Image-based multiplex immune profiling of cancer tissues: translational implications. A report of the International Immuno-oncology Biomarker Working Group on Breast Cancer

Recent advances in the field of immuno-oncology have brought transformative changes in the management of cancer patients. The immune profile of tumours has been found to have key value in predicting disease prognosis and treatment response in various cancers. Multiplex immunohistochemistry and immunofluorescence have emerged as potent tools for the simultaneous detection of multiple protein biomarkers in a single tissue section, thereby expanding opportunities for molecular and immune profiling while preserving tissue samples. By establishing the phenotype of individual tumour cells when distributed within a mixed cell population, the identification of clinically relevant biomarkers with high-throughput multiplex immunophenotyping of tumour samples has great potential to guide appropriate treatment choices. Moreover, the emergence of novel multi-marker imaging approaches can now provide unprecedented insights into the tumour microenvironment, including the potential interplay between various cell types. However, there are significant challenges to widespread integration of these technologies in daily research and clinical practice. This review addresses the challenges and potential solutions within a structured framework of action from a regulatory and clinical trial perspective. New developments within the field of immunophenotyping using multiplexed tissue imaging platforms and associated digital pathology are also described, with a specific focus on translational implications across different subtypes of cancer. © 2024 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

SORBET: Automated cell-neighborhood analysis of spatial transcriptomics or proteomics for interpretable sample classification via GNN

Spatially resolved transcriptomics or proteomics data have the potential to contribute fundamental insights into the mechanisms underlying physiologic and pathological processes. However, analysis of these data capable of relating spatial information, multiplexed markers, and their observed phenotypes remains technically challenging. To analyze these relationships, we developed SORBET, a deep learning framework that leverages recent advances in graph neural networks (GNN). We apply SORBET to predict tissue phenotypes, such as response to immunotherapy, across different disease processes and different technologies including both spatial proteomics and transcriptomics methods. Our results show that SORBET accurately learns biologically meaningful relationships across distinct tissue structures and data acquisition methods. Furthermore, we demonstrate that SORBET facilitates understanding of the spatially-resolved biological mechanisms underlying the inferred phenotypes. In sum, our method facilitates mapping between the rich spatial and marker information acquired from spatial 'omics technologies to emergent biological phenotypes. Moreover, we provide novel techniques for identifying the biological processes that comprise the predicted phenotypes.

Pitfalls in machine learning-based assessment of tumor-infiltrating lymphocytes in breast cancer: A report of the International Immuno-Oncology Biomarker Working Group on Breast Cancer

The clinical significance of the tumor-immune interaction in breast cancer is now established, and tumor-infiltrating lymphocytes (TILs) have emerged as predictive and prognostic biomarkers for patients with triple-negative (estrogen receptor, progesterone receptor, and HER2-negative) breast cancer and HER2-positive breast cancer. How computational assessments of TILs might complement manual TIL assessment in trial and daily practices is currently debated. Recent efforts to use machine learning (ML) to automatically evaluate TILs have shown promising results. We review state-of-the-art approaches and identify pitfalls and challenges of automated TIL evaluation by studying the root cause of ML discordances in comparison to manual TIL quantification. We categorize our findings into four main topics: (1) technical slide issues, (2) ML and image analysis aspects, (3) data challenges, and (4) validation issues. The main reason for discordant assessments is the inclusion of false-positive areas or cells identified by performance on certain tissue patterns or design choices in the computational implementation. To aid the adoption of ML for TIL assessment, we provide an in-depth discussion of ML and image analysis, including validation issues that need to be considered before reliable computational reporting of TILs can be incorporated into the trial and routine clinical management of patients with triple-negative breast cancer. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Spatial analyses of immune cell infiltration in cancer: current methods and future directions: A report of the International Immuno-Oncology Biomarker Working Group on Breast Cancer

Modern histologic imaging platforms coupled with machine learning methods have provided new opportunities to map the spatial distribution of immune cells in the tumor microenvironment. However, there exists no standardized method for describing or analyzing spatial immune cell data, and most reported spatial analyses are rudimentary. In this review, we provide an overview of two approaches for reporting and analyzing spatial data (raster versus vector-based). We then provide a compendium of spatial immune cell metrics that have been reported in the literature, summarizing prognostic associations in the context of a variety of cancers. We conclude by discussing two well-described clinical biomarkers, the breast cancer stromal tumor infiltrating lymphocytes score and the colon cancer Immunoscore, and describe investigative opportunities to improve clinical utility of these spatial biomarkers. © 2023 The Pathological Society of Great Britain and Ireland.

Abstract ED7-2: Multiplex spatial proteomic profiling

Multiplex spatial analysis is touted as a potential solution for tissue biomarkers for prognostic and companion diagnostic applications. Spatial analysis infers maintaining the location of a biomolecule so that both its amount and context can contribute to the information obtained. Immunohistochemistry (IHC) is the archetypal method of spatial profiling and the only application that has made it to the clinic. In this session, we will discuss technologies based on IHC, but that allow assessment of at least two, and as many as 100 or more proteins within the cell or molecular compartment that maintains spatial context. The simplest and most common approach is multiplex immunofluorescence, followed by a number of next generation techniques that use alternative labeling of the antibodies or cycling strategies or both to generate spatially informed data. Excluding Mass spectrometry-based methods, still in early stages, the highest “plex” method for protein is digital spatial profiling. Digital spatial profiling (DSP) uses a molecular method to define spatial compartments which is different from other high-plex methods that use cell segmentation, an image analysis method that largely depends on the presence of a cell nucleus in the image. As a result, the DSP method allows assessment of both cellular and non-cellular stromal compartments. DSP and other high-plex methods that will be discussed are thought to be best used as biomarker discovery tools, that might find biomarkers that can be measured with a more traditional approach for assessment in the clinic.

Citation Format: David Rimm. Multiplex spatial proteomic profiling [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr ED7-2.

The Effect of Black Cohosh on Ki67 expression and Tumor Volume: A Pilot Study of Ductal Carcinoma in Situ Patients

BACKGROUND

Black cohosh (BC) (Cimicifuga racemosa) may prevent and treat breast cancer through anti-proliferative, pro-apoptotic, anti-estrogenic, and anti-inflammatory effects. This study sought to evaluate the effect of BC on tumor cellular proliferation, measured by Ki67 expression, in a pre-operative window trial of ductal carcinoma in situ (DCIS) patients.

METHODS

Patients were treated pre-operatively for 2 to 6 weeks with BC extract. Eligible subjects were those who had DCIS on core biopsy. Ki67 was measured using automated quantitative immunofluorescence (AQUA) pre/post-operatively. Ki67, tumor volume, and hormone changes were assessed with 2-sided Wilcoxon signed-rank tests, α = .05.

RESULTS

Thirty-one patients were treated for an average of 24.5 days (median 25; range 15-36). Ki67 decreased non-significantly (n = 26; P = .20; median pre-treatment 1280, post-treatment 859; range pre-treatment 175-7438, post-treatment 162-3370). Tumor volume, estradiol, and FSH did not change significantly. No grade 3 or 4 adverse events were reported.

CONCLUSIONS

BC use showed no significant impact on cellular proliferation, tumor volume, or invasive disease upgrade rates in DCIS patients. It was well-tolerated, with no observed significant toxicities. Further study is needed to elucidate BC's role in breast cancer treatment and prevention.ClinicalTrials.gov Identifier: NCT01628536https://clinicaltrials.gov/ct2/show/NCT01628536.

Inhibition of renalase drives tumour rejection by promoting T cell activation

BACKGROUND

Although programmed cell death protein 1 (PD-1) inhibitors have revolutionised treatment for advanced melanoma, not all patients respond. We previously showed that inhibition of the flavoprotein renalase (RNLS) in preclinical melanoma models decreases tumour growth. We hypothesised that RNLS inhibition promotes tumour rejection by effects on the tumour microenvironment (TME).

METHODS

We used two distinct murine melanoma models, studied in RNLS knockout (KO) or wild-type (WT) mice. WT mice were treated with the anti-RNLS antibody, m28, with or without anti-PD-1. 10X single-cell RNA-sequencing was used to identify transcriptional differences between treatment groups, and tumour cell content was interrogated by flow cytometry. Samples from patients treated with immunotherapy were examined for RNLS expression by quantitative immunofluorescence.

RESULTS

RNLS KO mice injected with wild-type melanoma cells reject their tumours, supporting the importance of RNLS in cells in the TME. This effect was blunted by anti-cluster of differentiation 3. However, MØ-specific RNLS ablation was insufficient to abrogate tumour formation. Anti-RNLS antibody treatment of melanoma-bearing mice resulted in enhanced T cell infiltration and activation and resulted in immune memory on rechallenging mice with injection of melanoma cells. At the single-cell level, treatment with anti-RNLS antibodies resulted in increased tumour density of MØ, neutrophils and lymphocytes and increased expression of IFNγ and granzyme B in natural killer cells and T cells. Intratumoural Forkhead Box P3 + CD4 cells were decreased. In two distinct murine melanoma models, we showed that melanoma-bearing mice treated with anti-RNLS antibodies plus anti-PD-1 had superior tumour shrinkage and survival than with either treatment alone. Importantly, in pretreatment samples from patients treated with PD-1 inhibitors, high RNLS expression was associated with decreased survival (log-rank P = 0.006), independent of other prognostic variables.

CONCLUSIONS

RNLS KO results in melanoma tumour regression in a T-cell-dependent fashion. Anti-RNLS antibodies enhance anti-PD-1 activity in two distinct aggressive murine melanoma models resistant to PD-1 inhibitors, supporting the development of anti-RNLS antibodies with PD-1 inhibitors as a novel approach for melanomas poorly responsive to anti-PD-1.

