Selectivity of an Oncolytic Herpes Simplex Virus for Cells Expressing the DF3/MUC1 Antigen Free

Viruses have evolved extremely efficient mechanisms to infect cells, subvert cell defenses, deliver their genetic payload, express viral genes, and produce progeny virions. Most viruses used in clinical trials of gene therapy have been rendered replication-defective by genetic engineering such that they serve mainly as gene delivery vehicles and do not replicate outside of specialized packaging cell lines (1). There has been growing interest, however, in relying on the efficiency of viral replication itself as a means to destroy cancer cells in a process referred to as viral oncolysis (2). The safety and efficacy of this approach are dependent on selective viral replication in cancer cells rather than in normal cells.

Several strategies have been explored to restrict viral replication to neoplastic cells, including use of tumor-associated promoters to regulate expression of genes critical for viral replication. A promoter/enhancer sequence for the prostate-specific antigen (PSA) gene has been used to regulate adenoviral E1A expression to restrict its replication to PSA-positive cells (3). A promoter sequence of the α-fetoprotein gene has been used to regulate expression of both E1A and E1B55kD in an adenoviral mutant that replicates selectively in tumors that express α-fetoprotein (4). The E2F-responsive B-myb promoter has been used to regulate expression of a gene critical for replication of Herpes simplex virus 1 (HSV-1; Ref. 5).

DF3/MUC1 is a tumor-associated antigen that is overexpressed on many human carcinomas, including breast, pancreatic, and colon cancer (6, 7, 8). DF3/MUC1 transcript overexpression is observed in breast cancer (9), and the 5′ flanking region of the gene has been characterized (10). DF3/MUC1 gene expression is regulated by sequences between positions −598 and −485 bp upstream from the transcription start site. This promoter/enhancer has been used to regulate expression of E1A in an adenoviral mutant, Ad.DF3-E1, which replicates preferentially in DF3/MUC1-positive cancer cells (11).

HSV-1 is an effective oncolytic virus in animal models (12, 13, 14), and clinical studies of HSV-1 for oncolysis have been conducted. G207 is a replication-conditional HSV-1 mutant that has been administered to patients with recurrent malignant glioma (15). The HSV-1 mutant 1716 is defective in expression of HSV-1 γ₁34.5 (16) and has been administered to patients with recurrent malignant glioma in a clinical trial (17).

We have previously demonstrated that HSV-1 mutants defective in viral ribonucleotide reductase replicate preferentially in colon cancer liver metastases rather than normal liver because of higher mitotic activity and higher levels of functionally complementing cellular ribonucleotide reductase in the metastases (18). The destruction of these liver tumors is a result of viral replication rather than host-immune responses (19, 20). In this study we constructed and characterized a mutant HSV-1 in which the γ₁34.5 gene is regulated by a DF3/MUC1 promoter. Regulation of HSV-1 γ₁34.5 function results in preferential viral replication and oncolysis in cancer cells that express DF3/MUC1, restricted biodistribution in vivo, and less toxicity as assessed by LD₅₀. This HSV-1 mutant was effective against carcinoma xenografts in nude mice.

Vero African Green Monkey kidney cells were obtained from American Type Culture Collection (Manassas, VA). MC26 mouse colon carcinoma cells were obtained from the National Cancer Institute Tumor Repository (Frederick, MD). A375 human melanoma cells were provided by Isaiah Fidler (M. D. Anderson Cancer Center, Houston, TX), and MCF-7 cells were provided by Donald Kufe (Dana-Farber Cancer Institute, Boston, MA). SW1990 and CAPAN2 human pancreatic carcinoma cells were provided by Andrew Warshaw (Massachusetts General Hospital). Primary human hepatocytes were prepared as described previously (21). Human umbilical vascular endothelial cells were obtained from Cell Applications, Inc. (San Diego, CA). HSV-1 viruses F strain (wild-type HSV-1) and R3616 (defective γ₁34.5 expression) were provided by Bernard Roizman (University of Chicago, Chicago, IL; Ref. 22). MGH1 is a HSV-1 mutant defective in thymidine kinase (TK) expression and viral ribonucleotide reductase (ICP6) expression and was provided by E. Antonio Chiocca (Massachusetts General Hospital). Viruses were propagated and titered on Vero cells, and heat-inactivation of virus was performed as described (23).

