PD-L1 Immunohistochemistry Research Articles

Abstract 2034: Spatial transcriptomics identifies unique pharmacodynamic effects of checkpoint inhibitor treatment on the tumor microenvironment in NSCLC

Abstract Background: Immune checkpoint inhibitor (ICI) therapy has improved outcomes in non-small cell lung cancer (NSCLC), particularly in patients with high tumor infiltrating lymphocytes (TILs). Identifying the pharmacodynamic (PD) impact of ICI on immune cells in the tumor microenvironment (TME) can inform therapeutic development for patients with progressive disease. We sought to identify PD changes in the TME from patients before and after ICI therapy using a spatial transcriptomics platform that allows highly multiplexed profiling of 1,800 genes (Nanostring Digital Spatial Profiler [DSP]). Methods: Formalin fixed, paraffin embedded (FFPE) tumor tissue from 22 patients was sourced commercially. Patients were treated with neoadjuvant chemotherapy, then underwent a surgical tumor resection. After surgery an adjuvant chemotherapy was administered until progression; patients then received monotherapy ICI (nivolumab or pembrolizumab). Once progressed on ICI, another resection was performed. Patients were then treated with chemotherapy and followed until progression and/or death. The DSP technology was used to independently profile RNA from PanCK+ (tumor) and PanCK- (stroma) regions in the tissue based on fluorescence staining. Six circular regions of 600μm in diameter were selected for analysis using the GeoMx instrument; each area contained CD3+ cells. Additionally, immunohistochemistry for PDL1 and CD3 was performed, images were scored by a pathologist and analyzed with digital pathology algorithms. Results: Spatial transcriptomic analysis of pre- vs post-ICI treatment in stroma revealed significant increases in T cell genes (CD3E, TCRB, NKG7), T-cell activation markers (CD69, IFNG, OX40, GZMB, ICOS), costimulatory signaling (CD28, CD80, CD86), and immune checkpoints (CTLA4, TIGIT). Ayers et al., JCI 2017 identified 28 genes predictive of ICI response, 12 were significantly upregulated in the stroma post-ICI (26 were present in the DSP panel). We identified genes in the pre-ICI stromal microenvironment that were highly expressed in patients with partial response to ICI, the most significant genes were involved in immune regulation (IFIT1) and extracellular matrix remodeling (MMP3). In contrast, stromal genes highly expressed in patients with progressive disease were associated with T-cell maintenance (ETS1, IL7R, CCL19). Conclusions: In this study we used spatial transcriptomics to profile tissue regions where T-cells were in close proximity to tumor cells (microns). This focused, local PD analysis confirmed activated T-cells are present post-ICI. Ongoing studies in a larger cohort will be used to identify tumoral mechanisms of resistance and immune dysfunction to inform future therapeutics and combinations. Citation Format: Li Fan, Omar Jabado, Nora Pencheva, Patricia Coutinho de Souza, Brandon Higgs, Angelo Harris, Patrick Franken, Anantharaman Muthuswamy, Maria Jure-Kunkel, Suzana Couto, Kate Sasser, Mark Fereshteh. Spatial transcriptomics identifies unique pharmacodynamic effects of checkpoint inhibitor treatment on the tumor microenvironment in NSCLC [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2034.

Abstract 645: Heterogeneity of immunotherapy biomarkers in the TRACERx non-small cell lung cancer multi-region lung cancer cohort study

Abstract Immunotherapy containing regimens have become the standard of care first-line therapy for non-small cell lung cancer (NSCLC) in the metastatic setting and are being explored in the adjuvant and neo-adjuvant setting. However, as intratumor heterogeneity (ITH) is pervasive in cancer we investigated the impact of ITH on established and putative biomarkers predictive of immunotherapy response. We have used whole exome sequencing, RNA sequencing and PDL1 immunohistochemistry (IHC) derived from the multiregion treatment naïve TRACERx cohort to investigate this. Whole exome sequencing data was available for 432 primary tumours from 421 patients with greater than 1500 unique tumour regions. Tumour mutational burden (TMB) was calculated using best practice guidelines from the TMB harmonisation project. Using the FDA approved threshold of ≥ 10 mutations per megabase to define high TMB we found that 41% of patients with high TMB had at least one region with <10 mutation per megabase. This could result in misclassification of patients so we modelled increased sampling in this cohort and found that at least 5 spatially separated samples would be required to ensure no patients are misclassified. Contrary to the published literature we did not find TMB was confounded by tumour purity. Multi-region PDL1 IHC data using the FDA approved companion diagnostic assay (22c3) was available for 132 spatially separated tumour regions from 36 patients. Tumour regions were scored according to the established tumour proportion scoring system (≥50%, 49-1%, <1%). Using only a single sample 31% of patients would have been misclassified. Finally, we investigated several putative transcriptomic biomarkers including 10 single genes (e.g. CXCL9), 15 expression signatures (e.g cytolytic score, TIDE) and 2 tumour microenvironment classification systems. We found that 20-45% of these putative biomarkers were discordant when applied to multi-region data and this increased with the number of tumour regions per patients but was not impacted by tumour purity or sequencing depth. To investigate this further we compared the mean geneset RNA-ITH score between a geneset of all the immune biomarker genes with 100 randomly selected non-immune, but expressed, genesets. The immune biomarker gene set was subject to significantly more intra tumour heterogeneity of expression (Mean RNA-ITH Score 0.50 vs 0.31 p=6.4e-12). In conclusion we demonstrate significant heterogeneity of immune biomarkers in NSCLC and in particular a lack of robustness of expression-based predictors of response to immunotherapy. This highlights the need for biomarkers that are robust to intra-tumour heterogeneity and sampling bias or the need for more representative tumour sampling techniques. Citation Format: Crispin T. Hiley, Kevin Litchfield, Oriol Pich, David Moore, Cristina Naceur-Lombardelli, Selvaraju Veeriah, Maise Al Bakir, Simranpreet Summan, Kristiana Grigoriadis, Carlos Martinez Ruiz, Clare Puttick, Katey Enfield, Sophia Ward, Alexander Frankell, Dhruva Biswas, Rachel Rosenthal, Nicolai J. Birkbak, Mariam Jamal-Hanjani, Nicholas McGranahan, Charles Swanton, TRACERx Consortium. Heterogeneity of immunotherapy biomarkers in the TRACERx non-small cell lung cancer multi-region lung cancer cohort study [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 645.

