Conservation of Endo16 expression in sea urchins despite evolutionary divergence in both cis and trans-acting components of transcriptional regulation

Comparative studies have revealed that the level, timing and spatial expression of genes is subject to change during evolution. In many instances,a change in gene expression has been correlated with a particular change in the phenotype of an organism at an anatomical, physiological or behavioral level (e.g. Dudareva et al.,1996; Sinha and Kellogg,1996; Averof and Patel,1997; Schulte et al.,1997; Stern, 1998; Hariri et al., 2002). However,few studies have examined the molecular mechanisms by which patterns of gene expression have evolved both within and between closely related species. Changes in transcriptional regulation undoubtedly play a central role in generating different patterns of gene expression(Raff, 1996; Doebley and Lukens, 1998; Wray and Lowe, 2000; Carroll et al., 2001; Davidson, 2001). Changes in promoter sequence or in the activity of transcription factors can alter gene expression, which may have functional consequences during development (e.g. Stockhaus et al., 1997; Singh et al., 1998). Many human polymorphisms in promoter sequences affect transcription and are correlated with phenotypic consequences(Rockman and Wray, 2002). Alternatively, changes in transcriptional regulation can serve to maintain patterns of gene expression over evolutionary time scales(Piano et al., 1999; Ludwig et al., 2000).

Studying the evolution of transcriptional regulation requires a system in which one or more promoter sequences have been characterized in detail using biochemical and functional approaches(Wray et al., 2003). Most importantly, this system must be amenable to functional analysis of promoter sequences in multiple, closely related species. To date, relatively few studies have analyzed the functional consequence of evolutionary changes in transcriptional regulation (Franks et al.,1988; Li and Noll,1994; Ludwig et al.,1998; Ludwig et al.,2000; Shashikant et al.,1998; Singh et al.,1998; Crawford et al.,1999; Takahashi et al.,1999; Shaw et al.,2002; Tumpel et al.,2002). In this regard, sea urchins provide an outstanding system in which to study the evolution of transcriptional regulation. Eggs can be obtained in large quantities and develop synchronously upon fertilization,facilitating the collection of material for biochemical analyses. This has enabled researchers to characterize several promoter sequences in exceptional detail including CyIIIa (Calzone et al., 1988; Theze et al.,1990; Wang et al.,1995; Kirchhamer and Davidson,1996; Calzone et al.,1997; Coffman et al.,1996; Coffman et al.,1997) and Endo16 (Yuh et al., 1994; Yuh et al.,1996; Yuh and Davidson,1996; Yuh et al.,1998; Yuh et al.,2001a). Transient expression assays have proven remarkably successful for functional analysis of these promoter sequences in multiple species (reviewed by Kirchhamer et al.,1996). Moreover, the evolutionary history of sea urchins and other echinoderms is well characterized, allowing for interpretation of data in a phylogenetic context (Littlewood and Smith, 1995).

The Endo16 gene was originally isolated from Strongylocentrotus purpuratus by screening a gastrula stage cDNA library (Nocente-McGrath et al.,1989). In S. purpuratus, Endo16 is initially expressed throughout the vegetal plate of the hatched blastula(Nocente-McGrath et al., 1989; Ransick et al., 1993). Endo16 expression is downregulated in primary mesenchymal cells(PMCs) as they migrate away from the center of the vegetal plate to form the larval skeleton. During gastrulation, Endo16 is expressed throughout the invaginating archenteron. Endo16 expression is then downregulated in secondary mesenchymal cells (SMCs) as they migrate away from the anterior tip of the archenteron to form various cell types, including pigment cells,muscle cells and coelomocytes. At the end of gastrulation, Endo16expression is downregulated in the anterior third of the archenteron, which corresponds to the prospective foregut, as well as the posterior third of the archenteron, which corresponds to the prospective hindgut. Endo16expression thereby becomes restricted to the midgut of the pluteus larva.

Transient expression assays demonstrated that 2.2 kb of sequence immediately upstream of the transcriptional start site is sufficient to drive Endo16 expression (Yuh et al.,1994). Approximately 56 sites of specific DNA/protein interactions were mapped within this 2.2 kb region (Yuh et al., 1994) (Fig. 1A). These binding sites are clustered into six functionally distinct modules, which contribute in specific ways to the regulatory output of the Endo16 promoter (Yuh et al., 1996; Yuh and Davidson,1996) (Fig. 1B). The most proximal region of the promoter, module A, activates transcription in the vegetal plate and archenteron. Module B acts synergistically with module A to elevate levels of transcription in these regions. The activity of module A declines during gastrulation, and module B is responsible for maintaining Endo16 expression in the midgut of the pluteus larva. The binding sites responsible for shifting the spatial control of Endo16expression to module B have been identified(Yuh et al., 2001a)(Fig. 1C). The most distal region of the promoter, module G, acts synergistically with modules A and B to increase the rate of transcription by ∼4.2-fold throughout embryonic and larval development. Modules DC, E and F serve to confine Endo16expression to the endoderm: module DC represses transcription in PMCs, while modules E and F repress transcription in ectoderm adjacent to the vegetal plate. Finally, module A serves to communicate the integrated output of all modules to the basal promoter.

