Breast cancer is the most commonly diagnosed cancer in women, with 212,920 new cases estimated in the United States for 2006.1 Although many of these cases are fortunately detected at an early stage and can be cured, presentation or relapse with metastases is quite common and breast cancer is still the second most common cause of cancer related death in women with 40,970 cases predicted for 2006. Despite progress in chemotherapy, radiation therapy and monoclonal antibodies, metastatic breast cancer remains an essentially incurable disease and therefore new treatment alternatives are needed.
Oncolytic virotherapy utilizes the inherent ability of many viruses to lyse the cells that support their lytic lifecycle.2 New viruses released from dying cells spread to neighboring cells, thereby potentiating and propagating the therapeutic effect.3 Recombinant conditionally oncolytic adenoviruses are engineered to take advantage of the similarity between requirements of carcinogenesis and DNA virus replication, that is, the shutdown of major growth control pathways including p53/ARF, p16/Rb, Ras/MAPK and others.4, 5 Adenoviruses synthesize proteins that result in locking of the cell cycle into S-phase for effective viral replication.6, 7 Adenoviral E1A is the first protein to be synthesized upon infection and is normally required to release the transcription factor E2F-1 by binding to retinoblastoma (Rb) protein.4 Incorporation of a 24 bp deletion in the constant region 2 (base pairs 920–944) of adenoviral E1A results in dysfunctional E1A that can no longer bind Rb, rendering such viruses replication-incompetent in quiescent normal cells.6, 7 In the majority of human tumors, however, the Rb/p16 pathway is mutated allowing the viruses to replicate unhindered.8, 9
Efficient oncolysis by replicating adenoviruses is partly dependent on the expression of the primary coxsackie-adenovirus receptor (CAR) on cancer cells.10, 11, 12, 13 CAR expression has been reported to be variable and frequently low on various tumor types, including breast cancer.14, 15 CAR independent entry pathways can bypass this deficiency, and retargeting strategies that redirect viruses from their primary receptor to receptors preferentially expressed on cancer cells have been successfully utilized.11, 16 Transductional targeting can be achieved either by genetically modifying the fiber or other capsid proteins17, 18 or by using bispecific antibodies and fusion proteins that recognize both a specific viral structure and a specific receptor.13, 19, 20 Several studies have shown genetic incorporation of targeting motifs to be feasible in an adenoviral context.17, 21, 22 Arg-Gly-Asp (RGD) and polylysine motifs are interesting because their receptors, αvβ-integrins and polyanionic heparan sulfate proteoglycans, respectively, are abundantly expressed on tumor cells and roles in cancer progression have also been suggested.23, 24 Replacing the serotype 5 fiber knob with the serotype 3 knob results in CAR-independent infection, as Ad3 has a receptor distinct from CAR, that is not downregulated during carcinogenesis.21, 25, 26
Tissue macrophages, especially hepatic Kupffer cells, present an obstacle for systemic therapy with adenoviruses, because they clear virus rapidly from blood and unfavorable bioavailability results.27 Various chemical substances such as gadolinium chloride, polyinosinic acid, liposome encapsulated clodronate and empty liposomes have been studied for blocking of Kupffer cells.27, 28, 29 Theoretically, Kupffer cell inactivation might lead to improved bioavailability and enhanced infectivity of tumor metastases, which could enhance oncolysis following systemic treatment.
It is increasingly accepted that the environment of a tumor has a major effect on its behavior, including response to therapeutics.30 Therefore, subcutaneous tumors may not be optimal models for metastatic disease. Here, we developed two models that may more closely resemble the clinical disease. Hormone refractory breast cancer cells were grown in mammary fat pads of mice to model locally advanced disease. This model results in early lymph node metastases and subsequent lung and bone metastases. Further, we developed a highly aggressive lung metastatic model of hormone refractory breast cancer. Importantly, both models allow noninvasive green fluorescent protein (GFP) imaging of the tumor burden, which not only reduces the number of mice required for experiments but also facilitates repeated imaging of the same animals for detection of patterns of response and relapse.
