Osteoclasts Direct Bystander Killing of Bone Cancer

One strategy for cancer gene therapy involves delivery of the cytosine deaminase (CD) gene to sites of cancer and treatment with the CD prodrug 5-fluorocytosine (5-FC; refs. 1–8). CD catalyzes the deamination of nontoxic 5-FC to 5-fluorouracil (5-FU), a highly toxic chemotherapeutic agent. The resultant 5-FU is then further metabolized intracellularly to exert its cytotoxic effects, or can travel across the cell membrane and affect adjacent cells. This latter process is called bystander effect. Bystander effects occur when 5-FU released by CD-expressing cells diffuses into and kills adjacent unmodified cancer cells.

New methods for treating bone cancer are desperately needed. Survival following treatment of primary bone cancers is ∼60%, without any significant treatment advances in the last two decades. The prognosis is even worse in patients with cancer metastases to bone. The most recent advance in treating bone metastases is bisphosphonate therapy. Unfortunately, this treatment has exhibited no survival advantage and >50% of treated patients develop progression of bone cancers (9–11). Enzyme/prodrug gene therapy may provide much needed progress in treating bone cancers, but this approach has not yet been evaluated in experimental models of this devastating disease.

Experimental animal models are being developed where cells that home to or are maintained in the cancer microenvironment are used to deliver and provide CD-based cancer gene therapy. Embryonic endothelial progenitor cells have been used to deliver CD-based gene therapy to lung metastases, and neural progenitor cells have been used to deliver CD-based gene therapy to brain tumors (1, 7). Bone tumors stimulate osteoclast formation and promote osteoclastogenesis by recruitment and differentiation of monocytic lineage cells and their hematopoietic precursors (12–14). Osteoclast lineage cells home to and are activated at sites of bone cancer. This intrinsic homing mechanism introduces the possibility that osteoclast lineage cells may provide a basis for bone cancer–targeted gene therapy. To investigate this prospect, we determined if CD-expressing osteoclasts would direct the killing of bone cancer cells.

Transgenic mice (Tg/NCD) expressing the CD gene regulated by the osteoclastic tartrate-resistant acid phosphatase (TRAP) promoter (15) were used to determine if CD-expressing osteoclasts can direct the killing of cancer cells. In vitro experiments treated cocultures containing Tg/NCD osteoclast and tumor cells with 5-FC. In vivo experiments included 5-FC treatment of tumor-bearing Tg/NCD mice and 5-FC treatment of tumor-bearing C3H mice transplanted with Tg/NCD osteoclast precursor cells (Tg/NCD-C3H). Findings show that the CD/5-FC treatment strategy is an effective treatment in experimental models of bone cancer.

Cell culture. 2472 cell line, originally derived from a malignant tissue tumor (sarcoma) in a C3H mouse, was obtained from the American Type Culture Collection (Manassas, VA). 2472-DSR (DsRed2-containing, red-fluorescing sarcoma) was a gift from Dr. Paul Wacnik (Medtronic, Minneapolis, MN; ref. 16). Powdered medium was purchased from Sigma-Aldrich Chemical (St. Louis, MO) and sera were from Hyclone (Logan, UT). 2472 and 2472-DSR cells were maintained in NCTC-135 medium containing 10% horse sera, passaged once, and fed twice a week. MEM, α modification, containing 10% fetal bovine serum was used in all osteoclast generation assays.

In vitro osteoclastogenesis. Bone marrow cells from Tg/NCD, Tg/NCD-C3H, or C3H mice (4.5 × 10⁴ per well) were cultured in 96-well plates with 20 ng/mL receptor activator of nuclear factor κB ligand (RANKL; Amgen Corp, Thousand Oaks, CA) and 30 ng/mL macrophage colony stimulating factor (M-CSF; R&D Systems, Minneapolis, MN). Cultures were incubated at 37 C in 5% CO₂ for 6 days with addition of media containing fresh RANKL and M-CSF on day 3. After 6 days, wells were stained for TRAP (Sigma-Aldrich, St. Louis, MO) and/or the marker gene nerve growth factor receptor (NGFR).

