Workshop Meeting Report

Introduction

Analyzing biological mechanisms at a single-cell level can provide significant detail into cellular type, spatial structures and characterization, and cellular circuitry. Single-cell analytical concepts have existed for many years, however they were costly and inefficient as they had limited throughput. Technological advancements – including the advent of next-generation sequencing as well as single-cell flow cytometry and mass spectrometry – have increased yield and reduced costs, allowing for these methods to become more readily utilized in pre-clinical research. Advancements in single-cell technologies have had a significant impact on many medical fields, including cancer immunotherapy. Cancer immunotherapy is an expanding collection of treatments that primarily re-engage host immunity to target and destroy tumor cells. Many current immunotherapeutic strategies - including treatment with monoclonal antibodies, adoptive cell therapies, and vaccines – rely upon precise and specific interactions with immune or tumor markers, many of which have not been well-characterized mechanistically. As such, while immunotherapies have displayed beneficial and durable patient responses thus far, elucidation of basic biological mechanisms underlying targeted pathways and markers via single-cell techniques can greatly increase therapeutic efficacy. As such, The Society for Immunotherapy of Cancer (SITC) held a one-day program - titled Workshop on Single Cell Techniques in Immunology and Cancer Immunotherapy - as part of its 32^nd annual meeting. The goal of this meeting was to bring together experts in the field of cancer immunotherapy to discuss advances in single-cell technologies, and to educate attendees about potential applications. In this report, we summarize data presented at this meeting towards promoting incorporation of single-cell techniques into pre-clinical laboratories, drug-development programs, and clinical trials.

Understanding  immune checkpoint blockade response using single-cell techniques

PD-1/PD-L1 blockade has been very effective in treating a variety of malignancies, as evidenced by the number of recent FDA approvals. A thorough mechanistic understanding of the effects of checkpoint inhibition on T cell functionality, however, remains elusive. Nir Hacohen, PhD (Massacuhsetts General Hospital, Boston, MA, USA) believes that investigating the question using a systems immunology approach may help close this knowledge gap by identifying and characterizing specific mechanisms and pathways that contribute to immunotherapy efficacy. Hacohen’s group used single-cell RNAseq to analyze T cell gene expression profiles in samples collected from cancer patients treated with ICIs. Single cell analyses indicated that T cells have varied expression patterns that are more complicated than common labeling of CD8+, CD4+, etc. would suggest. The ratio of clusters with expression signatures of T cell exhaustion and memory were associated with therapeutic response. Specifically, patients who responded to CI therapies had gene expression signatures that were associated with an increase in T cell memory function compared to exhaustion. Conversely, signatures associated with exhaustion were over-represented compared to those associated with memory function in non-responders. In some samples, HLA loss was observed in non-responders who also had increased memory signatures, suggesting that alternative mechanisms and indicators of response also exist.

After identifying the aforementioned large-scale indicators of ICI response, Hacohen’s group worked to identify more specific factors driving these genetic programs. Transcription factor 7 (TCF7) has been previously implicated in altered T cell exhaustion and memory functionality. Using immunohistochemistry (IHC), researchers found that CD8+ and TCF7+ signals co-associated in tumor samples from ICI responders. In general, TCF7+ cells were increased in responders compared to non-responders. Additionally, the ratio of TCF7-/+ cells correlated with patient response to ICIs in a similar fashion to T cell memory and exhaustion genetic signatures (AUC = 0.92).

Anti-tumor immune responses and the outcome of cancer immunotherapy are integrally connected to cancer genomics, as neoantigens stemming from somatic mutations seem to shape immune responses and drive clinical benefit to these therapies.  To this end, tumor mutational burden (TMB) has been shown to predict response to immune checkpoint inhibitors (ICI) in a variety of tumors. An improved understanding of the genomic and neoantigen landscapes may therefore facilitate accurate prediction of response to these therapies. Unfortunately, the majority of cancer patients either do not benefit from immunotherapy or develop acquired resistance after an initial response highlighting the need to develop novel biomarkers that predict outcome. 

