Chemotherapy induces tumor clearance independent of apoptosis

Abstract

Dysregulation of apoptosis is associated with the development of human cancer and resistance to anti-cancer therapy. The ultimate goal of cancer treatment is to selectively induce cancer cell death and overcome drug resistance. A deeper understanding of how a given chemotherapy affects tumor cell death is needed to develop strategically designed anti-cancer agents. Here we utilize a xenograft mouse tumor system generated from genetically defined cells deficient in apoptosis to examine the involvement of multiple forms of cell death induced by cyclophosphamide (CP), a DNA alkylating agent commonly used in chemotherapy. We find that while apoptosis facilitates tumor regression, it is dispensable for complete tumor regression as other forms of cell death are activated. Sporadic necrosis is observed in both apoptosis-competent and deficient tumors evident by tumor cell morphology, extracellular release of high mobility group protein B1 (HMGB1), and activation of innate immune cells in CP treated tumors. Our findings indicate that in apoptosis-deficient tumors, necrosis may play a fundamental role in tumor clearance by stimulating the innate immune response.

Introduction

Apoptosis, or Type I programmed cell death, plays a crucial role in the clearance of cells that have potentially harmful genetic mutations. The ability of a cell to evade apoptosis is a crucial step in tumorigenesis (1–3). Despite the fact that human cancers are defective in their apoptotic pathways, DNA alkylating agents remain among the most effective chemotherapeutic agents used clinically (4), suggesting that alternative cell death pathways are activated. Five main types of cytocidal and cytostatic mechanisms have been described in the context of cancer therapy: apoptosis, necrosis, mitotic catastrophe, senescence, and autophagy (2, 5–7). However, a comprehensive understanding of tumor cell death has been lacking due to the complexity of a tumor’s response to chemotherapy. Presently, most mechanistic studies of the anti-cancer activities of DNA alkylating agents have been conducted using cultured cells. It is therefore important to develop an in vivo system to evaluate the contribution of the alternative cell death pathways in anti-cancer therapy. Apoptosis is mainly controlled by two sets of molecules: the Bcl-2 family of proteins, and the caspases. Deletion of the pro-apoptotic Bcl-2 family members Bax and Bak is sufficient to prevent apoptotic events initiated by mitochondrial membrane permeabilization (8). We utilized these cells to address the question as to how conventional chemotherapy induces cancer cell death in vivo.

Materials and Methods

Cell lines

Immortalized wild-type and bax^−/−bak^−/− murine embryonic fibroblasts (MEFs) (8) were retrovirally transformed with E1A and K-Ras. The cells were subcutaneously injected into the back of 6–8 week old male athymic nude mice (Taconic Farms). When tumors formed they were excised from the mice, minced, and digested with trypsin-EDTA (0.05%) and collagenase A (1 mg/mL) to generate stable tumor cell lines.

Antibodies

Bax (Santa Cruz), Bak (Upstate), Caspase 3 (Transducation Lab), cleaved Caspase 3 (Cell Signaling), γH2A.X (Cell Signaling), HMGB1 (Abcam), β-tubulin (Sigma), IL-1β (R&D systems), F4/80 (Serotec), Neutrophil (Serotec), PE-conjugated anti-Mac-1 (BD Biosciences), Alexa 488-conjugated anti-Mac-1 (BD Biosciences), APC-conjugated anti-F4/80 (eBiosciences), PE-conjugated anti-Gr-1 (BD Biosciences), and Alexa 488-conjugated anti-Gr-1 (Serotec).

Electron microscopy

The tumor samples for TEM were collected immediately after sacrificing the mouse to ensure tissue integrity. TEM was performed according to standard protocol by the Central Microscopy Imaging Center at Stony Brook University.

Xenograft mouse tumor experiments

Tumors were established by injecting 1×10⁶ tumor cells into the mid flanks of nude mice. When palpable tumors formed, mice were randomly grouped and either left untreated or treated via intraperitoneal injections of 170 mg/kg of cyclophosphamide monohydrate (CP) (Sigma) every 5 days. The tumor length (l) and width (w) were measured every 4–5 days with electronic calipers. Tumor volume (v) was calculated using the formula: v = (l×w²)/2. The animals bearing untreated tumors were sacrificed post tumor inoculation before tumors reached 4 cm³ in size in compliance with the Stony Brook University IACUC guidelines.

Statistical analysis

Data are represented as mean ± S.E.M. (standard error of the mean). Statistical analyses were performed using Microsoft Excel.

