Tumor-agnostic transcriptome-based classifier identifies spatial infiltration patterns of CD8+T cells in the tumor microenvironment and predicts clinical outcome in early-phase and late-phase clinical trials.

BACKGROUND: The immune status of a patient's tumor microenvironment (TME) may guide therapeutic interventions with cancer immunotherapy and help identify potential resistance mechanisms. Currently, patients' immune status is mostly classified based on CD8+tumor-infiltrating lymphocytes. An unmet need exists for comparable and reliable precision immunophenotyping tools that would facilitate clinical treatment-relevant decision-making and the understanding of how to overcome resistance mechanisms. METHODS: We systematically analyzed the CD8 immunophenotype of 2023 patients from 14 phase I-III clinical trials using immunohistochemistry (IHC) and additionally profiled gene expression by RNA-sequencing (RNA-seq). CD8 immunophenotypes were classified by pathologists into CD8-desert, CD8-excluded or CD8-inflamed tumors using CD8 IHC staining in epithelial and stromal areas of the tumor. Using regularized logistic regression, we developed an RNA-seq-based classifier as a surrogate to the IHC-based spatial classification of CD8+tumor-infiltrating lymphocytes in the TME. RESULTS: The CD8 immunophenotype and associated gene expression patterns varied across indications as well as across primary and metastatic lesions. Melanoma and kidney cancers were among the strongest inflamed indications, while CD8-desert phenotypes were most abundant in liver metastases across all tumor types. A good correspondence between the transcriptome and the IHC-based evaluation enabled us to develop a 92-gene classifier that accurately predicted the IHC-based CD8 immunophenotype in primary and metastatic samples (area under the curve inflamed=0.846; excluded=0.712; desert=0.855). The newly developed classifier was prognostic in The Cancer Genome Atlas (TCGA) data and predictive in lung cancer: patients with predicted CD8-inflamed tumors showed prolonged overall survival (OS) versus patients with CD8-desert tumors (HR 0.88; 95% CI 0.80 to 0.97) across TCGA, and longer OS on immune checkpoint inhibitor administration (phase III OAK study) in non-small-cell lung cancer (HR 0.75; 95% CI 0.58 to 0.97). CONCLUSIONS: We provide a new precision immunophenotyping tool based on gene expression that reflects the spatial infiltration patterns of CD8+ lymphocytes in tumors. The classifier enables multiplex analyses and is easy to apply for retrospective, reverse translation approaches as well as for prospective patient enrichment to optimize the response to cancer immunotherapy.

Tumor beta2-microglobulin and HLA-A expression is increased by immunotherapy and can predict response to CIT in association with other biomarkers.

Background: Downregulation of MHC class I expression and/or defects in the antigen presentation pathways are commonly reported in human cancers. Numerous studies previously have explored extensively the molecular mechanisms that underlie HLA-class I and Beta2-Microglobulin (B2M) downregulation. However, the techniques presently available to detect expression of MHC class I proteins lack the robustness, specificity and sensitivity needed for systematic integration and analysis in clinical trials. Furthermore, the dynamics of HLA-class I and B2M expression have not been comprehensively studied as a potential biomarker for immunotherapy. Methods: Using novel, validated, immunohistochemistry (IHC)-based methods for quantifying B2M and HLA-A in tumor samples from diverse cancer types, we have determined loss of B2M and HLA-A proteins in 336 archived, primary specimens and 329 biopsies from metastatic patients collected during Roche-sponsored Phase 1 clinical trials investigating novel immunotherapy candidates as monotherapy or in combination with CPI. Results: Up to 56% of cases with B2M or HLA-A loss were noted in the investigated tumor types. The frequency of loss was dependent on indication and stage of disease and revealed heterogeneous expression patterns across patients. B2M and HLA-A loss was increased in metastatic lesions compared to primary tumors, indicating selection of MHC class I low clones in metastatic and refractory tumor cells. High on-treatment B2M expression correlated with successful clinical outcome (RECIST), while high baseline B2M did not. A treatment-induced increase of B2M expression was noted in most of the patients with low B2M levels at baseline. The triple biomarker combination of B2M, CD8 and PDL1 strongly improved response prediction to cancer immunotherapy. Conclusion: Our results indicate that B2M and HLA-A loss occurs frequently in tumors and is reversed in most instances following immunotherapy which supports the conclusion that MHC class I loss is not the dominant resistance mechanism to CPI treatment. This investigation reveals a highly dynamic expression of HLA-A and B2M in tumors affected by indication, metastatic status, immunophenotype and immunotherapy treatment. Baseline expression levels of B2M on tumors may be of utility as a constituent of a biomarker panel used for selecting patients for immunotherapy clinical trials.

