Translational Biology: From Arabidopsis Flowers to Grass Inflorescence Architecture

One of the key events in plant development is the initiation of lateral organs from the flanks of the meristem. In grasses, the inflorescence meristem (IM) reiteratively initiates a series of lateral meristems with slightly different fates. Our understanding of the genes and networks that regulate grass inflorescence architecture has dramatically expanded due to significant advances in resources and tools. Many of the modules that regulate meristem fate in Arabidopsis (Arabidopsis thaliana) are also present in the grasses. Genetic networks that regulate IM size and floral organ fate are partially conserved between Arabidopsis and grasses, whereas genetic networks that regulate grass-specific meristems are either unique to grasses or have different functions in dicots.

Grasses have complex inflorescence architecture, with multiple meristem types. Maize (Zea mays) and rice (Oryza sativa) are the most well-studied grasses, and mutants that affect discrete stages of inflorescence development have been identified. In rice, the IM produces several primary branch meristems (BMs) before terminating. These primary branches initiate secondary branches, and ultimately spikelet meristems (SMs), which will form florets (Fig. 1A). Maize has two inflorescences: the ear, which arises in the axil of a vegetative leaf, and the tassel, located at the apex of the plant. The tassel IM first initiates several lateral meristems that become long branches and later switches to producing spikelet pair meristems (SPMs). SPMs are transient and produce a pair of SMs. SMs are also transient and produce sterile leaves, called bracts, and two floral meristems (FMs). The ear is unbranched and does not produce BMs (Fig. 1B). The SM is a meristem type unique to all grasses (Clifford, 1987).

Figure 1.

Grass inflorescence development. A, Inflorescence development in rice. The IM initiates primary branch meristems (PBMs), which initiate SMs. Each SM initiates a single FM. PBMs also initiate secondary branch meristems (SBMs). The PMB is converted to a terminal spikelet meristem (TSM). B, Inflorescence development in maize. The IM initiates SPMs, which form two SMs. Each SM initiates two FMs. Early stages of development in the tassel and ear, except, in the tassel, the IM also initiates BMs. Left, Diagram of a developing ear; right, schematic of maize inflorescence development. C, Diagram of a grass floret.

MERISTEM IDENTITY AND MAINTENANCE

Meristem identity and maintenance are regulated by the CLAVATA (CLV) and KNOX gene pathways. In Arabidopsis, signaling through the CLV pathway restricts expression of WUSCHEL (WUS), which defines the stem cell niche (Laux et al., 1996). In the absence of CLV1, CLV2, or CLV3, WUS expression expands and meristem size is increased (Brand et al., 2000; Schoof et al., 2000). Similar regulation is likely to occur in maize and rice. Mutations in the maize genes thick tassel dwarf1 (td1) and fasciated ear2 (fea2), which encode orthologs of CLV1 and CLV2, respectively, have larger IMs and increased spikelet pair density on the main rachis of the tassel and in the ear. Male flowers also produce more stamens, indicating that these genes also likely regulate FM size. Interestingly, the regulation of meristem size in vegetative meristems is apparently different than in IMs; both td1 and fea2 make fewer leaves, suggesting that they positively regulate meristem size in vegetative development (Taguchi-Shiobara et al., 2001; Bommert et al., 2005). In rice, mutants in the CLV1 and CLV3 homologs, FLORAL ORGAN NUMBER1 (FON1) and FON2, respectively, yield similar phenotypes, although the defect is restricted to the FM (Suzaki et al., 2004, 2006). Another CLV3-like gene, FON2-LIKE CLE PROTEIN1 (FCP1), regulates vegetative meristem size, suggesting FON2 and FCP1 have functionally diversified to regulate different meristem types (Suzaki et al., 2008). WUS belongs to a large gene family with members in rice and maize; however, a WUS homolog with similar expression or function has not yet been identified in grasses (Nardmann and Werr, 2006).

