Low-dose metronomic cyclophosphamide treatment mediates ... - Nature

Oncogene (2008) 27, 3729–3738

& 2008 Nature Publishing Group All rights reserved 0950-9232/08 $30.00 www.nature.com/onc

ORIGINAL ARTICLE

Low-dose metronomic cyclophosphamide treatment mediates ischemia-dependent K-ras mutation in colorectal carcinoma xenografts S Shahrzad1, S Shirasawa2, T Sasazuki3, JW Rak4 and BL Coomber1 1 Department of Biomedical Sciences, University of Guelph, Guelph, Ontario, Canada; 2Department of Cell Biology, School of Medicine, Fukuoka University, Fukuoka, Japan; 3Research Institute, International Medical Center of Japan, Tokyo, Japan and 4 Department of Pediatrics, Montreal Children’s Hospital, McGill University, Montreal, Quebec, Canada

Antiangiogenic therapies are promising approaches to cancer control, but the details of their effects on subsequent tumor progression are not fully understood. Such therapies have the potential to eventually generate extensive amounts of tumor ischemia, and we previously demonstrated that ischemic conditions induce K-ras mutations in cells with deficient mismatch repair (MMR) mechanisms. This suggested that similar effects on oncogene mutagenesis may accompany antiangiogenic therapy. To test this, MMR-deficient colorectal cancer cells (Dks-8) were xenografted into immune-deficient mice and treated with the antiangiogenic regimen of low-dose/ metronomic cyclophosphamide for 2 weeks followed by a 2-week recovery period without therapy. This treatment resulted in transient tumor growth inhibition, increased hypoxia, and decreased microvessel density, and cancer cells from treated tumors acquired activating mutations of the K-ras oncogene (K-rasG13D). In vitro exposure of Dks-8 cells to the active metabolite of cyclophosphamide (4-hydroxycyclophosphamide) had no effect on the K-ras status, indicating that there was no direct action of this alkylating agent on K-ras mutagenesis. In addition, cells sorted from hypoxic regions of Dks-8 tumors were enriched in K-rasG13D mutants. Collectively, our studies suggest that increases in tumor hypoxia induced by antiangiogenic treatment may lead to K-ras mutation and consequently tumor progression, especially in susceptible individuals. Oncogene (2008) 27, 3729–3738; doi:10.1038/sj.onc.1211031; published online 28 January 2008 Keywords: hypoxia; antiangiogenic therapy; K-ras mutation; tumor microenvironment; metronomic cyclophosphamide

Correspondence: Dr BL Coomber, Department of Biomedical Sciences, Ontario Veterinary College, University of Guelph, Guelph, Ontario, Canada N1G 2W1. E-mail: [email protected] Received 19 September 2007; revised 27 November 2007; accepted 6 December 2007; published online 28 January 2008

Introduction Angiogenesis, the development of new blood vessels from pre-existing capillary networks, plays an important role in the development and progression of cancer (Folkman, 1990). The induction of new blood vessels increases the opportunity for tumor cells to enter the circulation and metastasize (Liotta et al., 1974). The extent of angiogenesis is frequently related to cancer progression and in some cases may predict the probability of metastasis (Srivastava et al., 1988; Weidner et al., 1991). This linkage paved the way to clinical trials with antiangiogenic agents, which in some cases have dramatically changed the standards of care (Miller et al., 2005a, b; Motzer and Bukowski, 2006; Varker et al., 2007). In particular, bevacuzimab/avastin, a humanized monoclonal-neutralizing antibody directed against vascular endothelial growth factor has since 2004 become incorporated into the first-line therapies in metastatic colorectal carcinoma (CRC) (Hurwitz et al., 2004) and other malignancies (Gasparini et al., 2005). In addition to such targeted antiangiogenic agents designed to obliterate a well-defined molecular mechanism (for example, vascular endothelial growth factor/vascular endothelial growth factor receptor pathway), blood vessel inhibitory properties have also been uncovered in drugs with ostensibly unrelated activities, including cytotoxic anticancer chemotherapeutics (Browder et al., 2000; Klement et al., 2000). Studies with cyclophosphamide and several other agents revealed that when administered at low but frequent (metronomic) doses, they selectively target growing endothelial cells and block formation of tumor blood vessels (Browder et al., 2000; Klement et al., 2000). In spite of the unquestionable successes of some of these antiangiogenic strategies, the responses of individual patients vary markedly (Kerbel and Folkman, 2002; Yang et al., 2003; Miller et al., 2005a, b), in some instances amounting to overt resistance or lack of objective survival benefit (Jubb et al., 2006). While there may be several reasons why a subset of cancer patients exhibit resistance to antiangiogenic agents, it has recently become apparent that this property may reside in the genetic make up of cancer cells (Rak et al., 2006). For instance, oncogenic events such as loss of p53 tumor suppressor gene in CRC cells may render them

Antiangiogenic treatment and K-ras mutation S Shahrzad et al

3730

Results Accelerated tumor growth post-metronomic (LDM) therapy Dks-8 cells are K-raswt/null (Shirasawa et al., 1993), microsatellite unstable and deficient for the MMR enzyme MSH6 (Boyer et al., 1995). While originally reported to be non-tumorigenic (Shirasawa et al., 1993) Oncogene

