Cytotoxicity and Accumulation of Ganciclovir Triphosphate in Bystander Cells Cocultured with Herpes Simplex Virus Type 1 Thymidine Kinase-expressing Human Glioblastoma Cells¹

Enzyme-prodrug strategies have emerged as a mechanism to increase selectivity in cancer chemotherapy (1). Transfer of the HSV-TK³ gene to tumor cells followed by GCV treatment has been widely used in vitro, where it has been shown to be active in many different tumor types (2, 3, 4, 5, 6, 7). This strategy has resulted in impressive tumor regressions in several animal models, prompting clinical trials of HSV-TK/GCV therapy in patients with end-stage cancer including brain tumors (8). This impressive antitumor activity appears to be mediated through the activation of GCV to its triphosphate derivative. HSV-TK phosphorylates GCV to its monophosphate form (GCVMP), which is further phosphorylated by cellular kinases to GCV diphosphate and the presumed cytotoxic metabolite GCVTP (9, 10, 11, 12, 13). Because GCV is a better substrate for HSV-TK than mammalian nucleoside kinases (14), GCV cytotoxicity is selective for cells that express HSV-TK (2, 15). GCVTP mimics the endogenous DNA precursor dGTP and competes for mammalian DNA polymerases, resulting in inhibition of DNA synthesis and the incorporation of GCVMP into the nascent strand (9, 13, 16, 17, 18). Previous work in this laboratory showed that GCVTP elicits a unique, multi-log cell killing distinct from the 1–2 log cytotoxicity induced by most clinically effective nucleoside analogues (19). This multi-log cell killing occurs through a delayed mechanism that allows GCV-treated cells to complete one cell division and may be a consequence of GCVMP incorporation into the DNA template.

An important component of enzyme-prodrug antitumor therapy is the ability to kill cells that do not express the transgene, a process termed the bystander effect (20). This is critical for the clinical success of this approach, because currently available modes of gene transfer typically result in fewer than 10% of the cells expressing the gene of interest (8). For HSV-TK/GCV, this bystander effect has been effective both in vitro and in vivo. When only 1% of the cells in culture expressed HSV-TK, significant killing of bystander cells has been demonstrated (21, 22, 23). Expression of HSV-TK in as few as 10% of the tumor cells in animal models has resulted in complete tumor regression after GCV treatment (5, 24, 25, 26). In vivo, the immune system may play a role in this bystander cell killing (27, 28, 29, 30). However, induction of an immune response cannot account for bystander cytotoxicity in vitro, and another mechanism must account for efficient killing of bystander cells in culture and may also contribute to the bystander effect in vivo.

Whereas it is generally recognized that the killing of bystander cells is crucial to the success of enzyme-prodrug therapy, the actual mechanism by which this occurs has not been clearly identified. Several laboratories have demonstrated that the proximity of bystander cells to HSV-TK-expressing cells is important (31, 32, 33, 34). In addition, studies indicate that bystander cell killing appears to correlate with GJIC, a process that allows small molecules to diffuse between adjacent cells (22, 33, 35, 36, 37, 38, 39). These studies suggest that bystander cytotoxicity is mediated by GCV nucleotide transfer; however, there is little direct evidence documenting that such transfer occurs. Bi et al. (31) reported that tritium originally derived from GCV appeared in bystander cells cocultured with HSV-TK-expressing cells (31), although the chemical nature of the radioactivity in bystander cells was not identified. Ishii-Morita et al. (40) recently demonstrated the transfer of GCV nucleotides from human HSV-TK-expressing cells to murine bystander cells, although only a single incubation condition was evaluated, with no information on the kinetics or mechanism of this transfer. To optimize HSV-TK/GCV therapy, more detailed studies of the mechanism of bystander cell killing are needed using bystander and HSV-TK cells developed from the same parental cell line. We recently documented the accumulation of GCVTP in bystander cells cocultured with HSV-TK-expressing cells, in which both cell lines were derived from the same human colon carcinoma cell line (34). We have now expanded these studies to evaluate GCV cytotoxicity and nucleotide transfer in HSV-TK-expressing and bystander cells generated from a human glioblastoma cell line. In the results presented here, we have performed a detailed characterization of GCV nucleotide transfer with respect to the length of incubation, the concentration of GCV, and the number of HSV-TK-expressing cells. Furthermore, we have compared the bystander cytotoxicity and nucleotide transfer of GCV with those of another HSV-TK substrate, araT. These studies were facilitated by the development of a new technique reported here for separating bystander cells from HSV-TK-expressing cells based on the presence of GFP in the bystander population. The results provide important insight into the reason for the excellent cytotoxic activity of GCV in cultures composed of HSV-TK-expressing cells and bystander cells.