Abstract 761: Pan-cancer landscape and impact of PD-L1 and PD-L2 structural variation

Abstract Purpose PD-1 receptor and ligand interactions are the target of immunotherapies across 16 cancer types. Biomarkers that predict response to immunotherapy are microsatellite instability, tumor mutational burden and PD-L1 immunohistochemistry. Structural variations in PD-L1 (CD274) and PD-L2 (PDCD1LG2) have been observed across cancer, but the full landscape is unknown. Here we describe the genomic landscape of PD-L1 and PD-L2 structural variation, their potential impact on the tumor microenvironment and evidence that patients with these alterations can benefit from immunotherapy. Methods We analyzed sequencing data from cancer cases with PD-L1 (CD274) and PD-L2 (PDCD1LG2) structural variations across published data, The Cancer Genome Atlas (TCGA) and the Oncology Research Information Exchange Network. From TCGA we obtained copy number status through cBioPortal, gene expression using TCGABiolinks and PD-L1 structural variations identified by Kataoka et al. Gene fusions were detected in ORIEN using STAR-Fusion and Arriba. To evaluate immune signature enrichment we ran software ImSig on gene expression data, using Mann-Whitney tests to determine significant results. We curated literature of patients with structural variations in PD-L1 or PD-L2 receiving PD-1 immunotherapy. Findings From 18 studies and datasets we curated 319 cases with structural variations in PD-L1 and PD-L2. We observed breakpoint ‘hotspots’ in the untranslated regions (UTRs) of both genes including 70 duplications, 48 deletions, 78 inversions and 106 translocations. Leveraging TCGA, we found that PD-L1 amplified tumors had significantly upregulated PD-L1 expression and signatures for interferon signaling and immune cell proliferation, compared to PD-L1 copy neutral tumors, each p < 0.001. Similarly, in PD-L1 rearranged tumors we observed upregulation in PD-L1 expression and signatures for interferon signaling, macrophages, monocytes, T cells and immune cell proliferation, all p < 0.001 compared to PD-L1 copy neutral tumors. Further, retrospective review of 7 studies including patients with structural variations in PD-L1 or PD-L2 revealed >50% (43/77) response rate to immunotherapy. Implications Our evaluation of PD-L1 and PD-L2 structural variations show that the 3’ UTR is affected in hotspots involving a variety of structural variations. Our findings from TCGA suggest PD-L1 structural variation may play a role in driving expression of PD-L1 and immune dysregulation. Enriched interferon signaling in PD-L1 rearranged tumors is of interest as interferon exposure is known to drive PD-L1 and PD-L2 expression. Retrospective evidence from curated studies suggest that this genomic alteration could help identify candidates for PD-1 inhibitors. Based on these findings we propose further study to optimize detection of PD-L1 and PD-L2 structural variation in cancer and design of a pan-cancer prospective clinical trial to target these alterations. Citation Format: Emily L. Hoskins, Eric Samorodnitsky, Michele R. Wing, Julie W. Reeser, Julia Hopkins, Karthikeyan Murugesan, Zheng Kuang, Leah Stein, Zach Risch, Raven Vella, Serifat Adebola, Lianbo Yu, Anoosha Paruchuri, Richard S. Huang, Lee A. Albacker, Sameek Roychowdhury. Pan-cancer landscape and impact of PD-L1 and PD-L2 structural variation [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 761.

Abstract 471: AIM PD-L1-NSCLC: Artificial intelligence-powered PD-L1 quantification for accurate prediction of tumor proportion score in diverse, multi-stain clinical tissue samples

Abstract Introduction: Important immunotherapy drugs targeting PD-L1 are approved for first and second line treatment for various stages of NSCLC. Reproducible and precise evaluation of PD-L1 expression is essential to accurately evaluate patients’ eligibility for treatment and for enrollment in clinical trials. Current guidelines rely on pathologists to interpret tumor samples, which is challenging in part because different PD-L1 assays have distinct scoring criteria. As a result, determining eligibility by manual assessment can be inconsistent and inaccurate, leading to untreated patients. To support pathologist quantification of PD-L1 in clinical trials, PathAI has developed scanner-and antibody-agnostic machine learning (ML) models, AI-based histologist measurement of PD-L1 in NSCLC (AIM-PD-L1-NSCLC), for the quantification of PD-L1 expression in NSCLC using four PD-L1 immunohistochemistry (IHC) clones. Methods: AIM-PD-L1-NSCLC was trained using convolutional neural networks to identify and quantify PD-L1-positive cells in digitized whole slide images (WSI) of tissue samples. Models were developed using over 5,000 diverse clinical biopsies and resections, including primary and metastatic adenocarcinoma and squamous cell carcinoma samples collected from 10 clinical trials and from two clinical laboratories, each stained for PD-L1 with one of four IHC clones: SP263 (N=1,320), SP142 (N=1,829) (both Ventana Medical Systems Inc., Tucson, AZ), 28-8 (N=1,331), or 22C3 (N=843) (both Agilent Technologies, Santa Clara, USA). Slides were digitized using Aperio, Philips, and Ventana scanners, and WSI were split into training (N=3,818) and test (N=1,505) datasets. The training dataset was annotated by board certified pathologists (313,770 annotations) to label tissue regions and cells. Human Interpretable features representing the number of tumor cells were automatically extracted from the model and a slide level Tumor Proportion Score (TPS) calculated as the proportion of PD-L1+ cancer cells divided by total cancer cells in tumor regions. Model predicted slide level TPS were compared with the median TPS of five pathologists’ scores using intraclass correlation coefficient (ICC) statistics. Results: There was high concordance between ML model-predicted and median pathologists’ slide level TPS for all PD-L1 clones (ICC 0.93 (95% CI 0.90-0.94), and for each individual clone: 22C3 ICC 0.93 (95% CI 0.89-0.96); SP142 ICC 0.88 (95% CI 0.79-0.93); SP263 ICC 0.96 (95% CI 0.93-0.97; 28-8 ICC 0.90 (95% CI 0.85-0.93). Conclusions: AIM PD-L1 NSCLC is highly concordant with the gold standard pathologist consensus score across four PD-L1 clones in a large diverse dataset. This model could support patient enrollment and stratification in prospective clinical trials, as well as quality control of staining and pathology drift. Citation Format: Michael Griffin, Mevlana Gemici, Ashar Javed, Nishant Agrawal, Murray Resnick, Limin Yu, Sara Hoffman, Victoria Mountain, Jamie Harisiades, Megan Rothney, Benjamin Glass, Ilan Wapinski, Andrew Beck, Eric Walk. AIM PD-L1-NSCLC: Artificial intelligence-powered PD-L1 quantification for accurate prediction of tumor proportion score in diverse, multi-stain clinical tissue samples [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 471.

Abstract 1985: Ipilimumab and nivolumab induced reactivation of hepatitis B (HBVr) in patient with metastatic squamous cell lung cancer