700 Control 600

Tumor size (mm3)

refractory to ischemic conditions, that is, to the mechanism by which antiangiogenic agents exert their therapeutic influence (Yu et al., 2002). The natural gradients of oxygen, nutrients and pH caused by the limitations of tumor perfusion are thought to contribute to genetic alterations (Yuan and Glazer, 1998) and resistance to apoptosis (Galmarini and Galmarini, 2003). In addition to various selective mechanisms (Rak et al., 1996; Yu et al., 2001, 2002), we have recently reported that the ischemic tumor microenvironment may actually induce de novo K-ras mutations in CRC cells, notably those with a defect in DNA mismatch repair (MMR) mechanisms (Shahrzad et al., 2005). In this case, ischemic conditions downregulated one of the MMR proteins (MSH2) leading to an increase in frequency of K-ras mutations and eventual enrichment in mutant CRC cells (Shahrzad et al., 2005). Mutant K-ras occurs in approximately half of all CRC cases, 75–100% of pancreatic cancers and almost 50% of lung adenocarcinomas (Minamoto et al., 2002), and is associated with chemoresistance and radioresistance (Bernhard et al., 2000; Perona and Sańchez-Pe´rez, 2004), and the angiogenic phenotype (Rak et al., 1995). K-ras proto-oncogene becomes activated through a single-point mutation in one of the three critical codons 12, 13 or 61 (Bos, 1989). This renders the gene vulnerable to a defective MMR mechanism, which is involved in removing base mismatches that may occur spontaneously, or as a result of therapeutically induced DNA mutagenesis (Jiricny, 1998). We reasoned that an important context in which the aforementioned ischemic mechanisms may lead to increased K-ras mutagenesis is, in fact, during the course of antiangiogenic therapy. While moderate inhibition of aberrant growth of the tumor-associated vasculature may in some cases lead to transient improvement of tumor perfusion (Jain, 2005), a more complete therapeutic obliteration of this network implicitly brings about a state of ischemia, which may become lethal to some but not all cancer cells. It follows that such surviving cells may have pre-existing or acquired oncogenic alterations. Here we show that, indeed, xenografts of MMR-deficient CRC cells exposed to metronomic treatment with cyclophosphamide exhibit an increased frequency of K-ras mutations. These mutations were not recapitulated in vitro by treatment of these cells with an activated 4-hydroxy form of cyclophosphamide. We propose that the unappreciated consequences of antiangiogenic treatment may include an accelerated genetic progression of tumors with certain MMR defects.

Cyclophosphamide

*

500 400 300 200

2 weeks LDM CTX

*

100 0 60

70

80 90 100 Days after cancer cell injection

110

Figure 1 Effect of low-dose metronomic cyclophosphamide treatment on xenograft growth, showing mean values (s.e.) from Dks-8 xenografts. Significant differences between groups were seen on days 97 and 105 after injection of Dks-8 cells (N ¼ 10 for each group; *Po0.05), initially leading to tumor stasis (day 97; 1 week after cessation of treatment) and then to significant increase in tumor size in treated mice (day 105; that is, 2 weeks after cessation of treatment) compared to controls.

in our hands these cells demonstrate attenuated tumorigenicity (Shahrzad et al., 2005), and in this study formed palpable tumors in all mice 45 days after subcutaneous injection of 2 106 cells into the right flank of RAG-1null mice. When the average tumor size reached 35 mm3 mice were randomized into treatment and control groups and subjected to low-dose (metronomic) cyclophosphamide (LDM) for a period of 2 weeks followed by a 2-week break period. This study was performed in two similar trials; there were no statistically significant differences for the same treatment between trials, and so growth rates were calculated using pooled data from both trials. Growth rates of control and cyclophosphamide-treated tumors (N ¼ 10 for each group) were not significantly different during the 2-week period of LDM treatment, but there was a significant suppression of Dks-8 xenograft growth during the first week post-treatment (Po0.05; Figure 1). Surprisingly, the LDM-treated tumors displayed accelerated rates of growth approximately 1 week after cessation of drug exposure. Eventually, these tumors reached and then significantly exceeded the sizes of their control counterparts (Po0.05; Figure 1). LDM-induced tumor hypoxia To assess the effects of the LDM treatment at the microscopic level, we analyzed the extent of tumor hypoxia and microvessel density (MVD; number of blood vessels per mm2). Hypoxic areas, as determined by immunostaining for Hypoxyprobe-1 protein adducts, were wider and more extensive in the xenografts treated with LDM cyclophosphamide compared to controls (Figures 2a and b). LDM cyclophosphamide treatment increased the hypoxic area in Dks-8 xenografts by 170% on average, a difference that was statistically significant (Po0.05; Figure 2c). To confirm the antiangiogenic


3731

Figure 2 Increased tumor hypoxia and decreased blood vessel density was seen in low-dose metronomic (LDM) cyclophosphamidetreated xenografts collected 2 weeks post-treatment. Hypoxic areas (brown) were detected by anti-Hypoxyprobe-1 immunostaining; nuclei were counterstained with hematoxylin. (a) Representative control xenograft and (b) representative treated xenograft. Mean and s.e. for the average hypoxic area as a proportion of total cross-sectional area of Dks-8 xenografts (c). LDM cyclophosphamide treatment induced a significant increase in the proportion of tumor occupied by hypoxic cells (*Po0.05). Detection of blood vessel staining with anti-CD31 antibody (red) in a representative control xenograft (d) and a representative treated xenograft; (e) insets show 4,6-diamidino-2-phenylindole nuclear counterstain. (f) Mean (s.d.) for the average density of CD31-positive vessels per mm2 in treated and control Dks-8 xenografts. LDM cyclophosphamide induced a significant reduction in microvessel density compared to control tumors (*Po0.05).

action of LDM cyclophosphamide, MVD was scored in tumors by quantifying CD31 immunostained microvessels in control and treated tumors (Figures 2d and e). All microvessels were counted per tumor section, and there was a statistically significant 2.5-fold reduction in MVD in LDM cyclophosphamide-treated xenografts compared to control (Figure 2f; Po0.05). These differences in tumor hypoxia and MVD were still detected 2 weeks

after cessation of LDM cyclophosphamide treatment, highlighting the potent blockade of neovascularization induced by this approach. Changes in the K-ras status in LDM-treated tumors Despite the successful reduction in MVD and concomitant increased hypoxia seen with LDM cyclophosphamide, treated tumors displayed highly accelerated Oncogene


