Symplesiomorphies in the WUSCHEL clade suggest that the last common ancestor of seed plants contained at least four independent stem cell niches



  • Evolutionary studies addressing plant architecture have uncovered several significant dichotomies between lower and higher land plant radiations, which are based on differences in meristem histology and function. Here, we assess the establishment of different stem cell niches during land plant evolution based on genes of the stem cell-promoting WUSCHEL (WUS) clade of the WOX (WUSCHEL-related homeobox) gene family.
  • WOX gene orthology was addressed by phylogenetic analyses of full-length WOX protein sequences and cellular expression pattern studies indicate process homology.
  • Gene amplifications in the WUS clade were present in the last common ancestor (LCA) of extant gymnosperms and angiosperms. Whereas the evolution of complex multicellular shoot and root meristems relates to members in the WUS/WOX5 sub-branch, the evolution of marginal and plate meristems or the vascular cambium is associated with gene duplications that gave rise to WOX3 and WOX4, respectively. A fourth WUS clade member, WOX2, was apparently recruited for apical cell fate specification during early embryogenesis.
  • The evolution and functional interplay of WOX3 and WOX4 possibly promoted a novel mode of leaf development, and evolutionary adaptations in their activities have contributed to the great diversity in shape and architecture of leaves in seed plants.


Studies on the evolution of land plants have focused on key features of development, such as plant architecture, the histology of the shoot apical meristem (SAM), or the morphology, origin and formation of leaves, and have identified significant developmental dichotomies between land plant radiations: extant vascular plants, lycophytes and euphyllophytes (ferns and seed plants) grow and produce leaves by the activity of the sporophytic shoot apex, which in lycophytes and ferns depends on the function of apical initials, whereas the SAM of seed plants (angiosperms and gymnosperms) has a multicellular, layered histology and maintains a small stem cell population (Sussex, 1989). Stem cell homeostasis in the SAM of the model angiosperm Arabidopsis thaliana is controlled by a feedback loop between WUSCHEL (WUS) and CLAVATA (CLV) signalling (Brand et al., 2000; Schoof et al., 2000). WUS is transcribed in the organizing centre (OC) of the SAM and encodes a mobile homeodomain (HD) transcription factor (Mayer et al., 1998; Yadav et al., 2011), which promotes stem cell identity in a small population of cells above the OC. The stem cell number is limited by repression of WUS transcription, which involves heterodimeric transmembrane receptor kinases that respond to the CLAVATA3 (CLV3) ligand secreted by the stem cells (Trotochaud et al., 1999; Muller et al., 2008).

The multicellular and layered SAM of seed plants apparently coevolved with changes in the leaf developmental programme (Sussex, 1989; Harrison et al., 2005; Sanders et al., 2011). The sporophytic apex and leaves of most ferns grow by the activity of apical initials (Fig. 1), and stereotypic cell divisions create a repetitive array of several to many leaflets before the initial is ultimately consumed for a single leaflet or lobe (Sanders et al., 2011). In contrast to this acropetal mode of leaf maturation, leaves of angiosperms originate from a group of founder cells at the periphery of the multicellular SAM, and after an initial phase of synchronous cell proliferation, the leaf matures basipetally from the apical tip towards its base, where cells remain competent to proliferate and to manifest positional cues with regard to different leaf axes (Tsukaya, 2002). Although both fern and eudicot leaves are referred to as megaphylls, their morphogenesis differs profoundly in that cells proliferate either at the distal tip or at the leaf base, respectively, a difference that relates to the activity of apical or marginal initials in ferns or to the concerted function of marginal and plate meristems at the leaf base in eudicots (Fig. 1).

Figure 1.

Shoot apical meristem (SAM) histology and megaphyll growth in ferns and eudicots/Gnetales. The shoot apex of ferns grows by the activity of single apical initials (ai; upper left), whereas seed plant sporophyte development is based on multicellular layered meristems (upper right) with apical/basal (tunica/corpus) and radial zonation (central (CZ) and peripheral zone (PZ)). The CZ comprises the stem cell niche, whereas cells of the peripheral zone are recruited for leaf primordia (lp) development. Most fern leaves grow by the activity of leaf apical initials (lai; lower left), reminiscent of activity of the apical initial in the shoot apex, resulting in an acropetal gradient of leaf development, with growth ceasing first at the leaf base. By contrast, the diffuse growth of eudicots and Gnetum leaves is based on the activity of marginal and plate meristems at the leaf base (lower right) and results in a basipetal mode of leaf maturation, with growth ceasing first at the leaf tip. The arrows mark the direction of leaf growth; light grey stripes indicate growth zones, apical initials/stem cells are stained dark grey.

Traditional paleontological studies on leaf evolution have addressed the increase in the complexity of leaves from basal to higher land plants: microphylls or small leaves of lycophytes with a single vascular strand originated from tissue outgrowth (enation theory; Kenrick & Crane, 1997) whereas megaphylls according to the telome theory (Gifford & Foster, 1989) arose from planation, overtopping and webbing of lateral branches. However, these studies are usually based on interpretations of frond architecture and cannot elucidate the developmental potential of foliar meristems (Boyce & Knoll, 2002), which in Arabidopsis are responsible for leaf blade inception and expansion, respectively (Donnelly et al., 1999). By contrast, modern comparative evolutionary developmental biology (evo-devo) studies have addressed meristem evolution from a functional perspective rather than with regard to positional homology deduced from fossil records (Boyce & Knoll, 2002).