Results

Cyclophosphamide induces tumor regression in vivo independent of key apoptosis regulators

In order to study the contribution of different cell death pathways in the anti-tumor activity of chemotherapy, we first established an in vivo system to compare tumors derived from apoptosis-competent and deficient cells. E1A and K-Ras oncoproteins were used to transform genetically defined MEFs isolated from wild-type and bax^−/−bak^−/− mice (Fig. 1A). Moreover, the anti-apoptotic protein Bcl-xL was expressed in wild-type cells (Fig. 1B). Stable tumor cell lines were generated from these MEFs which maintained their respective genotypes (Fig. 1A and 1B).

Figure 1. DNA alkylating agents induce tumor regression independent of apoptosis.

Mouse embryonic fibroblasts (MEFs) were obtained from wild-type and bax^−/−bak^−/− mice and transformed with E1A and K-Ras. Transformed cells were injected into nude mice to generate tumors. The tumors were excised from the mice to generate stable tumor cell cultures. (A) Wild-type and bax^−/−bak^−/− tumor cells were plated and the cell density was determined on the indicated days. Data shown are the averages of four independent countings ± S.E.M. Cell lysates were made and probed for Bax and Bak by immunoblotting analysis. N.S., a non-specific band to verify equal loading. (B) Bcl-xL was expressed in the wild-type E1A and K-Ras-transformed MEFs. Lysates from vector control and Bcl-xL expressing cells were probed for Bcl-xL. Tom40 was used as a loading control. Wild-type and Bcl-xL-expressing cells were treated with staurosporine (STS, 0.5 μM) or etoposide (Eto, 25 μM) for 24 hours. Cell viability was measured by trypan blue exclusion. Data shown are the averages of three independent experiments ± S.E.M. (C) Transformed wild-type, bax^−/−bak^−/−, and Bcl-xL expressing MEFs were injected into nude mice to generate tumors. The tumors were excised from the mice to generate stable tumor cell cultures. One million of the established tumor cells were injected subcutaneously into nude mice. Mice were either left untreated or treated with 170 mg/kg of CP intra-peritoneally every 5 days, as indicated by the arrows. The tumor volume was calculated based on caliper measurement and plotted. Note that tumors of both genotypes regressed after CP treatment, regardless of their ability to die by apoptosis. (D) E1A and K-Ras-transformed wild-type and bax^−/−bak^−/− tumor cells were transfected with GFP and RFP, respectively. The tumors were injected bilaterally into nude mice and were left untreated (n = 5), or treated (n = 12) with 170 mg/kg of CP every 5 days. Tumors were visualized by the Maestro small animal imaging system. Representative mice are shown.

The stable tumor cells were injected into nude mice to test for their response to chemotherapeutic treatment in vivo. Tumors developed from all three cell lines (Fig. 1C). Nine days after tumor implantation, cyclophosphamide (CP) was injected into the tumor-bearing mice intraperitoneally at 170 mg/kg every 5 days. In response to CP, tumors derived from all three MEF lines stopped growing and progressively regressed, indicating that in vivo, DNA alkylating damage possesses anti-tumor activity independent of key apoptosis regulators Bax and Bak, and is unblockable by Bcl-xL (Fig. 1C).

In order to compare the response of apoptosis-proficient and deficient tumors to CP treatment in the same environment, the wild-type and bax^−/−bak^−/− tumor cells were labeled with GFP and RFP, respectively, and injected bilaterally into the same animal. The tumors were monitored by a fluorescence imager (Fig. 1D). The treated wild-type and bax^−/−bak^−/− tumors both grew for an additional 3 to 4 days after the first CP treatment, then began to regress. The treated bax^−/−bak^−/− tumors showed an initial delay in response to CP compared to the wild-type tumors, but most of these tumors eventually resolved to undetectable limits. Overall, in two separate experiments, 92% (11 out of 12) of the wild-type tumors and 50% (6 out of 12) of the bax^−/−bak^−/− tumors regressed to a size that was not detectable by fluorescence imaging. The 6 remaining treated bax^−/−bak^−/− tumors were still detectable at the end of the study (after day 50) but had progressively decreased from their peak volume. Importantly, when CP was discontinued on 6 mice whose wild-type and bax^−/−bak^−/− tumors regressed to a non-detectable limit, these mice remained tumor free for up to 9.5 months before they died due to natural causes. These results demonstrate that: 1) Bax/Bak-mediated apoptosis contributes to a more rapid tumor regression, as the bax^−/−bak^−/− tumors showed a delay in response; and 2) the apoptosis machinery may be dispensable for tumor regression in response to DNA alkylating agents, as the bax^−/−bak^−/− and Bcl-xL overexpressing tumors still regressed despite their defect in the apoptotic pathway.