3732

Dks-8

DLD-1

C1 C2 C3 C4 T1 T2 T3 T4

water

growth compared to control tumors in the 2-week period after cessation of therapy. Since we had previously shown that K-raswt/null Dks-8 xenografts develop mutations in codon 13 of this oncogene (Shahrzad et al., 2005), we explored the possibility that this event may have been enhanced in treated tumors. As expected, the Dks-8 cell line was K-raswt/null at codon 13. K-ras mutation in Dks-8 tumors was barely detectable using the single-step PCR-restriction fragment length polymorphism (RFLP) method in all tumor pieces (data not shown), but clear evidence of K-ras mutation was apparent when the two-step PCR-RFLP was employed. Control tumors showed some mutation and all LDM cyclophosphamide-treated tumors showed considerable enrichment for mutation at codon 13 compared to control tumors (Figure 3). Densitometry of fragments of size 165 bp of xenografts of cyclophosphamide-treated mice (N ¼ 10) in comparison with that of the control xenografts (N ¼ 10) provided semiquantitative values for the presence of mutant K-ras codon 13. Xenografts of the treated mice had threefold stronger bands at 165 bp compared to the controls. Amplified product was extracted from the agarose gels and sequenced as described above. Only transition mutations of the second base of codon 13 (GGC - GAC; glycine - aspartic acid) were found in all DNA extracted from undigested bands of those cells. Thus, a rapid accumulation of mutant cells was facilitated in tumors exposed to LDM-induced hypoxia. Using similar PCR-RFLP approaches (modified from Toyooka et al., 2003; Shahrzad et al., 2005), we further examined K-ras mutations at codon 61 in the xenografts. We found no evidence for de novo development of mutation at this additional codon in either control or treatment group of the xenografted tumors (Supplementary Figure 1). To originally generate the K-raswt/null cell line Dks-8, DLD-1 mutant K-ras was selectively disrupted by homologous recombination, a method that interferes with the detection of de novo development of mutation at codon 12 (Shirasawa et al., 1993); thus we could not test the xenografts for changes at this codon.

165 bp 137 bp

Figure 3 Two-step PCR-restriction fragment length polymorphism analysis of DNA from Dks-8 xenografts of low-dose metronomic cyclophosphamide-treated mice (T) versus control mice (C) was used to detect K-ras mutations at codon 13. Wildtype DNA produces a 137 bp band, and mutant DNA produces a 165 bp product. All 10 treated tumors had enhanced mutant K-ras compared to the low but detectable levels of mutant oncogene as seen in control tumors (four examples are shown). Cultured DLD-1 cells were used as positive control for K-ras codon 13 mutation, and Dks-8 cells as positive control for K-ras wild-type DNA. Oncogene

Nevertheless, this study does not intend to exclude the possibility of additional K-ras mutations at codon 12 in MMR-deficient cell lines due to ischemic conditions of the tumor microenvironment. Isolation of tumor cells harboring mutant K-ras oncogene from hypoxic tumor regions We postulated that cells harboring mutant K-ras occupy hypoxic regions of tumors and thereby could be isolated by using previously described cell sorting according to the gradient of Hoechst 33342 (Herman et al., 1989; Yu et al., 2002). Indeed, the Hoechst 33342 injection highlighted a highly fluorescent population of cells immediately surrounding the perfused vasculature. In contrast, weakly fluorescent cells were mainly derived from tumor regions more distal to the perfused blood vessels (Herman et al., 1989; Yu et al., 2002) (Figure 4a). Immunofluorescence staining of the sorted cells for hypoxic protein adducts indicated that those displaying the dimmest Hoechst fluorescence contained abundant hypoxic protein adducts, while cells with the brightest Hoechst fluorescence were negative in this regard (Figures 4b–e). This confirms that there is an inverse relationship between Hoechst staining intensity and the hypoxic status of cells sorted from our Dks-8 xenografts. We employed the two-step PCR-RFLP for selective amplification of the K-ras codon 13 mutation (as described above) on DNA obtained from our sorted cells. Little or no K-ras mutation was found in the DNA obtained from ‘Bright’ (normoxic) cells, however, the ‘Dim’ (hypoxic) cells demonstrated a clear band (165 bp) for codon 13 mutation, represented by a 165 bp amplicon (Figure 4f). Sequencing of these products revealed them to be transition mutations of the second base of codon 13 (GGC - GAC; glycine - aspartic acid). The detection limit of this two-step PCR-RFLP assay is one in 2 106 cells (as determined by serial dilutions of K-ras mutant (DLD-1) in wild-type cells (Dks-8); data not shown (Shahrzad et al., 2005)). This confirms that the K-rasG13D mutation identified in DNA from whole xenografts primarily occurred in hypoxic cells. Unchanged K-ras status in tumor cells exposed to mafosfamide in vitro To investigate whether direct interaction of cyclophosphamide with the cancer cell genome, for instance, as an alkylating agent, could be the cause of increased K-ras mutations observed in the LDM-treated mice, we employed an activated form of cyclophosphamide, mafosfamide (Takamizawa et al., 1975; Yu and Waxman, 1996). Dks-8 cells were incubated with the drug in culture and assayed for the rate of K-ras mutations. While we found cytotoxic effects of the drug at high concentrations (0.1 and 0.2 mg ml1), none of the incubations (that is, at 0.002, 0.02, 0.1 and 0.2 mg ml1 final concentration for up to 2 weeks) resulted in the development of detectable K-ras mutations, even after 2 weeks of recovery (Figure 5). These results suggest that the increased K-ras mutation observed


