Hypoxia-mediated Apoptosis from Angiogenesis ... - Hypoxia Imaging

[CANCER RESEARCH 59, 4882– 4889, October 1, 1999]

Hypoxia-mediated Apoptosis from Angiogenesis Inhibition Underlies Tumor Control by Recombinant Interleukin 121 Michael S. Gee, Cameron J. Koch, Sydney M. Evans, W. Timothy Jenkins, Charles H. Pletcher, Jr., Jonni S. Moore, Holly K. Koblish, Jungeun Lee, Edith M. Lord, Giorgio Trinchieri, and William M. F. Lee2 Biomedical Graduate Program [M. S. G., H. K. K.], Departments of Radiation Oncology [C. J. K., S. M. E., W. T. J.], Pathology and Laboratory Medicine [C. H. P., J. S. M.], and Medicine [W. M. F. L.], Cancer Center [J. S. M., W. M. F. L.], and Flow Cytometry and Cell Sorting Facility [C. H. P., J. S. M.], University of Pennsylvania, Philadelphia, Pennsylvania 19104; Department of Microbiology and Immunology and Cancer Center Immunology Program, University of Rochester, Rochester, New York 14642 [J. L., E. M. L.]; and The Wistar Institute, Philadelphia, Pennsylvania 19104 [G. T., W. M. F. L.].

ABSTRACT The role of angiogenesis inhibition in the antitumor activity of recombinant murine interleukin 12 (rmIL-12) was studied in K1735 murine melanomas, the growth of which is rapidly and markedly suppressed by rmIL-12 treatment. On the basis of the prediction that tumor ischemia should result from therapeutic angiogenesis inhibition, tumor cell hypoxia was evaluated as a marker of ischemia using the EF5 [2-(2-nitro-1Himidazol-1-yl)-N-(2,2,3,3,3-pentafluoropropyl)acetamide] approach. This method measures intracellular binding of the nitroimidazole EF5, which covalently binds to cellular macromolecules selectively under hypoxic conditions. Whereas 1 week of rmIL-12 treatment effectively inhibited K1735 cell-induced angiogenesis in Matrigel neovascularization assays, 2 weeks of treatment were needed before severe tumor cell hypoxia was detected in K1735 tumors. The hypoxia that developed was regional and localized to tumor areas distant from blood vessels. The great majority of severely hypoxic tumor cells were apoptotic, and in vitro studies indicated that the degree of hypoxia present within treated tumors was sufficient to trigger K1735 apoptosis. Tumor cell apoptosis was also prevalent in the first week of rmIL-12 treatment when few cells were hypoxic. In vitro studies indicated that this non-hypoxia-related apoptosis was induced directly by IFN-g produced in response to rmIL-12 administration. These studies reveal that rmIL-12 controls K1735 tumors initially by IFN-ginduced apoptosis and later by hypoxia-induced apoptosis. They also establish hypoxia as an expected result of tumor angiogenesis inhibition and a mediator of its therapeutic effect.

INTRODUCTION IL-123 has profound antitumor effects in a number of murine tumor models whether it is administered as the recombinant cytokine or secreted by engineered cells (Refs. 1–5; reviewed in Ref. 6). These effects are generally dependent upon T and NK cell production of IFN-g in response to IL-12 stimulation. Because IL-12 and IFN-g together induce Th1 differentiation of CD41 T cells, enhance CD81 T cell maturation and activation, and functionally activate NK cells (7), the antitumor activity of IL-12 has been attributed to the induction of antitumor immunity. Whereas this may be important for long-term tumor control by IL-12, its role in acute control of tumor growth is less certain. Recent studies demonstrated that therapeutically effective regimens of rmIL-12 impaired T cell-mediated immune responses Received 3/31/99; accepted 8/9/99. 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. 1 This work is supported by a NIH Medical Scientist Training Program grant (to M. S. G.), NIH Training Grant T32 AI-07285 (to J. L.), a grant from the Lucille B. Markey Trust (to J. S. M. and C. H. P.), and the following NIH grants: CA-74071 (to C. J. K., S. M. E. and W. T. J.); CA-28332 (to E. M. L.); CA-16520 (to J. S. M. and C. H. P.); AI-34412, CA-32898, and CA-20833 (to G. T.); and CA-77581 (to W. M. F. L.). 2 To whom requests for reprints should be addressed, at University of Pennsylvania, BRB II/III Room 330, 421 Curie Boulevard, Philadelphia, PA 19104. Phone: (215) 8980258; Fax: (215) 573-7912; E-mail: [email protected]. 3 The abbreviations used are: IL-12, interleukin 12; NK, natural killer; rmIL-12, recombinant murine IL-12; IP-10, IFN-inducible protein 10; Cy3, cyanine-3; TUNEL, terminal deoxynucleotidyl transferase-mediated nick end labeling; rIFN-g, recombinant IFN-g.

during and beyond the period of cytokine administration, even as tumor growth was controlled (8, 9). Impairment was attributable to suppression of T-cell mitogenic responses by macrophage-derived nitric oxide produced in response to IFN-g. Interestingly, the immunosuppression seen in rmIL-12-treated mice waned following cessation of therapy and was followed by enhanced antitumor immunity when this was tested at later time points. Tumor control and regression during rmIL-12 administration may be better explained by its ability to inhibit angiogenesis. This activity has been demonstrated in several model assays of neovascularization to involve IFN-g and IFN-g-dependent production of antiangiogenic chemokines such as IP-10 (10 –13). Production of these cytokines in turn leads to inhibition of angiogenesis by down-regulation of several proangiogenic factors, including tumor cell vascular endothelial growth factor production, matrix metalloproteinase activity (14), and expression of integrins involved in endothelial cell adhesion and survival (15). However, whereas rmIL-12 is known to inhibit angiogenesis in model assays, the issues of whether it inhibits angiogenesis within established tumors and what the consequences of tumor angiogenesis inhibition might be remain unaddressed. In this study, we assess the contribution of tumor angiogenesis inhibition to overall IL-12 antitumor efficacy. How angiogenesis inhibition by IL-12 controls tumor growth is important to our general understanding of the effects of antiangiogenesis therapy because the mechanism by which angiogenesis inhibitors cause tumor regression is not currently known. The maintenance of tumor dormancy by angiogenesis inhibitor therapy has been previously shown to be associated with an increase in tumor cell apoptosis (16, 17). Whereas no formal mechanism for this increased apoptosis has been proposed, it has been speculated that angiogenesis inhibition could restrict the supply of endothelial cell-derived paracrine factors required for tumor cell survival (16). Because effective inhibition of tumor angiogenesis by definition should result in tumor ischemia, we reasoned that the signal for tumor cell apoptosis could be a downstream consequence of severe ischemia. A reasonable candidate is hypoxia, which has been shown to be a potent inducer of apoptosis by a mechanism possibly involving HIF1a-mediated stabilization of p53 (18) and inhibition of Bcl-2 (19), and hypoxia-induced tumor cell apoptosis may explain how angiogenesis inhibitors control tumor growth. In this study, we examine K1735 murine melanomas undergoing rmIL-12 therapy to determine whether tumor hypoxia is induced by treatment and whether this plays a role in controlling tumor growth. Our results demonstrate that treatment results in severe tumor cell hypoxia, the appearance of which is delayed relative to the onset of angiogenesis inhibition. The development of hypoxia is selective, occurring only in tumor areas relatively removed from blood vessels, which reinforces its relationship with angiogenesis inhibition. The severe hypoxia that develops with therapy appears to mediate rmIL-12 antitumor activity by inducing widespread tumor cell apoptosis. Our results also reveal that, prior to the appearance of severe tumor hypoxia, rmIL-12 control of K1735 tumor growth is hypoxia inde-