One important, yet understudied aspect of ICI use is assessment of how therapy alters the cancer genome over the course of treatment. To address this, Valsamo Anagnostou, MD, PhD (Johns Hopkins University, Baltimore, Md, USA) presented research investigating tumor evolution under selective pressure of ICI therapy and novel mechanisms of resistance to these therapies. Pre-treatment and post-progression tumor samples from patients with non-small cell lung cancer (NSCLC) that progressed after an initial response were analyzed by whole exome sequencing, T cell receptor (TCR) sequencing and functional assays of T cell activation. Anagnostou’s group demonstrated that acquired resistance to immune checkpoint blockade can arise in association with the evolving genomic and neoantigen landscape in NSCLC [1]. More specifically, truncal and subclonal neoantigens were noted to be lost in the resistant tumors either through tumor cell elimination or chromosomal deletions.

In order to investigate the importance of eliminated neoantigens, peptides generated from the eliminated neoantigens were synthesized and their potential to elicit a clonotypic T cell amplification was tested in autologous T cell cultures. Neopeptides triggered neoantigen-specific clonal T cell expansion, suggesting that they generated functional immune responses and were important targets for achievement of the initial response to ICI. Taken together, these data suggest that response and emergence of acquired resistance to ICI therapy is shaped by changes in the mutation and neoantigen landscape in NSCLC.

Expanding on the study of clonal dynamics during ICI therapy, the group is now exploring non-invasive approaches that combine analyses of the evolving neoantigen landscape and the immune repertoire. These ongoing studies may lead to identification of robust biomarkers of response to ICI, allow the rational design of immuno-oncology treatment strategies and determine immediate next steps to overcome resistance to immune checkpoint blockade.

Circulating tumor DNA in single cell sequencing

Decisions as to whether to continue immune checkpoint therapies are standardly guided by changes in tumor size measured using clinical imaging techniques. Immunotherapy response, however, is difficult to interpret based on radiographic changes alone, as tumors often shrink slowly or appear transiently enlarged due to inflammation (pseudoprogression). A research team led by Abhijit Patel, MD, PhD (Yale University, New Haven, CT, USA) hypothesized that monitoring and quantifying changes in circulating tumor DNA (ctDNA) – only released during tumor cell death – could provide early evidence of immunotherapy response in patients with metastatic NSCLC. To overcome challenges in accurately measuring ctDNA, including DNA fragmentation and potential PCR/sequencing errors, the researchers developed a novel, ultrasensitive, highly-multiplexed NGS assay that yields a significantly reduced error rate (<1 per 100000 bases) and can analyze up to 100 samples per batch. The approach was validated in a preliminary study designed to measure both radiographic and ctDNA response in 28 patients with NSCLC receiving immune checkpoint inhibitors (ICIs). Overall, there was strong correlation between radiographic response to immunotherapy and reduction in ctDNA abundance (less than 50% of baseline level). As hypothesized, ctDNA levels fell significantly sooner than radiographic response became apparent (response median: 24.5 days vs 72.5 days, respectively). Importantly, improvements in overall survival (OS) and progression-free survival (PFS) were observed in patients who displayed a ctDNA response compared to those who did not (OS, p = 0.03; PFS, p = 0.007). Patel noted that while this study generated positive results, it had limitations including the lack of a validation cohort, and variability of blood collection time points and immunotherapy regimens. Nonetheless, these findings provide rationale for measuring ctDNA in conjunction with standard imaging techniques to provide an earlier and more comprehensive assessment of immunotherapy efficacy.

Assessing combination immunotherapeutic efficacy through cellular characteristics

