Single-cell spatial analysis may help predict response to neoadjuvant immunotherapy in triple-negative breast cancer

A next-generation technology that allows the study of protein expression at the single-cell level and the location of the cells within the tumor microenvironment (TME) was feasible and provided information on the benefit of adding the immune checkpoint inhibitor atezolizumab (Tecentriq) to chemotherapy as neoadjuvant treatment for patients with early high-risk and locally advanced triple negative breast cancer (TNBC), according to results presented at the San Antonio Breast Cancer Symposium, held December 7-10, 2021.

“We are experiencing a revolution in the technologies available for characterizing the molecular complexity of tumors,” said presenter Giampaolo Bianchini, MD, head of the Breast Cancer Group in the Department of Clinical Oncology at IRCCS Ospedale San Raffaele, Milan. “Among these, imaging mass cytometry allows us to collect unprecedented information about the heterogeneity of tumors and their surrounding microenvironment.”

Through imaging mass cytometry (IMC), it is possible to simultaneously analyze more than 40 markers in a single tissue section to identify the set of proteins present on individual cells, while accounting for their precise location within the tissue, Bianchini explained. IMC combines the principles of flow cytometry, which analyzes single cells or particles as they flow past single or multiple lasers, and mass spectrometry, which identifies the molecules present in a sample by accurately measuring their mass.

Emerging evidence has shown that TNBC tumors are infiltrated with mononuclear cells and lymphocytes. Combining immune checkpoint inhibition and chemotherapy demonstrated a significant benefit for high-risk TNBC patients in the KEYNOTE-522 trial, leading to the FDA approval of pembrolizumab (Keytruda) in combination with chemotherapy as neoadjuvant therapy in this setting.

“Unfortunately, one size does not fit all patients and it is possible that some of them may have responded to chemotherapy alone, while others who originally benefited from immunotherapy will eventually relapse. In addition, although immunotherapy is overall well tolerated, some rare but potentially serious immune-related side effects have been reported,” commented Bianchini. “For these reasons, biomarkers are urgently needed to help us identify the patients who will benefit the most from the addition of immunotherapy—potentially leading to chemotherapy de-escalation or chemo-free strategies, and those who will do well just with chemotherapy.”

Bianchini and colleagues investigated whether IMC could assist in the identification of ideal candidates for this therapeutic approach. They performed IMC analysis in the context of the phase III NeoTRIPaPDL1 trial, which was designed to evaluate the addition of atezolizumab (Tecentriq) to the chemotherapeutics carboplatin and nab-paclitaxel (Abraxane), compared with carboplatin and nab-paclitaxel only, as neoadjuvant therapy in patients with early high-risk and locally advanced TNBC who underwent surgery within six weeks of finishing the treatment.

“We assessed the predictive value of identifying the different phenotypes present within the tumor and the TME through single-cell analysis, and the relevance of the cell-cell interactions,” said H Raza Ali, MD, PhD, group leader at Cancer Research UK Cambridge Institute and University of Cambridge and a leading contributor to the study. “Physical interactions among cells are required for both immune activation and tumor cell killing, so information on the spatial organization of the tumor tissue is critical when studying the response to immunotherapy.”

The investigators successfully analyzed 43 proteins expressed on more than 1 million single cells identified in tissue samples collected through pre-treatment biopsies from 243 patients (representing 86.8 percent of the study population). For each sample, they generated three high-dimensional images that encompassed the tumor, tumor-stroma interface, and adjacent stroma. They investigated the association of protein expression on tumor and TME cells, cell phenotypes, and the spatial tissue organization with pathological complete response rate (pCR), defined as the absence of invasive cancer cells in tissue samples collected during surgery.

According to the authors, bulk protein expression analysis might deliver limited predictive information because it does not take into account the cell compartment in which each protein is expressed. For instance, assessment of Ki67 on TME cells and HLA-DR on epithelial cells seemed to provide more predictive information than the same biomarkers assessed in the whole tissue specimens.

By allowing for a precise identification of the different cell phenotypes, including cell type and functional state, this approach revealed the potential predictive role of the density of certain cell populations: high density of antigen presenting cells with high expression of PD-L1 and the immunosuppressive molecule IDO and of epithelial cells with high expression of the CD56 neuroendocrine marker was associated with higher pCR in patients who received atezolizumab plus chemotherapy but not in patients who only received chemotherapy.

In addition, high degree of spatial connectivity between epithelial cells and specific TME cells, for instance, CD8+ T cells with granzyme B or PD1 expression and features of exhaustion, correlated with a significant increase in the pCR rate after atezolizumab, whereas lower expression of these markers was associated with similar pCR rates between the atezolizumab arm and the chemotherapy only arm.

“Our results demonstrated that spatial data on the interactions among specific cells in the TME might be very informative about the benefit provided by an immune checkpoint inhibitor such as atezolizumab in addition to chemotherapy,” Bianchini commented. “This type of information can only be provided by technologies that allow us to simultaneously characterize the single cells and their spatial localization with precision.”

This approach also confirmed the extreme heterogeneity of TNBC, both in terms of tumor cell composition and in the amount, type, and functional state of the cells present in the TME.