Apoptosis is not the sole cause for cyclophosphamide-induced tumor regression

The observation that bax^−/−bak^−/− tumors regressed after CP-treatment suggests that apoptosis may be dispensable for chemotherapy-induced tumor regression. We performed transmission electron microscopy (TEM) analysis to examine the form of cell death (Fig. 2A). After CP treatment, both apoptotic cells and necrotic cells at various stages were observed in wild-type tumors. In bax^−/−;bak^−/− tumors, no apoptotic cells were apparent, while necrotic cells were commonly observed in CP-treated tumors. Some nuclei in the treated cells appeared to be larger and pleomorphic, indicative of senescence and/or mitotic catastrophe (9). In both wild-type and bax^−/−bak^−/− tumors treated with CP, lymphocytes, characterized by an electron dense cell body and high nucleus/cytoplasm ratio (10), were also observed (Fig. 2A).

Figure 2. Cyclophosphamide induces sporadic tumor necrosis.

E1A and K-Ras-transformed wild-type and bax^−/−bak^−/− tumor cells were injected bilaterally into the same nude mouse to allow tumors to form. When palpable tumors formed, the mice were treated with CP (170 mg/kg) every 5 days, or left untreated (Fig. 1). Tumors were obtained at various time points after CP treatment. (A) Tissues from tumors left untreated or treated with CP were subjected to transmission electron microscopic (TEM) analysis. Representative pictures are shown. Apoptotic cells (indicated by condensed chromatin), necrotic cells (indicated by disruption of cell structure, lack of chromatin condensation, and formation of a large amount of vacuoles), and leukocytes are observed (A, apoptotic cells; L, leukocytes; N, necrotic cells). Note that cells with large and pleomorphic nuclei indicative of senescence or mitotic catastrophe are present in the CP-treated tumors. (B) Sections from tumors untreated or treated with CP were assayed for cleaved Caspase 3, and for apoptotic DNA fragmentation using a TUNEL assay (Chemicon International). The dark brown staining indicates positive cells (apoptotic). Representative pictures of untreated tumors and tumors treated with CP for 1 day are shown. Similar results were obtained from tumors treated with CP 2, 3, 4, and 5 times and are not shown. Quantification was performed by counting positive cells under a microscope in randomly selected areas. Data shown represents averages of 5 independent countings ± S.E.M. (C) Tumor lysates from an untreated and two independent CP-treated animals (each received one treatment) were prepared and subjected to immunoblotting analysis for cleaved Caspase 3 and phosphorylated H2A.X. HMGB1 was probed as a control for equal loading. (D) Paraffin-embedded sections were made and assessed with IHC using an anti-HMGB1 antibody. Nuclear staining is present in the untreated tissue whereas cytoplasmic and extranuclear staining of HMGB1 is found in the tumor tissue treated twice with CP, indicative of necrosis. Similar results were obtained from tumors treated with CP 1, 3, 4, and 5 times and are not shown. Scale bar represents 20 μm.

To further determine if apoptosis occurred in tumor cells following CP treatment, frozen and paraffin-embedded tumor sections were subjected to immunohistochemistry (IHC) for cleaved Caspase 3 and terminal transferase dUTP end labeling (TUNEL) assay, both hallmarks for caspase-dependent apoptosis. The treated wild-type tumors displayed positive staining for cleaved Caspase 3 and TUNEL, whereas the treated bax^−/−bak^−/− tumor showed virtually no positive staining (Fig. 2B). The same effect was observed by immunoblotting for cleaved Caspase 3 using tumor lysates (Fig. 2C). An antibody against phosphorylated H2A.X (γH2A.X) revealed that both wild-type and bax^−/−bak^−/− tumors incurred a similar extent of DNA damage following CP treatment (Fig. 2C). Together with those shown in Fig. 1, these results indicate that while apoptosis contributes to CP-induced tumor regression, tumor cells deficient in apoptosis can still die via alternative forms of cell death.