Material and methods
GFP expressing human breast cancer cell line M4A4-LM3 is an aggressive metastatic, hormone refractory subline derived from MDA-MB-435.31 Human transformed embryonic kidney cell line 293 and hormone refractory breast cancer cell lines MDA-MB-435, ZR-75-1, CAMA-1, MDA-MB-436 and MCF-7 were obtained from the American Type Culture Collection (ATCC, Manassas, VA). All cell lines were maintained in the recommended conditions.
Replicating and nonreplicating viruses (Table I) were propagated on A549 and 293 cells, respectively. All viruses were purified and titered as described.18
Cells in triplicates were infected for 30 min in room temperature with virus diluted in 200 μl of growth medium with 2% fetal calf serum (FCS). Cells were washed once and complete growth medium was added. Cells were incubated for 24 hr at 37°C, after which luciferase assay was performed (Luciferase Assay System, Promega, Madison, WI); luciferase expression was expressed as relative light units (RLU). Background luciferase activities were subtracted from the data and results were calculated as RLU relative to control virus.
In vitro cytotoxicity assay
Cells in quadruplicate were infected with 0.1, 1 or 10 viral particles (VP)/cell for 1 hr at 37°C in 50 μl of growth medium with 2% FCS, and incubated thereafter with growth medium with 5% FCS. Cell viability was measured using the CellTiter 96 AQUAEOUS One Solution Cell Proliferation Assay (MTS assay; Promega) on day 7 (MCF-7, ZR-75-1), day 11 (MDA-MB-436, CAMA-1, M4A4-LM3) or day 12 (MDA-MB-435). The length of the experiment for each cell line was internally timed for maximum dynamic range with the positive and negative controls.
Orthotopic breast cancer model in nude mice
All animal experiments were performed according to the rules by the Provincial Government of Southern Finland. Pathogen free 3–4 weeks old female NMRI nude mice were purchased from Taconic (Ejby, Denmark) and quarantined for 2 weeks. Mice were injected orthotopically into the left and right uppermost mammary fat pad with 2 × 106 M4A4-LM3 cells and a tumor was allowed to develop. To study the metastasis of M4A4-LM3 cells into axillary lymph nodes, mice were killed, lymph nodes were dissected from mice and analyzed for GFP expression with the Xenogen IVIS 100 imaging system (Xenogen Corp., Alameda, CA) as recommended by the manufacturer.
Nude mice (n = 6) bearing M4A4-LM3 mammary fat pad tumors ∼1 cm in diameter were injected intravenously with viruses. Forty-eight hours later, mice were injected intravenously with 0.15 mg/g of body weight D-luciferin. Ten minutes later mice were killed and tumors imaged for luciferase expression with IVIS imaging system (Xenogen). Tumors and organs were collected and homogenized in Cell Culture Lysis Buffer (Luciferase Assay System, Promega) and luciferase activity was analyzed as before. RLU values were normalized to total protein content and results are presented as relative to virus only group, which have been given a value of 1.
Systemic treatment model of breast cancer lung metastasis
Nude mice (n = 10) were anesthetized using medetomidine and ketamine (Domitor® and Ketalar®, Orion Pharma, Espoo, Finland) and 2 × 106 GFP positive M4A4-LM3 cells were inoculated with a midaxillary percutaneous injection into the left lung (day 0). Starting on day 5, mice were injected intravenously with 2 × 1010 VP of Ad5.LacZ (a replication deficient control virus), Ad5-Δ24E3, Ad5.pK7-Δ24, Ad5/3-Δ24, Ad5-Δ24RGD or no virus. Treatments were repeated on days 10, 15 and 20.
In vivo imaging
Mice were imaged 10, 15 and 20 days after the injection of cells with the IVIS 100 imaging system (Xenogen). Images were obtained with a charge-coupled device camera cooled to −120°C with the field of view set to 15 cm. Fluorescence images were obtained with an open GFP excitation filter and a low level intensity, binning set to 6 and the exposure time to 1 sec. The photographic images used a 0.2 sec exposure time, 8 f/stop, binning set to medium and open filter. The fluorescence and grayscale images were overlaid using LivingImage software (Xenogen) and images were analyzed for GFP expression using Igor image analysis software Version 2.50 (Wavemetrics, Lake Oswego, OR). Images were displayed as pseudocolor images, where the data values (photons) are made to correspond to various colors (red for highest intensity, blue for lowest). For photon emission analysis fixed size regions of interest were drawn around the thorax and chosen threshold values were set for minimum and maximum emission measured. Values were normalized to values obtained from a region of interest with no tumor. Usually, an earlobe was used. Mean photon count of each group on day 10 is set as 100 % and mean photon count on days 15 and 20 is displayed relative to counts on day 10.