Abstract P5-13-26: The future of HER2-positive breast cancer patients might be written in miRNAs: An exploratory analysis from the NeoALTTO study

Importance. Neoadjuvant therapy with dual HER2 blockade improved pathological complete response (pCR) rate. Nevertheless, it would be desirable to identify patients exquisitely responsive to single agent trastuzumab to minimize or avoid overtreatment. Objective. To evaluate the predictive and prognostic value of basal primary tumor miRNA expression profile within the trastuzumab arm of the NeoALTTO study.Design, Setting and Participants. RNA samples from baseline biopsies were randomized into training (n =45) and testing (n =47) sets. After normalization, miRNAs associated with Event-free survival (EFS) and pCR were identified by univariate analysis. Multivariate models were implemented to generate specific signatures which were first confirmed, and then analyzed according to other clinical and pathological variables.Main outcomes and measures. Association between miRNA expression and pCR and/or EFS.Results. We identified a prognostic signature including hsa-miR-153-3p (HR 1.831, 95%CI: 1.34-2.50) and hsa-miR-219a-5p (HR 0.629, 95%CI: 0.50 - 0.78). For two additional miRNAs (miR-215-5p and miR-30c-2-3p), we found a statistically significant interaction term with pCR (p.interaction: 0.017 and 0.038, respectively). Besides, a two-miRNA signature was predictive of pCR (hsa-miR-31-3p, OR 0.70, 95%CI: 0.53 - 0.92, and hsa-miR-382-3p, OR: 1.39, 95%CI: 1.01 -1.91). Notably, the performance of this predictive miRNA signature resembled that of the genomic classifiers PAM50 and TRAR, and did not improve when the extended models were fitted.Conclusions and relevance. Analysis of primary tumor tissue miRNAs holds the potential of a parsimonious tool to identify patients with differential clinical outcomes after trastuzumab based neoadjuvant therapy. Trial registration. ClinicalTrials.gov Identifier: NCT00553358.

Table 1.MiRNAs associated with EFS. Results of the univariate Cox regression model- training set.miRNAsHR95% CIhsa-miR-12000.671(0.512; 0.879)ahsa-miR-1238-3p0.669(0.543; 0.826)ahsa-miR-12650.839(0.705; 0.997)ahsa-miR-129-5p0.785(0.628; 0.980)ahsa-miR-153-3p1.412(1.026; 1.943)ahsa-miR-15390.826(0.697; 0.979)ahsa-miR-1908-5p0.796(0.648; 0.978)ahsa-miR-205-5p0.737(0.553; 0.982)hsa-miR-219a-5p0.787(0.627; 0.989)ahsa-miR-25-5p0.566(0.391; 0.819)ahsa-miR-3000.581(0.399; 0.845)ahsa-miR-382-3p0.791(0.636; 0.985)ahsa-miR-4920.774(0.635; 0.944)ahsa-miR-519a-3p0.834(0.696; 0.998)ahsa-miR-551b-5p0.666(0.452; 0.981)ahsa-miR-5830.727(0.549; 0.963)ahsa-miR-6001.373(1.03; 1.829)ahsa-miR-6260.819(0.673; 0.998)ahsa-miR-651-5p0.611(0.382; 0.977)hsa-miR-7610.845(0.721; 0.991)ahsa-miR-891b0.658(0.479; 0.904)ahsa-miR-92a-2-5p0.787(0.637; 0.971)ahsa-miR-937-3p0.748(0.578; 0.967)aHR, Hazard Ratio; CI, Confidence Intervala microRNAs retained its statistically significance (at 10%) also when normalized for the overall mean.

Table 2.MiRNAs associated with pCR. Results of the univariate logistic model - training set.miRNAsOR95% CIhsa-miR-132-3p0.410(0.169; 0.991)ahsa-miR-23b-5p0.520(0.316; 0.856)ahsa-miR-31-3p0.566(0.351; 0.912)ahsa-miR-31-5p0.508(0.291; 0.887)ahsa-miR-330-3p0.501(0.251; 0.999)hsa-miR-34b-3p0.673(0.474; 0.956)ahsa-miR-382-3p1.392(1.006; 1.925)ahsa-miR-548j-5p1.702(1.090; 2.658)aOR, Odds Ratio; CI, Confidence Interval. a microRNAs retained its statistically significance (at 10%) also when normalized for the overall mean. 2

Citation Format: Marilena Valeria Iorio, Sara Pizzamiglio, Giulia Cosentino, Chiara M Ciniselli, Loris De Cecco, Alessandra Cataldo, Ilaria Plantamura, Tiziana Triulzi, Sarra El-abed, Yingbo Wang, Mohammed Bajji, Paolo Nuciforo, Jens Huober, Susan Ellard, David Rimm, Andrea Gombos, Mariagrazia Daidone, Paolo Verderio, Elda Tagliabue, Serena Di Cosimo. The future of HER2-positive breast cancer patients might be written in miRNAs: An exploratory analysis from the NeoALTTO study [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P5-13-26.

Artificial intelligence applied to breast pathology

The convergence of digital pathology and computer vision is increasingly enabling computers to perform tasks performed by humans. As a result, artificial intelligence (AI) is having an astoundingly positive effect on the field of pathology, including breast pathology. Research using machine learning and the development of algorithms that learn patterns from labeled digital data based on "deep learning" neural networks and feature-engineered approaches to analyze histology images have recently provided promising results. Thus far, image analysis and more complex AI-based tools have demonstrated excellent success performing tasks such as the quantification of breast biomarkers and Ki67, mitosis detection, lymph node metastasis recognition, tissue segmentation for diagnosing breast carcinoma, prognostication, computational assessment of tumor-infiltrating lymphocytes, and prediction of molecular expression as well as treatment response and benefit of therapy from routine H&E images. This review critically examines the literature regarding these applications of AI in the area of breast pathology.

240 Discovery of biomarkers of resistance to immune checkpoint blockade in non-small-cell lung cancer (NSCLC) using high-plex digital spatial profiling

Background

Despite the clinical effectiveness of Immune Checkpoint Inhibitors (ICI) in lung cancer, only around 20% remain disease free at 5 years. Predictive biomarkers for ICIs are neither sensitive nor specific. Here, we used the GeoMx Digital Spatial Profiler (DSP) (NanoString, Inc.) to analyze high-plex protein in a quantitative and spatially resolved manner from single formalin-fixed paraffin embedded tissue sections toward the goal of identification of new biomarkers with better predictive value.

Methods

Pre-treatment samples from 56 patients with NSCLC treated with ICI were collected, represented in Yale tissue microarray 471 (YTMA471), and analyzed. A panel of 71 photocleavable oligonucleotide-labeled primary antibodies (NanoString Human IO panel) was used for protein detection. Protein expression was measured in 4 molecularly defined tissue compartments, defined by fluorescence co-localization (tumor [panCK+], leukocytes [CD45+/CD68-], macrophages [CD68+] and an aggregate stromal immune cell compartment, defined as the sum of leukocyte and macrophage expression [panCK-/CD45+/CD68+]) generating 284 variables representing potential predictive biomarkers. Promising candidates were orthogonally validated with Quantitative Immunofluorescence (QIF). Pre-treatment samples from 40 patients with NSCLC (YTMA404) that received ICI, and 174 non-ICI treated operable NSCLC patients (YTMA423) were analyzed to provide independent cohort validation. All statistical testing was performed using a two-sided significance level of α=0.05 and multiple testing correction (Benjamini-Hochberg method, FDR < 0.1).

Results

Initial biomarker discovery on 284 protein variables were generated by univariate analysis using continuous log-scaled data. High PD-L1 expression in tumor cells predicted longer survival (PFS; HR 0.67, p=0.017) and validated the training cohort. We found 4 markers associated with PFS, and 3 with OS in the stromal compartment. Of these, expression of CD66b in stromal immune cells predicted significantly shorter OS (HR 1.31, p=0.016) and shorter PFS (HR 1.24, p = 0.04). Tertile analysis using QIF on all three tissue cohorts for CD66b expression, assessed by QIF, showed that CD66b was indicative but not prognostic for survival [discovery cohort, YTMA471 (OS; HR 3.02, p=0.013, PFS; HR 2.38, p=0.023), validation cohort; YTMA404 (OS; HR 2.97, p=0.018, PFS; HR 1.85, p=0.1), non-ICI treated cohort YTMA423 (OS; HR 1.02, p>0.9, PFS; HR 0.72, p=0.4)].

Conclusions

Using the DSP technique, we have discovered that CD66b expressed in the stromal immune [panCK-/CD45+/CD68+] molecular compartment is associated with resistance to ICI therapy in NSCLC. This observation was validated by an orthogonal approach in an independent ICI treated NSCLC cohort. Since CD66b identifies neutrophils, further studies are warranted to characterize the role of neutrophils in ICI resistance.

Acknowledgements

Dr Moutafi is supported by a scholarship from the Hellenic Society of Medical Oncologists (HESMO)

Ethics Approval

All tissue samples were collected and used under the approval from the Yale Human Investigation Committee protocol #9505008219 with an assurance filed with and approved by the U.S. Department of Health and Human Services

52 Characterization of the tumor microenvironment in melanoma using Multiplexed Ion Beam Imaging (MIBI)

Background

The complexity of the tumor microenvironment (TME) necessitates the application of high-dimensional methods that can spatially resolve the phenotypic heterogeneity that exists within tumors. MIBI, which combines time-of-flight secondary ion mass spectrometry (ToF-SIMS) with metal labeled antibodies to simultaneously image 40+ proteins at subcellular spatial resolution, was used to classify cell populations and their expression of immunoregulatory proteins within 54 melanoma samples.

Methods

A tissue microarray (TMA) comprised of 0.6 mm FFPE melanoma cores was stained with a panel of 30 metal labeled antibodies. The tissue was imaged using MIBI and multi-step processing was performed to create images of the samples. Single cell segmentation enabled enumeration of 32 cell populations and quantitative analyses were performed of both immune checkpoint expression and the spatial relationships between cells of different types.