The coding sequence of the γ₁34.5 gene was isolated from pBGL34.5 (5) as a Nco-SacI fragment and cloned into pLitmus28 (New England BioLabs, Beverly, MA) by use of the same restriction sites. The DF3/MUC1 promoter, provided by Donald Kufe (Dana-Farber Cancer Institute, Boston, MA), was isolated from pDF3 (11) as a SpeI-XhoI fragment and subcloned into pCRII (Invitrogen, San Diego, CA) in the same sites. A NheI-SpeI fragment containing autofluorescence protein (AFP) regulated by a cytomegalovirus (CMV) promoter was isolated and subcloned into the SpeI locus of this plasmid. A NsiI-XhoI restriction fragment of this plasmid containing AFP regulated by a CMV promoter and the DF3/MUC1 promoter was then subcloned into the same sites of pLitmus28 with the DF3/MUC1 promoter immediately upstream of the γ₁34.5 gene. This double-gene-expression cassette was then removed as a SpeI-KpnI fragment and subcloned into NheI and KpnI sites in pcDNA3.1(−) (Invitrogen). The cassette was then isolated as a BglII-PvuII fragment and subcloned into the BglII and SnaBI sites of HSV106 (kindly provided by Steven McNight, University of Texas Southwestern, Dallas, TX), such that the cassette is flanked by sequences of the HSV-1 TK gene. This plasmid was linearized with XbaI and cotransfected with R3616 viral DNA into Vero cells with Lipofectamine (Life Technologies, Inc., Gaithersburg, MD). Cells and media were collected 5–7 days after transfection when cytopathic effects were evident. Progeny virions were recovered from cells after three freeze-thaw cycles and then placed on a monolayer of Vero cells in the presence of ganciclovir. After the monolayer was overlaid with agarose, green fluorescent plaques were observed with fluorescence microscopy and selected as potential recombinants. Isolates were subjected to four rounds of plaque purification before their genetic identity was examined by Southern blot analysis.

Viral DNA was isolated after lysis of infected Vero cells with 0.5% SDS and proteinase K (500 μg/ml) by repeated phenol–chloroform extraction and ethanol precipitation. DNA was digested with PstI, separated by agarose gel electrophoresis, and transferred to a nylon membrane (Amersham Corp., Arlington Heights, IL). A probe to the TK gene was created by PCR amplification of HSV-1 DNA with the following primers: 5′-TACCCGAGCCGATGACTTACTG-3′ and 5′-CCAACACCCGTGCGTTTTATTC-3′. A probe to the DF3/MUC1 promoter was created by PCR amplification of HSV-1 DNA using the following primers: 5′-AGAAGGGTGGGGCTATTCCG-3′ and 5′-GCAGGTGACAGGTGACAAAACC-3′. PCR product were labeled with use of a random-prime labeling kit (Amersham Pharmacia Biotech, Piscataway, NJ) and purified in a spin column. After hybridization of the probe to the membrane, the membranes were washed and exposed to film.

Viral replication assays were performed as described (19). Briefly, 3 × 10⁶ cells were infected with 6 × 10⁶ plaque forming units (pfu) of virus for 2 h, at which time unadsorbed virus was removed by washing with a glycine–saline solution (pH 3.0). Forty h after infection the supernatant and cells were harvested, exposed to three freeze–thaw cycles to release virions, and titered on Vero cells. Viral cytotoxicity assays were performed as described (18). Briefly, cells were plated in 96-well plates at 5000 cells/well for 36 h. Virus was added at multiplicity of infection values ranging from 0.0001 to 10 and incubated for 6 days. The number of surviving cells was quantitated by a colorimetric 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. Experiments were performed three times in quadruplicate, and results of representative experiments are shown.

