Organ-specific genetic interactions between paralogues of the PXY and ER receptor kinases enforce radial patterning in Arabidopsis vascular tissue

Cell migration is fundamental to the development of animal body plans. By contrast, plant cell walls do not allow cells to migrate, and consequently plant growth and development is entirely a result of differential growth. As such, initiation and elaboration of plant organs occurs via coordinated changes to the orientation and occurrence of cell divisions, and by cell expansion. In Arabidopsis embryos, pattern is established early in development. Twenty-eight-cell embryos have already specified the provascular tissue that consists of four cells the centre of the embryo. A layer of endodermal tissue surrounds the provasculature, and an outer layer of epidermal cells has also been specified (ten Hove et al., 2015). Extra tissue types, cortex and pericycle, are subsequently derived from specific rounds of asymmetric cell division (Kajala et al., 2014). In the hypocotyl, the vascular tissue undergoes a transition from diarch to radial symmetry 6-10 days post-germination. Here, cells adjacent to the xylem divide to generate the vascular cambium (Smetana et al., 2019), such that the tissue pattern along the radial axis becomes epidermis-cortex-endodermis-pericycle-phloem-cambium-xylem. As the hypocotyl further expands, the epidermis and cortex are lost in a process that corresponds with periderm specification and proliferation at around 3 weeks post-germination (Wunderling et al., 2018). Thus, a morphology periderm-phloem-cambium-xylem is generated and maintained through the remainder of life of the plant (Chaffey et al., 2002).

Coordination of tissue expansion must occur as the hypocotyl transitions from diarch to radial symmetry, and as organs increase in size. This coordination must incorporate cell division, because cell numbers increase from tens to hundreds to thousands of cells. It must also incorporate cell size, which differs according to cellular function in differentiated cells. Consequently, the Arabidopsis hypocotyl represents an interesting model for studying how patterns are maintained through very large increases in plant size, a mechanism that is poorly understood. This organisation contrasts with that of the Arabidopsis inflorescence stem, where radial growth is much more limited. Here, radial pattern is defined in the rib zone below the shoot apex, rather than built upon a pre-existing embryonic pattern. The tissue types along the radial axis of the stem also differ. Epidermis, cortex and endodermis are present outside the phloem, procambium and xylem (Fig. 1A). Pith constitutes the cell type at the centre of the stem.

Evidence points to the presence of mechanisms that coordinate the order of tissue layers. In the Arabidopsis root, removal of the root tip results in a reorganisation of the organ to enable the formation of a new meristem. Strikingly, stable patterning of tissue layers is established in the reorganised tissue separately from the activity of the stem cell niche. This suggests that tissue layer organisation is independent of stem cell growth (Efroni et al., 2016). Non-cell autonomous signalling represents one mechanism through which tissue layer organisation could be coordinated. A ligand secreted by one tissue could provide positional information to a receptor located in an adjacent cell type. Ligand-receptor pairs that signal between tissue layers and are required for tissue layer organisation have been described. TRACHEARY ELEMENT DIFFERENTIATION INHIBITORY FACTOR (TDIF) is a ligand that is encoded for by three genes, CLAVATA3-LIKE/ESR-RELATED 41 (CLE41), CLE42 and CLE44. TDIF is excreted from the phloem and perceived by a receptor kinase, PHLOEM INTERCALATED WITH XYLEM (PXY), which is expressed in the cambium. Loss of TDIF-PXY results in a failure to correctly organise tissue layers in the vascular tissue. pxy mutants are characterised by disruption to the spatial separation of xylem, cambium and phloem. Reductions in cell division in the cambium and premature xylem differentiation are also a consequence of loss of PXY (Etchells and Turner, 2010; Fisher and Turner, 2007; Han et al., 2018; Hirakawa et al., 2010; Hirakawa et al., 2008; Ito et al., 2006; Kondo et al., 2014; Suer et al., 2011). TDIF-PXY signalling thus represents a mechanism through which differential growth in vascular tissue could be coordinated, regulating as it does, tissue organisation, cell division and differentiation.

TDIF-PXY genetically interacts with a second ligand-receptor pair to maintain the spatial separation of vascular tissues. In stems, the ERECTA (ER) receptor is expressed in the phloem, and its cognate ligands, CHALLAH-LIKE 2/EPIDERMAL PATTERNING FACTOR-LIKE 4 (CLL2/EPFL4) and CHALLAH (CHAL/EPFL6) are expressed in the endodermis (Abrash et al., 2011; Uchida et al., 2012). pxy er mutant stems show organisation defects greater than those of pxy single mutants (Etchells et al., 2013). Thus, in stems, the genetic interaction between EPFL-ER and TDIF-PXY represents a non-cell autonomous signalling system that organises tissue layers between endodermis, phloem, cambium and xylem. In hypocotyls, changes to the organisation of vascular tissues in er pxy hypocotyls are also apparent (Etchells et al., 2013). However, ER expression is reported to be much broader, being present in phloem, cambium and xylem parenchyma (Ikematsu et al., 2017). The spatial expression domains of CHAL and CLL genes have not been described in hypocotyls.