Abstract Background: Unique adverse events caused by immunological perturbation of immune checkpoint inhibitor (ICI) treatment were reported. Recently, it came into light that ICI can increase the risk of HBVr. Immune-mediated adverse events including immune-mediated hepatitis have been widely reported. However, HBVr is an extremely rare adverse event that has been seldom reported. Herein, we report the first case of ipilimumab/nivolumab induced HBVr in metastatic squamous cell lung cancer patient. Case Presentation: 71-year-old Asian gentlemen with stage IIIB poorly differentiated squamous cell carcinoma of the lung (PD-L1 immunohistochemistry 5-10%) was initially treated with 4 cycles of carboplatin/paclitaxel/pembrolizumab and one dose of maintenance pembrolizumab. The treatment was complicated by immune-related adverse events of thyroiditis and adrenal insufficiency and followed by surgical resection of the tumor. Unfortunately, he developed recurrent metastatic disease of the brain which provoked craniotomy and resection followed by multiple courses of radiation therapy. The patient was then started on treatment with ipilimumab/nivolumab maintenance therapy following 2 cycles of carboplatin/paclitaxel/nivolumab/ipilimumab combined therapy (every 3 weeks). Bevacizumab was added for cerebral edema associated with brain metastasis. The patient was initially noted to be positive for hepatitis B Antigen. However, he was asymptomatic and liver function test (LFT) and liver ultrasound were unremarkable. He had been monitored for LFT at every visit. After 5.5 months from starting the treatment (after seven cycles of ipilimumab/nivolumab and six cycles of bevacizumab), transaminitis was noted and treatment was held. He was admitted for abdominal pain with labs demonstrating hepatocellular injury with ALT 1761 units/L, AST 607 units/L, and total bilirubin 3.3 mg/dL. Hepatitis work-up revealed HBVr (positive for HBs Ag, hepatitis B viral load 11,745). Patient was started on tenofovir daily and LFT was normalized after three weeks without synthetic dysfunction or metabolic encephalopathy. Ipilimumab/nivolumab treatment was held for six weeks and reintroduced without any further significant events. Conclusion: It has recently been reported that pembrolizumab had a strong signal associated with HBVr. However, to our knowledge, this is the first case report for HBVr in squamous cell carcinoma of lung patient treated with ipilimumab/nivolumab. ASCO recommends universal screening for hepatitis B for patients anticipating systemic anticancer therapy. As illustrated by our case, it is important to have hepatitis B screening before starting systemic therapy including immunotherapy. Additional larger studies are warranted to validate the role of each type of ICI in HBVr. Citation Format: Leeseul Kim, Horyun Choi, Yeun Ho Lee, Jinah Kim, Young Kwang Chae. Ipilimumab and nivolumab induced reactivation of hepatitis B (HBVr) in patient with metastatic squamous cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1985.

Abstract 2027: Spatial omics and multiplexed imaging to discover new biomarkers of response or resistance to immune checkpoint inhibitors (ICI) in advanced non-small cell lung cancer (NSCLC)

Abstract Introduction: Despite the impressive outcomes with ICI in NSCLC, only a minority of the patients show long-term benefit from ICI. While PD-L1 immunohistochemistry is an approved companion diagnostic test, it is neither sensitive nor specific. Here, we describe the use of spatial transcriptomics using the GeoMx Digital Spatial Profiler (DSP) as a discovery platform to find biomarkers for ICI response or resistance. Methods: Pre-treatment samples from 56 patients with NSCLC treated with ICI were collected and analyzed in a tissue microarray (TMA) format in two-fold redundancy. The human whole transcriptome, represented by 18000 genes assessed by oligonucleotide-tagged in situ hybridization, was sequenced on the NovaSeq platform to quantify the mRNAs present in each region of interest. Three molecularly defined tissue compartments, defined by fluorescence co-localization (tumor [panCK+], leukocytes [CD45+/CD68-], macrophages [CD68+]) were generated to assess mRNA. All statistical testing was performed using a two-sided significance level of α=0.05. Result: 54000 gene variables were generated per case, from them about 27000 were analyzed after removing targets with expression lower than a prespecified frequency. Expression distribution of EPCAM in tumor, PTPRC in leukocytes and CD68 in macrophages confirmed that DSP successfully profiled the molecular compartments. Differential gene expression (DEG) analysis was performed for 6-month clinical benefit across two distinct tumor TMA spots per patient. This resulted in a DEG list of about 250, 125 and 122 genes in CK, CD45 and CD68 compartment respectively. When we used the intersection of significantly associated with benefit genes the candidate list was reduced dramatically to 5, 10 and 10 distinct genes in CK (CYBA, ESD, SH3BGRL3, TOP2B, UFM1), CD45 (ATRN, DGAT1, DNAAF9, GMPS, IFT20, MPHOSPH6, PKP4, SOCS2, TFAP4, TGFB3) and CD68 (ANKS3, ARFRP1, CTSD, DPYD, GTF3C5, HS2ST1, KIAA0895L, PIGR, SFTPA2, YEATS4) compartment respectively. From them, using the Cox Proportional-Hazards Model we further identified CYBA, ESD and UFM1 in CK, DGAT1 and IFT20 in CD45 and ANKS3 and DPYD in CD68 to predict survival as well. Conclusions: Using DSP technology allows rapid, patient-specific assessment of the transcriptome in TMAs by in situ hybridization in spatially defined molecular compartments. Here we show a small set of candidate genes that are associated with outcome in this ICI-treated cohort. By intersecting two non-adjacent cores of the same patient tissue sample, we have begun to dissect tumor heterogeneity and pilot biomarker candidate information with unique molecularly defined compartments for tumor cells, lymphocytes, and macrophages. Results from this cohort may lead to a novel, spatially defined, transcriptomic approach for developing new biomarkers for immunotherapy. Citation Format: Myrto Moutafi, Sandra Martinez-Morilla, Rolando Garcia-Milian, Thazin Nwe Aung, Ioannis Vathiotis, Niki Gavrielatou, Vasiliki Xirou, Leonidas Salichos, David L. Rimm. Spatial omics and multiplexed imaging to discover new biomarkers of response or resistance to immune checkpoint inhibitors (ICI) in advanced non-small cell lung cancer (NSCLC) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2027.

Abstract CT107: Correlation of biomarkers and clinical benefit of intratumoral BO112 and pembrolizumab in patients with anti PD1 refractory melanoma

