C-terminal binding protein: Biological Overview | Evolutionary Homologs | Regulation | Developmental Biology | Effects of Mutation | References

Gene name - C-terminal binding protein

Synonyms -

Cytological map position - 87D13--87D13

Function - transcription factor-interacting protein

Keywords - segmentation, dorsoventral patterning

Symbol - CtBP

FlyBase ID: FBgn0020496

Genetic map position - 3-[52]

Classification - C-terminal binding protein homolog

Cellular location - nuclear



NCBI links: Precomputed BLAST | Entrez Gene | UniGene |
BIOLOGICAL OVERVIEW

hairy is a Drosophila pair-rule segmentation gene that functions genetically as a repressor. To isolate protein components of Hairy-mediated repression, a yeast interaction screen was carried out and a Hairy-interacting protein was identified, the Drosophila homolog of the human C-terminal-binding protein (CtBP) (Poortinga, 1998). CtBP is the subject of this overview, but to put its function in perspective, it will be necessary to first discuss Hairy.

The Hairy and Enhancer of split proteins are characterized by two conserved domains, the Orange domain and the C-terminal conserved tetrapeptide WRPW. Functional studies have shown that both of these domains, as well as the bHLH domain, are needed for the proper function of these proteins. The Orange domain contributes to the functional specificity of Hairy/E(spl) proteins (Dawson, 1995 and Giebel, 1997). The WRPW motif has been shown to be necessary and sufficient for the recruitment of Groucho, a WD repeat-containing protein that is not able to bind DNA on its own but, when brought to an endogenous or heterologous promoter, serves as a strong repressor of transcription. Together, these results have led to the prevailing view that Hairy functions as a promoter-bound repressor. An intact bHLH region is required for Hairy to bind to specific DNA sites; once bound, Hairy then recruits the Groucho co-repressor protein (Poortinga, 1998).

Recruitment of Groucho does not account for all of Hairy's repressor properties. Hairy can function genetically as a repressor in the absence of the WRPW motif, and presumably in the absence of the Groucho co-repressor. The Orange domain has been shown to be required for proper function of Hairy and E(spl)m8. This suggests that Hairy is involved in separable repression mechanisms: repression in some cases requiring the bHLH and Orange domains, and in other cases, requiring the bHLH and C-terminal WRPW motif (Dawson et al., 1995). Thus, Hairy may function modularly, with the scope and specificity of its interactions dependent on the proteins recruited to its various conserved domains. Based on the expectation that Hairy works as part of a multiprotein complex, the yeast two-hybrid protein interaction system was employed to identify a Hairy-interacting protein, the Drosophila homolog of human C-terminal binding protein (CtBP), that interacts with a small, previously uncharacterized C-terminal region of Hairy (Poortinga, 1998).

Drosophila CtBP plays a role in embryonic segmentation. Reduction of maternal dCtBP results in severe segmentation defects visualized by cuticle preparations or engrailed staining. Expression of the primary pair-rule genes, eve and runt, as well as expression of hairy itself, is disrupted in embryos lacking maternal dCtBP, whereas the level and spatial positioning of gap gene expression appears normal. The primary pair-rule genes are required to establish each other's expression as well as to direct the striped expression of downstream secondary pair-rule genes, such as fushi tarazu. Similar to what is seen in hairy mutant embryos, Ftz stripes are expanded throughout the trunk region in embryos lacking maternal dCtBP. The physical interaction of CtBP with Hairy and CtBP's genetic interaction with hairy indicate a role for dCtBP in Hairy-mediated repression. The domain of Hairy that interacts with CtBP contains a five amino acid repeat (PLSLV) that bears sequence similarity to a six amino acid motif in the E1a C-terminus (PXDLSX) which interacts with mammalian CtBP (Poortinga, 1998).