Results

Tumor cells and immune cells represented 62% (1.4% - 92.2%) and 24% (2.0% - 92.1%), respectively, of the segmented cells from the melanoma samples. Fibroblasts, lymphatics, and blood vessels were also present at varying densities. Within the immune compartment, M2 macrophages, M2 monocytes, and monocyte-derived dendritic cells (mDCs) were most abundant, representing 36.3% (4.9% - 79.3%), 7.0% (0.0% - 30.6%), and 9.8% (0.0% - 34.6%), respectively, of the total immune cell infiltrate. Rare populations such as myeloid-derived suppressor cells (MDSCs) and regulatory T cells were at greater than 100 cells/mm2 in six samples and four samples, respectively. The presence of immune checkpoint markers (IDO-1, LAG3, PD-1, PD-L1, TIM-3) varied between populations and between samples. A minority of the samples showed expression of IDO-1 and PD-L1 on myeloid populations. Interestingly, among the myeloid subsets, M2 macrophages and monocyte-derived dendritic cells showed the most abundant PD-L1 expression. Although many samples had few T cell infiltrates, those that did showed expression of PD-1, LAG-3, and TIM-3 on T cell populations.

Conclusions

MIBI offers high-parameter tissue imaging, at sensitivity and resolution suited to understanding the complex tumor immune landscape, including the spatial relationships of immune and tumor cells and the expression of immunoregulatory proteins. Datasets that quantify population densities and immune checkpoint expression levels across tumor samples, such as the present one, can be used to test for associations to clinical variables and further understand the TME at the cellular level.

Interobserver Agreement of PD-L1/SP142 Immunohistochemistry and Tumor-Infiltrating Lymphocytes (TILs) in Distant Metastases of Triple-Negative Breast Cancer: A Proof-of-Concept Study. A Report on Behalf of the International Immuno-Oncology Biomarker Working Group

Patients with advanced triple-negative breast cancer (TNBC) benefit from treatment with atezolizumab, provided that the tumor contains ≥1% of PD-L1/SP142-positive immune cells. Numbers of tumor-infiltrating lymphocytes (TILs) vary strongly according to the anatomic localization of TNBC metastases. We investigated inter-pathologist agreement in the assessment of PD-L1/SP142 immunohistochemistry and TILs. Ten pathologists evaluated PD-L1/SP142 expression in a proficiency test comprising 28 primary TNBCs, as well as PD-L1/SP142 expression and levels of TILs in 49 distant TNBC metastases with various localizations. Interobserver agreement for PD-L1 status (positive vs. negative) was high in the proficiency test: the corresponding scores as percentages showed good agreement with the consensus diagnosis. In TNBC metastases, there was substantial variability in PD-L1 status at the individual patient level. For one in five patients, the chance of treatment was essentially random, with half of the pathologists designating them as positive and half negative. Assessment of PD-L1/SP142 and TILs as percentages in TNBC metastases showed poor and moderate agreement, respectively. Additional training for metastatic TNBC is required to enhance interobserver agreement. Such training, focusing on metastatic specimens, seems worthwhile, since the same pathologists obtained high percentages of concordance (ranging from 93% to 100%) on the PD-L1 status of primary TNBCs.

Quantitative Image Analysis for Tissue Biomarker Use: A White Paper From the Digital Pathology Association

Tissue biomarkers have been of increasing utility for scientific research, diagnosing disease, and treatment response prediction. There has been a steady shift away from qualitative assessment toward providing more quantitative scores for these biomarkers. The application of quantitative image analysis has thus become an indispensable tool for in-depth tissue biomarker interrogation in these contexts. This white paper reviews current technologies being employed for quantitative image analysis, their application and pitfalls, regulatory framework demands, and guidelines established for promoting their safe adoption in clinical practice.

PARP inhibitors in head and neck cancer: Molecular mechanisms, preclinical and clinical data

Poly (ADP-ribose) polymerase (PARP) inhibitors (PARPi) have revolutionized the treatment landscape in several cancers. PARPi increase DNA damage particularly in tumors with underlying defects in DNA repair. In addition to PARPi-induced DNA damage, PARPi enhance immune priming and induce adaptive upregulation of programmed death ligand 1 (PD-L1) expression. Patients with head and neck squamous cell carcinoma (HNSCC) are characterized by aberrant DNA repair pathways, including nucleotide excision repair (NER), base excision repair (BER) and DNA double-strand breaks (DSBs) repair and these deregulated repair mechanisms are implicated in both the pathogenesis of the disease and the outcome of therapy. Cisplatin represents the cornerstone of treatment of HNSCC and cisplatin resistance impedes successful treatment outcomes. To this end, research strategies that are testing modulation of cisplatin sensitivity by PARPi are of particular interest. Moreover, given the immune modulating effects of PARPi and the recent approval of Programmed Cell Death- 1 (PD-1) checkpoint inhibitors in HNSCC, the design of trials combining PARPi and PD-1 checkpoint inhibitors represent a rational research strategy. In this review, we summarize data supporting the integration of PARP inhibitors into HNSCC therapeutic strategy.

Abstract PO-003: Deep learning identifies conserved pan-cancer tumor features

Histopathological images are an integral data type for studying cancer. We show pre-trained convolutional neural networks (CNNs) can be systematically applied across cancer types, enabling comparisons to reveal shared spatial behaviors. We develop CNNs with a common architecture trained on 19 cancer types of The Cancer Genome Atlas (TCGA), analyzing 14459 hematoxylin and eosin scanned frozen tissue images. Our CNNs are based on the Inception-V3 network and classify TCGA pathologist-annotated tumor/normal status of whole slide images in all 19 cancer types with consistently high AUCs (0.995±0.008). Remarkably, CNNs trained on one tissue are effective in others (AUC 0.88±0.11), with classifier relationships recapitulating known adenocarcinoma, carcinoma, and developmental biology. Moreover, classifier comparisons reveal intra-slide spatial similarities, with an average tile-level correlation of 0.45±0.16 between classifier pairs on the TCGA test sets. In particular, the TCGA-trained classifiers had average tile-level correlation of 0.52±0.09 and 0.58±0.08 on hold-out TCGA lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) test sets, respectively. These relations are reflected on two external datasets, i.e., LUAD and LUSC whole slide images of Clinical Proteomic Tumor Analysis Consortium. The CNNs trained on TCGA achieved cross-classification AUCs of 0.75±0.12 and 0.73±0.13 on LUAD and LUSC external validation sets, respectively. These CNNs had average tile-level correlations of 0.38±0.09 and 0.39±0.08 on LUAD and LUSC validation sets, respectively. Breast cancers, bladder cancers, and uterine cancers have spatial patterns that are particularly easy to detect, suggesting these cancers can be canonical types for image analysis. This study illustrates pre-trained CNNs can detect tumor features across a wide range of cancers, suggesting presence of pan-cancer tumor features. These shared features allow combining datasets when analyzing small samples to narrow down the parameter search space of CNN models.

Citation Format: Javad Noorbakhsh, Saman Farahmand, Ali Foroughi pour, Sandeep Namburi, Dennis Caruana, David Rimm, Mohammad Soltanieh-ha, Kourosh Zarringhalam, Jeffrey H. Chuang. Deep learning identifies conserved pan-cancer tumor features [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PO-003.

Abstract PS2-03: Comparison of pathologist reads of sp142 and sp263 with quantitative measurement of protein and mRNA in triple negative breast cancer

Background: PD-L1 SP142 immunohistochemistry (IHC) assay has been approved as a companion test by the US Food and Drug Administration (FDA) to identify eligibility for atezolizumab therapy in patients with advanced triple negative breast cancer (TNBC) but a number of studies suggest the assay suffers from poor reproducibility. Using readings of 70 TNBC chromogenic FDA approved assays from 19 pathologists in a previous study as a baseline, we compared pathologist reads to quantitatively measured mRNA and protein expression Methods: Formalin-fixed paraffin-embedded (FFPE) slides representing primary invasive triple negative breast cancer (stage I-III) from 100 patients between 2012-16 were selected from the Yale Pathology archives. Slides were macrodissected to tumor enrichment for quantitative assessment of CD274 (PD-L1 mRNA) measured using a closed-system, real-time quantitative reverse transcription polymerase chain reaction (RT-qPCR) research use only (RUO)* prototype assay on the GeneXpert® instrument. We also measured protein expression levels using the AQUA method of quantitative immunofluorescence (QIF) in both the tumor and non-tumor compartments on full sections by QIF stained using SP142 in a lab derived test (LDT). The IHC stained slides were prepared using SP142 and SP263 assays prepared exactly according the FDA approved label followed by reading by 19 pathologists. This study was approved by Yale Human Investigation IRB protocol ID 9505008219.Results: Previous work from our group showed overall percent agreement for both the SP142 and SP263 IHC assays read by pathologists was in the 40-50% range. We used the median CD274 score to compare positive (IC≥1) vs negative (IC&lt;1) and found that the levels of mRNA were not statistically significantly different between the two categorical scores. However, quantitative measurement of protein expression (including both tumor and non-tumor regions) showed statistically significant differences between pathologist read PD-L1 positive/negative scoring for SP142 (p=0.0004) and SP263 (p=0.0185). Concordance of quantitative PD-L1 measurement between protein QIF scores and transcript RT-qPCR levels was modest, with a Spearman coefficient r = 0.18. Conclusions: By chromogenic IHC, using the FDA approved assays, pathologist read scoring shows no difference to mRNA for CD274. Quantitative continuous scoring of protein expression show that, on average, when pathologists score IC≥1, there is more protein present than when they score IC&lt;1. Further studies are needed to determine if RNA and protein for PD-L1 are concordant and to determine which assay(s) and cutoff values are best correlated with clinical outcomes on atezolizumab therapy.