[8-³H]GCV (12.4 Ci/mmol) and [methyl-³H(N)]thymine-1-β-d-arabinofuranoside (2.9 Ci/mmol) were obtained from Moravek Biochemicals, Inc. (Brea, CA). GCV (Cytovene) was a generous gift from Syntex (Palo Alto, CA). Propidium iodide, araT, and other nucleoside and nucleotide standards were purchased from Sigma Chemical Co. (St. Louis, MO). HPLC grade ammonium acetate and ammonium phosphate were purchased from J. T. Baker, Inc. (Phillipsburg, NJ). Ammonium phosphate was also obtained from Fisher Scientific (Pittsburgh, PA). All other chemicals were of the highest purity available.

U251 human glioblastoma cells were maintained in exponential growth in RPMI 1640 supplemented with 10% calf serum and 2 mm glutamine (Life Technologies, Grand Island, NY). To generate U251 clonal cell sublines that stably expressed HSV-TK (U251tk) or β-galactosidase (U251βgal) cDNA, cells were transduced with retrovirus vectors containing the corresponding cDNA. These vectors use the retrovirus long terminal repeat sequence as a promoter and also contain the cDNA for neomycin resistance (19, 34). Cells expressing GFP were generated by transfection with the pEGFP-N1 plasmid (Clontech Laboratories, Palo Alto, CA) and LipofectAMINE (Life Technologies). Cells expressing each transgene were selected with 400 μg/ml G418, and individual colonies were expanded and subsequently maintained in medium containing 200 μg/ml G418. Expression of HSV-TK was confirmed by assaying cell lysates for GCV phosphorylation (7), and β-galactosidase expression was verified by staining with the substrate X-gal. GFP fluorescence was visualized using a fluorescence microscope with excitation at 515 nm and emission at 525 nm.

Cytotoxicity was assessed in the U251βgal and U251tk cell lines by a standard colony formation assay as described previously (7). Exponentially growing cells were treated with the drug for up to 24 h, followed by plating 10–50 viable cells/well. Cells were allowed to grow for 7–11 days, at which time they were stained with crystal violet, and colonies of at least 50 cells were enumerated. In coculture experiments, U251βgal colonies were visualized by staining with X-gal and enumerated. Subsequently, the plates were stained with crystal violet to determine the total number of colonies formed, and the difference between the total number and the number of X-gal-stained colonies was used to calculate the survival of U251tk cells.

After incubation with GCV or araT, cells were harvested, and nucleotides were extracted with perchloric acid and neutralized as described previously (41). The phosphorylated derivatives of GCV and araT were separated from endogenous nucleotides and quantitated by strong anion exchange HPLC using a Waters (Milford, MA) gradient system composed of two Model 501 pumps, a U6K injector module, and a Model 996 photodiode array detector and controlled by Millenium 2010 software. Before injection, each sample was spun at 12,000 × g for 5 min to remove particulate matter, and the pH was adjusted to match the starting elution buffer pH. Samples were loaded onto a 5-μm Partisphere 4.6 × 250-mm strong anion exchange column (Whatman, Hillsboro, OR), and nucleoside triphosphates were eluted with a linear gradient of ammonium phosphate buffer ranging from 0.15 m (pH 2.8 or pH 3.6) to 0.6 m (pH 3.7 or pH 2.8). Nucleotide analogues were resolved from the endogenous nucleotides by these procedures. Nucleotides were identified based on their UV spectrum over the range of 200–355 nm and coelution with authentic standards. Cellular nucleotides were quantitated by comparing their peak areas with that of a known amount of the appropriate standard at wavelengths of 254 and 281 nm. The amount of GCVTP in U251tk or U251gfp cells separated after coculture was corrected for the low level (<3%) of contamination with the other cell type as determined by flow cytometric reanalysis.