The biochemical and functional studies described above, when combined with the experimental advantages of sea urchins, creates an excellent opportunity to analyze promoter evolution. We have therefore characterized the Endo16 promoter from a second sea urchin species, Lytechinus variegatus. Our results reveal a surprisingly strong dissociation between structure and function in this cis-regulatory system and provide insights into the evolutionary mechanisms that have operated on the Endo16 promoter during the past 35 million years.

L. vareigatus adults were collected by Jennifer Keller at the Duke Marine Laboratory (Beaufort, NC) or Susan Decker (Hollywood, FL), and maintained in an aquarium at room temperature. S. purpuratus adults were obtained from Marinus (Long Beach, CA) or Charles Hollahan (Santa Barbara, CA), and maintained in an aquarium at 9°C. Gametes were obtained by injecting adults with 0.55 M KCl. Following fertilization, the eggs were cultured at room temperature (L. variegatus) or 9°C (S. purpuratus) in artificial seawater until the desired stages.

RNA was isolated from gastrula-stage embryos using RNA STAT-60 (Tel-Test“B”, Friendswood, TX) and treated with DNase (Gibco BRL,Gaithersburg, MD). Reverse transcription (RT) was performed according to the instructions provided by the SuperScript Reverse Transcription kit (Gibco BRL). After the addition of a poly(A) tail, the cDNA was used to perform 5′ and 3′ RACE PCR. Primers were based on a partial cDNA sequence previously reported by Godin et al. (Godin et al., 1997) (GenBank Accession Number U89340). PCR products obtained by 5′ and 3′ RACE PCR were gel purified and ligated into pGEM-T vector (Promega, Madison, WI). Plasmid DNA was purified from transformed DH5α cells (Gibco BRL) and sequenced using an ABI Prism 3700 DNA Analyzer (PE Applied Biosystems, Foster City, CA). Sequences were assembled using Sequencher software (Gene Codes, Ann Arbor, MI).

Antisense and sense RNA probes were synthesized according to the instructions provided by the DIG RNA Labeling Kit (SP6/T7) (Roche,Indianapolis, IN) and stored in hybridization buffer (50 ng/μl) at-70°C. Sea urchin embryos were cultured to various stages of development and fixed for 2 hours in a solution containing 2.5% glutaraldehyde, 0.14 M NaCl and 0.2 M phosphate buffer, pH 7.4. The embryos were rinsed twice for∼15 minutes with buffer containing 0.3 M NaCl and 0.2 M phosphate buffer,pH 7.4, and dehydrated through 70% ethanol. Whole-mount in situ hybridization was performed using a protocol based on that of Zhu et al.(Zhu et al., 2001) with several modifications. One important modification was extending the incubation with PBST containing 5% sheep serum to ∼16 hours at 4°C. Images were recorded using a SPOT camera (Diagnostic Instruments, Sterling Heights,MI).

Genomic DNA was isolated from sperm by phenol-chloroform extraction followed by ethanol precipitation. LvEndo16 promoter sequence was obtained according to the instructions provided by the Universal GenomeWalker Kit (Clontech, Palo Alto, CA). In order to extend as far as 2.2 kb upstream of the transcriptional start site, three DNA walks were performed. Two rounds of amplification were performed for each DNA walk using nested primer pairs. Each promoter fragment was cloned and sequenced as described above. It is important to note that the promoter fragments overlapped by at least 50-100 bp. A 2337 bp sequence was assembled from overlapping fragments using Sequencher software. LvEndo16 intron sequence was amplified by PCR using primers flanking the position at which the first intron was predicted to occur based on the S. purpuratus sequence (GenBank Accession Number L34680). The sequence of the 5′ primer was 5′ AATGCGGAAGGAACTTTTTTGCTT and of the 3′ primer was 5′ GAAAGATCAAAGTCGGGAATCAT. The 468 bp product was cloned and sequenced as described above.

