QuPath Software Research Articles

Evaluation of spatial PD1 and PD-L1 expression in inflammatory bowel disease samples - a pilot study.

Alterations of PD1/PD-L1 pathway may be associated with an excessive inflammatory response in the intestinal wall in inflammatory bowel diseases (IBD). To evaluate the expression of PD-1 and PD-L1 in 4 compartments of intestinal wall (mucosa, submucosa, muscularis propria and lymphatic follicles), high-resolution immunohistochemically stained slides were obtained from formalin-fixed paraffin-embedded samples of 10 Crohn's disease (CD), 9 ulcerative colitis (UC) and 10 unaffected individuals cases. The levels of expression were quantified using the QuPath software. PD-1 was detected in lymphatic follicles in affected and unaffected tissue samples and in inflammatory infiltration in IBD. There was no difference between groups neither in PD-1 overall expression nor in individual compartments, with the exception of the mucosal expression. It was higher in the mucosa of CD patients comparing to controls, however this difference was marginal (p = 0.0461). PD-L1 was expressed in endothelium and mesenteric nervous plexi, consistently in each group. There were no significant differences in PD-L1 immunoreactivity in context of histologic compartment nor clinical diagnosis. The results suggest that PD-1 and PD-L1 expression in intestinal tissue is heterogeneous in the analysed groups, thus it may be dependent on individual characteristics.

Upregulation of HLA class II in pancreatic beta cells from organ donors with type 1 diabetes.

We aimed to characterise and quantify the expression of HLA class II (HLA-II) in human pancreatic tissue sections and to analyse its induction in human islets. We immunostained human pancreatic tissue sections from non-diabetic (n = 5), autoantibody positive (Aab+; n = 5), and type 1 diabetic (n = 5) donors, obtained from the Network of Pancreatic Organ Donors (nPOD), with HLA-II, CD68 and insulin. Each tissue section was acquired with a widefield slide scanner and then analysed with QuPath software. In total, we analysed 7415 islets that contained 338,480 cells. Widefield microscopy was further complemented by high resolution imaging of 301 randomly selected islets, acquired using a Zeiss laser scanning confocal (LSM880) to confirm our findings. Selected beta cells were acquired in enhanced resolution using LSM880 with an Airyscan detector. Further, we cultured healthy isolated human islets and reaggregated human islet microtissues with varying concentrations of proinflammatory cytokines (IFN-γ, TNF-α and IL-1β). After proinflammatory cytokine culture, islet function was measured by glucose-stimulated insulin secretion, and HLA-I and HLA-II expression was subsequently evaluated with immunostaining or RNA sequencing. Insulin-containing islets (ICIs) of donors with type 1 diabetes had a higher percentage of HLA-II positive area (24.31%) compared with type 1 diabetic insulin-deficient islets (IDIs, 0.67%), non-diabetic (3.80%), and Aab+ (2.31%) donors. In ICIs of type 1 diabetic donors, 45.89% of the total insulin signal co-localised with HLA-II, and 27.65% of the islet beta cells expressed both HLA-II and insulin, while in non-diabetic and Aab+ donors 0.96% and 0.59% of the islet beta cells, respectively, expressed both markers. In the beta cells of donors with type 1 diabetes, HLA-II was mostly present in the cell cytoplasm, co-localising with insulin. In the experiments with human isolated islets and reaggregated human islets, we observed changes in insulin secretion upon stimulation with proinflammatory cytokines, as well as higher expression of HLA-II and HLA-I when compared with controls cultured with media, and an upregulation of HLA-I and HLA-II RNA transcripts. After a long-standing controversy, we provide definitive evidence that HLA-II can be expressed by pancreatic beta cells from patients with type 1 diabetes. Furthermore, this upregulation can be induced in vitro in healthy isolated human islets or reaggregated human islets by treatment with proinflammatory cytokines. Our findings support a role for HLA-II in type 1 diabetes pathogenesis since HLA-II expressing beta cells can potentially become a direct target of autoreactive CD4+ lymphocytes.

Digital Pathology Enables Automated and Quantitative Assessment of Inflammatory Activity in Patients with Chronic Liver Disease.