After coculture, U251gfp and U251tk cells were harvested by trypsinization and washed once with PBS. Cells were resuspended in complete medium and sorted within 1 h of harvest. Cells were diluted to a concentration of 3 × 10⁶ cells/ml and separated using a Coulter Elite ESP cell sorter (Beckman-Coulter, Miami, FL). GFP-expressing cells were detected after excitation with a 488-nm argon laser using a 525/530 nm bandpass filter. GFP-expressing and non-GFP-expressing cells were analyzed separately and used to establish original gating conditions. These gating limits were used as part of the sort logic. To enhance the sorting purity, special circuits available in the Elite ESP cell sorter (the pulse pile up and the time of flight modules) were used. Use of these modules reduced sort contamination dramatically by improved coincidence elimination (pulse pile up) and aggregate removal (time of flight). For the determination of cell cycle position, U251gfp cells were stained with Hoechst 33342 after harvesting and analyzed by flow cytometry. For experiments using Hoechst 33342 in combination with GFP, dual argon lasers were used. Hoechst 33342 was excited with UV excitation (355–362 nm) from one argon laser, followed by 488 nm excitation of GFP with the second argon laser. These lasers were separated spatially by 40 μs, and the resulting signals were synchronized using the gated amplifier delay circuit on the ESP cell sorter. Hoechst 33342 fluorescence was detected through a 440 nm bandpass filter, correlated directly with the GFP expression level. Data were acquired to 10,000 intact cells, determined by light scatter gating, using the Coulter Elite acquisition software. Algorithms within the MultiCycle computer program (Phoenix Flow Systems, San Diego, CA) were used to determine the number of cells in each phase of the cell cycle based on the Hoechst 33342-generated DNA histogram data.

To determine whether the presence of HSV-TK-expressing cells incubated with GCV could induce cytotoxicity in non-HSV-TK-expressing bystander cells, U251tk cells were cocultured with either equal numbers of U251βgal cells (a 50:50 ratio) or 90% U251βgal cells (a 10:90 ratio) and incubated with GCV or araT for 24 h. As illustrated in Fig. 1, U251tk cells cocultured with U251βgal cells at a 50:50 ratio were exquisitely sensitive to GCV, with a >99.8% decrease in cell survival with 1 μm GCV. The sensitivity of U251tk cells under this coculture condition is similar to that observed previously in a culture of 100% U251tk cells (19). However, when cocultured with 90% bystander cells (a 10:90 ratio), 1 μm GCV decreased U251tk survival by only 54%, and 10 μm GCV was required to reduce survival by 96%. This suggests that the bystander cells rescued the U251tk cells from GCV toxicity in the 10:90 culture.

U251βgal cells cultured alone were marginally sensitive to GCV at concentrations of up to 100 μm (survival ≥ 82%). However, in a 50:50 coculture with U251tk cells, 3 μm GCV decreased survival by >99.5% in the U251βgal cells. Whereas substantial bystander cell killing occurred in the 10:90 coculture, it was less efficient, requiring 30 μm GCV to effect a 99.5% decrease in U251βgal cell survival. Thus, increasing the proportion of bystander cells to 90% decreased cytotoxicity in both cell types, such that the sensitivity of U251βgal cells was similar to that of U251tk cells in the 10:90 coculture.

Studies with araT demonstrated that this drug did not induce bystander killing, even at high concentrations (Fig. 1). In 50:50 cocultures of U251tk and U251βgal cells incubated with up to 1000 μm araT, no significant bystander cell killing was observed. Furthermore, the presence of an equal number of bystander cells reduced the toxicity of araT to the U251tk cells. In the 50:50 coculture, >60% of the HSV-TK-expressing cells survived incubation with 1000 μm araT, whereas in 100% U251tk cultures, <10% of the cells survived a similar incubation with araT (data not shown).