Sequences were aligned by ClustalX using default parameters(Thompson et al., 1997). This alignment was not significantly improved by reducing the gap penalty. Sequence similarity was calculated as the frequency of matching nucleotides for various regions of the Endo16 locus, excluding indels (insertions and deletions). At the present time, there are no generally accepted measures of sequence similarity that incorporate indels. Seqcomp analyses were performed to detect a specified number of matching nucleotides (f) in a sliding window of size N in a manner similar to Sonnhammer and Durbin(Sonnhammer and Durbin, 1995). Empirical work by Yuh et al. (Yuh et al.,2002) supports the calculations by Brown et al.(Brown et al., 2002) showing that random matches are expected at or below a 0.7 threshold, but none above 0.75 for a 20 bp window. A seqcomp analysis of the LvEndo16 and SpEndo16 promoter sequences was performed at a threshold (f) of 0.8 and a window size (N) of 20 bp. Seqcomp analyses of the LvEndo16promoter sequence with BAC sequence from S. purpuratus (Sp127I21_S)and of the SpEndo16 promoter sequence with BAC sequence from L. variegatus (Lv199M10_L) also were performed at a threshold (f) of 0.8 and a window size (N) of 100 bp. BAC sequences were obtained from the Sea Urchin Genome Project(http://sugp.caltech.edu:7000/resources/). Results of the seqcomp analyses were visualized on a dot plot and feature map using FamilyRelations (Brown et al.,2002). Similar results were obtained using identical parameters in the mVISTA program developed by Mayor et al.(Mayor et al., 2000) (not shown).

Endo16 promoter sequence was amplified by PCR as a single fragment(2,305 bp, S. purpuratus; 2,159 bp, L. variegatus) from genomic DNA using primers with restriction sites added to their 5′ ends in order to facilitate directional cloning. For S. purpuratus, the sequence of the 5′ primer was 5′GCGCGAATTCGTCGGTGACCTAATTTCCCTTGTT, and of the 3′ primer was 5′GCGCGGATCCCATCGTCTCAAAAATTAG. For L. variegatus, the sequence of the 5′ primer was 5′ GCGCGAATTCGAGCTTGTCAATGAGGGTAATTTT and of the 3′ primer was GCGCGGATCCCGACCAAGCAAAAAAGTTCC. The PCR products were cloned and sequenced as described above. The promoter fragments were excised from the pGEM-T vector (Promega) by restriction digestion with EcoRI and BamHI, and ligated into digested pEGFP-1 vector (Clontech). The ligation products were cloned and sequenced as described above. Promoter constructs were verified by restriction digestions and sequencing using primers based on the pEGFP-1 sequence. Prior to microinjection, the SpEndo16-GFP and LvEndo16-GFP promoter constructs were linearized upstream of the promoter fragment with SacI, and gel purified.

Eggs were de-jellied by incubating in artificial sea water, pH 5.0 for 3.5 minutes (S. purpuratus) or 1.5 minutes (L. variegatus). The eggs were then transferred to plastic petri dishes coated with protamine sulfate. S. purpuratus eggs were fertilized prior to microinjection in artificial sea water containing 0.2% PABA to prevent hardening of the fertilization envelope. Eggs were microinjected using a PLI-100 picospritzer(Medical Systems, Greenvale, NY) under an Axiovert S100 inverted microscope(Zeiss, Jena, Germany). Approximately 1500 molecules of linearized plasmid DNA were injected per egg in a 2 pl volume of solution containing a fivefold molar excess of HindIII-digested genomic DNA, as well as 0.12 M KCl and 30%glycerol. Following microinjection, the L. variegatus eggs were fertilized. Fertilized eggs were cultured at 9°C (S. purpuratus)or room temperature (L. variegatus) until the desired stages. Embryos and larvae were observed under a Axioskop MOT II microscope (Zeiss) equipped for fluorescence microscopy. Images were recorded using a Hamamatsu digital camera (Model #C4742-95-12R) (Hamamatsu City, Japan) and analyzed using Openlab 2.2.4 (Improvision, Lexington, MA). S. purpuratus embryos were cultured at 9°C and therefore, developed more slowly than L. variegatus embryos; however, images were recorded at equivalent developmental stages for both species.

Full-length LvEndo16 cDNA sequence was obtained by 5′ and 3′ RACE PCR using primers based on a partial cDNA sequence previously reported by Godin et al. (Godin et al.,1997). The full-length LvEndo16 cDNA sequence is 4544 bp in length and encodes a protein that consists of 1485 amino acids (data not shown). Whole-mount in situ hybridization was performed using an antisense riboprobe corresponding to nucleotides 1-943 of the coding region. No expression was observed in embryos or pluteus larvae that were hybridized with the corresponding sense riboprobe as a negative control (data not shown).

LvEndo16 is initially expressed throughout the vegetal plate of the hatched blastula (Fig. 2A). LvEndo16 expression is downregulated in PMCs as they ingress into the blastocoel (Fig. 2B). The PMCs lie at the center of the vegetal plate, so that LvEndo16 expression appears as a ring when viewed from the vegetal pole(Fig. 2a,b). During gastrulation, LvEndo16 is expressed throughout the invaginating archenteron (Fig. 2C), and continues to appear as a ring when viewed from the vegetal pole(Fig. 2c). LvEndo16expression is downregulated in SMCs as they migrate away from the anterior tip of the archenteron (Fig. 2D). LvEndo16 expression thus remains restricted to the endoderm throughout gastrulation (Fig. 2C,D). This pattern of Endo16 expression during embryonic development is conserved between S. purpuratus and L. variegatus (Fig. 3).