Leaf diversity based on modifications in meristem activity has been analysed in eudicots, and conserved or adapted regulatory nodes between simple and compound leaves have been identified (Blein et al., 2008; Piazza et al., 2010). Another approach has aimed to identify conserved mechanisms in micro- or megaphyllous leaf development (Bharathan et al., 2002; Harrison et al., 2005; Floyd & Bowman, 2010). For example, KNOTTED1-like homeobox (KNOX) genes encode a class of plant HD transcription factors that are expressed mutually exclusively to ASYMMETRIC LEAVES1, ROUGH SHEATH2, PHANTASTICA (ARP) genes in seed plant apices (Schneeberger et al., 1998; Waites et al., 1998; Byrne et al., 2000). Despite their presence in all vascular plant genomes, KNOX and ARP genes have been recruited differentially in lower and higher plant radiations. Those KNOX members expressed in the SAM of the extant lycophyte Selaginella kraussiana are not orthologous to those expressed in the SAM of euphyllophytes and are coexpressed with ARP orthologues in leaf primordia of lycophytes or ferns (Harrison et al., 2005). These data support differential recruitment of ancestral genes (i.e. homoplasy), which possibly relate to the independent origins of megaphylls within euphyllophytes, altered developmental programmes, but also inconsistencies in the histological classification of micro/megaphyllous leaves (Tomescu, 2009).

Megaphyllous leaves of gymno- and angiosperms most likely have a common origin (Tomescu, 2009) and characters shared by multiple taxa, and their most recent common ancestors in cladistics are synapomorphies or symplesiopmorphies, if shared with an earlier common ancestor. Based on the fossil record, complex laminated leaves evolved in the Paleozoic in at least four vascular plant lineages: progymnosperms (extinct), sphenopsids, ferns and seed plants. In each lineage, leaf complexity followed the same sequence of morphologies: from dichotomizing single-veined leaves, via multiveined leaves with divergent venation, to convergent and lastly reticulate venation patterns (Boyce & Knoll, 2002). The sequence of increasingly complex leaf architectures suggests that the last common ancestor (LCA) of all four clades had already evolved developmental mechanisms that served as the basis for parallel evolution. The venation pattern of a leaf reflects the mode by which it grows (i.e. marginal vein endings indicate marginal meristem activity), whereas a reticulate venation pattern with veins of multiple orders and free-ending internal veinlets characteristic of extant angiosperm species, at least of eudicots, also requires nonmarginal diffuse intercalary leaf growth (Boyce & Knoll, 2002) or plate meristem activity (Fig. 1). The long independent trajectory of angiosperms and gymnosperms provides a valuable resource with regard to leaf meristem evolution, because the major gymnosperm orders, Cycadales, Ginkgoales, Gnetales, Pinales and Cupressales (Christenhusz et al., 2011), display different modes of leaf growth. The genus Gnetum comprises c. 30–35 species of the order Gnetales and is the only nonangiosperm seed plant genus with a reticulate venation pattern (Fig. 1), whereas exclusively marginal vein endings are found in Cycads, Ginkgo and some conifers (Boyce, 2005).

Apart from the KNOX gene family mentioned earlier, members of the WUS-related homeobox or WOX gene family are particularly suitable for studies on meristem evolution. Beside its founding member WUS, several other family members are associated with specific stem cell niches: WOX5 in the quiescent centre of the root meristem (RM) in Arabidopsis, maize and rice (Kamiya et al., 2003; Nardmann et al., 2007; Sarkar et al., 2007); WOX4 in the vascular cambium of Arabidopsis, tomato, rice and maize (Ji et al., 2010b; Suer et al., 2011; Ohmori et al., 2013); WOX3 or PRESSED FLOWER (PRS) in lateral domains and margins of new organ primordia in Arabidopsis and maize (Matsumoto & Okada, 2001; Nardmann et al., 2004); and WOX1 in blade expansion of Petunia hybrida (MAEWEST; Vandenbussche et al., 2009), Medicago truncatula (STENOFOLIA) and Nicotiana sylvestris (LAM1; Tadege et al., 2011) or together with PRS/WOX3 in Arabidopsis (Nakata et al., 2012). In addition, WOX2 mediates apical cell fate after the first asymmetric division of the zygote during Arabidopsis embryogenesis (Breuninger et al., 2008) and this transcription pattern is conserved in the maize embryo (Nardmann et al., 2007). Some genes are functionally interchangeable when expressed under the appropriate promoter, that is, WUS and WOX5 can replace each other in the SAM or RM (Sarkar et al., 2007) and WUS can substitute for WOX3 in leaf margins (Shimizu et al., 2009). All these genes group in the WUS clade (Deveaux et al., 2008; Nardmann & Werr, 2012) and members of this clade are absent in the genomes of lower plant species (i.e. the moss Physcomitrella patens or the lycophyte Selaginella moellendorffii), but have recently been identified in two leptosporangiate ferns, Ceratopteris richardii and Cyathea australis (Nardmann & Werr, 2012). The cell type-specific expression of WOX1/WOX3 and WOX4 in leaf meristems and the vasculature system of eudicots, respectively, prompted us to search for orthologues in gymnosperms, as their expression patterns might provide insights into the evolution of different modes of euphyllophyte leaf development and venation patterns.

Here, we show that the genomes of Ginkgo biloba, Gnetum gnemon and Pinus sylvestris consistently contain WOX3 and WOX4 orthologues. The inclusion of the full-length gymnosperm WOX sequences and those of the basal angiosperm Amborella trichopoda refines phylogenetic reconstructions that robustly distingiush the WUS clade from two other sub-branches in the WOX gene family. This phylogeny, together with cellular expression studies, leads to four major conclusions: genes in the WUS clade were amplified before the separation of gymnosperms and angiosperms; the genome of the LCA of seed plants contained at least WOX2, WOX3 and WOX4 orthologues, in addition to a previously described WUS/WOX5 pro-orthologue (Nardmann et al., 2009); WOX2, WOX3 and WOX4 were subfunctionalized to the embryo proper, marginal meristems and the vascular system, respectively, at the base of seed plants and comprise discrete symplesiomorphic characters; and the shift of the concerted function of WOX3/WOX1 and WOX4 to nonmarginal leaf domains relates to the evolution of plate meristems, which enabled the new, basipetal mode of leaf development with reticulate venation pattern characteristic of eudicots and Gnetales.

