Steven Rosenberg gave an interesting talk at this year’s American Association for Cancer Research meeting (AACR 2014). He discussed various cell therapies that were developed at the National Cancer Institute (NCI). He began with a review of 3 trials in metastatic melanoma that used the patient’s own tumor infiltrating lymphocytes (TILs), isolated, expanded and re-injected, as the treatment. Ninety-three patients were enrolled in the trials. The partial response rate (PR) was 32% and the complete response rate (CR) was 22%. Notably, some of the CRs were durable; Dr Rosenberg went so far as to state that TIL therapy could be curative, albeit in a relatively low percentage of patients treated. In a new trial of 110 patients they are seeing similar results, including durable PRs.

Similar attempts to use TIL therapy in other solid tumors have mainly failed. So one interesting question, posed by Dr Rosenberg, is why do melanomas readily respond immune therapies? Such therapies include not just TIL-based treatment but also to high-dose IL-2, checkpoint inhibitors: blocking CTLA4, blocking the PD-1 pathway, even agonist anti-CD40 antibody (mAb) treatment. All of these therapies will activate cytotoxic T cells and should also activate the rest of the immune system either secondarily, or in the example of agonist anti-CD40 mAb therapy, directly.

Melanomas are unusual in the abundance of TILs that are found within the tumor and the tumor microenvironment. Rosenberg floated the “mutation” hypothesis to explain why TILs are abundant in melanoma: melanoma tumors are highly mutated, with an average of 34 mutations per individual patient tumor. The mutation hypothesis posits that it is the abundance of mutations and therefore mutated proteins that drive TIL accumulation, that is, the mutations produce antigenic protein fragments that can be presented in context of MHC (MHC class I and class II are complexes found on antigen-presenting cells that activate T cells).

If this hypothesis is correct than several predictions can be made. One is that we should be able to find antigenic peptides that activate the TILs from specific patients. Another is that the TILs should be disabled by the tumor or tumor microenvironment (this is already suggested by the success of immune checkpoint inhibitors like ipilimumab and nivolumab in melanoma). Indeed, TILs isolated from patient melanomas express multiple immune control pathways, both in the immune response inhibitory pathways (PD-1, CTLA4, TIM-3) but also immune response activation pathways (4-1BB, OX-40, CD25, CD28, CD27, CD70) and others (LAG-3). So, these calls appears primed to respond, but are held in check.

Further, the TILs are primed to respond, at least in part, to tumor-derived peptides. Dr Rosenberg and colleagues sequenced the tumors from individual patients and used an algorithm to scan the data and identify immunogenic peptide fragments. They then synthesized the peptides and ask whether any of them could stimulate patient TILs. For each patient they found several immunogenic peptides. They could then isolate the T cell receptor (TCR) that mediated that recognition, and use it in an expression construct to develop mutation specific T cells. Note here that it is the TCR on the T cell that interacts with the MHC complex on antigen-presenting cells to trigger T cell activation. We have moved now from bulk TILs expanded ex vivo and re-injected to patient-specific engineered T cells specific for tumor antigens. This TCR-based cell therapy has now shown activity beyond melanoma and may be useful for other solid tumors that contain large populations of TILs. Finally, it may also be feasible to use the TIL immunogenic peptide data to craft highly tumor specific CAR constructs, i.e. by raising the CAR Vh domain (engineered as a scFV) to tumor-mutated antigens.

There remain significant unanswered questions. Other tumor types carry very high mutational burdens but do not accumulate large numbers of TILs – why not? The expression of immune control pathways on TILs derived from melanomas is complex – how best to manipulate these pathways? Also, how do TIL immune control phenotypes vary among patients? The identification of patient-specific immunogenic peptides may be useful in moving tumor vaccine therapy forward – how best to incorporate this data? Finally, a theme we always return to – how should doctors and patients use TCR-based therapeutics in the context of other available therapies.