This assay was performed as described previously (20). Briefly, bacterial fusion proteins consisting of a His-tagged eIF-2α and glutathione-S-trans-ferase-tagged protein kinase R (PKR) were purified in nondenaturing conditions. Purified His-eIF-2α was reacted with glutathione-S-transferase-PKR in the presence of [γ³²P]ATP to radioactively label the His-eIF-2α. S10 fractions were prepared from infected or mock-infected cell lysates and reacted with the ³²P-labeled His-eIF-2α at 32°C for 1 or 5 min. The proteins were resolved on 10–20% gradient gels, and the ³²P remaining in eIF-2α-³²P was quantified by an image analyzer (ImageQuant; Molecular Dynamics, Sunnyvale, CA). Experiments were performed three times, and results of representative experiments are shown.

Cells were washed with cold PBS, incubated with 1 μg/ml mouse antihuman CD-227 (kindly provided from Dr. Kufe) on ice for 30 min, and again washed with cold PBS. Cells were incubated with FITC-conjugated goat antimouse IgG (Biosource, Camarillo, CA) on ice for 30 min, washed with cold PBS, and then analyzed by flow cytometry. Fluorescence Intensity was determined for 2 × 10⁴ cells.

Animal studies were performed in accordance with policies of the Massachusetts General Hospital. BALB/c (nu/nu) mice were obtained from Charles River Laboratories, Inc. (Wilmington, MA). LD₅₀ was assessed by a single tail vein inoculation of 1 × 10⁷, 5 × 10⁷, 1 × 10⁸, or 5 × 10⁸ pfu of R3616, RH105, DF3γ34.5, or F strain, after which the mice were followed for survival. In separate experiments, A375 or CAPAN2 tumors resected from mouse flanks were cut into 2-mm³ pieces and implanted s.c. in BALB/c (nu/nu) mouse flanks. Two weeks later, when the tumors measured 5 mm in diameter, 1 × 10⁸ pfu RH105, MGH1, or DF3γ34.5 or PBS was inoculated directly into the tumors (n = 6/group). Tumor sizes were measured every 5 days. Experiments were performed twice, and results of representative experiments are shown. In another set of experiments, F strain (5 × 10⁶ pfu), RH105 (1 × 10⁸ pfu), R3616 (1 × 10⁸ pfu), or DF3γ34.5 (1 × 10⁸ pfu) was inoculated into flank tumors. Four days later mice were sacrificed, and organs were harvested for analysis of extracted DNA by PCR amplification of HSV-1 sequences.

PCR amplification of HSV-1-specific sequences to investigate the biodistribution of HSV-1 in mice was performed as described (20). Forward oligonucleotide primer 5′-GGAGGCGCCCAAGCGTCCGGCCG-3′ and reverse oligonucleotide primer 5′-TGGGGTACAGGCTGGCAAAGT-3′ were used to amplify a 229-bp fragment of HSV-1 DNA polymerase gene. BALB/c mouse tissues were incubated in digestion buffer [10 mm Tris-HCl (pH 7.4), 5 mm EDTA, 0.5% SDS, 200 μg/ml proteinase K (pH 8.0)] at 56°C overnight. After phenol–chloroform (1:1) extraction, DNA was precipitated in 70% ethanol, lyophilized, and resuspended in distilled water. We then subjected 0.1 μg of DNA to PCR amplification. PCR reactions were performed in a 25-μl volume using rTth DNA polymerase according to the manufacturer’s instructions (Perkin-Elmer Applied Biosystems, Foster City, CA) in a DNA Thermal Cycler 480 (Perkin-Elmer Applied Biosystems). Cycling conditions were for 35 cycles of 95°C for 1 min, 60°C for 1 min, and 72°C for 1 min. Appropriate negative controls were used for all PCR reactions, and no contamination of reagents was detected.