Materials and Methods

Isolation and phylogeny of full-length WOX gene family members of gymnosperms and A. trichopoda

Full-length sequences of WOX proteins from the gymnosperms Ginkgo biloba L., Gnetum gnemon L. and Pinus sylvestris L. were established by cDNA cloning or genome walking, starting from the previously published homeobox sequences (Nardmann et al., 2009). Genome walking was performed with the GenomeWalker Universal Kit™ (Clontech; Takara Bio Inc., Otsu, Shiga, Japan) and rapid amplification of 5′- and 3′- cDNA ends (RACE) using the First-Choice RLM-RACE Kit (Ambion; Life Technologies Ltd, Carlsbad, CA, USA), according to the manufacturer protocols. The WOX sequences of the basal angiosperm Amborella trichopoda BAILL. were derived from the full genome sequence draft available at

The accession numbers of the new full-length sequences and the loci of the genome-derived sequences are given in Supporting Information, Table S1; other accession numbers or WOX IDs have been previously published (Nardmann et al., 2009; Tadege et al., 2011). Multiple sequence alignments were conducted using ClustalX ( and were refined using BioEdit (Tom Hall, IbisTherapeutics, Carlsbad, CA, USA) to correctly align regions of conservation such as the WUS- and the WUS/WOX5-box (Fig. S7). The Bayesian inference (BI) method was performed using MrBayes 3.1.2 ( with 1000 000 generations and four independent chains. Every 1000th tree was sampled and a consensus tree of 250 trees was determined using a burn-in of 250 (Fig. S1). The MEGA5.1 program (Tamura et al., 2011) was used for the maximum likelihood (ML) analysis with 1000 repetitions (Fig. S2).

In situ hybridization

Nonradioactive in situ hybridization followed the protocol of Jackson (1991); paraffin-embedded tissue was sectioned with a Leica RM 2145 rotary microtome (7–10 μm thickness; Leica, Wetzlar, Germany). Probes for in situ hybridization were cloned in sense or antisense orientation to the T7- or SP6-promoter, and digoxigenin-labelled RNA probes were obtained as described (Bradley et al., 1993). To avoid cross-hybridization, the WOX4 probe sequences exclude the homeobox, whereas the WOX2 probe contains a part of the homeobox and the WOX3 probes contain the whole open reading frames as a result of their short size. However, even the most similar GbWOX3A and GbWOX3B probes did not show any cross-hybridization (compare Fig. 4e–g), and cell type specificity of the antisense probes was generally controlled by sense probe hybridizations (data not shown). The CrWUL probe corresponded to bp 843–1580 of accession FR716458, the GbWOX3A probe to bp 1–529 of accession FM882125, the GbWOX3B probe to bp 1–542 of accession FM882126, the GbWOX4 probe to bp 597–1360 of accession HF564615, the GgWOX3 probe to bp 1–539 of accession FM882153, the GgWOX4 probe to bp 696 to 1278 of accession HF564612, the PsWOX3 probe to bp 1–444 of accession FM882158 and the PsWOX2 probe to bp 66–480 of accession FM882159. Images were taken using an Axioskop microscope equipped with an Axiocam camera (Carl Zeiss, Jena, Germany) and were processed using Adobe Photoshop CS2 (Adobe Systems, San Jose, CA, USA).



The simplified phylogenetic tree (Fig. 2a) depicts the division of the WOX family into major clades and sub-branches therein. The full tree, based on the BI method using the full-length WOX protein sequences of major lineages of the green plant tree (green algae, mosses, leptosporangiate ferns, gymnosperms and angiosperms), is given in Fig. S1. The tree was rooted using single WOX proteins from two extant green algae, Ostreococcus tauri and Ostreococcus lucimarinus, which are sisters to land plant WOX family members. The tree contains newly isolated sequences and full protein sequences, where previously only the HD sequences were available (Nardmann et al., 2009; Nardmann & Werr, 2012), from the gymnosperms G. gnemon (GgWOXX, GgWOXY, GgWOX2A, GgWOX2B, GgWOX4, GgWOX9, GgWOX13), G. biloba (GbWOX4, GbWOX9, GbWOX13) and P. sylvestris (PsWOX4) and from the basal angiosperm A. trichopoda (AmtWOX1–5, 9, 11, 13 and AmtWUS), which were derived from the full genome sequence draft available at The protein sequence-based phylogenetic tree (Figs 2a, S1) is supplemented by branch-specific sequence motifs and exon-intron structures of WOX genes from the WUS clade (Fig. 2b). For simplification, in respect to mono- and eudicots, only the exon-intron structures of the A. thaliana WOX genes are shown (Vandenbussche et al., 2009), as they represent the WOX gene structures typical of both angiosperm lineages.

Figure 2.

Phylogeny of the plant-specific WOX family. (a) Simplified Bayesian interference phylogenetic tree based on full-length protein sequences of WOX family members throughout the plant kingdom. The complete tree has been transferred to Supporting Information Fig. S1. Single protein sequences are denoted by smaller script, and sub-branches with several members are indicated by larger, underlined script. The presence (+) or absence (−) of stem cell-promoting members in gymnosperm or angiosperm genomes for each WUS clade sub-branch is indicated in the small table to the right of the tree. Posterior probability (pp) values are restricted to the nodes of major sub-branches; the pp-values that result from the omission of the SmWOXII sequence are in parentheses. All pp-values are included in Fig. S1. Abbreviations are as follows: Ot, Ostreococcus tauri; Ol, Ostreococcus lucimarinus; Sm, Selaginella moellendorffii; Ca, Cyathea australis; Cr, Ceratopteris richardii; Ps, Pinus sylvestris; Gg, Gnetum gnemon; Gb, Ginkgo biloba; Amt, Amborella trichopoda; At, Arabidopsis thaliana. The cladogram is a majority rule tree with a cutoff of 50% and was created using the FigTree program ( (b) Exon-intron structures of WOX genes in the WUS clade. Exons are boxed, lines indicate introns and are dashed when the length of the intron is unknown. The homeodomain and other protein sequence motifs are indicated by different colours: light blue, homeodomain; magenta, WUS-box; yellow, WOX1/WOX4 motif; light grey, WOX1/WOX2 motif; orange, EAR-like (WUS/WOX5) domain.