DNA alkylating damage triggers sporadic necrosis in vivo

TEM analysis indicated that necrosis may occur in CP-treated tumors (Fig. 2A). To further examine this, IHC was performed on CP-treated tumor tissues using an antibody against HMGB1 (high-mobility group box 1) protein. HMGB1 is a nuclear protein that binds tightly to chromatin in apoptotic cells, whereas during necrosis it is released into the extracellular environment (11). IHC analysis revealed nuclear staining of HMGB1 in untreated tumor tissues. In contrast, extracellular HMGB1 staining was observed in both wild-type and bax^−/−bak^−/− CP-treated tumors (Fig. 2D), indicating that necrosis occurs in response to CP in both apoptosis-proficient and deficient tumors. It is interesting to note that this CP-induced-necrosis is different from necrosis that is often observed in solid tumors with overgrowth. The latter constitutes the “necrotic centers” comprised of large amounts of necrotic cells resulting from limited oxygen and nutrient supplies due to tumor overgrowth and lack of vascularization. CP-induced necrotic cells were scattered and evenly distributed throughout the tumor mass, as judged by both TEM and HMGB1 staining (Fig. 2A and 2D), thus is referred to as “sporadic necrosis”.

To further confirm that CP can induce necrosis, we used a biologically active metabolite of CP, mafosfamide (MAF) (12) for in vitro cell culture studies. Similar to CP in vivo, MAF induced DNA damage indicated by the increase of γH2A.X (Fig. 3C). Treatment of wild-type, Bcl-xL-expressing, and bax^−/−bak^−/− tumor cells with MAF induced cell death in all three cell lines (Fig. 3A). Wild-type cells were more sensitive than Bcl-xL-expressing cells, which were more sensitive than bax^−/−bak^−/− cells, indicating that Bax/Bak deficiency suppresses apoptosis more efficiently than Bcl-xL overexpression (Fig. 3A). This was also indicated by the levels of Caspase 3 cleavage, which correlated to a cell’s ability to die by apoptosis. While apoptosis was virtually blocked in bax^−/−bak^−/− cells, Bcl-xL only shifts the balance between the anti-and pro-apoptotic Bcl-2 proteins and thus did not completely block apoptosis (Fig. 3D). Regardless, microscopic analysis revealed that the dead cells displayed features of necrosis such as plasma membrane dilation and disruption in all three cell lines (Fig. 3B). Additionally, HMGB1 was found in the cell culture medium from both MAF-treated apoptosis-proficient and deficient cells, indicative of necrosis (Fig. 3D). The more rapid release of HMGB1 in wild-type cells is probably due to “secondary necrosis”, since apoptotic cells are not engulfed and cleared as they are in vivo.

Figure 3. HMGB1 is released from tumor cells in response to DNA alkylating damage.

(A) Wild-type, Bcl-xL-expressing, and bax^−/−bak^−/− tumor cells were treated for 24 hours with increasing concentrations of mafosfamide (MAF). Cell death was measured by trypan blue staining, and expressed as average of 2 independent experiments ± S.E.M. (B) Wild-type, Bcl-xL-expressing, and bax^−/−bak^−/− tumor cells were treated with 12.5 μg/mL of MAF for indicated days. Cells were photographed under a phase-contrast filter. (C) Wild-type, Bcl-xL-expressing, and bax^−/−bak^−/− tumor cells were treated with 12.5 μg/mL of MAF for indicated hours. Cell lysates were made and immunoblotting was performed using an anti-γH2A.X antibody. β-tubulin was probed for equal loading. (D) Wild-type, Bcl-xL-expressing, and bax^−/−bak^−/− tumor cells were treated with 12.5 μg/mL of MAF for the indicated days. The cells and the cell culture media were collected separately and subjected to immunoblotting using an anti-HMGB1 antibody. The cell lysates were also probed with an antibody recognizing both pro- and cleaved Caspase 3. Note that HMGB1 is retained inside untreated cells, whereas it is found in the cell culture media of treated cells.

CP-induced cell death activates the innate immune response

A fundamental feature of apoptosis is that apoptotic cells are quickly engulfed as an intact corpse in vivo and thus do not trigger pro-inflammatory responses. In contrast, cells dying by necrosis release intracellular contents into the extracellular environment and cause pro-inflammatory responses. In TEM analysis, we observed the presence of leukocytes in both the wild-type and bax^−/−bak^−/− CP-treated tumors, suggesting infiltration of innate immune cells (Fig. 2A). This is further confirmed by IHC of the tumor sections and flow cytometry using single cell suspension prepared from the tumor tissues, using an antibody against F4/80, a pan macrophage marker, and an antibody against an allotypic marker of neutrophils (Fig. 4A and suppl. Fig. S1A).

Figure 4. Cyclophosphamide induces a pro-inflammatory response.

