Leaf Shape Evolution Through Duplication, Regulatory Diversification, and Loss of a Homeobox Gene

See allHide authors and affiliations

Science  14 Feb 2014:
Vol. 343, Issue 6172, pp. 780-783
DOI: 10.1126/science.1248384


In this work, we investigate morphological differences between Arabidopsis thaliana, which has simple leaves, and its relative Cardamine hirsuta, which has dissected leaves comprising distinct leaflets. With the use of genetics, interspecific gene transfers, and time-lapse imaging, we show that leaflet development requires the REDUCED COMPLEXITY (RCO) homeodomain protein. RCO functions specifically in leaves, where it sculpts developing leaflets by repressing growth at their flanks. RCO evolved in the Brassicaceae family through gene duplication and was lost in A. thaliana, contributing to leaf simplification in this species. Species-specific RCO action with respect to its paralog results from its distinct gene expression pattern in the leaf base. Thus, regulatory evolution coupled with gene duplication and loss generated leaf shape diversity by modifying local growth patterns during organogenesis.

Developmental Complexity

Although related, the plants Arabidopsis thaliana and Cardamine hirsuta have different sorts of leaves—one, a rather plain oval and the other, a complicated multipart construction. Comparing the development of the two leaf types, Vlad et al. (p. 780) uncovered a gene that regulates developmental growth. The C. hirsuta gene encoding the REDUCED COMPLEXITY (RCO) homeodomain protein arose through gene duplication and neofunctionalization, but was lost in the A. thaliana lineage. In C. hirsuta, RCO suppresses growth in domains around the perimeter of the developing leaf, yielding complex-shaped leaves. A. thaliana, lacking RCO, produces simple leaves. When RCO was expressed in A. thaliana, the leaves became more complex. Thus, the capacity to produce complex leaves remains, despite loss of the initiator.

Understanding how form evolves requires identifying the genetic changes underlying morphological variation between species and elucidating how those changes influence morphogenesis. In this work, we investigate this problem in the case of angiosperm leaves. Both simple and dissected leaves initiate as entire structures at the flanks of the pluripotent shoot apical meristem (1). However, in dissected leaves, elaboration of lateral growth axes after leaf initiation generates leaflets (2, 3). So far, no gene has been identified that expresses specifically at developing leaflets and is sufficient to convert a simple leaf into a more complex one. Rather, a key factor in leaflet formation is the reactivation of meristem genes in leaves, which suggests that evolutionary differences in leaf complexity arose through modification of activity of genes that influence meristem function (4). This view is also consistent with the evolutionary origin of leaves from branched shoots (5). To determine whether leaflet-specific factors exist, we conducted a genetic screen in Cardamine hirsuta, a dissected-leaf relative of the simple-leaf reference plant Arabidopsis thaliana (Fig. 1, A and B) (6). If such genes exist, then loss of their function might prevent leaflet formation without perturbing meristem function.

Fig. 1 Mapping of RCO and complementation of the rco mutant.

(A to C) Silhouettes of an A. thaliana simple leaf (A) with small marginal protrusions called serrations (red asterisk); a C. hirsuta dissected leaf (B) with lateral leaflets (black arrow) borne by petiolules (black arrowhead) and a terminal leaflet (TL); and a C. hirsuta rco mutant leaf (C), in which leaflets are converted to lobes (red arrowhead). (D) Schematic representation of genes in the RCO genetic interval predicted by sequence similarity to the A. thaliana genome. C. hirsuta orthologs are indicated with inverted commas; interval borders are marked with black flags; and shading indicates genes absent in A. thaliana. (E) Alignment of proteins encoded by LMI1 and LMI1-like genes in A. thaliana and C. hirsuta, respectively. The red arrow marks the last amino acid residue (Val124) of the truncated protein encoded by the rco mutant transcript; horizontal lines indicate the homeobox (red line) and homeobox-associated leucine zipper (blue line) domains. Amino acid residues are shown as single-letter abbreviations (25). (F to I) (F). Complementation to WT morphology (F) of the rco mutant phenotype (G) by transgenic expression of RCO (rco; RCOg) (H) but not ChLMI1 (rco; ChLMI1g) (I) genomic fragments. Fourth and fifth leaves are shown. Scale bars in (A to C) and (F to I), 1 cm.