The inclusion of the newly isolated full-length protein sequences from gymnosperms and Amborella in the phylogenetic analysis supports the subdivision of the WOX gene family into three branches: the WOX13, WOX9 and WUS clades (Deveaux et al., 2008; Nardmann & Werr, 2012). Divergence from the WOX13-type gene present in all analysed land plant genomes started in an ancestor of lycophytes: SmWOXII from Selaginella consistently locates outside of the ancestral WOX13 clade (Nardmann & Werr, 2012), but comprises a unique WOX family member. Its closest relatives are CrWOXA and CrWOXB from C. richardii, which, according to WOX9-specific signatures in the HD DNA-recognition helix, had been previously placed into the WOX9 lineage (Nardmann & Werr, 2012), but still lack the C-terminal motifs conserved in members of the seed plant-specific WOX9 clade that includes newly isolated gymnosperm representatives. The long independent evolutionary trajectory of SmWOXII possibly affects coalescent calculations; that is, the posterior probability (pp) values increase at subsequent branch points when the SmWOXII sequence is omitted from the phylogenetic reconstructions (Fig. S3). At the node that separates the WOX9 clade from the WUS clade, the pp-value increases from 0.65 to 0.98 and the pp-value changes even more markedly from 0.52 to 0.98 at the common branch point of the WUS, WOX1–5 orthologues. The branch point between the WOX9 clade and GgWOXY is unaffected (pp = 1), and GgWOXY is separated from the common node of all other WUS-clade members, including CrWUL and CaWUL from leptosporangiate ferns and GgWOXX, a second Gnetum peculiarity. Neither GgWOXX nor GgWOXY have orthologues in sequenced plant genomes and both represent either Gnetum-specific gene amplifications or are ancestral remnants lost in other plant radiations. Both GgWOXX and GgWOXY contain the WUS clade-specific WUS box; however, they differ significantly in their homeodomains from other WUS clade members – in the turn between helix 2 and the DNA recognition helix 3, the amino acid sequence GPIEPKN in GgWOXX is closer to the Gk/rIEGKN motif in true WUS relatives than it is to the motif GNVAGIN in GgWOXY. The position of GgWOXX and GgWOXY with regard to all other WUS clade members is interchanged in ML reconstructions, which otherwise result in a similar tree architecture (Fig. S2). Both BI and ML reconstructions robustly support separate branch points of the WOX9 clade and the WUS clade, with either GgWOXY (BI) or GgWOXX (ML) located outside the main WUS group, which indicates that the WUS/WOX9 clade bifurcation was apparently manifest in the LCA of seed plants. Whereas the origin of the seed plant-specific WOX9 clade is unclear, although CrWOXA and CrWOXB from C. richardii already contain WOX9-type HD sequences, both phylogenies (BI and ML) confirm that WUS clade members appeared with leptosporangiate ferns (Nardmann & Werr, 2012) and were amplified before seed plants. The phylogenetic reconstructions together with the conserved exon-intron structures (Fig. 2b) imply that the genome of the seed plant LCA contained orthologues of WOX2, WOX3 and WOX4 in addition to a WUS/WOX5 pro-orthologue.

A common origin of the WUS/WOX5 orthologues (Nardmann et al., 2009; Nardmann & Werr, 2012) is consistent with the conserved exon-intron structure of genes encoding the single gymnosperm members in the WUS/WOX5 clade and those of the angiosperm WUS and WOX5 orthologues. By contrast, although WOX1 and WOX3 group together in one sub-branch of the WUS clade, differences in the exon-intron structures and the conservation of discrete protein motifs do not support a simple duplication for WOX3 and WOX1, as was suggested by their redundant functions in Arabidopsis leaf margins (Vandenbussche et al., 2009; Tadege et al., 2011). This assumption neglected evidence for exon shuffling as provided by an N-terminal sequence motif (yellow box in Fig. 1b) that is conserved between WOX1 and WOX4 proteins, or a C-terminal domain (light grey box in Fig. 1b) shared between WOX1 and WOX2 proteins (Vandenbussche et al., 2009), which are both found in gymnosperm WOX4 and WOX2 orthologues (except GgWOX2A/2B), respectively, but are consistently absent in WOX3 orthologues. According to the organization of exons and protein sequence motifs, WOX1 might have evolved as a hybrid between N-terminal sequences of WOX4 and the WOX2 C-terminus, or by more complicated scenarios. Based on the absence of WOX1 orthologues in the genomes of three gymnosperms and its presence in A. trichopoda, WOX1 possibly evolved at the base of angiosperms and was preserved in the lineage to eudicots (Vandenbussche et al., 2009; Tadege et al., 2011), but apparently has been lost in monocots, which instead have manifested duplications of WOX3 orthologues (Nardmann et al., 2007, 2009). Alternatively, WOX1 orthologues might have escaped PCR-based detection in Pinus, Gnetum and Ginkgo and, consequently, the LCA of seed plants might have contained a WOX1 ancestor. None of these assumptions, however, affects the conclusion that at least a WUS/WOX5 pro-orthologue, as well as WOX2, WOX3 and WOX4 orthologues, was present in the LCA of extant seed plants.