In Arabidopsis, meristem maintenance requires SHOOTMERISTEMLESS (STM; Barton and Poethig, 1993). Strong stm alleles do not make an inflorescence, but weak alleles indicate that stm also functions in FMs (Lenhard et al., 2001; Kanrar et al., 2006; Scofield et al., 2007). A maize STM homolog with a similar function is knotted1 (kn1). Both STM and kn1 are expressed in all shoot meristems, but excluded from leaf primordia (Smith et al., 1992; Long et al., 1996). Maize kn1 mutant phenotypes are similar to Arabidopsis stm mutants; in some inbred backgrounds, kn1 mutants result in shootless seedlings, whereas in other inbreds, kn1 mutants result in decreased inflorescence branching (Vollbrecht et al., 1991, 2000; Kerstetter et al., 1997). OSH1 in rice is orthologous to maize kn1, with a similar expression pattern, but no known mutant phenotype.

REGULATION OF INFLORESCENCE BRANCHING

A class of maize mutations, called ramosa (ra), increases branching in the tassel and ear. Typically, mutant spikelet pairs are replaced by long branches with multiple spikelets and display additional reiterations of branching. ra1 encodes a zinc-finger transcription factor that is expressed at the base of SPMs (Vollbrecht et al., 2005). ra3 encodes a trehalose-P phosphatase with an overlapping expression domain (Satoh-Nagasawa et al., 2006). ra2 encodes a LATERAL ORGAN BOUNDARY (LOB) transcription factor that is expressed prior to lateral meristem initiation and thus prior to both ra1 and ra3. ra2 is also expressed at the position where BMs and SMs initiate (Bortiri et al., 2006), suggesting that it is not unique to spikelet pairs. ra2 has a similar expression pattern in other grasses (Bortiri et al., 2006), but no known mutant phenotypes. ra1 expression is decreased in ra2 and ra3 mutants, suggesting that both genes function to regulate ra1. Given the uniqueness of the spikelet pair to maize (and other members of the Andropogoneae) and the lack of ra1 in rice, a species lacking spikelet pairs (Vollbrecht et al., 2005), it may be that LOB genes function to specify spikelet pairs due to the unique contribution of ra1.

In Arabidopsis, the timing of the IM to FM transition is critical to determine inflorescence architecture. The timing of this transition is controlled by the antagonistic activities of two genes, LEAFY (LFY), which promotes FM identity (Weigel et al., 1992), and TERMINAL FLOWER1 (TFL1), which promotes an indeterminate state (Bradley et al., 1997). LFY and TFL homologs also play a role in inflorescence architectures in grasses. RNAi knockdowns of the rice LFY homolog, RFL, severely decrease panicle branching (Rao et al., 2008). Similarly, maize double mutants in the LFY co-orthologs zfl1 and zfl2 exhibit reduced tassel branching. In addition, zfl1 and zfl2 are required to promote proper floral organ identity and phyllotaxy (Bomblies et al., 2003). In Arabidopsis, TFL1 overexpression increases inflorescence branching (Ratcliffe et al., 1998), and similar phenotypes are observed when the rice TFL1 homologs RCN1 and RCN2 are overexpressed in Arabidopsis or rice, indicating a conserved role in the regulation of inflorescence architecture (Nakagawa et al., 2002).

SM IDENTITY AND DETERMINACY

In grasses, the SM initiates two bracts, called glumes, and a variable number of FMs. In maize, the SM initiates two FMs, and in rice, the SM initiates a single FM. Two types of APETALA2 (AP2) domain-containing transcription factors, ERF and AP2, regulate SM identity and determinacy in maize and rice. The AP2 domain is a DNA-binding domain (Ohme-Takagi and Shinshi, 1995), and ERF proteins contain one AP2 domain, whereas AP2-like proteins contain two (Sakuma et al., 2002; Magnani et al., 2004). Mutants in orthologous ERF genes, branched silkless1 (bd1) in maize and FRIZZY PANICLE1 (FZP1) in rice, exhibit defects in SM identity and determinacy. bd1 mutants initiate extra spikelets in the tassel, and in the ear, SMs are replaced with BMs (Chuck et al., 2002). Similarly, mutants in rice FZP1 form ectopic branches in place of spikelets (Komatsu et al., 2003).

Little is known about how bd1 and FZP1 regulate SM identity. Interestingly, bd1 and FZP1 mRNA are not expressed in the SM itself, but instead are expressed in a semicircular pattern at the base of the SM (Chuck et al., 2002; Komatsu et al., 2003). How, then, do bd1 and FZP1 impart SM identity? One possibility is that a secondary signal, such as a hormone or sugar, is a mobile signal that moves into the SM from the base where bd1/FZP1 is expressed. Alternatively, FZP1 and bd1 might not affect SM identity per se, but rather repress the formation and outgrowth of ectopic axillary meristems in the axil of the glume (Chuck et al., 2002; Komatsu et al., 2003).