The recessive mutant, rco (reduced complexity), converts the C. hirsuta adult leaf from dissected into a simple lobed leaf (Fig. 1, B and C) without affecting the number and positioning of leaves. Thus, RCO is required for leaflet development but not leaf initiation (fig. S1, A to G). RCO is a homeobox gene present in the genome of C. hirsuta but absent in A. thaliana, and the rco phenotype is caused by reduced gene function (Fig. 1, D to H, and fig. S1, H and I). Moreover, RCO is part of a tandem gene triplication, but the A. thaliana genome has only one of these genes, LMI1 (LATE MERISTEM IDENTITY 1), previously identified as a floral regulator (Fig. 1D) (7). Phylogenetic analysis revealed that RCO arose from the duplication of LMI1-type sequences within the Brassicaceae after the divergence of Aethionema and before the last common ancestor of Arabidopsis and Brassica (fig. S2). RCO duplicated further, yielding a cluster of three genes in C. hirsuta, A. lyrata, and Capsella rubella, though LMI1-like3 is not expressed in C. hirsuta, and concerted evolution may have influenced this gene cluster (fig. S1, I; fig. S2; and fig. S3). Secondary loss of RCO-type sequences left LMI1 as a singleton in the A. thaliana lineage (fig. S3). We found that simply increasing the dose of LMI1-type protein is not sufficient to suppress rco or increase leaf complexity in A. thaliana (Fig. 1I and fig. S4), suggesting that RCO function in leaflet development is borne out of its specific gene expression properties and/or protein function. Thus, RCO is a taxonomically restricted gene underlying a species-specific trait that distinguishes two species that diverged relatively recently (810).

We used reporter gene assays and RNA in situ hybridization to determine the expression pattern of RCO (Fig. 2, A to D, and fig. S5, A to F). RCO expression is restricted to developing leaves, in two small regions at the base of terminal and lateral leaflets, and is absent from the meristem-leaf boundary (Fig. 2, A to D; fig. S5, A to F; and fig. S6, A to C). By contrast, ChLMI1 is expressed in a near-complementary pattern to RCO in terminal and lateral leaflet margins and also in stipules and flowers, similar to its A. thaliana ortholog (Fig. 2, E to H; fig. S5G; and fig. S6, D to F) (7). Thus, RCO activity at the base of leaflets is required for leaflet formation, and its domain of expression distinguishes RCO from its paralog ChLMI1. The orthologous genes ChLMI1 and AtLMI1 share comparable expression domains, indicating that the distinct expression pattern of RCO reflects regulatory diversification from ChLMI1 (fig. S7). Consistent with this view, RCO 5′ regulatory sequences drive reporter gene expression in more proximal and internal regions of the A. thaliana leaf than those of ChLMI1, AtLMI1, and Aethionema arabicum LMI1 (Fig. 2, H to L; fig. S8, A to D; and fig. S9). Because LMI1 transcripts do not accumulate at the leaf base of A. arabicum, which is an early divergent member of the Brassicaceae, these comparisons indicate that the RCO expression pattern represents an evolutionary novelty that arose after gene duplication through neofunctionalization (11).

Fig. 2 The RCO expression pattern underlies its ability to promote leaf complexity.

(A to G) Complementary expression patterns of RCO and ChLMI1 shown by RCO::GUS (A and B) and ChLMI1::GUS (E and F) reporter gene analysis and by RNA localization of RCO (C and D) and ChLMI1 (G) at the shoot apex and fifth leaf of C. hirsuta. (H to K) GUS staining in AtLMI1::GUS (H), RCO::GUS (I), ChLMI1::GUS (J), and AaLMI1::GUS (K) A. thaliana leaves. (L) AaLMI1 RNA localization in vegetative Aethionema arabicum leaf. (M to O) Phenotype of leaves four to six of C. hirsuta wild type (M), rco mutant (N), and rco mutant complemented with an RCO::ChLMI1 transgene (O). (P to R) Rosettes of A. thaliana wild type (P) and transgenic RCOg (Q) and AlRCOg (R) plants. (S) GUS staining in AlRCO::GUS A. thaliana leaf. Asterisks indicate stipules. RNA localization images are minimal projections. Scale bars, 100 μm in (A) to (L) and (S); 1 cm in (M) to (R).