In the Arabidopsis genome, paralogues of both PXY and ER are present. The PXY family, hereafter referred to as PXf, comprises PXY, PXY-LIKE1 (PXL1) and PXY-LIKE2 (PXL2). TDIF is reported to bind the ligand binding pocket of PXL1 and PXL2 (Zhang et al., 2016), and pxl1 and pxl2 enhance the vascular organisation defects that are characteristic of pxy mutants (Etchells et al., 2013; Fisher and Turner, 2007). The ER paralogues are ER-LIKE1 (ERL1) and ERL2 (Shpak et al., 2004). The ERECTA family (ERf) have wide-ranging roles in regulation of plant growth and development. Redundantly, these three genes function in cell elongation, cell division, inflorescence architecture (Shpak et al., 2004; Torii et al., 1996), floral patterning (Bemis et al., 2013), shoot apical meristem fate (Kimura et al., 2018; Uchida et al., 2013) and stomatal spacing (Shpak et al., 2005). In the context of plant vascular development, they promote vascular expansion in the stem (Uchida and Tasaka, 2013). By contrast, in hypocotyls they repress radial expansion and also control the timing of xylem fibre formation (Ikematsu et al., 2017; Ragni et al., 2011). A hallmark of loss of ERf genes is an increase in cell size, particularly with respect to the radial axis (Shpak et al., 2004; Shpak et al., 2003).

In this article, we have investigated the genetic relationships between PXf and ERf receptors. We generated pxy pxl1 pxl2 er erl1 erl2 sextuple mutants using a combination of classical genetics and genome editing. In hypocotyls, the sextuple mutant failed to make the transition to secondary growth. Further analysis of these lines demonstrated that PXf and ERf genetically interact to coordinate tissue integrity at the levels of both cell size and cell division. Gene expression analysis in stems and hypocotyls suggested that members of PXY and ER gene families regulated expression of paralogues both within and between these families. However, this regulation was distinct in hypocotyls and stems. In stems, PXf and ER also influenced the expression of non-vascular-expressed EPFL4 and EPFL6. This suggests that coordination of growth regulators occurs between vascular and non-vascular tissue layers. Our results demonstrate that although interactions between members of both families are crucial in both stem and hypocotyl, the paralogues have specialised functions within vascular tissue of differing ontogenies.

In Arabidopsis stems and hypocotyls, tissue is arranged in concentric rings with the vasculature at the centre (Fig. 1A). PXY and ER genetically interact to control vascular development. In stems, ER ligands, CHAL and CLL2 (Abrash et al., 2011), are expressed in the endodermis whereas ER is expressed in the phloem (Uchida et al., 2012). TDIF-encoding genes are expressed in the phloem, and TDIF signals to PXY, which is expressed in the procambium (Etchells and Turner, 2010; Fisher and Turner, 2007; Hirakawa et al., 2008). In mature hypocotyls, endodermis is not present and the CLL2 and CHAL domains of expression are not known. To better understand the spatial relationships between the PXY and ER receptors and ER ligands in hypocotyls, we determined the CHAL and CLL2 expression pattern in 5-week-old plants using transcriptional reporters (Abrash et al., 2011). Both CHAL::GUS and CLL2::GUS lines demonstrated clear expression maxima both in xylem parenchyma and in the differentiating xylem adjacent to the cambium (Fig. 1B,C). Expression in the cambium itself was minimal. Thus, active ER ligand-receptor complexes occur in different locations in stems compared with hypocotyls. In stem tissue, active ER ligand-receptor complexes would be in the phloem, whereas in hypocotyls they must predominate in differentiating xylem.

To better understand the influence that ER might have upon PXY signalling, we tested whether expression levels of genes involved in PXY signalling differed in er mutants. We have previously shown that TDIF-encoding CLE41, CLE42 and CLE44 levels are unchanged in er (Etchells et al., 2013), so we analysed expression of the PXf family of receptors. qRT-PCR was used to test levels of PXf gene expression in stems and hypocotyls of 5-week-old wild-type and er plants (Fig. 1D-I). In hypocotyls, the level of PXf gene expression was unchanged in er mutants compared with wild type (Fig. 1G). By contrast, PXL1 and PXL2 expression, but not that of PXY was found to be elevated in er mutant stems (Fig. 1D). These observations suggest that ER signalling may regulate vascular development by setting PXL1 and PXL2 levels in the stem. They also underline that there are differences in regulatory relationships between patterning genes in stem and hypocotyl.

We sought to further investigate the role of PXL1 and PXL in vascular development. In transverse sections, pxl1 pxl2 double mutants were indistinguishable from wild type (Fig. S1); however, we have previously shown that pxl1 and pxl2 enhance the pxy phenotype (Etchells et al., 2013; Fisher and Turner, 2007) (Figs 2A,B and 3D). Thus, as PXL gene expression was observed to be elevated in er stems, but neither er (Fig. 1E-F) nor pxl1 pxl2 (Fig. S1) lines had vascular stem phenotypes except in a pxy mutant background, we addressed the function of PXL1 and PXL2 regulation by ER in the absence of pxy. er pxf quadruple mutants (er pxy pxl1 pxl2) were generated and compared with wild-type, pxy, er, er pxy and pxf lines. In inflorescence stems, er pxf lines had considerably fewer cells per vascular bundle than either pxf, er or pxy er counterparts (Fig. 2A; Tables S1 and S2). Therefore PXL1 and PXL2 do function redundantly with ER to regulate vascular proliferation in the stem, at least in the absence of PXY. In hypocotyls, a reduction in radial growth was observed in pxf er lines relative to pxf and er; however, pxf er and pxy er lines were indistinguishable (Fig. 3; Tables S1 and S3). Thus, pxl1 and pxl2 do not enhance pxy er hypocotyl phenotypes, a result consistent with our observation that PXL1 and PXL2 expression was unchanged in er mutant hypocotyls (Fig. 1G).