4882

IL-12 ANTIANGIOGENESIS CAUSES HYPOXIA-INDUCED APOPTOSIS

pendent and attributable to direct induction of tumor cell apoptosis by IFN-g. Thus, rmIL-12 is a complex antitumor agent that uses a combination of mechanisms to suppress tumor growth, and angiogenesis inhibitor therapy, in general, may control tumor growth largely through induction of tumor cell hypoxia and hypoxic cell death. MATERIALS AND METHODS Mice and Cell Lines. Female C3H/HeN mice, 6 – 8 weeks old, were purchased from Harlan Sprague Dawley (Indianapolis, IN). All animals were maintained in microisolator cages under sterile conditions. The K1735 murine melanoma cell line (20) was maintained in DMEM supplemented with 10% FCS and penicillin/streptomycin. Stably transfected N23 K1735 cells containing a dominant negative IFN-gR1 were described previously (12). In Vivo Studies. For tumor growth studies, 106 trypan blue-excluding K1735 cells were injected s.c. into the lower left flank. Injected cells were derived from low-passage frozen stocks that had been established in culture ,1 week prior to injection. rmIL-12 was administered i.p. on a five doses per week schedule (five daily injections of 125 ng, followed by 2 days of rest) for the durations indicated. Tumors were measured by calipers at regular intervals, and mice were euthanized according to guidelines established by the Institutional Animal Care and Use Committee. Serum IFN-g levels were assayed by RIA using the polyclonal antibody AN18 as the capture antibody and the monoclonal antibody XMG6 as the detection antibody. For EF5 [2-(2nitro-1H-imidazol-1-yl)-N-(2,2,3,3,3-pentafluoropropyl)acetamide] studies, the mice were given an i.v. injection of 0.25 ml of 10 mM EF5 in 0.9% saline 3 h prior to tumor excision. On the basis of the half-life of EF5 in circulation, this assures that virtually all free EF5 is cleared by the time tumors are excised and that no EF5 binding is due to hypoxia resulting from interruption of circulation associated with euthanasia or tumor excision. Matrigel Assay for Angiogenesis. In vivo Matrigel neovascularization assays were performed essentially as described previously (12). Briefly, C3H/ HeN mice were injected s.c. along the ventral midline with 0.5 ml of Matrigel (Collaborative Biomedical Products, Bedford, MA) mixed with 100 ml of PBS, either alone or with an angiogenic stimulus [200 ng of recombinant basic fibroblast growth factor (Upstate Biotechnology, Lake Placid, NY) and 1 3 106 K1735 or K1735.N23 cells]. In treated mice, rmIL-12 (125 ng) was injected i.p. daily on days 0 – 4. Cytokine ablation was achieved by injecting 0.5 mg of anti-IFN-g monoclonal antibody i.p. on days 21, 1, and 3. Matrigel pellets were harvested on day 6 and reliquefied by incubation at 4°C overnight in 300 ml of PBS. Matrigel neovascularization was quantitatively determined by measuring the hemoglobin content of the liquefied pellets (Drabkin’s method; Sigma Chemical Co., St. Louis, MO). Flow Cytometry. Tumor cell suspensions were prepared and stained for EF5 with the monoclonal antibody ELK 3-51 essentially as described (21). For single-color analysis, ELK 3-51 was conjugated to the fluorochrome Cy3, whereas for two-color analysis, it was conjugated to cyanine-5. Apoptosis was measured by TUNEL staining method according to the manufacturer’s instructions (Boehringer Mannheim, Indianapolis, IN). For dual EF5/TUNEL staining, cells were first TUNEL stained and then washed twice with PBS prior to overnight staining with ELK 3-51. Analysis of tumor cell proliferation was performed using a rapid propidium iodide-based staining protocol as described.4 Briefly, 2 3 106 tumor cells were incubated for 30 min at 37°C in 500 ml of staining solution [3% polyethylene glycol 6000, 50 mg/ml propidium iodide (Sigma), 1 mg/ml RNase A (Boehringer Mannheim), 0.1% Triton X-100, and 4 mM sodium citrate (pH 7.2)]. Five hundred ml of salt solution [3% polyethylene glycol 6000, 50 mg/ml propidium iodide, 0.1% Triton X-100, and 0.4 M NaCl (pH 7.2)] were then added, and the nuclei were stored at 4°C until analysis. Flow Cytometer Type and Laser Specifications. Flow cytometry was performed on a Becton Dickinson Immunocytometry Systems FACS Vantage and FACS Calibur. The FACS Vantage was equipped with a green laser (514 nm) to excite Cy3 fluorescence, whereas the FACS Calibur was equipped with blue (488 nm) and red (635 nm) lasers to excite fluorescein and cyanine-5 fluorescence, respectively. Standard collection optics were used to collect 4 K. D. Bauer, DNA analysis of tissue samples. Presented at the 1996 Annual Course in Flow Cytometry, Brunswick, ME, June 10 –14, 1996.