Kupffer cell inactivation in vivo
Nude mice (n = 4) bearing mammary fat pad tumors ∼1 cm in diameter were injected intravenously with either liposomes, polyinosinic acid (poly(I)) or gadolinium chloride (GdCl3). Liposomes (8 mg/mouse, prepared as described)28 were injected 2 hr prior, GdCl3 (0.3 mg/mouse) 24 hr prior35 and poly(I) (5 μg/g body weight)29 5 min prior to viral injection. These conditions attempted to exactly mimic previous successful reports. 5 × 1010 VP of Ad5lucI was injected intravenously and 48 hr later mice were killed and tumors collected. Luciferase expression from homogenized tissue was analyzed as RLU (see earlier) and results were calculated relative to results from the virus only group.
The effect of poly(I) on survival was studied in combination with Ad5.pK7-Δ24 treatment in the lung metastasis model. Ten nude mice were injected intravenously with either Ad5.pK7-Δ24 alone or with poly(I). Mice received the treatments on days 5, 10, 15 and 20 after injection of cells. In addition, 4 mice in each group received 1 × 109 VP of luciferase expressing Ad5.pK7 (GL) with Ad5.pK7-Δ24 to enable imaging of the fate of viruses after intravenous injection. Mice were injected intravenously with 0.15 mg/g of body weight D-Luciferin (Promega) and imaged for luciferase expression with the IVIS imaging system (Xenogen).
Gene transfer efficacy in vitro and in vivo and photon emission data was analyzed with 2-tailed Student's t test. Oncolytic cell killing efficacy data was analyzed by one-way ANOVA. A p value < 0.05 was considered statistically significant. Survival data was plotted into a Kaplan-Meier curve and groups were compared pair-wise with log-rank test. All analyses were done using SPSS 11.5.
Capsid modifications increase gene transfer to breast cancer cell lines
To study the effect of capsid modifications on gene transfer efficacy, we analyzed reporter gene activity mediated by Ad5/3lucI, Ad5lucRGD (contain luciferase), Ad5.pK7 (GL) and Ad5.RGD.pK7 (GL) (contain both GFP and luciferase as transgene) compared to control viruses Ad5lucI or Ad5 (GL) on 6 breast cancer cell lines (Fig. 1a). Ad5LucI is isogenic to Ad5/3lucI and Ad5lucRGD while Ad5 (GL) is isogenic to Ad5.pK7 (GL) and Ad5.RGD.pK7 (GL). Gene transfer with Ad5.pK7 (GL) was significantly and sometimes dramatically increased in M4A4-LM3, MDA-MB-435, MDA-MB-436, MCF-7, CAMA-1 and ZR-75-1 versus the respective Ad5 control. With Ad5/3lucI and Ad5lucRGD, similar results were seen in the same cell lines. Ad5.RGD.pK7 was generally less effective. In noncancer, CAR-positive 293 cells, differences in gene transfer efficacy compared to the control virus were mostly less than 3-fold (data not shown).
Capsid modifications increase gene transfer to clinical breast cancer samples
We analyzed the same panel of viruses on 3 clinical breast cancer samples fresh from patients with metastatic or advanced disease (Fig. 1b). With Ad5.pK7 (GL), gene transfer was increased up to 4-fold, over 6-fold and up to 2-fold in samples 1, 2 and 3, respectively. With Ad5/3lucI, gene transfer efficacy was increased up to 2-fold and 11-fold in samples 2 and 3, respectively, but in sample 1 there was no significant difference compared to the control virus. With Ad5.RGD.pK7 (GL) gene transfer efficacy was increased up to 2-fold in sample 2. With Ad5lucRGD, gene transfer efficacy was not increased in any of the samples.
