Cantils originate from the stem without an obvious junction but terminate at a distinct junction that holds a single pedicel at an angle (∼25-90°). The typical upward arc of longer cantils (Fig. 1E) is less obvious on shorter cantils (Fig. 1D), and, at its weakest, cantil growth is limited to a ‘cuff’ that comprises an unelongated cantil that encircles the entire pedicel base (Fig. 1E; see Fig. S1A-E for phenotype overview). In wild-type Col-0 plants, the magnitude and frequency of cantil display is most prominent nearest the M2-M3 junction (Fig. 1D and Fig. S1B-D), and characteristically decreases acropetally with pedicel position (Fig. 1F). SD-grown wild-type Wassilewskija (Ws) plants produce cantils that are morphologically similar to those of Col-0 (Fig. S1F), but do so 10-fold less frequently (Fig. S1G). Notably, the widely used Landsberg erecta (Ler) strain fails to produce any cantils or cuffs (Fig. S1G,H). When flowering in LD (16 h), wild-type Col-0 very rarely deviates from the archetypical form, e.g. producing only two cantils and three cuffs in >3635 inspected pedicels from 727 plants (Fig. S1G,I). No cantils were observed for Ws or Ler flowering in LD (Fig. S1G). These results suggest the variable penetrance and expressivity of wild-type cantil formation (Col-0>Ws>Ler and SD>LD) is controlled by distinct genetic elements.

Col-0 background gpa1 (Fig. 1G) and agb1 (Fig. 1H) mutants flowering in SD produce cantils with increased magnitude and frequency at each M3 inflorescence pedicel position relative to wild-type Col-0 (Fig. 1F, Fig. S1B-E). Null mutation of GPA1 or AGB1 extends the cantil developmental range to distal M3 nodes (Fig. 1G,H), with altered developmental patterning most evident distally adjacent to the M2-M3 junction (e.g. gpa1-3 in Fig. 1I, Figs S1E and S2A). Total cantil data from our two large-scale experiments showed that GPA1-null mutants in both the Col-0 (gpa1-3 and gpa1-4) and Ws (gpa1-1 and gpa1-2) backgrounds develop significantly more cantils in SD than do wild type (Fig. 1J). These differences are not explained by decreased co-florescence production (Fig. S2B) or by failed cuff elongation as all four gpa1 alleles also developed cuffs at a higher frequency (Fig. S2C). Similarly, SD-flowering AGB1 null mutants in the Col-0 background (agb1-1 and agb1-2) developed significantly more cantils than wild type (Fig. 1J,K). AGB1 also represses another phenotype, ectopic epidermal rippling (Fig. 1K,L), which formed at the M2-M3 junction in 84% of agb1 mutants (Fig. S2D). This rippling was often accompanied by very short internodes and a distal stem exiting the junction at an angle (Fig. 1L and Fig. S2E,F). Although primarily diagnostic for agb1, minor rippling and angled junctions appeared in 2% of GPA1 mutants (Fig. S2D). Like wild type, no cantils, ectopic rippling or altered stem angles were observed in gpa1-1, gpa1-3 or agb1-2 grown through flowering in LD (Fig. S2D), indicating these phenotypes are tightly linked to SD conditions. Under SD, cantil/plant averages were increased by 67% in gpa1 and 52% in agb1 mutants in the Col-0 background, and by a striking 612% in gpa1 in the Ws background (Fig. S2D).

Cantil repression by both GPA1 and AGB1 suggested signaling through one, or more, of the 12 heterotrimeric G protein complexes. Given that all three non-dissociable Gβγ dimers cannot form in agb1, signaling specificity was investigated in individual null mutants for the three Gγ-subunits (agg1-1c, agg2-1, agg3-1 and agg3-2) as well as the triple Gγ mutant. Both agg1 and agg2 recapitulated the Col-0 cantil frequency phenotype (Fig. S3A-C), while agg3-1 and agg3-2 mutants, contrastingly, increased cantil frequency (Figs. S3A,D) and recapitulated the ectopic rippling and stem angle defects of agb1 (Fig. S3E,F). The Gγ_1/2/3 triple mutant likewise morphologically recapitulated agb1 (Fig. S3G,H), and developed cantils at the same frequency as agb1-2 and the gpa1-3 agb1-2 double mutant (Fig. S3I), while XLG triple (xlg1-5 xlg2-1 xlg3-1) and RGS1 single (rgs1-1 and rgs1-2) mutants phenocopied Col-0 (Fig. S3J-M). Therefore, cantil repression and maintenance of the stereotypical wild-type inflorescence is mediated specifically by the subunits of a single G protein complex: GPA1-AGB1-AGG3.