emitted fluorescence. Photomultiplier tube voltages were adjusted daily by placing standard beads in the same channel. Standardization of EF5 fluorescence intensity was achieved using calibration beads (Flow Cytometry Standards Corporation, Research Triangle Park, NC). Isolation of tumor cells in cell suspensions was achieved by establishing gate parameters around in vitro cultured K1735 cells, which exhibit detectably higher forward and side scattering than erythrocytes and lymphocytes. Flow cytometric analysis of hypoxia and apoptosis was performed using CellQuest (Becton Dickinson, Mountain View, CA), whereas analysis of proliferation was performed using ModFit Version 2.0 (Verity Software, Topsham, ME). In Vitro Studies. For IFN-g stimulation studies, 5 3 105 K1735 cells were plated onto 10-cm dishes and incubated for 24 h prior to addition of rIFN-g (R&D Systems, Minneapolis, MN). IFN-g was replaced every 48 h. Hypoxia studies were performed essentially as described previously (21). Cells (7.5 3 105) were spot-plated onto 5-cm glass dishes in JRH 610 medium supplemented with penicillin/streptomycin and 10% FCS (complete JRH 610) and incubated overnight at 37°C. The following day, the dishes were placed in leak-proof aluminum chambers connected to a manifold in which the oxygen concentration was reduced to the appropriate level. For EF5 hypoxia studies, EF5 was added to the medium up to a final concentration of 0.10 mM immediately before induction of the appropriate level of hypoxia, and the cells were incubated for 3 h. For prolonged hypoxia studies, the medium was replaced with complete JRH610 supplemented with 20 mM HEPES, 0.1% glucose, and 0.008 N NaOH prior to hypoxic induction. pH and glucose measurements were taken at the beginning and end of the hypoxic period to ensure that adequate glucose was present in the medium and document any significant changes in acidity. Immunohistochemistry. Immunohistological staining for platelet/endothelial cell adhesion molecule (PECAM) and EF5 and imaging were performed essentially as described previously (22). Briefly, cryostat sections were incubated with 5 mg of rat antimouse PECAM-1 antibody (MEC 13.3; PharMingen, San Diego, CA), followed by 40 mg/ml peroxidase-conjugated mouse antirat IgG (H&L; Jackson Immunoresearch Laboratories, West Grove, PA). Following a PBS wash, sections were incubated with the chromagen 3-amino9-ethylcarbazole (Vector Laboratories, Burlingame, CA). Sections were then stained with ELK 3-51-Cy3. Normal tissues from tumor-bearing mice were also stained for EF5 to guarantee staining specificity (liver central vein hepatocytes stain positive, whereas s.c. tissue stains negative). Stained sections were imaged using an epifluorescence equipped Nikon microscope (320 objective), digitized, and image-analyzed using Image Pro software (Version 3.0; Media Cybernetics, Silver Spring, MD). Images from adjacent microscope fields were automatically acquired and digitally combined to form a 5 3 4 montage (corresponding to 2.9 3 1.7 mm of the section). Tumors in each treatment group were size-matched for comparison.

RESULTS K1735 Tumor Growth and Tumor Cell-induced Angiogenesis Are Suppressed by IL-12 Therapy in an IFN-g-dependent Manner. K1735 melanoma tumors were established in syngeneic C3H/ HeN mice by s.c. inoculation of 1 3 106 cultured cells and treated with rmIL-12 at the maximum tolerated dose (125 ng/day, 5 days/ week) for 1, 2, or 3 weeks after they became palpable (2–5-mm diameter). Tumor growth was progressive in PBS-treated mice but suppressed upon initiation of rmIL-12 therapy (Fig. 1A). Control of tumor growth by rmIL-12 was ablated by administration of a neutralizing monoclonal antibody to IFN-g (XMG6), indicating that it was dependent on IFN-g (Fig. 1B). Tumor control also required tumor cells to respond to IFN-g, shown by the unresponsiveness of N23 tumors to rmIL-12 treatment (N23 is a transfected K1735 clone overexpressing a dominant negative IFN-gR1 receptor; Ref. 12). Similarly, a 5-day course of rmIL-12 treatment (Fig. 2) inhibited Matrigel neovascularization induced by K1735 cells or a purified angiogenic factor, basic fibroblast growth factor (bFGF). This antiangiogenic effect was inhibited by coadministration of neutralizing antibodies to IFN-g (data not shown) and was not apparent when N23 cells were in the Matrigel (Fig. 2). Thus, rmIL-12 acts in an IFN-g-

4883


dependent manner to retard growth of established K1735 tumors and inhibit angiogenesis by these cells. Hypoxic Tumor Cells Arise in Avascular Tumor Regions during rmIL-12 Therapy. Although rmIL-12 has been shown to inhibit angiogenesis in model assays, its effect on angiogenesis within tumors and the functional consequences of these effects have not been studied. The expected effect of tumor angiogenesis inhibition is vascular insufficiency that should give rise to ischemia. We used tumor cell hypoxia, measured by binding of the polyfluorinated 2-nitroimidazole, EF5 (23), as an indicator of tumor ischemia. In vitro studies measuring EF5 binding to cultured K1735 cells under various O2 concentrations showed that EF5 binding is inversely correlated with O2 tension, with maximal discrimination of EF5 binding occurring between 0 and 2% O2 (Fig. 3). Thus, the range of O2 concentrations best determined by EF5 binding to K1735 cells is physiologically

Fig. 2. rmIL-12 inhibition of K1735 angiogenesis is IFN-g-dependent. C3H/HeN mice were injected s.c. with 0.5 ml of Matrigel containing 0.1 ml of PBS, either alone or mixed with an angiogenic stimulus. Both cellular (106 K1735 or N23 cells) and acellular (recombinant basic fibroblast growth factor) angiogenic stimuli were tested. Six days later, the Matrigel implants were excised and assayed for hemoglobin content using the Drabkin’s method (see “Materials and Methods”). The mice simultaneously received daily i.p. injections of either rmIL-12 (125 ng) or PBS vehicle for 5 consecutive days beginning on the day of Matrigel implantation. Each group contained three mice; their implants were assayed individually. Columns, means for each group; bars, SD.

