Conservation and divergence of APETALA1/FRUITFULL-like gene function in grasses: evidence from gene expression analyses

Summary

Duplicated APETALA1/FRUITFULL (AP1/FUL) genes show distinct but overlapping patterns of expression within rice (Oryza sativa) and within ryegrass (Lolium temulentum), suggesting discrete functional roles in the transition to flowering, specification of spikelet meristem identity, and specification of floral organ identity. In this study, we analyzed the expression of the AP1/FUL paralogues FUL1 and FUL2 across phylogenetically disparate grasses to test hypotheses of gene function. In combination with other studies, our data support similar roles for both genes in spikelet meristem identity, a general role for FUL1 in floral organ identity, and a more specific role for FUL2 in outer floral whorl identity. In contrast to Arabidopsis AP1/FUL genes, expression of FUL1 and FUL2 is consistent with an early role in the transition to flowering. In general, FUL1 has a wider expression pattern in all spikelet organs than FUL2, but both genes are expressed in all spikelet organs in some cereals. FUL1 and FUL2 appear to have multiple redundant functions in early inflorescence development. We hypothesize that sub-functionalization of FUL2 and interaction of FUL2 with LHS1 could specify lemma and palea identity in the grass floret.

Introduction

Several members of the MIKC MADS box family proteins are important regulators of inflorescence meristem, floral meristem and floral organ identity (Theissen et al., 1996; Yanofsky et al., 1990), making them prime candidates for involvement in diversification of floral form (Becker and Theissen, 2003). In the model species Arabidopsis thaliana, early-acting genes such as FRUITFULL (FUL), LEAFY (LFY) and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1/AGL20 (SOC1/AGL20) activate downstream targets involved in reproductive development (Blázquez et al., 1997; Ferrándiz et al., 2000; Onouchi et al., 2000; Samach et al., 2000; Simpson and Dean, 2002). Four classes of late-acting target genes (A, B, C and E) then function in combination to determine the identity of organs in each of four concentric floral whorls (Bowman et al., 1989; Coen and Meyerowitz, 1991; Honma and Goto, 2001; Pelaz et al., 2000; Riechmann and Meyerowitz, 1997; Weigel and Meyerowitz, 1994).

The putative A-class protein APETALA1 (AP1) is thought to interact with other MADS box proteins to confer sepal and petal identity (Mandel et al., 1992). However, the potentially confounding role of AP1 in specification of floral meristem identity, in addition to the lack of A-function mutants in core eudicots other than Arabidopsis (but see Benlloch et al., 2006), has led some authors to question the generality of this role across angiosperms (e.g. Schwarz-Sommer et al., 1990; Theissen et al., 2000) and even within Arabidopsis (Litt, 2007; Litt and Irish, 2003). AP1 homeotic mutants have leaf-like organs in the first whorl; these often produce secondary flowers in their axils (Irish and Sussex, 1990). More inflorescence meristems proliferate when ap1 mutations are combined with mutations in paralogous genes CAULIFLOWER (CAL) and FUL. All three genes thus act together during inflorescence development to specify the identity of meristems that give rise to flowers (Ferrándiz et al., 2000; Kempin et al., 1995; Mandel and Yanofsky, 1995). Later expression of AP1 in the sepals and petals may then be required to maintain floral identity (Litt, 2007), or may be involved in specification of outer perianth identity in combination with B- and E-class proteins (Kaufmann et al., 2005; Theissen, 2001; Theissen and Saedler, 2001).

AP1, CAL and FUL are members of the AP1/FUL-like MIKC gene subfamily, a clade of genes that have undergone numerous duplication events throughout the history of angiosperm diversification (Becker and Theissen, 2003; Litt and Irish, 2003; Preston and Kellogg, 2006; Purugganan et al., 1995; Shan et al., 2007). The ancestor of AP1 and CAL appears to have been duplicated during the whole-genome duplication at the base of Brassicaceae, while the ancestral gene of AP1/CAL and FUL duplicated around the origin of the core eudicots (Litt and Irish, 2003). Likewise, a whole-genome duplication at or near the base of grasses (Paterson et al., 2004; Yu et al., 2005) gave rise to two clades of grass AP1/FUL genes, which we have called FUL1 and FUL2 (Figure 1). FUL1 and FUL2 in rice (OsMADS14 and OsMADS15, respectively) are on chromosomes 3 and 7 (coordinates 30991977-30981554 and 475854-469473; http://www.tigr.org/tdb/e2k1/osa1/overview.shtml). These fall in regions identified as segmental duplications (http://www.tigr.org/tdb/e2k1/osa1/segmental_dup/500kb/segdup_500kb.shtml), providing genomic evidence that the gene duplication identified by the phylogeny corresponds to a genome-wide duplication event. Patterns of codon evolution and gene expression suggest diversification and partial redundancy of function of FUL1 and FUL2 following their duplication (Preston and Kellogg, 2006). This history of duplication events suggests that few AP1/FUL-like genes are truly orthologous across plant families, precluding one-to-one extrapolation of function between related genes (Malcomber et al., 2006). We are interested in exploring conservation and divergence of expression between AP1/FUL-like gene homologues of Arabidopsis and grasses, specifically to determine the role of FUL1 and FUL2 in inflorescence and floral development.