Abstract Background: There is a lack of predictive biomarkers for immunotherapy in melanoma. In terms of IHC, there is no strong rational to support the use of PD-L1 expression. BRAF mutations occur in 40-50% of melanomas and the MAPK pathway may also be activated by NRAS mutations. Patients harboring these mutations face usually a worse prognosis. BO-112 is a double stranded synthetic RNA formulated with polyethyleneimine (PEI) that, by mimicking the effect of a viral infection, mobilizes the immune system. Prior data from a phase I trial (NCT02828098) suggest that, when administered intratumorally, it causes an increase in CD8 infiltration and PD-L1 expression. The role of these and other biomarkers is being explored in the present phase II clinical trial. Study Design: Single arm study with BO-112 plus pembrolizumab (NCT04570332) in patients with advanced melanoma in progression to anti-PD1 therapy. Tumors were genotyped by next generation sequencing, whole exome sequencing and tumor mutation burden. Antitumor and immunological effects of the treatment in the tumor microenvironment were assessed by PDL1 and CD8 immunohistochemistry with a paired biopsy performed after 21 days of treatment. Results: A preliminary analysis has been performed, based on patients evaluable for clinical benefit (defined as response or stable disease>16 weeks). Samples from 35 patients have been analyzed, with 24 patients paired biopsies available. Patients with “cold” tumors (PD-L1 negative and CD8 low) at baseline had a trend to lack of clinical benefit. Only basal PD-L1 in the inflammatory component showed a statistically significant correlation with clinical outcome (4/20 (25%) tumors PDL1 IC negative had benefit versus 10/15 (67%) positive), p=0.0053. Fifteen patients had an increase in PD-L1 and 14 patients had increase in CD8 infiltrate after BO-112 treatment; the lack of increase in PDL1 and CD8 after treatment was also predictive of lack of response (p=0.04). BRAF/NRAS driver mutations correlated with positive outcome. Clinical benefit was observed in 4 of 17 (24%) patients not carrying activating mutations whereas 11 out of 18 (61%) patients with BRAF/NRAS activating mutations had clinical benefit (p=0.02), mainly in cutaneous histology (14% versus 65%, p=0.02). Mucosal melanoma patients (n=3) achieved an ORR 66.7% and DCR 100%. The two mucosal melanoma patients with partial response harbored SETD2 mutations and one of them showed extensive cyclic nucleotide-gated (CNGs), indicative of defects in DNA repair pathways. Regarding acral melanoma patients (n=9), no responses were observed, even in the single case with a NRAS mutation. The only patient achieving clinical benefit, with stable disease>16 weeks; had a unique mutation profile, with TP53 (inactivating) and KIT (activating) mutations. Conclusions: Patients basal mutant BRAF/NRAS could have more probability of benefit from BO-112 and pembrolizumab combination. PD-L1 and/or CD8 increase is an early marker of response. These findings could help to select patients in future clinical trials. Further investigation into predictive biomarkers is warranted. Citation Format: Ruth Roman, Philippe Saiag, Caroline Dutriaux, Luis de la Cruz Merino, Eduardo Castanon, Miguel F. Sanmamed, Caroline Robert, Juan F. Rodríguez-Moreno, Pablo Cerezuela-Fuentes, Ana M. Arance, María González Cao, Henry Montaudié, María Pilar López Criado, Julie Charles, Alfonso Berrocal, María del Carmen Álamo de la Gala, Enrique de Miguel, Elisa Funk-Brentano, Sorilla Prey, Eva Muñoz-Couselo, Delvys Rodríguez Abreu, Juan Martín-Liberal, Helena Escuin-Ordinas, Mark Branum, Sonia Macia, Marisol Quintero, Javier Sánchez López, Marya F. Chaney, Beatriz García-Peláez, Marta Vives-Usano, Miguel Ángel Molina, Stéphane Dalle, Iván Márquez-Rodas. Correlation of biomarkers and clinical benefit of intratumoral BO112 and pembrolizumab in patients with anti PD1 refractory melanoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr CT107.

Abstract CT112: AI-powered and manual assessment of PD-L1 are comparable in predicting response to neoadjuvant atezolizumab in patients (pts) with resectable non-squamous, non-small cell lung cancer (NSCLC)

Abstract Background: PD-L1 expression evaluated by immunohistochemistry (IHC) is a well-established predictor of anti-PD-L1/PD-1 cancer immunotherapy (CIT). The Phase II LCMC3 (NCT02927301) study evaluated pre-operative treatment (tx) with atezolizumab (anti-PD-L1) in pts with untreated early stage resectable NSCLC, achieving a 20% major pathologic response (MPR) rate (primary efficacy pts, n=143). A digital PD-L1 scoring method was developed to assess PD-L1 expression as a potential predictive marker for MPR in squamous and non-squamous tumor samples from LCMC3. Methods: Manual scoring was used to determine PD-L1 status on pre-tx biopsy samples using the tumor proportion score (TPS) with a positive threshold of TPS≥50 (22C3). Binary results were correlated with MPR and stratified by squamous/non-squamous histology. A digital pathology workflow for automated PD-L1 scoring was developed to yield a precise continuous PD-L1 TPS. Deep convolutional neural networks trained using pathologist annotations were used to detect individual cells within the tumor and tumor microenvironment and quantify their PD-L1 expression. These cell type predictions were used to compute a digital PD-L1 TPS. LCMC3 pts with available digital and manual PD-L1 scores were then used to assess the role of PD-L1 expression in predicting MPR. Results: PD-L1 scores were available for pre-tx biopsies from 108 pts. No significant difference in scores was seen between histological subtypes. At cutoff (Oct 15, 2021), TPS≥50 was seen in 41 (non-squamous, n=26 [39%]; squamous, n=15 [36%]) of 108 pts and was associated with MPR in non-squamous (odds ratio [OR], 28.6; P<0.001; Fisher’s exact test) but not squamous histology (OR, 1.3; P=1.0). Continuous digital PD-L1 scores (range: 0-100) were highly correlated with local manual PD-L1 scores (range: 0-100) for squamous (n=42, Pearson r=0.90, P<0.001) and non-squamous stained histology slides (n=66, Pearson r=0.90, P<0.001). Continuous digital and manual PD-L1 TPS on pre-tx biopsies (n=108) were predictive of MPR (digital: area under the receiver operating curve (AUROC)=0.678, logistic regression [LR] P=0.01; manual: AUROC=0.675, LR P=0.003). Strikingly, when pts were stratified by histology, PD-L1 scores were predictive of MPR from pre-tx biopsies for non-squamous samples (n=66; digital: AUROC=0.821, LR P=0.002; manual: AUROC=0.819, LR P=0.001) but not for squamous samples (n=42; digital: AUROC=0.519, LR P=0.93; manual: AUROC=0.506, LR P=0.90), despite no significant difference in MPR rate between the 2 groups. Conclusions: These findings support using digitally assessed PD-L1 IHC as a centralized and standardized scoring system and suggest that tumor histological subtype could be an important factor in the utility of PD-L1 as a predictive biomarker for neoadjuvant CIT in early stage NSCLC. Citation Format: John Abel, Christopher Rivard, Filip Kos, Guillaume Chhor, Yi Liu, Jennifer Giltnane, Sara Hoffman, Murray Resnick, Cyrus Hedvat, Amaro Taylor-Weiner, Farah Khalil, Alan Nicholas, Gregory A. Fishbein, Lynette M. Sholl, Natasha Rekhtman, Stephanie Hennek, Ilan Wapinski, Ann Johnson, Michael Montalto, Katja Schulze, Bruce E. Johnson, David P. Carbone, Konstantin Shilo, Andrew H. Beck, Sanja Dacic, William D. Travis, Ignacio Wistuba. AI-powered and manual assessment of PD-L1 are comparable in predicting response to neoadjuvant atezolizumab in patients (pts) with resectable non-squamous, non-small cell lung cancer (NSCLC) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr CT112.

Computer-assisted tumor grading, validation of PD-L1 scoring, and quantification of CD8-positive immune cell density in urothelial carcinoma, a visual guide for pathologists using QuPath

BackgroundAdvances in digital imaging in pathology and the new capacity to scan high-quality images have change the way to practice and research in surgical pathology. QuPath is an open-source pathology software that offers a reproducible way to analyze quantified variables. We aimed to present the functionality of biomarker scoring using QuPath and provide a guide for the validation of pathologic grading using a series of cases of urothelial carcinomas.MethodsTissue microarrays of urothelial carcinomas were constructed and scanned. The images stained with HE, CD8 and PD-L1 immunohistochemistry were imported into QuPath and dearrayed. Training images were used to build a grade classifier and applied to all cases. Quantification of CD8 and PD-L1 was undertaken for each core using cytoplasmic and membrane color segmentation and output measurement and compared with pathologists semi-quantitative assessments.ResultsThere was a good correlation between tumor grade by the pathologist and by QuPath software (Kappa agreement 0.73). For low-grade carcinomas (by the report and pathologist), the concordance was not as high. Of the 32 low-grade tumors, 22 were correctly classified as low-grade, but 11 (34%) were diagnosed as high-grade, with the high-grade to the low-grade ratio in these misclassified cases ranging from 0.41 to 0.58. The median ratio for bona fide high-grade carcinomas was 0.59. Some of the reasons the authors list as potential mimickers for high-grade cases are fulguration artifact, nuclear hyperchromasia, folded tissues, and inconsistency in staining. The correlation analysis between the software and the pathologist showed that the CD8 marker showed a moderate (r = 0.595) and statistically significant (p < 0.001) correlation. The internal consistency of this parameter showed an index of 0.470. The correlation analysis between the software and the pathologist showed that the PDL1 marker showed a robust (r = 0.834) and significant (p < 0.001) correlation. The internal consistency of this parameter showed a CCI of 0.851.ConclusionsWe were able to demonstrate the utility of QuPath in identifying and scoring tumor cells and IHC quantification of two biomarkers. The protocol we present uses a free open-source platform to help researchers deal with imaging and data processing in the surgical pathology field.