3733

I) Mixture Hypoxic adducts

Hoechst

II) Bright

DLD-1

Dks-8

Dim #4

Bright #3

Non-sorted #3

Dim #2

Bright #2

Non-sorted #2

Dim #1

Bright #1

Non-sorted #1

Ladder

III) Dim

165 bp 137 bp

Figure 4 Hoechst fluorescence detection and hypoxia immunostaining for sorted cells from xenografted tumors. (a) Histogram of ultraviolet Hoechst 33342 intensity fluorescence-activated cell sorting analysis of cell suspension from a Dks-8 xenograft showing fluorescence intensity distribution of unsorted cell (‘Mixture’), Hoechst positive cells (‘Bright’) and Hoechst negative cells (‘Dim’). Hypoxic protein adducts were detected in cytospins of sorted cells by Hypoxyprobe-1 immunostaining, and the same field imaged for both adduct fluorescence (red) and Hoechst signal (blue). There were no detectable hypoxic adducts in Hoechst Bright cells (panels b and c), while Hoechst Dim cells showed robust staining for hypoxic adducts (panels d and e). (f) Two-step PCR-restriction fragment length polymorphism (RFLP) analysis for K-ras mutation. Fragments of size 165 bp represent mutated K-ras codon 13, and fragments of size 137 bp represent wild type. Two-step PCR-RFLP clearly shows mutant K-ras in cells derived from the Hoechst dim cells of several tumors, and little or no detectable mutation in Hoechst bright cells from the same sorted tumor. Cultured DLD-1 cells were used as positive control for K-ras codon 13 mutation, and Dks-8 cells as positive control for K-ras wild-type DNA.

upon exposure to LDM in vivo were not the result of a direct mutagenic or cytotoxic effect of cyclophosphamide, but rather originated from indirect effects of the drug, notably angiogenesis inhibition and concomitant hypoxia. Discussion Despite effective angiogenic blockade, increased tumor progression has previously been reported in xenograft

studies of experimental tumor growth (Klement et al., 2000; Emmenegger et al., 2006). These observations, together with the results of clinical trials using antiangiogenic agents (Yang et al., 2003; Hurwitz et al., 2004; Miller et al., 2005a, b; Motzer and Bukowski, 2006; Varker et al., 2007) suggest a more complex impact of this treatment modality than previously anticipated. In most current clinical studies, antiangiogenic agents are used as adjuvant therapy in combination with chemotherapy or sometimes chemoradiotherapy. However, the relationship between Oncogene


3734 DLD-1 cells

Dks-8 cells

0.2 mg/mL

0.1 mg/mL

0.02 mg/mL

0.002 mg/mL

Water

DLD-1 cells

mafosfamide

165 bp 137 bp

Figure 5 Two-step PCR-restriction fragment length polymorphism (RFLP) analysis of cultured Dks-8 cells exposed to the active metabolite of cyclophosphamide, mafosfamide, for 2 weeks in vitro at concentrations ranging from 0.0 to 0.2 mg ml1. Fragments of size 165 bp represent mutated K-ras codon, and fragments of size 137 bp represent wild type. The active metabolite of cyclophosphamide did not lead to K-ras mutation in cultured Dks-8 cells after 2 weeks treatment, or after an additional 2-week recovery time in vitro without mafosfamide (data not shown). DNA from DLD-1 cells was used as positive control for mutant K-ras.

angiogenesis and tumor progression emerges as increasingly multifaceted. For instance, there are conflicting data regarding the influence of extensive vascularization on tumor progression (Koukourakis et al., 2000), and intriguing evidence that antiangiogenic therapy may paradoxically enhance the efficacy of standard cytotoxic approaches (Kerbel, 2006). Tumor progression may be promoted by both chronic (diffusion-limited) as well as acute (fluctuating) tumor hypoxia (Janssen et al., 2005; Rofstad et al., 2007), and antiangiogenic therapy increases the proportion of cells undergoing both types of hypoxia (Rofstad et al., 2007). Currently, intervention with antiangiogenic agents is performed in relatively late stages of disease, where there are more angiogenic and survival pathways involved and manipulations of any single factor (such as inhibition of vascular endothelial growth factor) likely become less critical for cancer progression (Kerbel et al., 2001; Kerbel and Folkman, 2002; Franco et al., 2006). Interestingly, addition of bevacizumab to first-line chemotherapy in patients with metastatic colorectal cancer increased response rates and prolonged overall survival (Hurwitz et al., 2004). Nevertheless, the response to these drugs varied significantly among patients, and in all studies performed to date numerous patients underwent an antiangiogenic regimen without apparent significant benefit in survival or time to progression (Yang et al., 2003; Miller et al., 2005a, b; Motzer and Bukowski, 2006; Varker et al., 2007). Therefore, it is important that we understand the causes of this variable drug response and discover the molecular indicators of individual response to antiangiogenic therapy to select an optimal regimen for each individual. Redundancy of angiogenic molecules and/or transient upregulation of endothelial survival factors (Kerbel et al., 2001), differing vascular dependence of cancer cells (Yu et al., 2001), and differing cancer cell characteristics, especially with regard to oncogene activation and tumor suppressor inactivation status Oncogene

(Yu et al., 2002), may account for the major causes of differential response to antiangiogenic approaches seen to date. The latter issue has been the focus of our study. We previously demonstrated that tumor progression is related to selection of cells with reduced vascular dependence (Yu et al., 2001), which may be one of the important mechanisms involved in tumor progression in the face of effective angiogenic blockade. We had previously shown that although inactivation of mutant K-ras in MMR-defective colorectal Dks-8 cells led to significantly diminished in vivo growth capacity, tumors eventually developed with a latency period of more than 30 days, and subsequently grew slowly. We showed that some of those cells had acquired a secondary mutation in the remaining K-ras allele in the tumor microenvironment (Shahrzad et al., 2005). Similar to those results, in this study we also detected the induction of K-ras mutation due to ischemic conditions found in tumors, and further confirmed that these mutations are primarily arising in hypoxic cells. Treatment of xenografts with LDM cyclophosphamide led to increased hypoxia and decreased MVD, which could explain the considerable increase in K-ras mutation that we also detected in these treated tumors. K-ras mutations result in production of constitutively activated K-ras proteins and continual transmission of signals allowing improper cell proliferation (Bos, 1989), and hence accelerated tumor growth. The effect of LDM cyclophosphamide treatment on K-ras mutation supports our previous finding that antiangiogenic therapies may lead to selection of hypoxia-resistant tumor cells as one mechanism of ‘resistance’ (Yu et al., 2002). The system employed in this study uses a MMRdeficient CRC cell line that has been artificially engineered to disrupt a previously mutant K-ras allele (Shirasawa et al., 1993). Since K-ras mutation is known to lead to other alterations in cancer cell behavior contributing to tumor progression (notably uncontrolled proliferation and enhanced production of angiogenic factors; Bos, 1989; Rak et al., 1995), it is possible that CRC cells that were always wild type for K-ras may behave differently under similar experimental conditions. Interestingly, the parental cells for this Dks-8 line (DLD-1 cells, which have one allele mutated) also show increased K-ras mutation upon exposure to ischemic conditions in vitro and in vivo (Supplementary Figure 2 and Shahrzad et al., 2005), suggesting that the mutagenic capacity of these cells is similar to their Dks-8 counterparts. Investigation of the factors affecting de novo K-ras mutations in other CRC cell lines wild type for this oncogene, and of the influence ischemia may have on the evolution of K-ras mutations in CRC patients receiving antiangiogenic therapy, are areas of investigation currently underway in our laboratory. Hereditary nonpolyposis colorectal cancer syndrome is caused by a germline mutation in a DNA MMR gene, most commonly affecting the genes MLH1 (40%), MSH2 (50%) and/or MSH6 (10%) (Mitchell et al., 2002; Peltomaki and Vasen, 2004). Individuals with