Figure 1.  Phylogram of AP1/FUL-like genes from taxa included in this study, showing that the duplication of FUL1 and FUL2 occurred around the base of grasses (based on Preston and Kellogg, 2006).

Grass inflorescences are branched structures that produce a variable number of short branches called spikelets. Differences in spikelet determinacy, organ number, and organ identity, size and shape are important to define taxonomic groups (Kellogg, 2000) and to determine crop yield. Each spikelet is composed of one or two leaf-like glumes, bearing one [e.g. barley (Hordeum vulgare)] to several [e.g. tef (Eragrostis tef)] florets (Clifford, 1987). Depending on the species, florets can be bisexual, unisexual or reduced. Bisexual florets consist of an outer leaf-like lemma and palea, two or three fleshy lodicules (modified petals) (Ambrose et al., 2000; Whipple et al., 2004, 2007), one or two whorls of stamens, and a gynoecium. Whereas lodicules, stamens and gynoecia can be compared directly with corresponding structures in other angiosperms, lemmae, paleae, glumes and the spikelet as a whole have no obvious counterpart within flowering plants (but see Whipple and Schmidt, 2006).

AP1/FUL-like genes in the grasses are expressed much earlier than those in the eudicots. Also, unlike AP1/FUL in eudicots, the genes in grasses are an early signal of the transition to flowering, as shown in diploid and hexaploid wheat (Triticum monococcum and T. aestivum, respectively; Danyluk et al., 2003; Trevaskis et al., 2003; Yan et al., 2003), H. vulgare (Danyluk et al., 2003; Trevaskis et al., 2003, 2006; Von Zitzewitz et al., 2005) and perennial ryegrass (Lolium perenne; Petersen et al., 2004, 2006). FUL1 (also known as VRN-1) is upregulated in leaves (Danyluk et al., 2003; Yan et al., 2003) and apices (Petersen et al., 2004; Yan et al., 2003) before the emergence of spikelet primordia, and antisense FUL1 wheat plants have significantly delayed flowering (Loukoianov et al., 2005; Murai et al., 2003; Trevaskis et al., 2003). FUL1 and FUL2 are also expressed in spikelet and floral primordia (Gocal et al., 2001; Greco et al., 1997), suggesting a further role in specifying spikelet and floral meristem identity.

In the few grass species investigated to date, expression of grass AP1/FUL-like genes does not fit with the classical A-function model from Arabidopsis. Instead, expression is broader than in just outer whorl organs. In Darnel ryegrass (Lolium temulentum), expression of both LtFUL1 (LtMADS1) and LtFUL2 (LtMADS2) is restricted to glumes, lemma, palea and possibly lodicules late in the development of spikelets, suggesting a redundant role for these genes in bract and perianth identity, and in meristem identity (Gocal et al., 2001). Expression of either gene under the AP1 promoter in ap1-15 Arabidopsis mutants results in some plants with normal sepals (LtFUL1) or plants with increased organ numbers and reduced axillary flowers (LtFUL2; Gocal et al., 2001). These different phenotypes may indicate overlapping but distinct roles for each gene in floral organ identity and floral meristem identity. Similarly, different roles in floral organ identity have been proposed for OsFUL1 (OsMADS14) and OsFUL2 (OsMADS15) in rice (O. sativa). OsFUL1 is expressed in all floral organs except lodicules (Moon et al., 1999; Pelucchi et al., 2002), whereas OsFUL2 is, like the Lolium genes, expressed only in the outer bracts and perianth (lodicules). Yeast two-hybrid experiments further demonstrate exclusive protein–protein interactions between OSFUL1 and grass SEP-like proteins OSMADS5 and OSMADS8, both of which are expressed in reproductive whorls (Cooper et al., 2003; Moon et al., 1999), consistent with an extended role for OsFUL1 in stamen and carpel organ identity (Malcomber and Kellogg, 2005).

Maize (Zea mays) has two FUL2 genes, which we call ZmFUL2a (ZmMADS3) and ZmFUL2b (ZAP1). Both are expressed throughout glumes, lemma and palea (Mena et al., 1995), and ZmFUL2a is only expressed in stamens at late stages of development, following organ differentiation (Heuer et al., 2001). Constitutive expression of ZmFUL2a results in undifferentiated floral organs in male spikelets and reduced male inflorescence branching (Heuer et al., 2001). This phenotype is attributed to interference with other MADS box proteins, for example through inappropriate dimerization at particular stages of development. This interference hypothesis predicts that FUL2 expression must be switched off during certain stages of floral meristem differentiation, or that FUL2 and interacting proteins should be co-expressed only in specific floral organs.