Analysis of 1382 cantils revealed GPA1 is uniquely responsible for repressing a distal terminus outgrowth that we call a spur because of its characteristic form (Fig. 2A versus 2B, Fig. S4A) and its diagnostic location: abaxially centered basal to the cantil-pedicel junction. Over 95% of all gpa1 cantils in the Col-0 background develop spurs (Fig. 2C,D), with 66.6% of all gpa1 cantils developing solitary spurs, and 28.9% developing spurs that are flanked by reduced, less organized, nub-like growths (Fig. 2C) that are also seen in agb1 (Figs S1A, S4B). agb1 mutants do not directly phenocopy gpa1 spurs, but produce relatively smaller, abaxially off-center, spur-like growths (Fig. S4C) that occur at low frequency (26.2%) (Fig. 2C). Solitary and coincident spur and nub patterning differences between gpa1 and agb1 (Fig. 2C) further indicate developmental partitioning between these G protein subunits for these traits. GPA1 mutants in the Ws background (gpa1-1, gpa1-2) (Fig. 2C-E) also produce spurred cantils, which were never observed in wild-type Col-0 or Ws (Fig. 2C,D).

In theory, the prominent gpa1 spurs could arise from loss of GPA1 or from enhanced activity of free AGB1-AGG_1/2/3 dimers; however, double (gpa1 agb1) and pentuple (gpa1 agb1 agg1 agg2 agg3) knockout lines still produced typical gpa1 spurs (Fig. S4D-F), demonstrating that GPA1 is the factor repressing spur formation. Shortened gpa1 cantils frequently show spur formation near the base, and cuffs can form spurs at the stem-cantil junction (Fig. S4G) or fused to the stem (Fig. S4H). When fully displaced, these spurs are abandoned on the stem as independent stem ‘spikes’ (Fig. 2F,G, Fig. S4H-K). When present, fully formed independent stem spikes always occurred below spur-free gpa1 cantils, cuffs or archetypal pedicels (n=95), but never below spurred cantils (n=510) or spurred cuffs (n=103), providing further evidence that gpa1 spurs and spikes share a common origin.

Cantils form early during inflorescence development, with immature cantils already visible at the onset of M2 elongation (Fig. S4L,M) and well before M3 internode elongation in the flower head. At the M3 node, pedicel elongation occurs prior to cantil elongation, which is followed by spur and stem-spike initiation and elongation (Fig. S4L-N). Cantil and spur tissue is delineated from pedicel tissue by the small cells (Stenvik et al., 2006) that identify a vestigial abscission zone (AZ) (Fig. 2H-J), which is characteristically found at the stem-pedicel boundary in archetypal M3 nodes (Cho and Cosgrove, 2000). Therefore, the AZ functionally demarcates the cantil-pedicel boundary and links the origins of the cantil to the stem proper.

We screened 10 additional wild-type Arabidopsis accessions for SD cantil production and found that cantil competency exhibited natural variant dependency, and ranged from 0.0 to 2.0 cantils/plant (Fig. 3A-E). Notably, MYB13 overexpression in the permissive Dijon-G ecotype produced structures reminiscent of spurred cantils (Kirik et al., 1998). These results indicate cantil formation is widely distributed across Arabidopsis accessions, and, as such, identifies a readily available genetic resource to identify additional regulatory factors. To this end, we returned our attention to Ler.

