ABSTRACT
The restriction of herpes virus latency to mammalian sensory ganglia has led to a search for tissue-specific regulatory molecules in these neurons which alter viral gene expression. We have recently shown that the POU-domain transcriptional regulator Brn-3.0 is abundantly expressed in the adult trigeminal ganglion. To begin to examine the hypothesis that Brn-3.0 might participate in the regulation of the HSV life-cycle, we used Brn-3.0 POU-domain protein as an affinity matrix, and biochemically screened the entire HSV genome for sites of Brn-3.0 binding. This screen identified several sites of the form T A/T A A T N A N T A/T, which significantly do not include the previously identified HSV octamer sequences. All of the selected sites occur in the <25% of the HSV genome which has not been assigned to open reading frames, suggesting that these sites may be transcriptional regulatory elements recognized by Brn-3.0 or another homeobox factor with similar DNA binding properties. However, these sites do not interact with Brn-3.0 with sufficiently high affinity to directly mediate transcriptional activation by Brn-3.0 alone in transfection assays. The experiments described also provide an effective general method for exhaustive screening of large viral genomes or sub-genomic fragments of eukaryotic DNA for sites of interaction with specific transcription factors.
Herpes simplex virus type I (HSV-1) produces cold sores or fever blisters in a primary infection of the skin. HSV-1 and other herpes viruses may also lie dormant or latent for extended periods without clinical manifestations, only to be reactivated at a later time. The site of the latent HSV-1 infection is the sensory neuron of the trigeminal ganglion, and in mouse models of HSV-1 infection, it has been demonstrated that the virus follows one of two pathways when it reaches the nerve. In most cases the virus productively infects neurons, which result in viral growth and probably death of the cell. In a minority of cases the virus can enter a latent state in the neuron, in which the lytic genes are not expressed and in which there is expression of a single transcription unit, the latency associated transcript (1 ,2 ).
In lytically-infected epithelial cells, the role of a widely-expressed POU-protein, Oct-1, has been well established. Oct-1 interacts with a tegument protein VP-16 at sites containing TAATGARAT sequences to stimulate expression of the five HSV immediate-early genes (3 -7 ). Although VP-16 is required for productive infection of neurons (8 ), peripheral sensory neurons express Oct-1 at much lower levels than epithelial cells. Oct-1 expression in these neurons is below the level of detection by in situ hybridization (9 ), and detectable at only low levels in sensitive DNA-binding assays (10 ), suggesting the involvement of alternate POU-factors or other gene products in neurons.
The tissue-specificity of HSV-1 latency has also led to a search for cellular factors in sensory neurons which may mediate the latency/reactivation process. It is clear that the immediate-early genes are not expressed during latency, and it has been proposed that neuronal forms of the Oct-2 protein repress the IE genes at the TAATGARAT sequences, allowing the latent state to be established (11 -19 ). However, these studies have been based largely on cell line models of sensory neurons, and recent studies have shown that Oct-2 is not detectable in mature sensory ganglia (10 ,20 ).
We have been engaged in studies of the Brn-3 family of transcription factors, consisting in mammals of the closely related genes Brn-3.0, Brn-3.1 and Brn-3.2 (9 ,21 -25 ; also designated Brn-3a, Brn-3c and Brn-3b respectively). The expression of these genes is largely restricted to the nervous system, and Brn-3.0 in particular is highly expressed in the sensory cranial and dorsal root ganglia. Mice homozygous for null mutations at the Brn-3.0 locus have defects in sensory neuronal development (26 ,27 ). This expression pattern and prior evidence for the role of POU-factors in the HSV life cycle suggest that Brn-3.0 might contribute to the unique properties of sensory neurons with respect to HSV infection.