Immune escape is a primary mechanism exploited by tumor cells to evade immunotherapeutic efficacy. Immune escape mechanisms reduce therapeutic efficacy in specific patients through loss of immunogenicity and/or enhancement of immunosuppression within the tumor microenvironment. Mechanisms for bypassing an immune response can be partially attributed to aspects of tumor biology including occlusion of T cell infiltration in the surrounding stroma. Many immune escape mechanisms, however, remain unclear, so a team led by Priti Hegde, PhD (Genentech Inc., San Francisco, CA, USA) used NGS to analyze tumor biopsies from patients with metastatic renal cell carcinoma (mRCC) enrolled in the IMmotion 150 clinical trial (NCT01984242). Patients received either standard of care (suntinib, 50 mg, four weeks on, 2 weeks off), PD-L1 inhibitor atezolizumab (1200mg IV Q3W), or combination therapy (atezolizumab + bevacizumab, bev: 15mg/kg Q3W) and were stratified by a T cell effector gene signature consisting of IFN-g and it’s downstream targets. Patients with gene signatures indicating low T effector cell presence derived no observable PFS or OS benefit from either immunotherapy regimen compared to sunitinib, while patients with high T-effector signatures treated with combination atezolizumab + bevacizumab demonstrated a 45% reduction in risk of death compared to those randomized to sunitinib (HR: 0.55; 95% CI: 0.32 – 0.95). In this cohort, atezolizumab monotherapy also demonstrated a modest reduction to risk of death compared to suntinib, but to a lesser degree than combination therapy (HR: 0.85; 95% CI: 0.50 – 1.43). To investigate these differences, Hegde’s group further stratified patients with high T-effector signatures by myeloid inflammation gene expression signatures, given the role of VEGF in promoting immunosuppression via myeloid derived suppressor cells. Patients with high myeloid inflammation and T-effector signatures who received combination atezolizumab + bevacizumab had increased PFS compared to patients receiving atezolizumab monotherapy (HR: 0.25; 95% CI: 0.1 – 0.6). Indeed, in this subgroup, atezolizumab monotherapy performed poorly with a median PFS of ~ 3 months compared to ~ 7 months for Sunitinib. Patients with a low myeloid inflammation signature among the T-effector signature positive population performed well on atezolizumab monotherapy alone, consistent with the combination of atezolizumab + bevacizumab, suggesting the combination of a doublet therapy in this population may not provide added benefit compared to monotherapy checkpoint inhibition alone.

Using single cell techniques and computational biology to classify immune cells

Computational decomposition of tumor-derived immune cells by single-cell RNAseq clusters gives rise to many different cell types. Correlation of these cell types with patient response to ICIs, however, reveals that only certain classifications of cells expand. For example, pre-clinical data suggests only specific subsets of CD8+ PD1+ cells expand during response to anti-PD-1 therapy, and other markers matter. As such, Dana Pe’er, PhD (Memorial Sloan Kettering Cancer Center, New York, NY, USA) discussed using computational approaches to better understand immune cell proliferation during ICI response using single-cell analyses. Pe’ers goal is to develop an ‘atlas’ that characterizes whole tumor immune cell ecosystems. To accomplish this, a novel droplet-characterization method was used to analyze the majority of CD45+ cells in tissue samples from patients with breast cancer. Immune cell subtypes were identified using a clustering program called BISCUIT (Bayesian Inference for Single-Cell Clustering and Imputing) to correct for RNAseq dropout normalize cell distribution across samples, revealing 82 cell type clusters (about 60,000 cells total) within the combined tumor, normal tissue, blood, and lymph node samples. Tumor samples showed increased T cell, myeloid cell, and macrophage cell states compared to normal tissue and blood. Additionally, novel, intratumoral cell type intersections existed that varied based on T cell activation and were distinct within each individual cluster. Other determinants of cell type included hypoxia and T cell exhaustion. Tissue histogenesis also played a role in differentiating cell types, where activation/hypoxia/exhaustion phenotypes are innately different. In certain cell clusters, such as those indicative of regulatory T cells, expression co-variation was also shown to drive cell type differentiation. 