The excellent killing at low GCV concentrations seen in U251βgal cells only when cocultured with U251tk cells suggested that there was a transfer of toxic GCV metabolites between the HSV-TK-positive and -negative cells. To determine conclusively whether this was true, we initiated studies to measure GCV nucleotide levels in both HSV-TK-expressing and neighboring bystander cells.

To measure GCV nucleotides in HSV-TK-expressing and non-HSV-TK-expressing cells that had been cultured together, it was necessary to develop a method to separate the two cell types with high purity. U251 cells that expressed GFP (U251gfp) were developed as bystander cells from the U251 wild-type cell line. This permitted separation of the autofluorescing U251gfp from the nonfluorescing U251tk cells by cell sorting. As illustrated in Fig. 2, A and B, a monoclonal population of U251gfp cells was readily distinguished from the U251tk cells on the basis of a more than 3-log greater fluorescence intensity at 525 nm. Fig. 2 C displays the analysis of U251tk and U251gfp cells that were harvested after coculture and sorted. A portion of each cell population was collected for further analysis. To ensure high purity of the U251gfp sorted cells, only the highest fluorescing portion of the population was collected. Reanalysis of each sorted fraction indicated that the U251tk population was >97% pure and the U251gfp population was >99% pure (data not shown). A similar sorting and collection of cells was performed using cocultures of U251βgal (nonfluorescing) and U251gfp cells. Reanalysis of the U251βgal cells again indicated a >98% purity. An aliquot of each sorted cell population was plated and stained with X-gal, indicating that >97% of the low GFP-fluorescing population were β-galactosidase-expressing cells, whereas <3% of the high GFP-fluorescing population were X-gal positive. Thus, two independent methods demonstrated that the U251gfp cells can be separated from nonfluorescing cells with >97% purity.

Because only the highest fluorescing U251gfp cells were collected (representing less than half of the total U251gfp cells separated), it was important to ensure that this represented a heterogeneous population and did not select for cells in a specific phase of the cell cycle. U251gfp cells were treated with Hoechst 33342 to stain the DNA before cell cycle analysis by flow cytometry. As illustrated in Fig. 3, representative samples of cells with low, mid-level, or high GFP fluorescence exhibited a similar cell cycle distribution. Thus, the collection of high GFP fluorescing cells did not bias the sample for a specific phase of the cell cycle. These studies demonstrated that U251gfp (bystander) cells were separated from U251tk cells with high purity, which allowed us to assess directly whether phosphorylated GCV could accumulate in U251gfp cells cocultured with U251tk cells.

To determine whether GCVTP accumulated in bystander cells, U251tk and U251gfp cells were cocultured in the presence of GCV. U251tk and U251gfp cells were combined at ratios of either 50:50 or 10:90 and allowed to grow for 2 days before drug addition. At that time, cells were examined under a fluorescence microscope, which indicated that the two cell types were still present at the desired ratio (data not shown). The cells in coculture appeared to be well mixed, and each culture was ≤ 50% confluent at the time of drug addition. After the drug incubation period, the cocultures were trypsinized, and the two cell types were separated and collected by cell sorting based on GFP fluorescence. Nucleotides in separated cell populations were then extracted and analyzed by HPLC. As illustrated in Fig. 4, GCVTP was readily distinguished from the endogenous nucleoside triphosphates after the incubation of U251tk cells with GCV. In U251gfp cells cultured alone with 10 μm [³H]GCV for 24 h, there was no detectable [³H]GCVTP (<6.9 pmol GCVTP/10⁷ cells). When U251tk and U251gfp cells were cocultured at ratios of 50:50 or 10:90 and incubated with 1, 3, or 10 μm GCV for 24 h, GCVTP was readily detected in both cell types (Fig. 5). Lower concentrations of GCV were not studied because radiolabeled drug would have been required for sensitivity, but radioactive samples could not be used with the flow cytometer. In the 50:50 culture, the levels of GCVTP in the U251tk cells were similar to those observed previously in 100% U251tk cultures incubated with similar concentrations of GCV (19). The amount of GCVTP measured in bystander cells increased with increasing concentrations of GCV, and in all cases, the GCVTP in bystander cells was less than that in the U251tk cells. Interestingly, the amount of GCVTP in both cell types was 3–8-fold lower in cultures with only 10% U251tk cells compared to those with 50% U251tk cells. However, the percentage of U251tk cells in the culture did not affect the proportion of GCVTP in bystander cells, which ranged from 30–80% of the amount in U251tk cells in both the 50:50 or 10:90 cocultures. These studies demonstrate that relatively low levels of GCVTP are sufficient to produce cytotoxicity ranging from 45% to >99% in bystander cells.