In this paper, we describe expression patterns of paralogous FUL1 and FUL2 genes among distantly related grass species, to test whether either or both might play a role in inflorescence meristem, spikelet meristem and floral organ identity. Taxa were chosen to represent the major lineages of the grass family. Based on previous studies summarized above, it has been suggested that (i) both genes are involved in the transition to flowering, (ii) both specify spikelet meristem identity, (iii) FUL2 specifies lemma, palea and lodicule identity (A-class function), and (iv) FUL1 specifies identity of all floral organs (E-class function). Each hypothesis predicts a particular expression pattern.

If both genes have a role in the transition to flowering, we expect gene expression in the shoot apical meristem during or just before the shift in morphology to an inflorescence meristem, and in tissues distinct from those destined to develop spikelet or branch meristems. If both genes confer spikelet meristem identity, we predict expression of both in spikelet meristems. If FUL2 specifies lemma, palea and lodicule identity (A-function), we expect the gene to be expressed in these organs. Finally, if FUL1 specifies all floral organs (E-function), we predict expression in all organs.

Results

FUL1 and FUL2 are expressed in inflorescence, inflorescence branch, and spikelet meristems

The transition from a juvenile to adult vegetative and then inflorescence apical meristem is marked by elongation of the shoot apical meristem in T. monococcum, oat (Avena sativa), green millet (Setaria viridis), sorghum (Sorghum bicolor) and Eleusine indica. The inflorescence meristem then produces either spikelet (T. monococcum) or branch (A. sativa, E. indica, S. bicolor, S. viridis) meristems on its flanks. In all species examined, FUL1 and FUL2 mRNA transcripts were most abundant in the apex of the shoot meristem (Figure 2a,b,e), but were also expressed in the developing spikelet or branch primordia (Figure 2). Gene expression was not detectable along the rachis (main inflorescence axis) or spikelet pedicels of any species (Figure 2). Because expression of both genes was identical, only expression for FUL1 is shown. In all cases, probes could distinguish between gene copies, as determined by Southern blot hybridizations (Figure S1). Furthermore, there was little or no hybridization with sense control probes for all genes tested (data not shown).

Figure 2.  FUL1 mRNA expression in the apex and branch or spikelet primordia of developing inflorescence meristems across distantly related grass species. (a) AsFUL1 expression in Avena sativa. (b) TmFUL1 expression in Triticum monococcum. (c) EiFUL1 expression in Eleusine indica. (d) SbFUL1 expression in Sorghum bicolor. (e) SvFUL1 expression in Setaria viridis. sam, shoot apical meristem; bm, branch meristem; sm, spikelet meristem; lf, leaf. Bars = 100 μm.

Comparison of FUL1 and FUL2 gene expression patterns in spikelets

Avena sativa and Triticum monococcum (Pooideae) have inflorescences comprised of pedicillate and sessile spikelets, respectively, with each spikelet containing 2–8 florets subtended by two glumes. Proximal florets in the spikelet are bisexual, and include a lemma, palea, two lodicules, three stamens and one gynoecium. Distal florets are generally under-developed and lack one or more floral organs. FUL1 (AsFUL1, TmFUL1) and FUL2 (AsFUL2, TmFUL2) of both species show the same pattern of expression within spikelets. Both genes are expressed at the apex of the spikelet meristem and throughout floret meristems (Figure 3). Following floral organ initiation, FUL1 transcripts are found in all floret organs of both complete and reduced florets (Figure 3a,b,e,f). At the same stage, FUL2 is expressed in all floret organs except carpels (Figure 3c,d,g–i). In stamens, expression of both genes is confined to the filaments and endothecial layers (not shown). In glumes, gene expression is strongest during early stages, becoming reduced after maturation (Figure 3).

Figure 3.  FUL1 and FUL2 mRNA expression in pooid grasses. (a) Triticum monococcum inflorescence with developing spikelets; TmFUL1 is expressed in all spikelet organ primordia. (b) Developing T. monococcum spikelet with floret meristems distal to differentiated florets; TmFUL1 is expressed in floral meristems, all floral organs and glumes. (c) T. monococcum inflorescence with developing spikelets; TmFUL2 is expressed in all spikelet organ primordia. (d) Complete floret of T. monococcum with TmFUL2 expression in all floral organs. (e) Spikelet of Avena sativa showing AsFUL1 gene expression in floret meristems and all floret organ primordia. (f) A. sativa spikelet; AsFUL1 is expressed in all developing floral organs. (g) Spikelet of A. sativa showing AsFUL2 gene expression in the floret meristem, lemma, palea and stamens. (h) Young A. sativa spikelet; AsFUL2 is highly expressed in stamens, but at a low level, if at all, in the gynoecium. (i) A. sativa spikelet; AsFUL2 is expressed in glumes, lemma, palea, lodicules and stamens. gl, glume; le, lemma; pa, palea; lo, lodicules; st, stamens; gy, gynoecium; sm, spikelet meristem; fm, floral meristems. Asterisks indicate floret meristems. Bars = 100 μm.