Although changes to vascular proliferation were apparent in er pxf inflorescence stems, by far the most dramatic defect was observed when the vascular bundle shape was assessed (Fig. 2B-F). In wild-type Arabidopsis stems, the distribution of vascular bundles is such that there is a greater distribution of vascular tissue along radial axis of the stem than along the tangential (Fig. 2C). We found the tangential:radial length ratio of wild-type vascular bundles to be 0.61. In pxf and pxy er lines, this ratio was 0.91 and 1.36, respectively. In er pxf stems, a dramatic redistribution of vascular cell types had occurred along the tangential axis (Fig. 2F), such that the ratio of tangential:radial length of vascular tissue was 2.30 (Fig. 2B; Table S1). In some plant stems, this led to an almost complete ring of vascular tissue, with phloem cells scattered around the circumference of the vascular cylinder (arrows in Fig. 2F), rather than present in discrete vascular bundles (Fig. 2C,D). Thus, PXL1 and PXL2 are crucial for regulating radial pattern in the stem, particularly in the absence of ER and PXY, and these data support the idea that ER and PXf constitute a mechanism for organising vascular cell layers.

Having observed that PXf genes were differentially expressed in er mutants (Fig. 1D), and that PXL1 and PXL2 contribute to the control of stem radial pattern (Fig. 2), we also sought to determine whether expression of members of the ER gene family were changed in response to perturbation of PXf or ER genes. In stems and hypocotyls of pxf lines, ER expression did not differ from wild-type levels, as determined by qRT-PCR. Expression levels of ERL1 and ERL2 were also indistinguishable in wild-type, er and pxf stems (Fig. 4A-C). By contrast, ERL1 expression was significantly reduced when er pxf lines were compared with er single mutants. Thus ERL1 expression in er mutants is maintained by the PXf in stems (Fig. 4A). Expression levels of the ER ligands that function in the stem, CHAL and CLL2, were also tested in this experiment, as was that of CLL1, which genetically interacts CHAL and CLL2 (Abrash et al., 2011; Uchida et al., 2012; Uchida and Tasaka, 2013). Inflorescence stem expression of CHAL and CLL2, but not that of CLL1, demonstrated significant reductions in expression in er pxf lines when compared with er (Fig. 4D-F). Thus, PXf and ER genetically interact to maintain EPFL ligand expression in stems in addition to that of their cognate receptor, ERL1 (Fig. 4A).

In hypocotyls, ERL1 acts redundantly with ER to repress hypocotyl growth and control the timing of xylem fibre differentiation (Ikematsu et al., 2017). ERL2 has not been assigned a function in hypocotyl development as its expression has been reported as absent from hypocotyls in 9-day-old seedlings and 3-week-old plants (Ikematsu et al., 2017; Uchida et al., 2013). To understand how PXY and ER might influence ERf expression, ERf:GUS reporter constructs (Shpak et al., 2004) were crossed into pxy and er mutants. To our surprise, in 5-week-old plants, we did detect ERL2::GUS reporter expression in the hypocotyls of wild type, which, at this growth stage, demonstrated a very similar pattern to that observed for ERL1 and ER (Fig. 5A,D,G). Thus, ERL2 expression is a feature of late hypocotyl development. ER, ERL1 and ERL2 expression was present in most hypocotyl cell types, with two maxima; the first in the cambium and xylem initials, and the second in the periderm (Fig. 5A,D,G; arrowheads). No change in the pattern of ERL1 or ERL2::GUS expression was observed in er mutants (Fig. 5C,F). However, the clearly defined expression maxima that were observed in ER::GUS, ERL1::GUS and ERL2::GUS lines in both wild type and er mutants, lacked definition in the absence of PXY (Fig. 5B,E,H). Here, for all three reporters, expression was observed to be more even across the hypocotyl, thus PXY signalling is either required to define ERf expression maxima, or cell types in which ERf are expressed are found throughout the hypocotyl in pxy mutants. The latter seems unlikely as there are fewer vascular cells in pxy mutants.

Having defined the pattern of ERf expression in a subset of genotypes, we sought to address changes to levels of ERf expression using qRT-PCR (Fig. 5I-K). In common with our observation in the stem (Fig. 4), hypocotyl ERL1 and ERL2 expression levels did not differ between wild-type, er and pxf lines (Fig. 5J-K). By contrast, a striking increase in ERL1 and ERL2 gene expression was observed in pxf er hypocotyls relative to all other genotypes tested (Fig. 5J,K). As such, opposite regulation of ERL1 and ERL2 by ER and PXf genes occurred in the hypocotyls (Fig. 4J-K) and stem (Fig. 2A-B). This highlights a difference in the nature of the PXf-ER genetic interactions in stems and hypocotyls. In hypocotyls, no changes were observed in levels of CHAL, CLL1 and CLL2 expression levels in er, pxf or er pxf lines (Fig. S2).