T-Cell Heterogeneity in Baseline Tumor Samples: Implications for Early Clinical Trial Design and Analysis.

Background: In early stage clinical trials, changes to levels of tumor infiltrating lymphocytes (TILs) in the tumor microenvironment (TME) are critical biomarkers of the mechanism of action of novel immunotherapies. However, baseline heterogeneity of tumor samples, both between and within patients, and the resultant impact on the validity of clinical trial data is not well defined. Here we identify and quantify the impact of baseline variables on the heterogeneity of FoxP3+ and proliferating CD8+ T-cells levels (MKi67+CD8A+) in the TME both between and within patients for the purpose of informing clinical trial design and analysis. Methods: We compared levels of FoxP3+ and MKi67+CD8+ cell densities (counts/mm2) from >1000 baseline tumor samples from clinical trials and commercially available sources. Using multivariate hierarchical regression techniques, we investigated whether inter-person heterogeneity of activated or regulatory T-cells could be attributed to baseline characteristics including demographics, indication, lesion type, tissue of excision, biopsy method, prior cancer treatment, and tissue type i.e., "fresh" or "archival" status. We also sought to characterize within-patient heterogeneity by lesion type and tissue type. Results: Prior cancer treatment with hormone therapy or chemotherapy that induces immunogenic cell death may alter the TME. Archival tissue is an unreliable substitute for fresh tissue for determining baseline TIL levels. Baseline and on treatment biopsies should be matched by lesion type to avoid bias.

Thymidine Metabolism as a Confounding Factor for 3'-Deoxy-3'-18F-Fluorothymidine Uptake After Therapy in a Colorectal Cancer Model.

Noninvasive monitoring of tumor therapy response helps in developing personalized treatment strategies. Here, we performed sequential PET and diffusion-weighted MRI to evaluate changes induced by a FOLFOX-like combination chemotherapy in colorectal cancer xenografts, to identify the cellular and molecular determinants of these imaging biomarkers. Methods: Tumor-bearing CD1 nude mice, engrafted with FOLFOX-sensitive Colo205 colorectal cancer xenografts, were treated with FOLFOX (5-fluorouracil, leucovorin, and oxaliplatin) weekly. On days 1, 2, 6, 9, and 13 of therapy, tumors were assessed by in vivo imaging and ex vivo analyses. In addition, HCT116 xenografts, which did not respond to the FOLFOX treatment, were imaged on day 1 of therapy. Results: In Colo205 xenografts, FOLFOX induced a profound increase in uptake of the proliferation PET tracer 3'-deoxy-3'-18F-fluorothymidine (18F-FLT) accompanied by increases in markers for proliferation (Ki-67, thymidine kinase 1) and for activated DNA damage response (γH2AX), whereas the effect on cell death was minimal. Because tracer uptake was unaltered in the HCT116 model, these changes appear to be specific for tumor response. Conclusion: We demonstrated that 18F-FLT PET can noninvasively monitor cancer treatment-induced molecular alterations, including thymidine metabolism and DNA damage response. The cellular or imaging changes may not, however, be directly related to therapy response as assessed by volumetric measurements.

Development of bispecific molecules for the in situ detection of protein-protein interactions and protein phosphorylation.

Investigation of protein-protein interactions (PPIs) and protein phosphorylation in clinical tissue samples can offer valuable information about the activation status and function of proteins involved in disease progression. However, existing antibody-based methods for phosphorylation detection have been found to lack specificity, and methods developed for examining PPIs in vitro cannot be easily adapted for tissues samples. In this study, we eliminated some of these limitations by developing a specific immunohistochemical staining method that uses "dual binders" (DBs), which are bispecific detection agents consisting of two Fab fragment molecules joined by a flexible linker, to detect PPIs and protein phosphorylation. We engineered DBs by selecting Fab fragments with fast off-rate kinetics, which allowed us to demonstrate that stable target binding was achieved only upon simultaneous, cooperative binding to both epitopes. We show that DBs specifically detect the activated HER2/HER3 complex in formalin-fixed, paraffin-embedded cancer cells and exhibit superior detection specificity for phospho-HER3 compared to the corresponding monoclonal antibody. Overall, the performance of DBs makes them attractive tools for future development for clinical applications.