WOX3 orthologues were recruited for leaf meristems in the LCA of seed plants

Gnetum gnemon, G. biloba and P. sylvestris represent three major gymnosperm lineages and display different modes of leaf growth, of which Gnetales is the only nonangiosperm lineage with diffuse growth of the leaf lamina (Boyce & Knoll, 2002). RNA in situ hybridizations on G. gnemon shoot apices revealed that GgWOX3 expression is first detectable in a ring-shaped domain at the periphery of the shoot apex (Fig. 3a), which resolves into two crescents that represent the anlagen of incipient leaflets in opposite phyllotaxy. Gnetum leaf primordia emerge by the activity of marginal meristems, which establish between 10 and 13 cell layers that taper towards the lateral margins (Tomlinson & Fisher, 2005). Early in blade expansion, GgWOX3 is transcribed in marginal initials before expression extends into subtending layers and marks a central stripe of two-cell-layer width (Fig. 3b; summarized in Fig. 5). Horizontal longitudinal sections beneath the plane of the marginal initials show continuity of this GgWOX3 expression domain, which marks the position of the plate meristem that promotes anticlinal divisions for leaf expansion (Tomlinson & Fisher, 2005); whereas expression is excluded from the midrib position at the distal leaf tip (Fig. 3c), plate meristem activity persists during subsequent stages of leaf development when seven major second-order veins are established at the flanks of the midrib and joined together by submarginal loops. GgWOX3 remains expressed in two central cell layers between the adaxial and abaxial leaf domains, where cell divisions are required for lamina expansion (Tomlinson & Fisher, 2005), but expression is eliminated at the position of established major second-order or developing higher-order veins (Fig. 3d). Thus, GgWOX3 expression starts with primordia initiation and subsequently marks marginal or plate meristems during leaf expansion (summarized in Fig. 5).

Figure 3.

Cellular expression patterns of gymnosperm WOX3 genes. RNA in situ hybridization performed with GgWOX3 (a–d), GbWOX3A (e), GbWOX3B (f, g) and PsWOX3 (h, i) on transverse (a, c–e, g, i) and longitudinal (b, f, h) sections of young side shoots (Gnetum gnemon and Ginkgo biloba) or seedlings (Pinus sylvestris). Note that the inset in (d) is a close-up of the blade region marked by a rectangle. SAM, shoot apical meristem; pm, plate meristem; mm, marginal mersitem; vb, vascular bundle; wg, wedge; id, indentation; mid, medial indentation; am, apical meristem. Bars, 50 μm.

In contrast to G. gnemon, the leaves of G. biloba grow exclusively by marginal meristems, which release cells in a basipetal direction that concomitantly expand laterally, their coordinated activities forming the fan-shaped Ginkgo leaf lamina, often with a prominent medial indentation (bilobed or biloba) or several wedges (Mundry & Stützel, 2004). The expression patterns of the two Ginkgo WOX3 paralogues, GbWOX3A and GbWOX3B, indicate subfunctionalization, as they are expressed in different parts of the developing leaf lamina. GbWOX3A marks the apical meristems of the several wedges of the emerging leaflets (Fig. 3e), whereas GbWOX3B is expressed complementary to GbWOX3A at the base of the medial indention (Fig. 3g) or in between future wedges (Fig. 3f). Modifications in marginal growth have morphological consequences (Boyce & Knoll, 2002), because, for example, alterations in the concerted GbWOX3A and GbWOX3B activities in the apical leaf meristems and at the indentation base, respectively, might relate to differences between adult and juvenile G. biloba leaves.

In the third gymnosperm, the conifer P. sylvestris, leaves are reduced to linear structures: cataphylls and needles (Boyce, 2005). Needles are borne in fascicles on short shoots and are surrounded by cataphylls at the base of the fascicle. Whereas the first two cataphylls are located opposite to each other, the consecutive foliar organs are arranged in a spiral phyllotaxy (Sacher, 1955b). PsWOX3 expression is first detectable in a few cells at the surface of the SAM periphery (Fig. 3h), where leaf initiation begins with a series of periclinal cell divisions. In the early cataphyll, PsWOX3 transcription is found in apical initials (Fig. 3h), which establish a group of ground tissue cells that increase in number and subsequently result in the outgrowth of the cataphyll. At the time of apical growth initiation, lateral growth also starts via the activity of marginal initials (Sacher, 1955a), which are again marked by PsWOX3 expression (Fig. 3i). In conclusion, the expression patterns of WOX3 orthologues in developing leaves of three gymnosperm representatives reflect growth characteristics with regard to the activity of marginal initials, marginal meristems or plate meristems.

Recruitment of WOX4 orthologues to the vascular cambium predated seed plants

Similar cellular expression studies revealed association of WOX4 with the vascular cambium. In G. gnemon, GgWOX4 expression is first detectable in the midrib region (Fig. 4a) and then in the second-order veins flanking the midrib on either side (Fig. 4b). During blade expansion, GgWOX4 marks the reticulate pattern of the higher-order veins (Fig. 4c), which intersperse plate meristems marked by GgWOX3 activity (Fig. 4d). Both expression patterns are mutually exclusive in consecutive sections hybridized with the GgWOX4 or the GgWOX3 probe (Fig. 4e–h). Expansion of the G. gnemon leaf, therefore, involves the concerted activity of GgWOX3 in plate meristems and GgWOX4 in the vascular cambium, which relates to parenchymal and vascular cell fates.

Figure 4.