Traditional histological evaluation for grading liver disease severity is based on subjective and semi-quantitative scores. We examined the relationship between digital pathology analysis and corresponding scoring systems for the assessment of hepatic necroinflammatory activity. A prospective, multicenter study including 156 patients with chronic liver disease (74% nonalcoholic fatty liver disease-NAFLD, 26% chronic hepatitis-CH etiologies) was performed. Inflammation was graded according to the Nonalcoholic Steatohepatitis (NASH) Clinical Research Network system and METAVIR score. Whole-slide digital image analysis based on quantitative (I-score: inflammation ratio) and morphometric (C-score: proportionate area of staining intensities clusters) measurements were independently performed. Our data show that I-scores and C-scores increase with inflammation grades (p < 0.001). High correlation was seen for CH (ρ = 0.85–0.88), but only moderate for NAFLD (ρ = 0.5–0.53). I-score (p = 0.008) and C-score (p = 0.002) were higher for CH than NAFLD. Our MATLAB algorithm performed better than QuPath software for the diagnosis of low-moderate inflammation (p < 0.05). C-score AUC for classifying NASH was 0.75 (95%CI, 0.65–0.84) and for moderate/severe CH was 0.99 (95%CI, 0.97–1.00). Digital pathology measurements increased with fibrosis stages (p < 0.001). In conclusion, quantitative and morphometric metrics of inflammatory burden obtained by digital pathology correlate well with pathologists’ scores, showing a higher accuracy for the evaluation of CH than NAFLD.

Quantitative analysis of KLF4 and SOX2 expression in oral carcinomas reveals independent association with oral tongue subsite location and histological grade.

Stemness factors associated with tumorigenesis in different types of cancers have not been specifically studied in oral tongue SCC (OTSSC). Here, we aimed to quantify expression levels and distribution of KLF4 and SOX2, two relevant stemness factors, in oral SCC including OTSCC samples from different subsites. We determined KLF4 and SOX2 expression levels by immunostaining 35 biopsies of OSCC. Stained wholeslide images were digitized and subjected to automatic cell detection and unbiased quantification using Qupath software. We found statistically significant reduction in KLF4 positive cells density (p= 0.024), and fraction (p= 0.022) in OTSCC from tongue borders compared with other tongue subsites. Instead, quantitative SOX2 analysis did not show differences in expression levels between OTSCC from the borders versus OTSCC developed in others subsites. Notably SOX2 expression was revealed increased in moderately and poorly differentiated OSCC compared with well differentiated ones (positive cells density p= 0.025, fraction p= 0.006). No significant correlation between KLF4 and SOX2 expression was observed, neither in OSCC nor in OTSCC. KLF4 and SOX2 exhibit opposite expression profiles regarding subsite localization and differentiation level in OSCC. Our study prompts future OTSCC prospective studies looking for clinical prognosis to incorporate detailed subsite information in the analysis.

IL-17 is expressed on beta and alpha cells of donors with type 1 and type 2 diabetes

PurposeIL-17 is an important effector cytokine driving immune-mediated destruction in autoimmune diseases such as psoriasis. Blockade of the IL-17 pathway after the initiation of insulitis was effective in delaying or preventing the onset of type 1 diabetes (T1D) in rodent models. Expression of IL-17 transcripts in islets from a donor with recent-onset T1D has been reported, however, studies regarding IL-17 protein expression are lacking. We aimed to study whether IL-17 is being expressed in the islets of diabetic donors. MethodsWe stained human pancreatic tissues from non-diabetic (n = 5), auto-antibody positive (aab+) (n = 5), T1D (n = 6) and T2D (n = 5) donors for IL-17, Insulin, and Glucagon, and for CD45 in selected cases. High resolution images were acquired with Zeiss laser scanning confocal microscope LSM780 and analyzed with Zen blue 2.3 software. Cases stained for CD45 were also acquired with widefield slide scanner and analyzed with QuPath software. ResultsWe observed a clear cytoplasmic staining for IL-17 in insulin-containing islets of donors with T1D and T2D, accounting for an average of 7.8 ± 8.4% and 14.9 ± 16.8% of total islet area, respectively. Both beta and alpha cells were sources of IL-17, but CD45+ cells were not a major source of IL-17 in those donors. Expression of IL-17 was reduced in islets of non-diabetic donors, aab+ donors and in insulin-deficient islets of donors with T1D. ConclusionOur finding that IL-17 is expressed in islets of donors with T1D or T2D is quite intriguing and warrants further mechanistic studies in human islets to understand the role of IL-17 in the context of metabolic and immune stress in beta cells.