CtBP is involved in transcriptional repression mediated by Knirps and Snail. Knirps and Snail are thought of as short-range repressors, acting over distances of less than 100 base pairs to quench upstream activators or the core transcription complex. This form of repression allows enhancers to work independently of one another to direct complex, additive patterns of gene expression, including the seven-stripe patterns of even-skipped and hairy expression. A P-DLS-K sequence is present in the repression domains of Knirps and Snail, and the latter protein also contains the related sequence P-DLS-R. Gene dosage assays suggest that Knirps and Drosophila CtBP interact in vivo. Embryos that are heterozygous for the knirps9 null mutation exhibit occasional defects in the even-skipped expression pattern, including reduced staining of stripe 5. Combining the Drosophila CtBP and knirps9 mutation results in more severe disruptions in the eve pattern, including the fusion or loss of stripes 4 and 6. The latter knirps9/CtBP transheterozygous phenotype is virtually indistinguishable from that observed for knirps mutant embryos. These results raise the possiblility that Drosophila CtBP is part of a larger co-repressor complex that assembles on the Knirps/DNA template. Snail might also require CtBP, because embryos derived from dCrBP mutant oocytes exhibit dorsoventral patterning defects. The C-terminal repression domain of a third short-range repressor, Kruppel, contains a sequence that is related to the mammalian and Drosophila CtBP interaction sequence -- P-DLS-H. Mutations in this sequence nearly abolish Kruppel-mediated repression in human osteocarcinoma cells (Nibu, 1998a).

CtBP may mediate repression throught the enzymatic modification of chromatin because it sequence is related to D-isomer 2-hydroxy acid dehydrogenases. Despite this rather unexpected homology, immunolocalization assays indicate that the Drosophila CtBP protein accumulates in nuclei. Perhaps CtBP cause local changes in chromatin structure by introducing subtle changes in core histones. Alternatively, it is possible that CrBP is a component of an enzymatic cascade that modulates the activities of histone deacetylases or other co-repressor proteins (Nibu, 1998a).

CtBP-independent repression in the Drosophila embryo

There are three mechanisms of transcriptional repression in eukaryotes. The first is quenching, whereby repressors and activators co-occupy closely linked sites and then the repressor inhibits adjacent activators. The second is direct repression, in which repressors block the function of the core transcription complex. The third is competition, in which repressors compete with activators for a common DNA-binding site. Previous studies have shown that the Drosophila CtBP corepressor (dCtBP) is essential for the quenching activity of three short-range sequence-specific repressors in the early Drosophila embryo: Krüppel, Knirps, and Snail. This study demonstrates that dCtBP is dispensable for target enhancers that contain overlapping activator and repressor binding sites. However, it is essential when Krüppel and Knirps repressor sites do not overlap activator sites but are instead located adjacent to either activators or the core promoter. These findings provide evidence that competition is distinct from quenching and direct repression. Quenching and direct repression depend on dCtBP, whereas competition does not (Nibu, 2003 ).

Krüppel is a zinc finger DNA-binding protein that is composed of 502 aa residues. The quenching activity of the C-terminal repression domain (aa 402 to 502) requires a dCtBP interaction motif located at amino acids (aa) 464 to 470. Another repression domain has been identified in cultured cells. It is located between aa 62 and 92 and does not contain a dCtBP interaction motif. A transgenic embryo assay was used to determine whether this N-terminal repression domain might be a source for CtBP-independent repression in early embryos (Nibu, 2003).

A Gal4-Krüppel fusion protein containing aa 402 to 502 created gaps in the staining patterns directed by st2.UAS-st3-lacZ, NEE.UAS-lacZ, and NEE.UAS-twi-lacZ. The st2.UAS-st3-lacZ reporter gene contains Gal4 UAS binding sites near the distal eve stripe 2 enhancer (st2). NEE.UAS-lacZ reporter gene is driven by a modified 200-bp rhomboid rhomboid lateral stripe neurectoderm (NEE) enhancer that contains three Gal4 binding sites and three Dorsal activator sites. This reporter gene is normally activated in the ventral mesoderm. For the st2.UAS-st3-lacZ (st3 is the eve stripe 3 enhancer) and NEE.UAS-twi-lacZ reporter genes, repression was observed only for the staining pattern produced by the enhancer containing UAS binding sites. For example, the binding of the Gal4-Krüppel fusion protein to the stripe 2 enhancer does not alter expression from the neighboring stripe 3 enhancer. Similarly, the binding of the fusion protein to the rhomboid NEE enhancer does not alter expression from the twist enhancer. Substitutions in three of the amino acid residues within the dCtBP interaction motif (PEDLSMH to AAALSMH) eliminate the repression activity of an otherwise normal Gal4-Krüppel fusion protein (Nibu, 2003).