Immunocytochemistry for the marker gene, human NGFR, was done in wells using a standard three-step avidin-biotin complex procedure. Briefly, monoclonal mouse anti-human NGFR (clone 20.4; provided by Dr. Paul Orchard, University of Minnesota, Minneapolis, MN) was diluted 1:500 and incubated at 37°C for 1 hour. Biotinylated goat anti-mouse immunoglobulin G (IgG; Rockland, Gilbertsville, PA) was used as the secondary antibody (1:750 dilution, 30 minutes at 37°C) and horseradish peroxidase (HRP) avidin-biotin complex (HRP-ABC, Vector Labs, Burlingame, CA) was used as the tertiary reagent.

In vitro CD enzyme assay. CD enzyme activity was determined spectrophotometrically as previously described (17). In brief, cell lysates were incubated with 5 mmol/L 5-FC for 2 hours. Aliquots were removed at 0, 1, and 2 hours to measure conversion of 5-FC to 5-FU. Samples were read at 255 and 290 nm. 5-FC and 5-FU levels were calculated based on the extinction coefficients and the differences in UV spectra: 5-FC (mmol/L) = 0.119 × A₂₉₀ − 0.025 × A₂₅₅, and 5-FU (mmol/L) = 0.185 × A₂₅₅ − 0.049 × A₂₉₀.

In vitro osteoclast cytotoxicity assay. Bone marrow cells from Tg/NCD, Tg/NCD-C3H, or C3H mice (five replicate wells per dose) were cultured under osteoclastogenic conditions in 96-well plates with increasing concentrations of 5-FC (0.01-1.0 mmol/L). Cultures were demi-depleted, media were replaced after 72 hours, and cells incubated an additional 3 days (8). TRAP enzyme activity (solution assay) was measured with a colorimetric assay on day 6 (18). Cytotoxicity was determined by calculation of percent alive, as measured against wells containing no drug. The effect of a pulse of 5-FC was determined by culturing bone marrow cells (five replicate wells per dose) under osteoclastogenic conditions in 96-well plates for 4 days. Increasing concentrations of 5-FC (0.01-1.0 mmol/L) were added, cultures were allowed to incubate for 3 hours, and then all media were replaced. Osteoclasts were assessed at 24, 48, and 72 hours after the 5-FC pulse using the TRAP solution assay.

In vitro tumor killing by osteoclasts. To assess the effect of 5-FC on osteoclast-directed killing of tumor cells, five replicate wells containing Tg/NCD or C3H bone marrow cells and 20 ng/mL RANKL plus 30 ng/mL M-CSF were cultured in 96-well plates for 3 days. Tumor cells (2472-DSR, 2 × 10³ per well) were added on day 3 with fresh media plus cytokines and increasing doses of 5-FC (0.01-1.0 mmol/L). Cultures were demi-depleted, media and 5-FC were replaced after 48 hours, and cells incubated an additional 2 days. Fluorescence was quantitated on day 7 using a fluorescent plate reader with 540-nm excitation and 585-nm emission filters. Cytotoxicity was determined by calculating percent fluorescence generated by tumor cells, as measured against wells containing no drug.

Effect of 5-FC on Tg/NCD osteoclasts. Previously described transgenic mice (Tg/NCD) on a C3H background were used (15). These mice have a fusion gene containing a marker gene (NGFR)(N) and the CD gene under regulation of the osteoclast TRAP gene promoter. A minimum of four Tg/NCD mice was used for each treatment group. Two experiments were done to assess the effect of 5-FC on Tg/NCD osteoclasts in vivo. A single dose of 5-FC, 400 mg/kg (dissolved as 12.5 mg/mL in 0.9% sterile NaCl, injection grade), was given and the effect on osteoclasts and bone marrow cells examined at 4 and 7 days after treatment. In the second experiment, four daily doses of 5-FC were given and the effects evaluated on day 5. Femora were fixed in Z-fix (Anatech, Battle Creek, MI) for 4 hours, decalcified in neutral 10% EDTA for 10 days, and processed for histology. Bone marrow cells were harvested from tibias and cultured under osteoclastogenic conditions to assess osteoclast progenitors.