The PXf promotes radial growth in hypocotyls (Etchells et al., 2013; Fisher and Turner, 2007; Hirakawa et al., 2008) (Fig. 3D; Tables S1 and S3), whereas ER and ERL1 signalling represses it (Ikematsu et al., 2017). Thus, our gene expression data demonstrating that PXf plays a part in repression of ERL gene expression in hypocotyls (Fig. 5J,K) are consistent with existing phenotypic data because the PXf might be expected to repress expression of negative regulators of hypocotyl radial growth. In addition to repressing radial growth, ER and ERL1 have also been described as preventing premature fibre formation, as er erl1 hypocotyls develop fibre cells in the location where parenchyma are present in wild type. ERL2 was thought not to function in the hypocotyl given its very low expression levels in the early stages of development (Ikematsu et al., 2017). Because we found ERL2 to be expressed in hypocotyls at 5 weeks (Fig. 5D,K), we tested whether ERL2 functioned similarly to ERL1 by analysing er erl2 lines. Neither change to fibre formation, nor to hypocotyl radial growth were observed (Fig. S3); thus, in contrast to ERL1 (Ikematsu et al., 2017), a function for ERL2 in hypocotyl development is not apparent in a double mutant background with er.

To address the function of the elevated ERL gene expression that we observed in pxf er hypocotyls (Fig. 5J,K), we removed ERL gene function from this genotype by generating pxf er erl1, pxf er erl2 and pxf erf quintuple and sextuple mutants. PXY and ERL1 are tightly linked on chromosome 5, separated by just 270 kb, so to overcome this linkage we employed a CRISPR/cas9 construct that contained two guide RNAs against ERL1 (Fig. S4). Thus, pxf er erl1 and pxf erf plants were generated by genome editing. Secondary growth in these lines and controls was determined by measuring the hypocotyl radius in 6-week-old plants (Fig. 6A, Fig. S5C). Radii of pxf er and pxf er erl1 lines did not show a significant difference. By contrast, radii of pxf er erl2 and pxf erf hypocotyls were significantly smaller than those of pxf er erl1 plants (Fig. 6A). Thus, ERL1 and ERL2 expression is required in pxf er hypocotyls to maintain hypocotyl growth rates; however, pairwise comparisons suggested that ERL2 played a greater role than ERL1 in this respect, as pxf er erl1 hypocotyls were larger than those of pxf er erl2 lines.

During vascular cylinder development in the embryo, the hypocotyl forms in a diarch pattern with a row of xylem cells that are flanked by two phloem poles (Dolan et al., 1993). As secondary growth proceeds, this organisation develops radial symmetry with phloem present around the circumference of the vascular cylinder (Chaffey et al., 2002). We analysed hypocotyl morphology in 5-week-old plants. Strikingly, development was perturbed to such a degree in pxf erf mutants that the position of the original phloem poles remained apparent (arrows in Fig. 6F; see Fig. S5 for higher magnification). This demonstrates that vascular development was retarded to such a degree that these plants could not make the transformation to true radial growth. Such phenotypes were not observed in pxf, erf or pxf er erl2 lines (Fig. 6B-E).

Next, we looked to identify recent cell divisions in our mutant hypocotyls by analysing thin sections. In wild-type and erf lines, cell divisions were always oriented perpendicular to the hypocotyl radial axis (Fig. 6B,C, arrowheads). This aspect of normal vascular development known to perturbed in lines that lack pxy and its paralogues (Fig. 6D) (Fisher and Turner, 2007). Recent cell divisions were clearly identifiable in the absence of the PXf, ER and ERL2, and they remained present, albeit lacking orientation and at a much reduced frequency in pxf erf lines (Fig. 6E-F). Thus, although not an absolute necessity for formation of either phloem or xylem vessels, these receptor-kinase families are absolutely essential in specifying their positioning and in coordinating cell division in a manner that allows organised radial expansion and pattern maintenance (Fig. 6).

One common characteristic of mutants with reduced cell division is an increase in cell size, relative to wild-type plants. This compensates for fewer cells, such that final organ size is often similar to that of wild-type plants (Horiguchi and Tsukaya, 2011). In the course of our hypocotyl analysis, cell sizes and shapes appeared to differ among our mutant lines, and, in particular, cells in pxf lines appeared larger than those of other lines (Fig. 3A,B). Consequently, cell morphology was calculated from images of anatomical sections by selecting cell representatives from the different genotypes and using a MATLAB code to analyse the cells as connected components with measurable features (Fig. S6A,B). Cell area and perimeter were investigated for xylem vessels, fibres, xylem parenchyma and phloem cells in wild-type, pxf, pxf er erl1, pxf er erl2, and pxf erf lines (Fig. 7) with one exception. Fibre morphology could not be assessed in pxf erf, as insufficient fibre cells were present (Fig. 6F). In hypocotyls, all pxf cell types tested demonstrated increases in cell perimeter relative to wild type (Fig. 7; Table S4). pxf er and pxf cells demonstrated no statistically significant differences in vessel, fibre and phloem cell perimeters, but pxf er xylem parenchyma perimeters were smaller than those of pxf lines. Strikingly, removal of further members of the ERf restored vessel, parenchyma and phloem cell perimeters to wild-type sizes (Fig. 7A,B,D; Table S4). Thus members of the ERf are required to promote increases in cell size in the absence of PXf.