3735

hereditary nonpolyposis colorectal cancer are at increased risk for several types of cancer, with the highest life-time risks for CRC (80%), endometrial cancer (40–60%), ovarian cancer (10–15%), cancer of the small intestine and upper urothelial cancer (Aarnio et al., 1999). In addition, somatic MMR defects occur in a subset of certain sporadic tumor types, for example, in 15–20% of gastrointestinal and endometrial cancer, which are mostly caused by somatic hypermethylation of the MLH1 promoter (Kane et al., 1997; Esteller et al., 1998). Cancers with defective MMR system may be considered susceptible to hypoxia-mediated K-ras mutation. Over 95% of hereditary nonpolyposis colorectal cancer tumors can be characterized by widespread microsatellite instability and 90% by loss of immunohistochemical expression of the MMR protein affected (Halvarsson et al., 2004). Accurately predicting the effects of antiangiogenic therapy on different cancers is important for the selection of optimal treatment strategies and regimens. Our results suggest that identifying individuals who are susceptible for development of hypoxia-mediated K-ras mutation might be a useful addition to the clinical diagnosis of colorectal cancer. Thus, patients presenting with tumors still wild type for K-ras but with impaired MMR may benefit from a different therapeutic regime than those whose tumors are already K-ras mutant. One possibility might be to combine antiangiogenic approaches with hypoxic cell cytotoxins, a new class of drugs that specifically target hypoxic cells (Brown, 2000; Brown and Wilson, 2004), to reduce the incidence of K-ras mutagenesis and its concomitant effects on tumor progression during antiangiogenic therapy.

Materials and methods Abbreviated Materials and methods are presented here; full detailed descriptions of additional methods and materials are located in Supplementary Materials. Cell lines The human DLD-1 CRC cell line (Dexter et al., 1981) harbors one mutant K-rasG13D allele, and was obtained from the American Type Culture Collection. DLD-1 K-rasG13D was selectively disrupted by homologous recombination to generate the K-raswt/null cell line Dks-8 (Shirasawa et al., 1993). Both DLD-1 and Dks-8 cells are microsatellite-unstable and deficient for the MMR enzyme MSH6 (Boyer et al., 1995). Cells were maintained in standard culture conditions: DMEM (Sigma, Oakville, ON, Canada) supplemented with 10% heatinactivated fetal bovine serum, 50 mg ml1 gentamicin, 1 mM sodium pyruvate and 1.25 mg ml1 amphotericin B as 5 ml ml1 of fungizone antimycotic (Invitrogen/Gibco, Burlington, ON, Canada). Cells were cultured at 37 1C in a humidified atmosphere containing 5% CO2. To exclude the possibility that these knockout cells were contaminated by rare spontaneous reversion K-ras mutants, a monoclonal Dks-8 cell population was obtained by dilution cloning, and the wild-type K-ras allele was confirmed by two-step PCR-RFLP. Each experiment in this study was performed with freshly expanded cells from cryopreserved stock of this clone, and K-ras status was confirmed as wild type by two-step PCR-RFLP of cells

used in each experiment (xenograft production and in vitro exposures). Cyclophosphamide trial All in vivo procedures were performed according to the guidelines and recommendations of the Canadian Council of Animal Care (CCAC), and approved by the University of Guelph local Animal Care Committee. The study was performed in two independent trials. We established xenograft tumors by injecting 2 106 Dks-8 cells into the right flank of RAG-1null mice (Chen et al., 1994). Tumor growth was assessed by caliper measurements every 3–4 days, and tumor size was calculated using the equation: volume ¼ length width2 0.5 (Shahrzad et al., 2005). When the average tumor size reached 35 mm3 (76 days) for this slow-growing, poorly tumorigenic cell line (Shahrzad et al., 2005), mice were randomly allocated into two groups: treatment and control. In the first trial there were four (N ¼ 4) mice in each group, and in the second trial there were six mice (N ¼ 6). For the treated group, mice were given drinking water with 200 mg ml1, a dose of approximately 30 mg kg1 day1, of cyclophosphamide (Sigma) for 2 weeks (water was replaced twice weekly; control mice received normal drinking water). After 2 weeks, the treatment was discontinued, and both groups of mice were kept for an additional 2 weeks under standard husbandry conditions. At the end of this second 2-week period, mice received intraperitoneal injection of Hypoxyprobe (150 mg kg1; Calbiochem/ VWR, Mississaga, ON, Canada) 1 h prior to euthanasia, and intravenous injection of Hoechst 3342 dye (0.25 mg per mouse; Sigma) 5 min prior to euthanasia. Mice were euthanized by CO2 asphyxia followed by cervical dislocation. Tumors were removed from the mice and cut into small pieces, which were snap-frozen in cryomatrix and stored at 80 1C for later sectioning or snap-frozen and stored at 80 1C for subsequent DNA isolation. Hypoxia measurements Cryosections of tumor xenografts were fixed in 4% paraformaldehyde, and immunostained for hypoxic adducts using anti-Hypoxyprobe-1 followed by biotin-conjugated goat antimouse for 3-amino-9-ethyl-carbazole histochemistry. Slides were coded and scored in a semiblinded manner, and the areas of hypoxia were scored and assessed by Optimas 6.2 image analysis software (MediaCybernetics, San Diego, CA, USA). The total hypoxic area for each section was divided by the cross-sectional area to obtain the proportion of hypoxia in each tumor. Blood vessel density The vessels in the tumor tissues were stained using an antibody to the endothelial marker CD31 (DeLisser et al., 1997) with rat anti-mouse CD31 primary antibody (PECAM-1) and donkey anti-rat Alexa Fluor 546-conjugated secondary antibody; nuclei were counterstained with 4,6-diamidino-2-phenylindole. Slides were coded and scored in a semiblinded manner and the total number of blood vessels for each section was divided by the cross-sectional area to obtain the MVD (number of blood vessels per mm2). Cell sorting We used our previously reported approach to sort cells from Dks-8 xenografts based on their proximity to perfused blood vessels (Herman et al., 1989; Yu et al., 2002). Briefly, the cells most proximal to perfused vessels were labeled by cardiac injection of 0.25 mg per mouse Hoechst 33342 dye, in 200 ml volume, 5 min before euthanasia. After enzymatic Oncogene