“The predictive information we obtained through IMC complemented what can be derived with commonly used immune biomarkers such as PD-L1 expression or the amount of stromal tumor-infiltrating lymphocytes. In addition, we found that several immune-related gene expression signatures that capture immune cell types and function were less informative than the corresponding biomarkers assessed by IMC,” Bianchini said.

The complexity of the IMC technology led to questioning whether it could be applied to large series of tumor samples, such as those collected in routine practice. “In our study, we demonstrated that this disruptive technology can be successfully applied to samples prospectively collected in large clinical trials, paving the way for its broad implementation in cancer research to aid precision immunology,” added Bianchini.

According to the authors, all the findings of this study will require independent validation. In addition, a formal adjustment for multiple comparisons was not applied, calling for caution in the interpretation of the results. Finally, the reproducibility and applicability of this technology outside of the research setting must still be investigated.

This study was funded by Associazione Italiana per la Ricerca sul Cancro (AIRC), Cancer Research UK, The Breast Cancer Research Foundation, Fondazione Michelangelo, Fondazione Gianni Bonadonna, and an unrestricted grant from Roche and Celgene.

Bianchini has received honoraria from Amgen, AstraZeneca, Chugai, Daiichi Sankyo, EISAI, Exact Science, Gilead, Eli Lilly, MSD, Novartis, Pfizer, Roche, Sanofi, and Seagen outside of the present work.

Abstract

GS1-00

Single-cell spatial analysis by imaging mass cytometry and immunotherapy response in triple-negative breast cancer (TNBC) in the NeoTRIPaPDL1 trial

Background: Immune checkpoint inhibitors are effective in early and advanced TNBC, however only a minority of patients benefit making precision immune-oncology a major unmet need. Imaging mass cytometry (IMC) enables high dimensional tissue imaging at subcellular resolution for assessment of TNBC ecosystems, providing information on cell type composition, functional status, and spatial organisation.

Methods: InNeoTRIP patients with TNBC were randomized to eight cycles of nab-paclitaxel/carbo (CT) with/without atezolizumab (CTA). Forty-four proteins spanning cancer cells and the tumor microenvironment (TME) were assessed on pre-treatment biopsies (n=243/280; 86.8% evaluable after QC). FFPE samples were labelled with antibodies conjugated to isotopically pure rare earth metal reporters and profiled at one micron resolution by IMC. For each sample, we have generated three high dimensional images that encompass the tumor, tumor-stroma interface and adjacent stroma. We investigated the association of protein expression assessed separately for epithelial and TME cells, cell phenotypes, and spatial architectures with PD-L1 status (Ventana SP142), stromal TILs, TNBC types and pathological complete response (pCR). 237 patients (84.6%) have both IMC and RNA-seq available allowing for comparison with gene signatures derived from HALLMARK, ConsensusTME immune cell types, and Nanostring.

Results: Across 243 samples we identify just over one million single cells. By supervised clustering, we defined 37 robust cell phenotypes. PD-L1-positive tumors, high stromal TILs and TNBC type were characterized by extreme heterogeneity and unique cell-type and spatial TME composition. Several biomarkers demonstrated a significant test for interaction. Considering protein expression, GATA3 and CD20 on TME, HLA-DR in epithelial cells and Ki67 assessed both in epithelial and TME, had a significant test for interaction (p <0.05). For all these biomarkers, high expression (above median) was associated with an increase of pCR of >10% in favour of atezolizumab, whereas lower expression group demonstrated a similar pCR rate among arms. Two cell phenotypes, PD-L1+IDO+ antigen presenting cells (APCs) and CD56+ neuroendocrine (NE) epithelial cells, had a significant test for interaction. Higher expression of these biomarkers was associated with higher likelihood of pCR in CTA arm, but not in CT arm. For example, PD-L1+IDO+APCs in the CTA arm were associated with pCR proportions of 64.6% and 24.6% for above- and below-median groups respectively (OR4.5 [2.01-10.1], p<0.001). Most of these tests of interaction retained significance after adjustment by PD-L1status and stromal TILs. Notably, none among 61 gene-expression based immune-related pathways and 7 proliferation-related signatures demonstrated a significant test of interaction. Results of systematic multi-tiered image analysis at the levels of cell-cell interactions and recurrent higher order multicellular complexes defining TNBC ecosystems identified by graph-based methods will be presented at the meeting.

Conclusions: Imaging mass cytometry provides a more comprehensive overview of TNBC heterogeneity at a single-cell level with spatial resolution. Bulk protein or gene expression might deliver limited predictive information because it does not consider the cell compartment of expression. Precise cell phenotyping highlights the predictive role ofPD-L1+IDO+APCs and CD56+NE epithelial cells. Overall, we demonstrated that IMC is feasible in a large, randomized trial and provides independent predictive information on immune checkpoint inhibitors benefit to PD-L1, TILs and gene-expression profiles.

Contact:

Julia Gunther
American Association for Cancer Research
[email protected]
Office: 215-446-6896
Cell: 770-403-7690