By the end of gastrulation, LvEndo16 expression is downregulated in the anterior third of the archenteron, the prospective foregut(Fig. 2E). This decline of LvEndo16 expression in the prospective foregut occurs as the archenteron bends to make contact with the oral ectoderm. LvEndo16continues to be expressed in the middle third of the archenteron, the prospective midgut (Fig. 2E). LvEndo16 expression also continues to be expressed in the posterior third of the archenteron, the prospective hindgut(Fig. 2E). By the time that the post-oral arms begin to extend from the pluteus larva, LvEndo16expression in the prospective foregut has completely disappeared(Fig. 2F,G). However, LvEndo16 expression persists in both the midgut and hindgut of the pluteus larva until at least the four-arm stage(Fig. 2H-J). This persistent transcription in the hindgut constitutes a difference in the pattern of Endo16 expression between S. purpuratus and L. variegatus during larval development(Fig. 3).

SpEndo16 expression can be driven by only 2.2 kb of sequence immediately upstream of the transcriptional start site(Yuh et al., 1994). In the present study, 2337 bp of LvEndo16 sequence was assembled from overlapping fragments generated by a series of `walks' upstream of the transcriptional start site (Fig. 4) (GenBank Accession Number AY292383). The LvEndo16promoter sequence then was amplified as a single fragment (∼2.2 kb) that included the basal promoter, and cloned into the promoterless pEGFP-1 vector. The LvEndo16 promoter sequence was inserted upstream of the EGFP gene to create a reporter construct referred to as LvEndo16-GFP.

Microinjection of LvEndo16-GFP into L. variegatus eggs drives GFP expression in a pattern that recapitulates the results of whole-mount in situ hybridization described above(Fig. 2). Fluorescence was consistently observed in a few cells located in the vegetal plate of the hatched blastula (Fig. 5A). These cells contributed to fluorescent patches within the invaginating archenteron (Fig. 5B). Fluorescence was maintained in the midgut of the pluteus larva until at least the four-arm stage (Fig. 5C,D). It is important to note that fluorescence also was observed in the hindgut(Fig. 5D), consistent with the fact that the endogenous gene is expressed in this region of the endoderm in L. variegatus but not S. purpuratus(Fig. 3). Ectopic fluorescence was rarely detected in the ectoderm, PMCs or SMCs. Furthermore, no fluorescence was detected upon microinjection of a promoterless construct containing the EGFP gene into L. variegatus eggs as a negative control. These results indicate that the 2.2 kb upstream fragment contains most or all of the LvEndo16 promoter region.

Microinjection of DNA into sea urchin eggs produces mosaic expression(Arnone et al., 1997). In our hands, this method produced between one and six patches of fluorescent cells per embryo in which fluorescence was detected. We estimate that microinjection of LvEndo16-GFP into L. variegatus eggs produced fluorescence in ∼10% of the resulting embryos. These numbers are smaller than those reported by Arnone et al.(Arnone et al., 1997) in their studies of the sm50 and cyIIa genes in S. purpuratus perhaps because we used a different GFP vector to create fusion proteins. It is also possible that the efficiency of transient incorporation may differ between species. Because of the mosaic incorporation,it is difficult to quantitate the results of these experiments in terms of cell types expressing GFP. In contrast to CAT assays in which the level of transcription within a batch of embryos can be precisely measured, these experiments serve to define the spatial pattern of LvEndo16expression. In this regard, we focused on studying the spatial specificity of cis-regulatory elements, as has been carried out in several previous studies(e.g. Ludwig et al., 1998; Takahashi et al., 1999; Spitz et al., 2001; Tumpel et al., 2002; Yuh et al., 2001b). Future work using CAT reporter constructs will allow us to explore the kinetics of LvEndo16 transcription as was done for the SpEndo16 promoter after its initial characterization by Yuh et al.(Yuh et al., 1994).

Alignment of the Endo16 promoter sequences revealed that module A,the most proximal ∼350 bp of the promoter, is well conserved between S. purpuratus and L. variegatus(Fig. 6). By contrast, upstream modules B through G are not conserved (sequence not shown). Although sequences upstream of module A were difficult to align, it is clear that modules B-G are significantly more divergent than module A. Specifically, module A contains only 11 indels (insertions and deletions), ranging from 1-5 bp in length,whereas the best alignment of modules B through G contains considerably more indels, ranging from 1 to 18 bp in length.