Fig. 1. rmIL-12 therapy of K1735 tumors. Female C3H/HeN mice bearing established s.c. K1735 tumors (,100 mm3) following inoculation of 1 3 106 cultured cells were treated with daily injections of rmIL-12 (125 ng/mouse/day, 5 days/week) for 1, 2, or 3 weeks. Tumor size was measured with calipers and tumor volumes were estimated (5 pd3/6, where d refers to the dimensions; tumors generally had the same perpendicular and bidirectional dimensions). A, the increase in tumor volume (log10 scale) over time of individual PBS-treated tumors (solid black lines) and rmIL-12-treated tumors (dashed gray lines; n 5 3 for all groups, except controls, which had n 5 4). Arrowheads, days of PBS or rmIL-12 injection. B, data points, increase in tumor volume as a multiple of the starting tumor volume is graphed as a function of time (n 5 3); bars, SD. Top, the effect of giving anti-IFN-g (XMG6) antibody (solid black lines) or normal rat antibody (NRA; dashed black lines) on K1735 tumor response to rmIL-12 therapy; bottom, the effect of rmIL-12 (solid black lines) or PBS (dashed black lines) treatment on N23 non-IFN-gresponsive K1735 tumors. For reference, gray solid lines show the mean for rmIL-12treated K1735 tumors, and gray dashed lines show the mean for PBS-treated K1735 tumors (data from A, with bars omitted for clarity).

Fig. 3. In vitro binding of EF5 to K1735 cells under various oxygen concentrations. K1735 tumor cells cultured in glass dishes under defined O2 concentrations with 100 mM EF5 for 3 h were fixed, labeled with Cy3-conjugated anti-EF5 (ELK 3-51) monoclonal antibody, and analyzed by flow cytometry. Cy3/EF5 fluorescent intensity (F, means; bars, SD) is shown as a function of O2 concentration (indicated).

4884


Fig. 4. rmIL-12 therapy results in delayed development of tumor hypoxia. A, K1735 tumors from mice receiving rmIL-12 or PBS were labeled with EF5 in vivo (i.v. injection of 0.25 ml of 10 mM EF5 3 h prior to tumor excision) and excised for flow cytometric analysis following dissociation and staining with ELK 3-51-Cy3 monoclonal antibody. Flow cytometry histograms of representative tumors treated with PBS or rmIL-12 for 1, 2, or 3 weeks are shown. Top horizontal axis, the approximate oxygen concentration based on in vitro correlation of EF5 fluorescence intensity versus oxygen concentration as described in Fig. 3. Bottom horizontal axis, absolute EF5 fluorescence intensity. B, histograms showing means (columns) and SD (bars) of percentage oxic (M) and % hypoxic (,1% O2; ^) tumor cells from K1735 and N23 tumors treated for the indicated periods with PBS or rmIL-12 (n 5 3).

relevant because half-maximal cellular hypoxic responses occur at an O2 concentration of 1.5% (24). To assess the oxygenation of tumor cells in vivo, we injected K17352 tumor-bearing mice treated with rmIL-12 or untreated with EF5 i.v. 3 h prior to tumor excision. Tumor cell suspensions were prepared, stained with a fluorescently tagged monoclonal antibody directed against EF5, and analyzed by flow cytometry (Fig. 4A). Cells from untreated K1735 tumors varying from 2- to 16-mm diameter exhibited a uniform pattern of low EF5 staining similar to that seen in K1735 cells cultured at 10 –20% O2; only 1–2% of cells in these tumors showed EF5 staining at levels equivalent to that of cells cultured at ,1% O2 (Fig. 3). Tumors treated with rmIL-12 for 1 week also appeared to be uniformly oxygenated with minimal EF5 staining, although there was a shift toward a lower median O2 percentage. Tumors receiving 2 weeks of rmIL-12 treatment, however, exhibited a peak of EF5-staining cells (median O2, ;0.1%) not seen in either of the previous treatment groups. This hypoxic peak, which comprised 22% of total tumor cells after 2 weeks of treatment, grew to include the vast majority of tumor cells (80%) after 3 weeks of therapy (Fig. 4B). The progressive development of K1735 tumor hypoxia with rmIL-12 therapy is consistently and reproducibly observed (note the SDs displayed in Figs. 4B and 6B, in which n 5 3 for each group). Hypoxia did not develop when anti-IFN-g antibody was given with rmIL-12 therapy, suggesting that rmIL-12 induction of tumor hypoxia, like its ability to inhibit angiogenesis, is IFN-g dependent. In addition, rmIL-12 failed to induce significant hypoxia in N23 tumors (Fig. 4B), just as it had failed to inhibit N23 tumor cell-induced Matrigel angiogenesis (Fig. 2). The spatial distribution of hypoxic tumor cells relative to blood supply was examined in histological sections of K1735 tumors after staining for EF5 and CD31 (PECAM; Fig. 5A). These results supported those of the flow studies, showing that untreated tumors and tumors treated with rmIL-12 for 1 week have few, if any, visible areas of EF5-stained cells and that tumors treated for 2 weeks contain large patches of EF5-stained cells. The close correlation in EF5 binding obtained by cell suspension and histological studies of rmIL-12treated K1735 tumors suggests that EF5 binding of tumor cell suspensions accurately reflects events in vivo. Examination of microscopic fields from multiple tumors revealed that hypoxia selectively developed in regions of the tumor that were devoid of PECAMstaining vessels (Fig. 5B). This is expected because O2 delivery in these regions depends upon extravascular diffusion, a less efficient process than direct capillary exchange, making them most susceptible

to changes in O2 supply. Together, these studies show that treatment with rmIL-12 leads to the development of tumor cell hypoxia in a manner consistent with its angiogenesis-inhibitory activity. Significantly, the appearance of severe tumor hypoxia was delayed relative to the ability of rmIL-12 to inhibit tumor cell-induced Matrigel angiogenesis (Fig. 2) and suppress tumor growth (Fig. 1A), both of which occurred during the first week of therapy. Hypoxia Resulting from rmIL-12 Therapy Induces Tumor Cell Apoptosis. Because severe hypoxia has been shown to induce tumor cell apoptosis (18, 19), we examined whether rmIL-12-induced tumor cell hypoxia was associated with apoptosis by two-color flow cytometric analysis using TUNEL staining to identify apoptotic cells. In untreated tumors, only 2% of tumor cells were apoptotic. In contrast, 40% of tumor cells were apoptotic after 2 weeks of rmIL-12 therapy, when tumors were significantly hypoxic (Fig. 6, A and B). Apoptosis was strongly associated with hypoxia (20% of normoxic cells were apoptotic, whereas .80% of hypoxic cells were apoptotic), so that, although normoxic cells predominated in these tumors, the majority of apoptotic cells were in the hypoxic population. This trend continued in tumors treated for 3 weeks with rmIL-12, in which 67% of tumor cells were apoptotic and, because hypoxic cells were predominant, the vast majority of apoptotic cells were hypoxic. The relationship between hypoxia and apoptosis during rmIL-12 treatment was examined by testing whether hypoxia could induce apoptosis in K1735 cells in vitro. When cells were incubated under a range of O2 concentrations (10, 1, and 0.1% O2) for 16 or 32 h, marked apoptosis was seen in cells grown at 0.1% O2 (Table 1), which was the median O2 concentration of the hypoxic tumor cell population in vivo after 2 weeks of rmIL-12 therapy (Fig. 4A). By 32 h, .70% of cells in 0.1% O2 were apoptotic, consistent with the 70 – 80% of hypoxic cells observed to undergo apoptosis in vivo (Fig. 6B). In the final 8 h of incubation in 0.1% O2, the pH of the culture medium decreased from pH 7.1 to 6.8. However, acidosis was not the cause of apoptosis in these cells, because culturing K1735 cells at pH 6.8 in air for up to 16 h did not induce significant apoptosis (data not shown). These results strongly suggest that, in vivo, rmIL-12 treatment suppresses K1735 tumor growth via the development of severe hypoxia, which triggers tumor cell apoptosis. rmIL-12 Also Induces Early IFN-g-dependent Tumor Cell Apoptosis Independent of Hypoxia. The observation that rmIL-12 suppressed tumor growth during the first week of therapy prior to the development of severe hypoxia was still unexplained. When tumors treated for 1 week were examined, we found that the level of apoptosis