A phase II pharmacodynamic study of erlotinib in patients with advanced non-small cell lung cancer previously treated with platinum-based chemotherapy.

PURPOSE: To examine potential markers of clinical benefit and the effects of erlotinib on the epidermal growth factor receptor (EGFR) signaling pathway in advanced non-small cell lung cancer patients refractory to platinum-based chemotherapy. EXPERIMENTAL DESIGN: Patients were given erlotinib (150 mg/d). Tumor biopsies were done immediately before treatment and in a subgroup of patients after 6 weeks' treatment. RESULTS: Of 73 evaluable patients, 7 (10%) had partial response and 28 (38%) had stable disease. In 53 patients with baseline tumor samples, no relationship was observed between pretreatment levels of EGFR, phosphorylated (p)-EGFR, p-AKT, p-mitogen-activated protein kinase (MAPK), or p27 and clinical benefit (i.e., response, or stable disease >/=12 weeks). Tumors from 15 of 57 patients had high EGFR gene copy number, assessed using fluorescence in situ hybridization (FISH positive), 10 of whom had clinical benefit, compared with 5 of 42 FISH-negative patients. FISH-positive patients had longer median progression-free [137 versus 43 days, P = 0.002; hazard ratio (HR), 0.37] and overall (226 versus 106 days, P = 0.267; HR, 0.70) survival than FISH-negative patients. In paired biopsy samples from 14 patients, p-EGFR (P = 0.002), p-MAPK (P = 0.001), and Ki-67 (P = 0.025) levels were significantly reduced after 6 weeks' treatment. Apoptosis was significantly increased in patients with clinical benefit (P = 0.029), and may be a marker of clinical benefit. CONCLUSION: In this study, EGFR FISH-positive status was associated with improved outcome after erlotinib therapy. Erlotinib led to reduced levels of p-EGFR, p-MAPK, and Ki-67, and stimulated apoptosis in tumor samples from patients with clinical benefit.

Peptide-beta2-microglobulin-MHC fusion molecules bind antigen-specific T cells and can be used for multivalent MHC-Ig complexes.

Recombinant soluble MHC molecules are widely used for visualization, activation and inhibition of antigen-specific immune responses. Using a genetic approach, we have generated two novel peptide-beta2-microglobulin-MHC constructs. We have linked the MHC molecule with the peptide of interest, without limiting the recognition by the cognate TCR. This molecule can also be joined with the IgG heavy chain resulting in a dimeric MHC-Ig fusion protein. These molecules bind antigen-specific T cells with high specificity and sensitivity, therefore, providing a valuable tool for detection as well as enrichment of antigen-specific T cells.

Characterization of chimeric enzymes between caprine arthritis--encephalitis virus, maedi--visna virus and human immunodeficiency virus type 1 integrases expressed in Escherichia coli.

In order to investigate the functions of the three putative lentiviral integrase (IN) protein domains on viral DNA specificity and target site selection, enzymatically active chimeric enzymes were constructed using the three wild-type IN proteins of caprine arthritis-encephalitis virus (CAEV), maedi-visna virus (MVV) and human immunodeficiency virus type 1 (HIV-1). The chimeric enzymes were expressed in Escherichia coli, purified by affinity chromatography and analysed in vitro for IN-specific endonuclease and integration activities on various DNA substrates. Of the 21 purified chimeric IN proteins constructed, 20 showed distinct site-specific cleavage activity with at least one substrate and six were able to catalyse an efficient integration reaction. Analysis of the chimeric IN proteins revealed that the central domain together with the C terminus determines the activity and substrate specificity of the enzyme. The N terminus appears to have no considerable influence. Furthermore, an efficient integration activity of CAEV wild-type IN was successfully demonstrated after detailed characterization of the reaction conditions that support optimal enzyme activities of CAEV IN. Also, under the same in vitro assay conditions, MVV and HIV-1 IN proteins exhibited endonuclease and integration activities, an indispensable prerequisite of domain-swapping experiments. Thus, the following report presents a detailed characterization of the activities of CAEV IN in vitro as well as the analysis of functional chimeric lentiviral IN proteins.