*For Research Use Only - Not for use in diagnostic procedures. Not approved or reviewed by any regulatory body

Citation Format: Swati Gupta, Vesal Yaghoobi, Aileen Fernandez, Leena McCann, Jodi Weidler, Michael Bates, Emily Reisenbichler, David Rimm. Comparison of pathologist reads of sp142 and sp263 with quantitative measurement of protein and mRNA in triple negative breast cancer [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PS2-03.

Abstract PS5-08: Comparison of PD-L1 protein expression between primary tumors and metastatic lesions in triple negative breast cancers

Background: Several recent studies that compared small cohorts of metastatic and primary lesions, suggested substantial heterogeneity in tumor infiltrating lymphocyte count, immune gene expression and PD-L1 protein expression across different metastatic sites and between primary breast cancers and metastasis. Understanding the frequency of PD-L1 positivity rates across different tissue sites can indicate differences in the immune microenvironment and may also guide biopsy site selection. We compared PD-L1 positivity on immune cells and tumor cells in primary and metastatic triple negative breast cancer tumors (TNBC).Methods: A retrospective data analysis of the Foundation Medicine PD-L1 IHC database was conducted on 340 cases of TNBC. PD-L1 positivity was determined by IHC using SP142CDx. Results are reported as percent of PD-L1 stained immune cells (IC) in the tumor area. A tumor was considered PD-L1 positive if ≥ 1% IC stained positive with PD-L1. As an exploratory analysis, PD-L1 positivity of tumor cells (TC) was also assessed. PD-L1 percent positive staining results are reported as means with 95% CI. The proportion of PD-L1 positive and negative IC and TC in primary tumors vs metastatic sites was compared using Chi-Square test. Prism 8 was used for all data analysis.Results: All patients were female, with median age 56 years (range 26-89); 179 samples were from primary tumors and 161 from metastatic lesions, representing 15 different tissue sites. Overall, PD-L1 expression on immune cells was statistically significantly more frequent in primary tumors compared to metastatic sites (63.7% [n=114] vs 42.9% [n=69]), p&lt;0.0001). This was driven by the lower PD-L1 positivity rates in skin (23.8%, 95% CI: 8.22% - 47.2%), liver (17.4%, 95%CI: 5.00% - 38.8%) and bone (16.7%, 95%CI: 2.10% - 48.4%) metastases. Lung (68.8%, 95% CI: 41.3 - 90.0), soft tissues (65.2%, 95% CI: 42.7 - 83.6) and lymph nodes 51.1%, 95% CI (35.8 - 66.3) had PD-L1 % positivity rates similar to primary tumors. PD-L1 expression was rare on tumor cells in both the breast and metastatic sites (8.3% vs 4.3%, p=0.13). Conclusion: We observed substantial heterogeneity in PD-L1 positivity rates across metastatic sites. Lung, soft tissues and lymph node metastases had PD-L1 % positivity rates that were similar to that of primary tumors whereas skin, liver and bone metastases had significantly lower PD-L1 % positivity rates. These results raise the possibility that response to immune therapy could depend on the location and the PD-L1 positivity of the metastatic site. Limited current experience in breast cancer is not sufficient to correlate tumor response with PD-L1 expression in metastases, but as more patients receive treatment, this could be examined in the future.

Table 1: Sample Characteristics and % PD-L1 positivity on immune cellsSample typeTotal N (%)N PD-L1 positive (%, 95% CI)Primary Tumor179 (52.6)114 (63.7%, 56.2% - 70.7%)Metastatic Lesion161 (47.4)69 (42.9%, 35.1% - 50.9%)Sites of MetastasesN (% of metastatic samples)N PD-L1 positive (%, 95% CI)Lung16 (10.0)11 (68.8%, 41.3% - 90.0%)Soft Tissues23 (14.3)15 (65.2%, 42.7% - 83.6%)Lymph Nodes45 (28.0)23 (51.1%, 35.8% - 66.3%)Skin21 (13.0)5 (23.8%, 8.22% - 47.2%)Liver23 (14.3)4 (17.4%, 5.00% - 38.8%)Bone12 (7.5)2 (16.7%, 2.10% - 48.4%)Brain9 (5.6)5Mediastinum4 (2.5)1Pleura2 (1.2)0Muscle1 (&lt;1)0Omentum1 (&lt;1)1Ovary1 (&lt;1)0Pelvis1 (&lt;1)0Retroperitoneum1 (&lt;1)0Adrenal Gland1 (&lt;1)1

Table 2: Comparison of PD-L1 positivity in primary versus metastatic sitesTissuePDL1+ Immune CellPDL1- Immune CellP valuePDL1+ Tumor CellPDL- Tumor CellP valuePrimary114650.0001151640.1313Metastasis69927154

Citation Format: Mariya Rozenblit, Richard Huang, Natalie Danziger, Brian Alexander, Shakti Ramkissoon, Kim Blenman, Jeffrey Ross, David Rimm, Lajos Pusztai. Comparison of PD-L1 protein expression between primary tumors and metastatic lesions in triple negative breast cancers [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PS5-08.

Biomarkers in Precision Cancer Immunotherapy: Promise and Challenges

The rapid expansion of modern cancer immunotherapeutics has led to a dramatic improvement in patient survival and sustained remission for otherwise refractory malignancies. However, a significant limitation behind these current treatment modalities is an irregularity in clinical response, which is especially pronounced among checkpoint inhibition. This unpredictability leads to significant side effects, financial costs, and health care burden, with unsatisfactory clinical benefit in the majority of treated patients. Additionally, although ongoing studies and trials investigate the use of multiple biomarkers predictive of patient response or harm, none of these are comprehensive in predicting potential benefit. This unmet need for validated biomarkers is largely secondary to a prohibitive complexity within tumor parenchyma and microenvironment, dynamic clonal and proteomic changes to therapy, heterogenous host immune defects, and varied standardization among sample preparation and reporting. Herein, we discuss current advantages of predictive biomarkers, as well as limitations in their clinical use and application. We also review future directions, ideal characteristics, and trial design needed for proper precision immuno-oncology and biomarker development.

Deep learning-based cross-classifications reveal conserved spatial behaviors within tumor histological images

Histopathological images are a rich but incompletely explored data type for studying cancer. Manual inspection is time consuming, making it challenging to use for image data mining. Here we show that convolutional neural networks (CNNs) can be systematically applied across cancer types, enabling comparisons to reveal shared spatial behaviors. We develop CNN architectures to analyze 27,815 hematoxylin and eosin scanned images from The Cancer Genome Atlas for tumor/normal, cancer subtype, and mutation classification. Our CNNs are able to classify TCGA pathologist-annotated tumor/normal status of whole slide images (WSIs) in 19 cancer types with consistently high AUCs (0.995 ± 0.008), as well as subtypes with lower but significant accuracy (AUC 0.87 ± 0.1). Remarkably, tumor/normal CNNs trained on one tissue are effective in others (AUC 0.88 ± 0.11), with classifier relationships also recapitulating known adenocarcinoma, carcinoma, and developmental biology. Moreover, classifier comparisons reveal intra-slide spatial similarities, with an average tile-level correlation of 0.45 ± 0.16 between classifier pairs. Breast cancers, bladder cancers, and uterine cancers have spatial patterns that are particularly easy to detect, suggesting these cancers can be canonical types for image analysis. Patterns for TP53 mutations can also be detected, with WSI self- and cross-tissue AUCs ranging from 0.65-0.80. Finally, we comparatively evaluate CNNs on 170 breast and colon cancer images with pathologist-annotated nuclei, finding that both cellular and intercellular regions contribute to CNN accuracy. These results demonstrate the power of CNNs not only for histopathological classification, but also for cross-comparisons to reveal conserved spatial behaviors across tumors.

Abstract PD1-01: Durvalumab (MEDI4736) concurrent with nab-paclitaxel and dose dense doxorubicin cyclophosphamide (ddAC) as neoadjuvant therapy for triple negative breast cancer (TNBC)

Background: The goal of this Phase I/II trial (NCT02489448) was to assess the safety and efficacy of concurrent durvalumab with weekly nab-paclitaxel (100 mg/m2) x 12 followed by ddAC x 4 as neoadjuvant therapy for stage I-III TNBC and to identify biomarkers of response. The primary efficacy endpoint was pathologic complete response (pCR: ypT0,is/N0). Methods: The Phase I portion of the trial assessed two dose levels of durvalumab 3 and 10 mg/kg q 2 weeks. The trial followed Simon’s two step design, with early stopping for futility if &lt; 7 of the first 20 patients achieve pCR. PD-L1 expression on pretreatment biopsies was assessed with chromogenic immunohistochemistry using the SP263 antibody. PD-L1 positivity was determined by consensus review of 2 pathologists (E.R., D.R.) and staining &gt;1 % on immune and tumor cells was considered positive. Tumor infiltrating lymphocyte (TIL) count was assessed on H&E stained slides using QuPath v0.2.0 open source digital image analysis software platform and a breast cancer specific scoring algorithm (CL11NN). TIL count was expressed as TIL:TIL+Tumor cells x 100. Results: 57 patients were enrolled and evaluable (n=4 at 3 mg/kg, n=53 at 10 mg/kg dose). No dose limiting toxicities were observed during the Phase I portion, the final pCR rate is 44% (95% CI:30%-57%). 18 patients (31%) experienced grade 3/4 adverse events (AE), most frequently neutropenia (n=5). Possibly immune related grade 3 or 4 AEs included Guillain-Barre syndrome (n=1), hypothyroidism (n=1), colitis (n=1), hyperglycemia (n=1). 14 (24%) patients received &lt; 9 of the planned 12 cycles of durvalumab. No perioperative adverse events were seen. Fifty patients had baseline PD-L1 IHC results available (n=7 QC failure), 19 (38%) were PD-L1 positive. The pCR rates were 55% (95% CI: 36%-73%) versus 21% (95% CI: 6%-45%) in the PDL-1 positive and negative groups, respectively (p=0.03). Digital stromal TIL counts were available on 52 patients, there was no significant difference in TIL count between the response groups. Conclusion: Concomitant administration of durvalumab with weekly nab-paclitaxel and sequential ddAC neoadjuvant chemotherapy resulted in a pCR rate of 44%. pCR rate was higher in PD-L1 positive patients (55%) than PD-L1 negative (21%) cancers.

