A Gene Regulatory Network Subcircuit Drives a Dynamic Pattern of Gene Expression

See allHide authors and affiliations

Science  02 Nov 2007:
Vol. 318, Issue 5851, pp. 794-797
DOI: 10.1126/science.1146524


Early specification of endomesodermal territories in the sea urchin embryo depends on a moving torus of regulatory gene expression. We show how this dynamic patterning function is encoded in a gene regulatory network (GRN) subcircuit that includes the otx, wnt8, and blimp1 genes, the cis-regulatory control systems of which have all been experimentally defined. A cis-regulatory reconstruction experiment revealed that blimp1 autorepression accounts for progressive extinction of expression in the center of the torus, whereas its outward expansion follows reception of the Wnt8 ligand by adjacent cells. GRN circuitry thus controls not only static spatial assignment in development but also dynamic regulatory patterning.

The genomic regulatory code that controls the specification of the future skeletogenic, gut endoderm, and nonskeletogenic mesodermal components of the sea urchin embryo is embodied in a gene regulatory network (GRN). The GRN states the interactions of about 50 genes encoding transcription factors, as determined in an extensive perturbation analysis along with other data (1, 2). The subcircuits of this network control the establishment of transient regulatory states in the spatial domains of the developing embryo. Here we consider a GRN subcircuit, the function of which is to direct a dynamically expanding ring or torus of regulatory gene transcription early in sea urchin embryogenesis. Transcription of the torus regulatory genes begins at the vegetal pole of the egg in the newly born fourth-cleavage micromeres. These cells give rise to the skeletogenic lineages of the embryo. Transcription of the earliest torus genes starts at about 6 hours after fertilization, then extends to the adjacent ring of mesodermal blastomeres in the early blastula stage (12 hours), and finally encompasses the precursor cells that will generate the gut just before mesenchyme blastula stage (>18 hours) (Fig. 1A). However, within a few hours after these genes are first activated, their expression is extinguished, first in the skeletogenic domain and then in the mesodermal domain.

Fig. 1.

Moving-torus gene expression pattern. (A) Representation of expression pattern of blimp1 or wnt8 genes (red). The innermost cells are skeletogenic micromeres; the red ring in the second drawing shows mesoderm cells (prospective secondary mesenchyme); the outer ring is definitive endoderm. Expression of the blimp1 gene begins in the micromeres around 6 hours after fertilization and appears in the adjacent tier of mesodermal cells by 12 hours. Soon after, expression disappears from the micromeres. By 18 hours, expression of blimp1 begins in the adjacent presumptive endoderm lineage and disappears from the mesodermal cells. (B) GRN subcircuit including otx, blimp1, and wnt8 genes; blimp1b indicates the early isoform of the blimp1 gene (5); nb-TCF, complex of nuclear β-catenin and TCF transcription factor; GSK3, enzyme normally responsible for β-catenin clearance, the activity of which is inhibited as a consequence of reception of the Wnt8 signal ligand. In the absence of nuclearized β-catenin, a Groucho/TCF complex forms instead (12) and acts as a dominant repressor at both the wnt8 and blimp1 loci (dark blue barred stems). nb-TCF is inhibited from forming by GSK3, the biochemical mechanism of which is symbolized by the solid circle. Positive inputs from Blimp1 and nb-TCF control wnt8 transcription (4), whereas both nb-TCF and Otx are required for blimp1 expression; blimp1 is subject to autorepression via two Blimp1 target sites. (C) Expression of wnt8 and blimp1, visualized by WMISH. By fifth cleavage, wnt8 transcript is evident in the four micromeres at the vegetal pole. One cleavage later (sixth), blimp1 transcripts are present in the micromeres. After this, wnt8 and blimp1 are expressed in the same territories.

Determination of the GRN underlying endomesodermal development in the sea urchin embryo (1, 2) has revealed that the key driver of the dynamic torus pattern is the GRN subcircuit shown in Fig. 1B. To understand the operation of this subcircuit, it is important to note that the cis-regulatory control apparatuses of both wnt8 and blimp1 function as AND operators (3); that is, wnt8 expression requires both β-catenin/TCF and Blimp1 inputs (4) and blimp1 expression requires both β-catenin/TCF and Otx inputs (table S1). Morpholino-substituted antisense oligonucleotide (MASO) targeting blimp1 mRNA (5) or wnt8 mRNA (4) blocks endomesoderm specification.

Expression of the wnt8 gene illustrates the canonical torus pattern and directly controls its expansion. Several different wnt genes are expressed in the sea urchin embryo (6). Although earlier evidence from sea urchin and Xenopus indicated that Wnt8 is probably responsible for driving progressive β-catenin nuclearization during cleavage (79), we found that Wnt8 is responsible for producing the β-catenin/TCF input, which, according to cis-regulatory analysis, is obligatory for blimp1 expression (table S1). Thus, MASO repression of Wnt8 expression eliminates 80 to 98% of early blimp1 expression (fig. S1).