Wild-type and bax^−/−bak^−/− untreated and CP-treated tumor tissue was assessed for innate immune cell infiltration. IHC and flow cytometry were used to detect the infiltration of macrophages (A). Tissue sections from animals that received 1, 2, 3, 4, or 5 CP-treatments were assessed for macrophage (F4/80) infiltration. Similar results were obtained at the various treatment points. Photographs of tissues from untreated or CP-treated (twice) mice are shown. Positive cells were counted in 5 randomly selected microscopic fields. Data shown are the average ± S.E.M. Additionally, single tumor cell suspensions were obtained from the untreated and treated tumors by digesting the tumor tissues with trypsin-EDTA (0.05%) and collagenase A (10 μg/mL). Tumor cell suspensions were subjected to flow cytometry analysis. Due to the expression of GFP and RFP in the wild-type and bax^−/−bak^−/− tumors, respectively, PE-conjugated Mac-1 (BD Biosciences) was used for the wild-type, and Alexa-488-conjugated Mac-1 (BD Biosciences) was used for the bax^−/−bak^−/− tumor cells. Both samples were co-stained with APC-conjugated F4/80 (eBiosciences). (B) IHC was performed using an anti-IL-1β antibody to detect activated leukocytes in the tumor tissue. (C and D) Tissue lysates were made from both wild-type and bax^−/−bak^−/− tumors that were left untreated or treated once (1x) or twice (2x) with CP. One-hundred μg of tumor lysates were run on an enzyme-linked immuno sorbent assay (ELISA, R&D systems) to determine the levels of IL-1β (C) and tumor necrosis factor α (TNFα) (D). The amount of IL-1β and TNFα in tumor tissues were calculated and expressed as picogram per milligram of tumor protein. Each data point represents tumors isolated from two independent mice and each sample was run in duplicate. Shown is the average ± S.E.M.

Both apoptotic and necrotic cells can attract phagocytes. However, it is generally accepted that phagocytes such as macrophages secrete immune-suppressive cytokines upon engulfing apoptotic cells, whereas those encountering necrotic cells secrete pro-inflammatory cytokines (13, 14). To examine this issue, we determined whether the leukocytes recruited to the tumor tissue produced pro-inflammatory cytokines. In both wild-type and bax^−/−bak^−/− tumors, CP treatment increased the number of interleukin- 1β (IL-1β) positive cells and the levels of both IL-1β and TNFα (Fig. 4B–D). To further investigate if macrophage activation contributes to tumor clearance, we used gadolinium chloride (GdCl₃) that has been shown to deplete peripheral blood macrophages (15, 16). We tested the effect of GdCl₃ on E1A/K-Ras transformed Bcl-xL-expressing tumor cells. Animals bearing these tumors were treated with CP alone, or in combination with GdCl₃ every 3 days by intravenous (IV) injection. A significant level of macrophage depletion in the peripheral blood was achieved in animals treated with GdCl₃ (Suppl. Fig. S1B). Around day 25, the tumors in mice treated with both CP and GdCl₃ began to increase in size (Suppl. Fig. S1C). This suggests a critical correlation of the innate immunity in DNA alkylating damage-induced tumor clearance. Taken together, these results strongly indicate that CP-induced tumor cell death is not exclusively apoptotic, but includes necrosis, and can activate a pro-inflammatory response.

Discussion

A greater understanding of how tumor cells die in response to chemotherapy is likely to reveal new approaches to induce tumor cell death. In the present study, we utilize a genetically defined apoptosis-deficient cell system to establish a xenograft mouse tumor model. We examine the contribution of different forms of cell death in tumor regression induced by CP. We find that while apoptosis facilitates a more rapid tumor regression, it is dispensable for complete tumor regression.

Necrosis is often observed in solid tumor areas where vascularization lags behind tumor tissue growth. Using bax^−/−bak^−/− cells in which apoptosis is genetically inhibited, we find that sporadic necrosis is induced by CP treatment. Importantly, we notice features of necrosis not solely in the apoptosis-deficient bax^−/−bak^−/− cells, but also in the apoptosis-competent wild-type cells. This is different from spontaneous necrosis resulting from tumor overgrowth and lack of vascularization, that promotes chronic inflammation and stimulates further tumor growth (17). We demonstrate that HMGB1 is liberated from the nucleus, consistent with previous studies showing that HMGB1 released by necrotic cells confers pro-inflammatory responses (11). Thus, sporadic necrosis is not merely a death mechanism observed in the absence of apoptosis. Rather, it may play an important physiological role even in apoptosis-proficient cells by directly inducing tumor cell death and by activating the innate immunity.