Wedge biopsies consisting of colon carcinoma liver metastases and surrounding normal liver were obtained fresh from the operating room in accordance with a protocol approved by the Massachusetts General Hospital Institutional Review Board. Small blood vessels were immediately cannulated and perfused with PBS. Through these same cannulas, either DF3γ34.5 or F strain (1 × 10⁸ pfu of either) was slowly injected into the liver section and allowed to dwell for 5 min. Uniform distribution throughout both normal liver and the metastases was confirmed with an injection of methylene blue. The liver was then sectioned into pieces measuring 5 mm in thickness and cultured in hepatocyte medium with 20% fetal bovine serum. Forty-eight h later portions of either the normal liver or liver metastases were dissected free, weighed, and homogenized in PBS containing collagenase (1 mg/ml). Dilutions were plated on monolayers of Vero cells to determine the titer of infectious virus in the tissues. This experiment was performed in four human liver specimens.

Two nonparametric statistical analyses, the log-rank test and Peto–Wilcoxon test, were used to compare survival between groups (InStat; Graphpad Software, New York, NY).

To develop a HSV-1 mutant with attenuated replication in MUC1-negative cells, we used a DF3/MUC1 promoter sequence to regulate γ₁34.5 expression. We first constructed a plasmid containing an expression cassette in which γ₁34.5 expression regulated by a DF3/MUC1 promoter sequence and an AFP gene is driven by a strong immediate-early CMV promoter (Fig. 1). Using homologous recombination techniques, we introduced this dual gene cassette into the TK locus of R3616, which is a HSV-1 mutant that harbors deletions in both γ₁34.5 loci (22). After four rounds of plaque purification, one isolate was selected and designated DF3γ34.5. The genotype of this mutant was confirmed by Southern blot analysis.

To assess γ₁34.5 function in HSV-1 infected cells, we analyzed eIF-2α dephosphorylation in infected cells. eIF-2α is normally phosphorylated by PKR in response to HSV-1 infection, and HSV-1 γ₁34.5 interacts with cellular protein phosphatase 1α to dephosphorylate eIF-2α to block the shutoff of protein synthesis (Fig. 2). We compared γ₁34.5 function in cells infected by F strain (wild-type HSV-1), R3616 (γ34.5-deficient), and DF3γ34.5 (γ34.5 expression regulated by the DF3/MUC1 promoter). A375 melanoma cells and Vero cells are DF3/MUC1-negative, whereas CAPAN2 cells are DF3/MUC1-positive (Fig. 3,A). As expected, eIF-2α dephosphorylation activity was observed after infection with wild type (F strain) in all three cell lines because this virus expresses wild-type γ₁34.5 (Fig. 3 B). Also as expected, no eIF-2a dephosphorylation activity was observed after infection with R3616 (defective γ₁34.5) in any of the three cell lines. Of note, eIF-2α dephosphorylation activity was observed with DF3γ34.5 infection only in the DF3/MUC1-positive CAPAN2 cells, and not in the DF3/MUC1-negative A375 or Vero cells. This eIF-2α dephosphorylation pattern observed in cells infected by DF3γ34.5 is consistent with regulation of γ₁34.5 expression by the DF3/MUC1 promoter and represents one of the fundamental mechanisms by which we intend to regulate viral replication in this engineered HSV-1 mutant.

We compared DF3γ34.5 replication with that of other HSV-1 mutants in DF3/MUC1-negative and -positive cells. HSV-1 mutants used for comparison were R3616 (γ₁34.5⁻), RH105 (U_L23⁻), and MGH1 (γ₁34.5⁻ and U_L39⁻; Table 1). The intended phenotype of DF3γ34.5 is for its replication to be more attenuated than that of R3616 and RH105 in DF3/MUC1-negative cells because of its deficient expression of both TK and γ₁34.5 in these cells. DF3γ34.5 should also replicate as robustly as either R3616 or RH105 in DF3/MUC1-positive cancer cells because of its expression of functional γ₁34.5 gene product in these cells.