The TIL data remind us that tumors raise an immune response to tumors, and this has implications for the re-emerging tumor vaccine field. Perhaps these mutated tumor antigens could be used in the context of tumor vaccination. There were several talks at AACR14 describing successful application of tumor vaccines in early phase clinical trials. There have been high-profile failures in this space – GSK’s phase 3 bust with their MAGE-A3 vaccine being a notable recent example. But sticking to melanoma, we see a few strong signals emerging.

Roger Perlmutter updated results from Amgen’s Phase 3 trial with T-Vec, which was initiated during his tenure (he is now at Merck). The T-Vec program was brought into Amgen with the $1 billion buyout of BioVex. T-Vec is a engineered viral vaccine that can infect and then replicate in tumor cells, pumping out the pleiotropic, immune-system priming growth factor GM-CSF along with encoded antigen. The injection is given at accessible tumor sites, e.g. in the skin, causing the melanoma to shrink. Importantly, not just the injected tumors, but tumors distant from the injection site responded, indicating that a systemic immune response had been triggered. T-Vec was compared to GM-CSF injection alone. While the overall response rate was high (about 60%) the interesting data are the comparisons of duration of response.

The response can be traced to cytotoxic T cells. These initially resemble patient TILs. However, after immunization these T cells have up-regulated immune response proteins (CD28, CD137, CD27, GITR) and down-regulated immune checkpoint proteins (PD-1, CTLA4, Lag3, TIM-3). So this immunization protocol is resetting the T cell phenotype, from immunosuppressed or anergic, to immune-competent and activated. This biological response is likely driven by the effect of GM-CSF on monocytes, macrophages and related cells. The mechanism of action bears further study.

We have not seen enough data yet to determine if there will be long-term responders (those that contribute to the “long tail” phenomena on OS curves) as we see in the immune checkpoint inhibitor trials. Regardless, Amgen is moving forward with clinical trials of T-Vec in combination with anti-CTLA4 mAb (Vervoy^tm, from Bristol-Myers Squibb) and with anti-PD-1 mAb MK-3475, in collaboration with Merck.

Lindy Durrant and colleagues from the University of Nottingham used a different approach to engage the immune system in the vaccine setting. They developed SCIB1, a DNA immunotherapy that encodes epitopes from gp100 and TRP-2 (melanoma antigens) into a human IgG1 antibody (honestly I need to understand better how they engineered this). The DNA vaccine is electroporated directly into muscle weekly x 3 and then at 3 months and 6 months. The transfection results in expression of the construct that is then taken up by Fc-receptor bearing cells via the CD64 Fc-receptor. CD64+ cells include monocytes, macrophages, dendritic cells and other immune cells. This Phase 1 study was designed as a 3×3 dose escalation study with an expansion cohort at the maximum tolerated dose, determined to be 4mg. Stage III and Stage IV melanoma patients were enrolled. 19/20 patients were shown to have an immune response to vaccination. There was a clear dose response. In the expansion cohort (n = 14) all patients showed an immune response despite expression of PD-L1 on tumor cells. Epitope recognition by both CD4 and CD8+ T cells was observed. Median survival of the expansion cohort is currently 15 months.

While this is a small early stage trial, such results are dramatic and highlight the concept that productively engaging the immune response requires recruitment of the patient’s antigen presenting cell populations (as noted above in the T-Vec example, this is what GM-CSF does). The tumor cell profile data hint at the potential use of PD-1 pathway blockade as a co-therapy for this DNA vaccine approach.

For smaller companies developing cancer vaccine modalities the potential to develop their technology alongside immunotherapy agents should be attractive. While PD-1 and CTLA4 targeting antibodies remain one obvious approach, data presented at AACR suggest that immune activating pathways (GITR, OX40 and others) might also be useful in the context of immune vaccine approaches. The trick will be to aim carefully.

We’ll follow up with a look at immune activation pathways.

stay tuned.