High-resolution measurements of blimp1 mRNA by quantitative real-time fluorescence polymerase chain reaction (fig. S2, A and B) show that a small amount of blimp1 mRNA is present maternally; however, there is no maternal wnt8 mRNA. In the early-cleavage embryo, β-catenin localizes to the nucleus, and by fourth cleavage, β-catenin can be visualized in the newly born micromere nuclei (7). By fifth cleavage, the wnt8 gene is activated in the micromeres (Fig. 1C). Because β-catenin/TCF and Blimp1 are the required inputs into the wnt8 gene, maternal Blimp1 factor must be available, consistent with the evidence that this gene is maternally expressed (fig. S2A). When the wnt8 gene begins to be transcribed, its response to its own signal transduction system produces a positive feedback circuit between adjacent endomesodermal cells that both produce and receive Wnt8 (1, 4). Otx protein is also nuclearized initially in the micromeres early in cleavage (10), hence it is available ab initio. blimp1 transcription is activated one cleavage after wnt8 transcription (Fig. 1C and fig. S2A). Activation must depend on the enhanced level of the β-catenin/TCF input driven by wnt8 transcription itself. Once both genes are transcribed in the same cells (i.e., from sixth cleavage on), the subcircuit architecture (Fig. 1B) indicates that the patterns of expression of the mutual regulatory partners, blimp1 and wnt8, should be similar. This was confirmed by whole-mount in situ hybridization (WMISH) (Fig. 1C), and their patterns of expression are equally represented in Fig. 1A.

An essential design feature of the relevant blimp1 cis-regulatory module is that it includes autorepression sites (5). The architecture of the subcircuit in Fig. 1B suggests that autorepression of the blimp1 gene, some hours after its activation, could account for the progressive clearance of both blimp1 and wnt8 transcripts from the center of the moving torus of regulatory gene transcription. A series of cis-regulatory reengineering experiments showed that this is indeed the mechanism of clearance. We used a blimp1 cDNA expression construct that produces normal blimp1 mRNA under control of the cis-regulatory module responsible for early blimp1 expression (5) (Fig. 2, A and B). When the cis-regulatory autorepression sites were mutated (Fig. 2B), the construct produced patches of mesodermal blimp1 transcript lying within the endogenous (mesodermal) blimp1 clearance zone, whereas the control generated only the normal torus pattern of expression (Fig. 2C, first three columns). We used these constructs to test whether, as predicted, ectopic redeployment of blimp1 mRNA was sufficient to cause continued expression of wnt8 in mesodermal territories. Endogenous wnt8 gene expression monitored by WMISH, as well as a coinjected wnt8 BAC-GFP (bacterial artificial chromosome–green fluorescent protein) knock-in reporter, produced persistent mesodermal expression in experimental embryos engineered to express blimp1 in the mesoderm (Fig. 2C, fourth and fifth columns). [See (11) and fig. S3 for quantitative data from these experiments, including an enhanced GFP mRNA output from the wnt8 BAC-GFP construct.]

Fig. 2.

Experimental demonstration of spatial blimp1 autorepression. (A) Genomic locus. Red boxes, exons; bent arrow, start of transcription; light blue boxes, noncoding sequence patches displaying high conservation between the sea urchin species Lytechinus variegatus and Strongylocentrotus purpuratus. Only the exons of the early-expressed (blimp1b) isoform are shown (5). (B) Expression constructs. “ExoBlimp” combines three conserved regions that are sufficient to reproduce correct blimp1 expression and to drive expression of blimp1 cDNA in the normal pattern of expression. In the “mutExoBlimp” construct, the two Blimp target sites indicated are mutated, destroying the autorepression function; the partially overlapping TCF sites and a third possible Blimp site proximal to the Blimp-TCF cluster remain intact. (C) Expression of genes and constructs indicated in upper right corner of each panel in embryos bearing ExoBlimp (top row) or mutExoBlimp (bottom row). LV, lateral view; VV, vegetal view; VLV, oblique lateral-vegetal view. Stable incorporation of injected constructs is mosaic in sea urchins (18); therefore, not all cells in the center of the ring are uniformly stained in mutExoBlimp embryos. In ExoBlimp embryos, expression of blimp1 mRNA from the construct is superimposed on the endogenous torus pattern of blimp1 expression. Wnt8→GFP denotes a BAC containing the wnt8 gene that contains a GFP reporter sequence in place of exon 1 of the gene, produced by in vitro recombination (11). Blimp→GFP denotes a construct similar to mutExoBlimp [compare with (B)], except that the blimp1 cDNA has been replaced by a GFP sequence.