Regulation of SM determinacy by AP2 transcription factors in maize involves two related genes, indeterminate spikelet1 (ids1) and sister of ids1 (sid1; Fig. 2). ids1; sid1 double mutants do not make FMs, but instead initiate many bract-like organs, eventually terminating in an ovule-like structure that is probably not the product of a FM (Chuck et al., 2008). Thus, ids1 and sid1 are necessary for FM initiation. In rice, mutants in the ids1-like gene SUPERNUMERARY BRACT (SNB) resemble ids1; sid1 double mutants: They initiate extra bract-like organs and delay the transition of SM to FM (Lee et al., 2007). Notably, snb mutants do eventually make a FM.

Figure 2.

MIR172 regulates ids1 and sid1 to control FM initiation. A, MIR172 negatively regulates ids1 and sid1, which promote FM fates. In addition, IDS1 negatively regulates sid1 mRNA accumulation. B, In wild type (left), MIR172 restricts IDS1 expression, and two FMs are formed. In Ts6 or ts4 mutants (right), IDS1 expression is increased and expanded, resulting in extra FMs. LFM, Lower floral meristem.

IDS1/SID1 is also sufficient to promote the SM to FM transition. Two mutants that increase IDS1 expression, tasselseed4 (ts4) and Ts6, initiate extra florets (Chuck et al., 2007). During normal development, MIR172 restricts the domain and level of IDS1/SID1 expression and thus restricts the number of FMs. ts4 encodes a MIR172 family member that negatively regulates AP2 genes, including ids1 and sid1 (Chuck et al., 2007, 2008). Ts6 harbors a mutation in the MIR172 complementarity site of ids1 mRNA, resulting in increased IDS1 expression (Chuck et al., 2007). Thus, up-regulation of IDS1, either by removing its negative regulator or rendering the ids1 mRNA immune to negative regulation, results in extra florets. Adding to the complexity, IDS1 negatively regulates sid1 mRNA accumulation, which might explain why ids1 single mutants initiate extra florets (Chuck et al., 2008). Interestingly, ts4 and Ts6 mutants also fail to abort carpels in the tassel, indicating that the sex determination and SM determinacy pathways intersect, although the relationship between these to pathways is unclear (see below).

FLORAL ORGAN IDENTITY AND PATTERNING

One of the most significant advances in the past 20 years in plant biology is the ABC model of floral development. This simple and elegant model posits that transcription factors act in a combinatorial manner to achieve floral patterning (Coen and Meyerowitz, 1991). The major class A, B, and C genes have been identified and, except for the class A gene AP2, encode members of the MIKC type of MADS-box transcription factors. The model has been expanded to include class D genes, which promote ovule development and class E, or SEPALLATA (SEP), genes, which act as cofactors to A, B, C, and D class genes (for review, see Theissen, 2001). The basic ABCDE model is depicted in Figure 3A. A major challenge for grass biologists is determining how this model applies to monocot flowers, which display unique floral morphologies compared to dicots.

Figure 3.

ABC model of floral development. The basic ABC model posits that class A genes specify whorl 1 organs, class A and B genes specify whorl 2 organs, class B and C genes specify whorl 3 organs, and class C genes specify whorl 4. The expanded ABCDE model includes class D genes that promote ovule identity and class E genes that act as cofactors for the class A, B, C, and D genes. Class A genes are depicted in green, class B in orange, class C in blue, class D in purple, and class E in yellow. Solid colors indicate functional data; shaded colors indicate that function is hypothesized based on expression data or phylogenetic analysis. Color gradients illustrate subfunctionalization of duplicated genes. A, Model of floral development in Arabidopsis. B, Model of floral development in rice. C, Model of floral development in maize. See text for details.

The grasses possess a unique floral structure, the floret (Fig. 1B). Florets contain carpels and stamens, like their dicot counterparts; however, they lack petals and sepals. Surrounding the sex organs are lodicules, and two bract-like organs, the palea and lemma. Lodicules are thought to correspond to petals in dicots (see below). The corresponding dicot organs to palea and lemma, however, remain controversial; palea and lemma may represent unique grass structures.