Fig. 6 demonstrates that the accumulation of GCVTP in the U251gfp cells occurred readily, with detectable levels of phosphorylated GCV apparent within 4 h after drug addition. The amount of GCVTP in both the U251tk and U251gfp bystander cells increased continuously throughout the 24-h incubation period, without evidence of saturation. Furthermore, the proportion of GCVTP in the bystander cells compared to that in U251tk cells was relatively constant throughout the incubation (54–71% of the amount in U251tk cells).

After demonstrating that GCVTP could accumulate in U251gfp cells after coculture with U251tk cells and GCV, we wished to determine whether this was a property unique to GCV, or whether other nucleotides phosphorylated selectively by HSV-TK could be transferred. araT is a HSV-TK substrate with no detectable phosphorylation in cultures of 100% U251gfp cells (data not shown). However, when equal numbers of U251tk and U251gfp cells were cocultured in the presence of 1000 μm araT for 24 h, araTTP was apparent in both cell types after separation (Fig. 7). The proportion of araTTP in bystander cells was <12% of that in U251tk cells, substantially less than the proportion of GCVTP observed in bystander cells compared to U251tk cells. Although the actual amount of araTTP in bystander cells was similar to that of GCVTP in some cases, previous studies have shown that this amount of araTTP is not toxic to U251 cells (7), accounting for the lack of bystander killing in cocultures incubated with araT. Nonetheless, the results demonstrate that both GCVTP and araTTP are present in bystander cells after coculture with HSV-TK-expressing cells and the parent nucleoside, indicating that both GCV and araT nucleotides can be transferred from U251tk to non-HSV-TK-expressing bystander cells.

Despite a wealth of reports on the ability of HSV-TK/GCV to produce toxicity to both HSV-TK-expressing as well as non-HSV-TK-expressing bystander cells in coculture (2, 5, 23, 25, 26, 30, 31, 33, 42, 43, 44), there is little concrete evidence to demonstrate that this cytotoxicity is caused by the transfer of phosphorylated GCV metabolites. One problem in acquiring such data is the lack of methodology to rapidly separate two cell populations that have been cultured together to allow an analysis of nucleotide metabolism. Recently, Ishii-Morita et al. (40) described a procedure to separate human HSV-TK-expressing cells from murine bystander cells by antibody tagging the human cells and removing them with immunomagnetic beads. Although these investigators were able to identify GCV monophosphate, diphosphate, and triphosphate in bystander cells after coculture, the separation technique required several hours and two rounds of purification to achieve a high purity for each cell population. In addition, this method relied upon xenogeneic cell populations for separation. We recently reported another immunomagnetic technique in which cells were engineered to express a single chain antibody permitting their selection with hapten-coated beads (34). This technique required only a single round of selection to generate a >97% pure population of bystander cells after coculture, and this method was more rapid than that of Ishii-Morita et al. (40). However, the purity of the cells that did not express the antibody was generally <80%. Here we report another method that relies on the unique fluorescence of cells expressing GFP, facilitating their separation from nonfluorescing cells by flow cytometry. Compared to our previous technique, the GFP method has the advantage of sorting both the GFP-expressing and non-GFP-expressing populations with high purity, and larger numbers of cells can be processed rapidly. However, the GFP technique is limited by the inability to use the radiolabeled drug for cells undergoing flow cytometric analysis and cell sorting, which lowered the sensitivity of the system somewhat and precluded measurement of the low amount of GCVMP in DNA. Both the immunomagnetic and GFP methods are rapid, producing separated cells within 1 h of harvest for analysis. Thus, these techniques should be useful to investigators in a variety of areas who wish to isolate individual populations of cells after coculture.