Capsid-modified viruses display efficient oncolysis of breast cancer cells in vitro
To evaluate the potency of oncolytic viruses with the same capsid modifications, a cell viability assay was performed. In all but one cell line (MDA-MB-436), significantly more effective cell killing was observed with 1 or more of the capsid-modified oncolytic viruses compared to the /isogenic control virus with a wild type fiber (Fig. 2). In CAMA-1, MCF-7, MDA-MB-435, M4A4-LM3 and ZR-75-1, Ad5/3-Δ24 was more effective than Ad5-Δ24E3. In MCF-7, MDA-MB-435, M4A4-LM3 and ZR-75-1, Ad5.pK7-Δ24 was more effective than Ad5-Δ24E3; and in MCF-7, MDA-MB-435 and M4A4-LM3, Ad5-Δ24RGD was significantly more effective than Ad5-Δ24E3.
Intravenous biodistribution in tumor-bearing mice was analyzed in a model of locally advanced breast cancer where M4A4-LM3 cells were injected into two uppermost mammary fat pads (Figs. 3a and 3b). Following intravenous virus injection, most of the bioluminescence was observed in the liver, but tumor transduction could also be demonstrated (Fig. 3c, 4 uppermost tumors). In this model, tumor cells spread rapidly into axillary lymph nodes, which can be detected with GFP emission and due to the enlarged size of the nodes (Fig. 3d). To study biodistribution on a tissue level, homogenized organs were analyzed for luciferase expression (Fig. 3e). There were no significant differences in biodistribution between unmodified virus and any of the capsid-modified viruses. Tumor-to-liver gene delivery ratios were also analyzed, and despite a trend favoring some of the capsid-modified viruses, there were no significant differences (Fig. 3f).
Effect of polyinosinic acid, gadolinium chloride or liposomes on transductional efficacy and survival in vivo
Kupffer cells may be responsible for the majority of clearance of adenovirus from blood.27 To test if substances reported to block Kupffer cell action could enhance gene transfer efficacy in vivo, we injected poly(I), GdCl3 or liposomes intravenously into mice bearing mammary fat pad tumors, followed by intravenous injection of adenovirus. Poly(I) resulted in a significant increase in transgene expression in tumors (p < 0.05) (Fig. 4a).
To evaluate the effect of poly(I) on the fate of intravenously injected virus, we co-injected luciferase expressing virus with an oncolytic virus, and after 72 hr imaged living mice for photon emission. Without poly(I), strong photon emission was seen in the liver, but no photon emission could be detected when poly(I) was used (Fig. 4b).
To test whether improved tumor transduction and reduced liver transduction would translate into improved survival, we combined pre-injection of poly(I) to injection of Ad5.pK7-Δ24 in a systemic treatment model of advanced lung metastatic hormone refractory breast cancer. Surprisingly, poly(I) resulted in reduced survival versus Ad5.pK7-Δ24 alone (p < 0.01) (95% CI) (Fig. 4c). Median survival in the combination treatment group was 19 days and in the single treatment group 32 days.
Capsid-modified viruses in a systemic treatment model of orthotopic hormone refractory metastatic breast cancer
To test the efficacy of the capsid-modified oncolytic viruses in a systemic treatment model of disseminated breast cancer, we developed a new model featuring GFP positive hormone refractory M4A4-LM3 cells growing in the left lung. A large metastatic tumor initially forms in the injected lung and subsequently disseminates to the other lung, the heart and regional lymph nodes of the mediastinum, usually leading to severe weight loss and death of the mouse within 3–4 weeks. We treated the mice with intravenous virus injections on days 5, 10, 15 and 20 after the injection of cells. Treatment with capsid-modified oncolytic viruses resulted in significant survival benefit when compared to either mock (p < 0.01, p < 0.01 and p < 0.001 with Ad 5/3-Δ24, Ad5.pK7-Δ24 and Ad5-Δ24RGD, respectively) or Ad5.LacZ (all significant) (Fig. 5a). Treatment with Ad5-Δ24RGD resulted in significant survival benefit in comparison to Ad5-Δ24E3 treatment (p = 0.0136). Ad5/3-Δ24 and Ad5.pK7-Δ24 were equally effective to this highly potent positive control. The median survival of Ad5-Δ24RGD, Ad5.pK7-Δ24, Ad5/3-Δ24 and Ad5-Δ24E3 treated mice was 39, 32, 31 and 28 days, respectively.