In DF3/MUC1-negative cells, we observed that DF3γ34.5 replication was as attenuated as that of MGH1 replication (double mutant) and more attenuated than both R3616 and RH105 (Fig. 3 C). In the DF3/MUC1-positive cells, DF3γ34.5 replication was equivalent to that of the TK-defective RH105 in CAPAN2 cells, but was one-half log order less than RH105 in the other DF3/MUC1-positive cells (MCF-7 and SW1990). Of note, based on its design, DF3γ34.5 replication should not exceed that of RH105; in DF3/MUC1-positive cells DF3γ34.5 replication is expected to at best be equivalent to that of RH105. We also observed that DF3γ34.5 replication was approximately one log order greater than that of R3616 in DF3/MUC1-positive cells and one log order less than that of R3616 in MUC1-negative cells. These results suggest that DF3γ34.5 replication is regulated by restriction of γ₁34.5 function to DF3/MUC1-positive cells as intended, thereby resulting in attenuated replication in DF3/MUC1-negative cells relative to DF3/MUC1-positive cells.

We next examined whether DF3γ34.5-induced cytotoxicity in vitro also followed a pattern that correlated with DF3/MUC1 expression. Cells were infected with each of the HSV-1 mutants at increasing multiplicity of infection values, and cell survival was assessed 6 days later. In the DF3/MUC1-positive cells, DF3γ34.5 was as oncolytic as RH105 (γ₁34.5⁺) and more oncolytic than R3616 (γ₁34.5⁻). In the DF3/MUC1-negative cells, DF3γ34.5 was more attenuated than both RH105 and R3616 and as attenuated as the double mutant MGH1 at low multiplicity of infection values (Fig. 3 D). These data are consistent with DF3γ34.5 replication and oncolysis being dependent on DF3/MUC1 expression.

It was not possible to examine the specificity of the human DF3 promoter in mouse models; accordingly, we devised an assay to examine in human tissues how this promoter regulates DF3γ34.5 replication. Human liver biopsy specimens, each containing a small colon carcinoma metastasis and surrounding normal liver, were harvested fresh from the operating room. As expected, immunohistochemical staining revealed DF3/MUC1 expression in the metastases and not in the normal liver (data not shown). Small blood vessels were cannulated to inject either F strain or DF3γ34.5. The specimen was then cut into pieces measuring 5 mm in thickness and incubated in medium for 72 h, at which time examination of the slices infected with DF3γ34.5 revealed green fluorescence (indicative of viral replication and marker transgene expression) preferentially in the metastases rather than the normal liver (Fig. 4,A). Moreover, although the titers of F strain recovered from normal liver tissue were similar to titers recovered from liver metastases, titers of DF3γ34.5 were two log orders lower in the normal liver than in the metastases (Fig. 4 B). These data demonstrate preferential replication of DF3γ34.5 in liver metastases compared with normal liver after intravascular perfusion of human liver specimens.

We examined antineoplastic efficacy in vivo against DF3/MUC1-positive CAPAN2 flank tumors and DF3/MUC1-negative A375 flank tumors. Tumors were directly inoculated with virus every third day for a total of four inoculations, and tumor sizes were measured every 5 days. In both tumor models, DF3γ34.5 demonstrated greater antineoplastic effects than that of mock-infected medium (control mice; Fig. 5). In addition, the antineoplastic effect of DF3γ34.5 was greater than that of the control RH105 virus in DF3/MUC1-positive CAPAN2 tumors and less than that of RH105 in DF3/MUC1-negative A375 tumors. As expected in the A375 tumors, DF3γ34.5 was as attenuated as the double mutant MGH1. These data are consistent with our in vitro observations and provide further evidence that DF3γ34.5 replication and consequent antineoplastic activity are regulated by the DF3/MUC1 promoter.

