The Floral Regulator LEAFY Evolves by Substitutions in the DNA Binding Domain

See allHide authors and affiliations

Science  08 Apr 2005:
Vol. 308, Issue 5719, pp. 260-263
DOI: 10.1126/science.1108229


The plant-specific transcription factor LEAFY controls general aspects of the life cycle in a basal plant, the moss Physcomitrella patens. In contrast, LEAFY has more specialized functions in angiosperms, where it specifically induces floral fate during the reproductive phase. This raises the question of a concomitant change in the biochemical function of LEAFY during the evolution of land plants. We report that the DNA binding domain of LEAFY, although largely conserved, has diverged in activity. On the contrary, other, more rapidly evolving portions of the protein have few effects on LEAFY activity.

LEAFY (LFY) is found in all land plants, which evolved during the past 400 million years. The proteins are remarkably well conserved with two blocks of similarity, in the N- and C-terminal regions (the N and C domain) (Fig. 1A). All missense mutations identified in Arabidopsis mutant screens map to these two domains, which make up a little more than half of the protein (table S1 and fig. S1). Between nonflowering and flowering plants, there is very little sequence conservation outside the N and C domains. In Arabidopsis, the nuclear LFY protein binds sequences in the enhancers of several floral homeotic genes, including APETALA1 (AP1) (14). The missense changes found in mutant alleles very much reduce in vitro DNA binding to the AP1 promoter (fig. S1). Deletion analyses identify a minimal DNA binding domain from amino acids 320 to 507 (numbering refers to consensus sequence), which includes the highly conserved C domain (Fig. 1A and fig. S1). Although the N domain is not essential for DNA binding, DNA binding is compromised in a deletion derivative that retains part of the N domain, as well as in a protein with a point mutation in the N domain. These observations suggest that the N domain regulates the activity of the DNA binding domain proper (fig. S1).

Fig. 1.

Phylogenetic relationships of LFY sequences. (A) Aligned amino acid sequences of LFY (Arabidopsis), UNI (pea), ALF (petunia), PRFLL (pine), WelNDLY (Welwitschia), CrLFY2 (fern Ceratopteris), and PpLFY1 (moss Physcomitrella) (16). A composite numbering with gaps (dashes) is indicated on top; a gapless count is shown to the right. Unless specified, absolute numbering is used in descriptions. Shaded residues are present in at least 70% of sequences. Green indicates identical residues, and cyan, those with similar biochemical properties. N and C domains are overlined in black and red, respectively. PpLFY1 amino acids that were mutated (Fig. 3) are indicated by arrowheads. (B) Phylogenetic tree of 48 LFY sequences spanning the four major clades of extant land plants. Only bootstrap support above 50% is shown (italic numbers). Color code indicates ability to complement an Arabidopsis lfy mutant (green, full; orange, partial; red, no complementation).

There is generally only a single copy of LFY in angiosperms (Fig. 1B). In species with lineage-specific duplications, these do not seem to have diverged in function (5, 6). Nonflowering plants appear to have additional copies, but these are also all closely related and there is no evidence for major subfunctionalization (7, 8). The strong sequence conservation of the DNA binding domain suggested that the molecular function of LFY is conserved as well. To test this assumption, we linked LFY cDNAs from 14 species to the Arabidopsis LFY promoter and introduced them into a strong lfy mutant. The 14 species represent the three main taxa with known LFY homologs: ferns and mosses, gymnosperms, and angiosperms (Fig. 1B).

Angiosperm genes fully complement lfy mutant, whereas gymnosperm genes provide only partial rescue (Fig. 2A), with PRFLL from pine (9) having more activity in this assay than WelNDLY from Welwitschia (7) (table S3). The LFY and NDLY clades represent a gymnosperm-specific duplication event, with the NDLY lineage having been lost in angiosperms (10). Among homologs from the most basal group, the fern genes CrLFY2 and AlLFY4 (8) have some rescue ability, although less than the gymnosperm genes, whereas the moss genes PpLFY1 and PpLFY2 (8) are inactive. This gradient of complementation reflects the phylogenetic distance from angiosperms (Fig. 1B) and suggests that a continuum of discrete and nonneutral changes, rather than a sudden modification, is responsible for changes in function.