The data presented here demonstrated that bystander killing of human glioblastoma cells was dependent on the number of HSV-TK-expressing cells in coculture and the concentration of GCV, which is consistent with previous reports from this and other laboratories (23, 25, 34, 42). In cultures with equal numbers of U251tk and U251βgal cells, at least three times more HSV-TK-expressing cells were killed compared to bystander cells for each GCV concentration. However, when U251tk and U251βgal cells were cultured at a ratio of 10:90, the cytotoxicity was similar. This likely reflects the decreased accumulation of GCVTP in each cell type cultured at the 10:90 ratio compared to the 50:50 ratio, as discussed below. The ability of cells deficient in a metabolic pathway to rescue neighboring cells has been well documented (45). Along those lines, other investigators have noted that the sensitivity of cells to GCV is diminished in the presence of a large bystander population, presumably through the depletion of GCV nucleotides in the HSV-TK-expressing cells (33), which is consistent with our findings. No bystander cell killing was observed with araT, which was previously found to be a weaker cytotoxic agent than GCV in 100% HSV-TK cultures (7, 19). Therefore, for the studies shown here, significant bystander cell killing appears to require a highly cytotoxic drug.

The use of GFP-expressing cells as the bystander population permitted their rapid separation from HSV-TK-expressing cells for the subsequent detection of GCV nucleotides. These studies demonstrated that the accumulation of phosphorylated GCV in bystander cells occurred readily at all concentrations tested, suggesting that the transfer of GCV nucleotides is rapid. The proportion of GCVTP in bystander cells ranged from 30–80% of that in U251tk cells; due to the highly potent nature of GCVTP reported previously (19), these levels were sufficient to induce good killing in bystander cells.

When U251tk and U251gfp cells were cocultured at a ratio of 10:90, markedly less GCVTP accumulated in both cell types compared to that present in the cells cultured at the 50:50 ratio. This suggests that in the 10:90 coculture, the rate of phosphorylation of GCV in the HSV-TK-expressing cells was unable to keep up with the rate of transfer of GCV nucleotides to bystander cells. In support of this hypothesis, we reported previously that HSV-TK-expressing cells accumulated GCVTP linearly for approximately 8–12 h, at which point the amount of this nucleotide reached a plateau (19). In contrast, the HSV-TK cells cocultured with an equal number of bystander cells showed a continuous rise in GCVTP levels over the entire 24-h incubation period (Fig. 6), indicating that an equilibrium between the formation and elimination of GCVTP was not achieved. Although the levels of GCVTP at the end of the 24-h incubation period were similar in U251tk cells cultured alone or with an equal number of U251gfp cells, it required an additional 12–16 h to achieve the same GCVTP levels for the cells in coculture. When U251tk cells accounted for only 10% of the culture, it would be expected that facile transfer of GCVTP from a small number of HSV-TK-expressing cells to a large number of bystander cells would make it more difficult for the U251tk cells to achieve and maintain high levels of GCVTP.

Measurable levels of araTTP were observed in U251tk and bystander cells after coculture with araT. However, the amount of araTTP in the U251tk cells was markedly lower than that observed previously in 100% cultures of U251tk cells incubated with araT. In a previous report, we demonstrated that incubation of U25tk cells with 1000 μm araT resulted in >4000 pmol araTTP/10⁷ cells within 4 h (19). In contrast, <800 pmol araTTP/10⁷ cells accumulated in the U251tk cells cocultured with the U251gfp cells (Fig. 7). We also note that the proportion of araTTP in bystander cells was considerably lower than that observed with GCVTP. These differences may be explained in part by loss of araTTP during the sorting procedure, because araTTP has a much more rapid initial half-life than GCVTP (1.9 versus 11 h, respectively; Ref. 19). In addition, the half-life of araT nucleotides may be shorter in bystander cells than in U251tk cells due to the lack of HSV-TK to rephosphorylate araT derived from degraded araTTP. If the transfer of araT nucleotides to bystander cells is dependent on a concentration gradient, then rapid degradation of araT nucleotides in bystander cells would continually force more phosphorylated araT into the bystander cells and therefore deplete the U251tk cells of phosphorylated araT. Alternatively, the lower proportion of araTTP relative to GCVTP in bystander cells may indicate that GCV nucleotides are transferred more readily, although this explanation cannot account for the lower levels of araTTP in the U251tk cells in coculture. Thus, it seems most likely that a more rapid degradation of araTTP accounts for its lower proportion in bystander cells.