O26 Intertumoural heterogeneity in the structure of the neurovascular unit in glioma: a retrospective observational cohort study

Abstract Introduction The neurovascular unit (NVU) is implicated in glioma tumourigenesis and treatment resistance but intertumoural heterogeneity is not well characterised. We evaluated the structure of the glioma NVU across tumour grades, in primary and recurrent disease. We determined association to overall survival. Method Ethics approval was obtained (15/ES/0094). A tissue microarray was constructed using biopsies from 137 patients and was immunostained for endothelial cells (CD31), extracellular matrix (collagen-IV), pericytes (PDGFR-β and α-SMA) and tight junctions (claudin-V). The immunostaining was assessed using QuPath software and compared by one-way analysis of variance with Tukey post-hoc testing. Association to overall survival was by univariate Cox regression with hazard ratios (HR). Result Primary grade IV tumours had greater CD31+ than grade II tumours (P = 0.046) or recurrent tumours (P = 0.013). After normalising to CD31+, collagen-IV+ was greater in recurrent tumours than in primary grade IV tumours (P = 0.013). PDGFR-β+ was greater in primary grade II tumours than in primary grade III tumours (P = 0.002), primary grade IV tumours (P &amp;lt; 0.001) or recurrent tumours (P &amp;lt; 0.001). Normalised α-SMA+ was greater in recurrent tumours than in primary grade IV tumours (P = 0.021). CD31 + (HR: 1.028, P = 0.004) and collagen-IV + (HR: 1.014, P = 0.008) were negative prognostic factors. After normalisation, PDGFR-β + (HR: 0.975, P = 0.004), α-SMA + (HR: 0.971, P = 0.020) and claudin-V + (HR: 0.994, P = 0.021) were positive prognostic factors. Conclusion Neurovascular remodeling is a feature of malignancy and recurrent tumours have altered neurovascular phenotypes. Markers for pericytes, extracellular matrix and tight junctions supplement classical grading features and predict tumour behavior. Take-home Message Neurovascular remodeling is a feature of malignancy and recurrent tumours have altered neurovascular phenotypes. Markers for pericytes, extracellular matrix and tight junctions supplement classical grading features and predict tumour behavior.

Whole-Slide Image Analysis of Human Pancreas Samples to Elucidate the Immunopathogenesis of Type 1 Diabetes Using the QuPath Software.

Type 1 diabetes is a chronic disease of the pancreas characterized by the loss of insulin-producing beta cells. Access to human pancreas samples for research purposes has been historically limited, restricting pathological analyses to animal models. However, intrinsic differences between animals and humans have made clinical translation very challenging. Recently, human pancreas samples have become available through several biobanks worldwide, and this has opened numerous opportunities for scientific discovery. In addition, the use of new imaging technologies has unraveled many mysteries of the human pancreas not merely in the presence of disease, but also in physiological conditions. Nowadays, multiplex immunofluorescence protocols as well as sophisticated image analysis tools can be employed. Here, we described the use of QuPath—an open-source platform for image analysis—for the investigation of human pancreas samples. We demonstrate that QuPath can be adequately used to analyze whole-slide images with the aim of identifying the islets of Langerhans and define their cellular composition as well as other basic morphological characteristics. In addition, we show that QuPath can identify immune cell populations in the exocrine tissue and islets of Langerhans, accurately localizing and quantifying immune infiltrates in the pancreas. Therefore, we present a tool and analysis pipeline that allows for the accurate characterization of the human pancreas, enabling the study of the anatomical and physiological changes underlying pancreatic diseases such as type 1 diabetes. The standardization and implementation of these analysis tools is of critical importance to understand disease pathogenesis, and may be informative for the design of new therapies aimed at preserving beta cell function and halting the inflammation caused by the immune attack.

Digital image analysis of immunohistochemistry KI-67 using QuPath software in breast cancer

Background: Today, pathology services are more developed for quantitative diagnostic evaluation. The quantitative diagnostic evaluation requires detailed accuracy and can be done using digital image analysis (DIA). Assessment of the Ki67 labelling index (LI) in breast carcinoma needs to be done quantitatively. A visual evaluation of Ki67 LI using light microscopy has high inter-observer variability. The evaluation of Ki67 LI could be done digitally with the DIA technique to overcome the inter-observer variability. The DIA technique is carried out by counting the Ki67 LI manually or automatically with bioimage analysis software. QuPath is one of the bioimage analysis software, has characteristics of cross-platform, intended for bioimage analysis and digital pathology. Objective: This study aims to compare the manual and automatic calculation of Ki67 LI digitally. Methods: This study was a cross-sectional study; a total of 240 digital Ki67 images from 30 slides were analyzed by counting manually and automatically using QuPath. Results: Statistical analysis using the T-test showed no significant difference between the manual and automatic counting of Ki67 LI (p = 0,801, α = 0,05). Conclusion: Digital image analysis using QuPath can be used to calculate the Ki67 LI automatically.

