The safety of oncolytic viruses for treatment of cancer has been shown in clinical trials while antitumor efficacy has often remained modest. As expression of the coxsackie-adenovirus receptor may be variable in advanced tumors, we developed Ad5-D24-RGD, a p16/Rb pathway selective oncolytic adenovirus featuring RGD-4C modification of the fiber. This allows viral entry through alpha-v-beta integrins frequently highly expressed in advanced tumors. Advanced tumors are often immunosuppressive which results in lack of tumor eradication despite abnormal epitopes being present. Granulocyte-macrophage colony stimulating factor (GMCSF) is a potent activator of immune system with established antitumor properties. To stimulate antitumor immunity and break tumor associated immunotolerance, we constructed Ad5-RGD-D24-GMCSF, featuring GMCSF controlled by the adenoviral E3 promoter. Preliminary safety of Ad5-D24-RGD and Ad5-RGD-D24-GMCSF for treatment of human cancer was established. Treatments with Ad5-D24-RGD (N = 9) and Ad5-RGD-D24-GMCSF (N = 7) were well tolerated. Typical side effects were grade 1-2 fatigue, fever and injection site pain. 77% (10/13) of evaluable patients showed virus in circulation for at least 2 weeks. In 3 out of 6 evaluable patients, disease previously progressing stabilized after a single treatment with Ad5-RGD-D24-GMCSF. In addition, 2/3 patients had stabilization or reduction in tumor marker levels. All patients treated with Ad5-D24-RGD showed disease progression in radiological analysis, although 3/6 had temporary reduction or stabilization of marker levels. Induction of tumor and adenovirus specific immunity was demonstrated with ELISPOT in Ad5-RGD-D24-GMCSF treated patients. RGD modified oncolytic adenoviruses with or without GMCSF seem safe for further clinical development.
Inadequate treatment options are available for cancer patients with advanced metastatic disease refractory to current therapies. One experimental strategy with an increasing amount of clinical evidence is oncolytic adenoviruses.1
Replication of oncolytic virus per se kills tumor cells but oncolysis can also activate the immune system which may play a role in tumor control.2–4 Advanced tumors are nevertheless often immunosuppressive resulting in lack of tumor eradication.5 Tumor infiltrating dendritic cells (DCs) play a key role in immunological defence against cancer but may not be optimally activated in the tumor microenvironment due to lack of costimulatory molecules.6 Granulocyte-macrophage colony stimulating factor (GMCSF) is a potent activator of DCs and a local production of GMCSF at the tumor site has been shown to be involved in the activation of DCs.7 This is particularly appealing in the presence of a strong “danger” signal and abundant tumor epitopes, both of which result from viral oncolysis.
Recombinant GMCSF has been used extensively in cancer patients and thus the safety of the approach is well established.8 Nevertheless, toxicity can occur after systemic administration of large doses, while insufficient local concentration at tumors may have limited activity of the approach.9 Therefore, randomized phase III trials combining chemotherapy and GMCSF have shown conflicting results.9, 10 In contrast, when GMCSF has been used locally, preliminary results seem more promising.8, 11 Importantly, promising safety and efficacy data have been obtained in clinical trials featuring oncolytic viruses coding for GMCSF.12–15
Here, after establishing the safety of integrin targeted RGD modified oncolytic adenoviruses (exemplified by Ad5-D24-RGD) for treatment of human cancer, we generated Ad5-D24-RGD-GMCSF, an RGD modified oncolytic adenovirus coding for human GMCSF. A 24bp constant region 2 deletion in E1A16, 17 was introduced for p16/Rb pathway selectivity. In this article, we report the first human data with these viruses.
