The Power Of The Csregulatory Module

Their internal designs endow czs-regulatory elements with the power to generate unique spatial interpretations from what is often rather sloppily graded input information, or from overlap of individually crude regulatory domains. The developmental process depends on this property of czs-regulatory systems, and thus on their exact DNA sequence features. Understanding development obviously requires appreciation of these structure/function relationships. But that is not all: these structure/function relationships also define the evolutionary problem of how given czs-regulatory elements originated in the genomes of taxa to which they are specific.

Diverse cis -Regulatory Outputs from a Simple Input

The declining ventral-to-dorsal gradient of nuclearized Dl factor in Drosophila embryos can be regarded as a simple input, in that it appears to a first approximation to have a diffusion-limited form, and it has a unique, localized source. This gradient results from the activation of the Toll receptor along the ventral surface of the egg, initially triggered by the regional proteolytic activation of an extracellular ligand (Stein and Niisslein-Vollard, 1992). Diffusion of the Toll pathway signaling components ultimately results in the phosphorylation and then degradation of a protein to which the maternal Dl factor is bound in the cytoplasm. It is thereby released for transit into the nuclei. In this way a graded ventral-to-dorsal concentration cline of nuclear Dl protein is produced (Anderson et al., 1985; Steward and

Maternal Gene Toll And Dorsal

Govind, 1993; Steward, 1989; Rusch and Levine, 1996). This is revealed by immunological staining in Fig. 2.6A (Rushlow et al., 1989).

We have already met two as-regulatory elements that respond in distinct ways to this simple D1 input, viz the zen repressor element, and the rho neuroectoderm element (Fig. 2.3). But these are only two of a series of diverse as-regulatory designs, all of which enable their genes to "read" the D1 gradient, and each of which produces a different output (Jiang and Levine, 1993; reviewed by Rusch and Levine, 1996; Huang et al., 1997). The genes that encode the transcriptional regulators Twist and Snail (Sna) are expressed only in the ventral mesodermal domain: twist because its czs-regulatory element includes a series of low-affinity binding sites for the Dl activator so that they are occupied only at high D1 concentrations; and the sna gene because its as-regulatory element includes both low-affinity Dl target sites and also sites for the positively acting Twist factor. Endogenous sna expression in a normal egg is illustrated in Fig. 2.6B. The

FIGURE 2.6 Diverse c/s-regulatory responses to normal and reoriented gradients of Dorsal in the Drosophila embryo. (A) The endogenous ventral-to-dorsal gradient of nuclear Dorsal (Dl) transcription factor. The factor is displayed In syncytial embryos by fluorescence immunocytology. The left panel is 12th cycle; nuclearization has just begun. The right panel is 13th cycle. Higher nuclear Dl concentrations are ventral. Nuclei lacking Dl appear dark, and solid red Indicates cytoplasmic Dl. [(A) Adapted from Rushlow et al. (1989) Cell 59, 1165-1177.] (B) Endogenous snail expression in ventral mesodermal domain displayed by in situ hybridization in a normal embryo. Snail expression pattern depends on a c/s-regulatory element that contains sites for Dl and Twist activators (Ip et al., 1992a). (C) Reoriented anterior-to-posterior gradient of nuclearized Dl, displayed in a late syncytial-stage embryo by yellow-green immunofluorescence (cf. A). The view is dorsolateral, and the highest Dl concentration is anterior (the endogenous Dl gradient is invisible). In the anterior third of the egg the Dl factor is almost entirely localized to the nuclei; it is partly nuclear and partly cytoplasmic in the middle region; and it is largely cytoplasmic at the posterior end. (D) Expression of snail gene visualized by in situ hybridization, in egg with an anterior-to-posterior Dl gradient superimposed on the endogenous gradient (cf. B). (E) Anterior expression of a lacz construct driven by c/s-regulatory elements containing two low affinity sites for Dl from the twist gene, visualized by hybridization of lacz mRNA. (F) Anterior expression of a derivative twist lacz construct driven by two copies of the element in (E), but altered so that the Dl sites bind the factor with higher affinity (twi*); the anterior cap of expression now extends further in the posterior direction. (G) Expression of endogenous short gastrulation (sog) gene in a normal embryo, displayed by in situ hybridization. The gene is expressed in a broad pair of lateral stripes overlying the mesodermal domain. (H) Expression of sog in an embryo bearing a reoriented Dl gradient, as in (C). In this egg the endogenous gradient has been prevented from forming by mutational interference with the endogenous Toll signaling process. [(B-H) From Huang et al. (1997) Genes Dev. I I, 1963-1973.]