Automated PD-L1 Scoring for Non-Small Cell Lung Carcinoma Using Open-Source Software

PD-L1 expression in non-small cell lung cancer (NSCLC) is predictive of response to immunotherapy, but scoring of PD-L1 immunohistochemistry shows considerable interobserver variability. Automated methods may allow more consistent and expedient PD-L1 scoring. We aimed to assess the technical concordance of PD-L1 scores produced using free open source QuPath software with the manual scores of three pathologists. A classifier for PD-L1 scoring was trained using 30 NSCLC image patches. A separate test set of 207 image patches from 69 NSCLC resection cases was used for comparison of automated and manual scores. Automated and average manual scores showed excellent correlation (concordance correlation coeffecient = 0.925), though automated scoring resulted in significantly more 1–49% scores than manual scoring (p = 0.012). At both 1% and 50% thresholds, automated scores showed a level of concordance with our ‘gold standard’ (the average of three pathologists’ manual scores) similar to that of individual pathologists. Automated scoring showed high sensitivity (95%) but lower specificity (84%) at a 1% threshold, and excellent specificity (100%) but lower sensitivity (71%) at a 50% threshold. We conclude that our automated PD-L1 scoring system for NSCLC has an accuracy similar to that of individual pathologists. The detailed protocol we provide for free open source scoring software and our discussion of the limitations of this technology may facilitate more effective integration of automated scoring into clinical workflows.

412 Neurobehavioural and neuropathological consequences of fetal growth restriction in a rabbit model

To characterize the behavioural and neuropathological effects in a uteroplacental ligation FGR rabbit model. At gestational day (GD) 25, time mated does underwent partial uteroplacental vessel ligation (UPVL) in one randomly chosen horn, leaving the contralateral foetuses as internal controls. Does were delivered at GD 30 by C-section, and neurobehavioural assessment (NBA) was performed 24h after delivery. Thereafter neonates were transcardially perfused with heparinised saline and 4% paraformaldehyde. Brains were further immerse-fixated for 72 hrs, weighed, embedded and serial sectioned. Slides were stained with Cresyl violet, scanned, and neuron count was performed manually using QuPath software by an evaluator blinded to the specimen allocation. Brain slides from 12 cases and 12 controls were selected according to sample quality. Normal distribution was assessed by Shapiro-Wilk test and data presented as a mean with standard deviations or median and interquartile ranges. The comparison was done by unpaired students t-test or Mann Whitney test. A p-value < 0.05 was considered significant. Neonatal survival was 45% (27/60) in the ligated and 85% (44/52) in the non-ligated horn. UPVL resulted in lower birth weight (35.00±7.14 g vs. 39.05±9.94 g; p<0.05), and no difference in brain-to-body weight ratio (0.046±.005 vs. 0.044±.006; p=0.7). FGR resulted in significantly lower motor (19±4.25 vs. 20±2, p=0.007) and sensory scores (12±4 vs. 14±1.25, p=0.003) on early NBA, and was associated with lower neuron density in the prefrontal cortex (227.8±67 vs. 327±115 n/mm2; p<0.05), caudate nucleus (533.3±85 vs. 444.2±50 n/mm2; p<0.01), and dentate gyrus (222.5±34 vs. 292±49 n/mm2; p=0.001). FGR is associated with functional neurological impairment and structural alterations in the prefrontal cortex, basal ganglia and hippocampus in this rabbit model

Tumour Stroma Ratio Assessment Using Digital Image Analysis Predicts Survival in Triple Negative and Luminal Breast Cancer