In order to further understand the significance of promoter divergence,sequence similarity was calculated for various regions of the Endo16locus between the two species. Nucleotide identity within module A is 73%,which is comparable with nucleotide identity within the coding sequence. This indicates a similar level of functional constraint on the evolution of these two regions of the locus. As expected, nucleotide identity within binding sites (86%) is higher than within non-binding site nucleotides (69%) of module A. There is a decline in sequence similarity upstream of module A: 55% in module B, and less than 50% within modules DC-G. The first intron, which should be evolving neutrally due to the fact that it contains no functional binding sites (Yuh et al.,1994), has a sequence similarity of 54% (sequence not shown). Thus, modules B-G appear to be evolving neutrally as well.

Surprisingly, none of the binding sites identified within modules B through G of the SpEndo16 promoter can be identified in the LvEndo16promoter, nor in the 5′ UTR, first intron, or coding sequence(Fig. 7A,B). It is important to bear in mind that more than one nucleotide can often fit the consensus sequence for a particular binding site. For example, the SpEndo16promoter contains multiple binding sites for GCF1 and CG. The sequences for many of these binding sites differ slightly within S. purpuratus, but still fall within a well-defined consensus sequence(Yuh et al., 1998). Several programs, including PipMaker (Schwartz et al., 2000), were employed to search for binding sites in the LvEndo16 promoter. Other regions of the locus were also examined in both the 5′ and 3′ orientation, as there can be drastic changes in the order and spacing of binding sites during the evolution of cis-regulatory elements (Wray et al., 2003). It remains possible that variants of binding sites from modules B-G occur within the LvEndo16 promoter, but if so, they have diverged considerably in sequence and perhaps relative position. In any case, such sites were not detected using algorithms to search for consensus sequences based on the SpEndo16 promoter.

These findings are illustrated by a dot plot(Fig. 7C) and a series of feature maps (Fig. 7D-F)generated by FamilyRelations to visualize the results of a seqcomp analysis(Brown et al., 2002). Seqcomp is a relatively new program for comparative analyses that has been optimized for large sequences and can identify conserved sequences of a defined length without regard to spacing or orientation, a capability that is particularly important when examining non-coding regions. First, a pairwise comparison of the SpEndo16 and LvEndo16 promoter sequences was performed using a threshold of 0.8 and a window size of 20. In the case of the dot plot,the LvEndo16 and SpEndo16 promoter sequences are shown on the x- and y-axes, respectively, with regions of aligned sequence indicated as dots. Most of the dots occur in the upper, right corner of the graph, corresponding to module A of the Endo16 promoter(Fig. 7C). In the feature map,the SpEndo16 and LvEndo16 promoter sequences are parallel with one another and red lines indicate regions of conservation. Most of the lines occur at the right end of the feature map, once again corresponding to module A of the Endo16 promoter(Fig. 7D).

To test the possibility that modules B-G are separated from module A by a large insertion in the 5′ flanking region in L. variegatus, we compared the known Endo16 promoter sequences with BAC sequences containing the Endo16 locus. Modules B-G do not appear to be located further upstream of the isolated 2.2 kb sequence in L. variegatus, as evidenced by a pairwise comparison of the SpEndo16 promoter sequence with a ∼22 kb BAC sequence from L. variegatus that contains the LvEndo16 locus. In this case, the analysis was performed using a threshold of 0.8 and a larger window size of 100 in order to avoid noise from repetitive elements. The feature map shows only one region of strong conservation that corresponds to module A of the Endo16 promoter(Fig. 7E). The same parameters were applied to a pairwise comparison of the LvEndo16 promoter sequence with a ∼50 kb BAC sequence from S. purpuratus that contains the SpEndo16 locus. In this case, the feature map shows two regions of conservation that correspond to module A of the Endo16promoter as well as a microsatellite consisting of TAC repeats(Fig. 7F).

To investigate whether there have been evolutionary changes in the set of transcription factors that bind to the Endo16 promoter, reciprocal cross-species transient expression assays were performed. These experiments tested whether the SpEndo16 promoter can drive correct expression in L. variegatus and whether the LvEndo16 promoter can drive correct expression in S. purpuratus. Endo16 promoter sequence from one species (donor) was microinjected into the egg of the other species(host), and GFP expression was observed in the resulting embryos and larvae by fluorescence microscopy. The pattern of GFP expression was interpreted in the context of the expression and sequence data obtained for each species, as well as data from microinjection of the Endo16 promoter into eggs of the same species. As described above, microinjection of LvEndo16-GFP into L. variegatus eggs produced a pattern of GFP expression that recapitulated the results of in situ hybridization(Fig. 8J-L). Microinjection of SpEndo16-GFP into S. purpuratus eggs produced a nearly identical pattern of GFP expression; however, no fluorescence was observed in the hindgut (Fig. 8A-C). This latter result is consistent with studies by Yuh et al.(Yuh et al., 1994). No fluorescence was detected upon microinjection of a promoterless construct into eggs of either species as a negative control.