On the basis of our observation that DF3γ34.5 replication is more attenuated than that of R3616 and RH105 in MUC1-negative cells, we examined whether this correlated with a lower LD₅₀ after i.v. inoculation. Cohorts of mice received injections of F strain, DF3γ34.5, R3616, or RH105. Mice receiving the higher doses commonly developed paralysis followed by death within 5 days (Table 2). As has been demonstrated previously, wild-type F strain was the most virulent, with all mice rapidly dying at the lowest dose examined, whereas the single-site mutants R3616 and RH105 were more attenuated in virulence. Notably, the LD₅₀ of DF3γ34.5 was one-half to one full log order greater than that of mutants R3616 and RH105. This reduction in pathological virulence of DF3γ34.5 observed after i.v. inoculation was associated with a more restricted biodistribution as assessed by PCR analysis of harvested organs (Table 3).

Replication-competent viruses have many advantages over replication-defective viruses for cancer therapy applications. Because progeny virions can infect adjacent cells, it is not necessary to infect all tumor cells initially, and vector distribution increases over time (24). In contrast, the distribution of transgene expression is much more restricted after direct intratumoral inoculation of replication-defective vectors. Another benefit of replication-competent viruses is that the maximum “dose” is greater than the input dose as a result of in vivo amplification. Incorporation of a therapeutic transgene within the genome of a replication-competent virus permits a two-pronged cancer therapy strategy: oncolysis by viral replication together with the effects of therapeutic transgene expression (25, 26). However, careful selection of therapeutic transgenes is necessary to avoid the problem of antagonism between transgene expression and viral replication (25, 27). The combination of viral oncolysis with therapeutic transgene expression may reduce the risk of emergence of tumor cell resistance to therapy.

One of the greatest challenges faced in the field of viral oncolysis is the development of successful strategies to maximize viral replication in tumor cells and minimize replication in normal cells. Several approaches to restrict viral replication to cancer cells have been examined. The simplest strategy is to inoculate the virus directly into the tumor. This approach has several drawbacks, including the inability to treat radiographically or visually occult lesions and the inability to distribute the virus homogeneously throughout the tumor. Another strategy involves modulation of the interaction between virus and cell surface receptors to permit viral entry into tumor cells but not normal cells. The most well-known example of this strategy is modification of the adenovirus fiber to overcome tumor cell down-regulation of the viral entry receptor CAR (28). A third strategy involves exploitation of the natural properties of some viruses to infect and replicate specifically within cancer cells. The natural selectivity of Newcastle disease virus and vesiculostomatitis virus for cancer cells appears to be a result of defects in the IFN signaling pathways that are commonly present in cancer cells but intact in normal cells (29, 30). Another strategy involves removal of genes from a virus that are critical for replication in normal cells but whose absence is functionally complemented in neoplastic cells. The E1b 55-kDa protein is not expressed in cells infected by the adenoviral mutant Onyx-015 (31). In the absence of E1B55kD protein, viral replication is attenuated except in cells in which the p53 pathway is already disrupted.

In this study we restricted viral replication by regulation of the HSV-1 γ₁34.5 gene by a promoter sequence for a tumor-associated antigen. γ₁34.5 plays a critical role in aiding HSV-1 to subvert an important cellular defense after infection, i.e., PKR activation (32, 33). The importance of this defense mechanism against viral infection has been affirmed by the observation that most viruses have incorporated strategies to overcome the shutoff of protein translation that accompanies PKR activation. Adenovirus expresses VAI RNA to inhibit PKR activation (34). Similarly, human immunodeficiency virus produces TAR RNA, which performs a function similar to that of VAI RNA (35). Influenza virus stabilizes a cellular inhibitor of PKR that forms after infection, thereby functionally inhibiting PKR (36, 37), and as another example, the E3L and NS5A proteins that are expressed by HCV are known inhibitors of PKR (38). HSV-1 circumvents the consequences of PKR activation by expression of γ₁34.5 (Fig. 2), which has sequences homologous with the GADD34 protein (39).