Forward and reverse genetic approaches have identified several genes required for floral development in grasses. Not surprisingly, some of the genes identified correspond to B, C, D, and E class genes in dicots. However, forward genetics has also identified a number of floral regulators that do not have a functional dicot counterpart and appear to have unique functions in grass floral development.

In Arabidopsis, the class A genes AP1 and AP2 specify the outer two whorls, sepals and petals. AP1 has an additional role in promoting the transition to flowering. AP1 homologs have been identified in grasses, and, despite the lack of mutants, the available data do not support a role in floral patterning. AP1 homologs are expressed in the FM of diverse grass species, consistent with a function in transition to flowering, as in Arabidopsis. The general expression of AP1 throughout the spikelet is thought to be ancestral and thus is inconsistent with strict class A function (Preston and Kellogg, 2006). In addition, expression of a Lolium AP1 homolog in Arabidopsis ap1 mutants does not rescue ap1 organ identity defects (Gocal et al., 2001). Thus, AP1 genes do not appear to have strict class A function in grasses. However, true class A mutants have not been defined outside of Arabidopsis and the roles of AP1 and AP2 in floral organ identity may not be as clear as predicted by the ABC model (see Litt and Irish, 2003; Preston and Kellogg, 2006).

In Arabidopsis, the class B genes AP3 and PISTILLATA (PI) specify whorl 2 (with class A genes) and whorl 3 (with class C genes) organs. In contrast to class A genes, the function of at least one class B gene is clearly conserved between dicots and the grasses. Mutants in two AP3 homologs, silky1 (si1) in maize and SUPERWOMAN1 (SPW1) in rice, result in homeotic transformations of stamens to carpels and lodicules to palea-like organs (Ambrose et al., 2000; Nagasawa et al., 2003). Thus, AP3-like genes are required to promote whorl 2 and 3 identities in grasses. The homeotic transformation of lodicules to palea-like structures supports the hypothesis that lodicules correspond to whorl 2 organs or petals in dicots. Furthermore, the maize genes si1 (AP3) and zmm16 (PI) have similar biochemical activities as their Arabidopsis counterparts and can rescue the corresponding mutants in Arabidopsis, providing additional evidence of class B conservation (Whipple et al., 2004). Two rice genes, OsMADS2 and OsMADS4, are similar to the other class B gene, PI; however, OsMADS2 appears to be more important in whorl 2 and OsMADS4 in whorl 3. RNAi knockdowns of OsMADS2 affect lodicule development, but do not affect stamen development (Prasad and Vijayraghavan, 2003; Yadav et al., 2007). Furthermore, RNA expression patterns are consistent with OsMADS2 functioning in lodicule development, and OsMADS4 functioning in stamen development (Yadav et al., 2007).

A single class C gene, AGAMOUS (AG), specifies whorl 3 (with class B genes) and whorl 4 organs in Arabidopsis. In addition to its role in floral organ identity, AG also promotes FM determinacy. In grasses, the AG gene has been duplicated and the two class C functions have largely been subfunctionalized to separate genes. For example, rice contains two AG homologs, OsMADS3 and OsMADS58. OsMADS3 mutants transform stamens to lodicules, but have only minor defects in FM determinacy. In contrast, OsMADS58 RNAi knockdowns have only minor defects in floral organ identity, but greatly affect FM determinacy (Yamaguchi et al., 2006). Thus, OsMADS3 has a central role in floral organ identity, whereas OsMADS58 has a central role in FM determinacy. Similarly, mutants in the maize agamous homolog zag1 (for zea agamous1) lack FM determinacy, but do not affect floral organ identity (Mena et al., 1996). Mutants in the zag1 duplicate zmm2 have not been identified, but zmm2 expression patterns are consistent with class C function (Mena et al., 1996). Notably, available mutants do not support a role for grass class C genes in carpel identity.

Class D genes specify ovule identity. Cosuppression of FBP7 and FBP11 in Petunia transforms ovules into carpelloid structures (Angenent et al., 1995), and overexpression of FBP11 results in ectopic ovules on sepals and petals (Colombo et al., 1995). In Arabidopsis, shatterproof1 (shp1) shp2 seedstick1 (stk1) triple mutants also transform ovules into carpelloid structures (Pinyopich et al., 2003), indicating these genes function redundantly to promote ovule identity. SHP1 and SHP2 are more closely related to AG, and STK1 groups with the class D genes of Petunia. Rice contains two putative class D genes, OsMADS13 and OsMADS21. osmads21 mutants do not have a mutant phenotype; however, osmads13 mutants convert ovules into carpelloid structures, indicating that class D function is in part conserved between dicots and the grasses. osmads13 mutants also make excess carpels, indicating that OsMADS13 also plays a role in FM determinacy (Dreni et al., 2007).

