Evidence for a Bidirectional Element Located Downstream from the Herpes Simplex Virus Type 1 Latency-Associated Promoter That Increases Its Activity during Latency

Plasmids.

The plasmids used in this study are displayed in Fig. 1A. Plasmid pHB18 was derived, with minor modifications, from plasmid pJA1 described previously (23). It contains the LAP extending from a SmaI site to a SacII site (positions 117938 to 118843 on the HSV-1 genome) upstream from the reporter gene lacZ (from pCH110; Pharmacia). Plasmids pHB22F and pHB22R were derived from pHB18 by first inserting the LTE sequence from a PstI site to a HpaI site (positions 118862 to 120303) as a BamHI-BglII fragment into the BglII site of the intron sequence (GGATCCAGGTAAGCCTAGATCTCTAACCATGTTCATGCCTTCTTTTCCTAGGATCC) either in the forward (pHB22F) or in the reverse (pHB22R) orientation and then inserting the LTE-intron sequence as a BamHI-BamHI fragment into a BamHI site between LAP and lacZ of pHB18. Plasmid pHB23 was similar to pHB22F, but the LTE sequence was replaced by an inverted 855-bp DNA fragment internal to the coding sequence of the I-Sce1 gene of Saccharomyces cerevisiae. Plasmids pHB17.1, pHB12F, and pHB12R are the counterparts of pHB18, pHB22F, and pHB22R, respectively, with the LAP being replaced by the cytomegalovirus (CMV) enhancer-promoter. Plasmid pHB30F was derived from pHB17.1 by inserting the LTE sequence (BamHI-BglII fragment) alone in the forward orientation into the BamHI site between the CMV promoter and the lacZ gene of pHB17.1. Plasmids pHB16, pHB19F, and pHB19R are the counterparts of pHB18, pHB22F, and pHB22R, respectively, with the LAP being replaced by the HSV-1 thymidine kinase (TK) promoter (from a PvuII site at position 48108 to a BglII site at position 47855).

Cells and viruses.

African green monkey kidney (Vero), rabbit skin (RS), and ND7 cells were maintained in Dulbecco modified Eagle medium plus 10% fetal bovine serum, in Earle's minimum essential medium plus 5% newborn calf serum, and RPMI 1640 medium plus 10% fetal bovine serum, respectively. All media were also supplemented with penicillin and streptomycin and were buffered with sodium bicarbonate. Vero cells were used for selection, propagation, and titer determination of the recombinant viruses. RS and ND7 cells were used for gene expression analysis following transfection of plasmids or productive infection.

The recombinant viruses described in this report, hereafter designated KOS18, KOS22F, and KOS22R, were constructed by recombination at the gC locus between the parental virus KOS dl1.8 HSV-1 (kindly provided by David Leib [22]) and plasmids pHB18, pHB22F, and pHB22R, respectively. These viruses were selected, plaque purified, and produced as previously described (10).

Animal inoculation.

All animals were manipulated using institutionally approved animal welfare procedures. Experiments were carried out essentially as previously described (10). Briefly, 6-week-old female Swiss Webster mice were anesthetized by ether inhalation and a 10% NaCl solution was injected subcutaneously in the footpads. After 3 h, the mice were anesthetized by intraperitoneal injection of pentobarbital (60 mg/kg). The skin of the footpads was removed, and 10⁷ PFU of the viral stocks was applied on each footpad. The infections were carried out for at least an hour.

Histochemical staining.

At the times indicated in Fig. 5, infected mice were euthanized by Halothane inhalation and exsanguinated by transcardiac perfusion with phosphate-buffered-saline (PBS) through the left ventricle after cutting the right atrium. The mouse tissue was fixed with a PBS solution containing 2% formaldehyde and 0.2% glutaraldehyde. The L3 to L6 dorsal root ganglia (DRG) were surgically removed and further fixed for 15 min in the same fixative solution. Histochemical staining was conducted exactly as previously described (10).