The one cell type that was the exception to this cell size regulation was xylem fibres. Here, the increase in fibre perimeter that was characteristic of pxf mutant hypocotyl cells was not rescued by erf mutants (Fig. 7C; Table S4). These observations were supported by cell area measurements. For the four cell types tested, pxf cell areas were larger than those of wild-type plants but, with the exception of fibres, removal of er erl1 or er erl2 from pxf restored cell areas to those observed in wild type (Fig. 7; Table S5).

Xylem cells are characterised by rigid secondary cell walls, so we hypothesised that parenchyma may be subject to changes in cell shape to accommodate the increased xylem cell size. To test this hypothesis, we calculated the ellipticity of the parenchyma and other hypocotyl cell types by determining their major to minor axis ratios in wild-type, pxF and pxF er erl2 lines. However, this parameter varied little between genotypes (Fig. S6C-F).

To complete our analysis of pxf erf sextuple mutant morphology, we examined vascular tissue in inflorescence stems. Inflorescence stem vascular morphology was similar in pxf erf lines and pxf er erl2 counterparts (Fig. 8). Both were characterised by very large reductions in vascular bundle size. Characteristic xylem and phloem cell types were present, but only very small xylem vessels were observed, relative to those found in wild-type, erf and pxf lines (Figs 8D,E and 9A; Tables S6 and S7). Furthermore, tissue layer organisation defects were apparent beyond those previously observed. In particular, the clearly defined organisation of endodermal and adjacent phloem cap cells were lacking, with the phloem cap appearing to extend into the cortex (Fig. 8D) or be absent altogether (Fig. 8E). Thus, tissue layer defects occurred outwith vascular cell types. These similarities in vascular morphology were independent of plant size because gross morphology of pxf erf sextuple lines was considerably smaller than pxf er erl2 counterparts (Fig. S7).

Having observed large reductions in xylem vessel size in stems (Fig. 9A), we tested whether PXf and ERf genes genetically interacted to control cellular morphology of other vascular cell types in the stem (Fig. 9B,C). In stems, xylem vessels and cells in the phloem were smaller in pxf lines than in wild type, as determined by measuring both cell perimeter and cell area, and in contrast to measurements in the hypocotyl. Removing ER from pxf lines resulted in no change to the size of these cells, but loss of ERL2 from pxf er plants caused a further reduction in cell size (Fig. 9A,C; Tables S6 and S7). Thus, in phloem and xylem vessels, pxf and erf families interact to maintain cell size. Xylem fibre sizes differed from this trend. Here, pxf er cells were significantly larger than wild type, but this phenotype was suppressed in pxf er erl2 plants as fibre perimeter and area was unchanged from wild type (Fig. 9B; Tables S6 and S7). We were unable to assess fibre morphology in pxf erf vascular bundles, as too few were identifiable in these lines (Fig. 8E). Taken together, our results demonstrate that a genetic interaction between PXf and ERf signalling coordinates organ size at the level of cell size, in addition to coordination of proliferation and pattern maintenance in both stems and hypocotyls.

Plant growth and development require coordination between expanding tissue layers, particularly where tissue types are organised in concentric rings. Clearly, expansion of inner layers must be coordinated with expansion of outer layers. How does coordination between tissue layers occur? It was proposed some time ago that the ERf could perform this function (Shpak et al., 2004), and this initial suggestion has subsequently been supported by observations that, in the inflorescence stem, endodermis derived EPFL ligands signal to ER in the phloem to regulate cell division in the adjacent procambium (Uchida et al., 2012; Uchida and Tasaka, 2013) (Fig. 10A). Our observation that PXL expression is higher in the stem of er mutants (Fig. 1D) suggests that these endodermis-derived signals could act through ER to attenuate PXf-regulated vascular expansion (Fig. 10A). The alternative conclusion would be that PXL expression is higher in er mutants due to a change in stem morphology, but we regard this as unlikely for two reasons. First, there are negligible differences in vascular proliferation and organisation in er stem vascular tissue compared with wild type that could account for such changes in gene expression (Figs. 1E,F and 2A,B). Second, there is clear evidence that pxl1 and pxl2 genetically interact with er. This interaction is apparent in a pxy mutant background, as pxf er lines demonstrated fewer cells in stem vascular bundles than either pxy er or pxf lines (Fig. 2; Table S2).

Our experimentation with pxf er lines led to observations that PXf receptors, redundantly with ER, are required for normal expression levels of ERL receptors and their EPFL ligands in the stem (Fig. 3). As CLL2 and CHAL are endodermis expressed, changes in the expression levels of these genes could be due to coordination of vascular tissue expansion in stems across multiple tissue layers via a series of feedback loops (Fig. 10). As the endodermal stem layer remains clearly defined in er pxf lines, it is unlikely that the reduction in CHAL/ and CLL2 expression in these lines is due to the disruption of endodermal cell fate (Fig. 2F). However, owing to severe disruptions to vascular morphology adjacent to the endodermis, we cannot rule out that such changes are a consequence of the disruption to xylem, phloem and procambium organisation. Disruption to pxf er quadruple mutants was severe to such a degree that in stems, vascular tissue was no longer found in discrete bundles, but scattered around the stem adjacent to the endodermis (Fig. 2).