The widely used Ler laboratory strain harbors an x-ray induced mutation of ERECTA (ER) (Torii et al., 1996; Zapata et al., 2016), which encodes a leucine-rich repeat receptor-like kinase (RLK) that directly influences lateral SAM signaling and architecture (Mandel et al., 2014; Zhang et al., 2021). Ler plants never produced any cantils, cuffs or ectopic rippling while flowering in either SD or LD (Fig. 1M, Fig. S1G,H), and RLKs have recently been implicated as phosphoregulators of plant heterotrimeric G proteins (Chakravorty and Assmann, 2018), suggesting ER may function as a direct positive regulator of cantil formation. Indeed, we found that the Col-0 background er-105 knockout phenocopies Ler and fails to produce cantils or cuffs (Fig. 3F), consistent with ERECTA enabling cantil formation in SD.

We genotyped and phenotyped a segregating, all Col-0 background agb1-2 (−/−) ER/er-105 (+/−) self-cross population (n=47) and found that all of the plants producing cantils and rippled M2-M3 junctions (76.6%, n=36) had at least one copy of wild-type ER (Fig. 3G). All 11 of the double homozygous agb1-2 er-105 segregants failed to produce any of the agb1 traits, demonstrating that erecta is functionally epistatic to agb1 and ERECTA functions as a global regulator of cantil formation.

Cantil expressivity is promoted by short photoperiod, but the underlying signaling response could arise from contemporaneous photoperiod sensing while flowering in SD, or from the temporally delayed transition to flowering that occurs in SD. Neither GPA1 nor AGB1 influence flowering time in SD (Fig. S5A,B) or LD (Trusov et al., 2008; Urano et al., 2016b); therefore, we assessed null mutants for FLOWERING LOCUS T (FT), a protein that functions as the predominant floral inducer after translocation from rosette leaves to the SAM (Pin and Nilsson, 2012). Col-0 FT null plants (ft-10) characteristically exhibit delayed flowering in LD (Yoo et al., 2005). We report that ft-10 plants exhibit delayed flowering in SD (Fig. S5C) and produce cantils in both SD and LD conditions, with equivalent frequency (Fig. 4A).

We used the Col-0 ft-10 photoperiod-independent cantil competency as a tool to evaluate the observed SD dependence of gpa1 cantil, spur and spike formation in LDs. Long-day ft-10 control plants (Fig. 4B,C) produced cantils morphologically identical to Col-0, while, in three experiments, up to 100% of LD gpa1-3 ft-10 plants produced spurred cantils (Fig. 4D-F). Long-day gpa1-3 ft-10 also produced stem-spikes in 61.1-68.2% of plants (Fig. 4F,G). Therefore, like cantil formation, spur and spike development is not directly dependent on photoperiod, but on flowering delays. Surprisingly, up to 59.1% of LD gpa1-3 ft-10 plants also phenocopied the agb1/agg3 trait of ectopic rippling (Fig. 4D-F), and the rippling frequency followed the spurs/plant and cantils/plant frequencies (Fig. 4F,G). Together, these results provide evidence that the observed photoperiod dependence of natural ‘wild-type’ cantil formation, and the altered morphology of G-protein mutants, is in fact a function of delayed flowering rather than of day length per se.

Our work directly identifies the existence of a regulatory network (Fig. 4H) that controls organogenesis at the M2-M3 junction and ensures archetypal pedicel emergence directly from M3 nodes. Wild-type Arabidopsis accessions are innately capable of producing cantils in SD, and FT mutation reveals that delayed flowering is the key expressivity determinant. The entire cantil signaling network is subject to global positive regulation by ERECTA, while negative regulation is provided by the specific GPA1-AGB1-AGG3 heterotrimeric G protein complex subunits, which repress cantil formation and suppress cantil spurs and stem spikes. The loss of GPA1 directly recapitulates the distal protrusions produced by both ROT4 peptide (Ikeuchi et al., 2011) and MYB13 transcription factor (Kirik et al., 1998) overexpression. Furthermore, the loss of AGB1 directly recapitulates the aberrant M2-M3 morphology and altered stem angle produced by TAX1 peptide overexpression (Colling et al., 2015). Potentially, those overexpression phenotypes reflect perturbed G protein signaling through an unknown mechanism.