Programmed Death-Ligand 1 Expression in Lymphovascular Tumor Emboli in Lung Cancer

IntroductionProgrammed death-ligand 1 (PD-L1) expression determined by immunohistochemistry is the most widely used biomarker for predicting response to immune checkpoint inhibitors. The characteristics of PD-L1 expression in tumor cells inside lymphovascular spaces are largely unknown. Although PD-L1 expression in circulating tumor cells within vascular spaces had been studied, results were conflicting due to lack of standardized PD-L1 expression assessment. MethodsWe investigated PD-L1 expression in lung cancer primary tumor tissue, lymphovascular tumor emboli, and lymph node metastasis using the standard PD-L1 immunohistochemistry 22C3 pharmDx assay. PD-L1 expression was scored in the primary tumor, lymphovascular emboli, and lymph node metastasis by a pathologist using the tumor proportion score (TPS). ResultsWe collected and analyzed surgical specimens from 36 patients with lung cancer with lymph node metastasis. In the primary tumor, 64% of cases were PD-L1 negative (TPS < 1%), 25% were PD-L1 low (TPS 1%–49%), and 11% were PD-L1 high (TPS ≥ 50%). In contrast, in lymphovascular tumor emboli, 89% of cases were PD-L1 negative, 11% were PD-L1 low, and none were PD-L1 high. In lymph node metastases, 72% of cases were PD-L1 negative, 17% were PD-L1 low, and 11% were PD-L1 high. ConclusionsWe observed a significant decrease of PD-L1 expression in lymphovascular tumor emboli compared with that in primary tumors (p = 0.002). Whether such differences are related to intrinsic tumor cell heterogeneity or extrinsic factors such as the microenvironment warrants further investigation.

Immuno-multiple reaction monitoring (iMRM) for quantitation of PD-L1 and PD-1-signaling proteins in non-small cell lung carcinoma (NSCLC).

2627 Background: More accurate predictive biomarkers of response to checkpoint inhibitors (CPIs) still is a major unmet need in oncology. PD-L1 immunohistochemistry (IHC) limitations include its analytical variability and the post-translational modifications of PD-1 signaling-associated proteins like glycosylation. Moreover, PD-L1 IHC is an imperfect surrogate of the tumor immune microenvironment, and immunoscoring is important but difficult to assess in a clinical setting. Proteomic based technologies can overcome these challenges, but the low concentration of these proteins and the presence of high background noise in formalin-fixed paraffin embedded (FFPE) tumors were limiting obstacles. In this study, we evaluate the benefit of a new approach we used with anti-peptide antibodies to purify surrogate peptides, while liquid chromatography (LC) was coupled to multiple reaction monitoring mass spectrometry (iMRM) to improve specificity and precision of protein quantitation. Methods: To determine the concentration of PD-L1, PD-1, PD-L2, NT5E, LCK and ZAP70, we used unique and well detectable proteolytic peptides as surrogates. In a refined protocol, we optimized protein extraction and digestion, peptide immuno-enrichment, LC and MRM parameters to maximize recovery, increase target-specific signal and reduce noise. Plus, we assessed the glycosylation status of PD-L1, PD-L2, and PD-1. The entire workflow was fully validated using 31 NSCLC FFPE tumors. PD-L1 quantitation by iMRM was compared to PD-L1 IHC clone 22C3. Results: On average, 71±29 µg (n = 52) of protein could be extracted from each 1–3 mm3 NSCLC tumor FFPE core. The optimized iMRM method allowed the quantitation of PD-L1 and PD-1 down to 21 amol on-column. Inter- and intra-day repeatability were well below FDA guidelines (coefficients of variation [CV] &lt; 20%) with average CVs of 5.2±4.0% (intra-day) and 4.5±2.6% (inter-day). Sample storage had no significant effect on peptide quantitation. The final multiplexed iMRM assay enables quantitation all targets and glycosylation states for &gt; 40 samples in only 3 days (including external calibration and quality controls) and was used to quantify the PD-1/PD-L1 axis proteins successfully in all 31 NSCLC FFPE tumors. PD-L1 expression ranged from 2 amol/μg to 61 amol/μg of total protein. As expected, iMRM results correlated moderately (R = 0.56, ρ &lt; 0.001) with PD-L1 IHC. PD-L1 glycosylation status ranged from 99.9±0.2%, and therefore did not explain the discrepancies between IHC and iMRM for these samples. Conclusions: Herein a robust iMRM workflow was developed for the quantitation of the PD-1/PD-L1 axis in FFPE. This proof-of-concept supports that MS-based assay can provide otherwise unavailable data (e.g., PD-L1 concentration, glycosylation status). CPI treated patient tumors are being currently processed to validate the predictive value of the assay.

Centrally determined PD-L1 among non-eligible for cisplatin-based chemotherapy recurrent/metastatic head and neck patients (R/M HNSCC pts): Pathology outcomes from the NIVOTAX trial.