Assessing T cell functionality at the single cell level

As T cell classification becomes increasingly complex, assessment of T cell functionality may prove useful in guiding immunotherapy treatment selection. Sean Mackay (Isoplexis, Branford, CT, USA) discussed efforts to refine polyfunctional strength – an index assessing T cell functionality - to more accurately predict response to immune checkpoint inhibitors (ICIs). Isoplexis has developed a chip-based technology that allows for single-cell capture into individual chambers and subsequent detection of highly multiplexed (up to 40+ proteins) cytokine profiles via an antibody barcode array. The ability to look at more secreted cytokines per single cell allows for more accurate stratification of cytokine responses than bulk analyses. Using this technology, cytokine signals are assessed using ELISA based methods to reveal active processes mediated by the individual T cell. T cell data can then be evaluated in conjunction with phenotypes such as therapeutic response. Both signal frequency and intensity are measured to provide a quantitative output used to develop functional response profiles, or polyfunctional strength index (PSI). In one example, tumor-infiltrating lymphocytes (TILs) within melanoma samples from patients treated with anti-PD-1 and/or anti-CTLA-4 agents were enriched using anti-CD8 and anti-CD4 microbeads, and stimulated with anti-CD3 to induce cytokine secretion. Stimulated cells were then analyzed using this technology, and results indicated that responders (n = 7) had an increased polyfunctionality (> 1 cytokine detected, p = 0.05). PSI was also increased in responders compared to non-responders (p = 0.03). Effector function signals – including granzyme B, IFN-ƴ, TNF-α, MIP-1α, and perforin – made up the majority of the observed markers in patients who experienced response. Of note, the presence of highly polyfunctional subsets was present in responding patients and absent in non-responding patients treated with anti-CTLA-4 and/or anti-PD-1 treatments. Polyfunctional strength index has also been used to analyze CAR T pre-infusion products, where increased cytokine secretion correlated with patient response to therapy (p = 0.01).

Other techniques, including mass cytometry and NGS, can also be used to assess T cell functionality. Ana Anderson, PhD (Harvard Medical School, Boston, MA, USA) described methods combining these techniques to determine T cell functionality at a single-cell level. TILs from B16F10 melanoma samples were initially analyzed using CyToF, then clustered samples based on results assessing immune stimulatory and inhibitory molecules. Analyses revealed that CD4+ and CD8+ T cells expressed a module of co-inhibitory (eg. TIM-3, PD-1, LAG3, and TIGIT) and co-stimulatory molecules (eg. OX-40, GITR, and 41BB). Some co-inhibitory and co-stimulatory molecules were notably excluded from the shared module.

Previous studies indicated that colon carcinoma tissue contains CD8+ T cells with functionality determined by TIM-3 and PD-1 expression patterns [2]. TIM-3-/PD-1- cells display markers indicating increased cytotoxicity, while TIM-3+/PD-1+ cells demonstrate reduced cytotoxic marker expression and increased IL-10. TIM-3-/PD-1+ cells displayed an intermediate phenotype regarding cytotoxicity. Based on these observations, Anderson’s group analyzed the transcriptomes of these observed CD8+ populations. Ten clusters of gene expression were constructed using expression patterns of ~3000 differentially expressed genes. Of interest, one gene cluster had genetic signatures indicating both T cell exhaustion and activation, highlighting a conundrum: exhaustion and activation mechanisms appear to be intimately intertwined. Upon further investigation of this gene cluster, increased expression of metallothionein-1 (MT-1) was observed. To better understand whether MT-1 actively contributes to regulation of T cell activation/exhaustion, Anderson’s group analyzed MT-1 deficient mice for tumor growth phenotypes. Her group demonstrated that MT-1 deficient mice displayed increased control of melanoma tumor growth. Additionally, MT-1 overexpression abrogated tumor control. Analyses of CD8+ T cells from MT-1 deficient mice displayed no differences in TIM-3 or PD-1, but did reveal an increase in effector markers, indicating that MT-1 deficiency uncouples the TIM-3/PD-1 surface-based phenotype from cell functionality. Researchers are now working to verify MT-1 involvement, as well as to characterize T cell populations in the tumor microenvironment in an attempt to better understand receptor expression in relation to activation/exhaustion phenotypes.

Using single-cell analyses of protein/neoantigen structural characteristics

Mass spectrometry can be used to characterize specific protein interactions, such as neoantigen/MHC associations. James Heath, PhD (California Institute of Technology, Pasadena, CA, USA) described efforts to better understand the implications of tumor neoantigen quality to interactions with MHC molecules. A primary challenge for neoantigen characterization is that the number of HLA-specific cells is low in liquid samples and expansion of these cells post-collection can distort results. Thus, Heath’s group has developed a library-based method that uses neoantigen-specific MHC tetramers, which are attached to magnetic nanoparticles using a DNA oligomer that acts as a barcode for the antigen under investigation. Magnetic beads are then mixed with patient TILs or peripheral blood mononuclear cells, separated and analyzed using a single-cell chip. Barcodes are read after cell capture to determine neoantigen structure. Heath’s group validated the neoantigen identification capabilities of this process by purifying T cells from the single-cell chip, and subsequently cloning the associated receptor, demonstrating neoantigen binding in vitro. Moving forward, a dropseq-like method has been recently incorporated into the single-cell chip technologies to allow for RT PCR of individual T cells, and to determine TCR alpha/beta chain sequences that are matched to the specific neoantigen/MHC.