There is a second potential dCtBP interaction motif, located between aa 414 and 420 (PLDLSED), that weakly binds dCtBP in vitro. However, this second motif is not sufficient to support discernible repression activity in vivo. These results suggest that most or all of the repression activity of the Gal4-Krüppel 402-502 fusion protein resides within the major dCtBP interaction motif between amino acid residues 464 and 470. Moreover, repression is not observed for a Gal4-Krüppel fusion protein that contains the N-terminal repression domain (aa 62 to 92). These results suggest that the C-terminal dCtBP motif mediates most or all of the quenching activity in the early embryo (Nibu, 2003).

The proximal UAS site within the NEE.UAS-lacZ reporter gene is located 120 bp 5' of the core promoter, slightly beyond the range of Krüppel-mediated repression. In contrast, the UAS sites map within 50 bp of critical Dorsal sites within the NEE. Thus, repression of the reporter gene is most likely due to quenching rather than the direct repression of the core promoter. Another lacZ reporter was created to investigate this issue, NEE-5xUAS-lacZ. The most distal UAS site is located 250 bp 5' of the most proximal Dorsal binding site within the modified 700-bp NEE enhancer, while the most proximal UAS site is located just 57 bp 5' of the transcription start site of the hsp70 promoter. The Gal4-Krüppel 402-502 fusion protein attenuates lacZ expression. This direct repression is not obtained with the mutagenized fusion protein lacking the dCtBP interaction motif or with a fusion protein containing the N-terminal repression domain. These results suggest that the C-terminal dCtBP interaction motif is essential for both quenching and direct repression (Nibu, 2003).

Previous studies suggest that Krüppel mediates quenching by recruiting dCtBP to distal enhancers, such as the eve stripe 2 enhancer. An NEE-lacZ reporter gene that contains two synthetic Krüppel recognition sequences located 50 bp 5' of the most distal Dorsal binding site and 50 bp 3' of the most proximal site was created. This enhancer lacks the native Snail repressor sites and therefore directs lacZ expression in both lateral and ventral regions of early embryos. lacZ staining was diminished in central regions due to the localized expression of the Krüppel repressor. This gap in the pattern was eliminated in Kr1/Kr1 mutant embryos. Krüppel also failed to repress the reporter gene in mutant embryos derived from dCtBP germ line clones. These results indicate that dCtBP+ gene activity is required for the quenching activity of the Krüppel repressor (Nibu, 2003).

Subsequent experiments were done to determine whether dCtBP is required for the direct-repression activity of Krüppel and another short-range repressor, Knirps. lacZ transgenes with either Krüppel or Knirps binding sites located near the core promoter were examined. Both transgenes contain two tandem copies of the 250-bp twist proximal enhancer placed either upstream or downstream of rhomboid lateral stripe enhancers (NEE). In wild-type embryos, the enhancers direct additive patterns of expression in the lateral neurogenic ectoderm and ventral mesoderm. A single Krüppel binding site located 75 bp 5' of the transcription start site was sufficient to create a central gap in both staining patterns. Staining directed by the tandem twist enhancers was nearly eliminated, whereas the lateral stripe produced by the rhomboid NEE was diminished. Repression of the twist pattern is almost certainly due to direct repression, since the solo Krüppel site maps more than 800 bp from the nearest Dorsal activator site in the twist enhancer. Krüppel-mediated repression is lost when the transgene is introduced into embryos obtained from dCtBP germ line clones. There is no longer a central gap in the staining pattern. Moreover, there is a fusion of the expression patterns directed by the twist and NEE enhancers due to a loss in the activity of the Snail repressor. Normally, Snail binds to the NEE enhancer and represses expression in the ventral mesoderm, thereby restricting the staining pattern to lateral stripes in the neurogenic ectoderm. The broad uniform staining pattern obtained in dCtBP mutants suggests that the dCtBP corepressor is required for the direct repression of the core promoter (Nibu, 2003).

Similar results were obtained with the Knirps repressor. In this case, two tandem Knirps binding sites were placed 55 bp 5' of the transcription start site. In wild-type embryos, there is a clean gap in both the NEE-mediated lateral stripes and the twist-mediated staining pattern in the ventral mesoderm. This gap coincides with the site of Knirps expression in the presumptive abdomen. As seen for Krüppel, the gap in the staining patterns disappears in dCtBP mutant embryos. These results suggest that dCtBP is required for the direct repression activities of both Krüppel and Knirps (Nibu, 2003).