Citation Format: Lajos Pusztai, Emily Reisenbichler, Yailai Bai, Neal Fischbach, Justin Persico, Kerin Adelson, Anamika Katoch, Nina Horowitz, Donald Lannin, Brigid Killelea, Anees Chagpar, Courtney Frederick, Trisha Burello, Kim Blenman, David Rimm, Andrea Silber. Durvalumab (MEDI4736) concurrent with nab-paclitaxel and dose dense doxorubicin cyclophosphamide (ddAC) as neoadjuvant therapy for triple negative breast cancer (TNBC) [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr PD1-01.

Abstract P3-09-05: Predictive markers of response to durvalumab concurrent with nab-paclitaxel and dose dense doxorubicin cyclophosphamide (ddAC) neoadjuvant therapy for triple negative breast cancer (TNBC)

Background: The goal of this analysis was to identify immune cell types and mRNA expression signatures that are associated with pathologic complete response (pCR: ypT0/N0) to neoadjuvant durvalumab concurrent with weekly nab-paclitaxel (100 mg/m2) x 12 followed by ddAC x 4 chemotherapy in stage I-III triple negative breast cancer (TNBC).

Methods: Pre-treatment core needle biopsies were obtained from 57 patients (n=25 pCR, n=32 residual disease (RD)) who participated in Phase I/II clinical trial (NCT02489448) for RNA sequencing and histology assessment. Formalin-fixed paraffin-embedded (FFPE) core biopsies were stained by immunofluorescence for CD8, CD68 and cytokeratin, images were acquired on the Vectra-Polaris system and analyzed using InForm software with the adaptive segmentation into three compartments: Tumor, Stroma, and All (Tumor + Stroma). Correlation between immune cell density in the tissue compartments and pCR was assessed. RNA was extracted from core biopsies collected into RNAlater and poly-A enriched mRNAs were sequenced on Illumina platform using NovaSeq paired-end, 100bp fragments, with a depth of 50 million reads. Six specimens failed QC due to insufficient tumor cells in the sample. Correlation between pCR and immune markers and previously published immune, proliferation and DNA damage response deficiency (DDRD) gene signatures were assessed using logistic regression. Due to high overlap of genes across signatures, no multiple testing correction was done. Differentially expressed genes were found using DESeq2 R package with Benjamini-Hochberg correction for multiple testing. Broadside, an interaction mining tool, was used to extract sets of combinations of genes associated with pathologic response.

Results: In cases with pCR, we detected significantly higher overall CD8 cell density and a trend for higher CD8 in tumor and stroma compartments. There was no significant difference in CD68 cell density between response groups. At the RNA level, high expression of DDRD, IFNγ, T-cell, B-cell, dendritic cell, M1 macrophage signatures and the tumor inflammation gene signature were significantly associated with higher rate of pCR. High expression of epithelial mesenchymal transition and TGFβ and IL8/VEGF gene signatures were associated with higher rate of residual disease. Proliferation gene signatures were not associated with response in this TNBC population. At individual gene level, the highest pCR predictive value was observed for the STAT1 + PCF11 and LAMP3 + SDR39U1 doublets.

Conclusion: High CD8 cell density, high expression of a broad range of immune gene expression signatures and DNA damage response deficiency are associated with greater sensitivity to neoadjuvant anti-PD-L1 and chemotherapy.

Citation Format: Kim RM Blenman, Xiaotong Li, Michal Marczyk, Tess O'Meara, Vesal Yaghoobi, Vignesh Gunasekharan, Tristen Park, David Rimm, Lajos Pusztai. Predictive markers of response to durvalumab concurrent with nab-paclitaxel and dose dense doxorubicin cyclophosphamide (ddAC) neoadjuvant therapy for triple negative breast cancer (TNBC) [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P3-09-05.

Abstract 2842: Vesal Yaghoobi

Background: It has been shown that even after adjusting for clinical stage at presentation and socio-economic variables, Triple Negative Breast Cancer (TNBC) still has a worse overall outcome in African American (AA) compared to Non-African American (NAA) patients. The abundance and composition of immune cells in the tumor microenvironment is a powerful prognostic factor in TNBC. Our goal was to assess if differences exist in CD8, CD68 and PD-L1 protein expression between AA and NAA TNBC. We hypothesize that the microenvironment of African AA TNBC patients may be different than that of NAAs.

Method: We selected N=43 AA and N=43 NAA TNBC samples from the Yale Pathology archives that were matched by diagnosis date. We measured CD8, CD68 and PD-L1 protein expression in both the tumor and stromal compartments in whole slides from formalin fixed paraffin embedded (FFPE) tissues using multiplexed quantitative immunofluorescence (QIF). The average of each marker expression was calculated in all Fields of View (FOV), and the top 10% brightest FOV of each slide (i.e. hotspot).

Results: The frequency of macrophages, as assessed by the expression of CD68 was significantly higher in AA compared to NAA. This was seen by overall assessment for all FOV (mean NAA=2273au versus mean AA=3627au; p = 0.0052), and in hotspot FOVs (mean NAA=4858au versus mean AA=6371au; p = 0.0411), and also for assessments in tumor and stromal subcompartments. Expression of CD8 positive cytotoxic T cells was significantly lower in AA compared to NAA, but only when measuring hotspots in the tumor compartment (p=0.017).

Conclusion: The significantly higher CD68 and lower CD8 expression in AA compared to NAA TNBC might contribute to a more immune attenuated microenvironment. We are currently further characterizing the macrophage polarity and the cytokine milieu of these samples.

Citation Format: Vesal Yaghoobi, Vasiliki Pelekanou, Tess O'Meara, Andrea Silber, Lajos Pusztai, David Rimm. Vesal Yaghoobi [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2842.

Abstract 3151: Quantitative measurement of Siglec-15 expression in non-small cell lung cancer and its association with PD-L1, B7-H4 and tumor infiltrating lymphocytes

Introduction: Siglecs are sialic acid-binding transmembrane receptors that regulate the functions of innate and adaptive immune system through the recognition of glycan ligands. Siglec-15 has been recently described as an immune suppressive molecule that shares molecular structure with PD-L1, but its role in NSCLC is still unknown. In this study, we determined Siglec-15 expression in three retrospective NSCLC cohorts and evaluated its association with mutation status, major clinicopathologic characteristics and survival.

Experimental procedures: We used multiplexed automated quantitative immunofluorescence (QIF) to develop a validated assay for Siglec-15 measurement and used it to assess Siglec-15 expression and its association to major clinicopathologic variables and survival in three NSCLC cohorts of over 600 patients. Additionally, we investigated the correlation of Siglec-15 expression with tumor-infiltrating lymphocytes, PD-L1, B7-H3 and B7-H4 tumor expression.

Results: In our study, Siglec-15 tumor positivity was found in 12.5%, 13.7% and 22.8% of NSCLC cohort patients. Siglec-15 expression was higher in EGFR mutant and EGFR/KRAS wild type tumors compared to KRAS mutants but was not associated with any other major clinicopathological characteristics. Siglec-15 followed a mutually exclusive pattern of expression with PD-L1, B7-H3 and B7-H4 (co-expression in 3.1%, 13.2% and 7.7% of the patients respectively), while high levels were not correlated with lymphocyte infiltration or T-cell activation, suggesting a different upregulation mechanism not mediated by IFN-gamma.

Conclusions: Siglec-15 is a protein expressed with high frequency in NSCLC. Co-expression of Siglec-15 with PD-L1, B7-H3 and B7-H4 is relatively rare, suggesting that Siglec-15 has distinct and non-redundant features compared to other B7 family members.

Citation Format: Maria Toki, Jon Zugazagoitia, Mehmet Altan, Linda Liu, Nikita Mani, Yuting Liu, Konstantinos Syrigos, Lieping Chen, Solomon Langermann, Roy Herbst, David Rimm. Quantitative measurement of Siglec-15 expression in non-small cell lung cancer and its association with PD-L1, B7-H4 and tumor infiltrating lymphocytes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 3151.

Abstract A09: Impaired HLA Class I antigen processing and presentation as a mechanism of acquired Rrsistance to immune checkpoint inhibitors in lung cancer

Immune checkpoint inhibitors (ICIs) including programmed death 1 (PD-1) and programmed death ligand 1 (PD-L1) antagonist antibodies for lung tumors mark a new era of cancer therapeutics harnessing a patient’s immune system to induce durable antitumor responses. Despite their impressive activity in some patients, challenges remain as most tumors exhibit primary or acquired resistance (AR) to ICIs, and the biologic mechanisms for resistance are poorly understood. Using a collection of 14 ICI-resistant lung cancer samples, we investigated whether alterations in genes encoding components of the HLA Class I antigen processing and presentation machinery or interferon signaling pathways play a role in AR to PD-1 or PD-L1 antagonistic antibodies. Although we did not detect any recurrent mutations or copy number changes in our AR cohort, we noted one case of an acquired homozygous loss of B2M that resulted in lack of HLA class I expression on the cell surface in both the patient’s tumor sample and the corresponding patient-derived xenograft (PDX). Downregulation of B2M was also found in two additional PDXs established from ICI-resistant tumors. To test if B2m expression confers sensitivity to ICIs in vivo, we used a CRISPR-mediated approach to knockout B2m in an ICI-sensitive, murine lung cancer cell line and transplanted the cells into syngeneic, immunocompetent mice. We found that the B2m knockout tumors were resistant to ICI therapy in contrast to the wild type tumors. Furthermore, RNA-seq analysis of a subset of the samples in the acquired resistance to ICI cohort revealed an inflammatory tumor microenvironment with significant upregulation of the inhibitory receptors LAG-3 and PD-1 in tumor-infiltrating T cells at ICI resistance. Overall, these findings provide a novel system for the evaluation and screening of candidate genes for their ability to mediate AR to ICIs in vivo. Moreover, they also provide evidence for the disruption of HLA Class I antigen processing and presentation as a mechanism for escape from ICIs in lung cancer and provide information on the immune microenvironment in ICI-resistant tumors.