Here we have used the POU-domain of the Brn-3.0 protein in a novel affinity strategy to search the HSV genome exhaustively for Brn-3.0 recognition sites. Brn-3.0 interacts with approximately eight sites in the HSV genome which contain the consensus sequence T A/T A A T N A N T A/T. These sites occur exclusively in the 20% of the HSV genome which has not been assigned to open reading frames, and are frequently found near the 3' terminus of HSV late gene transcribed sequences. The affinity of Brn-3.0 for these sequences is significantly higher than its affinity for the previously characterized Oct-1/VP-16 (TAATGARAT) sites. However, our prior studies of functional Brn-3.0 recognition elements suggest that these sites probably do not have high enough affinity to mediate direct transcriptional activation by Brn-3.0, and thus do not conclusively establish a role for Brn-3.0 in the HSV life cycle. The techniques used to detect Brn-3.0 binding sequences in the HSV genome provide a convenient screening method which could be applied to any transcriptional regulator.
Human Oct-2 and murine Brn-3.0 (3 ) cDNA sequences corresponding to the POU-specific plus POU-homeodomains were ligated into the pGEX expression vector. Protein expression and purification were performed as previously described (25 ) and the fusion proteins were used uncleaved. The glutathione-S-transferase protein alone exhibited no activity in binding viral sequences or in gel shift assays.
Rabbit skin cells were infected with strain 17+ at a multiplicity of infection of 1 p.f.u./cell for 48 h. Infected cells were pelleted at 4oC and lysed in 5 ml hypotonic lysis buffer (10 mM Tris 8.0, 10 mM EDTA, 1% NP-40 and 0.5% deoxycholate) for 10 min. Nuclei were pelleted and discarded, and the HSV virions from the cytoplasmic supernatant were extracted with phenol-chloroform and ethanol-precipitated to yield viral genomic DNA.
Viral DNA (5 [mu]g) was digested to completion with MspI and HinPI in buffers supplied by the manufacturer (New England Biolabs). The digested product was end-labeled with 32P to follow yield, and incubated in a binding mix containing 20 mM Tris, pH 8.0, 0.1 mg/ml poly(dI[middot]dC) (Pharmacia), 0.025 mg/ml poly(dA[middot]dT), 5 mM MgCl2, 10% glycerol, 100 mM KCl, 0.2 mM EDTA, 1 mM DTT, 0.1 mg/ml bovine serum albumin and 40 [mu]g Brn-3.0 POU-domain GST fusion protein. After 20 min at room temperature, 100 [mu]l of 1:1 slurry of glutathione-agarose (Sigma) was added, incubated on ice for 30 min, then briefly centrifuged and washed twice in the binding buffer. Bound DNA was extracted with phenol-chloroform, and the binding reaction was repeated with the selected product. Final yield was ~0.5% of the initial mass of genomic DNA.
Oligonucleotide gel shift assays were performed as previously described (22 ), except with the omission of salmon sperm DNA. Oligonucleotide competition assays were performed under conditions of oligonucleotide excess, with 2.5 * 10-11M active Brn-3.0 protein, 2.5 * 10-9 labeled CRH site and 1-400 nM of competitor oligonucleotide in a 20 [mu]l reaction volume. The near-saturation of Brn-3.0 protein with oligonucleotide under conditions of the competition assay was confirmed by the negligible increase in 32P binding following the addition of 10 nM radiolabeled CRH oligonucleotide. As predicted, addition of an equimolar amount (2.5 nM) of unlabeled CRH oligonucleotide displaced 40-50% of the bound label, and the relative affinities of competitor oligonucleotides were estimated based on the concentration of competitor which inhibited binding to the same extent as the equimolar concentration of cold CRH oligonucleotide (I0.5).
The HSV-derived competitor oligonucleotide sequences, with the sequence of viral origin in upper case, are: 101a, gatctGGTATGGTAATTAGAAACCggatc; 101b, gatctCCATTAATGAGTTTCCggatc; 107s, gatctGATGTTAATAAATAACACATAggatc; 123s, gatctGTGTTTTAATCAATAAAAGACCACggatc; 128s, gatctGTACCCTTAATAAATTTTACAAACggatc; 170s, gatctCATACCTAAATAAATAAAAACCggatc.