Material and Methods
To generate Ad5-RGD-D24-GMCSF, human GMCSF was amplified from pORF-hGMCSFv.12 (Invitrogen, Carlsbad CA). GMCSF was subcloned into pTHSN and recombined with a rescue plasmid featuring RGD-4C in the fiber HI-loop to generate pAd5-RGD-D24-GMCSF. Linearized plasmid was transfected into A549 cells for amplification and rescue. A nucleic acid sequence encoding GM-CSF is in the place of the deleted gp19k/6.7K in the E3 region. Deletions are not expected to reduce potency but may increase tumor selectivity.18 Gp19k has a role in downregulation of MHCI molecules, but as MHCI expression is typically deficient in tumors,19 this function is not required. In contrast, lack of MHCI blocking could enhance clearance from normal tissues.
Ad5-D24-RGD is intact in E3. Ad5-D24-RGD,20 Ad5Luc1 and Ad5lucRGD21 have been published previously. Clinical grade virus was produced according to the principles of good manufacturing practices (GMP) by Oncos Therapeutics, Inc (Helsinki, Finland).
Lung adenocarcinoma A549 and breast carcinoma MDA-MB-436 cell lines were acquired from ATCC (American Type Culture Collection) and cultured as recommended. Clinical grade A549 cell banks were developed by Oncos Therapeutics according to the principles of GMP.
Characterization of Ad5-RGD-D24-GMCSF in vitro
Lung cancer (A549) and breast cancer (MDA-MB-436) cells were infected and cell viability was analysed 6 days later with the Cell Titer 96 AQueous One Solution Cell Proliferation Assay (MTS assay; Promega, Stockholm, Sweden).
To verify GMCSF production, A549 cells were infected with phosphate buffered saline (mock) or Ad5-D24-RGD-GMCSF at 1000 VP/cell in 2% DMEM. GMCSF concentration in the supernatant was analyzed with Cytometric Bead Array (CBA) Human Soluble Protein Flex Set (Becton Dickinson, Franklin Lakes, NJ). The biological activity of adenovirus-coded GMCSF has been shown earlier.22
Baseline characteristics of patients are listed in Table 1 and prior therapies in Supporting Information Table 1. Patients were treated with a single round of virus at doses from 2x10e10 to 9x10e11 VP (Table 2). Four fifths of the dose was given intratumorally (or intraperitoneally for patients with peritoneal disease) using ultrasound guidance and one fifth intravenously. There is quite a lot of preclinical data suggesting that intravenous injection might have utility.23, 24 We have also preclinical data suggesting that iv + it delivery of a GMCSF expressing oncolytic adenovirus can result in antitumor efficacy.25 The study was completed according to Good Clinical Practice and the Declaration of Helsinki. This Advanced Therapy Access Program (ATAP) is approved by the Medicolegal Department of the Finnish Ministry of Social Affairs and Health and the Gene Technology Board, and is regulated by Finnish Medicines Agency Fimea.
Table 1. Summary of baseline characteristics
Table 2. Baseline characteristics and description of treatment
Side effects were graded according to CTCAE v3.0. Blood samples were collected at various time points after the treatment. Differences in sampling schedule were due to practical issues (weekends, holidays) and patient preference (e.g., unwilling to travel for providing blood sample). Patients were imaged before and after treatment and RECIST (Response Evaluation Criteria in Solid Tumors) criteria were used to evaluate antitumor efficacy.26 Ad5-D24-RGD-GMCSF treated patients were also administered metronomic low-dose cyclophosphamide (50mg/d, starting one week before virus and continued until progression) to downregulate regulatory T-cells,27 which have been proposed counterproductive for induction of antitumor immune responses.28
Neutralizing antibody titer determination
Neutralizing antibody titers (NAb) were determined as described.29
Viral DNA determination from serum samples
Total DNA was extracted and viral loads were determined as described.29 All available samples for all patients were analyzed. Positive samples were confirmed by real-time PCR using LightCycler480 SYBR Green I Master mix (Roche, Mannheim, Germany) and primers specific for adenovirus and RGD or GM-CSF sequences (Ad5-a forward primer 5′- ACAAACGCTGTTGGATTTATGC-3′ and RGD reverse primer 5′-GATGGGCAGAAACAGTCTCC-3′; Ad5-b forward primer 5′-AAACACCACCCTCCTTACCTG-3′ and GMCSF reverse primer 5′-TCATTCATCTCAGCAGCAGTG -3′). The limit of quantification for the assay was 500 VP/ml of serum.