The stem-like nature and origin of the cantil invites comparison with the pea tendril; the tendril is anatomically similar to the basal rachis (Gould et al., 1994; Tattersall et al., 2005) and exhibits a basal junction lacking axillary meristems (Tattersall et al., 2005), yet is defined as an organ (Hofer and Ellis, 1998; Hofer et al., 2009). Cantil displacement of the vestigial abscission zone from the main stem, coupled with the shared origin of the spur and stem spike, indicate that delayed flowering consequentially induces domains of extended activity in the phase-transitioning SAM. These cantil-related domains provide new targets for understanding and manipulating plant architecture.

In summary, our results lay the foundation for a new area of study, conditional organogenesis at the M2-M3 transition. We identify short photoperiod as the permissive factor, delayed flowering as the cause, and FT, ER and GPA1-AGB1-AGG3 as specific genetic determinants.

Arabidopsis thaliana plants were used for all aspects of this study. Columbia (Col-0), Wassilewskija (Ws) and Landsberg erecta (Ler) are widely available accessions, and were initially obtained from the Arabidopsis Biological Resource Center (ABRC), which was also the source for additional wild-type accessions (Dijon-G, Tsu-1, Sei-0, Kyoto, Cvi-1, Sp-0, W1-0, C-24, Est-1 and Bch1-3).

Columbia (Col-0) background G protein mutants (gpa1-3, gpa1-4, agb1-1, agb1-2 and agg3-1) were individually backcrossed twice to Col-0 and segregated prior to phenotypic quantification. An agg1-1c agg2-1 mutant originally obtained from the Botella lab (University of Queensland, Brisbane, Australia) was backcrossed to Col-0 to isolate the single agg mutants, which were subsequently backcrossed once more to Col-0. Seed bulking was performed before the large-scale experiments, and all of the G-protein mutant parent plants, as well as the wild-type Col-0, Ws and Ler parent plants were periodically genotyped to verify the absence of other contaminating mutations. The gpa1-3 agb1-2 double mutant was derived using backcrossed parents. The gpa1-3 agb1-2 agg1-1c agg2-1 agg3-1 pentuple mutant was generated by crossing gpa1 agb1 with agg1-1c agg2-1 agg3-1. RGS1 mutants (rgs1-1, rgs2-1) were obtained from ABRC, ft-10 is a GABI-Kat line obtained from NASC and the er-105 line was a kind gift from the Keiko Torii lab (University of Texas, Austin, USA).

Seeds were surface sterilized and plated on 0.5× MS medium 0.7% agar plates supplemented with 1% sucrose. After 3 days of stratification at 4° C, the plates were set vertically for germination and exposed to 8 h light/16 h dark at 20°C for 7-10 days, but not longer. Seedlings were transferred to well-watered soil in trays covered by transparent lids, which were adjusted the following day to allow adequate air transfer and moisture release. Vented transparent lids were used to protect the seedlings from growth chamber air flow for ∼2 weeks, and after the seedlings hardened, the lids were removed. Seedlings that failed to thrive were removed. Likewise, weak or otherwise atypical juvenile or mature plants were excluded before the onset of flowering.

Plants were grown in Conviron walk-in chambers with parameters for SD experiments set to 8 h light (140 µmol photons m⁻² s⁻¹) at 21°C, 50% humidity followed by 16 h dark at 19°C, 60% humidity. LD experiments were performed under the same conditions except the light cycle was set to 16 h light/8 h dark. To avoid potential analytical issues caused by comparing data from multiple plantings with differing phenotypic expressivity, we set up a large-scale experiment to simultaneously assess the phenotypes of nine genotypes with 36 seedlings each (324 seedlings in one initial planting). This was replicated in a second large-scale experiment, bringing the total to 648 plants. To validate that our observed phenomena were not unique to a specific growth facility, we performed the two large-scale experiments in two different Conviron walk-in growth chambers. Chamber 1 (Model #MTPS72) has horizontal air handling and precise humidity control, while Chamber 2 (Model #C813) is over twice the size, has air handling via ceiling vents and experiences variable 85-90% humidity in the dark cycle. Additional experiments and validation were performed in Conviron Model #MTPS144 chambers, as well as in similarly set reach-in chambers (pH Environmental Model #CEC38-15-G) with humidity that typically ranged from 35-70% depending on watering status.