Comparison of GgWOX4 and GgWOX3 transcription patterns and expression domains of GbWOX4 and CrWUL. In situ hybridization using antisense probes of GgWOX4 (a–c, f, h), GgWOX3 (d, e, g), GbWOX4 (i, j) and CrWUL (k, l) on longitudinal and transverse sections of Gnetum gnemon side shoots (a, b) as well as frontal sections of leaves from Gnetum (c–h), Ginkgo biloba (i, j) and Ceratopteris richardii (k, l). (e–h) Side-by-side sections hybridized with GgWOX3 (e) or GgWOX4 (f) and their corresponding close-ups (g, h). mr, midrip; vb, vascular bundle; vl, veinlet; pm, plate merstem; wg, wedge; id, indentation; vi, vascular initial. Bars, 50 μm.

By contrast, the G. biloba leaf displays a convergent venation pattern with open vascular veins that extend into the leaf margins, and GbWOX4 transcripts are restricted to one or a few marginal cells in expanding leaf lobes early during development (Fig. 4i,j). GbWOX4 apparently marks the vascular initials of leaf margins, which probably interfere with the recruitment of cellular descendants for the parenchymal cell fate via GbWOX3A, the expression of which extends into submarginal layers of leaf protrusions (compare Fig. 4i with Fig. 3f; summarized in Fig. 5). Later in development, GbWOX4 expression only remains detectable in the vascular cambium of the stem (Fig. S4). In the conifer P. sylvestris, resins associated with developing vascular strands cause a background signal that interferes with cellular expression studies; however, CrWUL, the single member in the WUS clade of the leptosporangiate fern C. richardii (Nardmann & Werr, 2012), is transiently expressed in the developing vasculature (Fig. 4k,l), which suggests that the association of genes in the WUS clade with the vascular cambium possibly predated seed plants.

Figure 5.

Expression patterns of WOX3 and WOX4 orthologues are adapted to different modes of leaf development during seed plant evolution. (a) In Gnetum gnemon, WOX3 (green) and WOX4 (red) mark the marginal meristem (mm) as well as the plate meristem (pm) and the major veins (mv) in young leaf primordia. In older leaf primordia, WOX3 expression is restricted to the plate meristems, which are interspersed by developing veinlets (vl), where WOX4 is transcribed. The upper part displays a medial longitudinal section and the lower part a transverse section of a young (left) and an older (right) leaf primordium. The preservation of WOX1/3 and WOX4 expression patterns between eudicots and Gnetum leaves with reticulate venation patterns implies process homology. (b) The fan-shaped leaf of Ginkgo biloba (right) grows by marginal meristems, and the two Ginkgo WOX3 paralogues, GbWOX3A and GbWOX3B, have been subfunctionalized to the apical meristems (green) of the several wedges (wg) or to the medial (mi) and the minor indentions (id; orange) of the emerging leaflets. The convergent venation pattern of the Ginkgo leaf with open vascular veins arises from vascular initials (vi) marked by GbWOX4 (red) in the margins of the leaf primordia. (c) Competing gymnosperm phylogenies: according to Lee et al. (2011), Gnetales locate at the base of extant gymnosperms (indicated by black lines), whereas other studies indicate a close evolutionary relationship between the Gnetales and Pinaceae (Gnepine hypothesis; Zhong et al., 2011), as indicated by the broken grey arrow. If Gnetales represent a basal gymnosperm lineage, the diffuse mode of leaf development and the similarities in WOX gene expression patterns suggest process homology and preservation of gene functions between eudicots and Gnetum. According to the Gnepine hypothesis, however, these similarities would result from convergent evolution.


The genome of the LCA of seed plants contained at least four members in the WUS clade

The presence of WUS/WOX5, WOX3, WOX4 and WOX2 subclades in seed plants, in contrast to only single members in the genomes of C. richardii and C. australis (Nardmann & Werr, 2012), implies that gene amplifications in the WUS clade occurred after leptosporangiate ferns separated from the LCA of seed plants, which agrees with a postulated whole genome duplication (WGD) before the origin seed plants (Jiao et al., 2011). Discrete WUS and WOX5 functions, serving stem cell niches of the angiosperm SAM and RM, respectively (Nardmann et al., 2009), possibly relate to a second WGD postulated at the base of flowering plants (Jiao et al., 2011), although some uncertainty exists with respect to the HD sequences in the single gymnosperm WUS/WOX5 clade members. The HDs in Gnetum and Ginkgo orthologues contain a WUS-specific extra amino acid residue, whereas the Pinus HD is reminiscent of the WOX5 HD (Nardmann et al., 2009). In contrast to the WUS/WOX5 bifurcation, WOX1 and WOX3 cannot have originated simply from the duplication of a common ancestor, as both differ in protein and exon structure. The presence of WOX1 in A. trichopoda, sister to all other angiosperms, and its apparent absence in three gymnosperms suggest that WOX1 occurred at the base of flowering plants, although in the absence of gymnosperm genome sequences we cannot exclude that WOX1 orthologues evolved earlier and escaped PCR-based detection.

With regard to WGDs, a striking difference exists between the numbers of genes fixed in the WUS clade and those in the ancestral WOX13 branch. This indicates either that seed plants preserved gene duplicates in the case of WUS clade members, but lost corresponding duplicates of the ancestral WOX13-type, or that the selective increase in the number of stem cell-promoting WOX family members might relate to the amplification of single genes or small genome fragments, as described for other gene classes that are important in plant evolution (Cibrian-Jaramillo et al., 2010). Two additional genes within the WUS clade, GgWOXX and GgWOXY, are without orthologues in other seed plant genomes and indicate that the genome of the LCA of seed plants might have contained and even tolerated a higher number of genes in the stem cell-promoting WUS clade. This assumption gains support by the recent positioning of G. gnemon to the base of gymnosperms according to nuclear transcriptome data (Lee et al., 2011), although this view contrasts with previous phylogenetic estimations (i.e. the Gnepine hypothesis (see comparison in Fig. 5c)) that group Gnetales as sister to Pinaceae (Zhong et al., 2011). However, regardless of its position within the gymnosperms, the example of G. gnemon demonstrates that amplifications in the WUS clade have been tolerated during seed plant evolution.