4885


Fig. 5. Histological analysis of rmIL-12-treated tumors for hypoxia and vascularity. Tumors were excised from K1735-bearing mice treated with rmIL-12 for 0, 1, or 2 weeks; frozen; and stained for EF5 and PECAM, as described in “Materials and Methods.” A, stained sections of size-matched PBS-treated (left), 1-week rmIL-12-treated (middle), or 2-week rmIL-12-treated (right) K1735 tumors. The same sections were stained for EF5 (top) and PECAM (bottom). B, superimposition of the EF5 staining pattern over the equivalent PECAMstained section from three individual tumors treated with rmIL-12 for 2 weeks, demonstrating that hypoxia develops selectively in regions lacking vessels. Far right, taken from the 2-week rmIL-12treated tumor shown in A to confirm the fidelity of the superimposition.

in these tumors (52%) was comparable with that present at later treatment time points (Fig. 6, A and B). However, this apoptosis occurred in the absence of hypoxia, indicating that rmIL-12 was inducing hypoxia-independent as well as hypoxia-dependent tumor cell apoptosis. Several observations led us to suspect that IFN-g caused this hypoxia-independent apoptosis. (a) IFN-g levels were highest in the first week of rmIL-12 treatment when this apoptosis was present. (b) This apoptosis was abrogated by antibody neutralization of IFN-g (Fig. 6C), and non-IFN-g-responsive N23 tumors treated with rmIL-12 for 1 week did not exhibit significant apoptosis. (c) Previous studies had shown that IFN-g can induce apoptosis in vitro in a signal transducers and activators of transcription (STAT)-dependent manner (25). To test this hypothesis, we cultured K1735 cells with different concentrations of murine rIFN-g for 48 or 96 h (Table 2; 96 h approximates the time that tumor cells are exposed to high IFN-g levels in vivo during a 5-day course of rmIL-12 therapy). The percentage of apoptotic cells increased with dose and duration of rIFN-g treatment to a high of 41% with 100 ng rIFN-g/ml for 96 h, which approximates the percentage of apoptotic cells seen in vivo after a week of rmIL-12 treatment (Fig. 6B). That apoptosis was not induced in N23 cells under the same conditions indicates that this is likely a direct effect of IFN-g on K1735 cells. Thus, hypoxia-independent apoptosis induced after 1 week of rmIL-12 therapy can be attributed to direct IFN-g induction of K1735 tumor cell apoptosis. Apoptosis involved 52, 26, and 29% of the nonhypoxic tumor cells after 1, 2, and 3 weeks of rmIL-12 treatment, respectively. The reduction in apoptotic fraction of nonhypoxic tumor cells after the first week of treatment was statistically significant (P , 0.05 for 1 versus 2 weeks and P , 0.01 for 1 versus 3 weeks by Student’s t test) and may be explained by the significantly lower levels of IFN-g induced by rmIL-12 after the first week (Fig. 6B). This reduction in hypoxia-independent, IFN-g-dependent apoptosis was offset, however, by the concomitant rise in hypoxia-dependent apoptosis, so that

a significant fraction of tumor cells was apoptotic throughout the 3 weeks of rmIL-12 treatment. Both apoptosis mechanisms contribute to rmIL-12 control of K1735 tumor growth, but each predominates at a different time. To be certain that apoptosis accounted for rmIL-12 suppression of tumor growth, we also evaluated changes in tumor cell proliferation by propidium iodide staining for DNA content. These studies revealed no increase in S-phase fraction with rmIL-12 treatment to offset the increases in apoptosis (data not shown). In fact, the fraction of cells in S-phase dropped by ;35% in tumors treated for 1 or 2 weeks and by ;65% in tumors treated for 3 weeks. These decreases can be explained by the rise in the percentage of apoptotic tumor cells (which should not be cycling) observed with rmIL-12 treatment and probably do not represent therapeutic suppression of proliferation by healthy tumor cells. DISCUSSION The goal of this study was to examine the effects of rmIL-12 angiogenesis inhibition within tumors to understand how this property of rmIL-12 contributes to its overall antitumor efficacy. Hypothesizing that angiogenesis inhibition would lead to tumor cell ischemia and knowing that ischemia engenders cellular hypoxia, we examined hypoxia, measured by EF5 binding, as a possible physiological consequence of tumor angiogenesis inhibition that could be responsible for its antitumor effect. Intracellular binding of EF5, a polyfluorinated 2-nitroimidazole, is a well-established marker of cellular hypoxia (23). EF5 under hypoxic conditions is metabolically activated and forms covalent adducts with cellular macromolecules that can be detected by monoclonal antibodies. Its binding to cultured tumor cells has been shown to be a quantitative method for determining intracellular oxygen concentration (26). In addition, a recent comparison of EF5 binding with Doppler ultrasound in tumors demonstrated that