The preceding experiments suggest that dCtBP is required for both quenching and the direct repression of the core promoter. A synthetic lacZ reporter gene was prepared to determine whether Krüppel can mediate repression by competition and, if so, whether dCtBP is required for this repression. A 14-bp oligonucleotide that contains overlapping Dorsal and Krüppel binding sites was synthesized. Each subunit of the Dorsal homodimer binds to an inverted half-site: GGG...CCC. Krüppel binds DNA as a monomer, and the core recognition sequence includes the CCC Dorsal half-site. This short sequence also contains an optimal Bicoid binding site (GGATTA). This motif is located between the two half-sites of the Dorsal recognition sequence and overlaps the Krüppel consensus sequence (Nibu, 2003).

Gel shift assays were done to determine whether Dorsal and Krüppel bind the synthetic 14-bp sequence in a mutually exclusive manner. A 30-bp fragment that contains the 14-bp sequence along with 8 bp of flanking sequence at each end was synthesized. In the first set of experiments, a full-length Krüppel protein produced in E. coli was mixed with the 30-bp fragment and fractionated on an agarose gel. A shifted Krüppel-DNA complex was observed. The addition of increasing amounts of the Dorsal DNA-binding domain (Dl DBD; aa 1 to 403) resulted in the gradual loss of this complex. A new complex that is identical in size to those obtained with the Dorsal protein alone was observed. These results suggest that high concentrations of the Dorsal DNA-binding domain can displace Krüppel (Nibu, 2003).

Similar results were obtained in reciprocal DNA-binding assays. In this case, the shifted Dorsal-DNA complex was formed in the absence of Krüppel. The addition of increasing amounts of the Krüppel protein resulted in the gradual loss of the Dorsal-DNA complex. A new complex was obtained that has the same size as the one observed with increasing amounts of Krüppel in the absence of the Dorsal protein. These results suggest that increasing amounts of Krüppel can displace Dorsal-DNA complexes. Thus, the gel shift assays indicate mutually exclusive binding of Dorsal and Krüppel to the overlapping binding sites contained within the 14-bp fragment (Nibu, 2003).

Transient-transfection assays were used to determine whether the Krüppel DNA-binding domain is sufficient to mediate transcriptional repression. Six tandem copies of the synthetic oligonucleotide used in the preceding DNA-binding assays were attached to an eve-luciferase reporter gene containing the minimal eve promoter. This reporter gene was introduced into mbn-2 cultured cells (a Drosophila blood cell line) along with various expression vectors containing Dorsal or Krüppel coding sequences. An expression vector containing the full-length Dorsal coding sequence (Dl FL) produced a 6 fold induction in luciferase activity. However, an expression vector containing the Krüppel DNA-binding domain (Kr DBD; aa 217 to 401) reduced luciferase activity to background levels. This reduction in reporter gene expression was not obtained with a Krüppel expression vector that contained a single amino acid substitution in the zinc finger DNA-binding domain (Kr9 DBD). These results suggest that Krüppel can repress the synthetic enhancer by simply binding DNA and excluding the Dorsal activator. Repression does not depend on Krüppel protein sequences that map outside the DNA-binding domain. Subsequent experiments were done to determine whether Krüppel can mediate repression by competition in transgenic embryos (Nibu, 2003).

Either 6 or 14 tandem copies of the 14-bp synthetic enhancer sequence were attached to a lacZ reporter gene containing the minimal, 42-bp eve promoter region. Similar results were obtained with both fusion genes, and most of the following results were obtained with individual strains carrying the transgene with six copies attached. The transgene exhibits a combinatorial pattern of lacZ staining in wild-type (yw) embryos. Staining is first detected in the anterior 40% of 120-min embryos, presumably in response to the broad Bicoid activator gradient and is also detected in both anterior regions and along the entire length of the ventral mesoderm. Mesoderm expression was first seen at the time when the maternal Dorsal protein is released from the cytoplasm and enters nuclei. During cellularization, staining is lost in central regions, presumably due to the onset of Krüppel expression. In addition, there is a refinement in the anterior staining pattern, so that it becomes restricted to the anterior one-fourth of the embryo and exhibits a reasonably sharp posterior border. This staining pattern persists during gastrulation and germ band elongation (Nibu, 2003).