DF3/MUC1 overexpression is observed in many human carcinomas, and mRNA overexpression has been observed in breast carcinomas (40). Abe and Kufe (10) identified elements in the DF3/MUC1 5′ region responsible for regulating transcription. The strength and specificity of this promoter has been demonstrated previously by its ability to appropriately restrict transgene expression to DF3/MUC1-positive cells (41). Selectivity of this promoter sequence has also been demonstrated in an adenoviral construct in which E1A expression was regulated by the DF3/MUC1 promoter (11).

We observed that replication of the HSV-1 mutant DF3γ34.5 is attenuated in DF3/MUC1-negative cells relative to DF3/MUC1-positive cells. This attenuation was associated with a more restricted pattern of biodistribution in mice after treatment of flank xenografts and was also associated with a higher LD₅₀ dose in mice. An important limitation in the interpretation of these data is that the human DF3/MUC1 promoter is not expected to function in mice as it does in humans. The hypothesis that the attenuated toxicity of DF3γ34.5 in humans mirrors that observed in mice necessarily requires examination in a clinical trial. Human gene therapy phase trials are exceedingly costly; we therefore developed an assay to examine viral replication in human tissues. Our observation that DF3γ34.5 replication is attenuated in normal human liver relative to colon cancer liver metastases after perfusion of a portion of the organ lends credence to the notion that DF3γ34.5 would behave similarly after intravascular administration into patients’ livers. We used this organ perfusion experimental model to examine HSV-1 replication; however, it is clearly applicable to examination of other viruses and other therapeutic agents. Conceivably, similar models using portions of other human organs can be developed.

DF3γ34.5 itself is not suitable for examination in clinical trials because it is not susceptible to ganciclovir or acyclovir as a result of inactivation of its TK gene. Sensitivity to these therapeutic agents is an important safety feature to limit unwanted viral replication. We selected the TK locus for homologous recombination because of the ease with which recombinants can be selected with ganciclovir and because we are interested mainly in testing principles. Despite the availability of other antiherpetic agents to which these TK-defective viruses should be sensitive, DF3γ34.5 is not suitable for clinical trials without repair of the TK gene.

The strategic decision of which promoter to use is important to the success of this strategy. For this study, we selected a DF3/MUC1 promoter sequence that has previously been demonstrated to effectively regulate adenoviral replication (11). The choice of location in the HSV-1 genome in which to place the heterologous promoter is equally important. Use of a carcinoembryonic antigen (CEA) promoter in the U_L39 locus to regulate ICP4 expression does not result in preferential HSV-1 replication in CEA-positive cells (42). cis interactions in the region of the promoter may affect the specificity of transcriptional regulation. Others have successfully regulated gene expression within the U_L23 (TK) locus (43), and this observation strongly influenced our decision to use this locus. Finally, the strategic decision of which HSV-1 gene to regulate with the heterologous promoter is important; the gene product ideally should be one whose absence is not effectively complemented in normal cells.

Replication-competent viruses offer several advantages over replication-defective viruses for cancer gene therapy applications. The success with which replication-competent viruses can treat cancer will very likely be dependent on the ability to achieve replication preferentially in neoplastic cells rather than normal cells. Our results demonstrate that the DF3/MUC1 promoter regulates γ₁34.5 expression within the context of HSV-1 replication in a manner that effectively attenuates viral replication in DF3/MUC1-negative cells but permits effective destruction of tumors. Because DF3/MUC1 is overexpressed in a broad spectrum of carcinomas, this approach to viral oncolysis may be broadly applicable.