Bone cancer model. A minimum of four Tg/NCD, Tg/NCD-C3H, or control mice were used for each tumor or treatment group. Intraosseous injections of tumor cells were done as previously described (19). Treatment with 5-FC commenced 6 days after tumor cell injection and the animals were sacrificed 4 days later. Femora were snap frozen in liquid nitrogen for mRNA analysis or processed for histology.

Bone marrow transplantation. Tg/NCD mice were used as bone marrow donors for tail vein transplants into C3H mice. Mice were lethally irradiated (900 cGy, Cs) 24 hours before marrow infusion of 1 × 10⁶ to 4 × 10⁶ bone marrow cells, and designated Tg/NCD-C3H. Engraftment was evaluated at 4, 8, and 16 weeks post-transplant measuring bone mRNA for the transgene NCD. Bones were prepared for histology or mRNA analysis. Cells cultured under osteoclastogenic conditions were evaluated for NGFR immunocytochemistry and 5-FC cytotoxicity.

Determination of transgene mRNA. Expression levels of mRNA were determined using two-step quantitative real-time PCR, with SYBR green as a detection method as previously described (15). Three sets of primer pairs were used: (a) NCD (transgene), CAGAACAAGACCTCATAGCC (forward), GACATCCGCCAATAGGAACA (reverse); (b) endogenous TRAP CCAATGCCAAAGAGATCGCC (forward), TCTGTGCAGAGACGTTGCCAAG (reverse); and (c) hypoxanthine phosphoribosyltransferase (HPRT), GTAATGATCAGTCAACGGGGGAC (forward), CCAGCAAGCTTGCAACCTTAACCA (reverse). HPRT was used for comparison and determination of relative ratio between samples. Quantitative real-time PCR analysis was done with the ABI PRISM 7900 Sequence Detection System instrument and software (PE Applied Biosystems, Inc., Foster City, CA). Thermal cycling conditions were set according to the directions of the manufacturer. Quantitated mRNA values were normalized to the amounts of HPRT mRNA and results are given as fold induction.

Histology, histomorphometry, and immunohistochemical analysis. Femora were cut at 5 μm. Routine H&E staining was done on all specimens, and four replicate sections were used for histomorphometric measurements (20). Images of whole femora were acquired and analyzed as previously described (20). Briefly, histomorphometric analysis recorded areas within bone containing marrow cells, reparative granulation tissue, or tumor. TRAP stain was done on serial sections and osteoclasts counted from images acquired at ×40 in the distal femur. Counts were normalized per millimeter of bone area. Area measurements of individual osteoclasts were done from images acquired at ×200. At least 10 osteoclasts were measured per section and the values were averaged. Immunohistochemistry for NGFR was done on deparaffinized and hydrated sections of bone. Staining was a standard three-step avidin-biotin complex procedure. Briefly, monoclonal mouse anti-human NGFR was diluted 1:1,000 and incubated at 4°C overnight. A biotinylated goat anti-mouse IgG [F′(ab)₂; Rockland, Philadelphia, PA] was used as the secondary antibody (1:1,500 dilution, 1 hour) and HRP-ABC was the tertiary reagent.

Statistical analysis. Data are presented as mean ± SD. Statistical significance was determined by Student's t test. P < 0.05 was considered statistically significant.

Detection of NCD protein and enzyme activity in vitro. We first sought to determine if osteoclasts generated from Tg/NCD bone marrow expressed NGFR. Microscopic inspection of osteoclasts formed from Tg/NCD bone marrow showed abundant NGFR protein by immunocytochemistry (Fig. 1A and B). Assay of osteoclast cultures for CD enzyme activity indicated a high level of conversion of 5-FC to 5-FU. Specifically, cytosolic lysates from day 6 osteoclast cultures converted 5-FC to 5-FU at a rate of 31.2 ± 10.3 μmol/h/mg cytosolic protein (n = 4). This value was higher than the conversion from untreated bone marrow lysates, 4.8 ± 5.0 μmol/h/mg. Neither C3H bone marrow nor osteoclasts showed conversion of 5-FC to 5-FU.