Fig. 2.

Rescue of Arabidopsis lfy mutants by LFY homologs. (A) Scanning electron micrographs of inflorescence meristems (top) and mature flowers (bottom). In wild type (wt) and plants expressing pea UNI, the inflorescence meristem (asterisk) produces floral meristems, which differentiate into flowers composed of sepals (se), petals (pe), stamens (st), and carpels (ca). Plants expressing moss PpLFY1, like nontransgenic lfy mutants, have bracts (br) subtending secondary inflorescences, which replace early-arising flowers (arrowhead). Later-arising flowers consist of organs with mixed sepal and carpel identity (arrowhead) and unfused carpels (ca*). Welwitschia WelNDLY and fern CrLFY2 partially rescue this phenotype. Later-arising flowers of lfy WelNDLY plants have an almost wild-type complement of organs, whereas CrLFY2 flowers have spirally arranged organs that include petaloid stamens (arrowhead). Scale bar is 200 μm. (B) Synthetic Venn diagram of genes induced by Arabidopsis LFY. Ellipses represent the sets of genes induced by the different homologs (see Fig. 1B for color code).

We used microarrays to investigate in more detail the different activities of the homologs. Floral development was synchronized by transfer of plants from short days to long days, which induces LFY promoter activity (11, 12). Of the 16 genes responding most strongly to Arabidopsis LFY, 15 and 13 are significantly induced by the angiosperm homologs UNI and ALF, respectively. WelNDLY, a gymnosperm representative, induces two targets, whereas CrLFY2, a fern gene, induces only one target. None of the LFY targets respond to PpLFY1 from moss (Fig. 2B and fig. S2).

For orthologs of animal HOX proteins, altered activity in cross-species experiments has been traced back to changes outside the DNA binding domain proposed to affect the transcriptional activation or repression potential (1315). To determine whether a similar scenario applies to LFY, we fused LFY homologs to the VP16 activation domain and tested whether they could, like an Arabidopsis LFY-VP16 fusion, induce expression of yeast reporters under the control of LFY binding sites from the homeotic genes AP1 or AGAMOUS (AG) (1, 2). The ability of the different VP16 fusions to interact with the AP1 and AG sites and activate the yeast reporters parallels their rescue activity in plants. Of the proteins from nonflowering plants, PRFLL is as effective as Arabidopsis LFY, whereas WelNDLY and CrLFY2 are substantially less active. Moss PpLFY1, which is inactive in the transgenic plant assay, is also inactive in yeast, indicating that the failure to complement Arabidopsis lfy mutants is not simply caused by a change in transcriptional activation potential (Fig. 3A and fig. S3). Thus, in an a-minima model, declining ability to replace Arabidopsis LFY in plants is caused by a progressive failure to interact with canonical LFY binding sites.

Fig. 3.

Mapping of differences in activity of LFY homologs to DNA binding domain. (A) DNA binding activity of LFY proteins measured in a yeast reporter assay with the use of reporters containing trimerized binding sites from AP1 or AG (1, 22). Error bars indicate standard deviation. (B) In vivo activity of chimeric proteins. Chimeras of Arabidopsis LFY (green) and fern CrLFY2 (orange) are shown as diagrams on top, with conserved N and C domains shown as boxes; sequence identity with Arabidopsis LFY is indicated. Below, partially dissected flowers of lfy mutants expressing the different sequences are compared. Scale bar is 200 μm. (C) DNA binding affinity of LFY, PpLFY1, and PpLFY1 mutants assayed in yeast with the AP1 binding site. (D) Images of lfy plants expressing wild-type PpLFY1 (WT) or PpLFY1 with D394H and C427R substitutions. Organs with petaloid and stamenoid characters are only seen with the D394H mutant. Scale bar is 200 μm.