e18015 Background: PD-L1 is a biomarker of response to anti-PD1 in HNSCC. CA209-7HE trial (NCT04282109) is a randomized, open-label, multicenter, phase II trial including first line R/M HNSCC pts non-eligible for cisplatin-based chemotherapy. Pts were stratified according to Karnofsky performance status, HPV and tissue PD-L1 by Combined Positive Score (CPS). Subjects were randomized to either paclitaxel + nivolumab or paclitaxel + cetuximab. Primary objective was 2-year overall survival. We present the data on centrally determined PD-L1 patterns and its correlation with pt clinical characteristics. Methods: From 176 preliminary samples, only 141 were analyzed. 10 were excluded for not fulfilling quality requirements and 25 belonged to non-eligible pts. 74% (104) were paraffin blocks and 26% (37) tissue slides. 72.3% (102) were primary tumor biopsies, 12.7% (18) local soft tissue relapses, 9.2% (13) lymph node metastases and 5.6% (8) distant metastases. 46.6% (66) were collected from irradiated areas. Hematoxylin-eosin stain confirmed sample representativeness. At least 100 tumor cells/sample was required. Immunohistochemistry for PD-L1 was performed using DAKO PD-L1 IHC 22C3. Means and proportions were dropped for descriptive analyses. Multivariate logistic regresion was used to identify variables associated with higher CPS. Results: Clinical characteristics included 73% (103) males; 77% (108) were either former [56% (78)] or current smokers [21.2% (30)] and clinical stage was recurrent in 48.9% (69), metastatic alone in 29.8% (42) and both recurrent and metastatic in 21.3% (30). Diagnoses included 35.4% (51) oral cavity, 32.3% (45) oropharynx (only 2 pts HPV positive), 16.3% (23) larynx and 15.6% (22) hypopharynx. Mean time to CPS assessment was 5.25 days (CI 95% 4.95 – 5.55). CPS distribution was as follows: 21.7% (31) were CPS &lt; 1, 42.8% (60) were CPS 1-20 and 35.4% (50) were CPS &gt; 20. Multivariate model revealed that oral cavity tumors showed an increased likelihood of CPS &gt; 20 (OR 2.35, 95%CI 1.46 - 3.78; p &lt; 0.001) which was also seen among non-smokers (OR 3.37, 95% CI 1.16 - 9.73; p = 0.02) and pts with recurrent and metastatic disease (OR 2.75, 95% CI 1.03- 7.32; p = 0.04). Oropharyngeal tumors showed lower probability for CPS &gt; 20 (OR 0.35, 95% CI 0.14 - 0.89; p = 0.03). Neither prior irradiation nor origin of the biopsy were associated with any significant variation in CPS. Conclusions: Central assessment of CPS is feasible with no significant treatment delay. HPV negative oropharyngeal tumors were not associated with higher CPS. We also found that oral cavity tumors, non-smokers and pts with recurrent and metastatic disease were independently associated with an increased probability of higher CPS. Future results will elucidate the promising role of combinatorial regimens with immunotherapy in these subpopulations. Clinical trial information: NCT04282109.

Impact of PD-L1 expression on conventional urothelial bladder carcinoma (UCB) genomic alteration (GA) profile.

e16535 Background: Immunohistochemistry (IHC) to determine PD-L1 expression level has been used as a biomarker to predict immune checkpoint inhibitors response in UCB. We hypothesized that the GA profiles would differ between UCB featuring high vs negative PD-L1 expression. Methods: 102 cases of advanced UCB with known PD-L1 tumor cell expression underwent hybrid-capture based comprehensive genomic profiling to evaluate all classes of genomic alteration (GA). Tumor mutational burden (TMB) was determined on up to 1.1 Mbp of sequenced DNA and microsatellite instability (MSI) was determined on 114 loci. Tumor cell (TC) PD-L1 expression was determined by IHC (Dako 22C3). Only PD-L1 high (H) (≥50% TC expression) and negative (N) (0% TC expression) cases were included in this study. Results: Overall, only 2 (8.3%) of the 24 PD-L1H UCB featured CD274 ( PD-L1) amplification (mean 19 copies) and none of 78 PD-L1N had CD274 amplification (P = .05). The gender, age and GA per tumor frequencies were similar in the groups. When compared with the PD-L1H UBC cases, FGFR3 GA were significantly more frequent in the UBC PD-L1N cases (p = .02). Currently “untargetable” GA that were more frequent in the PD-L1H UBC, but did not reach significance, included TP53, TERT and RB1. MTAP loss, a potential target for PRMT5 and MTA2 inhibitors, were 3X more frequent in the PD-L1N UBC. ERBB2 amplification and ERBB3 and PIK3CA short variant (SV) GA were more frequent in the PD-L1N UBC with differences not reaching significance. The mean gLOH scores were similar in both groups. Other ICPI-associated potential biomarkers, including MSI status, TMB level and GA in PBRM1, STK11 and MDM2 were not significantly different in the groups. For UCB cases where a mutational signature could be determined, 15/33 (45%) of PD-L1H and 34/112 (30%, p = 0.14) of PD-L1N UCB featured an APOBEC gene signature; 79% of PD-L1H and 83% of PD-L1N featured European ancestry (p = 0.61). Conclusions: PD-L1H and PD-L1N subtypes of UCB differ in their genomic profiles. PD-L1N UCB features greater frequencies of potentially “targetable” GA, including FGFR3, ERBB2, ERBB3 and PIK3CA. PD-L1 IHC may thus not only play a role in the selection of ICPI for advanced UCB but also in designing trials that may combine ICPI with targeted therapies.[Table: see text]

Effective biomarker testing rates in a large U.S. community practice.

e21093 Background: Molecular testing is an integral part of care for patients with non-small cell lung cancer (NSCLC). Professional guidelines advise broad molecular profiling to identify driver mutations that have either FDA approved therapies or available clinical trials. Testing is recommended for at least 8 biomarkers plus PD-L1 immunohistochemistry and obtaining biomarker testing results back before initiation of treatment is critical. However, many patients with NSCLC remain untested or receive inadequate treatment prior to obtaining a test result. We performed a retrospective chart review to assess real world effective biomarker testing rates and practice patterns to understand the current biomarker testing scenario in a large US community practice, Florida Cancer Specialists (FCS). Methods: Advanced NSCLC patients (Stage IV; n = 361) who received care in the FCS network and were diagnosed or received first-line treatment on or after 1/1/2020 were included in this retrospective observational study. Data was extracted from the electronic medical records in participating practices as well as via abstraction of patient records when available. Patient-level data was anonymized and reported in aggregate. All data collected for this analysis was evaluated in accordance with regulatory requirements and applicable guidance for electronic records. We determined the percentage of biomarkers tested, orders placed and received before treatment, test ordered after treatment, and the overall effective testing rate, defined as orders placed and result received before treatment start/stage IV patients treated. Results: The median patient age was 70 years (range, 41-90), 188 (52%) were male, and 277 (77%) were Caucasian. 270 (75%), had adenocarcinoma histology, and 273 (76%) had an ECOG status of 0 or 1. Biomarker testing was ordered for 285 patients (79%). EGFR tests were more likely to be ordered before the start of treatment than other biomarkers. The median time between biomarker test ordering to treatment start was 15 days. 217 patients received test results before treatment start (effective testing rate = 60%). 6% of patients had tests ordered later in their workup, while 13% had tests ordered early, but received test results after treatment start. 21% of patients were not tested for any biomarker. 85% of stage IV patients tested received multi-gene tests. 71% of patients had PD-L1 testing and less than 2% received PD-L1 testing alone. Conclusions: This real-world study showed that 85% stage IV NSCLC patients in a large US community practice network underwent multigene biomarker testing and 60% received their test results prior to treatment start. The median time between biomarker test ordering to treatment start was 15 days. Strategies to improve time to biomarker test order, enhance percentage of patients tested, and improvement of percentage of patients with test results prior to initiation of therapy are ongoing.

Pan-cancer landscape of CD274 (PD-L1) and PDCD1LG2 (PD-L2) structural variations.