In Arabidopsis, SEP1 to SEP4 function redundantly as class E genes. Class E genes function as cofactors with class A, B, and C genes, and in the absence of all four SEP genes, floral organs are transformed into leaf-like structures (Ditta et al., 2004). leafy hull sterile1 (lhs1), which encodes the SEP-like gene, OsMADS1, is the only reported class E mutant in grasses. lhs1 mutants produce fewer stamens and transform lemma, palea, and lodicules into leaf-like structures (Jeon et al., 2000). OsMADS1 RNAi knockdowns have a more severe phenotype in which floral organs in all four whorls are transformed into leaf-like structures (Prasad et al., 2005). The grass SEP lineage is complex and gene expression patterns are variable, suggesting that grass SEP-like genes potentially fulfill more diverse developmental functions than in Arabidopsis (Malcomber and Kellogg, 2004).

Non-MADS-box genes also play key roles in floral development. Two rice mutants, drooping leaf1 (dl1) and aberrant panicle organization1 (apo1), resemble class C mutants, suggesting that they regulate class C genes or that they have some class C function. DL1 is a candidate carpel identity gene in rice; dl1 mutants convert carpels to stamens (Nagasawa et al., 2003). dl1 encodes a YABBY transcription factor, most similar to CRABSCLAW (CRC) in Arabidopsis (Yamaguchi et al., 2004). CRC also has a role in carpel and ovule development, but, unlike DL1, does not play the central role in carpel identity (Bowman and Smyth, 1999). DL1 and the class B gene, SPW, are mutually antagonistic, and this antagonism is critical to set up the boundary between whorls 3 and 4. No carpel identity genes have been identified in maize.

The apo1 mutant also phenotypically resembles class C mutants. apo1 mutants make extra lodicules at the expense of stamens, suggesting stamens are converted to lodicules. In addition, apo1 mutants make extra carpels, implicating apo1 in FM determinacy, another class C function. Consistent with this phenotype, expression of the class C gene, OsMADS3, is reduced in apo1 mutants, indicating that APO1 positively regulates class C gene expression (Ikeda et al., 2005, 2007). APO1 encodes an F-box protein, similar to UNUSUAL FLORAL ORGANS (UFO) in Arabidopsis (Ikeda et al., 2007). In contrast to APO1, UFO is required to activate class B genes (Lee et al., 1997). Thus, whereas UFO and APO both play key roles in floral development and are likely to have similar biochemical functions, their roles in the floral regulatory network are distinct.

SEX DETERMINATION

Grasses exhibit a variety of sexual systems, including bisexual and unisexual flowers. Plants that make unisexual flowers are most commonly monoecious (male and female flowers on the same plant, but separate inflorescences) or dioecious (male and female flowers on separate plants). Maize is monoecious and the only grass for which significant genetic and molecular data on sex determination exist.

In maize (and in other grasses with unisexual flowers), flowers are initially bisexual, but carpel and stamen primordia arrest in male and female flowers, respectively. The rich history of maize genetics has produced a collection of sex determination mutants, giving insight into the molecular regulation of this process. In ts1, ts2, Ts3, ts4, Ts5, and Ts6, carpels do not abort in the tassel (Veit et al., 1993; Irish et al., 1994). Mutants that affect GA synthesis, including anther ear1 and the dwarf mutants, do not abort stamens in the ear (Emerson and Emerson, 1922; Bensen et al., 1995). Finally, carpels inappropriately abort in ears of silkless1 (sk1) mutants (Jones, 1925).

ts2, ts4, and Ts6 have been cloned. ts2 encodes a short-chain dehydrogenase (DeLong et al., 1993), but the substrate for this enzyme and how it functions in the molecular regulation of sex determination are unknown. As discussed above, ts4 encodes MIR172, and Ts6 harbors a mutation in the MIR172 complementarity site of ids1. ts1 is necessary for ts2 expression, and sk1 is thought to protect carpel primordia in the ear from ts2-mediated cell death (Calderon-Urrea and Dellaporta, 1999). Malcomber and Kellogg (2006) identified ts2 homologs in various grasses and asked whether specific variants were associated with unisexual floral development. However, they found that ts2 homologs likely have a broader role in programmed cell death and are not specific to sex determination. Indeed, no single sex-determining gene is likely to be responsible for the variety of sexual systems observed in the grasses, and experiments will have to be conducted in multiple species.