Effect of 5-FC on Tg/NCD osteoclasts in vitro. Having confirmed the presence of NCD fusion protein in vitro, we examined the effect of 5-FC on osteoclast formation and on the survival of established osteoclasts. To determine if 5-FC affected osteoclast formation, osteoclast formation assays were done with Tg/NCD bone marrow and culture systems were exposed to increasing doses of 5-FC. As expected, exposure of culture systems from C3H mice to 5-FC doses as high as 1,000 μmol/L had no effect on osteoclast number. In contrast, dramatic reduction in osteoclast number was seen at all doses of 5-FC added to culture systems from Tg/NCD mice. Specifically, the lowest dose (10 μmol/L) reduced osteoclast number by 20%, whereas the highest dose (1,000 μmol/L) reduced osteoclast number by 90% (Fig. 1C). The estimated ED₅₀ was 100 μmol/L.

To evaluate the effect of 5-FC on established osteoclasts, osteoclast formation assays were exposed to 5-FC for 3 hours, 4 days after assay initiation. Osteoclast number was evaluated 24, 48, and 72 hours after the 3-hour pulse exposure to 5-FC. Again, treatment of culture systems from C3H mice with 5-FC had no effect on osteoclast number. Findings showed no effect of 5-FC on osteoclasts from Tg/NCD mice after 24 hours but showed significant reduction in osteoclast number 48 and 72 hours after exposure (Fig. 1D). Interestingly, ED₅₀ estimates were significantly higher (1,000 and 500 μmol/L for 48- and 72-hour exposures, respectively) compared with systems treated at culture initiation.

Effect of 5-FC on cocultures containing Tg/NCD osteoclasts and tumor cells. We next determined if Tg/NCD osteoclasts could direct bystander killing of tumor cells in vitro. Tg/NCD bone marrow cultures containing established (day 3) osteoclasts were inoculated with 2472-DSR cells and cocultures were exposed to increasing doses of 5-FC. As expected, culture systems containing osteoclasts from C3H mice exhibited robust cancer cell growth and included healthy-appearing osteoclasts. In contrast, culture systems treated with 5-FC showed microscopic evidence of decreased number of cancer cells and contained osteoclasts with fainter TRAP staining and cell membrane disruptions (data not shown).

When compared with dark-field images from cocultures not exposed to 5-FC, culture systems treated with 5-FC had dramatically fewer DsRed2-expressing cells (Fig. 2A and B). Quantitative analysis revealed a 5-FC dose-dependent elimination of tumor cells (Fig. 2C). There was a 40% reduction in tumor cell number in the presence of 25 μmol/L 5-FC and an 80% reduction in tumor cell number when treated with 250 μmol/L 5-FC. The estimated ED₅₀ in this system was ∼50 μmol/L. These findings show osteoclast-directed bystander killing in vitro. 5-FC treatment of cocultures derived from wild-type bone marrow at doses up to 1,000 μmol/L had no effect on tumor cell number. 2472 cells treated with 5-FU, the active metabolite of 5-FC, were sensitive to 5-FU with an ED₅₀ of 2 μmol/L (6).

Effect of 5-FC on osteoclast precursor cells in vivo. We next determined the effect of 5-FC on osteoclast precursors in vivo. A single dose of 5-FC had a profound effect on Tg/NCD osteoclast precursor cells. Bone marrow cells collected 4 and 7 days after 5-FC treatment were grown under osteoclastogenic conditions, assayed for TRAP activity, and compared with cultures from untreated animals. Microscopic examination revealed fewer and smaller TRAP-positive cells in osteoclastogenic cultures from Tg/NCD mice 4 and 7 days after 5-FC treatment (Fig. 3). Culture systems from 4- and 7-day marrow exhibited a profound reduction in TRAP-positive cells and contained only occasional multinucleated cells. NGFR-expressing cells were present, but TRAP enzyme activity was dramatically reduced to 11% and 19% of control (untreated) enzyme levels for assays from marrow collected in mice 4 or 7 days after 5-FC treatment, respectively. These data show a profound killing effect of 5-FC in vivo on Tg/NCD osteoclast progenitor cells.