Serum cytokine analysis
Cytokine analysis was performed with BD Cytometric Bead Array (CBA) Human Soluble Protein Flex Set. All available samples for all patients were analyzed.
Assessment of tumor and adenovirus specific immunity by ELISPOT
Tumor biopsy was not allowed in this ATAP study and therefore peripheral blood mononuclear cells (PBMCs) were used to evaluate the induction of tumor and adenovirus specific immunity after treatment. PBMCs were isolated by Percoll gradient. Cells were frozen in CTL-CryoABCTM serum-free media (Cellular Technology, Cleveland, Ohio). For adenovirus ELISPOT, cells were stimulated with the HAdV-5 Penton peptide pool (ProImmune, Oxford, UK). For survivin, BIRC5 PONAB peptide was used (ProImmune). No pre-stimulation of PBMCs was done to avoid artificial or incorrect signals and to ensure adequate viability of cells which might be compromised during prolonged culture.
All analyses were done with SPSS 15.0 software for Windows. Correlation between input viral dose and viral genomes in the serum was analyzed with Pearson Correlation analysis.
Preclinical characterization of Ad5-D24-RGD and Ad5-RGD-D24-GMCSF
Construction and preclinical testing of Ad5-D24-RGD has been described previously.20, 30 Oncolytic potency of Ad5-RGD-D24-GMCSF was verified to be equal or nearly equal as wild type on most tumor cell lines. However, in some conditions wild type virus was more potent while on some cell lines RGD-modified oncolytic viruses have earlier shown to be more potent30, 31 (Supporting Information Fig. 1A). Measurable levels of GMCSF were seen in the supernatant after infection with Ad5-RGD-D24-GMCSF (Supporting Information Fig. 1B).
Toxicity and biodistribution studies with Ad5-D24-RGD have been performed previously32 and the safety of systemic GMCSF in humans is well established.33 As human GMCSF is not active in mice, and tropism (determined by the capsid) of Ad5-RGD-D24-GMCSF is identical to Ad5-D24-RGD, Ad5-RGD-D24-GMCSF behaves identically to Ad5-D24-RGD in available preclinical test systems.20, 30, 32
Given the multiple tumor types that the patients had, one way to view the data is as a set of case reports. All patients experienced mild to moderate adverse events (AE) after treatment, typically fatigue and fever (Table 3). 5 patients treated with Ad5-D24-RGD had grade 3 AEs. Four were asymptomatic haematological findings (alkaline phosphatase increase, hyponatremia, INR increase, lymphocytopenia) while one patient (Y36) had grade 3 fever for two days.
Table 3. Adverse events
For Ad5-RGD-D24-GMCSF treated patients, one (C143) displayed asymptomatic grade 3 hyponatremia and another (C151) experienced grade 4 pulmonary embolism 3 weeks after treatment with 9x10e11 VP. Anti-thrombotic medication was started and the patient recuperated without sequelae. It is unclear if the virus played a role in this event, as thrombotic events are quite common in late stage cancer patients and have been reported in up to 15% in colorectal cancer patients.34 In contrast, there is no data linking adenovirus to thrombosis.
IL-6 or TNF-alpha have been proposed sensitive markers of adenovirus toxicity.35 IL-6 and IL-8 showed a considerable variation at a baseline between patients whereas baseline values for IL-10, TNF-alpha and GMCSF were less variable and often low or undetectable (Supporting Information Table 2).
4 out of 14 evaluable patients (both viruses) showed a measurable concentration of serum GMCSF at the baseline. 4 out of 9 and 3 out of 5 patients treated with Ad5-D24-GMCSF and Ad5-RGD-D24-GMCSF, respectively, showed small increases in GMCSF concentration after the treatment. Overall, no major changes in cytokines were seen, suggesting good safety (IL-6, TNF-alpha), and localization of GMCSF production to the tumor (Ad5-RGD-D24-GMCSF patients).