Citation Format: Katherine Hastings, Scott Gettinger, Choi Jungmin, Anna Truini, Ila Datar, Ryan Sowell, Anna Wurtz, Weilai Dong, Guoping Cai, Mary Ann Melnick, Joseph Schlessinger, Sarah Goldberg, Anne Chiang, Ignacio Melero, Jackeline Agorreta, Luis Montuenga, Richard Lifton, Soldano Ferrone, Paula Kavathas, David Rimm, Susan Kaech, Kurt Schalper, Roy Herbst, Katerina Politi. Impaired HLA Class I antigen processing and presentation as a mechanism of acquired Rrsistance to immune checkpoint inhibitors in lung cancer [abstract]. In: Proceedings of the AACR Special Conference on Tumor Immunology and Immunotherapy; 2017 Oct 1-4; Boston, MA. Philadelphia (PA): AACR; Cancer Immunol Res 2018;6(9 Suppl):Abstract nr A09.

Abstract 1695: A multi-institutional study to evaluate automated scoring of immunohistochemistry slides for assessment of programmed death-ligand 1 (PD-L1) expression in non-small cell lung cancer

Purpose: Assessment of PD-L1 expression is a critical part of patient management for immunotherapy. However, studies have shown that pathologist-based analysis lacks reproducibility, especially for immune cell expression. The purpose of this study was to validate and to assess reproducibility of the automated Optra image analysis for PD-L1 IHC for both tumor cells and immune cells.

Experimental Design: We compared conventional pathologists' scores (3 board-certified pathologists active in routine signout of these cases) for both tumor and immune cell positivity separately, using 22c3 antibody on the Dako Link 48 platform for PD-L1 expression in non-small cell lung carcinoma (NSCLC). We examined interpathologist PD-L1 expression scoring variability for both, tumor and immune cells using ordinal tumor proportion scores for tumor cells and continuous percentage positive scores. The cohort included 230 NSCLCs obtained from Yale School of Medicine, Department of Pathology archives. We assessed interpretation first by pathologists and secondly by the Optra PD-L1 image analysis scores for both tumor and immune cells. The Intra Class Correlation (ICC) for each pathologist was measured to assess variability between pathologists in scoring both tumor and immune cells. The concordance between pathologists using digital manual reads of PD-L1 staining percentages and Optra PD-L1 image analysis quantitative scores was then assessed using the Lin's concordance correlation coefficient for both tumor and immune cells.

Results: Intraclass correlation coefficients to evaluate the correlation between pathologists for tumor cell (ICC = 0.750) showed an excellent concordance but lower concordance for immune cell scoring (ICC = 0.4). To compare the pathologist scores to the Optra automated system, the scores from the 3 pathologists were averaged to produce a single conventional read score. The Lin's concordance correlation coefficient between the conventional read and the machine score was 0.83 for tumor cells and 0.6 for immune cell population in intra- and peritumoral compartments. This is considered excellent agreement for tumor cells and good concordance for immune cells.

Conclusion: The interpathologist assessment seen in this study is similar to previously reported studies where agreement is higher in tumor cells than immune cells. The Optra PD-L1 image analysis showed concordance with the pathologists' average scores that were comparable to interpathologist scores. This suggests promise for automated assessment of PD-L1 in NSCLC. These results justify similar studies with immunotherapy-treated patients with known outcomes.

Citation Format: Clive Taylor, Anagha P. Jadhav, Abhi Gholap, Gurunath Kamble, Jiaoti Huang, Allen Gown, Isha Doshi, David Rimm. A multi-institutional study to evaluate automated scoring of immunohistochemistry slides for assessment of programmed death-ligand 1 (PD-L1) expression in non-small cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1695.

Mutation profiles in early-stage lung squamous cell carcinoma with clinical follow-up and correlation with markers of immune function

Background

Lung squamous cell carcinoma (LUSC) accounts for 20–30% of non-small cell lung cancers (NSCLCs). There are limited treatment strategies for LUSC in part due to our inadequate understanding of the molecular underpinnings of the disease. We performed whole-exome sequencing (WES) and comprehensive immune profiling of a unique set of clinically annotated early-stage LUSCs to increase our understanding of the pathobiology of this malignancy.

Methods

Matched pairs of surgically resected stage I-III LUSCs and normal lung tissues (n = 108) were analyzed by WES. Immunohistochemistry and image analysis-based profiling of 10 immune markers were done on a subset of LUSCs (n = 91). Associations among mutations, immune markers and clinicopathological variables were statistically examined using analysis of variance and Fisher’s exact test. Cox proportional hazards regression models were used for statistical analysis of clinical outcome.

Results

This early-stage LUSC cohort displayed an average of 209 exonic mutations per tumor. Fourteen genes exhibited significant enrichment for somatic mutation: TP53, MLL2, PIK3CA, NFE2L2, CDH8, KEAP1, PTEN, ADCY8, PTPRT, CALCR, GRM8, FBXW7, RB1 and CDKN2A. Among mutated genes associated with poor recurrence-free survival, MLL2 mutations predicted poor prognosis in both TP53 mutant and wild-type LUSCs. We also found that in treated patients, FBXW7 and KEAP1 mutations were associated with poor response to adjuvant therapy, particularly in TP53-mutant tumors. Analysis of mutations with immune markers revealed that ADCY8 and PIK3CA mutations were associated with markedly decreased tumoral PD-L1 expression, LUSCs with PIK3CA mutations exhibited elevated CD45ro levels and CDKN2A-mutant tumors displayed an up-regulated immune response.

Conclusion(s)

Our findings pinpoint mutated genes that may impact clinical outcome as well as personalized strategies for targeted immunotherapies in early-stage LUSC.

Abstract PD6-08: Analysis of immune infiltrates (assessed via multiplex fluorescence immunohistochemistry) and immune gene expression signatures as predictors of response to the checkpoint inhibitor pembrolizumab in the neoadjuvant I-SPY 2 trial

Background: Pembrolizumab (Pembro), an anti-PD-1 immune checkpoint inhibitor, has been approved for the treatment of a variety of cancers including melanoma, non-small cell lung cancer, head and neck squamous cell carcinoma, and urothelial carcinoma. Pembro was recently evaluated in HER2- breast cancer patients in the neoadjuvant I-SPY 2 TRIAL and graduated in the triple negative (TN), HR+HER2-, and HER2- signatures. HER2- patients were randomized to receive Pembro+paclitaxel followed by doxorubicin/cyclophosphamide (P+T -&gt; AC) vs. T -&gt; AC. We and others have shown that TN breast cancers tend to have high numbers of immune infiltrates, including T cells and tumor associated macrophages (TAMs). We evaluated expression signatures representing 14 immune cell types (TILs, T cells, CD8 T cells, exhausted T cells, Th1, Tregs, cytotoxic cells, NK, NK CD56dim, dendritic cells, mast cells, B cells, macrophages, and neutrophils) as specific predictors of response to Pembro.

Methods: Data from 248 patients (Pembro: 69; controls: 179) were available. Pre-treatment biopsies were assayed using Agilent gene expression arrays. Signature scores are calculated by averaging cell type specific genes. All I-SPY 2 qualifying biomarker analyses follow a pre-specified analysis plan. We used logistic modeling to assess biomarker performance. A biomarker is considered a specific predictor of Pembro response if it associates with response in the Pembro arm but not the control arm, and if the biomarker x treatment interaction is significant (likelihood ratio test, p&lt;0.05). This analysis is also performed adjusting for HR status as covariates, and within receptor subsets. For successful biomarkers, we use Bayesian modeling to estimate the pCR rates of 'predicted sensitive' patients in each arm. Our statistics are descriptive rather than inferential and do not adjust for multiplicities of other biomarkers outside this study.

Results: 10 out of the 14 cell-type signatures tested are associated with response in the Pembro arm. Higher expression levels of 9 of these cell-type signatures are associated with higher pCR rates (T cells, exhausted T cells, Th1, cytotoxic cells, NK, NK CD56dim, dendritic cells, B cells, and macrophages), whereas higher mast cell signature expression is associated with non-pCR. Interestingly, many of these same signatures also associate or trend towards association with response in the control arm; and in a model adjusting for HR status, only 3 of these signatures (Th1, B cells and dendritic cells) show significant interaction with treatment. Within the whole population and the TN subtype, the dendritic cell signature is the strongest predictor of specific response to Pembro (OR/1SD: 4.04 and 4.4, LR p &lt; 0.001 overall and in TN). Although other immune signatures (T cells, exhausted T cells, NK, and macrophages) also associate with response in the Pembro arm in the TN subtype, only the dendritic cell and Th1 signatures have a significant interaction with treatment. In contrast, in the HR+HER2- subtype, only 3 signatures (Th1, B cells, and mast cells) associate with response to Pembro; but none of these signatures have significant interaction with treatment. Of note, in both the Pembro and control arms, HR+HER2- patients with higher average mast cell marker expression have lower pCR rates (OR/1SD: 0.33 and 0.51, LRp: 0.006 and 0.04 in Pembro and control arm).

Conclusion: As expected, multiple immune cell expression signatures are predictive of response in the Pembro arm; but only dendritic cells and Th1 cells are specific to Pembro in both the population as a whole and the TN subtype. Interestingly, the presence of mast cells may impede response, especially in HR+HER2- patients. Correlation of these signatures with multiplex-IF immune markers is pending.