The transgene was introduced into different mutant backgrounds in order to confirm that the synthetic enhancer is regulated by Bicoid, Dorsal, and Krüppel. The anterior staining pattern is eliminated when the transgene is introduced into embryos derived from females homozygous for a null mutation in bicoid. However, staining persists in ventral regions in response to the Dorsal gradient. The loss of staining in the anterior regions correlates with an anterior expansion of the Krüppel expression pattern in bicoid mutants. The maternal Dorsal gradient is eliminated in embryos derived from females that are homozygous for a null mutation in gastrulation defective (gd7/gd7). lacZ staining in the ventral mesoderm of these mutants is lost. However, staining persists in anterior regions, presumably in response to the Bicoid gradient, which is unaffected in gd mutants. The transgene was also crossed into Kr1/Kr1 mutant embryos. The central gap of repression seen in wild-type embryos is essentially abolished in Kr mutants. There may be a subtle attenuation in central regions due to the low levels of Krüppel protein that are retained in this mutant (Kr1 is not quite a null allele. The anterior staining pattern directed by the Bicoid gradient may be a bit broader in Kr mutants than in wild-type embryos, suggesting that the Krüppel repressor might help refine the pattern. These results indicate that the artificial enhancer is activated by Bicoid and Dorsal but repressed by Krüppel. Competition is the likely form of repression since the Krüppel repressor sites directly overlap the Bicoid and Dorsal activator sites (Nibu, 2003).

One of the central goals of this study was to determine whether Krüppel requires dCtBP when it mediates repression by competition. This issue was investigated by crossing the transgene into mutant embryos derived from germ line clones produced in dCtBP/+ females. Krüppel continues to induce a central gap of repression in these mutants. In fact, the repression obtained in dCtBP mutants is comparable to that observed in wild-type embryos. These results provide a clear example of Krüppel-mediated repression in the absence of the dCtBP corepressor. In contrast, Krüppel fails to repress transcription in dCtBP mutants when Krüppel and Dorsal sites do not overlap (Nibu, 2003).

This study provides evidence for two distinct mechanisms of short-range repression, corepressor-dependent (quenching and direct repression) and corepressor-independent (competition) repression. In addition, this is the first demonstration that transcriptional repression by competition does not require a corepressor in transgenic Drosophila embryos. dCtBP is dispensable when Krüppel binding sites directly overlap Dorsal activator sites. However, dCtBP is essential for repression when the Krüppel and Dorsal sites are nonoverlapping and can be coordinately occupied. The previous analysis of eve stripe 2 regulation led to the proposal that the Krüppel repressor establishes the posterior stripe 2 border via competition. Two of the Krüppel repressor sites contained within the stripe 2 enhancer overlap Bicoid activator sites. Subsequent studies led to the surprising observation that Krüppel binding sites need not overlap activator sites in order to mediate transcriptional repression (Nibu, 2003).

There are three Krüppel binding sites in the minimal, 480-bp eve stripe 2 enhancer. Two of the sites directly overlap Bicoid activator sites. In both cases, it is likely that the binding of the Krüppel repressor precludes the binding of Bicoid. This type of simple competition is probably not restricted to the regulation of eve stripe 2. For example, one of the mixed Bicoid/Krüppel binding sites in the stripe 2 enhancer is conserved in a newly identified ftz enhancer, which appears to be activated by Bicoid but repressed by Krüppel (V. Calhoun and M. Levine, unpublished data reported in Nibu, 2003). The two enhancers contain the same composite recognition sequence, ACGGATTAA. Repression by competition probably governs, in part, the regulation of the rhomboid lateral stripe enhancer (NEE) since some of the Snail repressor sites directly overlap critical Dorsal and basic helix-loop-helix activator sites (Nibu, 2003).