Despite identifying potentially immunogenic neoantigens using the above process, it remains unclear which isolated TCR/neoantigen will garner the best therapeutic response. Heath hypothesized that the presence of a ‘catch bond’, which is a bond that can be formed during TCR –peptide(p)MHC dissociation, could potentially increase binding interactions to facilitate increased immune response. Using molecular dynamics simulations, experimentally characterized catch bonds were modeled by applying sheer force to disrupt the TCR-pMHC binding.  Under sheer, the simulations revealed that when specific salt bridge or hydrogen bonding interactions break, they can be reformed from different bonding partners, leading to extended engagement between the T cell and the antigen-presenting cell.  Whether such simulations can be used to predict agonist TCR-peptide/MHC interactions from, for example, sequence data, remains to be seen.

The role of the gut microbiome in response to immunotherapy

Systemic immunity and tumor characteristics each play an important role in ICI resistance, but very little is understood about the contribution of the host environment. Dysbiosis of the human microbiome has recently been implicated in the etiology of a variety of diseases, including cancer, but relatively little is known about the influence of microbial composition and diversity on response and resistance to immunotherapy. Jennifer Wargo, MD, MMSc (MD Anderson Cancer Center, Houston, TX, USA) presented research investigating the role of the human gut microbiome on anti-PD-1 efficacy. Wargo’s group measured the composition and diversity of microbiome samples in 233 patients with metastatic melanoma, before and after anti-PD-1 treatment, using 16SrRNA sequencing or whole genome sequencing (WGS). Results indicated significantly greater gut microbiome diversity in responders than non-responders (p < 0.01). Stratifying patients into high-, intermediate-, or low-microbiome diversity groups showed that patients with a highly diverse gut microbiome have significantly longer PFS than patients with intermediate- or low-diversity (p < 0.05 each). Specifically, responders had increased abundance of Clostridia, while non-responders had a preponderance of Bacteriodales species. PFS was longer in patients with high abundance of the Clostridiales member Feacalibacterium compared to patients with low abundance (p < 0.05). Additionally, patients with high abundance of Bacteriodales had reduced PFS compared to patients with low abundance (p < 0.05). Of note, responders also had increased CD8+ T cell proliferation (p < 0.05). Interestingly, increased abundance of Clostridiales correlated with an increase in observable tumor T cell markers. Increased abundance of Bacteriodales, however, correlated with reduced tumor T cell infiltration. Recently, Wargo’s group has shown that mice transplanted with microbiomes from ICI responders have less tumor growth compared to mice transplanted with microbiota from non-responders. As such, research done by Wargo’s group supports further investigation into the efficacy of this therapeutic approach in humans. 

Conclusions

Pre-clinical research will be critical for enhancing patient response rates and reducing toxicities when receiving immunotherapy. Improvements upon current immunotherapies will require small-scale optimization that will require the development of novel, single-cell technologies. Research discussed at this workshop detailed significant advancements in single-cell methods that will help elucidate specific mechanisms regarding tumor biology and immunity. Further incorporation of these novel technologies into pre-clinical research and drug-development programs will provide a significant impact to the field of cancer immunotherapy.

References

1. Anagnostou V, Smith KN, Forde PM, Niknafs N, Bhattacharya R, White J, Zhang T, Adleff V, Phallen J, Wali N et al: Evolution of Neoantigen Landscape during Immune Checkpoint Blockade in Non-Small Cell Lung Cancer. Cancer Discov 2017, 7(3):264-276.
2. Sakuishi K, Apetoh L, Sullivan JM, Blazar BR, Kuchroo VK, Anderson AC: Targeting Tim-3 and PD-1 pathways to reverse T cell exhaustion and restore anti-tumor immunity. J Exp Med 2010, 207(10):2187-2194.