Chasmanthium latifolium (Centothecoideae) has pedicillate spikelets, each containing 4–24 florets subtended by two glumes. Reduced florets are found both proximally and distally to complete florets. Proximal reduced florets consist solely of sterile lemmas. Complete florets are bisexual, comprising a lemma, palea, two lodicules, one stamen and one gynoecium. In early development, both C. latifolium FUL1 (ClFUL1) and FUL2 (ClFUL2) are expressed in glumes (not shown), floret meristems, and in the spikelet apex (Figure 4a,c). ClFUL1 is expressed in all floral organ primordia at early stages of development, and in maturing paleae, lodicules, stamens and gynoecia (Figure 4a,b). No expression is detectable in maturing lemmae or mature glumes (Figure 4b). ClFUL2 is expressed in early lemmae and paleae, but is undetectable in lodicules, stamens or gynoecia (Figure 4c,d). Following organ maturation, ClFUL2 expression is also absent from lemmae and glumes (Figure 4d).

Figure 4.  FUL1 and FUL2 mRNA expression in centothecoid and chloridoid grasses. (a) Developing Chasmanthium latifolium spikelet; ClFUL1 is expressed in spikelet and floral meristems, and in lemma and palea primordia. (b) C. latifolium spikelet showing ClFUL1 expression in all floral organs except maturing lemmae. (c) Developing C. latifolium spikelet; ClFUL2 is expressed in spikelet meristems, and lemma and palea primordia. (d) Developing C. latifolium spikelet with floret meristems distal to differentiated florets; ClFUL2 is expressed in all paleae and young lemmae only. (e) Young Eleusine indica spikelet; EiFUL1 is expressed in the spikelet meristem, lemmae and paleae. (f) Spikelets of E. indica with developing floret meristems distal to differentiated florets; EiFUL1 is expressed in the floret meristem (marked with an asterisk), paleae and early lemmae. (g) Close-up of anther stage E. indica florets; EiFUL1 is expressed in paleae, but not well-developed lemmae, lodicules or stamens. (h) Young E. indica spikelet; EiFUL2 is expressed in the spikelet meristem and lemmae. (i) Spikelets of E. indica with developing floret meristems distal to differentiated florets; EiFUL2 is expressed in paleae and early lemmae. gl, glume; le, lemma; pa, palea; lo, lodicules; st, stamens; gy, gynoecium; sm, spikelet meristem; fm, floral meristems. Asterisks indicate floret meristems. Bars = 100 μm.

Eleusine indica (Chloridoideae) has sessile spikelets containing 3–15 florets subtended by two glumes. Reduced florets are found distal to complete florets. Complete florets comprise a lemma, palea, two lodicules, three stamens and one gynoecium. Expression of E. indica FUL1 (EiFUL1) and FUL2 (EiFUL2) is similar throughout spikelet development (Figure 4e–i). In early developing spikelets, gene expression is observed in glume primordia (not shown) and spikelet meristems (Figure 4e,h). Later in development, both genes are expressed in floret meristems and the outer organs of the floret (lemma and palea). Neither gene transcript is detectable in lodicules, stamens, carpels or maturing glumes (Figure 4f,g,i).

Setaria viridis (Panicoideae) inflorescences bear pedicillate spikelets, each subtended by one to several undifferentiated branches (bristles; see Doust and Kellogg, 2002). Each spikelet comprises an upper complete floret and a lower reduced floret, both subtended by two glumes. The complete floret has a lemma, palea, two lodicules, three stamens and one gynoecium. The expression patterns of S. viridis FUL1 (SvFUL1) and FUL2 (SvFUL2) are distinct. SvFUL1 is expressed in glumes, and all floral organs of the upper and lower florets (Figure 5a,b). SvFUL2 is also expressed in glumes, lemmae, paleae and lodicules, but is undetectable in stamens or gynoecia (Figure 5c–e). Neither gene is expressed in bristles (Figure 5a–e).

Figure 5.  FUL1 and FUL2 mRNA expression in panicoid grasses. (a) Young Setaria viridis spikelet; SvFUL1 is expressed in the upper and lower floret meristems, but not in bristles. (b) S. viridis inflorescence with developing spikelets; SvFUL1 is expressed in all spikelet organ primordia. (c) Young S. viridis spikelet; SvFUL2 is expressed in lemmae and paleae, but comparatively weakly, if at all, in floral meristems. (d) S. viridis inflorescence with developing spikelets; SvFUL2 is expressed in glumes, lemmae, paleae and lodicules, but not in stamens, carpels or bristles. (e) Developing S. viridis spikelet; gene expression in the lower glume, lower floret meristem, lemma/palea, and lodicules of upper floret. (f) Sessile (upper) and pedicillate (lower) spikelets of Sorghum bicolor, showing SbFUL1 gene expression in all floral organs of the upper and lower florets. (g) Sessile spikelet of S. bicolor; SbFUL1 is expressed in the lemma, palea, lodicules and stamens. (h) Sessile spikelet of S. bicolor; SbFUL2 is expressed in outer floral organs only. gl, glume; le, lemma; pa, palea; st, stamens; gy, gynoecium; fm, floral meristem; br, bristle. Bars = 100 μm.