3133 Background: PD-1 receptor and PD-L1 ligand interactions are the target of immunotherapies across multiple tumor types. Established biomarkers that predict response to immunotherapy are microsatellite instability, tumor mutational burden, and PD-L1 immunohistochemistry. Structural variations in CD274 (PD-L1) and PDCD1LG2 (PD-L2) have been observed in cancer, but the comprehensive landscape is unknown. Here we describe the genomic landscape of CD274 and PDCD1LG2 structural variations, their potential impact on the tumor microenvironment, and evidence that patients with these alterations can benefit from immunotherapy. Methods: We analyzed sequencing data from 514 cancer cases with CD274 and PDCD1LG2 structural variations across 25 publications and data sources, including large pan-cancer sources: Foundation Medicine, Inc (FMI), The Cancer Genome Atlas (TCGA), and Oncology Research Information Exchange Network. To evaluate immune signature enrichment, we ran the software ImSig on gene expression data. We curated literature reporting clinical outcomes of patients harboring structural variations in CD274 and PDCD1LG2. Results: From 25 studies and datasets, we curated 514 cancer cases with structural variations in PD-L1 and PD-L2, including 158 duplications, 126 deletions, 97 inversions, 178 translocations, and 96 unclassified structural variations, totaling to 655 events. We observed breakpoint ‘hotspots’ in the 3’-untranslated regions (UTRs) of PD-L1 and PD-L2. Leveraging TCGA data, we observed, in CD274-rearranged tumors, significant upregulation in PD-L1 and PD-L2 expression and signatures for interferon signaling, macrophages, monocytes, T cells, and immune cell proliferation (each p &lt; 0.001, compared to CD274 non-rearranged, copy neutral tumors). Furthermore, retrospective review of 12 studies that included patients with structural variations in CD274 or PDCD1LG2, including duplications, inversions and copy number amplifications, revealed a 73% (52/71) response rate to PD-1 immunotherapy with durable responses. Conclusions: Our evaluation of CD274 and PDCD1LG2 structural variations shows that the 3’-UTR is frequently affected and is associated with increased expression of ligands and immune signatures. Enriched interferon signaling in CD274-rearranged tumors is of particular interest, as interferon exposure is known to drive PD-L1 and PD-L2 expression. Retrospective evidence from curated studies suggests that these genomic alterations could identify candidates for PD-1 or PD-L1 immunotherapy. We expect that these findings will better define CD274 and PDCD1LG2 structural variations in cancer and support our pan-cancer prospective clinical trial to target these alterations.

Distinct mutational landscapes characterize melanomas metastatic to different anatomical sites.

9562 Background: Despite revolutionary advances in systemic therapies for melanoma, many patients with metastatic disease have limited treatment options and some sites of disease remain particularly challenging to control, such as brain and liver metastases. We sought to define anatomical site-specific mutational profiles of melanoma metastases from which potentially novel personalized therapeutic opportunities may be developed. Methods: Targeted exome genomic profiling was performed via the FoundationOneCDx platform and known or likely pathogenic variants retained for analysis. PD-L1 immunohistochemistry was performed with the Dako 22C3 assay with PD-L1+ defined as tumor proportion score ≥1. Tumor mutational burden (TMB)-high was defined as TMB ≥10mut/Mb. UV signature was calculated using established algorithms. Aberrations in 23 genes potentially actionable in melanoma ( BARD1, BRAF, BRCA1/2, CDKN2A, DDX3X, FANCC, HRAS, IDH1, KIT, KRAS, MAP2K1, NF1, NRAS, PALB2, PIK3CA, PPP6C, PTEN, RAC1, RAD51C/D, RB1, TP53) were compared across metastatic sites including skin, lymph node, lung, liver, and brain. Results: A total of 4918 cutaneous-type melanoma tumors was evaluated, including 2854 skin lesions (primary/metastasis) and metastases from lymph nodes (n = 858), liver (n = 194), lung (n = 342), and brain (n = 200). The commonly mutated genes were CDKN2A (2220/4918, 45.1%), BRAF (2148/4918, 43.7%), NRAS(1347/4918, 27.4%), NF1 (1038/4918, 21.1%), TP53 (1234/4918, 25.1%) and PTEN (1038/4918, 13.5%). Compared with skin lesions, metastases to the lung were enriched for variants affecting NF1 (OR 2.57, p = 8.99e-13), TP53 (OR 1.65, p = 3.04e-04) and depleted in NRAS (OR 0.49, p = 9.00e-06) and BRAF (OR 0.70, p = 1.13e-02). Lung metastases were associated with higher prevalence of UV signatures, PD-L1 positivity and high TMB. PTEN variants were enriched in both brain (OR 3.00, p = 2.38e-08) and small intestine (OR3.46, p = 6.63e-3) metastases. Brain metastases were also enriched for CDKN2A variants (OR 1.56, p = 4.50e-02). Lymph node metastases had lower rates of UV signatures, NF1 variants, or high TMB, but were associated with PD-L1 positivity. Positional analysis of variants revealed that several, including NRAS, MAP2K1, KRAS, HRAS, IDH1, RAC1, GNA11 and GNAQ were highly restricted to hotspot loci, without definite variation by metastatic site. Conversely, variants affecting CDKN2A, PTEN, TP53 and NF1 were distributed widely across the gene, with high levels of non-hotspot mutations observed in skin lesions, suggesting that such non-hotspot subclones contribute less to distant metastatic potential and are lost during disease evolution. Conclusions: We observed distinct and organ site-specific mutational patterns in patients with metastatic melanoma. These data raise the possibility of metastatic phenotype-directed therapy to improve the personalization and outcomes of treatment for this disease.

Endocervical adenocarcinoma biomarkers for immunotherapy: Descriptive analysis.

e17526 Background: Adenocarcinoma and adenosquamous carcinoma of the cervix are subtypes of HPV related cervical cancer. As virally mediated tumors, they have become an obvious choice for immunotherapy treatment approach. Currently, there is little information about the presence of immunological tumor markers among patients with endocervical adenocarcinomas. We hypothesized cervical adenocarcinomas highly express immunotherapy markers. Here we characterize the protein status of immunotherapy markers in a large single institution dataset. Methods: An IRB approved retrospective review of cervical adenocarcinoma and adenosquamous carcinoma was performed. We gathered original slides for each patient and performed immunohistochemistry (IHC) for tumor markers CD8, PDL1, TIM3 and LAG3. Demographic and pathologic data was captured. Patients were categorized by positivity percentage for each marker. Descriptive statistics and univariate analysis were used. Results: A total of 97 patients met criteria for the study and slides were available for staining in 68 patients. Demographic data showed a median age of 42 years old with majority of patients identifying as White at 66%, followed by Hispanic (13%), Asian (10%), other (7%) and Black (1%). Of the entire cohort, 75% were adenocarcinoma, 25% were adenosquamous carcinoma. Among those with recorded status, 82% were HPV positive, 14% had metastasis, and 13% had recurrence. On univariate analysis, adenocarcinoma was significantly more common in patients who were LAG3 &lt; 1% or PDL &lt; 1% (p = 0.002, p = 0.007). Patients who were LAG3 &lt; 1% or PDL &lt; 1% were also less likely to have Grade 1 tumors (p = 0.006, 0.03). When analyzing by respective immune marker percentage group there were no statistically significant differences in LVSI, depth of invasion, stage, or recurrence rate. By IHC, 68% were PDL1 &gt; 1% positive, 63% were LAG3 &gt; 1% positive, 100% were TIM3 &gt; 1% positive, 100% were CD8 &gt; 1% positive. Seven of the 68 patients had a recurrence. Of those patients with recurrence, one was treated with immunotherapy. Conclusions: A large proportion of cervical adenocarcinomas displayed immunotherapy targets by IHC suggesting that combined immunotherapy agents may be useful in this population. Further study is needed to assess whether IHC is predictive of outcome or response.[Table: see text]

Multimodal integration of radiology, pathology, and genomics for prediction of response to PD-1 blockade in patients with non–small cell lung cancer.