Virus presence in the circulation
For Ad5-D24-RGD treated patients, a correlation between input viral dose and the highest viral load in the serum was seen (p= 0.041). In all seven evaluable cases, virus was detected in serum on day 11 or later and the latest time point for the positive measurement was day 48 (Table 4).
With Ad5-RGD-D24-GMCSF, 6 out of 7 patients had virus present in serum in at least one measurement later than day 1, but viral loads were lower in comparison to day 1 for all except O130, who had 18 473 VP/ml on day 8 after treatment. The latest time for virus detection was day 64.
Induction of neutralizing antibodies
All patients after treatment by 1–3 weeks, except K117, had an increase in their NAb titer (Table 4). NAb titers remained high until the end of follow-up for all patients except C151, who showed a minor decrease after 6 weeks.
Adenovirus specific T cell response
As hexon peptide pool usable in ELISPOT is not available, we used a penton pool to stimulate PBMCs. Interestingly, baseline Ad responses (before virus treatment) were well in line with those reported in healthy donors.36 3 out of 7 patients (M126, S135, C145) showed an increased number of circulating adenovirus penton recognizing PBMCs after treatment (Fig. 1a), while 3 (K117, O130, C151) showed a decrease. Patient C143 had no detectable numbers before or after the treatment.
Tumor-specific immune response
Patient O130 showed induction of survivin specific T-cells 1 month after Ad5-RGD-D24-GMCSF treatment and they were detectable also two months after virus injection (Fig. 1b). Four patients (S135, M126, K117, C151) showed a decrease while two patients (C145, C143) had no detectable numbers.
Antitumor responses and survival
All patients had progressive disease before treatment. 5 out of 9 patients treated with Ad5-D24-RGD (S23, O26, Y36, C47, E40) were evaluable according to RECIST and all showed progressive disease (PD) (Table 4). For Ad5-RGD-D24-GMCSF, 6 out of 7 patients were evaluable, and three cases (C143, O130, C151) were scored as stable disease (SD). The three other cases (M126, S135, C145) showed PD.
For Ad5-D24-RGD, tumor markers were evaluable for 6 patients (Table 4). In two cases (O26, R42), tumor markers decreased 15% and 14%, respectively, which was interpreted as a minor response (MR). Patient Y36 showed a 1% increase in Ca19-9 after 49 days and was thus scored as SD. In all of these cases the tumor markers were progressing before therapy. Two cases (C31, E40) showed continuous increase in their tumor markers and were scored as PD.
Tumor markers were evaluable for three Ad5-RGD-D24-GMCSF treated patients and one MR (C151; 22% decrease in CEA in two weeks), one SD (O130; no change in Ca12-5 for 1 month) and one PD (C143; 92% increase in CEA in 1.5 months) was seen. Ovarian cancer patient O130 treated with Ad5-RGD-D24-GMCSF was alive and well at the end of follow-up (378 days) (Supporting Information Fig. 2).
Pretreatment analysis of tumor samples for prediction of treatment efficacy
In three cases (K117, C143, C151), fresh pretreatment malignant pleural effusions were available for transduction analysis (Fig. 2a). In all cases, gene delivery with Ad5lucRGD could be achieved. C143 and C151 were evaluable for efficacy by RECIST and both had stabilization of disease which had been progressing before therapy (Table 4). These preliminary data suggest that the ex vivo transduction assay might be interesting for testing of prediction of clinical utility but further studies are needed.
Post-treatment quantitation of virus at the tumor versus in serum
Oncolytic viruses are expected to replicate locally at tumors. This was testable in two patients, whose pleural effusion samples were available after treatment (Fig. 2b). For K117, 456 923 VP/ml were found at the tumor on day 63 while no genomes were detected in serum. In M126, the respective numbers were 528 131 VP/ml at the tumor and <500 VP/ml in the serum on day 63.