3736 disaggregation of each Dks-8 xenograft into a single-cell suspension, samples were fixed in 10% formalin and sorted by fluorescence-activated cell sorting (using a 350–70 nm UV channel with a FACSVantageSE instrument); five Dks-8 xenografts were used for this part of the study. Sorted cells were stained with anti-Hypoxyprobe-1 antibody and secondary anti-mouse Cy3. The slides were washed and mounted using a fluorescent mounting medium (DakoCytomation, Mississaga, ON, Canada). Remaining sorted cells were pelletted and their DNA was extracted for K-ras mutation analysis. In vitro exposure to cyclophosphamide (mafosfamide) To investigate whether the direct DNA alkylating action of cyclophosphamide plays a role in mutation of the K-ras gene, Dks-8 cells were exposed to a broad range of concentrations in vitro (0.002, 0.02, 0.1 and 0.2 mg ml1 final concentration). Cyclophosphamide must be metabolically converted into the active 4-hydroxylated intermediate by liver microsomal mono-oxygenases to exert its effects (Chang et al., 1993; Yu and Waxman, 1996; Ren et al., 1997). The 4-hydroperoxycyclophosphamide is a preactivated derivative of cyclophosphamide that hydrolyzes spontaneously in aqueous solution to form 4-hydroxycyclophosphamide. Mafosfamide (D-17272, NIOMECH, IIT GmbH, Bielefeld, NRW, Germany) is a cyclophosphamide derivative that rapidly generates 4-hydroperoxycyclophosphamide after aqueous dissolution (Takamizawa et al., 1975; Zon et al., 1984). Dks-8 cells were seeded into 100 mm plates at a density of 1 106 cells per plate, and incubated under standard conditions overnight. Thereafter, six plates were assigned to each of two groups—control, and drug treatment—and the cells received fresh media with mafosfamide every 3 days for 2 weeks. After mafosfamide treatment, three plates from each group were trypsinized, and cells were washed twice using a large volume of phosphate-buffered saline; washed cell pellets were kept at 80 1C until DNA isolation. The remaining plates were passaged and seeded at 1 106 cells per 100 mm plate, cultured under standard conditions in drug-free media for an additional 2 weeks, then harvested as described above. Each experiment was repeated at least twice; the results for the

samples exposed to each condition were compared with each other, and with those of simultaneous control (no mafosfamide) samples. K-ras mutation analysis DNA from DLD-1 cells, which are heterozygously mutated in codon 13 (Shirasawa et al., 1993) were used as a positive control for codon 13 K-ras mutation. DNA was extracted from cultured cells and xenografted tissues, and K-ras mutation status at codon 13 and 61 was determined as described (Toyooka et al., 2003; Shahrzad et al., 2005). For codon 13, if there is a mutation in either of the first two bases, the mutant PCR fragment will not be cut by XcmI, and will remain at its original size of 165 bp. To enhance the 165 bp products, which represent the mutated DNA, a second PCR-RFLP using the products of the first digestion as templates was performed. The same primers from the first PCR could thus more selectively amplify the undigested mutated DNA from the first digestion. Mutated K-ras was confirmed and identified by sequencing. K-ras codon 61 letter 1 and 2 mutations were determined by a two-step PCR-RFLP analysis, and letter 3 mutations were determined by a singlestep PCR-RFLP (modified from Toyooka et al., 2003; Shahrzad et al., 2005). Statistical analysis Quantitative data are presented as means of several independent measurements±s.d. Statistical analysis (t-test) was used to determine differences between means within experiments, with a two-tailed level of significance (Po0.05). Acknowledgements We thank Courtney Stone and Kanwal Minhas for technical support, Barb Mitchell for mouse husbandry and the members of the Coomber laboratory for their valuable suggestions. This work was supported by Canadian Cancer Society (National Cancer Institute of Canada Grant no. 016117 to BL Coomber and JW Rak).