expression of the rho gene higher up on the embryo (i.e., more dorsally), where nuclear Dl concentrations are lower, is due not only to higher affinity D1 sites, but also to the inclusion of sites for additional bHLH factors that synergize with Dl (viz, Scute and Daughterless; see Fig. 2.3A). The as-regulatory elements of the twist, sna, and rho genes include the specific apparatus that is necessary and sufficient for their diverse interpretations of the Dl gradient.

The autonomous responses of the czs-regulatory elements of these and other genes have been demonstrated, very dramatically, by redirecting the Dl gradient with respect to the egg axes (Huang etal., 1997). A gene encoding a constitutively active Toll receptor was introduced under the control of a maternally active cis-regulatory element, and the encoded mRNA was also equipped with the 3' trailer sequence of the bed mRNA, which mediates mRNA localization to the anterior pole. The result is an anterior-to-posterior gradient of nuclear Dl. This is seen by immunofluorescence in Fig. 2.6C (compare Fig. 2.6A). cz's-Regulatory systems responding to Dl which normally generate stripes of expression parallel to the A/P axis now produce stripes parallel to the D/V axis. Expression of the endogenous sna gene in such an egg is shown in Fig. 2.6D. An anterior cap of sna expression appears, and indeed these cells are respecified as mesoderm (Huang et al., 1997). Anterior expression is superimposed on the normal ventral sna mRNA stripe. However, note that a gap appears between the ectopic and the original sna expression domains, betraying the existence of a repressor which is apparently also under Dl control, and which delimits the distal boundary of sna expression (i.e., what would normally be the dorsal boundary). Figure 2.6E illustrates the behavior of a different expression construct bearing the low affinity Dl-binding sites from the twist gene, in an embryo in which the highest levels of nuclear Dl are at the anterior end: the result is an anterior cap of expression. But if the sequences of these Dl target sites are altered so that they bind Dl with higher affinity, expression of the construct spreads in the posterior direction (Fig. 2.6F). A further example is given in Figs. 2.6G and H. The normal lateral band of expression of the endogenous short gastrulation (sog) gene, is shown in (G); and when the Dl gradient is reoriented to form an anterior to posterior cline, a circumferential band of sog expression instead appears (H; in this case in an embryo lacking the endogenous D/V Dl gradient). These examples illustrate a profound principle of development: diversity in the structure of as-regulatory elements is the cause of diversity in gene expression patterns that arise in response to much less diverse prior regulatory states.

Direct Integration of Noncoincident Spatial Inputs

There is a class of czs-regulatory elements that function only where two different transcription factors overlap, each representing a different prior regulatory pattern. In this way these elements create new domains of gene expression. Such elements are "wired" so that they are incapable of responding adequately to either input alone (at normal levels of input activity), but when both are present they stimulate meaningful transcription. In biological terms, these elements are used to integrate different developmental processes. For instance field "A" could arise as a stripe of gene expression along the A/P axis of an organism, while field "B" is independently set up orthogonal to field A, along the D/V axis. Where they intersect a gene encoding a transcription factor required for specification of a new population of cells is activated. Or field A might define the progenitors of a new structure while field B overlaps field A only in its anterior portion, thus defining this region of the structure. The spatial coordinates of many developmental processes in both insects and vertebrates appear to be set in this way, though only a few have been analyzed at the cz's-regulatory level.