Citation Format: Campbell M, Yau C, Borowsky A, Vandenberg S, Wolf D, Rimm D, Nanda R, Liu M, Brown-Swigart L, Hirst G, Asare S, van't Veer L, Yee D, DeMichele A, Berry D, Esserman L. Analysis of immune infiltrates (assessed via multiplex fluorescence immunohistochemistry) and immune gene expression signatures as predictors of response to the checkpoint inhibitor pembrolizumab in the neoadjuvant I-SPY 2 trial [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr PD6-08.

Assessing Tumor-Infiltrating Lymphocytes in Solid Tumors: A Practical Review for Pathologists and Proposal for a Standardized Method from the International Immuno-Oncology Biomarkers Working Group: Part 2: TILs in Melanoma, Gastrointestinal Tract Carcinomas, Non-Small Cell Lung Carcinoma and Mesothelioma, Endometrial and Ovarian Carcinomas, Squamous Cell Carcinoma of the Head and Neck, Genitourinary Carcinomas, and Primary Brain Tumors

Assessment of the immune response to tumors is growing in importance as the prognostic implications of this response are increasingly recognized, and as immunotherapies are evaluated and implemented in different tumor types. However, many different approaches can be used to assess and describe the immune response, which limits efforts at implementation as a routine clinical biomarker. In part 1 of this review, we have proposed a standardized methodology to assess tumor-infiltrating lymphocytes (TILs) in solid tumors, based on the International Immuno-Oncology Biomarkers Working Group guidelines for invasive breast carcinoma. In part 2 of this review, we discuss the available evidence for the prognostic and predictive value of TILs in common solid tumors, including carcinomas of the lung, gastrointestinal tract, genitourinary system, gynecologic system, and head and neck, as well as primary brain tumors, mesothelioma and melanoma. The particularities and different emphases in TIL assessment in different tumor types are discussed. The standardized methodology we propose can be adapted to different tumor types and may be used as a standard against which other approaches can be compared. Standardization of TIL assessment will help clinicians, researchers and pathologists to conclusively evaluate the utility of this simple biomarker in the current era of immunotherapy.

Assessing Tumor-infiltrating Lymphocytes in Solid Tumors: A Practical Review for Pathologists and Proposal for a Standardized Method From the International Immunooncology Biomarkers Working Group: Part 1: Assessing the Host Immune Response, TILs in Invasive Breast Carcinoma and Ductal Carcinoma In Situ, Metastatic Tumor Deposits and Areas for Further Research

Assessment of tumor-infiltrating lymphocytes (TILs) in histopathologic specimens can provide important prognostic information in diverse solid tumor types, and may also be of value in predicting response to treatments. However, implementation as a routine clinical biomarker has not yet been achieved. As successful use of immune checkpoint inhibitors and other forms of immunotherapy become a clinical reality, the need for widely applicable, accessible, and reliable immunooncology biomarkers is clear. In part 1 of this review we briefly discuss the host immune response to tumors and different approaches to TIL assessment. We propose a standardized methodology to assess TILs in solid tumors on hematoxylin and eosin sections, in both primary and metastatic settings, based on the International Immuno-Oncology Biomarker Working Group guidelines for TIL assessment in invasive breast carcinoma. A review of the literature regarding the value of TIL assessment in different solid tumor types follows in part 2. The method we propose is reproducible, affordable, easily applied, and has demonstrated prognostic and predictive significance in invasive breast carcinoma. This standardized methodology may be used as a reference against which other methods are compared, and should be evaluated for clinical validity and utility. Standardization of TIL assessment will help to improve consistency and reproducibility in this field, enrich both the quality and quantity of comparable evidence, and help to thoroughly evaluate the utility of TILs assessment in this era of immunotherapy.

Whole-exome sequencing and immune profiling of early-stage lung adenocarcinoma with fully annotated clinical follow-up

Background

Lung adenocarcinomas (LUADs) lead to the majority of deaths attributable to lung cancer. We performed whole-exome sequencing (WES) and immune profiling analyses of a unique set of clinically annotated early-stage LUADs to better understand the pathogenesis of this disease and identify clinically relevant molecular markers.

Methods

We performed WES of 108 paired stage I-III LUADs and normal lung tissues using the Illumina HiSeq 2000 platform. Ten immune markers (PD-L1, PD-1, CD3, CD4, CD8, CD45ro, CD57, CD68, FOXP3 and Granzyme B) were profiled by imaging-based immunohistochemistry (IHC) in a subset of LUADs (n = 92). Associations among mutations, immune markers and clinicopathological variables were analyzed using ANOVA and Fisher's exact test. Cox proportional hazards regression models were used for multivariate analysis of clinical outcome.

Results

LUADs in this cohort exhibited an average of 243 coding mutations. We identified 28 genes with significant enrichment for mutation. SETD2-mutated LUADs exhibited relatively poor recurrence- free survival (RFS) and mutations in STK11 and ATM were associated with poor RFS among KRAS-mutant tumors. EGFR, KEAP1 and PIK3CA mutations were predictive of poor response to adjuvant therapy. Immune marker analysis revealed that LUADs in smokers and with relatively high mutation burdens exhibited increased levels of immune markers. Analysis of immunophenotypes revealed that LUADs with STK11 mutations exhibited relatively low levels of infiltrating CD4+/CD8+ T-cells indicative of a muted immune response. Tumoral PD-L1 was significantly elevated in TP53 mutant LUADs whereas PIK3CA mutant LUADs exhibited markedly down-regulated PD-L1 expression. LUADs with TP53 or KEAP1 mutations displayed relatively increased CD57 and Granzyme B levels indicative of augmented natural killer (NK) cell infiltration.

Conclusion(s)

Our study highlights molecular and immune phenotypes that warrant further analysis for their roles in clinical outcomes and personalized immune-based therapy of LUAD.

Abstract P6-10-02: MHC-II positive breast tumors are more immunogenic and may preferentially select for LAG-3-positive tumor immune infiltrates

Background: Lymphocyte-activation gene 3 (LAG-3) is a T-cell checkpoint regulator and a current target in immunotherapy trials. LAG-3's main ligand is MHC class II (MHC-II), to which it binds with higher affinity than CD4. Binding of LAG3 to MHC-II antigen-presenting cells negatively regulates cellular proliferation, activation, and homeostasis of T cells, similarly to CTLA-4 and PD-1, suggesting that antibodies targeting LAG-3 may demonstrate similar anti-tumor immune effects.

Hypothesis: We recently reported an association of MHC-II on tumor cells and its involvement in mediating sensitivity to PD-1/PD-L1 monoclonal antibodies. MHC-II demonstrates a strong bimodal expression pattern on tumor cells from a variety of tissues, including those of the breast. In breast cancer patients, tumor-specific MHC-II expression on TNBCs is correlated with a 'hot' immune environment. We hypothesized that 1) MHC-II expression may drive potent anti-tumor immune responses and 2) MHC-II-positive tumors that generate immunotolerance may develop a specific immune checkpoint dependency on LAG-3, since LAG-3 is the inhibitory receptor for MHC-II-mediated antigen presentation.

Methods: To determine the functionality of MHC-II in driving anti-tumor immune responses, we constitutively expressed the MHC-II master regulator CIITA in MMTV-neu mouse tumor cells and determined their ability to form tumors in immunocompetent syngeneic hosts. To evaluate the association of MHC-II+ tumors with LAG-3 expression, we evaluated LAG-3-positivity by immunohistochemistry (IHC) in lymphocytic infiltrates in a series of 111 post-NAC TNBC specimens from patients with residual disease remaining after presurgical chemotherapy. Tumor-infiltrating lymphocytes (TILs) were scored by H&E, PD-L1 and MHC-II (HLA-DR) were scored in the stroma and tumor compartments using automated quantitative immunofluorescence (AQUA).

Results: Enforced expression of MHC-II via constitutive expression of CIITA caused rejection in 60% of mice, while only 11% of mice rejected MMTV-neu tumors expressing the vector control (Fisher's exact p=0.04). All rejecting mice were immune to rechallenge with parental (non-CIITA-expressing) MMTV-neu cells, suggesting a memory effector response.

Clinically, 11/102 patients (10.8%) had LAG-3+ immune cells in their tumor. LAG-3+ tumors were strongly correlated with MHC-II positivity in tumor cells (p&lt;0.0001). Presence of LAG-3+ cells also correlated strongly with overall TILs (p&lt;0.0001), and PD-L1 expression on TILs (p&lt;0.02). Since the likelihood of identifying LAG3+ lymphocytes is confounded by the inclusion of poorly-infiltrated tumors, we performed a subset analysis on only those tumors with substantial TILs (&gt;20%). When this subset was analyzed, LAG-3 positivity retained its association with tumor MHC-II expression (p=0.0001), while the association of LAG-3 with stromal PD-L1 was reduced below the level of significance (p=0.052).

Conclusions: MHC-II expression causes increased immune activation in breast cancers, consistent with our previous findings. MHC-II positivity in breast tumors may identify a population with preferential dependence on the LAG-3 checkpoint, which may be important for future immunotherapy trials.

Citation Format: Balko JM, Loi S, Giltnane JM, Combs S, Estrada MV, Sanchez V, Rimm D, Sanders ME, Salgado R, Gomez H, Johnson DB. MHC-II positive breast tumors are more immunogenic and may preferentially select for LAG-3-positive tumor immune infiltrates [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P6-10-02.