References Aarnio M, Sankila R, Pukkala E, Salovaara R, Aaltonen LA, De la Chapelle A et al. (1999). Cancer risk in mutation carriers of DNA-mismatch-repair genes. Int J Cancer 81: 214–218. Bernhard EJ, Stanbridge EJ, Gupta S, Gupta AK, Soto D, Bakanauskas VJ et al. (2000). Direct evidence for the contribution of activated N-ras and K-ras oncogenes to increased intrinsic radiation resistance in human tumor cell lines. Cancer Res 60: 6597–6600. Bos JL. (1989). ras oncogenes in human cancer: a review. Cancer Res 49: 4682–4689. Boyer JC, Umar A, Risinger JI, Lipford JR, Kane M, Yin S et al. (1995). Microsatellite instability, mismatch repair deficiency, and genetic defects in human cancer cell lines. Cancer Res 55: 6063–6070. Browder T, Butterfield CE, Kra¨ling BM, Shi B, Marshall B, O’Reilly MS et al. (2000). Antiangiogenic scheduling of chemotherapy improves efficacy against experimental drug-resistant cancer. Cancer Res 60: 1878–1886. Brown JM. (2000). Hypoxic cytotoxic agents: a new approach to cancer chemotherapy. Drug Resist Updat 3: 7–13. Brown JM, Wilson WR. (2004). Exploiting tumour hypoxia in cancer treatment. Nat Rev Cancer 4: 437–447. Oncogene

Chang TKH, Weber GF, Crespi CL, Waxman DJ. (1993). Differential activation of cyclophosphamide and ifosfamide by cytochromes P-450 2B and 3A in human liver microsomes. Cancer Res 53: 5629–5637. Chen J, Shinkai Y, Young F, Alt FW. (1994). Probing immune functions in RAG-deficient mice. Curr Opin Immunol 6: 313–319. DeLisser HM, Christofidou-Solomidou M, Strieter RM, Burdick MD, Robinson CS, Wexler RS et al. (1997). Involvement of endothelial PECAM-1/CD31 in angiogenesis. Am J Pathol 151: 671–677. Dexter DL, Spremulli EN, Fligiel Z, Barbosa JA, Vogel R, VanVoorhees A et al. (1981). Heterogeneity of cancer cells from a single human colon carcinoma. Am J Med 1: 949–956. Emmenegger U, Morton GC, Francia G, Shaked Y, Franco M, Weinerman A et al. (2006). Low-dose metronomic daily cyclophosphamide and weekly tirapazamine: a well-tolerated combination regimen with enhanced efficacy that exploits tumor hypoxia. Cancer Res 66: 1664–1674. Esteller M, Levine R, Baylin SB, Ellenson LH, Herman JG. (1998). MLH1 promoter hypermethylation is associated with the microsatellite instability phenotype in sporadic endometrial carcinomas. Oncogene 17: 2413–2417.


3737 Folkman J. (1990). What is the evidence that tumors are angiogenesis dependent? J Natl Cancer Inst 82: 4–6. Franco M, Man S, Chen L, Emmenegger U, Shaked Y, Cheung AM et al. (2006). Targeted anti-vascular endothelial growth factor receptor-2 therapy leads to short-term and long-term impairment of vascular function and increase in tumor hypoxia. Cancer Res 66: 3639–3648. Galmarini CM, Galmarini FC. (2003). Multidrug resistance in cancer therapy: role of the microenvironment. Curr Opin Investig Drugs 4: 1416–1421. Gasparini G, Longo R, Toi M, Ferrara N. (2005). Angiogenic inhibitors: a new therapeutic strategy in oncology. Nat Clin Pract Oncol 2: 562–577. Halvarsson B, Lindblom A, Rambech E, Lagerstedt K, Nilbert M. (2004). Microsatellite instability analysis and/or immunostaining for the diagnosis of hereditary nonpolyposis colorectal cancer? Virchows Arch 444: 135–141. Herman TS, Teicher BA, Holden SA, Collins LS. (1989). Interaction of hyperthermia and radiation in murine cells: hypoxia and acidosis in vitro, tumor subpopulations in vivo. Cancer Res 49: 3338–3343. Hurwitz H, Fehrenbacher L, Novotny W, Cartwright T, Hainsworth J, Heim W et al. (2004). Bevacizumab plus irinotecan, fluorouracil, and leucovorin for metastatic colorectal cancer. N Engl J Med 350: 2335–2342. Jain RK. (2005). Normalization of tumor vasculature: an emerging concept in antiangiogenic therapy. Science 307: 58–62. Janssen HL, Haustermans KM, Balm AJ, Begg AC. (2005). Hypoxia in head and neck cancer: how much, how important? Head Neck 27: 622–638. Jiricny J. (1998). Replication errors: cha(lle)nging the genome. EMBO J 17: 6427–6436. Jubb AM, Oates AJ, Holden S, Koeppen H. (2006). Predicting benefit from anti-angiogenic agents in malignancy. Nat Rev Cancer 6: 626–635. Kane MF, Loda M, Gaida GM, Lipman J, Mishra R, Goldman H et al. (1997). Methylation of the hMLH1 promoter correlates with lack of expression of hMLH1 in sporadic colon tumors and mismatch repair-defective human tumor cell lines. Cancer Res 57: 808–811. Kerbel R, Folkman J. (2002). Clinical translation of angiogenesis inhibitors. Nat Rev Cancer 2: 727–739. Kerbel RS, Yu J, Tran J, Man S, Viloria-Petit A, Klement G et al. (2001). Possible mechanisms of acquired resistance to anti-angiogenic drugs: implications for the use of combination therapy approaches. Cancer Metastasis Rev 20: 79–86. Kerbel RS. (2006). Antiangiogenic therapy: a universal chemosensitization strategy for cancer? Science 312: 1287–1295. Klement G, Baruchel S, Rak J, Man S, Clark K, Hicklin DJ et al. (2000). Continuous low-dose therapy with vinblastine and VEGF receptor-2 antibody induces sustained tumor regression without overt toxicity. J Clin Invest 105: R15–R24. Koukourakis MI, Giatromanolaki A, Sivridis E, Fezoulidis I. (2000). Cancer vascularization: implications in radiotherapy? Int J Radiat Oncol Biol Phys 48: 545–553. Liotta LA, Kleinerman J, Saidel GM. (1974). Quantitative relationships of intravascular tumor cells, tumor vessels, and pulmonary metastases following tumor implantation. Cancer Res 34: 997–1004. Miller KD, Chap LI, Holmes FA, Cobleigh MA, Marcom PK, Fehrenbacher L et al. (2005a). Randomized phase III trial of capecitabine compared with bevacizumab plus capecitabine in patients with previously treated metastatic breast cancer. J Clin Oncol 23: 792–799. Miller KD, Trigo JM, Wheeler C, Barge A, Rowbottom J, Sledge G et al. (2005b). A multicenter phase II trial of ZD6474, a vascular endothelial growth factor receptor-2 and epidermal growth factor receptor tyrosine kinase inhibitor, in patients with previously treated metastatic breast cancer. Clin Cancer Res 11: 3369–3376.