Immunohistochemistry.

At the times indicated in Fig. 7, mice were euthanized by carbon dioxide inhalation and transcardiac perfusion with PBS (pH 7.2) followed by perfusion with 4% paraformaldehyde in PBS. The dorsal root ganglia from three experimental animals were then removed, combined, and immersion fixed at 4°C in 4% paraformaldehyde for 1 h. Fixed tissue was then equilibrated with 20% sucrose, embedded in OCT, and frozen in liquid nitrogen (23). Serial sections (6 μm) were cut from frozen ganglia and collected as five alternative sets onto a Superfrost plus slide (Fisher). The tissue slides were stored at −20°C.

For dual-immunofluorescence studies of β-galactosidase and HSV antigen expression, tissue sections were first incubated for 1 h at room temperature with rabbit anti-β-galactosidase antiserum (Cappel, Durham, N.C.) diluted 1:700 in PBS. Tissue sections were then sequentially incubated with biotinylated goat anti-rabbit immunoglobulin G (Vector Labs, Burlingham, Calif.) for 1 h, rhodamine600 avidin D (Vector) diluted 1:1,200 for 40 min, and biotinylated goat anti-avidin D diluted 1:1,200 for 40 min. After being washed, the sections were blocked with 10% normal rabbit serum for 10 min and incubated for 40 min with both fluorescein isothiocyanate-conjugated rabbit anti-HSV-1 (Dako Corp., Carpinteria, Calif.) diluted 1:100 and rhodamine600 avidin D diluted 1:1,200. Sections were then washed and mounted with coverslips using Vectashield (Vector). Stained slides were evaluated using a Nikon fluorescence photomicroscope.

Quantification of β-galactosidase activity.

Tissue culture plates (60 mm) containing 1 × 10⁶ (transfection) or 3 × 10⁶ (infection) RS (fibroblast) or ND7 (neuronal) cells were either transfected with 2 μg of DNA of the indicated plasmids previously mixed with 10 μl of Lipofectin as specified by the manufacturer (Gibco) or infected at a multiplicity of infection of 3 PFU/cell. At 2 days postinfection (p.i.) or at 2 days posttransfection the cells were harvested in 200 μl of PBS and immediately frozen.

Unfixed L3 to L6 DRG from infected mice (three mice per time point per virus) were removed as indicated above and immediately frozen. Cellular extracts were produced by grinding these DRG in 200 μl of ice-cold PBS in a 0.1-ml Dounce homogenizer. Prior to use, the cells were frozen and thawed three times and cellular debris were briefly microcentrifuged. Aliquots (10 to 40 μl) of the supernatants were incubated in the previously described in vitro β-galactosidase assay (9), using chlorophenolred-β-d-galactopyranoside (CPRG; Sigma) as the substrate.

Validation of the intron strategy.