Microinjection of SpEndo16-GFP into L. variegatus eggs resulted in fluorescence in a few cells located in the vegetal plate of the hatched blastula (Fig. 8G). Patches of fluorescent cells were later observed in the invaginating archenteron (Fig. 8H),consistent with the pattern of Endo16 expression as characterized by in situ hybridization in each species(Nocente-McGrath et al., 1989; Ransick et al., 1993). Fluorescence was maintained in the midgut of the pluteus larva until at least the four-arm stage (Fig. 8I). However, fluorescence was not observed in the hindgut, where Endo16is normally expressed in L. variegatus(Fig. 2H-J). Interestingly,fluorescence was consistently observed in SMCs during gastrulation(Fig. 8H). At later stages of development, fluorescence was restricted to pigment cells(Fig. 8I), one of several cell types that are derived from SMCs (Gibson and Burke, 1985). Ectopic fluorescence was strictly confined to the pigment cells, with no fluorescence detected in the ectoderm, PMCs, or other SMC derivatives. It is important to note that microinjection of the Endo16 promoter into eggs of the same species did not produce ectopic fluorescence in the SMCs or any other cell type.

Microinjection of LvEndo16-GFP into S. purpuratus eggs resulted in a pattern of GFP expression similar to that observed in the reciprocal experiment. Fluorescence was observed in the vegetal plate of the hatched blastula, and later in the invaginating archenteron(Fig. 8D,E). In addition,fluorescence was observed in the midgut of the pluteus larva until at least the four-arm stage (Fig. 8F). Fluorescence was not observed in the hindgut, consistent with the endogenous pattern of SpEndo16 expression. Unlike the reciprocal experiment,ectopic fluorescence was not observed in the SMCs or any other cell type. These data are summarized in Fig. 9.

Our analysis of the Endo16 promoter reveals an unexpectedly complex evolutionary dynamic. Capitalizing on detailed biochemical and functional analyses of the Endo16 promoter in the purple sea urchin, S. purpuratus (Yuh et al.,1994; Yuh et al.,1996; Yuh and Davidson,1996; Yuh et al.,1998; Yuh et al.,2001a), we have analyzed the structure and function of this promoter in a second sea urchin species, L. variegatus. The LvEndo16 cDNA sequence encodes a large 4.6 kb protein with several motifs, suggesting a role in cell adhesion(Soltysik-Espanola et al.,1994). Indeed, experiments using antisense morpholinos indicate that Endo16 may be required for the dynamic changes in cell adhesion that occur during gut morphogenesis (L.A.R. and G.A.W., unpublished). Remarkably,the Endo16 promoter displays a mosaic pattern of evolution, with only module A being conserved between the two species. Reciprocal cross-species transient expression assays indicate that the set of transcription factors that bind to the Endo16 promoter has also diverged to some extent. Nonetheless, LvEndo16 is expressed in a pattern similar to that observed in S. purpuratus, suggesting that stabilizing selection has acted on the transcriptional output of the Endo16 promoter throughout the past 35 million years.

Yuh et al. (Yuh et al.,1994) have demonstrated that Endo16 expression is regulated by 2.2 kb of sequence immediately upstream of the transcriptional start site. This sequence contains at least 56 transcription factor binding sites that are clustered into six functionally distinct modules that regulate the level, timing and spatial transcription of Endo16 in S. purpuratus. We have shown that 2.2 kb of sequence immediately upstream of the transcriptional start site is sufficient to drive Endo16expression throughout embryonic and larval development in L. variegatus as well. Although the pattern of Endo16 expression is similar between S. purpuratus and L. variegatus(Fig. 3), our data demonstrate that drastic changes have evolved in the Endo16 promoter since these two species diverged. Of the entire Endo16 promoter, only the most proximal region, module A, is conserved between the two species(Fig. 7).

These results indicate that different regions within the Endo16promoter are under different levels of functional constraint. Specifically,module A appears to be under a much higher level of functional constraint than the rest of the promoter. It is not surprising that certain modules of the Endo16 promoter are more conserved than others because they perform different functions. Modularity in cis-regulatory sequences allows changes in gene expression to evolve in one tissue independently of another, and has been proposed to facilitate the evolution of morphological diversity (Kitchhamer et al., 1996; Gerhart and Kirschner, 1998; Carroll et al., 2001). Within the Endo16 promoter, the conservation of module A makes functional sense given its essential roles in relaying the integrated output of all modules to the basal promoter and serving as the primary activator of Endo16 expression during embryogenesis(Yuh et al., 1998). Nucleotides within binding sites are more conserved than those not in binding sites presumably because they are directly responsible for activating Endo16 expression. This pattern of functional constraint on binding sites versus non-binding sites has been noted for a few genes (e.g. Core et al., 1997). It is likely that negative selection has maintained functionally important binding sites within module A of the Endo16 promoter since S. purpuratus and L. variegatus last shared a common ancestor.