An implication of this study is that the residual activity of the Krüppel repressor observed in dCtBP mutants might be due to repression by competition. For example, Krüppel can repress the hairy stripe 7 enhancer when misexpressed throughout early embryos using the heat-inducible hsp70 promoter. This repression is retained in dCtBP mutants. Moreover, a mutant form of Krüppel that lacks the dCtBP interaction motif can repress hairy stripe 7 expression. hairy stripe 7 is activated, at least in part, by Caudal and repressed by Krüppel. Interestingly, five Krüppel binding sites directly overlap Caudal activator sites within the hairy stripe 7 enhancer. Similar arguments apply to the Knirps repressor, which helps establish the posterior border of eve stripe 3. The stripe 3 pattern expands in kni-/kni- mutant embryos but is essentially unchanged in dCtBP mutants. Knirps repressor sites might overlap critical activator sites, such as binding sites for D-Stat or an unknown activator(s) within the stripe 3 enhancer. Previous studies suggest that Brinker can also function independently of corepressors when bound to sites that directly overlap critical Smad activator sites within cis regulatory regions of Dpp target genes. Direct evidence for simple competition was obtained in transient-transfection assays. The Krüppel DNA-binding domain is sufficient to inhibit activation of the synthetic enhancer by Dorsal in cultured mbn-2 cells (Nibu, 2003).

The results reported in this study exclude another possible explanation for the residual activity of the Krüppel and Knirps repressors in dCtBP mutants: direct repression of the core promoter. In principle, direct repression could involve distinct corepressor proteins. If so, then target genes that contain promoter-proximal Krüppel and Knirps binding sites might be repressed in dCtBP mutants. However, the lacZ fusion genes containing either a single Krüppel site or two tandem Knirps sites located near the transcription start site are no longer repressed in dCtBP mutants. Thus, the possibility is favored that the residual Krüppel and Knirps repression activities depend on competition between overlapping activator and repressor binding sites within selected target enhancers (Nibu, 2003).

The demonstration that both quenching and direct repression require dCtBP raises the possibility that these two seemingly distinct forms of repression employ similar mechanisms. At least three types of models come to mind. First, dCtBP could disrupt physical interactions between upstream activators and the RNA polymerase II transcription machinery/mediator complex at the core promoter. Perhaps dCtBP masks or modifies the activation domains of upstream activators. However, this model can account for quenching but not direct repression. A second type of model involves local chromatin modification. dCtBP contains a well-conserved dehydrogenase catalytic center and binds NADH. Perhaps dCtBP modifies proteins such as histones and helps condense DNA within the limits of a nucleosome. In Saccharomyces cerevisiae, the Rpd3 histone deacetylase (HDAC) causes histone deacetylation over a distance of just two nucleosomes. A third model is that dCtBP 'poisons' the RNA polymerase II transcription machinery and impedes its binding, assembly, or function at the core promoter. This poisoning can be accomplished by placing dCtBP-dependent repressors near the core promoter or by looping distal enhancers to the promoter. According to the latter model, the linkage requirement seen for short-range repressors (they must bind within 100 bp of adjacent activators) might reflect a reliance of the repressors on linked activators in order to loop to the core promoter (Nibu, 2003).


GENE STRUCTURE

By screening early embryonic cDNA libraries with the two-hybrid cDNA insert as a probe, four different cDNA classes have been identified. Each of these four classes encodes the first 376 amino acids of dCtBP then undergoes alternative splicing so that the most C-terminal amino acids (5-83 amino acids), as well as the 3' non-coding regions, are different (with the two-hybrid cDNA having an additional 10 amino acids). cDNAs corresponding to all transcripts identified by Northern analysis have not yet been identified. However, a dCtBP subclone containing only the ATG to the junction where the sequence divergence occurs (amino acids 1-376) retains full interaction with Hairy, indicating that all protein isoforms should interact with Hairy (Poortinga, 1998).


PROTEIN STRUCTURE

Amino Acids - 383 (Nibu, 1998a) 381 and 459 (Poortinga, 1998)

Structural Domains

A sequence database search reveals 63% sequence identity over the entire h-C28 clone (amino acids 93-343), as compared with the human CtBP, a 48 kDa phosphoprotein. Based on the high degree and extent of this homology, it is concluded that h-C28 encodes the Drosophila homolog of the human CtBP protein. Human CtBP was identified as a protein that binds the C-terminus of the adenovirus E1a oncoprotein (Boyd, 1993, Schaeper, 1995 and Poortinga, 1998).


C-terminal binding protein: Evolutionary Homologs | Regulation | Developmental Biology | Effects of Mutation | References

date revised: 5 June 98

Home page: The Interactive Fly © 1995, 1996 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.