We previously showed that a DNA fragment 3′ of the LAP but within the LAT transcription unit was able to support long-term expression of the LAP during latency (23). Because the lacZ mRNA was relatively unstable, leading to a signal barely detectable by the RNase protection assay, it was difficult to quantify the effect of the LTE on the activity of LAP. Another difficulty of the previous assay system is that the presence of the LTE sequence, inserted between the promoter and the lacZ coding region, prevented translation of the lacZ mRNA (23). To address these problems, we inserted the LTE DNA fragment (positions 118862 to 120303) into a cassette flanked by the consensus splice donor, branch point, and splice acceptor sites and cloned the LTE/intron sequence between the promoters and the lacZ reporter gene (see Materials and Methods) (Fig. 1A). This construction should allow the LTE region to function while allowing translation of the reporter gene, which would increase the sensitivity of our assays. To demonstrate the effect of the splicing signals on mRNA translation, we inserted the LTE, either alone or within an intron, between the CMV immediate-early enhancer-promoter and the lacZ gene. We then transfected the plasmids in RS cells and analyzed the reporter gene expression using an in vitro enzymatic assay to quantify the β-galactosidase activity. As shown in Fig. 2, β-galactosidase activity is very high for transfection of a control plasmid in which the CMV immediate-early enhancer is driving the lacZ gene, pHB17.1. Insertion of the LTE fragment without flanking intron signals, pHB30F, blocked the majority of lacZ expression (compare pHB17.1 to pHB30F). By contrast, insertion of the LTE as an intron between the promoter and the lacZ gene resulted in transient-transfection assays with high levels of β-galactosidase activity. This was true whether the LTE was inserted in the forward (natural) direction, pHB12F, or in the reverse direction, pHB12R. In each case, the level of expression was very similar to that obtained with the CMV-driven lacZ gene, pHB17.1. In other words, insertion of the LTE within the context of an intron allowed efficient translation of the lacZ mRNA but did not decrease or increase the level of expression compared to that obtained with a plasmid lacking the LTE and splicing signals. From these results, we conclude that insertion of the LTE without splicing signals prevented efficient expression of a downstream coding sequence (as stated previously) and that the LTE/intron sequence merely restored the same level of CMV promoter-enhancer activity. Therefore, we could apply this approach to other experiments to further analyze the role of the LTE.

The LTE region has an enhancer activity for the LAP.

To construct viruses containing the LTE/intron in the context of the LAP, we first constructed a set of plasmids and examined their expression in transient-transfection assays. In our previous work, we had inserted a similar but slightly shorter LTE DNA fragment between the LAP and the lacZ gene. As explained in the introduction, this LTE fragment blocked translation of the lacZ mRNA (24). We again transfected this construct, under the CMV enhancer, in pHB30 (see above) and again observed a strong block of translation, which was alleviated when the LTE fragment was placed in an intron. Because we had already shown that the LAP-LTE constructs were not translated, we excluded them from further study here. Instead, a control plasmid, pHB18, was constructed, in which the LAP was placed in front of the lacZ gene (Fig. 1A). The activity of this plasmid was compared to that of two others in which the LTE was inserted in the context of an intron between the LAP and the lacZ gene, in either the forward (pHB22F) or the reverse (pHB22R) direction (Fig. 1A). These three plasmids were transfected into ND7 cells, a neuronal cell line representing a fusion between neuroblastoma cells and a primary culture of neurons originating from DRG. Surprisingly, plasmids containing the LTE within an intron (pHB22F and pHB22R) showed a high increase in lacZ expression compared to the plasmid (pHB18) lacking an LTE (Fig. 3A). Thus, the result of inserting the LTE DNA fragment into an intron between the LAT promoter and lacZ was to increase lacZ expression. To show that this increased expression was due specifically to the LTE, a DNA fragment from the I-Sce1 gene of S. cerevisiae was inserted in the reverse orientation into the intron in place of the LTE, leading to plasmid pHB23. The transient-expression experiment performed with pHB18, pHB22F, pHB22R, and pHB23 in ND7 cells showed that this I-Sce1 DNA fragment was not able to increase the expression of lacZ (Fig. 3B). Therefore, the LTE sequence is able to specifically increase the expression of the reporter gene when placed in an intron between LAP and lacZ, in a manner independent of its orientation. Considering that the LTE, when not placed in the context of an intron, was unable to increase the expression of either the LAT or the CMV enhancer (23) (Fig. 2), it is unlikely that the increased expression of pHB22F and pHB22R over the control pHB18 is due to the activity of a promoter within the LTE. Furthermore, if this activity were due to a promoter, one would expect it to be directional; thus, the fact that both orientations of the LTE increased the level of expression of the LAT promoter also suggests that the LTE does not contribute a promoter-like activity in these transfections.

The LTE enhancer activity is not tissue specific or promoter specific.