In Xenopus embryos a gene encoding a homeodomain transcription factor called Siamois is activated on the future dorsal side when transcription resumes at the midblastula stage. Siamois is required for specification of the Spemann organizer (Lemaire et al., 1995; Carnac et al., 1996; reviewed by Kimelman, 1999). Its transcription is confined to the vegetal dorsal quadrant by a combination of positive and negative interactions. The gene is repressed on the ventral side by a ubiquitous maternal Tcf factor (Tcf3) which acts by way of the Groucho corepres-sor. Activation requires two spatially confined positive inputs that function syner-gistically in the siamois cis-regulatory element where they overlap (as well as a third positive regulator that is ubiquitous). These are a vegetal input downstream of a TGF(3 signaling system; and a dorsal input consisting of a f5-catenin-Tcf3 complex. P-catenin interferes with Gro function and in complex with Tcf3 acts instead as an activator (Brannon et al, 1997; reviewed by Kimelman, 1999). (3-catenin activity is confined to the dorsal side by a mechanism which is triggered by the cortical rotation toward that side (reviewed by Moon and Kimelman, 1998). Immediately downstream of the siamois gene lie other regulatory genes, one of which is goosecoid. The goosecoid (gsc) gene is expressed in cells that will form the head organizer, and which themselves are progenitors of dorsal mesoderm (Har-land and Gerhart, 1997). Its locus of expression is also defined by integration of two inputs within the relevant gsc cis-regulatory element: this element responds directly to a vegetal TGF(3 family signal, and also to the dorsal input provided by Siamois or another similar factor, Twin (Watanabe etal, 1995; Laurent etal, 1997). Both the siamois and goosecoid cz's-regulatory elements thus include mechanisms for integration of vegetal and dorsal signals (Moon and Kimelman, 1998; Kimelman, 1999). The same form of "and" logic continues to be utilized downstream. Xnr3, a gene encoding a TGF|3 family ligand expressed in the Spemann organizer, is also controlled by such a system. In this case the inputs, both required, are a Tcf factor plus another regulator that is expressed zygotically only at gastrula stage (McKendry et al., 1997). Thus we have a series of multiple input systems, each adding spatial, or spatial plus temporal, regulatory value to what was there before.

Integrative summing of diverse spatial regulatory inputs is utilized over and over in Drosophila development. Pair-rule and gap genes are regulated in this way in the syncytial embryo (Baumgartner and Noll, 1991; Háder et al., 1998; Rivera-Pomar and Jackie, 1996; Wu and Lengyel, 1998). Another example is the integration of Hedgehog and Wingless signaling inputs by a dpp enhancer active in wing imaginal discs (Hepker etal., 1999)- czs-Regulatory elements that are targets of box gene cluster regulators often display this same kind of program feature. For example, the teashirt gene, which encodes a Zn-finger transcriptional regulator required for trunk specification, is expressed in trunk epidermis in response to Antennapedia plus an epidermis-specific factor; and in thoracic mesoderm in response to Ultrabithorax (Ubx) plus a mesoderm-specific factor (McCormick et al., 1995). Similarly, the midgut regulatory element of the dpp gene responds positively to the combined inputs of Ubx plus a visceral mesoderm-specific factor (Capovilla etal., 1994; Manak etal., 1995).

A visually beautiful example of cz's-regulatory integration of geometrically distinct input patterns is illustrated in Fig. 2.7 (Kim et al., 1996, 1997a,b; Klein and Martinez Arias, 1999). The vestigial (vg) gene encodes a transcriptional coactivator which works together with the Scalloped transcription factor (Haider et al., 1998). The vg gene is expressed and is required for growth in all imaginal disc cells of the wing pouch. But its control is more mosaic than meets the eye, in that its expression throughout the wing pouch depends on two separate regulatory modules. A "boundary enhancer" initially sets up expression along the D/V boundary, in response to Notch signaling, and along the A/P boundary as well. The cross-like pattern generated by a lacz construct controlled by this element is shown in Fig. 2.7A. Expression in the remainder of the wing pouch is controlled by the "quadrant enhancer." This element integrates two signals, viz, a Dpp signal emanating from the A/P boundary, and (indirectly) a second signal emanating from D/V boundary. A Mads box factor which mediates the Dpp signal response binds within this element, and its interaction is required for function (Kim et al., 1997a,b). The integrative response of the quadrant enhancer can be seen in Figs. 2.8B and C: note the initial pattern of expression, which begins just where the A/P and D/V boundaries intersect, and then spreads concentrically outward. The geometry of the expression pattern described by the quadrant enhancer follows from its activation in response to dual, orthogonal inputs.

c¿s-Regulatory elements that integrate inputs can be thought of as control units which execute logic functions. But in essence, that is what all cz's-regulatory elements do: they are genomic sequences that specify logic operations.