Sorghum bicolor (Panicoideae) inflorescences contain pedicillate–sessile pairs of spikelets, each containing two florets subtended by two glumes. Both florets of the pedicillate spikelet are reduced, whereas only the lower floret of the sessile spikelet is reduced to a sterile lemma. The complete upper floret of the sessile spikelet is bisexual, with a lemma, palea, two lodicules, three stamens and one gynoecium. The expression patterns of S. bicolor FUL1 (SbFUL1) and FUL2 (SbFUL2) differ (Figure 5f–h). SbFUL1 is expressed in glumes, and all floral organ primordia of pedicillate and sessile spikelets (Figure 5f). Expression appears to be maintained in these organs as they mature (Figure 5g). SbFUL2 is also expressed in glumes, paleae and young lemmae, but not in lodicules, stamens, gynoecia or maturing lemmae (Figure 5h).

Reconstruction of ancestral AP1/FUL expression before and after gene duplication

Maximum parsimony ancestral state reconstruction (Maddison and Maddison, 2003) immediately prior to the AP1/FUL duplication estimates the ancestral AP1/FUL gene as expressed in spikelet meristems, glumes, outer spikelet organs and stamens, with the ancestral state for gynoecia being equivocal (Figure 6). Following gene duplication, FUL1 and FUL2 expression was conserved in spikelet meristems, glumes, lemmae and paleae across grasses. All changes in gene expression occurred within the lodicules and reproductive organs, and are estimated as losses of expression.

Figure 6.  Parsimony ancestral state reconstruction of FUL1 and FUL2 expression in spikelets across grasses. The tree topology shows the duplication event at the base of grasses giving rise to paralogous gene lineages FUL1 (upper) and FUL2 (lower). Speciation events following the evolution of grasses resulted in orthologous genes within each clade. These relationships track the species phylogeny. FUL1 gene expression is shown in red, FUL2 in blue, and the hypothetical ancestor of FUL1/FUL2 in green (gray in the gynoecium is ambiguous). Black organs denote no gene expression in terminal taxa or the hypothesized loss of gene expression along particular branches. Large circles at the end of terminal branches represent conserved expression in spikelet/floral meristems where known. Spikelet organs from outside to inside are glumes (outer left and right curved lines), lemma (lower curved line), palea (upper ridged curved line), lodicules (small curved line), stamens (figures of eight) and gynoecium (inner circle). Organ numbers are not conserved across taxa.

FUL1 expression in lodicules is lost independently along the lineages leading to O. sativa and E. indica. Likewise, FUL1 expression in stamens is also lost in the lineage leading to E. indica. Expression of FUL2 in lodicules and stamens was lost once prior to the split between E. indica (Chloridoideae), C. latifolium (Centothecoideae) and the S. bicolor–S. viridis subfamily (Panicoideae). Loss of FUL2 expression in stamens is also estimated for the lineages leading to O. sativa and L. temulentum. The direction of change in gynoecial expression cannot be determined.

Discussion

In this study, we analyzed the expression of recently duplicated genes FUL1 and FUL2 across phylogenetically disparate grasses to test hypotheses of gene function. Specifically, we tested the hypotheses that FUL1/FUL2 genes are involved in the transition to flowering and have a redundant role in specifying the identity of spikelet meristems, that FUL2 has strict A-class function in specifying the identity of lemma, palea and lodicules, and that FUL1 has a general E-class function, contributing to the identity of all floral organs. In combination with other studies, our data support multiple developmental roles for FUL1 and FUL2 early in inflorescence and late in floral organ development, but reject a strict role for FUL2 as an A-class gene. We discuss these results in the context of a general model of grass spikelet development, particularly by comparing expression patterns of FUL1/FUL2 with those of phylogenetically and functionally related genes from the same grass species (LHS1, Figure 7).

Figure 7.  Heterogeneous expression patterns of FUL1, FUL2 and LHS1 in spikelets of diverse grass taxa. Only taxa with expression data for all three genes are shown. Expression of FUL1 is shown in red (upper boxes), that of FUL2 in blue (middle boxes) and that of LHS1 in green (lower boxes). White boxes indicate no expression. GL, glumes; LE, lemma; PA, palea; LO, lodicules; ST, stamens; GY, gynoecium.