Oriented cell divisions and the development of organ boundaries in the rib zone of the shoot apical meristem, from which stem vascular tissue is derived, have been reported to be regulated by a homeodomain transcription factor, REPLUMLESS (RPL). Pertinent to the results obtained here, RPL was found to occupy the promoters of PXY, CLE41, CLE42, ER, ERL1, ERL2 and CHAL in ChIP-Seq experiments (Bencivenga et al., 2016). RPL is localised to the cytoplasm unless present in a heterodimer with class I KNOX protein, such as BREVIPEDICELLUS (Bhatt et al., 2004). rpl bp double mutants, particularly those in the Ler background that lacks a functional copy of ER, demonstrate considerable defects in vascular development (Etchells et al., 2012; Smith and Hake, 2003). Thus, events in the rib zone that are controlled by RPL could set up the initial pattern in the stem. Our genetic analysis demonstrates that however the pattern is initiated, it is maintained by interacting signalling pathways characterised by members of the ERECTA and PXY families.

Evidence that mechanisms exist to adjust cell morphology in order to maintain tissue size and organisation include the observation that cell expansion differs according to the rate of cell division. Here, overall organ size in mutants with fewer cells is often comparable to or only subtly different from those of wild-type plants due to an increase in cell size (De Veylder et al., 2002; Hemerly et al., 1999; Shpak et al., 2004; Ullah et al., 2001). Furthermore, such mechanisms can act non-cell autonomously. Expression of KRP1 reduces cell division (Hemerly et al., 1995). When it is specifically expressed in the epidermal cell layer, concomitant changes to palisade cell size and density also occur (Lehmeier et al., 2017). Thus, where the cell cycle has been manipulated in one cell layer, influence on cell size and organisation occurs in adjacent tissues, contributing to tissue integrity. We found that the interaction between PXf and ERf was crucial to regulation of cell size in multiple cell types. The ability to adjust cell size to compensate for the profound reductions in cell division in pxf er lines was particularly dependent on ERL2 (Figs 7 and 9). This is in contrast to the consequences of losing the ERECTA family alone, as cell size adjustments are a feature of erf mutants (Shpak et al., 2004). However, the influence of ERL2, ER and ERL1 differed by cell type and organ. In hypocotyls, vascular cells were larger in either pxf or pxf er lines compared with wild type. In hypocotyl xylem vessels, parenchyma and phloem cells, this increase in size was dependent on ERL gene expression as increases in cell size were lost in pxf er erl1 and pxf er erl2 lines (Fig. 7A,B,D). In stem vascular bundles, the only cell type with an increase in size in response to fewer to cell divisions were the fibres. This phenotype was also suppressed by removal of ERL genes. These observations support the idea that one function of the genetic interaction between ERf and PXf is coordination of tissue expansion. We propose that with these signalling mechanisms removed, the positional information that must be interpreted for cell morphology adjustments to occur is missing.

In stems, the receptor kinases that are the focus of this study are expressed in discrete domains. By contrast, in hypocotyls, expression patterns of ER and PXY overlap on the xylem side of the cambium (Hirakawa et al., 2008; Ikematsu et al., 2017; Shi et al., 2019; Smetana et al., 2019). As the domain of ERL gene expression is expanded in pxy mutants (Fig. 5B,E,H), the presence of PXL receptors in cells that also express ERf proteins is increasingly likely. A direct interaction between members of these receptor families is therefore possible. A recent in vitro global analysis of receptor kinase interactions did not include direct interactions between ERf and PXf family members because putative interactions did not pass cut-offs for inclusion in the high confidence bidirectional dataset (Smakowska-Luzan et al., 2018). Nevertheless strong in vitro interactions between ER and PXY, and ER and PXL1 proteins were observed in one orientation, and between PXL2 and ERL2 in both directions [Youssef Belkhadir, personal communication; data available in BAR ePlant (Waese et al., 2017)]. The determination of whether these interactions are genuine and, if so, the circumstances under which they occur in vivo will be an important focus for future research. ERf activity in the epidermis has previously been reported to be buffered by a second receptor, TOO MANY MOUTHS (TMM). Loss of this buffering in tmm mutants leads to opposite stomatal spacing phenotypes in spatially separate cotyledon, where stomata cluster, compared with hypocotyls where stomata are absent. Differing ligand availability in cotyledon and hypocotyl is thought to account for this difference (Abrash et al., 2011). CHAL and CLL2 have been demonstrated to act as ERf ligands in the inner tissues of stems (Uchida et al., 2012). We have shown that these ligands are expressed in developing xylem in hypocotyls (Fig. 1B,C). Thus, in stem vascular tissue, active ligand-ER complexes most likely reside in the phloem, whereas in hypocotyls they would be predominant in xylem initials. It remains to be determined whether the difference in ERL gene regulation by ER and PXf in stem and hypocotyl could be due to differing complements of co-receptors and ligands in these differing locations.