To evaluate whether protein sequence specificity also contributes to RCO function, we expressed ChLMI1 under the regulatory regions of RCO and found that this transgene complemented the rco mutant phenotype (Fig. 2, M to O). Thus, RCO and ChLMI1 proteins are functionally equivalent in this developmental context, and these results suggest that the species-specific action of RCO in leaflet formation reflects diversification of gene expression from its paralog ChLMI1. The absence of RCO, a leaf complexity gene, from the A. thaliana genome suggests that RCO plays a key role in shaping leaf diversity. If so, and given that the simple leaf shape of A. thaliana is evolutionarily derived (12), introducing RCO into A. thaliana should reverse some of the effects of evolution and increase leaf complexity. As predicted, A. thaliana transgenic lines carrying a C. hirsuta RCO genomic clone (RCOg) produced deep lobes never seen in the wild type, suggesting that RCO is sufficient to increase A. thaliana leaf complexity (Fig. 2, P and Q, and fig. S10). These transgenic lines lacked pleiotropic effects, consistent with a specific role of RCO in leaf margin morphogenesis (fig. S10). Furthermore, introducing a genomic clone of the RCO ortholog of A. lyrata (AlRCOg), a lobed leaf species, into A. thaliana also produced lobed leaves (Fig. 2R). This phenotype is likely to result from AlRCO protein activity in the leaf base because the AlRCO::GUS reporter is expressed in a similar domain to the C. hirsuta RCO gene (Fig. 2S and fig. S8D). The finding that RCO::GUS and AlRCO::GUS are expressed in the A. thaliana leaf base, despite the absence of an RCO-type gene, suggests that at least part of the ancestral regulatory landscape that promotes leaf complexity through RCO activation has been retained in this species and highlights the importance of the leaf base as an organizing region for leaf growth (13). Collectively, these results indicate that localized RCO action is a key factor in determining leaf shape complexity in the Brassicaceae, and loss of this gene contributed to leaf simplification in A. thaliana.

We next considered how RCO regulates leaflet development. Auxin is required for leaflet development, and failure to organize discrete auxin activity maxima along the leaf margin reduces leaf complexity (3, 14). However, the auxin activity marker DR5 and the auxin efflux carrier PINFORMED1 were similarly expressed in leaf primordia of rco and the wild type, with sequential discrete auxin activity maxima forming at the leaf margin in both lobes and leaflets (Fig. 3, A and B, and fig. S11). Thus, RCO does not contribute to the establishment or maintenance of local auxin activity maxima that control leaflet initiation. RCO is expressed immediately adjacent to leaflet primordia (Fig. 2, A to D, and fig. S5, A to F), so drawing on classic ideas of leaf shape patterning (15), we hypothesized that RCO may influence growth locally to enable separation of individual leaflets. To test this hypothesis, we used MorphoGraphX software to analyze time-lapse images of leaflet growth (16) (Fig. 3, C to H). In wild-type (WT) plants, we observed that cell expansion and proliferation is inhibited in the marginal region between initiating leaflets (Fig. 3, D and E, and fig. S12, A to C). By contrast, in the rco mutant these cells proliferate and grow faster than in the wild type, filling up the space between leaflets (Fig. 3, G and H, and fig. S12, A to C). Conversely, cells within leaflets grow and proliferate fast in both genotypes (Fig. 3, C to H, and fig. S12, A to C). Thus, rco mutant leaflets initiate and grow in a comparable way to the wild type but fail to separate properly from each other due to incomplete growth repression at their boundaries, resulting in a simplified leaf. We also analyzed A. thaliana leaves expressing RCOg and found significant repression of growth and cell proliferation adjacent to emerging serrations that transformed these small protrusions into deep lobes (fig. S12, D to F, and fig. S13). Thus, RCO contributes to growth repression between adjacent leaflets.

Fig. 3 RCO represses growth at the boundary between leaflets and does not influence auxin-based patterning.

(A and B) DR5::VENUS expression (yellow) and chlorophyll autofluorescence signal (blue) in the seventh leaf of WT (A) and rco (B) C. hirsuta. (C to H) Time-lapse of developing lateral leaflets in WT (C to E) and rco (F to H) fifth leaves. Propidium iodide–stained leaf cells (green) are shown in (C) and (F) for each time point. Heat maps of relative surface area increase over 48 hours of growth (color bar: percentage increase) for lateral leaflets are shown in (D) and (G). Heat maps of cell proliferation over 48 hours [color bar: number of cells (n) originating from one initial cell] for lateral leaflets are shown in (E) and (H). White dotted lines denote leaf margins; white dotted rings indicate areas with excess growth and cell proliferation in the rco mutant. Scale bars, 100 μm in (A) and (B); 30 μm in (C) to (H).