A cis-Regulatory Logic Device

Portions of the endoló cz's-regulatory system of Strongylocentrotus are to date the most extensively explored of any, with respect to the functional meaning of each interaction that takes place within them. What emerges is almost astounding: a network of logic interactions programmed into the DNA sequence that amounts essentially to a hardwired biological computational device.

As will be recalled, endoló is expressed in the endoderm of the embryo. At first glance its regulation appears a rather typical tissue-specific process, and indeed, it probably is rather typical. But the number and variety of regulatory transactions required for this gene to achieve its expression profile is surprising. The spatial

FIGURE 2.7 Integrative control of the Drosophila vestigial gene. (A) Expression of a c/s-regulatory construct in the wing blade region of a third instar wing imaginal disc. In this construct a lacz reporter is controlled by the "boundary enhancer," a c/s-regulatory module from the 2nd intron of the vestigial (vg) gene. The boundary enhancer initiates expression along the D/V boundary in response to Notch signaling. The expression pattern marks the D/V boundary (horizontal, ventral to top) and the A/P boundary (vertical, anterior to left), lacz expression is shown by

(Continues)

FIGURE 2.7 Integrative control of the Drosophila vestigial gene. (A) Expression of a c/s-regulatory construct in the wing blade region of a third instar wing imaginal disc. In this construct a lacz reporter is controlled by the "boundary enhancer," a c/s-regulatory module from the 2nd intron of the vestigial (vg) gene. The boundary enhancer initiates expression along the D/V boundary in response to Notch signaling. The expression pattern marks the D/V boundary (horizontal, ventral to top) and the A/P boundary (vertical, anterior to left), lacz expression is shown by

(Continues)

pattern of endol6 activity is shown in Fig. 2.8A (Ransick et al., 1993). It is not known whether the protein has earlier functions as well, but the gene is activated long before there is a midgut, or any gut at all, in mid-to-late cleavage, during or soon after the initial processes of endoderm specification (Godin et al., 1996). Figure 2.8A1 shows the initial expression of endol6 in the vegetal plate of a blastula-stage embryo, which is comprised of the cells that will give rise to the endodermal and mesodermal cell types of the embryo, except for skeletogenic cells. The skeletogenic progenitors (plus a few cells of the future coelomic sacs) appear as the nonexpressing patch within the ring of expressing endomesodermal cells in Fig. 2.8A2. We can see here that there are two boundaries that must be established: one within, between the endomesodermal cells and the skeletogenic cells; and one without, between the endomesodermal cells and the overlying ectoderm. It will not surprise that these boundaries depend on spatial repression (Yuh and Davidson, 1996). The subsequent panels of Fig. 2.8A show that endol6 is later expressed throughout the archenteron, remaining silent in the skeletogenic mesenchyme and the surrounding ectoderm. Expression is then extinguished in the foregut, and later in the hindgut, while intensifying in the midgut, its terminal locus of activity (Nocente-McGrath et al, 1989; Ransick et al, 1993).