Our observation that ERL gene expression is de-repressed in the absence of PXf and ER in hypocotyls (Fig. 5J-K) supports the idea that these components genetically interact. Perhaps the most striking of our findings was the observation that ER and PXf regulation of ERL gene expression in the hypocotyl occurred in a manner opposite to that observed in the stem (compare Figs 4A,B and 5J,K), where ER and PXf combine to repress ERL gene expression. Thus, while PXf and ERf are required non-cell autonomously for tissue organisation and expansion in both stems and hypocotyls, the regulatory networks through which development is controlled in these two organs differs (Fig. 10). One explanation for differences in regulation is that tissue layer organisation varies by location. In hypocotyls, cambium division must occur concomitantly with factors that control periderm division. By contrast, in stems there is no such continually expanding tissue outside the vascular tissue, so vascular proliferation in stems must be under much tighter regulation.

Factors controlling the transition to secondary growth in Arabidopsis hypocotyls have recently been described. It first arises in cells adjacent to xylem, and central to this transition was an accumulation of auxin and expression of HD-ZipIII transcription factors. These factors, in turn, activate expression of PXY signalling (Smetana et al., 2019). Nevertheless, pxy mutants, and indeed pxf triple mutants, do ultimately make the transition to secondary growth (Figs 3 and 6). Thus, other factors must act with PXY to regulate the transition secondary growth and radial pattern in hypocotyls. pxy er double mutants, erf triple mutants, pxf er quads, and both pxf er erl1 and pxf er erl2 quintuple lines all made the transition to full secondary growth (Figs 4 and 6). By contrast, pxf erf sextuple mutants did not. As such, these lines demonstrated a phenotype that, as far as we are aware, has never previously been described. The observation of this novel phenotype further supports the idea that these receptor families coordinate development through a genetic interaction, and that the phenotypes cannot be explained simply by a correlative loss of cell division-promoting factors. Thus, PXf and ERf signalling act redundantly to regulate radial growth transition; consequently, complete removal of PXf and ERf families results not only in prominent proliferation defects, but also in dramatic defects to patterning (Fig. 6).

In Arabidopsis, stem and hypocotyl differ in that the hypocotyl undergoes radial growth, but the vast majority of the stem does not. Radial hypocotyl growth is largely the consequence of expansion of a pattern that is laid down in the embryo, but in stems, de novo patterning must occur below the shoot apical meristem. Nevertheless, in both stem and hypocotyl, the xylem, (pro)cambium and phloem must be specified in adjacent tissue layers in a coordinated manner. Our mutant analysis demonstrates that interactions between PXf and ERf are central to maintaining this organisation by regulating cell division (Figs 6 and 8) and coordinating cell size (Figs 7 and 9) in these different contexts.

AGI accession numbers for the genes studies in this manuscript are as follows: At3g24770 (CLE41), At5g61480 (PXY), At1g08590 (PXL1), At4g28650 (PXL2), At2g26330 (ER), At5g62230 (ERL1), At5g07180 (ERL2), At4g14723 (CLL2/EPFL4), At3g22820 (CLL1/EPFL5) and At2g30370 (CHAL/EPFL6).

For qRT-PCR, RNA was isolated using Trizol reagent (Life Technologies) prior to DNAse treatment with RQ1 (Promega). cDNA synthesis was performed using Tetro reverse transcriptase (Bioline). All samples were measured in technical triplicates on biological triplicates. qPCR reactions were performed using qPCRBIO SyGreen Mix (PCR Biosystems) using a CFX connect real-time system (Bio-Rad) with the standard sybr green detection programme. A melting curve was produced at the end of every experiment to ensure that only single products were formed. Gene expression was determined using a version of the comparative threshold cycle (Ct) method using average amplification efficiencies of each target, as determined using LinReg PCR software (Ramakers et al., 2003). Samples were normalised to 18S rRNA or ACT2. Primers for qRT-PCR are described in Table S1. Significant differences in gene expression were identified with ANOVA and an LSD post-hoc test for four-way comparisons or using Student's t-test for two-way comparisons.

Previously described parental lines pxy-3 pxl1-1 pxl2-1 (referred to hereafter as pxf) and pxy-5 er-124 (Etchells et al., 2013) were crossed to generate pxy-3 pxl1-1 pxl2-1 er-124 (er pxf). The quadruple mutants were selected in F3 by PCR using primers listed in Table S8. To generate pxf er erl2 quintuple mutants, parental lines er-105 erl1-2/+ erl2-1 (erf) (Shpak et al., 2004) and pxy-3 pxl1-1 pxl2-1 (Etchells et al., 2013) were crossed. Plants homozygous for er were selected by visual phenotype in the F2, which was also sprayed with glufosinate to select for plants carrying an erl2-1 allele. Families homozygous for glufosinate resistance in the F3 were screened for pxy-3, pxl1-1 and pxl2-1 to generate pxf er erl2. er and erl2 mutants were subsequently confirmed by PCR.

erl1 genome edited lines were generated using an egg cell-specific CRISPR/Cas9 construct (Wang et al., 2015; Xing et al., 2014). Briefly, target sequences TCCAATTGCAGAGACTTGCAAGG and TCTTGCTGGCAATCATCTAACGG were identified using the CRISPR-PLANT website (Xie et al., 2014) and tested for off-targets (Bae et al., 2014). Primers incorporating the target sequences (Table S8) were used in a PCR reaction with plasmid pCBC-DT1T2 as a template to generate a PCR product incorporating a guide RNA against ERL1. A golden gate reaction was used to incorporate the purified PCR product into pHEE2E-TRI. The resultant ERL1 CRISPR/cas9 clone was transferred to Arabidopsis by floral dip (Clough and Bent, 1998). erl1^GE mutants were selected in the T1 generation by sequencing PCR products generated from primers specific to ERL1 genomic DNA that flanked the guide RNA target sites.