Expression of FUL1 and FUL2 in inflorescence meristems suggests a role in the transition to flowering in all grasses

In this study, we found strong expression of FUL1 and FUL2 in T. monococcum, A. sativa and E. indica inflorescence meristems, consistent with roles for both genes in the transition to flowering. This developmental role is distinct from specification of spikelet or floral meristem identity. The inflorescence meristem of T. monococcum and A. sativa ultimately becomes converted to a spikelet, but the inflorescence meristem of S. viridis does not. Thus strong FUL1/2 expression in the inflorescence meristem is not simply a harbinger of spikelet formation. In both T. monococcum (Shitsukawa et al., 2006) and L. temulentum (Gocal et al., 2001), FUL1 and FUL2 mRNAs are abundant in the inflorescence meristem, whereas the putative spikelet meristem identity genes WHEAT FLORICAULA/LEAFY (WFL) and LtLFY, respectively, are expressed only within lateral regions of the inflorescence meristem, presumably corresponding to areas destined to become spikelet meristems. In the non-grass monocot Dendrobium grex (Orchidaceae), the AP1/FUL-like gene DOMADS2 is detectable in the shoot apical meristem following the morphological phase change to reproductive development but before initiation of floral primordia (Yu and Goh, 2000).

Identical FUL1 and FUL2 expression in floret meristems is consistent with a conserved role in floral meristem identity

In all grasses examined, FUL1 and FUL2 were expressed in floret meristems. While loss-of-function mutants are unavailable in the grasses, the expression pattern is consistent with a conserved role for these genes in floral meristem identity.

Mutations in the related Arabidopsis genes AP1, CAL and FUL cause loss of floral meristem identity and proliferation of inflorescence meristems in positions normally occupied by developing flowers. Likewise, mutations in AP1/FUL-like genes of other core eudicots, including SQUAMOSA of snapdragon (Antirrhinum majus;Carpenter et al., 1995; Huijser et al., 1992), PEAM4 of pea (Pisum sativum;Berbel et al., 2001) and MtPIM of Medicago truncatula (Benlloch et al., 2006), result in replacement of flowers with leafy shoots. Thus, functional studies in the core eudicots and expression studies in the grasses suggest that the role of AP1-like and FUL-like genes in floral meristem identity has been conserved since the origin of the entire AP1/FUL-like gene family.

Unlike eudicots and other monocots, initiation of grass floral meristems is preceded by the formation of spikelet meristems. Each spikelet meristem gives rise to one or two bracts (glumes), and then initiates one to several floral meristems, each giving rise to a single flower. In all grasses examined, FUL1 and FUL2 were expressed in spikelet meristems.

Expression patterns support a possible role for FUL1 in floral organ identity (E-class function)

Based on studies in O. sativa and L. temulentum, FUL1 was hypothesized to play a general role in floral organ identity (E-class function) (Cooper et al., 2003; Lim et al., 2000; Malcomber and Kellogg, 2005; Moon et al., 1999). Our data support this hypothesis. However, FUL1 cannot always be necessary for specification of inner floral whorls as E. indica lacks FUL1 expression in lodicules, stamens and pistils. In addition, the basal grass Pharus lacks a functional copy of FUL1 but still produces typical spikelets (Preston and Kellogg, 2006). It is possible that other MADS box proteins (e.g. FUL3 or SEP-like) with similar protein-binding affinities replace FUL1 in the inner floral whorls of E. indica and all whorls of Pharus.

Similarities in expression domains between FUL2 and AP1 are convergent

Some of the earliest AP1/FUL-like genes to be investigated in the grasses were those of Lolium and Oryza (Gocal et al., 2001; Moon et al., 1999; Pelucchi et al., 2002), in which expression of FUL2 occurs in the lemmae, paleae and lodicules, similar to AP1 in eudicots. Our expanded sample, however, reveals that this similarity is in fact an evolutionary convergence. In most grasses, FUL2 expression does not fit the pattern expected for eudicot-like A-class function. FUL2 is not expressed in lodicules of S. bicolor, C. latifolium or E. indica, and thus cannot be involved in specification of inner perianth/lodicule identity. Furthermore, wider expression than predicted is found in the subfamily Pooideae (represented by A. sativa and T. monococcum), with FUL2 being expressed in every organ of the spikelet except carpels.

Gene expression throughout the spikelet (except possibly carpels) appears to be the ancestral state prior to the FUL1/FUL2 duplication. This is consistent with evidence from monocots such as African blue lily (Agapanthus africanus), tiger lily (Lilium lancifolium) and Virginia spiderwort (Tradescantia virginiana), where AP1/FUL genes are expressed in all four floral whorls (Preston and Kellogg, 2006). Our analysis also suggests that general expression throughout the spikelet (except carpels) is ancestral for FUL2, indicating that loss of expression domains in various lineages occurred within the evolutionary history of the grasses.