Effect of 5-FC on Tg/NCD osteoclasts in vivo. Having shown profound effects of 5-FC treatment on osteoclast precursor cells, we next determined the effect of 5-FC on Tg/NCD osteoclasts in vivo. Two treatment schemes were tested. In the first, mice received a single 400 mg/kg 5-FC treatment. Femora from 5-FC-treated Tg/NCD mice were analyzed 4 and 7 days after treatment and findings were compared with untreated mice. Findings showed no change in osteoclast number 4 or 7 days after 5-FC treatment when compared with pretreatment (untreated) values. Mean osteoclast number from animals 4 and 7 days after 5-FC treatment was 21.0 ± 5.4/mm² and 26.6 ± 3.6/mm², respectively (n = 6), compared with 23.5 ± 6.6/mm² (n = 6) for osteoclast number in untreated mice. Analysis of osteoclast size revealed similar findings. Mean osteoclast area was unchanged 4 or 7 days after treatment when compared with pretreatment (untreated) values. Mean osteoclast area from animals 4 and 7 days after 5-FC treatment was 156 ± 21 and 188 ± 51 μm², respectively, compared with 154 ± 35 μm² for osteoclasts in untreated mice.

In the second treatment scheme, four mice received four daily doses of 5-FC and were evaluated 24 hours after the final dose (day 5). Findings showed no change in osteoclast number at day 5 compared with untreated Tg/NCD mice (22.0 ± 2.6/mm² versus 23.5 ± 6.6/mm²) but did reveal a significant increase in mean osteoclast area among 5-FC-treated mice compared with osteoclast area from untreated animals (219 ± 34 versus 154 ± 35 μm²; P < 0.05). Taken in total, analysis of the effect of 5-FC on osteoclasts in Tg/NCD animals showed no effect on osteoclast number and a significant increase (40%) in osteoclast size after four daily doses.

TRAP and NCD gene expression at sites of bone cancer. As NCD gene expression by osteoclast lineage cells is based on expression of the TRAP promoter, we next sought to examine the effect of 2472 tumors on TRAP and NCD gene expression in Tg/NCD mice. Control or tumor-injected femora were evaluated for levels of endogenous TRAP mRNA and NCD mRNA. As expected for this osteolytic tumor, tumors induced a 4-fold increase in TRAP mRNA compared with control femora (relative ratio, 11.7 ± 4.1 versus 2.9 ± 0.5; n = 4). Tumors also induced an increase in NCD mRNA when tumor-bearing and control femora were compared (relative ratio, 0.67 ± 0.09 versus 0.41 ± 0.13). These findings showed significantly increased (P < 0.05) NCD gene expression in tumor-bearing femora and provided rationale to pursue the hypothesis that 5-FC treatment will influence bone tumors in Tg/NCD mice.

Effect of 5-FC treatment on bone tumors in Tg/NCD mice. Mice with femoral 2472 tumors were treated with 5-FC and profound killing effects were observed. Histologic analysis (Fig. 4) recorded areas within bone containing marrow cells, reparative granulation tissue, or tumor. Femora from untreated mice (n = 6) were composed predominantly of tumor (91 ± 9%) with limited marrow (9 ± 11%) and no granulation tissue. In contrast, femora from tumor-bearing Tg/NCD mice treated with 5-FC (n = 10) were predominantly composed of granulation tissue (54 ± 32%) or marrow (38 ± 32%) and significantly less tumor (7 ± 18%). Differences in tumor area and area of granulation tissue were significantly different between mice that did or did not receive 5-FC treatment (P < 0.05). Based on previous work, this significant increase in granulation tissue represents killing of tumor that was present before 5-FC treatment (21). In addition, it is important to note that 8 of 10 mice treated with 5-FC had no evidence of tumor.

Examination of osteoclasts at sites of tumor revealed expression of the NCD protein and a modest reduction in osteoclast number. Immunohistochemistry detected NCD protein in osteoclasts at sites of tumor, with the fusion protein detected along that portion of the polarized osteoclast cell membrane opposite the bone surface (Fig. 5). Osteoclasts were abundant at sites of tumor (35.4 ± 7.6 osteoclasts/mm²) in untreated mice, but were reduced at sites of tumor in 5-FC-treated animals (24.0 ± 10.1 osteoclasts/mm²; P < 0.05). Based on data presented earlier, where after four doses of 5-FC, bones without tumor had 22.0 ± 2.6 osteoclasts/ mm², it is most likely that the reduction in osteoclast number in tumor-bearing, 5-FC-treated mice represents the influence of decreased tumor burden and not an effect of 5-FC.