Other model binding sites include: octamer/TAATGARAT (O/T, ref. 28 ), gggccgtGC
Sites of Brn-3.0 binding in the HSV genome were identified in vitro by Brn-3.0 affinity selection of restriction digest fragments of HSV strain 17+ genomic DNA. For the selection process, HSV genomic DNA was digested with restriction enzymes HinPI (GCGC) and MspI (CCGG). These restriction enzymes were selected because they provided fragments in the desired 50-500 bp range, because the GC-rich recognition sequence is unlikely to disrupt AT-rich Brn-3.0 binding sites, and because they yield identical 5'-CG extensions compatible with ligation into the ClaI site of the polylinker of the cloning vector pBKSII. Following digestion, genomic fragments were end-labeled with 32P, then incubated with Brn-3.0 or Oct-2 fusion proteins (Materials and Methods). Brn-3.0-DNA complexes were separated on glutathione-agarose beads, washed, and precipitated for electrophoretic analysis or cloning.
The initial analysis of the HSV genomic fragments bound by Brn-3.0 and Oct-2 was performed by denaturing polyacrylamide gel electrophoresis of the selected product (Fig. 1 ). The stringency of the binding reaction was adjusted by varying the concentration of poly(dA[middot]dT) and by repeating the selection procedure until bands representing only the highest-affinity bound fragments were observed. In the selected DNA, genomic fragments were observed which bound uniquely to Brn-3.0, uniquely to Oct-2, or to both proteins. Frequently, fragments which bound to both proteins at low stringency bound to only one protein at high stringency. In at least one case, high affinity interaction of a genomic fragment (clone 101) with both Brn-3.0 (site 101b) and Oct-2 (site 101a) resulted from discrete high affinity recognition sites which occurred in the same genomic region, not from similar affinity for a single site. Although we cannot rule out some sites which have a similar high affinity for both Oct-2 and Brn-3.0, in general, their highest affinity binding sites in the HSV genome appear to be distinct.
The restriction of herpes simplex virus latency to the sensory ganglia suggests that these neurons have unique properties with respect to HSV infection, conferred by cellular factors. The cooperative interaction of the POU-domain factor Oct-1 with the virion protein VP-16 in promoting lytic infection through activation of the viral immediate-early genes is well established. Several prior studies have suggested that the B-cell transcription factor Oct-2 mediates herpes virus latency by repressing viral immediate-early gene expression via octamer/TAATGARAT motifs (9 -14 ,27 ). These studies have employed polymerase chain reaction to identify multiple splice forms of Oct-2 in sensory neuron-related cells, and have suggested that these specific neuronal splice forms of Oct-2 repress expression of viral immediate-early genes via octamer sites. However, recent studies (10 ,20 ) have failed to detect Oct-2 message and protein expression in adult trigeminal and dorsal root ganglia, and thus a role for Oct-2 in the viral life cycle in sensory neurons appears unlikely.
In contrast to Oct-2, Brn-3.0 is activated early in the genesis of the sensory ganglia (25 ), and persists in the adult (9 ,20 ,22 ). Three members of the Brn-3 family are expressed in the adult trigeminal ganglion, but expression of Brn-3.0 predominates. We have used the POU-domain of Brn-3.0 to screen the HSV genome for Brn-3.0 binding sites, and identified seven sites of high-affinity Brn-3.0 binding. For several reasons, it is highly likely that this affinity screen has identified the highest affinity binding sites for Brn-3.0 in the HSV genome. Most of the sites identified were obtained multiple times in the screen, the number of clones obtained corresponds approximately to the number of high affinity bands seen on electrophoretic analysis of the selected DNA, and a data base search for additional sites based on the consensus sequence derived yielded only one additional site. To our knowledge, this is the first exhaustive screen of a genome of this size for the recognition element of a specific class of transcription factors.
All seven of the Brn-3.0 binding sites identified in the screen (eight including database screening) are located outside the known HSV open reading frames. Because ~78% of the HSV sequence has been assigned to open reading frames (30 ), this is highly unlikely to have occurred by chance (P ~5 * 10-6 for eight sequences). The Brn-3.0 binding sites are either preferentially included in non-coding sequence or excluded from coding sequence. However, the specific location of the sites with respect to known HSV genes is less revealing. One of the selected clones (101) encompasses part of the ICP0 gene promoter, a known regulatory region. Several others lie between open reading frames for HSV late genes (Table 1 ), but are not associated with known regulatory elements.