Two of the Ad5-Δ24RGD treated mice were alive when the experiment was terminated on day 60. Upon autopsy, both of them had residual tumor remaining. Despite differences in survival, no significant differences in mean fluorescence emission were detected (Fig. 5b). However, there seemed to be a trend for low fluorescence on day 10, or decreasing fluorescence between days 10 and 15 in responding mice (Fig. 6). In nonresponding or untreated mice, there was a continuous increase in fluorescence which seemed to correlate with short survival.
Adenoviruses are safe for treatment of human cancer, regardless of the route of administration.36 Although efficacy has been seen following local administration,37, 38 nonreplicating adenoviral agents have not yet demonstrated systemic therapeutic efficacy. This is in contrast to preclinical data and somewhat disappointing, as most end-stage cancer patients die of systemic disease, which may necessitate systemic delivery. The less than convincing clinical results are a direct consequence of the negligible number of tumor cells that are transduced following intravenous administration. An advantage gained by using replication competent oncolytic viruses is the potential for local amplification within the tumor, following initial transduction of a small number of cells. Dissemination of virus to distant or nearby sites may also occur. Importantly, systemic delivery of large doses of oncolytic adenovirus has been remarkably safe, but efficacy or tumor transduction have not been demonstrated yet.39 This may be due to the low oncolytic potency of the only agent tested so far. Lack of the primary adenovirus receptor CAR on clinical tumors,40, 41 or binding of adenovirus to blood factors42 may also contribute to the lack of positive correlative data.
With regard to breast cancer specifically, deficiency or abnormal spatial expression of CAR may be a frequent phenomenon.43 Ineffective CAR-mediated gene delivery would not only hamper initial vascular infection of tumor cells, but would also affect lateral dispersion in the tumor mass. To avoid this, various CAR independent entry pathways have been exploited.19, 44 Here, we utilized 4 different capsid modifications. An RGD-4C motif in the fiber HI loop, polylysine in the fiber C-terminus, both combined, or a chimeric Ad5/3 fiber to study whether they would result in enhanced gene transfer and subsequently enhanced oncolysis in the context of disseminated hormone refractory breast cancer. Indeed, we saw enhanced gene transfer in various breast cancer cell lines tested in vitro (Fig. 1a). Enhancement of transductional efficacy with each capsid modification was cell line specific, implying varying receptor levels on different cell lines. Because gene delivery is the most relevant endpoint here, we elected to study functional transgene delivery directly instead of receptor expression. As cell lines cultured for a large number of passages may not resemble the initial clinical tumor very closely, we felt it was important that our cell line data was confirmed with three unpassaged clinical specimens fresh from patients (Fig. 1b). The level of infectivity enhancement was less dramatic with the clinical samples, which may be due to infiltrating stromal and inflammatory cells. Also analysis of clinical specimens is more complex than cell lines, leading to less stringent conditions which can dilute differences between viruses.
We saw no significant differences in vivo in gene transfer efficacy with capsid-modified viruses, neither into orthotopic mammary fat pad tumors nor into any of the normal organs tested. Serotype 5 adenovirus has been shown very safe in clinical cancer trials,45 and therefore a similar biodistribution seen with the capsid-modified agents is promising with regard to tolerability in humans. Moreover, capsid-modified viruses showed a promising but statistically insignificant trend for a favorable tumor to liver transduction profile when compared to virus with an unmodified capsid, suggesting the potential for a good systemic efficacy versus toxicity.
To study the oncolytic potency of capsid-modified viruses in circumstances that mimic the clinical situation as closely as possible, we developed a murine model of advanced hormone refractory breast cancer metastatic to the lungs. GFP positive M4A4-LM3 cells were utilized to allow noninvasive imaging of disease progression (Figs. 6a–6f). Injection into the left lung causes severe lung metastatic disease with subsequent dissemination into the mediastinum and regional lymph nodes, resulting in death in just a few weeks unless treated. Intravenous injections of Ad5-Δ24RGD resulted in significantly enhanced survival versus a highly potent positive control virus Ad5-Δ24E3 (Fig. 5). Ad5-Δ24E3 contains the complete E3 region and is therefore even more potent that Ad5-Δ24 and dl922-947, both of which have been suggested for human trials due to their high efficacy and a promising preclinical toxicity profile.6, 7 Also Ad5.pK7-Δ24 and Ad5/3-Δ24 resulted in enhanced survival in comparison to mock treated or Ad5.LacZ treated mice, but there was no significant difference in comparison to Ad5-Δ24E3.