To determine if the LTE functions in a tissue-specific manner, rabbit skin cells (fibroblasts) were transfected with the three plasmids, pHB18, pHB22F, and pHB22R. The results showed a similar fold increase in lacZ gene expression for both pHB22F and pHB22R over pHB18 in these cells to that in ND7 cells (Fig. 3A). Therefore, the enhancer function of the LTE sequence seemed not to be specific to a particular cell line.

To study a potential effect of the LTE region on another promoter than LAP, another set of plasmids was also constructed. Plasmids pHB16, pHB19F, and pHB19R, bearing the lacZ gene downstream from an early viral promoter, the TK promoter (position 48108 to position 47855), alone (pHB16) or with the LTE inserted in the forward (pHB19F) or reverse (pHB19R) orientation into the intron (Fig. 1a) were then transfected into RS cells. The results of this transient-expression experiment showed that lacZ expression from pHB19F and pHB19R was significantly higher than that from pHB16 (Fig. 3C), suggesting that the LTE enhancer can function with another promoter than LAP and is therefore not specific to a promoter.

The LTE region increases gene expression in an orientation-dependent manner during productive infection in vitro.

To study the effect of the LTE region on the activity of the LAT promoter in the context of the virus, the three plasmids pHB18, pHB22F, and pHB22R were inserted into the gC locus of HSV-1. Because of the extensive homology to the two copies of the LTE and LAP in the long repeat regions, HSV-1 KOS dl1.8, a LAT region deletion virus described previously by Leib et al. (22), was used as the parental strain. Recombinant viruses were screened for the insertion of the lacZ gene and plaque purified a total of five times. The DNA structure of each recombinant virus was verified by Southern blot hybridization (data not shown). RS cells and ND7 cells were infected at a multiplicity of infection of 3 PFU per cell. At 2 days later, the infected cells were harvested and β-galactosidase activity was determined using an in vitro enzymatic assay. As shown in Fig. 4, there was no significant difference in lacZ expression between KOS18 and KOS22R. Only KOS22F-infected cells showed an increase beyond the control KOS18-infected cells. Since it has been shown that the ICP4 protein down regulates the LAP during productive infection by binding to the ICP4 binding motif at the LAP cap site (1, 2, 14), we assume that LAP is repressed in all three viruses during the lytic cycle. Therefore, the lacZ expressions observed with KOS18 and KOS22R corresponded to a background expression from cells infected with lacZ coding sequence-containing viruses. As expected from our previous study of LAP activity, the background expression from KOS18, KOS22F, and KOS22R is much higher in ND7 than in RS cells (9).

lacZ expression arising from KOS22F-infected cells was significantly higher than that from KOS18- and KOS22R-infected cells in the same infected cell line, indicating that the lacZ gene is highly expressed during productive infection in vitro only when the LTE is inserted into the intron in the forward orientation. These results, markedly different from those obtained in transient expression with the plasmids, showed that lacZ gene expression is dependent on the orientation of the LTE. The LTE contains LAP2, a viral promoter that is active only in the lytic cycle (17). Although LAP2 expression could explain the increased expression from virus 22F, it is several hundred bases from the splice acceptor site and would require the utilization of a cryptic splice donor site to avoid several ATG codons upstream of the lacZ gene. Perhaps a more likely reason for increased expression from the 22F virus is the presence of a TATA element just upstream of the HpaI site, only a short distance from the end of the LTE. This TATA box would be activated by ICP4, rather than repressed; furthermore, this element is very far away from the lacZ gene in virus KOS22R.

The LTE region increased gene expression during acute and latent infection in vivo.

To examine the effect of the LTE on LAP expression in vivo, mice were infected via the footpads of the rear limbs. On days 4 and 12, DRG were removed and analyzed for β-galactosidase activity by in situ staining. As shown in Fig. 5, whole ganglia stained for β-galactosidase activity showed a large increase in KOS22F- and KOS22R-infected ganglia compared to KOS18-infected ganglia at 4 days p.i. By 12 days p.i., both KOS22F and KOS22R showed less total activity than on day 4 but substantially more activity than KOS18 on day 12, for which no detectable expression was observed.