4886


EF5 binding is correlated with decreased blood flow (27), thus providing a link between reduced tumor perfusion and hypoxia. Our results support the validity of this approach. Both rmIL-12 angiogenesis inhibition (measured in Matrigel neovascularization assays) and induction of tumor cell hypoxia were dependent on IFN-g and required tumor cells to respond to IFN-g. This latter requirement is likely related to IFN-g-induced tumor cell IP-10 production because K1735 cells produce IP-10 in response to IFN-g stimulation, whereas N23 cells do not (data not shown). The observed regions of tumor cell hypoxia were localized to relatively avascular areas of the tumor. The only difference between the two parameters is that, whereas angiogenesis inhibition was seen in the first week of rmIL-12 treatment, the appearance of severe tumor cell hypoxia was delayed. However, we believe that this delay can be explained physiologically and is consistent with a causal relationship. In a well-oxygenated tumor such as K1735, there are no hypoxic regions, indicating that oxygen delivery to tumor cells is uniformly adequate. Curtailment of new vessel development will likely reduce tumor perfusion gradually, so that there will be a temporal lag before oxygen delivery is reduced to a critical level and becomes manifest as severe hypoxia. In support of this model, median K1735 tumor cell oxygenation decreased after the first week of rmIL-12 treatment (Fig. 4A), although severe hypoxia was not yet apparent. Whereas we are confident from Matrigel neovasularization assays that rmIL-12 inhibits new blood vessel development, we do not know at this time whether it also has effects on existing vessels. Our studies suggest that angiogenesis inhibition leads to tumor cell apoptosis through the agency of hypoxia. Therapeutic efficacy by antiangiogenic agents is associated with increased tumor cell apoptosis (16, 17), but the inducer of this apoptosis has yet to be identified. On the basis of the results with rmIL-12, we propose that inhibition of tumor angiogenesis produces severe tumor hypoxia (or exacerbates it if tumor hypoxia exists prior to therapy) and that hypoxia is largely responsible for the tumor cell apoptosis and tumor growth arrest seen with angiogenesis inhibitor therapy. This does not preclude a contribution from other apoptosis stimuli to tumor stasis, but hypoxiadependent apoptosis may be an inevitable consequence of tumor ischemia from angiogenesis inhibition, provided that the tumor cells are susceptible. Of course, ischemia from angiogenesis inhibition brings about physiological changes in addition to hypoxia, such as acidosis or hypoglycemia, that may also contribute to tumor cell apoptosis. For example, acidosis has been demonstrated to induce apoptosis in a hypoxia-independent manner in vitro (28). In our tumor Table 1 Induction of K1735 apoptosis in vitro by hypoxiaa % apoptotic cells 10% O2

1.0% O2

0.1% O2

Cell type

16 h

32 h

16 h

32 h

16 h

32 h

K1735

2.7

5.0

4.0

7.3

29

72

a

Fig. 6. Development of apoptosis and hypoxia in rmIL-12-treated K1735 tumors. K1735 tumors from mice receiving rmIL-12 or PBS were labeled with EF5 in vivo, excised, and dissociated for flow cytometric analysis of hypoxia (ELK 3-51-cyanine-5) and apoptosis (TUNEL staining with fluorescein-dUTP). A, flow cytometry scatterplots from representative tumors treated with PBS or with rmIL-12 for 1, 2, or 3 weeks. B, histograms showing means (columns) and SD (bars) of percentage oxic and hypoxic (,1% O2) tumor cells (M, n 5 3). ^, means (columns) and SD (bars) of percentage apoptotic cells among oxic and hypoxic cell populations. The total percentage of apoptotic tumor cells and the average serum IFN-g level (taken at day 4 during the final week of rmIL-12 therapy and determined by RIA) are listed above the graph. C, histograms showing percentage apoptotic cells in K1735 and N23 tumors treated with rmIL-12 for 1 week; mice bearing the K1735 tumors were also given either a normal rat antibody (NRA) or a monoclonal antibody (XMG6) against IFN-g.

K1735 cells were cultured on 6-cm glass dishes under various O2 concentrations for indicated durations of time. Apoptosis was determined by TUNEL staining and analyzed by flow cytometry. Each value listed is a percentage of total cells analyzed. Table 2 Induction of K1735 apoptosis in vitro by IFN-ga % apoptotic cells 0 ng/ml Cell type K1735 N23

10 ng/ml

50 ng/ml

100 ng/ml

2 days

4 days

2 days

4 days

2 days

4 days

2 days

4 days

0.3

1.7

6.4

23

17

30

28 5.8

41 12

a K1735 cells were cultured on 10-cm plastic dishes with various concentrations of IFN-g for indicated durations of time. Apoptosis was determined by TUNEL staining and analyzed by flow cytometry. Each value listed is a percentage of total cells analyzed.

4887


model, however, acidosis seems to contribute little to tumor cell apoptosis induced by hypoxia. rmIL-12 controls K1735 tumor growth from the outset of therapy, unlike agents such as angiostatin and endostatin, which are suspected to act purely as angiogenesis inhibitors and which control tumor growth only after 1 or 2 weeks of therapy (29, 30). This was a clue that another antitumor mechanism is responsible for the initial phase of rmIL-12 tumor control, which we suggest is direct induction of tumor cell apoptosis by IFN-g. An earlier study had demonstrated this to be an effect of high-dose IFN-g stimulation in vitro (25), but the studies presented here suggest a role in suppressing tumor growth in vivo. IFN-g has pleiotropic effects on cells, many of which are indirect and mediated by other factors, making it difficult to prove that hypoxia-independent, IFN-gdependent tumor cell apoptosis observed in vivo results from the apoptosis-inducing activity of IFN-g. However, the absence of this type of apoptosis in N23 tumor cells, which should resist direct effects of IFN-g but not those of intermediary factors induced by IFN-g, argues that the apoptosis results from the direct action of IFN-g on tumor cells. We do not know whether the concentrations of IFN-g necessary to stimulate apoptosis in vitro are present in the local tumor environment in vivo. However, we do know that other proinflammatory cytokines likely to be induced by rmIL-12 in vivo (e.g., tumor necrosis factor-a) act synergistically in vitro with IFN-g to induce K1735 tumor cell apoptosis (data not shown). The presence of these other cytokines reduces the concentration of IFN-g needed to induce apoptosis, increasing the likelihood that effective induction of tumor cell apoptosis by IFN-g occurs with cytokine concentrations that are achieved in vivo. This mechanism explains the immediate suppression of K1735 tumor growth seen with rmIL-12 administration and warrants consideration as a mechanism underlying the rapid response of other tumors to rmIL-12 therapy. A question that arises is which host cells are responsible for the IFN-g produced during the period of rmIL-12 administration. We believe that the best candidates for primary acute mediators of rmIL-12 antitumor activity and producers of IFN-g are NK cells. These have been shown to be the first cells to infiltrate tumors in response to IL-12 therapy, where they appear to mediate both tumor cell killing and angiogenesis inhibition (31, 32). In addition, administration of antibodies directed against NK cells during the first week of rmIL-12 therapy is associated with a reduction in systemic levels of IFN-g and a loss of angiogenesis inhibitory activity (data not shown). Finally, rmIL-12 used in the manner that generally produces the greatest acute antitumor effect (high doses administered frequently) has been shown to transiently but profoundly inhibit T-cell mitogenesis and antigen-specific responses (8, 9), making T cells unlikely mediators of rmIL-12 acute antitumor effects. However, further studies are required to define the role of different effector cell populations during IL-12 therapy more completely. In summary, examination of the time course of rmIL-12 effects on the physiology of therapeutically responsive K1735 melanoma tumors reveals that two distinct apoptosis mechanisms suppress tumor growth at different periods during treatment. IFN-g-induced apoptosis is initially active but subsequently declines. This decline is accompanied by a corresponding rise in hypoxia-induced apoptosis, which results in a relatively stable level of tumor cell apoptosis that suppresses tumor expansion throughout treatment. Beyond rmIL-12 therapy, these studies suggest that hypoxia-induced tumor cell apoptosis may be a general mechanistic pathway whereby treatment with antiangiogenic agents leads to tumor stasis or regression. The involvement of hypoxia carries potentially important consequences and implications for this form of therapy. For example, hypoxia is known to induce a variety of