To avoid possible nutrient deficiencies arising from nutrient leaching or nutrient depletion during the long growth period, plants were grown in Miracle-Gro potting mix for the large-scale experiments. Soil moisture was regularly checked and the soil was not allowed to fully dry between waterings. As expected for most plant growth traits (e.g. leaf size, plant height and seed yield), cantil expressivity follows plant health and vigor, but fertilized soil is not necessary to observe cantils as they were also observed in plants grown on 1:1 mixes of Miracle-Gro:MetroMix 360, on MetroMix 360 alone and on Sunshine LC1 mix alone. This is exemplified by the comparison of the ∼1.8 cantils/plant formed by ft-10 grown under SDs and LDs in Miracle-Gro (Fig. 4A), and an additional LD ft-10 control planted in MetroMix 360, which produced 0.95±0.12 cantils/plant (mean±s.e.m., n=36 plants). Growth in miniature pots, or in other conditions that negatively influence plant vigor, can reduce or negate cantil development. As expected for a trait related to flowering, which molecularly initiates prior to visual appearance, inadvertent stress (e.g. drought, flooding or rosette leaf sampling) to juvenile or mature plants can strongly influence cantil frequency, even if the plants appear ‘recovered’ for several weeks.

After full elongation of the primary inflorescence and browning of the first siliques (17-19 weeks after germination for Col-0), the primary inflorescences were removed for inspection under a 1-3.5× dissecting scope. Relevant morphological characteristics were hand-recorded for each assayed plant (short cantils, long cantils, cuffs, spurs, nubs, positions and rippling, etc.) before digital tabulation. At least the first eight or nine pedicels were inspected for each plant, as cuffs were occasionally identified up to nine pedicels away from the M2-M3 junction. Inspection of distal pedicels was also necessary as relatively distal cuffs or very short cantils were observed to occur acropetally to the first 1-2 archetypal pedicels, albeit very rarely. After full inspection of ∼200 long photoperiod grown Col-0 plants, the inspection of LD grown plants was reduced to only the first five pedicels. Likewise, only the first five pedicels were inspected for long photoperiod flowering Ws and Ler. Representative images were taken through the eyepiece tube of a Bausch & Lomb 0.7-3.0× stereozoom microscope using a Canon G5 camera and a Cole Parmer fiber optic illuminator (Model 41500-50). In many cases, co-florescence below the M2-M3 junction were trimmed for image clarity, leaving a blunt end subtended by a single cauline leaf visible at the bottom of the figure panel: see Fig. 1D,K,L; Fig. 2F; Fig. 3B-E; Fig. 4B-C; Fig. S1C,D; Fig. S3B,C,K,M. For comparison, two visible co-florescences subtended by a single cauline leaf are untrimmed in Fig. 1C but trimmed in Fig. 1D. In Fig. 1E, a trimmed co-florescence is visible but the cauline leaf is out of frame.

Recorded data were digitized into Microsoft Excel 2016 and imported into GraphPad Prism 8.0 for statistical analysis. As plant sample sizes were unequal in most cases (e.g. large-scale experiments with 31-36 plants per genotype), statistical analysis was performed using one-way ANOVA with a mixed-effects model and Dunnett's correction for multiple comparisons. Multiple comparison tests were restricted to a single wild-type accession control and its associated mutants (e.g. Col-0, gpa1-3 and gpa1-4). Comparison of ft-10 versus gpa1-3 ft-10 LD cantil frequency was performed using an all-against-all one-way ANOVA using Tukey's correction for multiple comparisons. Unpaired t-tests were employed for two-sample analysis. Graphs were prepared using Prism 8.0.

We thank Dr Dharmendra Singh, previously from Prof. Timothy McNellis's lab (Pennsylvania State University), members of Prof. Nina Federoff's lab (Pennsylvania State University) and others for early independent confirmation of cantil morphogenesis in their own growth facilities using their own standard protocols. We thank Dr Ángel Ferrero-Serrano for assistance with image collection and Dr David Chakravorty for comments on the manuscript. We thank Profs Keiko Torii (er-105) and Phillip Wigge (ft-10) for seeds.