RGD-4C modification of the fiber HI-loop allows adenovirus to enter cells through αvβ integrins which has been shown to enhance transduction of several human cancer cell lines20, 30, 37, 38 and to improve efficacy of oncolytic adenovirus in xenograft murine models30, 37, 38 and in primary tumor specimens.39 The utility of the approach has been demonstrated especially in cancer cells low in coxsackie-adenovirus receptor (CAR).37, 40 Clinical evidence suggests that CAR expression is highly variable and often low in advanced tumors.41 In contrast, abundant expression of αvβ integrins is a common feature in many cancers.20, 30, 37, 38, 40 Here, pre-treatment pleural effusion cells isolated from patients showed efficient ex vivo transduction with RGD modified adenovirus.
After a single treatment with Ad5-D24-RGD or Ad5-RGD-D24-GMCSF, virus was seen in serum for 2 to 5 weeks in 77% of evaluated patients. In line with earlier studies suggesting that prolonged presence of the virus in circulation, and especially titer increases over day 1 values, may be indicators of active virus replication in tumors,42, 43 our data suggests that RGD modified viruses are capable of prolonged replication in cancer patients.
A recently published exploratory study showed safety of an unarmed RGD modified oncolytic adenovirus (ICOVIR-5) in treatment of four children with refractory metastatic neuroblastoma.44 Similar safety was seen in our patients treated with Ad5-D24-RGD. These findings are in line with two other recent reports where unarmed capsid modified oncolytic adenoviruses were used for treatment of advanced and refractory cancer patients.29, 45
Ad5-RGD-D24-GMCSF is the first armed integrin targeted adenovirus used in humans. Treatment was generally well tolerated and side effects were similar to what was seen with unarmed viruses. Similarly with previous experience suggesting no definite connection between viral dose and antitumor responses, because of virus replication,15, 29 no correlation was seen in our patients either. Thus, the initial quantity of virions in the tumor may not be the factor that determines activity or lack thereof. Instead, immunological or tumor specific factors (such as intratumoral complexities and stromal areas) might play a more important role. Further studies are needed to understand these issues better and human data may be most useful as model systems suitable for oncolytic adenoviruses are not very good in reproducing conditions present in the microenvironment of human tumors. The easiest way to improve tumor transduction, immune response and efficacy might be to administer virus several times.
Our data is in accord with previous reports suggesting that GMCSF as a viral transgene is relatively well tolerated.12-15 In comparison to systemic dosing of recombinant protein, where side effects can become limiting,9 the use of a viral vector is attractive because it facilitates high local concentration of the transgene product at the tumor while systemic concentrations remain low. This may be particularly relevant in the context of GMCSF, which has well established antitumor activity when present at sufficient concentrations at the tumor,11 while high systemic concentrations have been proposed to harbor the potential for unwanted immunological effects.46 Viral delivery has the additional advantage of long-term expression without the need for daily administration.
In Ad5-RGD-D24-GMCSF, transgene expression is controlled by the E3 promoter, and therefore GMCSF production should be restricted to sites allowing virus replication.47 In support of this notion, no significant increases in systemic levels of GMCSF were detected over time, nor were there differences between Ad5-RGD-D24-GMCSF and Ad5-D24-RGD treated patients.
Ad5-D24-RGD was designed with optimal oncolytic potency in mind (Bauerschmitz Cancer Res 2002). In contrast, Ad5-RGD-D24-GMCSF (reported here for the first time) was armed with GMCSF to achieve an anti-tumor immune response. One of the key lessons learned in the field of cancer vaccines in recent decades has been that immune suppressiveness mediated by tumors is a major obstacle.48 Therefore, we combined Ad5-RGD-D24-GMCSF treatments with metronomic low-dose cyclophospamide, known to be effective in reducing the number and activity of regulatory T-cells.49, 50 However, this resulted in uneven use of cyclophophamide between the cohorts which might contribute to efficacy data. Thus, as always in non-randomized series where the main endpoint is safety, it is important not to make any final conclusions on efficacy.