In the LCA of angiosperms, WUS and WOX5 were apparently subfunctionalized to organizers of shoot and root meristem stem cell niches, respectively, in contrast to gymnosperms, where cellular expression patterns of the single WUS/WOX5 clade member in shoot and root apices are still elusive despite their transcripts being detectable by reverse transcription polymerase chain reaction (RT-PCR; Nardmann et al., 2009). However, in G. gnemon, GgWUS/WOX5 is expressed in young strobili and incipient leaf primordia (Nardmann et al., 2009), where its expression domain overlaps with the initial GgWOX3 pattern elaborated here (Fig. 3a). The presence of WOX3 and WOX4 throughout angiosperms and gymnosperms and their conserved cell type-specific expression patterns indicate that their recruitment for leaf and vascular meristems, respectively, was manifest in the LCA of seed plants. Furthermore, in situ hybridizations performed on P. sylvestris embryos with the PsWOX2 probe (Fig. S6), showed transcription in the apical embryo proper and an absence in the subtending suspensor, an expression pattern reminiscent of the WOX2 patterns in maize and Arabidopsis few-celled embryos (Haecker et al., 2004; Nardmann et al., 2007). Therefore, in addition to WOX3 and WOX4, the recruitment of WOX2 for the specification of apical embryonic cell fate was possibly fixed at the base of seed plants. Based on angiosperm models, WOX3 and WOX4 functions essentially contribute to: the maintenance of marginal and/or plate meristems, the latter of which in eudicots is associated with WOX1 function; the competence to differentiate mesenchymal from vascular cell fates; and the coordinated interplay of mesenchymal and vascular tissue during lamina expansion. Modulations in WOX3/WOX1 and WOX4 activity might account for the variability in leaf shapes and venation patterns observed in extant seed plants, and their symplesiomorphy implies that this plasticity evolved by the use of a substantially amplified gene repertoire in the WUS clade at the base of gymnosperms and angiosperms that was apparently not present in the progenitor of ferns and seed plants.

The interplay of two stem cell niches enables leaf expansion in gymnosperms and angiosperms

In Arabidopsis, the synergistic function of WOX3/PRS and WOX1 in the middle domain of the emerging leaflet is required for blade expansion (Nakata et al., 2012), whereas, as a result of redundancy, single mutants have only subtle leaf phenotypes. By contrast, predominant WOX1 activity in central mesenchymal layers accounts for severe defects in lamina outgrowth in stenofolia or maewest mutants of Medicago truncatula or Petunia hybrida (Tadege et al., 2011), and in analogy the ns1/2 mutant phenotypes in maize uncovered a WOX3-derived instructive signal from the L1 to subtending cell layers for the outgrowth of basal leaf margins (Scanlon, 2000; Nardmann et al., 2004; Shimizu et al., 2009). The expression pattern of GgWOX3 from the margins into the two-celled mesenchymal stripe in the Gnetum leaf combines those of the angiosperm WOX3 and WOX1 orthologues, and a noncell-autonomous WOX3 signal might enable concerted cell division in the 10–13 clonal layers of the Gnetum leaf (Tomlinson & Fisher, 2005).

Loss of WOX4 function in Arabidopsis results in the accumulation of undifferentiated ground tissue and the reduction of differentiated phloem and xylem in leaves, stems and roots (Ji et al., 2010b). In the Arabidopsis stem, WOX4 is required to reprogram mesenchymal cells for interfascicular cambial fate during secondary thickening growth (Suer et al., 2011). Thus, WOX4 implements vascular cell fates in the Arabidopsis model, consistent with the WOX4 expression patterns in Gnetum and Ginkgo. The GgWOX3 and GgWOX4 expression patterns not only outline a dynamic interplay in the expanding G. gnemon leaf (schematically depicted in Fig. 5a), but also indicate a striking conservation of cell type specificity and layering between the Gnetum and the eudicot leaf (Vandenbussche et al., 2009; Tadege et al., 2011; Nakata et al., 2012).

In contrast to the reticulate venation pattern in Gnetum, the G. biloba leaf has a convergent venation pattern and vascular strands extend into the apical or lateral leaf margins (summarized in Fig. 5b). The GbWOX3A expression domain in apical protrusions is interspersed by GbWOX4 transcriptional activity in a few marginal epidermal cells, most likely the termini of prospective vascular strands (summarized in Fig. 5b). The interdependence of marginal meristem activity and vascular development is substantiated by the reduced number of vascular strands in ns1/ns2 double-mutant maize leaves (Scanlon et al., 1996), and marginal WOX3 activity is thus required for the development of lateral veins in the typical parallel pattern of grass leaves. This is reminiscent of WOX1 defects in stenofolia mutants, which outline the importance of plate meristem function in establishing the correct venation pattern in the eudicot M. truncatula (Tadege et al., 2011).

In Pinus, WOX3 expression pre-patterns the position of the incipient leaf primordium, similar to ZmWOX3A/B, OsWOX3 or GgWOX3 expression in cells of the P0 primordium at the circumference of the SAM. PsWOX3 later marks the apical and marginal initials of the developing linear cataphyll similar to NS1/NS2 or OsNS transcription in the leaf margins of maize or rice (Nardmann et al., 2004, 2007). Thus, the altered mode of leaf development in grasses and Ginkgo compared with the Gnetales and eudicots correlates with a lack of plate meristem activity marked by WOX3 in Gnetum or WOX1 in eudicots. Irrespective of the unresolved origin of WOX1, the preserved cell type specificity of the WOX3 and WOX4 transcription patterns, together with their symplesiomorphy and adaptations to altered leaf developmental programmes (schematic comparison of the G. gnemon and G. biloba patterns in Fig. 5a,b), suggests that common ancestry relates to process homology in extant gymno- or angiosperms.