NCD expression following transplantation of Tg/NCD bone marrow. Having shown that 5-FC treatment of Tg/NCD mice kills bone cancer, we next sought to determine if transplantation of C3H mice with Tg/NCD bone marrow provided animals expressing the NCD transgene in bone. Bone marrow transplantation was done and bone expression of NCD mRNA was analyzed in Tg/NCD-C3H mice 4, 8, and 16 weeks following transplantation. NCD gene expression in bones from Tg/NCD-C3H mice increased at each of the three time points, but did not achieve donor (Tg/NCD) levels, with NCD mRNA levels in Tg/NCD-C3H bones varying between 30% and 60% of levels measured in Tg/NCD bones (data not shown).

Immunocytochemical evaluation of osteoclastogenic cultures from Tg/NCD-C3H bone marrow revealed detection of the NCD fusion protein (Fig. 6A) and treatment of osteoclastogenic cultures with 5-FC resulted in substantial reduction in osteoclasts (Fig. 6B). To this end, 5-FC treatment of C3H bone marrow had no effect on osteoclast formation, but similar treatment of osteoclastogenic cultures from Tg/NCD and Tg/NCD-C3H bone marrow had profound effects on osteoclast formation. Findings from each of these culture systems were significantly different from C3H controls but were indistinguishable from each other. Both culture systems from Tg/NCD mice and Tg/NCD-C3H mice had 25% fewer osteoclasts compared with controls when exposed to 10 μmol/L 5-FC, and both culture systems experienced maximal reduction (80%) when exposed to 1,000 μmol/L 5-FC.

Taken in total, these findings show significant NCD gene expression and CD enzyme activity following transplantation of Tg/NCD bone marrow into C3H mice, and provide rationale to test the hypothesis that 5-FC treatment will influence bone tumors in Tg/NCD-C3H mice.

Effect of 5-FC treatment on bone tumors in Tg/NCD-C3H mice. 5-FC treatment had a significant effect on bone tumors in Tg/NCD-C3H mice. Histologic analysis recorded areas containing marrow cells, reparative granulation tissue, or tumor. As described above, femora from untreated mice were composed predominantly of tumor (91 ± 9%) with limited marrow (9 ± 11%) and no granulation tissue. When compared with femora from untreated mice, tumor-bearing femora from 5-FC-treated Tg/NCD-C3H mice (n = 6) had a 45% reduction in tumor area (91 ± 9 versus 51 ± 29; P < 0.05) and a significant increase in reparative granulation tissue (0 ± 0 versus 23 ± 20%; P < 0.05). The reduction in tumor area and the increase in granulation tissue represent killing of bone tumors by 5-FC treatment. When compared with data from 5-FC-treated Tg/NCD mice, 5-FC-treated Tg/NCD-C3H had less tumor killing. Five of six 5-FC-treated Tg/NCD-C3H mice had reduced tumor compared with untreated controls.

Examination of osteoclasts at sites of tumor revealed a modest reduction in osteoclast number in 5-FC-treated Tg/NCD-C3H animals compared with untreated tumor controls (19.8 ± 8.9 versus 35.4 ± 7.6, respectively; P < 0.05). There was no difference in osteoclast number at sites of tumor between 5-FC-treated Tg/NCD-C3H mice and 5-FC-treated Tg/NCD mice. As stated earlier, it is likely that this reduction in osteoclast number represents decreased tumor burden and not an effect of 5-FC.

In this study, we explored the possibility that osteoclast lineage cells can direct CD-mediated bystander killing of bone cancer. Experiments were done in vitro and in vivo using Tg/NCD transgenic cells or animals transplanted with Tg/NCD bone marrow. Findings in vitro and in vivo showed that CD-expressing osteoclast lineage cells can direct killing of bone cancer cells.