The pattern of Endo16 expression is similar in S. purpuratus and L. variegatus despite the fact that only module A of the Endo16 promoter is conserved. It has been postulated that selection for compensatory mutations is a primary mechanism by which patterns of gene expression are conserved for long periods of evolutionary time(Ludwig et al., 2000). Several studies provide support for this idea (e.g. Ludwig and Kreitman, 1995; Maduro and Pilgrim, 1996; Tamarina et al., 1997; Ludwig et al., 1998; Piano et al., 1999; Takahashi et al., 1999; Ludwig et al., 2000; Tumpel et al., 2002). Functional compensation appears to have also evolved within the Endo16 promoter, although the changes are more extensive than in any of these previously known cases.

Several pieces of evidence are relevant to understanding the genetic basis for conservation of function despite such divergence in sequences. Yuh and Davidson (Yuh and Davidson,1996) demonstrated that microinjection of a GFP reporter construct containing only module A drives GFP expression in the vegetal plate and archenteron, but is not sufficient to maintain expression in the midgut of the pluteus larva in S. purpuratus(Yuh and Davidson, 1996). Despite the fact that only module A is conserved, the 2.2 kb region immediately upstream of the transcriptional start site of the LvEndo16 gene is sufficient to drive later phases of LvEndo16 expression. It is possible that module A is entirely responsible for the pattern of LvEndo16 expression, although this seems unlikely given its inability to drive larval expression in S. purpuratus. It is also possible that binding sites could not be identified upstream of module A within the LvEndo16 promoter because of unrecognized variation in their consensus sequences. Alternatively, the remaining region of the 2.2 kb region of the LvEndo16 promoter may contain binding sites for a different set of transcription factors that are functionally equivalent to those in modules B-G of the SpEndo16promoter. That is, during the evolution of the Endo16 promoter, some binding sites may have been replaced by others that generate a similar pattern of Endo16 expression. The transcription factors that interact with the Endo16 promoter may have co-evolved to maintain this pattern of Endo16 expression, as has been documented for the bicoidpromoter in insects (Shaw et al.,2002). In any case, the SpEndo16 and LvEndo16promoter sequences are very different, yet generate a similar pattern of Endo16 expression. Although this situation suggests the operation of stabilizing selection, we cannot rule out the possibility that drift or directional selection have been important contributors until data are obtained for additional species.

Although the pattern of Endo16 expression is generally conserved,transcription persists only in the midgut of the pluteus larva in S. purpuratus (Nocente-McGrath et al.,1989; Ransick et al.,1993), but in both the midgut and hindgut of the pluteus larva in L. variegatus. This difference in transcriptional regulation may have evolved in several different ways. The SpEndo16 and LvEndo16promoters may contain binding sites for different transcription factors involved in segmentation of the tripartite gut. Alternatively, the expression and/or activity of these transcription factors may be different between the two species. For example, the transcription factor UI binds within module B of the SpEndo16 promoter, and is directly responsible for maintaining SpEndo16 expression in the midgut of the pluteus larva(Yuh et al., 1998). Although a binding site for the transcription factor UI could not be identified within the LvEndo16 promoter, it is possible that LvEndo16expression persists in the hindgut due to expansion of the spatial domain of UI expression in L. variegatus. Another possibility is the existence of a transcription factor that represses Endo16 expression, and is expressed in the hindgut of S. purpuratus but not L. variegatus.

Binding sites within modules B-G of the SpEndo16 promoter do not appear to be present in any region of the LvEndo16 locus including the 2.2 kb region that was shown to drive the correct pattern of GFP expression (Fig. 7). This result suggests that Endo16 expression is regulated, at least in part, by a different set of transcription factors in S. purpuratusand L. variegatus. Indeed, reciprocal injection of the Endo16 promoter between the two species revealed differences in the expression and/or activity of transcription factors that bind to the Endo16 promoter.

Microinjection of SpEndo16-GFP into L. variegatus eggs,as well as microinjection of LvEndo16-GFP into S. purpuratuseggs, produced fluorescence in the vegetal plate and archenteron(Fig. 9B,D). This result is consistent with the fact that module A is responsible for activating Endo16 expression in these regions(Yuh et al., 1996; Yuh and Davidson, 1996). Moreover, this most proximal region of the Endo16 promoter is conserved between S. purpuratus and L. variegatus. A few nucleotide substitutions and indels occur within known transcription factor binding sites of module A (Fig. 6). Some of these changes occur within multiply represented binding sites for the `structural' protein GCF1, which stabilizes DNA looping(Zeller et al., 1995). However, a few changes occur within binding sites for proteins with a regulatory function. These changes may have been tolerated because they have little or no effect on DNA/protein interactions, a possibility that can be tested with mobility shift assays.