Interaction of FUL2 with LHS1 could specify lemma and palea identity

FUL2 is consistently expressed in glumes, lemmae and paleae, but the pattern is variable within inner organs (lodicules, stamens and carpels). This variable pattern of expression is analogous to that of the closely related grass SEP-like gene, LEAFY HULL STERILE1 (LHS1), which is expressed in the lemma and palea of all grasses investigated, is always absent from glumes, and is highly variable within lodicules, stamens and carpels (Malcomber and Kellogg, 2004; Reinheimer et al., 2006).

When the expression patterns of FUL2 and LHS1 are compared, both are co-expressed in lemmae and paleae, but they are never expressed simultaneously in glumes, lodicules, stamens or carpels. This is true for all grasses for which we have complete expression data (O. sativa, A. sativa, E. indica, C. latifolium and S. bicolor; Figure 7) (Malcomber and Kellogg, 2004; Moon et al., 1999; Pelucchi et al., 2002; Prasad et al., 2001; Reinheimer et al., 2006; this study). Together these grasses represent a divergence period of approximately 50 million years (Wolfe et al., 1987). In contrast, although FUL1 genes are also co-expressed with FUL2 and LHS1 in lemmae and paleae, they are often co-expressed with LHS1 in non-bracteate organs of the inner floral whorls (Figure 7).

We know that LHS1 and FUL proteins can interact. Yeast two-hybrid studies in O. sativa have demonstrated protein–protein interactions between LHS1 (expressed in lemma/palea) and both OsFUL1 (expressed in all but lodicules) and OsFUL2 (expressed in lemma/palea and lodicules) (Cooper et al., 2003; Lim et al., 2000). OsFUL1 also forms dimers with the rice SEP-like proteins OsMADS5, 7 and 8, whereas OsFUL2 only dimerizes with OsMADS7 (Cooper et al., 2003). In addition, in O. sativa, mutations in LHS1 result in leafy lemma and palea (Jeon et al., 2000; Prasad et al., 2005), so LHS1 function is necessary for lemma/palea identity.

MADS box proteins function as dimers or higher-order complexes (De Folter et al., 2005; Egea-Cortines et al., 1999; Immink et al., 2002; Krizek and Meyerowitz, 1996). Grasses have more MADS box genes than Arabidopsis and other eudicots, allowing many more possibilities for protein interactions (Malcomber and Kellogg, 2004). Thus, protein interactions, rather than protein identity per se, may be important for determining the identity of grass organs in general.

Implications for inflorescence development and evolution of modified outer whorl organs

The AP1/FUL-like gene family of transcription factors has a complex history of redundancy and neo-functionalization following duplication events within several angiosperm lineages. The few studies in basal angiosperms, basal eudicots and monocots suggest that the ancestral expression pattern of AP1/FUL-like genes was throughout flowers (Kim et al., 2005; Litt and Irish, 2003; Preston and Kellogg, 2006). We suggest, as have others, that the ancestral role for AP1/FUL-like genes was floral meristem identity and floral organ development, with the classical A-function (perianth identity) being derived in core eudicots.

In contrast, early expression of FUL1/FUL2 in inflorescence meristems in grasses is novel. We suggest that the deployment of these genes in the transition to flowering is derived in grass. Finally, distinct patterns of expression in spikelets support functional divergence of the genes in specifying floral organ identity.

Based on evidence from previous studies and discrete co-expression of LHS1 and FUL2 in lemma/palea across grasses (Malcomber and Kellogg, 2004; Reinheimer et al., 2006; this study), we propose a model for the evolution of outer floral organ morphology resulting from gene duplication. The duplication event giving rise to the FUL1 and FUL2 gene lineages occurred at or near the base of grasses. Models of sequence evolution suggest that changes along branches leading to FUL1 and FUL2 occurred under relaxed selection, an outcome predicted by functional redundancy (Preston and Kellogg, 2006). However, all non-synonymous codon changes occurred solely along the branch leading to FUL2, and a quarter of these are conserved across grasses (Preston and Kellogg, 2006). We previously suggested that this pattern of codon substitution is consistent with neo- or sub-functionalization of FUL2.

Because AP1/FUL genes are transcription factors, protein sequence evolution may change protein-binding affinities, downstream targets, or both. Yeast two-hybrid experiments in O. sativa suggest that OsFUL2 interacts with many of the proteins (LHS1, SEP3, AGL6, FUL1, FUL2) (Cooper et al., 2003; Lim et al., 2000; Moon et al., 1999) predicted from similar studies of Arabidopsis AP1 (Honma and Goto, 2001; Pelaz et al., 2001). However, this is only a subset of proteins that interact with OsFUL1. OsFUL1 also interacts with the grass SEP-like proteins OSMADS5 and OSMADS8, neither of which is expressed in lemma or palea (Cooper et al., 2003; Moon et al., 1999). Thus, in terms of protein-binding affinity, FUL2 may be sub-functionalized to interact with proteins expressed within outer floral organs. This may explain why ectopic expression of ZmFUL2a in spikelet organs leads to lack of organ differentiation in whorls also expressing LHS1 (Cacharrón et al., 1999), i.e. all organs except glumes (Heuer et al., 2001).