As Gnetum of the order Gnetales is the only nonangiosperm seed plant genus with a diffuse mode of lamina expansion (Boyce & Knoll, 2002), the question of homologous processes that involve orthologous protein functions relates to the phylogenetic position of Gnetales within seed plants. Recent studies based on nuclear genome transcriptome data position Gnetales as a basal gymnosperm order (Lee et al., 2011); the conservation of the WOX3 and WOX4 transcription patterns and the diffuse mode of leaf growth between Gnetum and Arabidopsis would therefore support process homology based on distinct WOX3 and WOX4 functions at the base of seed plants. By contrast, the Gnepine hypothesis, considering morphological criteria and conservation of the plastid genome, groups Gnetales as sister to Pinaceae (Zhong et al., 2011), which would only be compatible with convergent evolution (indicated in Fig. 5c). The WOX phylogeny cannot resolve this discrepancy; however, the preservation of the WOX3 and WOX4 transcription patterns in a central two-cell-layered stripe and in interspersed developing vascular strands, respectively, of the expanding leaves of Gnetum and eudicots (Arabidopsis, Medicago) would require the convergent evolution of multiple characters. Although formally not excluded, this simultaneous convergence in our view appears less likely than considering a basal position of Gnetales within gymnosperms as supported by the nuclear transcriptome (Lee et al., 2011).

Evolutionary aspects

The presence of WOX4 and WOX3 orthologues before the angiosperm/gymnosperm split and the deep branches in the land plant phylogeny do not allow the order of WOX3 and WOX4 evolution to be easily distinguished. However, the finding that the single WUS lineage member of the leptosporangiate fern C. richardii, CrWUL, is expressed not only in pluripotent descendants of the root apical initial (Nardmann & Werr, 2012) but also in the vasculature suggests that new additions in the WOX family phylogeny were possibly recruited into the vascular system before seed plants. By contrast, we could not find any evidence for CrWUL transcription in shoot or leaf apical initials (Nardmann & Werr, 2012). The uniqueness of the WOX4 gene function in the WUS clade was shown by promoter-swap experiments in Arabidopsis, as the prs mutant phenotype is fully rescued when either the WOX3 or the WUS protein is expressed from the PRS promoter (Shimizu et al., 2009), but only partially complemented when WOX4 is similarly expressed (Ji et al., 2010a).

Amplifications and subfunctionalizations of a WUS clade progenitor might have resulted in a WOX4 sub-branch that preserved an ancestral character in specifying the vascular cell fate and in other clades where neofunctionalization is more evident. Whereas the association of WOX2 with apical embryonic cell fate and WOX3 with cell types that give rise to marginal and/or plate meristems possibly predated the gymnosperm/angiosperm split, previous analyses revealed that members of the WUS/WOX5 subclade might have been subject to significant adaptations during seed plant evolution (Nardmann & Werr, 2006; Nardmann et al., 2009). For example, expression of the single G. gnemon WUS/WOX5 gene has been detected in incipient leaf primordia at the periphery of the vegetative shoot apex (Nardmann et al., 2009) where it overlaps with that of GgWOX3 (Fig. 3a), but the GgWUS/WOX5 and GgWOX3 transcription patterns differ substantially during the reproductive phase (Fig. S5).

Concerning leaf development, the spatial and temporal proximity of WOX3 and WOX4 activity raises the question of crosstalk between distinct cell types. As the activities of members of different WOX family clades, WUS, WOX4 and WOX5 of the WUS clade (Brand et al., 2000; Stahl et al., 2009; Hirakawa et al., 2010; Ohmori et al., 2013) and WOX8 of the WOX9 clade (Fiume & Fletcher, 2012), are regulated by CLAVATA3/ESR-related (CLE) peptides, the CLE-dependent feedback loops possibly predated the evolution of genes in the WUS clade, which is consistent with the presence of CLE and ancestral WOX13-type genes in basal land plant genomes. Sub- or neofunctionalization of newly evolved WOX family members therefore possibly occurred within existing regulatory networks and the secretion or noncell-autonomy of CLE peptides might still synchronize WOX3 and WOX4 activity at the boundary of the mesenchyme or vasculature either by repressing (Brand et al., 2000; Stahl et al., 2009; Hirakawa et al., 2010) or by promoting (Fiume & Fletcher, 2012) gene expression.

In conclusion, symplesiomorphies in the WUS clade imply that the LCA of seed plants contained at least ancestors of WOX2, WOX3, and WOX4, in addition to the previously described WUS/WOX5 pro-orthologue (Nardmann et al., 2009), which were possibly associated with discrete stem cell niches. Whereas the recruitment of WUS or WOX5 function for stem cell organizers in the multicellular shoot or root apex, respectively, might have occurred at the base of angiosperms, the fixation of WOX3 and WOX4 and their recruitment for marginal and plate meristems or the vascular cambium, respectively, predated the gymnosperm/angiosperm split. The WOX3 and WOX4 symplesiomorphies are linked to a major seed plant invention, the basipetal mode of leaf growth, where the striking conservation of expression patterns during lamina expansion in G. gnemon and eudicot models suggests homologous processes. Gene amplifications in the WUS clade, and their fixation and neofunctionalization into discrete stem cell niches apparently distinguished the LCA of seed plants from a common progenitor with ferns. The multiplicity of stem cell niches present at the base of seed plants conceivably provided the potential for adaptations that facilitated the evolution of diverse architectures and shapes characteristic for extant seed plants.


We thank Dr J. Chandler for critically reading the manuscript; H. Shahbodaghi-Rueckert for technical assistance; and J. Hintzsche, S. Wirges, and C. Wienhold for providing gymnosperm plant material. This project was supported by the Deutsche Forschungsgemeinschaft through WE 1262/7-1 and SFB680.