It is unlikely that the Brn-3.0 binding sites in the HSV genome have occurred entirely by chance. Because the HSV genome consists of 68.3% GC residues (30 ), the probability of an A residue at a given position is 0.1585. Similarly, the probability of T is 0.1585, and an ambiguous A/T is 0.318. If these base pairs are assumed to occur at random throughout the HSV genome, the consensus sequence selected by Brn-3.0: T A/T A A T N A N T A/T, which encompasses clones 101b, 107, 123 and 128, plus one additional site identified by database search, would be expected to occur with a probability of (0.1585)6 * (0.317)2 = 1.6 * 10-6, or ~1/600 000. Because both strands of the HSV genome represent ~300 000 bp, this random probability is equivalent to 0.5 occurrences per genome, compared to the actual rate of five occurrences per genome. Thus the consensus sequence occurs with ~10 times the expected frequency. However, the non-random frequency of Brn-3.0 binding sites may occur for reasons which are irrelevant to Brn-3.0 binding. For instance, some of these sites overlap putative polyadenylation signals for the adjacent genes. However, the conserved polyadenylation site A A/T T A A A occurs >100 times in the genome and is not, by itself, sufficient to confer high-affinity Brn-3.0 binding.
In spite of the specificity of binding of these sites, and their occurrence in the HSV genome at far greater than expected frequencies, the Brn-3.0 recognition sites identified here are not likely to directly mediate transcriptional activation in vivo by Brn-3.0 acting as sole tissue-specific transactivator. Recent results have demonstrated that the optimum recognition site for proteins of the Brn-3 family, including Brn-3.0, Brn-3.2 and the Caenorhabditis elegans homologue Unc-86, contain the core recognition sequence GCATAATTAAT, with some flexibility in the occurrence of A or T residues at positions 5, 7 and 9 of this sequence (31 ). Furthermore, to effectively mediate transcriptional activation by Brn-3.0, recognition sequences which deviate from the optimal site must form complexes with Brn-3.0 which have dissociation constants within ~10-fold of the ideal sequence. In contrast, EMSA and footprinting methods may detect specific binding to sites with 100-1000-fold higher dissociation constants than the optimal sequence, demonstrating that these methods are not sufficiently specific to identify transcriptionally active sites for Brn-3.0. The CRH model site used here binds Brn-3.0 with 4-fold lower affinity than the optimal site, and transactivation by Brn-3.0 on this site has been demonstrated (22 ,31 ). However, none of the selected viral sites, which exhibit affinities for Brn-3.0 that are 20-100-fold lower than the CRH site, binds to Brn-3.0 with sufficient affinity to activate transcription in epithelial (CV-1) cell transfection models.
It remains to be seen whether the sites of Brn-3.0 interaction with the HSV genome are physiologically relevant to the viral life cycle and latency process. It appears unlikely that Brn-3.0 acts alone to modify viral gene expression at these sites, but it may function in concert with a cellular or viral partner, or it may act indirectly on viral gene expression. Another possibility is that the sites identified by the Brn-3.0 affinity screen may represent physiologically relevant sites for interaction with yet-unidentified cellular factors with DNA-binding characteristics somewhat similar to Brn-3.0, such as the products of other homeobox-containing genes. In either case the experiments described illustrate an efficient method for exhaustively screening large viral genomes, or sub-genomic fragments of eukaryotic DNA such as cosmids or yeast artificial chromosomes, for specific transcription factor recognition sites, which will facilitate the testing of specific models of gene regulation.
Supported in part by the Pfizer New Faculty Award, the Howard Hughes Fellowship for Physicians, and Department of Veterans Affairs Merit funding (E.E.T.). E.E.T. is a NARSAD Young Investigator and is supported by the Scottish Rite Schizophrenia Research Program. Work on Brn-3.0 recognition sequences in the HSV genome was initiated in the laboratory of Dr M.G.Rosenfeld.
REFERENCES