Consistent with the observation of lower araTTP levels, cytotoxicity was low in both the U251tk and U251βgal cells. Previous studies indicated that U251tk cells alone incubated with 1000 μm araT killed 95% of the cells in the culture, more than twice the cell killing observed here when U251tk cells were cocultured with U251βgal cells. Furthermore, the lack of bystander cytotoxicity in cocultures incubated with araT appears to be due to the low levels of araTTP that accumulated in bystander cells, levels that were not associated with cytotoxicity in U251tk cells as reported previously (19).

The results suggest that GCV and araT nucleotides were transferred from the HSV-TK-expressing cells to non-HSV-TK-expressing bystander cells in coculture. Whereas these studies measured the amount of GCV and araT nucleotides that appeared in bystander cells, the actual metabolites transferred and the mechanism by which the transfer occurred were not determined. Two mechanisms have been proposed to account for the transfer of GCV nucleotides from HSV-TK-expressing cells to bystander cells. Freeman et al. (25) have suggested that apoptotic bodies formed in HSV-TK-expressing cells may contain phosphorylated GCV, and that phagocytosis of these particles into neighboring bystander cells would result in the appearance of GCV nucleotides in bystander cells. This does not appear to be an important mechanism in the U251 glioblastoma cells, because GCV nucleotides were observed in bystander cells within 4 h after coculture, days before significant apoptosis was observed in this cell line (19). More data implicating GJIC as the mechanism of GCV nucleotide transfer in vitro have accumulated recently (22, 31, 35). Gap junctions are formed by hexameric connexin proteins connecting adjacent cells (31, 46). These protein pores allow molecules of <1 kDa to diffuse between connected cells across a concentration gradient. The characteristics of the appearance of GCVTP in U251gfp cells are consistent with GJIC transfer because both araT and GCV nucleotides have molecular masses of <1 kDa. In addition, the amount of GCVTP that was observed in bystander cells appeared to be a constant percentage of that in the U251tk cells at different concentrations and lengths of incubation, which is as expected if the extent of transfer was dependent simply on a concentration gradient. Furthermore, the transfer of araT nucleotides was demonstrated, indicating that the mechanism of transfer is not selective for GCV nucleotides. We have evaluated the GJIC capacity of U251 cells by measuring the transfer of Lucifer yellow dye from one microinjected cell to the surrounding cells. These studies demonstrated that approximately 80% of the U251 cells communicate with neighboring cells.⁴ Thus, the data presented are consistent with the transfer of GCV and araT nucleotides through GJIC, although further work is necessary to conclusively identify the mode of transfer.

Given the importance of bystander killing to the success of enzyme-prodrug gene transfer protocols, it is necessary to have a thorough understanding of the mechanism by which the transfer of GCV nucleotides occurs. We have recently noted that in two populations of colon carcinoma cells, a GJIC-deficient cell line showed better bystander killing and higher transfer of GCV nucleotides than a GJIC-competent cell line (34). Whereas it is possible that a low, undetectable level of GJIC was responsible for the nucleotide transfer, these data may suggest a new mechanism for the transfer of cytotoxic nucleotides between cells. In the data presented here, at least 20% of the U251βgal cells appeared incapable of GJIC, yet >99% could be killed when cocultured with U251tk cells and GCV. More data are necessary to determine whether GJIC alone can account for this high level of cytotoxicity to bystander cells, or whether other mechanisms may also play a role. Whereas several groups have shown that the transfer of connexin proteins can improve bystander killing with HSV-TK/GCV (35, 36, 38, 39, 47), our data suggest that other mechanisms are also important. We are now attempting to identify the mechanism by which the GJIC-deficient colon carcinoma bystander cell line accumulated GCVTP and determine whether it is a mechanism used by other tumor cell types such as U251 glioblastoma cells. Greater understanding of the process by which the highly potent GCV nucleotides are transferred to bystander cells may lead to novel mechanisms to enhance bystander killing.