According to the architecture of the subcircuit in Fig. 1B, positive spatial input into the blimp1 gene is provided by β-catenin/TCF (i.e., in response to Wnt8 signaling), because the other positive input, Otx, is continuously available throughout the whole region. To test this, we used a blimp1 GFP reporter bearing mutated Blimp1 target sites. This construct cannot autorepress, but it displays a normal pattern of expression if the location of blimp1 transcript is normal (Fig. 2C, sixth column, top). Therefore, the restricted pattern of endogenous β-catenin/TCF, due to the restricted domain of Wnt8 expression, suffices for the restricted spatial expression of the blimp1 gene. But when the restriction of Wnt8 expression in the mesoderm was relaxed by introduction of ectopic blimp1 mRNA, the expression of the blimp1 GFP construct lacking Blimp1 target sites was also relaxed (Fig. 2C, sixth column, bottom). These experiments show that the cause of progressive vegetal clearance of wnt8 expression is the restricted localization of the Blimp1 input, which is due entirely to blimp1 autorepression, as portrayed in the network subcircuit of Fig. 1B.

The Fig. 1B subcircuit architecture is directly authenticated at the cis-regulatory level (3, 6, 7) and in this work. Its design ordains its function. It consists of two partially overlapping feedback loops, both of which are subject to an autorepression function, one directly and one indirectly. One loop is signal-mediated: Reception of Wnt8 ligand in recipient cells produces the active β-catenin/TCF transcription factor complex that is required for expression of the wnt8 gene itself. In the endomesoderm, whether or not there is also an autocrine component, adjacent cells are indeed linked through this signal-driven transcription loop [the “community effect” (1)]. The cis-regulatory system of the wnt8 gene operates as an AND processor, in the sense that it requires both the Blimp1 and β-catenin/TCF inputs for function. Thus, the requirement for Blimp1 links it obligatorily to the second feedback loop. The second loop consists of the requirement for Blimp1 as a driver of wnt8 expression and the reciprocal requirement of β-catenin/TCF for blimp1 expression. The AND processor of the blimp1 gene similarly links it into the first loop by its obligatory requirement for the β-catenin/TCF input, but the other partner here is the Otx activator. If both cis-regulatory systems did not include AND gates dependent on the β-catenin/TCF input, the subcircuit would not work. The subcircuit has conditional operating features; that is, its behavior depends on the particular inputs it sees. The TCF input can function either negatively or positively, because (except in cells receiving Wnt8 signal) it binds the transcriptional repressor Groucho (12). This keeps the whole subcircuit quiet in the ectoderm. Its second conditional “off” function is the autorepression of the blimp1 gene, which from time-course data depends on a certain accumulation of blimp1 mRNA (and factor) (fig. S2C).

The subcircuit acts to produce the moving torus of gene expression, as summarized in Fig. 3. Measurements of the transcript concentrations indicate that during this phase the blimp1 gene is producing about 50 transcripts per cell-hour [this is only a few percent of the maximum possible transcription rate (13)]. Blimp1 factor eventually reaches a level where it acts to repress its own transcription when there could be as many as ∼1500 molecules per nucleus, given sea urchin translation rates (14) [more than sufficient for target site occupancy by the typical transcription factor (13), particularly in the small-micromere nuclei]. The Blimp1 factor then disappears from these cells and wnt8 gene expression turns off as the reinforcing feedback loop is broken. The half-life of blimp1 transcripts is about 1.5 hours in the micromeres and 2.5 hours in the mesoderm, versus a default average of 3 to 5 hours for all polysomal sea urchin embryo mRNAs (15). Meanwhile, however, the Wnt8 ligand has diffused to the next tier of cells, the future mesoderm (middle tier in Fig. 3). The intercellular diffusion rate of Wnt8 plus the molecules to which it is bound is most unlikely to be rate-limiting, given the very small intercellular distance and the rates that have been assumed for this process by others (16). Upon receipt of the Wnt8 signal, the subcircuit is thereby activated within the mesodermal territory, and the same cycle of events runs its course in this tier of cells, where it operates with very similar kinetics (Fig. 1 and fig. S2C). Subcircuit reactivation in the cells on the inside of the torus of gene expression (by inward signaling) is not observed: Once blimp1 transcription goes off, it stays off. Blimp1, a SET domain protein, could be silencing its own locus by recruitment of factors such as histone methyltransferases (17). The subsequent disappearance of nuclear β-catenin from the cells within the torus could also result in Groucho repression (7).

Fig. 3.

Summary of mechanism by which the dynamic, concentrically expanding torus of wnt8 and blimp1 expression is generated. The drawing shows a seventh-cleavage embryo viewed from the vegetal pole to illustrate the radial concentric organization. However, the events indicated in the numbered key begin at fourth cleavage and extend out to mesenchyme blastula stage (18 to 24 hours).

The genomic regulatory code is a static linear structure, whereas embryonic development is intrinsically a process driven by dynamically changing regulatory states. Here, we resolved a gene regulatory network subcircuit that combines these aspects in one small apparatus. The subcircuit's kinetics depend on the synthesis and turnover rates of the relevant mRNAs and proteins, as well as on the affinities of the transcription factors for their cis-regulatory target sites (13). What the apparatus does, however, depends on the genomic cis-regulatory sequence of the blimp1 and wnt8 genes, where its unique features, its feedback loops, AND gates, and autorepression function are encoded.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Table S1


References and Notes

View Abstract

Navigate This Article