Minamoto T, Ougolkov AV, Mai M. (2002). Detection of oncogenes in the diagnosis of cancers with active oncogenic signaling. Expert Rev Mol Diagn 2: 565–575. Mitchell RJ, Farrington SM, Dunlop MG, Campbell H. (2002). Mismatch repair genes hMLH1 and hMSH2 and colorectal cancer: a HuGE review. Am J Epidemiol 156: 885–902. Motzer RJ, Bukowski RM. (2006). Targeted therapy for metastatic renal cell carcinoma. J Clin Oncol 24: 5601–5608. Peltomaki P, Vasen H. (2004). Mutations associated with HNPCC predisposition—update of ICG-HNPCC/INSiGHT mutation database. Dis Markers 20: 269–276. Perona R, Sańchez-Pe´rez I. (2004). Control of oncogenesis and cancer therapy resistance. Br J Cancer 90: 573–577. Rak J, Coomber B, Yu JL. (2006). Oncogenes and tumor suppressor genes in therapeutic resistance: the role of evolving interrelationship between cancer cells and their host tissues. In: Tiecher B (ed). Cancer Drug Discovery and Development: Cancer Drug Resistance. Humana Press Inc.: NJ, USA, pp 67–103. Rak J, Filmus J, Kerbel RS. (1996). Reciprocal paracrine interactions between tumour cells and endothelial cells: the ‘angiogenesis progression’ hypothesis. Eur J Cancer 32A: 2438–2450. Rak J, Mitsuhashi Y, Bayko L, Filmus J, Shirasawa S, Sasazuki T et al. (1995). Mutant ras oncogenes upregulate VEGF/VPF expression: implications for induction and inhibition of tumor angiogenesis. Cancer Res 55: 4575–4580. Ren S, Yang JS, Kalhorn TF, Slattery JT. (1997). Oxidation of cyclophosphamide to 4-hydroxycyclophosphamide and deschloroethylcyclophosphamide in human liver microsomes. Cancer Res 57: 4229–4235. Rofstad EK, Galappathi K, Mathiesen B, Ruud EB. (2007). Fluctuating and diffusion-limited hypoxia in hypoxia-induced metastasis. Clin Cancer Res 13: 1971–1978. Shahrzad S, Quayle L, Stone C, Plumb C, Shirasawa S, Rak JW et al. (2005). Ischemia-induced K-ras mutations in human colorectal cancer cells: role of microenvironmental regulation of MSH2 expression. Cancer Res 65: 8134–8141. Shirasawa S, Furuse M, Yokoyama N, Sasazuki T. (1993). Altered growth of human colon cancer cell lines disrupted at activated Kiras. Science 260: 85–88. Srivastava A, Laidler P, Davies RP, Horgan K, Hughes LE. (1988). The prognostic significance of tumor vascularity in intermediatethickness (0.76–4.0 mm thick) skin melanoma. A quantitative histologic study. Am J Pathol 133: 419–423. Takamizawa A, Matsumoto S, Iwata T, Tochino Y, Katagiri K, Yamaguchi K. (1975). Studies on cyclophosphamide metabolites and their related compounds. 2. Preparation of an active species of cyclophosphamide and related compounds. J Med Chem 18: 376–383. Toyooka S, Tsukuda K, Ouchida M, Tanino M, Inaki Y, Kobayashi K. (2003). Detection of codon 61 point mutations of the K-ras gene in lung and colorectal cancers by enriched PCR. Oncol Rep 10: 1455–1459. Varker KA, Biber JE, Kefauver C, Jensen R, Lehman A, Young D et al. (2007). Randomized phase 2 trial of bevacizumab with or without daily low-dose interferon alfa-2b in metastatic malignant melanoma. Ann Surg Oncol 14: 2367–2376. Weidner N, Semple JP, Welch WR, Folkman J. (1991). Tumor angiogenesis and metastasis—correlation in invasive breast carcinoma. N Engl J Med 324: 1–8. Yang JC, Haworth L, Sherry RM, Hwu P, Schwartzentruber DJ, Topalian SL et al. (2003). A randomized trial of bevacizumab, an anti-vascular endothelial growth factor antibody, for metastatic renal cancer. N Engl J Med 349: 427–434. Yu JL, Rak JW, Carmeliet P, Nagy A, Kerbel RS, Coomber BL. (2001). Heterogeneous vascular dependence of tumor cell populations. Am J Pathol 158: 1325–1334. Yu JL, Rak JW, Coomber BL, Hicklin DJ, Kerbel RS. (2002). Effect of p53 status on tumor response to antiangiogenic therapy. Science 295: 1526–1528. Oncogene


3738 Yu L, Waxman J. (1996). Role of cytochrome P450 in oxaza phosphorine metabolism. Deactivation via N-dechloroethylation and activation via 4-hydroxylation catalyzed by distinct subsets of rat liver cytochromes P450. Drug Metab Dispos 24: 1254–1262. Yuan J, Glazer PM. (1998). Mutagenesis induced by the tumor microenvironment. Mutat Res 400: 439–446.

Zon G, Ludeman SM, Brandt JA, Boyd VL, Ozkan G, Egan W et al. (1984). NMR spectroscopic studies of intermediary metabolites of cyclophosphamide. A comprehensive kinetic analysis of the interconversion of cis- and trans-4-hydroxycyclophosphamide with aldophosphamide, and the concomitant partitioning of aldophosphamide between irreversible fragmentation and reversible conjugation pathways. J Med Chem 27: 466–485.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc).

Oncogene