Our in vitro studies showed that treatment of Tg/NCD osteoclasts with 5-FC killed osteoclasts and showed that 5-FC treatment of cocultures containing Tg/NCD osteoclasts and tumor cells killed tumor cells via bystander effects. These findings are similar to a recent report that osteoclast precursor cell lines and bone marrow cells transduced with retrovirus containing the CD gene kill tumor cells in vitro (8). In that report, following RANKL-stimulated osteoclast formation, 5-FC treatment of cocultures containing CD-expressing primary bone marrow or RAW 267.4 osteoclasts and tumor cells eliminated tumor cells via bystander killing. It is noteworthy that, in this study, 5-FC sensitivity of osteoclast lineage cells was related to culture duration. That is, as cells matured in culture, they were less sensitive to 5-FC. In this study, treatment of osteoclast formation cultures at the time of culture initiation resulted in a reduced number of osteoclasts, and 5-FC treatment of culture systems 4 days after culture initiation resulted in a time- and dose-dependent elimination of osteoclasts. Interestingly, the ED₅₀ for 5-FC treatment after 4 days in culture was 500 μmol/L and represented a 5-fold increase compared with the ED₅₀ for osteoclast formation cultures treated at culture initiation. This also suggests that 5-FC sensitivity of osteoclast lineage cells is reduced as these cells mature.

Our in vitro findings from 5-FC-treated osteoclast formation cultures suggest that the sensitivity of Tg/NCD osteoclast lineage cells diminishes with osteoclast precursor cell maturation and multinucleation. As the dominant mechanism mediating the influence of 5-FU is disruption of DNA synthesis, it follows that 5-FC/CD is expected to be most effective at those points in osteoclast formation, which are most dependent on cell division (22). Osteoclast formation has been described as having proliferation and formation phases (23). The proliferative phase occurs during the first 3 to 4 days, and then the formation (fusion) phase begins. Our findings described here and in other systems suggest decreased 5-FU sensitivity during the formation phase (8).

In vitro bystander killing directed by CD-expressing nonmalignant cells found within the tumor microenvironment has been reported in other tumor coculture systems. Cultures containing osteosarcoma cells and CD-expressing embryonic endothelial progenitor cells (eEPC-CD) and cultures containing CD-expressing neuroprogenitor cells and glioma cells exhibited significant tumor cell killing on exposure to 5-FC (1, 7).

Our data show that 5-FC treatment had a profound effect on osteoclast precursor cells in vivo, but had no effect on osteoclast number. 5-FC treatment of Tg/NCD mice caused a profound reduction in the capacity of bone marrow cells collected 4 and 7 days after 5-FC treatment to form osteoclasts in vitro, but did not have a significant effect on osteoclast number. This apparent paradox of effects on osteoclast precursors but not on mature osteoclasts most likely reflects the fact that osteoclast precursor cells are engaged in cell division and thus more sensitive to 5-FU, whereas the mature osteoclasts are quiescent and less sensitive to 5-FU (24).

To date, essentially all efforts to enhance CD-directed bystander killing have focused on augmenting gene expression by addition of downstream genes such as uracil phosphoribosyl-transferase (25), tandem prodrug systems like herpes simplex virus (HSV) thymidine kinase/ganciclovir (26), or translocation proteins (HSV-1 tegument protein vp22) to transport 5-FU to adjacent cells (27). Our findings support a new approach—identification of 5-FU-resistant CD-expressing delivery cells. The observation that mature osteoclasts are relatively insensitive to CD/5-FC treatment in vivo is very important; it identifies osteoclasts as a promising CD gene delivery system.