Automated measurement of estrogen receptor in breast cancer: a comparison of fluorescent and chromogenic methods of measurement

Whereas FDA-approved methods of assessment of estrogen receptor (ER) are 'fit for purpose', they represent a 30-year-old technology. New quantitative methods, both chromogenic and fluorescent, have been developed and studies have shown that these methods increase the accuracy of assessment of ER. Here, we compare three methods of ER detection and assessment on two retrospective tissue microarray (TMA) cohorts of breast cancer patients: estimates of percent nuclei positive by pathologists and by Aperio's nuclear algorithm (standard chromogenic immunostaining), and immunofluorescence as quantified with the automated quantitative analysis (AQUA) method of quantitative immunofluorescence (QIF). Reproducibility was excellent (R(2)>0.95) between users for both automated analysis methods, and the Aperio and QIF scoring results were also highly correlated, despite the different detection systems. The subjective readings show lower levels of reproducibility and a discontinuous, bimodal distribution of scores not seen by either mechanized method. Kaplan-Meier analysis of 10-year disease-free survival was significant for each method (Pathologist, P=0.0019; Aperio, P=0.0053, AQUA, P=0.0026); however, there were discrepancies in patient classification in 19 out of 233 cases analyzed. Out of these, 11 were visually positive by both chromogenic and fluorescent detection. In 10 cases, the Aperio nuclear algorithm labeled the nuclei as negative; in 1 case, the AQUA score was just under the cutoff for positivity (determined by an Index TMA). In contrast, 8 out of 19 discrepant cases had clear nuclear positivity by fluorescence that was unable to be visualized by chromogenic detection, perhaps because of low positivity masked by the hematoxylin counterstain. These results demonstrate that automated systems enable objective, precise quantification of ER. Furthermore, immunofluorescence detection offers the additional advantage of a signal that cannot be masked by a counterstaining agent. These data support the usage of automated methods for measurement of this and other biomarkers that may be used in companion diagnostic tests.

Abstract 89: Whole-exome sequencing and immune profiling of early-stage lung adenocarcinoma

PURPOSE: Lung adenocarcinomas (LUADs) lead to the preponderance of deaths attributable to lung cancer. We performed whole-exome sequencing (WES), comprehensive immune profiling and clinicopathological analysis of LUADs to better understand the molecular pathogenesis of this disease and identify clinically relevant molecular markers.

METHODS: We performed WES of 108 paired surgically resected stage I-III LUADs and normal lung tissues using the Illumina HiSeq 2000 platform. Additionally, ten immune related markers (PD-L1, PD-1, CD3, CD4, CD8, CD45ro, CD57, CD68, FOXP3 and Granzyme B) were profiled by imaging-based immunohistochemistry in a subset of LUADs (n = 92). Associations among mutations, immune markers and clinicopathological variables were analyzed using ANOVA and Fishers Exact tests. Cox proportional hazards regression models were employed for multivariate analysis of clinical outcome.

RESULTS: LUADs in this cohort exhibited an average of 243 coding mutations per tumor. We identified 28 genes with significant enrichment for mutation. SETD2-mutant LUADS exhibited relatively poor recurrence-free survival (RFS) and mutations in STK11 and ATM were associated with poor RFS in KRAS-mutant tumors. EGFR, KEAP1 and PIK3CA mutations were predictive of poor response to adjuvant therapy. Immune marker analysis demonstrated that PD-L1 expression was increased in smoker compared to non-smoker LUADs and, along with other immune markers, was positively correlated with somatic mutation burden. Moreover, immune marker levels including PD-L1 were elevated in TP53-mutant LUADs. In contrast, STK11 and U2AF1 mutant tumors exhibited a suppressed immune response and LUADs with PIK3CA mutations exhibited markedly decreased tumoral PD-L1 expression.

CONCLUSION: Our study highlights mutations that may impact clinical outcome and personalized strategies for immune-based therapy of early-stage LUAD patients.

Citation Format: Humam Kadara, Murim Choi, Jiexin Zhang, Edwin Parra Cuentas, Jaime Rodriguez Canales, Stephen Gaffney, Zi-Ming Zhao, Carmen Behrens, Junya Fujimoto, Chi-Wan Chow, Neda Kalhor, Cesar Moran, David Rimm, Stephen G. Swisher, Don L. Gibbons, John V. Heymach, Edward Kaftan, Jeffrey Townsend, Thomas J. Lynch, Joseph Schlessinger, J. Jack Lee, Richard Lifton, Ignacio I. Wistuba, Roy S. Herbst. Whole-exome sequencing and immune profiling of early-stage lung adenocarcinoma. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 89.

Abstract S5-01: Whole exome sequencing of pre-treatment biopsies from the neoALTTO trial to identify DNA aberrations associated with response to HER2-targeted therapies

Background: We examined if alterations in nucleic acid variants, genes, pathways, and overall mutational load and clonal entropy are associated with pathologic complete response (pCR) and survival after neoadjuvant anti-HER2 therapies in the NeoALTTO trial.

Methods: Whole exome sequencing was performed of 203 baseline biopsies with outcome information. The mean nucleotide coverage was 150x with &gt;90% of target bases showing &gt; 30x coverage in &gt; 99% of samples. Somatic mutations were called by MuTect and indels by Strelka, using pooled reference normal DNA. Significantly mutated genes (FDR&lt;10%) were identified by MutSigCV. Mutations in 714 canonical biological pathways were assessed and mutational load and genome clonal entropy (MATH) were calculated. Association with pCR and survival were evaluated by logistic regression adjusted for ER status and Cox-proportional hazards regression.

Results: Only 12 genes had mutation rates significantly above background and among these only PI3KCA was associated with lower pCR rate (OR=0.42, p=0.019). Genes with somatic mutations in more than 10 patients were also assessed, but none were associated with pCR or survival. Clonal entropy or adjusted mutation load also did not correlate with response. Mutations in 33 pathways showed significant association with response in the entire cohort. In the trastuzumab arm, 23 of the 33 pathways showed an association with response but none was independent of PIK3CA mutation. We constructed "PIK3CA-gene network" that included all unique genes (n=439) from theese 23 pathways. Of the 66 patients in the trastuzumab arm, 50 carried at least one mutation in one of the 439 genes and among these only 2 achieved pCR (4%) compared to 9 of 16 pCR (56%) among the wild type (OR=0.035; p &lt; 0.001). The same genes/mutations had little impact on pCR in the lapatinib arm (pCR 20%). In the lapatinib arm, mutations in 3 pathways conferred higher probability of pCR. The "Regulation of RhoA activity" pathway, had the most significant association with pCR in the entire cohort (OR=3.77, p=0.0009) and in the lapatininb (pCR 67% vs 17%, OR=14.8, p=0.008) and lapatinib + trastuzumab (OR=3.0, p=0.06) arms, but not in the trastuzumab arm (OR=1.4, p=0.7). Event free and overall survival were also significantly higher in patients who had mutations in this pathway. Twenty seven of the 48 genes in this pathway had mutations affecting 33 patients but different genes were affected in different individuals.

Conclusions: There are no high frequency recurrent single mutations associated with response to HER2-targeted therapies, other than PIK3CA. We identified several biological pathways, including RhoA activity, and a network of PIK3CA associated genes that are significantly associated with response when affected by mutations, however, different genes are mutated in different individuals.

Citation Format: Pusztai L, Shi W, Jiang T, Nuciforo P, Holmes E, Harbeck N, Sotiriou C, Rimm D, Hatzis C, de la Peña L, Armour A, Piccart-Gebhart M, Baselga J. Whole exome sequencing of pre-treatment biopsies from the neoALTTO trial to identify DNA aberrations associated with response to HER2-targeted therapies. [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr S5-01.

Abstract P5-07-12: Local nuclear architecture features from H&E images predict early versus distant recurrence in lymph node negative, ER+ breast cancers

Introduction: Breast cancer (BCa) Patients with ER+ tumors that are lymph node negative (LN-) typically receive hormonal therapy. There is a need to identify ER+ LN- patients that will not benefit from adjuvant chemotherapy and will respond to hormonal therapy alone. Oncotype DX, a quantitative prognostic and predictive gene assay, provides a recurrence score that has been correlated with distant and early recurrence. In this work we present an approach that employ computer extracted features of nuclear architecture and morphology from routine H&E slides alone that can distinguish early and distant recurrence in ER+ breast cancers. By constructing graph networks within epithelium and stroma regions, built using nuclei as vertices and edge connections between proximal nuclei, local nuclear architecture can be quantitatively characterized. Hosoya index (HI) (originally introduced for analysis of chemical bonds) is a measure of a bond (in this context nuclei connections in a graph). In this work, we leverage HI to measure structural similarities of graphs across the populations that are indicative of recurrence in LN- ER+ breast cancer tissue microarray (TMA) images.

Design: In this study we considered two tissue microarrays (TMAs) comprising 453 early-stage lymph-node negative (LN-) estrogen receptor positive (ER+) breast cancer (BCa) patients (diagnosed with invasive ductal carcinoma), with a total of N=90 patients experiencing lifetime distant recurrence and N=343 patients who did not. All TMA cores were digitized at 20x magnification (0.33 um/pixel spatial resolution) using a digital whole-slide scanner. Each nucleus was identified via an automated computerized image analysis algorithm developed by our group. Then, using a cluster cell graph that encodes a link between a pair of nodes based on proximity, a series of graphs are constructed for a TMA. A HI value was then assigned to each local graph. A support vector machine classifier was trained in conjunction with the distribution of HI values for the early and distant recurrence cases on the training TMA (n=243, 50 early recurrences). Independent validation of the SVM classifier was performed on the second TMA (n=210, 40 early recurrences).

Results: For the LN- ER+ breast cancer dataset, our method was able to distinguish tumors with early and distant recurrence with an accuracy of 75.4%, a positive predictive value of 78.6% and a negative predictive value of 76.4%. The separation between the Kaplan-Meier curves for early and distant recurrence of LN-, ER+ breast cancers on the validation set was statistically significant (p &lt; 0.00102).

Conclusion: Based only on tiny H&E punches, a computer-aided morphometric classifier appears to identify lymph node negative, ER+ breast cancers with a low likelihood of recurrence. With further validation, this approach could be developed into an image based assay which could serve as a lower cost alternative to Oncotype DX.

Citation Format: Ali S, Rimm D, Ganesan S, Madabhushi A. Local nuclear architecture features from H&E images predict early versus distant recurrence in lymph node negative, ER+ breast cancers. [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr P5-07-12.