Because the conserved N and C domains had been implicated in DNA binding, we suspected that the changes in activity are caused by divergence in these two domains, rather than by changes in the surrounding sequences. To test this hypothesis, we swapped the N and C domains between Arabidopsis LFY and CrLFY2 (Fig. 3B). Across the entire sequence, the LFY-CrLFY2 chimera, in which the N and C domains are derived from fern CrLFY2, is more similar to Arabidopsis LFY than the CrLFY2-LFY chimera (77% versus 67% sequence identity to Arabidopsis LFY). Nevertheless, it is the less-similar CrLFY2-LFY chimera that provides almost complete rescue when introduced into Arabidopsis lfy mutants, whereas the other chimera has very little activity, comparable to CrLFY2 itself (Fig. 3B). Thus, changes in the highly conserved N and C domains are responsible for most of the functional differences between the proteins from fern and Arabidopsis. We confirmed that these differences were caused by differential DNA binding activities with the yeast assay described above (fig. S4). Combination of the N and C halves of Arabidopsis LFY and CrLFY2 showed that the C domain, which corresponds to the minimal DNA binding domain, is primarily responsible for the divergence in function (fig. S5).

We found two amino acid substitutions, His394→Asp394 (H394D) and Arg427→Cys427 (R427C) (16), which discriminate between the DNA binding domains of proteins at opposite ends of the functional spectrum, Arabidopsis LFY and moss PpLFY1. We created versions of PpLFY1 in which these positions were individually changed to the angiosperm sequence and tested them again in the yeast assay. PpLFY1 (D394H) but not PpLFY1 (C427R) partially activated transcription (Fig. 3C and fig. S3). Similarly, in transgenic plants PpLFY1 (D394H) but not PpLFY1 (C427R) provided partial LFY activity (Fig. 3D). That a one–amino acid change is sufficient to have PpLFY1 bind a canonical LFY binding site indicates that position 394 is crucial for DNA binding.

The D394H substitution appears to be restricted to true mosses (10) and is not found in the liverworts Riccia and Marchantia. We consider it unlikely that the aspartate (D) at position 394 in moss PpLFY1 simply inactivates the protein, because moss plants lacking both PpLFY genes have dramatic developmental defects (17). We therefore conclude that PpLFY1 has most likely a different DNA binding specificity than its angiosperm counterpart (10). The gradient of activity among nonflowering plants furthermore suggests a systematic change in the DNA binding specificity of LFY homologs.

In animals, the relationship between molecular evolution of developmental regulators and morphological changes is best understood for HOX and PAX homeodomain proteins. Modification of expression patterns can lead to changes in morphology (18), but changes in the proteins themselves may be important as well. Functional differences in the activity of HOX orthologs from different species map to the repression and activation domains rather than the DNA binding domain (13, 14). In contrast, our results emphasize the LFY DNA binding domain as a source of functional variation across species. Examples of molecular evolution of transcription factors in flies and worms that implicated the DNA binding domains as the source of variation have been interpreted as divergence within families of related multicopy genes. Gene duplication followed by subfunctionalization (19), as described for the PAX family (20), is unlikely to apply to LFY, because there has been no radiation of the LFY family during the evolution of land plants.

In a broader perspective of functional evolution of LFY among land plants, two scenarios can be considered. LFY might control similar networks of genes in nonflowering and flowering plants, with coevolution of target sequences and LFY DNA binding specificity. If this is the case, one still needs to postulate that these networks have been modified, because the primary target of LFY in angiosperms, AP1, is restricted to flowering plants (21). Alternatively, there may have been a complete change of LFY function between basal taxa and flowering plants, in which an initial, albeit gradual change in biochemical activity was the prerequisite for recruitment and/or intercalation of new targets, such as AP1.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Tables S1 to S4


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article