Reciprocal injection also produced fluorescence in the midgut of the pluteus larva (Fig. 9B,D). Yet,module B, which was shown to maintain SpEndo16 expression in this region of endoderm (Yuh et al.,1998), is not present in L. variegatus. Thus, it appears as if changes have evolved within the Endo16 promoter to maintain the regulatory output of module B even in the absence of any obvious sequence similarity. Interestingly, the fact that the SpEndo16 promoter correctly drives GFP expression in the midgut of L. variegatusindicates that the appropriate transcription factors are expressed in both species in a conserved manner. If this were not the case, GFP reporter expression would not mimic the expression of the endogenous gene in reciprocal cross-species microinjection experiments. For example, microinjection of the CyIIIa promoter from S. purpuratus into L. variegatus eggs resulted in ectopic CAT activity in several cell types(Franks et al., 1988).

Fluorescence was not detected in the hindgut upon microinjection of SpEndo16-GFP into L. variegatus eggs(Fig. 9C). Microinjection of LvEndo16-GFP into S. purpuratus eggs also failed to produce fluorescence in the hindgut, despite the fact that LvEndo16 is expressed in this region of endoderm (Fig. 9B). Either the appropriate transcription factors are not present in this region of S. purpuratus, or there has been a change in the activity of co-factors that are required for these transcription factors to bind to the LvEndo16 promoter.

Interestingly, microinjection of SpEndo16-GFP into L. variegatus consistently produced ectopic fluorescence in the SMCs and their descendents, the pigment cells (Fig. 9C). By contrast, microinjection of LvEndo16-GFP into S. purpuratus did not produce ectopic fluorescence(Fig. 9B). These data suggest that L. variegatus and S. purpuratus use different mechanisms to repress Endo16 expression in the SMCs. The transcription factors that normally repress SpEndo16 expression in the SMCs may not be present in L. variegatus. However, any transcription factors that normally repress LvEndo16 expression in the SMCs must be present in S. purpuratus. Alternatively, it is possible that there are no binding sites within the LvEndo16 promoter capable of activating LvEndo16 expression in the SMCs and other nonendodermal cell types.

Thus, it appears as though compensatory changes have evolved that lie both cis and trans to the Endo16 gene. Only a few studies have analyzed promoter sequences in the context of another species to determine the extent to which the corresponding transcription factors have co-evolved(Klueg et al., 1997; Takahashi et al., 1999; Shaw et al., 2002). For example, Takahashi et al. (Takahashi et al., 1999) performed reciprocal injections of the brachyury promoter in two species of ascidians, Ciona intestinalis and Halocynthia roretzi. Extensive changes have evolved in the brachyury promoter, although it activates notochord-specific expression in both species(Corbo et al., 1997; Takahashi et al., 1999). Microinjection of the C. intestinalis brachyury promoter into H. roretzi eggs produced ectopic lacZ expression in other mesodermally derived tissues, suggesting that there have also been alterations in the set of transcription factors that bind to the brachyurypromoter. Most other studies carried out unidirectional analysis of promoter sequences in the context of another species (e.g. Franks et al., 1988; Ludwig et al., 1998; Ludwig et al., 2000; Shashikant et al., 1998), and may therefore have missed finding evidence for trans components to changes in transcriptional regulation.

In summary, this study combines expression, sequence and functional data to analyze changes in cis-regulatory sequences that influence transcription. Data from additional species of sea urchins will help provide a more complete understanding of how changes in transcriptional regulation relate to the evolution of morphological diversity. In addition, site-directed mutagenesis and biochemical assays will allow us to test the functional consequences of specific nucleotide substitutions and indels on Endo16 expression both within and between closely related species.

We thank Cyndi Bradham (Duke University) and members of the Wray laboratory(Jim Balhoff, Chisato Kitazawa, Ann Klatt, Margaret Pizer and Matt Rockman)for their insightful comments on a draft of this manuscript. We are very grateful to Eric Davidson, Andy Cameron and Cathy Yuh (CalTech) for their helpful advice, and for providing us with unpublished data. Finally, we are very grateful to Titus Brown (CalTech) for assisting us with the seqcomp analysis, and Hyla Sweet (Carnegie Mellon University) for advice regarding in situ hybridization. This work was supported by NASA grant NAG-2-1377 and NSF grant IBN-96346 awarded to Gregory Wray.