Figure 2.8B is a protein-binding map of the endol6 as-regulatory system, which is included in a 2.3 kb fragment of DNA that suffices to reproduce the normal pattern of expression when linked into an expression vector (Ransick et al., 1993; Arnone et al, 1997). Thirteen different proteins bind with high specificity within this region (Yuh et al, 1994), and it has been divided experimentally into six functional regions (G-A). The role played by each region is indicated briefly below the map (Yuh and Davidson, 1996; Yuh et al, 1996). The boundary beyond which expression is precluded in the skeletogenic cells requires immunofluorescence. [(A) From Kim et al. (1996) Nature 382, 133-138; copyright Macmillan Magazines, Ltd.] (B) Expression of lacz construct controlled by the quadrant enhancer, early third instar. This enhancer integrates distinct inputs, one emanating from the A/P boundary of the wing blade portion of the imaginal disc, and the other from the D/V boundary. The A/P boundary input depends on Dpp, which is expressed along this boundary (see text), and is mediated by target sites for Mads transcription factors, lacz is shown in blue; endogenous vg gene expression in red. At this early stage vg expression is mainly due to the boundary element, and activity of the quadrant element, blue and purple, is just beginning, at the intersection of D/P and A/P boundaries. [(C) From Kim et al. (1996) Nature 382, 133-138; copyright Macmillan Magazines, Ltd.] (D) Final pattern of quadrant enhancer expression, visualized as in (B). The Lacz product now fills all four quadrants of the wing blade (blue overlying red). The dual spatial source of these inputs is clearly implied by the central location of the early pattern shown in (A). The total pattern of vg expression in the wing blade region of the disc is the sum of the outputs of the quadrant and boundary enhancers. [(D) From Kim et al. (1997a) Nature 388, 304-308; copyright Macmillan Magazines, Ltd.]

repressive interactions in the DC region; and for the boundary with the overlying ectoderm interactions in both regions E and F are required. The protein responsible for repression in the F region is a factor of a class which is usually associated with signaling systems, viz, a Creb factor, so the boundary with the ectoderm is at least in part likely to be established by a signaling interaction (as implied by other evidence as well; Davidson et al., 1998). The endol6 system includes three positive regulatory regions. These are the relatively unimportant Module G, which acts as a booster; Module B, which is solely responsible for the late rise in midgut-specific expression; and Module A. Module B is activated by an endo-derm-specific regulator (UI), the same that causes cylla gene expression in the gut late in development (Arnone and Davidson, 1997). Module A acts as the central processing unit for the whole upstream system, G-B, and it is on structure/ function relations within Module A and its intricate linkages to Module B that we now focus.

Module A carries out multiple functions. The basal promoter (Bp) of the endol6 gene consists of the sites where the transcription apparatus assembles plus a few proximal target sites (Fig. 2.8B). The Bp is entirely promiscuous with respect to the inputs it will service, and it is almost inactive without upstream inputs (Yuh et al., 1996; R.A. Cameron and E.H. Davidson, unpublished data). The only inputs the Bp receives from the whole endol6 czs-regulatory system are normally channeled through Module A; their processed result is the output of Module A. The value of this output determines whether transcription of the gene will occur, and if so at what rate, in every cell throughout embryogenesis (Yuh etal., 1996,1998). Module A also serves as a terminus for the upstream modules that mediate spatial repression, viz, E, F, and DC, so that in its absence they do not work. The outputs of these repressor subsystems can also be considered inputs to Module A. As we have already seen, the skeletogenic cell repressors in region DC do not act as transcriptional silencers (or long-range repressors; Fig. 2.4), but are rather dedicated to control of the positive regulatory activity of Module A. The positive regulatory function of Module A is mediated by an Otx factor which binds within it (Yuh et al., 1998; Li etal., 1999), and the repressive controls on spatial expression that Module A mediates are necessary because Otx is active in many regions of the embryo (Mao et al., 1996; Li et al., 1997). However, later in development the DC, E, and F regions can all be discarded, as can Module G, and a construct consisting only of B and A runs accurately in the midgut at almost the same rate as does the whole system. By this stage the system utilizes only the gut-specific positive input of Module B. But this requires two additional functions of Modules A and B. The first is a switch which is sensitive to the level of input of the positive regulator of Module B, and when this becomes significant the switch turns off the input of the Otx regulator, ensuring that only the input of Module B will count. The second is an amplification function by which Module A steps up the amplitude of the Module B input. These various functions of Modules A and B could never have been enumerated except by assessment of the functional meaning of each of its target sites (Yuh et al., 1998, 2001).

Was this article helpful?

0 0

Post a comment