For spatial expression of ERf genes in pxy or er, previously described ER::GUS, ERL1::GUS and ERL2::GUS reporters were used (Shpak et al., 2004). These were crossed to pxy-3 or er-124. pxy mutants were selected in the F2 using primers described Table S8. Reporter lines were picked that also demonstrated GUS expression as judged by GUS histochemical staining, and the presence of GUS reporter construct was subsequently confirmed by PCR using primers (Table S8). For determination of ER-ligand expression, previously described CHAL::GUS and CLL2::GUS lines were used (Abrash et al., 2011).

Vascular morphology was assessed using tissue embedded in JB4 resin. For vascular bundles, inflorescence stem tissue from 0.5 cm above the rosette was assessed. Tissue was fixed in FAA, dehydrated in ethanol and infiltrated with JB4 infiltration medium, prior to embedding. Sections (4 µm) taken using a Thermo Fisher Scientific Finesse ME 240 microtome were stained in 0.02% aqueous Toluidine Blue and mounted with histomount.

GUS-stained tissue was harvested to ice-cold phosphate buffer. Samples were treated with ice-cold acetone for 5 min and then returned to phosphate buffer. GUS staining buffer (50 mM phosphate buffer, 0.2% triton, 2 mM potassium ferrocyanide, 2 mM potassium ferricyanide and 2 mM X-Gluc) was added and samples were infiltrated using a vacuum, before incubation overnight at 37⁰C. Samples were progressively incubated in: FAA, then 70%, 85% and 95% ethanol for 30 min each prior to embedding in Technovit 7100 according to the manufacturer's instructions. Embedded samples were allowed to polymerize at room temperature for 2 h and at 37°C overnight. The inhibition layer was removed by wiping with a lint-free cloth. Samples were sectioned, counterstained with 0.1% Neutral Red and mounted using histomount.

To capture measurements for the cell perimeters and areas, images from six different individuals were selected for each genotype tested. A minimum of 10 cells of each cell type (xylem vessels, xylem fibres, phloem and parenchyma) were selected from a wedge with a 60° central angle from each image (Fig. S6A). Cells of each type were selected along the full length of the radial axis to ensure that cells of all sizes and phenotypic variation were represented. A MATLAB code (available on request) was generated to extract the intrinsic properties of each cell type. To that end, the code was designed to split each image into binary sub-images, wherein the interior of the cell type of interest was represented as white objects on black background (Fig. S6B). The cells (the white objects) from each image were then analysed as connected components of the image and their area and perimeter extracted. To remove noise, i.e. data obtained from objects that were wrongly classified as connected components within the algorithm (e.g. stray pixels), the code was devised to discard data that yielded unrealistically small values for perimeter and area (perimeter value of 0 µm, area smaller than 1 µm²). The data were converted from pixels to μm using a calibration factor, in order to yield results consistent with laboratory observations. For each cell type, an equal number of cells was selected on a random basis from each plant within each genotype to avoid small variations between the number of representatives obtained from each individual plant.

To test the significance of the variation between the cell areas and perimeters between the different genotypes, a nested ANOVA was performed in R at 5% significance level. To perform the nested ANOVA, the data were classified according to genotype (treatment) and plant ID (plants within that treatment), with the response variable either the area or perimeter. A post-hoc Tukey HSD test was performed to determine the significance of the pairwise differences between the means of the areas/perimeters between the different genotypes. To determine the reliability of the results, the residuals from the data were tested for normality. Histograms and quantile-quantile plots for the residuals of each cell type were used to judge the distribution, followed by a Shapiro-Wilk normality test. The residuals for the data for all cell types withstood the Shapiro-Wilk normality test at 5% significance level, confirming that the results of the ANOVA analysis were reliable.

Mean hypocotyl diameters were measured using callipers. The radius was calculated from hypocotyl images of six plants from each genotype. A MATLAB code was used to measure the length of the shorter radius. The length of the radii in pixels was subsequently converted to μm. A Lilliefors test at 5% significance level was used to confirm that the radii for each genotype were normally distributed. A one-way ANOVA followed by a post-hoc Tukey HSD test was used to determine pairwise variation between the means.

We thank Miguel de Lucas, Keith Lindsey and Jen Topping for critical reading of the manuscript, and Youssef Belkhadir for comments on the preprint. The authors are grateful to Keiko Torii for sharing er and erl mutants, and ERf reporter lines, and to the Nottingham Arabidopsis Stock Centre for providing other genetic resources.