In radiological evaluation, 3 out of 6 evaluable patients treated with Ad5-RGD-D24-GMCSF showed stabilization of disease previously progressing. It may be possible that GMCSF mediated immune activation contributed to these findings, as suggested by preclinical studies.22 Accumulating evidence suggests that tumor antigens released from dying cancer cells are available for antigen presenting cells which can activate cytotoxic T-lymphocytes for killing of non-infected tumor cells.4 The “danger signal” provided by oncolysis might be amplified further by locally produced GMCSF and this could lead to enhanced antitumor immune attack. Further assessment of the GMCSF mediated immunological responses would require additional studies with immune responsive tumor types such as melanoma or renal carcinoma. However, the final confirmation of the role of GMCSF could be done only in the randomized clinical trial.
Patient O130 showed an increase in survivin-recognizing T-cells suggesting induction of tumor-specific immunity. This patient also had the highest serum virus load among Ad5-RGD-D24-GMCSF treated patients both on day 1 and one week later. O130 also had a delayed peak of virus in serum (day 8), whereas other patients had their highest titer shortly after the treatment (day 1). Interestingly, patient O130 showed also stabilization of disease for 8 months and she was alive and well more than 1 year after treatment, which is unusual in patients with metastatic chemotherapy refractory tumors.
Whether the induction of adenovirus-specific T-cells, observed in 3 out of 7 patients, is beneficial vis-a-vis efficacy, remains to be resolved in further clinical studies. Nevertheless, it has been suggested that antiviral immune response may be an important part of the overall antitumor effect mediated by oncolytic viruses.2–4
Interestingly, the amount of circulating virus-specific and/or tumor-specific T-lymphocytes decreased in 5 out 7 patients after treatment. Lack of induction of T-cell responses would be expected to cause no changes in T-cell circulation. Thus, one explanation for the finding is that the cells were recruited to the tumor site. Not opposing this hypothesis, patient C151 whose disease stabilized showed a decrease in both survivin- and adenovirus-specific T-cells in the blood. However, as we only analyzed circulating PBMCs, no conclusions on cell numbers at tumors can be made from this data.
In line with numerous earlier studies,12, 14, 15 a clear induction of neutralizing antibodies was seen in all patients and 66% of patients had detectable levels already at baseline. Previous studies with oncolytic viruses suggest that pre-existing immunity against virus might increase the safety of the treatment14 but it has not been shown to limit efficacy.15 Moreover, no data is available on the role of potentially beneficial antitumor antibodies, which might be induced as a result of oncolytic virus treatment.
Several species of oncolytic viruses have been armed with GMCSF and tested in clinical trials. The data suggests good safety for this class of drugs and the emerging evidence of antitumor efficacy and tumor-specific immune activation hold promise for further development.13, 15 The massive tumor volume often seen in patients with advanced disease may be an important hindrance for oncolytic virus therapies. In contrast to artificial tumor models in animals, human tumors are often quite heterogeneous. Many normal cell types, such as endothelial cells and fibroblasts, are present and may not allow virus replication. Antiviral immune activity might hamper vascular infection of metastases despite extended presence of virus in circulation. These issues are important reasons for attempting to obtain systemic efficacy by recruiting the immune system, instead of relying on oncolysis alone.
Considering that all of the patients in this series had advanced disease progressing after heavy pretreatment, stable disease in half of the Ad5-RGD-D24-GMCSF treated patients may be promising but leaves plenty of room for further optimization of the treatment protocol. Aspects appealing for testing in future studies include repeated injections, combining with other therapies and treatments given earlier and to lower tumor load patients, whose tumors might be less immune suppressive.
The authors thank Dr. Saila Eksymä-Sillman, Dr. Marina Rosliakova, Mrs. Jenni Kylä-Kause and other personnel at International Comprehensive Cancer Center Docrates and at Eira Hospital, Helsinki, Finland. Akseli Hemminki is K. Albin Johansson Research Professor of the Foundation for the Finnish Cancer Institute. A.H. is co-founder and shareholder in Oncos Therapeutics Inc.