Interestingly, several ts mutants affect branching as well as sex determination, including Ts3, ts4, and Ts6. While we have some insight into the role of ts4 and Ts6 in branching (Chuck et al., 2007), we do not know downstream targets that are important in sex determination. GA treatment mimics the ts phenotype (Nickerson, 1959) and is also critical for stamen arrest in the ear. A more complete understanding of sex determination will require cloning additional sex determination genes, and a challenge for the future is understanding how these how these modules link together into a coherent regulatory network to regulate sex determination.

REGULATORY NETWORKS AND INFLORESCENCE ARCHITECTURE

Developmental biology has long focused on identifying individual genes and studying their role in development. Genome-scale experiments in models such as Arabidopsis permit us to assemble these genes into pathways and even networks that control developmental processes. Recently, large-scale experiments have yielded similar data in other species, such as maize and rice, which have distinct inflorescence morphologies. Inflorescence morphology is determined by the architecture of the underlying gene regulatory network and differences in morphologies reflect the differences in the network architecture (Prusinkiewicz et al., 2007). Taking into consideration the broader themes of network biology provides insight into inflorescence development and the evolution of different morphologies.

A network is defined as the connections between nodes (Barabasi and Oltvai, 2004); nodes can be genes, cells, systems, etc. Cellular networks are scale free; that is, a few nodes have many connections, called hubs, but most nodes have only a few connections (Barabasi and Oltvai, 2004). Hubs are more likely to appear early in the history of the network (Barabasi and Albert, 1999). Thus, hubs are likely to be conserved between species. One implication of scale-free networks is that disruption of most nodes does not perturb the network; however, disruption of a hub greatly perturbs the network. Forward genetic screens are likely to uncover hub genes that are conserved in the regulatory networks of different species. Because the hubs are likely to be the same, morphological diversity must be rooted in the connections between hubs and other nodes.

Transcriptional control of floral development by MADS-box proteins provides an example of how different interactions between nodes can lead to morphological diversity. The protein-protein interactions of MADS-box proteins have been intensively studied in Arabidopsis (de Folter et al., 2005). While a similar endeavor has not been undertaken in a grass, data suggest that some interactions are conserved, while others are novel (see Malcomber and Kellogg, 2005). The composition of specific protein complexes determine DNA-binding specificity and downstream target genes. A key to understanding different floral morphologies will be determining the protein interaction network and how it maps onto an underlying transcriptional network.

Another key aspect of networks is modularity. Regulatory modules from one part of the network can be moved to another part of the network where they fulfill novel functions. An excellent example in the inflorescence is MIR172 regulation of AP2 genes. In Arabidopsis, this regulatory module is required for floral organ identity and FM size (Zhao et al., 2007), whereas in maize it is critical for sex determination and SPM determinacy (Chuck et al., 2007, 2008).

Maize has recently undergone whole-genome duplication (Swigonova et al., 2004) and thus provides a unique opportunity to study how gene duplication events affect regulatory networks. Duplicated genes, or nodes, provide redundancy and therefore network robustness. For example, in maize, ids1 and sid1 function redundantly to promote the SM to FM transition. In addition to redundancy, node (gene) duplication can also lead to subfunctionalization. These duplicated nodes will eventually form unique connections to other nodes. In Arabidopsis, the class C gene, AG, provides both class C functions, whereas in maize and rice, AG has been duplicated and no single gene provides all class C function.

Experiments in Arabidopsis have provided a framework to understanding developmental processes in other species. However, in different species, new modules arise for new structures (e.g. the ra genes), modules attain new functions, and nodes form new connections. Indeed, regulatory networks vary even within species, as evidenced by the dramatic phenotypic variations of some maize mutants (e.g. kn1) in different inbred backgrounds. Ultimately, comparisons of regulatory networks both within and between species will aid in our understanding of evolution of different morphologies.