9064 Background: Immunotherapy is now given to almost all patients with advanced non-small cell lung cancer (NSCLC). However, developing robust biomarkers to predict benefit remains challenging. We set out to evaluate the predictive capacity of integrating medical imaging, histopathologic, and genomic features to develop a multimodal biomarker for immunotherapy response. Methods: We used baseline data that is routinely obtained during diagnostic clinical workup at a single center in patients with NSCLC and known outcomes following immunotherapy. The multimodal dataset included DNA alterations from NGS, CT scan images, and digitized PD-L1 immunohistochemistry (IHC) slides. Guided by experts in each field, we developed a workflow to extract data for each patient and used an attention-gated machine learning approach to integrate the features into a risk prediction model. Results: Our cohort included 247 patients with advanced NSCLC who received immunotherapy and had complete radiology, pathology, genomic, and clinical data. The patient cohort was 54% female, had a median age of 68 years (range 38-93), and 88% patients had a smoking history. Responders (CR/PR) vs non-responders (SD/PD) showed a median PFS and OS of 2.7 months (95% CI 2.5-3.0) and 11.4 months (95% CI 10.3-12.8), respectively. Of all patients, 187 (76%) had segmentable disease on chest CT scans. We used a radiomics approach and aggregated the average individual lesion predictions to construct patient-level response predictions which resulted in an overall AUC = 0.65, 95% CI 0.57-0.73. We next studied digitized FFPE slides of pre-treatment PD-L1 IHC staining of tumor specimens. Overall, 52% of slides showed PD-L1 tumor proportion score (TPS) ≥ 1% and were used to extract IHC-texture, a novel spatial characterization of PD-L1 staining. Logistic regression modeling on IHC-texture resulted in prediction accuracy of AUC = 0.62 (95% CI 0.51-0.73) which was inferior to the pathologist-assessed PD-L1 TPS (AUC = 0.73, 95% CI 0.65-0.81). We next implemented a dynamic deep attention-based multiple instance learning model with masking to evaluate the impact of combining features from all modalities. Our multimodal model (AUC = 0.80, 95% CI 0.74-0.86) outperformed unimodal measures, including tumor mutational burden (AUC = 0.61, 95% CI 0.52-0.70) and PD-L1 TPS (AUC = 0.73, 95% CI 0.65-0.81). Conclusions: Our study is a proof of concept for using multimodal features to improve prediction of immunotherapy response over standard-of-care approaches in patients with NSCLC using expert-guided machine learning.

Clinical outcomes and immune markers by race in a phase I/II clinical trial of durvalumab concomitant with neoadjuvant chemotherapy in early-stage TNBC.

516 Background: The incidence of triple negative breast cancer (TNBC) is higher among Black or African American (AA) women, yet AA patients (pts) are underrepresented in clinical trials, exemplifying racial disparity in oncology. We conducted a phase I/II trial to assess the safety and efficacy of durvalumab concurrent with weekly nab-paclitaxel and dose dense doxorubicin/cyclophosphamide (ddAC) neoadjuvant therapy for stage I-III TNBC. The primary efficacy endpoint was pathologic complete response (pCR; ypT0/is,N0) rate. Given the unclear efficacy and safety of immunotherapy in AA pts with breast cancer, we extended our accrual to recruit AA pts, with the goal of evaluating the association between racial groups and PD-L1 expression, stromal tumor infiltrating lymphocytes (sTILs), toxicities, treatment response and survival. Methods: Our study population included 67 pts. PD-L1 immunohistochemistry results and sTIL counts were available on 59 and 60 pts, respectively. Chi-Squared test was used to evaluate associations between race and baseline characteristics. Cox proportional hazards model was used to assess association between AA race and overall survival (OS) and event free survival (EFS), adjusting for age, comorbidities and pCR status. Multivariate logistic regression analyses were used to evaluate the association between race and pCR, development of immune-related adverse events (irAEs) and breast cancer recurrence. Results: Twenty-one pts (31%) self-identified as AA. No significant associations between AA race and baseline body mass index (BMI; p=0.075), Charlson comorbidity index (p=0.32), tumor stage (p=0.40), grade (p=0.54), PD-L1 status (0.92) and sTIL count (p=0.57) were observed. pCR rates did not significantly differ between AA and non-AA pts: 9/21 (43%) AA vs. 22/48 (48%) non-AA (p=0.71). 3-yr OS was 87% in the non-AA versus 81% in the AA cohort (HR 1.72, 95% CI 0.481-6.136; p=0.405); 3 yr EFS were 78.3% and 71.4% in non-AA and AA pts respectively. (HR 1.451, 95% CI 0.524-4.017; p=0.474). Pts with pCR were more likely to remain event-free at 3 yrs, irrespective of race (HR 0.234, 95% CI 0.066-0.829; p=0.024). In multivariate logistic regression analyses, lack of pathologic response (OR for pCR 0.17, 95% CI 0.03-0.7; p=0.02) and node positive status (OR 4.13, 95% CI 1.05-19.88; p=0.05) were associated with recurrence. The incidence of irAEs was similar between AA and non-AA pts and no significant associations were found between irAEs and pathologic response. Conclusions: pCR rates after neoadjuvant immunotherapy and chemotherapy were similar in AA and non-AA pts. Stromal TILs, PD-L1 status, 3yr OS and EFS, and the frequency of irAEs were also similar. These results suggest that when patients receive identical treatment and are monitored closely, disparities in outcomes can be mitigated or abolished. Clinical trial information: NCT02489448.

PD-L1 (SP142 and 22C3) immunohistochemistry in clinical metastatic triple-negative or low hormone receptor breast carcinomas: experience from a large academic institution

The discovery of new immune checkpoint molecules has led to the emergence of new treatments for metastatic breast carcinoma. We aimed to investigate PD-L1 (SP142 and 22C3) expression in 77 clinical metastatic triple-negative or low hormone receptor (<10%) breast carcinomas (TNBC/LHRBC). SP142 was positive in 23.4%of cases (18/77). SP142-positive (SP142+) cases showed lower liver metastasis and androgen receptor expression, but increased tumor-infiltrating lymphocytes than SP142-negative (SP142-) cases in univariate analysis, but only increased tumor-infiltrating lymphocytes in multivariate analysis. 22C3 testing was available in 21 cases including 14 with combined positive score (CPS) ≥10 (9 SP142- and 5 SP142+) and 7 with CPS <10 (7 SP142-). Ten (13%) patients received immunotherapy including 8 SP142+ (7 with atezolizumab and 1 with pembrolizumab) and 2 SP142- cases (2 with pembrolizumab). Survival data showed a trend of increased survival rate in SP142+ (72.2%) when compared to SP142- patients (55.9%). Our study provides unique insights into the distribution of PD-L1 staining in metastatic breast carcinomas in a real-world clinical setting.