Functional analyses of grass AP1/FUL genes expressed in the Arabidopsis ap1-15 mutant also indicate functional divergence in regulation of downstream genes (Gocal et al., 2001). Expression of LtFUL2 from L. temulentum under the A. thaliana AP1 promoter rescues ap1-15 floral organ number in outer floral whorls, but these organs are bract-like. In contrast, the same construct with L. temulentum LtFUL1 does not rescue floral organ number, but occasionally produces normal sepals (Gocal et al., 2001). These data suggest that FUL2 controls downstream regulators involved in outer floral organ development, but not sepal identity, whereas FUL1 interacts with proteins in all whorls to promote organ development, without specifying organ identity. We propose that FUL2 interacts with LHS1 to specify outer floral organ identity, and that evolution of FUL2 has led to the modification of sepals/outer tepals to bract-like lemmae and paleae.

Experimental procedures

Plant materials

Material from Avena sativa (CIav9401Ogle), Chasmanthium latifolium (Malcomber3117), Eleusine indica (PI217609), Setaria viridis (PI204624), Sorghum bicolor (Malcomber3116) and Triticum monococcum (PI427802) was collected from plants grown at the University of Missouri – St Louis under standard greenhouse conditions. Seeds were obtained from the United States Department of Agriculture (USDA) or from pre-existing seed stocks.

Southern blot hybridization

To determine the specificity of probes for in situ hybridization, Southern blot hybridization was carried out. Total DNA was extracted from 250 mg of leaf tissue according to the method described by Doyle and Doyle (1987). Approximately 10 mg of total DNA was digested with the enzymes BamHI, EcoRI, HindIII, and NcoI or PstI, run on a 0.8% agarose gel overnight. Digested DNA was blotted onto a nylon membrane and hybridized with ³²P-dCTP-labeled in situ hybridization probes (see below) for 16 h at 65°C. To remove non-specific binding of the probes, each membrane was washed twice with 0.1x wash buffer at 65°C, according to the method described by Laurie et al. (1993).

In situ hybridization

Tissue from various developmental stages of vegetative and inflorescence development was fixed in FAA (47.5% v/v ethanol, 5% v/v acetic acid, 3.7% v/v formaldehyde; Sigma; http://www.sigmaaldrich.com/) using vacuum infiltration, and dehydrated into paraffin wax according to the method described by Jackson (1991). Ribbons of 8 μm sections were cut, mounted on Probe-On-Plus microscope slides (Fisher Scientific; http://newfishersci.com/wps/portal/HOME), and left to dry at 37°C overnight.

Gene-specific probe templates were generated using a nested approach. The first round of PCR used primers complementary to the 5′ end of the C-terminus [AP1.1.1 (5′-GAGAAGCAGAAGGCCCA-3′) for FUL1, or AP1.2.1 (5′-GAACTKKYRGAGAGGCAGAAGGC-3′) for FUL2] and the 3′ end of the 3′ UTR (polyT-adaptor, 5′-CCGGATCCTCTAGAGCGGCCGCTTTTTTTTTTTTTTTTT-3′). The sec round of PCR used primers complementary to the 5′ end of the 3′ UTR (ZAP-FUL3, 5′-ATGGATGCTGAGCMMGYTC-3′) and the 3′ end of the 3′ UTR. Probes were approximately 300 bp in length and showed approximately 35% nucleotide difference between the two AP1/FUL copies. Each reaction was run for 30 cycles with an annealing temperature of 55°C on an MJ Research PTC-200 thermocycler (GMI Inc.; http://www.gmi-inc.com). PCR products were gel-purified through a QIAquick spin column (Qiagen Inc.; http://www.qiagen.com/), sub-cloned into the pGEM-T easy vector (Promega Corp.; http://www.promega.com/), and sequenced to determine identity and orientation. Sense and antisense riboprobes were generated using T7 and SP6 Megascript in vitro transcription kits (Ambion; http://www.ambion.com) with digoxygenin-labeled UTP (Roche; http://www.roche.com) according to the manufacturer’s instructions. Probe hydrolysis was performed according to the method described by Jackson (1991). Probe hybridization, washing, immunolocalization and photography were performed according to the methods described by Jackson et al. (1994) and Malcomber and Kellogg (2004). Photographs were imported into Adobe Photoshop and adjusted for contrast, brightness and color balance.

Acknowledgements

We would like to thank the Missouri Botanical Garden and the United States Department of Agriculture for access to plant materials, Simon Malcomber (California State University, Long Beach, USA) for help with in situ hybridizations, and Robert Marquis (University of Missouri, St Louis, USA), Robert Schmidt (University of California, San Diego, USA) and Peter Stevens (University of Missouri, St Louis, USA), for helpful comments on an earlier version of this manuscript. This work was supported by National Science Foundation grant DBI-0110189 to E.A.K.