One major limitation of CD-based bystander killing–dependent gene therapy strategies, in general, has been the death of prodrug-converting cells following exposure to 5-FC (1, 6, 7). This prodrug-induced death of the delivery cell has been called the “Good Samaritan” effect (28). This circumstance creates the obvious disadvantage of reducing both the 5-FU levels and duration of 5-FU exposure to cancer cells. It has been reasoned that quiescence in CD-expressing cells decreases the sensitivity of cell-based CD delivery systems to the cytotoxic effects of 5-FU and prolongs the production and diffusion of 5-FU into surrounding tumor cells (1). To this end, cells such as differentiated neural cells and endothelial cells have effectively been used to enhance CD-directed bystander killing (1, 7). Our data indicate that mature osteoclasts exhibit criteria desired in cell-based, bystander killing gene therapies. That is, mature osteoclasts are not proliferating, are relatively insensitive to CD/5-FC, and survive 5-FC treatment in vivo. A final observation defining osteoclasts as a desirable bone-targeting gene delivery system is the finding that osteoclast size increased following multiple 5-FC treatments. As osteoclast size has previously been correlated with osteoclast activation (29), this finding suggests that TRAP promoter–based therapeutic gene expression may likewise be enhanced.

5-FC treatment had significant tumor-killing effects in Tg/NCD mice and in mice transplanted with Tg/NCD bone marrow. Tumor killing was more dramatic in Tg/NCD mice and is likely explained by the observed reduction in CD gene expression in Tg/NCD-C3H compared with Tg/NCD mice. In vivo CD-directed bystander killing has been reported in experimental models of lung metastases and brain tumors (7). In that report, mice with established lung metastases received tail vein injection of eEPC-CD cells. The eEPC-CD cells preferentially homed to sites of hypoxic lung tumors and 5-FC treatment resulted in improved survival. Fifty percent of animals not receiving 5-FC treatment survived 25 days compared with 50% of animals surviving 35 days with 5-FC treatment (7). Coinjection of rat brains with C6 glioma cancer cells and CD-expressing ST14A neuroprogenitor cells resulted in brain tumors that responded to 5-FC treatment. Animals treated with 5-FC had a 50% reduction in tumor area (1).

We observed fewer osteoclasts at sites of tumor in 5-FC-treated mice compared with tumor-bearing animals not exposed to the prodrug. Interestingly, the number of osteoclasts in Tg/NCD or Tg/NCD-C3H mice with bone cancer was indistinguishable from non-tumor-injected controls and was significantly less than the number in untreated bone cancers. One explanation for this finding is the simple fact that reduced tumor size results in reduced tumor-induced osteoclastogenesis. This observation has previously been reported with the study of CD-transduced bone tumors (6). A second explanation is the possibility that proliferating (5-FU-sensitive) osteoclast precursor cells are the cells that tumors induce to form osteoclasts. Such cells would be eliminated by 5-FU production. In contrast, preexisting mature osteoclasts are less sensitive to 5-FU and therefore survive exposure.

The challenges facing translation of these findings to human therapy include delivery of the CD gene to sites of bone cancer and subsequent CD gene transcription at those sites. Options to consider in the future for gene delivery include systemic infusion or intratumoral injection of molecularly engineered CD-expressing osteoclast precursor cells. Whereas cell-based CD gene delivery has not been reported in humans, direct injection of constructs containing the CD gene has been reported in human clinical trials for prostate and breast cancer, and therapeutic gene expression has been detected (30, 31). Options to consider for optimizing and localizing CD gene expression at sites of tumor include development of gene promoters that permit spatial and/or temporal control. Two interesting possibilities include hypoxia- and radiation-activated promoters. Hypoxia-activated promoters have been described and would be preferentially expressed at sites of tumor hypoxia (32, 33). Radiation-activated promoters could also be considered as bone cancer is commonly treated with radiation, and 5-FU can function as a radiation enhancer (34–36). In fact, recent work describes the CD/5-FC enzyme/prodrug system as capable of enhancing the killing effect of radiation on breast cancer tumors in bone (24, 36).

In conclusion, we report that osteoclasts exhibit features critical for design of cell-based CD/5-FC cancer gene therapy. Following CD gene insertion, these cells produce CD, convert 5-FC to 5-FU, direct bystander killing of cancer cells, and are relatively insensitive to 5-FU themselves. We also provide the initial in vivo experimentation proving the concept that osteoclast-based CD/5-FC treatment strategy reduces the size of bone cancer tumors.

Grant support: National Cancer Institute grant CA90434, the National Institute of Arthritis, Musculoskeletal and Skin Diseases grant AR47302, and the Roby C. Thompson Endowment.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.