cellular changes that can be adaptive for tumors, such as alterations in glucose metabolism and increased vascularity (33). In addition, hypoxia confers resistance to the cytotoxic effects of ionizing radiation (34), suggesting that care is warranted if angiogenesis inhibitors are to be used in combination with radiation therapy. Hypoxia has also been shown to select for cells with diminished apoptotic potential and p53 mutation (35), which could lead to expansion of tumor cell populations that are more therapeutically refractory (36, 37). Although these considerations may only be applicable with chronic hypoxia, longterm maintenance therapy with angiogenesis inhibitors has been proposed (30, 38). Additionally, tumors may vary in their susceptibility to the apoptotic effects of hypoxia, which may be a basis for variation in tumor response to antiangiogenesis therapy. Finally, if severe tumor hypoxia results from effective inhibition of tumor angiogenesis and mediates its antitumor effect, measurements of tumor hypoxia may be a sensitive and meaningful way to monitor clinical therapy with antiangiogenic agents. ACKNOWLEDGMENTS We thank William N. Procopio for figure design and construction, M. Wysocka for immunological and IFN-g RIA reagents, and the Genetics Institute (Andover, MA) for rmIL-12. We thank Newman Yeilding and Denise LaTemple for critical reading of this manuscript.

REFERENCES 1. Brunda, M. J., Luistro, L., Warrier, R. R., Wright, R. B., Hubbard, B. R., Murphy, M., Wolf, S. F., and Gately, M. K. Antitumor and antimetastatic activity of interleukin 12 against murine tumors. J. Exp. Med., 178: 1223–1230, 1993. 2. Zitvogel, L., Tahara, H., Robbins, P. D., Storkus, W. J., Clarke, M. R., Nalesnik, M. A., and Lotze, M. T. Cancer immunotherapy of established tumors with IL-12. Effective delivery by genetically engineered fibroblasts. J. Immunol., 155: 1393– 1403, 1995. 3. Tahara, H., Zitvogel, L., Storkus, W. J., Zeh, H. J., McKinney, T. G., Schreiber, R. D., Gubler, U., Robbins, P. D., and Lotze, M. T. Effective eradication of established murine tumors with IL-12 gene therapy using a polycistronic retroviral vector. J. Immunol., 154: 6466 – 6474, 1995. 4. Coughlin, C. M., Wysocka, M., Trinchieri, G., and Lee, W. M. F. The effect of interleukin 12 desensitization on the antitumor efficacy of recombinant interleukin 12. Cancer Res., 57: 2460 –2467, 1997. 5. Coughlin, C. M., Wysocka, M., Hurzawa, H. L., Lee, W. M. F., Trinchieri, G., and Eck, S. L. B7–1 and interleukin 12 synergistically induce effective antitumor immunity. Cancer Res., 55: 4980 – 4987, 1995. 6. Brunda, M. J. Role of IL-12 as an anti-tumor agent: current status and future directions. Res. Immunol., 146: 622– 628, 1995. 7. Trinchieri, G. Interleukin-12: a proinflammatory cytokine with immunoregulatory functions that bridge innate resistance and antigen-specific adaptive immunity. Annu. Rev. Immunol., 13: 251–276, 1995. 8. Kurzawa, H., Wysocka, M., Aruga, E., Chang, A. E., Trinchieri, G., and Lee, W. M. F. Interleukin 12 enhances cellular immune responses to vaccination only after a period of suppression. Cancer Res., 58: 491– 499, 1998. 9. Koblish, H. K., Hunter, C. A., Wysocka, M., Trinchieri, G., and Lee, W. M. F. Immune suppression by recombinant interleukin (rIL)-12 involves interferon g induction of nitric oxide synthase 2 (iNOS) activity: inhibitors of NO generation reveal the extent of rIL-12 vaccine adjuvant effect. J. Exp. Med., 188: 1603–1610, 1998. 10. Voest, E. E., Kenyon, B. M., O’Reilly, M. S., Truitt, G., D’Amato, R. J., and Folkman, J. Inhibition of angiogenesis in vivo by interleukin-12. J. Natl. Cancer Inst., 87: 581–586, 1995. 11. Sgadari, C., Angiolillo, A. L., and Tosato, G. Inhibition of angiogenesis by interleukin-12 is mediated by the interferon-inducible protein-10. Blood, 87: 3877–3882, 1996. 12. Coughlin, C. M., Salhany, K. E., Gee, M. S., LaTemple, D. C., Kotenko, S., Ma, X. J., Gri, G., Wysocka, M., Kim, J. E., Liu, L., Liao, F., Farber, J. M., Pestka, S., Trinchiert, G., and Lee, W. M. F. Tumor cell responses to IFN-g affect tumorigenicity and response to IL-12 therapy and antiangiogenesis. Immunity, 9: 25–34, 1998. 13. Majewski, S., Marczak, M., Szmurlo, A., Jablonska, S., and Bollag, W. Interleukin-12 inhibits angiogenesis induced by human tumor cell lines in vivo. J. Invest. Dermatol., 106: 1114 –1118, 1996. 14. Dias, S., Boyd, R., and Balkwill, F. IL-12 regulates VEGF and MMPs in a murine breast cancer model. Int. J. Cancer, 78: 361–365, 1998. 15. Ruegg, C., Yilmaz, A., Bieler, G., Bamat, J., Chaubert, P., and Lejeune, F. J. Evidence for the involvement of endothelial cell integrin aVb3 in the disruption of the tumor vasculature induced by TNF and IFN-g. Nat. Med., 4: 408 – 414, 1998. 16. O’Reilly, M. S., Holmgren, L., Chen, C., and Folkman, J. Angiostatin induces and sustains dormancy of human primary tumors in mice. Nat. Med., 2: 689 – 692, 1996.