To quantify the lacZ gene expression and to examine the effect of the enhancer over a longer period of latency, we extended the period of infection. At 4, 12, and 28 days p.i., unfixed L3 to L6 DRG were removed and used to produce cellular extracts, which were subsequently analyzed for β-galactosidase activity in an in vitro quantitative assay (CPRG assay). The results showed that at 4 days p.i., β-galactosidase activity from KOS22F-infected DRG was 18-fold higher than from KOS18-infected DRG and 3-fold higher than from KOS22R-infected DRG (Fig. 6). At 12 days p.i., however, expression from both KOS22F- and KOS22R-infected DRG were clearly reduced by about 50%. As expected, by 12 days p.i., lacZ gene expression originating from KOS18-infected DRG was not significantly different from the background level. Also, at 28 days p.i., no β-galactosidase activity was observed for KOS18-infected DRG. This has been a consistent observation in our laboratory, i.e., that just pairing the upstream LAP region to the lacZ gene does not result in long-term expression in vivo (9, 22). By contrast, β-galactosidase activity could still be detected at 28 days p.i. for both KOS22F- and KOS22R-infected DRG, although the level of expression was markedly reduced compared to the expression at 4 days p.i.

It has been our experience that by 42 days p.i., the level of LAP activity in vivo has essentially reached a plateau. This is true for measuring the levels of the LAT intron, or when the LAP was paired with a long terminal repeat element from a retrovirus to generate long-term expression from a reporter virus (24). To verify that the level of expression of LAP-LTE activity was relatively constant, mouse ganglia latently infected with KOS22F and KOS22R were periodically removed and stained for activity. A level of expression similar to that at 28 days p.i. was also observed in KOS22F- and KOS22R-infected ganglia at 60 days and at 6 months p.i. A picture of the 6-month data is included in Fig. 5G. Thus, the presence of the LTE/intron sequence initially stimulated a very dramatic increase in lacZ gene expression in vivo, which subsequently declined to a lower level. The fact that the neurons remained blue when infected by a virus harboring the LTE/intron sequence could be interpreted in different ways and is discussed below.

It was clear from the data shown in Fig. 5 and 6, that both KOS22F- and KOS22R-infected ganglia expressed far more β-galactosidase protein than did KOS18-infected ganglia. To ensure that the difference in gene expression observed in KOS18-, KOS22F-, and KOS22R-infected ganglia was not due to differences in the abilities of the viruses to grow in vivo, the titers of each virus per time point per mouse were determined. As shown in Table 1, titers of KOS18, KOS22F, and KOS22R ranged from 1.8 × 10³ to 3.1 × 10³ at 4 days p.i., and as expected, no infectious virus could be recovered at 12 or at 28 days p.i. Thus, the differences in expression observed on days 4 and 12 were not due to differences in viral growth.

Immunostaining for β-galactosidase and viral antigens in infected DRG.

Margolis et al. have previously shown that by 4 days p.i. most neurons expressing β-galactosidase from the LAT promoter were in the latent phase of infection and did not express viral antigen, although a small percentage of neurons expressed both β-galactosidase and viral antigens (24). However, we showed earlier in this report that KOS22F virus was able to express β-galactosidase protein in lytically infected cells in culture. Therefore, it was important to know if the cells staining for β-galactosidase at 4 or 12 days p.i. were lytically or latently infected. In the present experiment, DRG infected for 4 days (KOS22F) or for 12 days (KOS22F and KOS22R) were sectioned and double stained using antisera against β-galactosidase and viral antigens. As qualitatively illustrated in Fig. 7, no viral antigen-positive neurons were detected at 12 days after KOS22F or KOS22R infection, confirming our previous data that no infectious virus could be recovered at that time (Table 1). By contrast, at 4 days p.i., KOS22F- and KOS22R-infected ganglia harbored both lytically and latently infected neurons (Fig. 7).