With regard to mechanisms for the high efficacy of Ad5-Δ24RGD, αvβ3 integrin upregulation and activation has been reported to contribute to metastasis in human breast cancer.46 It has also been suggested that αvβ3 integrins are present in a nonactivated state in the parental MDA-MB-435 breast cancer cell line, but in an activated state in various in vivo selected cell variants,46 such as M4A4-LM3.31 This might help explain the superiority of Ad5-Δ24RGD over Ad5-Δ24E3 in vivo.
RGD-4C was initially described as a peptide with tropism to breast cancer vasculature.47 Intravenously administered virus would meet the tumor vasculature before the actual tumor, which might help explain the efficacy of Ad5-Δ24RGD. Further, this virus is not CAR binding ablated, which suggests that intratumoral dissemination and cell killing can occur also through CAR, if available.
Despite low CAR mediated gene transfer and cell killing in vitro (Figs. 1a and 2a), Ad5-Δ24E3 was nearly as effective as Ad5.pK7-Δ24 and Ad5/3-Δ24 in vivo. We have demonstrated that this is due to upregulation of CAR with this cell line in vivo (Ranki et al., unpublished). The frequency and generalizability of this phenomenon remains to be studied.
Liver Kupffer cells are the main mechanism responsible for the rapid clearance of intravenously injected adenoviruses from the circulation.27, 48 Various chemical substances have been studied for Kupffer cell inactivation, and enhancement of gene transfer efficacy in target tissue after adenoviral injection has been reported to occur.27, 28, 29 We demonstrated a 10-fold increase in gene transfer into tumor tissue after intravenous injection of poly(I) preceding intravenous injection of Ad5lucI (Fig. 4a). Further, liver transduction was reduced, as expected (Fig. 4b). Nevertheless, these phenomena did not translate into enhanced survival, when poly(I) injections were combined with therapeutic doses of Ad5.pK7-Δ24 in a systemic treatment model of metastatic hormone refractory breast cancer. Instead, some mice died soon after injection, suggesting possible treatment related toxicity. This may be related to the already compromised cardiovascular system of mice with advanced lung metastases (Fig. 4c). However, toxicity was not very frequent nor does it seem to explain the unimpressive survival of poly-I treated mice that experienced no treatment related toxicity. Further studied are needed to clarify why reduced liver transduction and increased transduction of mammary fat pad tumors was not associated with increased systemic efficacy. One possibility is that poly(I) somehow blocks intratumoral dissemination or prevents oncolysis. Nevertheless, despite some reports describing successful inactivation of Kupffer cells,27 it is noteworthy that none have described a therapeutic advantage. Our results confirm this intriguing discrepancy between transduction and efficacy data. Therefore, the question of systemic oncolytic potency may not be solvable by blocking of Kupffer cells alone.
In summary, we found that capsid modifications can be used for enhancing gene transfer to hormone refractory breast cancer cells. This in turn translated into enhanced oncolytic potency in vitro and enhanced survival of mice in vivo. Importantly, we saw enhanced gene transfer also to fresh patient samples. The most effective in the panel of viruses was Ad5-Δ24RGD, which is therefore particularly appealing for translation into phase I breast cancer clinical trials. In fact, the NCI is currently involved in setting up ovarian cancer and glioma trials with this virus, which might facilitate extension into breast cancer trials, pending acceptable initial safety data. In addition to Ad5-Δ24RGD, many of the other oncolytic adenoviruses had activity in the models studied. It would be useful if a number of agents with nonidentical capsids would be available for treatment of patients. Following a neutralizing antibody response which is likely after treatment of immune competent animals or humans, it might be interesting to study retreatment with a virus with a different capsid in a “sero-switch” approach.49 Neutralizing antibodies are conformation sensitive and even small changes such as described here are sufficient to avoid neutralization and efficacy is retained.16, 18, 50