4888


17. Holmgren, L., O’Reilly, M. S., and Folkman, J. Dormancy of micrometastases: balanced proliferation and apoptosis in the presence of angiogenesis inhibition. Nat. Med., 1: 149 –153, 1995. 18. An, W. G., Kanekal, M., Simon, M. C., Maltepe, E., Blagosklonny, M. V., and Neckers, L. M. Stabilization of wild-type p53 by hypoxia-inducible factor 1a. Nature (Lond.), 392: 405– 408, 1998. 19. Carmeliet, P., Dor, Y., Herbert, J. M., Fukumura, D., Brusselmans, K., Dewerchin, M., Neeman, M., Bono, F., Abramovich, R., Maxwell, P., Koch, C. J., Ratcliffe, P., Moons, L., Jain, R. K., Collen, D., and Keshet, E. Role of HIF-1a in hypoxiamediated apoptosis, cell proliferation and tumour angiogenesis. Nature (Lond.), 394: 485– 490, 1998. 20. Kripke, M. Speculations on the role of ultraviolet radiation in the development of malignant melanoma. J. Natl. Cancer Inst., 63: 541–545, 1979. 21. Evans, S. M., Jenkins, W. T., Joiner, B., Lord, E. M., and Koch, C. J. 2-Nitroimidazole (EF5) binding predicts radiation resistance in individual 9L s.c. tumors. Cancer Res., 56: 405– 411, 1996. 22. Lee, J., Fenton, B. M., Koch, C. J., Frelinger, J. G., and Lord, E. M. Interleukin 2 expression by tumor cells alters both the immune response and the tumor microenvironment. Cancer Res., 58: 1478 –1485, 1998. 23. Lord, E. M., Harwell, L., and Koch, C. J. Detection of hypoxic cells by monoclonal antibody recognizing 2-nitroimidazole adducts. Cancer Res., 53: 5721–5726, 1993. 24. Jiang, B. H., Semenza, G. L., Bauer, C., and Marti, H. H. Hypoxia-inducible factor 1 levels vary exponentially over a physiologically relevant range of O2 tension. Am. J. Physiol., 271: C1172–C1180, 1998. 25. Chin, Y. E., Kitagawa, M., Kuida, K., Flavell, R., and Fu, X. Y. Activation of the STAT signaling pathway can cause expression of caspase 1 and apoptosis. Mol. Cell. Biol., 17: 5328 –5337, 1997. 26. Koch, C. J., Evans, S. M., and Lord, E. M. Oxygen dependence of cellular uptake of EF5 [2-(2-nitro-1H-imidazol-1-yl)-N-(2,2,3,3,3-pentafluoropropyl) acetamide]: analysis of drug adducts by fluorescent antibodies vs bound radioactivity. Br. J. Cancer, 72: 869 – 874, 1995.

27. Evans, S. M., Laughlin, K. M., Pugh, C. R., Sehgal, C. M., and Saunders, H. M. Use of power Doppler ultrasound-guided biopsies to locate regions of tumour hypoxia. Br. J. Cancer, 76: 1308 –1314, 1997. 28. Schmaltz, C., Hardenbergh, P. H., Wells, A., and Fisher, D. E. Regulation of proliferation-survival decisions during tumor cell hypoxia. Mol. Cell. Biol., 18: 2845–2854, 1998. 29. O’Reilly, M. S., Boehm, T., Shing, Y., Fukai, N., Vasios, G., Lane, W. S., Flynn, E., Birkhead, J. R., Olsen, B. R., and Folkman, J. Endostatin: an endogenous inhibitor of angiogenesis and tumor growth. Cell, 88: 277–285, 1997. 30. Boehm, T., Folkman, J., Browder, T., and O’Reilly, M. S. Antiangiogenic therapy of experimental cancer does not induce acquired drug resistance. Nature (Lond.), 390: 404 – 407, 1997. 31. Pham-Nguyen, K. B., Yang, W., Saxena, R., Thung, S. N., Woo, S. L., and Chen, S. H. Role of NK and T cells in IL-12-induced anti-tumor response against hepatic colon carcinoma. Int. J. Cancer, 81: 813– 819, 1999. 32. Yao, L., Sgadari, C., Furuke, K., Bloom, E. T., Teruya-Feldstein, J., and Tosato, G. Contribution of natural killer cells to inhibition of angiogenesis by interleukin-12. Blood, 93: 1612–1621, 1999. 33. Guillemin, K., and Krasnow, M. A. The hypoxic response: huffing and HIFing. Cell, 89: 9 –12, 1997. 34. Horsman, M. R., and Overgaard, J. The oxygen effect. In: G. G. Steel (ed.), Basic Clinical Radiobiology, pp. 132–140. New York: Oxford University Press, 1997. 35. Graeber, T. G., Osmanian, C., Jacks, T., Housman, D. E., Koch, C. J., Lowe, S. W., and Giaccia, A. J. Hypoxia-mediated selection of cells with diminished apoptotic potential in solid tumours. Nature (Lond.), 379: 88 –91, 1996. 36. Lowe, S. W., Ruley, H. E., Jacks, T., and Housman, D. E. p53-dependent apoptosis modulates the cytotoxicity of anticancer agents. Cell, 74: 957–967, 1993. 37. Lowe, S. W., Bodis, S., McClatchey, A., Remington, L., Ruley, H. E., Fisher, D. E., Housman, D. E., and Jacks, T. p53 status and the efficacy of cancer therapy in vivo. Science (Washington DC), 266: 807– 810, 1994. 38. Kerbel, R. S. Inhibition of tumor angiogenesis as a strategy to circumvent acquired resistance to anticancer therapeutic agents. BioEssays, 13: 31–36, 1991.

4889