We found that ChLMI1 rescues the rco phenotype when expressed from the RCO promoter and the smooth leaf margin of A. thaliana lmi1 when expressed in its genomic context (Fig. 2O and fig. S14). Growth derepression probably contributes to this lmi1 phenotype and that of a classical pea mutant, in which mutation in an unusual LMI1-like gene converts filamentous leaf tendrils into laminate leaflets (7, 17). Additionally, both LMI1 and RCO repress growth when overexpressed in A. thaliana (fig. S15). To understand the degree of conservation of the growth-regulating function of LMI- and RCO-type genes during crucifer evolution, we evaluated the ability of selected genes from the phylogeny (fig. S2) to modify A. thaliana leaf shape when expressed under the control of the RCO promoter. With the exception of the RCO-B gene of C. rubella, all sequences assayed produced deep lobes in A. thaliana leaves (fig. S16). Because these sequences included LMI1 from the early divergent crucifer Aethionema and the basal eudicot Aquilegia, it follows that the potential for LMI1 protein to repress growth evolved before the appearance of RCO in the Brassicaceae, and probably before the split of eudicots from other seed plants. Within the Brassicaceae, evolution of RCO through gene duplication created a new version of these growth repressors that is active in the morphogenetically important leaf base, thus contributing to diversification of leaf shape (Fig. 4).

Fig. 4 Evolution of RCO and its consequences for diversification of leaf shape in crucifers (based on phylogeny presented in figs. S2 and S3).

The genome of Aethionema arabicum, a simple-leaf early divergent crucifer, contains a single LMI1-type gene. RCO-type genes arose from duplication of an LMI1-type gene after the divergence of Aethionema from core Brassicaceae. The genomes of dissected-leaf C. hirsuta and lobed-leaf A. lyrata contain both LMI1-type and RCO-type genes. In the lineage that gave rise to A. thaliana, the RCO-type gene was secondarily lost, contributing to evolution of a simple leaf. Consistent with a role for RCO in promoting leaf complexity, removing its activity from C. hirsuta in the rco mutant leads to leaf simplification, whereas introducing RCO-type function into A. thaliana results in a more complex leaf. Silhouettes are from adult leaves and are not to scale. A. lyrata and C. hirsuta also contain LMI1-like3, a third copy in this gene cluster that arose through duplication of RCO-type but is not shown in this diagram because it is not expressed in C. hirsuta (fig. S1) and has not been characterized functionally.

Leaflet formation in C. hirsuta requires the RCO homeobox gene that arose through gene duplication and is only expressed at initiating leaflets. Thus, evolutionary changes in leaf complexity can arise through factors distinct to those acting at the shoot apical meristem. RCO does not appear to act through the well-characterized auxin-based patterning that underpins leaflet and serration formation (Fig. 3, A and B, and fig. S11), or to regulate transcription of CUP-SHAPED COTYLEDON (CUC) or KNOTTED1-like homeobox (KNOX) genes that influence this patterning (fig. S17) (3, 18, 19). One possibility is that RCO acts parallel to or downstream of these genes to regulate leaf complexity, thus providing a means to uncouple growth and patterning inputs during evolution. RCOg expression in A. thaliana leaves transforms serrations into deep lobes by locally repressing growth adjacent to each serration (fig. S13); however, complete transformation into leaflets may require other genes that are active in C. hirsuta but not in A. thaliana leaves, such as CUC1 or KNOX genes (20, 21). RCO-mediated shape diversification follows a broad principle of regulatory evolution: that morphological diversity is driven by changes in gene expression that minimize fitness costs by circumventing pleiotropy (22, 23). It will be interesting to explore whether the high tendency toward gene duplication in plants (24) was a major driver for evolution of regulatory variants that underlie trait diversification between species, as we have shown here for RCO.

Supplementary Materials

Materials and Methods

Figs. S1 to S17

Table S1 to S4

References (2655)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.
  3. Acknowledgments: We thank A. Hudson, G. Coupland, Y. Eshed, and N. Gompel for critical comments on the manuscript; P. Sarchet and S. Broholm for mutant screening and providing the mapping cross; I. A. Al-Shehbaz for helpful discussions on leaf shape in Aethionema; S. Hake, M. Aida, G. Coupland, S. Sampson, and D. Wagner for materials; and R. Berndtgen, S. Höhmann, E. Rabbinowitsch, J. Baker, and Z. Lewis for technical assistance. This work was supported by Biotechnology and Biological Sciences Research Council grants BB/H011455/1 (M.T.) and BB/H006974/1 (M.T. and A.H.), Deutsche Forschungsgemeinschaft ‘Adaptomics’ grant TS 229/1-1 (M.T. and A.H.) and grant SFB 680 (M.T.), the Gatsby Charitable Foundation (M.T.), Human Frontier Science Program grant RGP0047/2010 (M.T.), a core grant from the Max Planck Society (M.T.), and NSF Plant Genome Research Mid-Career Investigator Award 1238731 (C.D.B.). M.T. also acknowledges support of the Cluster of Excellence on Plant Sciences. The data reported in this paper are tabulated in the supplementary materials. Sequences have been submitted to GenBank for archiving under the following accession numbers: ChLMI1, KF939590; RCO, KF939591; and LMI1-like3, KF939592.
View Abstract

Navigate This Article