Simple SummaryTumour stroma is known to predict outcome and play an important role in the growth and spread of solid tumours and their response to therapy. In breast cancer, there is evidence that the tumour stroma ratio (TSR) can predict outcome in aggressive triple negative breast cancer, but its value for the more common hormone receptor positive breast cancer is unclear. Using computerised image analysis and machine learning algorithms, we show that TSR is an important factor in predicting outcome for triple negative disease and hormone receptor positive cancer. However, its influence on good or poor outcome appears to depend on tumour type and the relative predominance of the stromal component. By better understanding the role of the tumour stroma in cancer growth, and its response to treatment, this study may help support the role of TSR as a new prognostic marker for breast cancer to guide clinical decision making.We aimed to determine the clinical significance of tumour stroma ratio (TSR) in luminal and triple negative breast cancer (TNBC) using digital image analysis and machine learning algorithms. Automated image analysis using QuPath software was applied to a cohort of 647 breast cancer patients (403 luminal and 244 TNBC) using digital H&E images of tissue microarrays (TMAs). Kaplan–Meier and Cox proportional hazards were used to ascertain relationships with overall survival (OS) and breast cancer specific survival (BCSS). For TNBC, low TSR (high stroma) was associated with poor prognosis for both OS (HR 1.9, CI 1.1–3.3, p = 0.021) and BCSS (HR 2.6, HR 1.3–5.4, p = 0.007) in multivariate models, independent of age, size, grade, sTILs, lymph nodal status and chemotherapy. However, for luminal tumours, low TSR (high stroma) was associated with a favourable prognosis in MVA for OS (HR 0.6, CI 0.4–0.8, p = 0.001) but not for BCSS. TSR is a prognostic factor of most significance in TNBC, but also in luminal breast cancer, and can be reliably assessed using quantitative image analysis of TMAs. Further investigation into the contribution of tumour subtype stromal phenotype may further refine these findings.

Abstract 2607: Integrin Beta4/Paxillin mediated chemoresistance in lung adenocarcinoma

Abstract Introduction: Non-small cell lung cancer (NSCLC) accounts for 85% of the diagnosed lung cancers, of which 40% are lung adenocarcinoma (LUAD). These adenocarcinomas are often diagnosed with advanced or metastatic disease and are typically treated with a combination of cisplatin and immunotherapy. In most cases, malignant cells exposed to cisplatin adapt and become less susceptible to the anti-proliferative and cytotoxic effects of the drugs. Thus, many cisplatin-treated patients experience therapeutic failure and tumor recurrence. Therefore, there is a dire need for a deeper understanding of chemoresistance and the identification of prognostic and predictive markers to discern responders from non-responders. Here, we interrogate the role of Integrinβ4 (ITGB4) and Paxillin (PXN) in developing cisplatin resistance that are components of the focal adhesion complex. Experimental Procedures: Validated lung adenocarcinoma cancer cell lines were engineered to express nuclear specific red fluorescence protein for live tracking of cell proliferation, wound healing and cell death using InCucyte Live Cell Imaging System. Cell line-derived spheroids were generated to recapitulate in vivo conditions and spheroid integrity was monitored over time. Changes in gene expression were checked using western blotting or qPCR. Mitochondrial activity was analyzed using Seahorse and genomic stability was measured using γH2AX staining. Immunohistochemical analysis of the stained tissue section was done using the QuPath software. Summary: Cells with higher expression of PXN and ITGB4 are ~50% less sensitive to cisplatin as compared to low expressing cells. Silencing both PXN and ITGB4 in resistant cells inhibited their proliferation by 6-fold, migration by 30 - 40% and increase cisplatin sensitivity by ~ 60%, in 2D and 3D cultures. Differential gene expression analysis revealed significant enrichment of down regulated genes belonging to pathways like TargetsV1, G2-M checkpoint and E2F targets. VDAC1 and USP1 of MYC target V1 pathway were downregulated ~70% and ~ 50% respectively. Knocking down USP1 increased number of γH2AX foci counts, suggesting increased susceptibility to DNA damage. Knocking down VDAC1 increase mitochondrial respiration ~ 30% and increase ROS production by 3-fold. H3K27 acetylation dropped by ~70% at USP1 promoter site in double knocked down cells, inhibiting its transcription. Cisplatin treatment on a heterogeneous population enriched a fraction of cells expressing higher ITGB4 by 2-fold. Immunohistochemical analysis of tumor tissue sections revealed heterogeneity in expression of PXN, ITGB4 or both. Similar, data was also observed in exosomes isolated from patient serum. Conclusion: ITGB4 and PXN co-expression helps cancer cells to maintain their genomic stability and mitochondrial function, thus helping it to evade drug toxicity. NSCLC adenocarcinomas harboring high expression of them are likely to develop drug resistance. Citation Format: Atish Mohanty, Arin Nam, Alex Pozhitkov, Lu Yang, Michael Nelson, Xiwei Wu, Prakash Kulkarni, Erminia Massarelli, Ravi Salgia. Integrin Beta4/Paxillin mediated chemoresistance in lung adenocarcinoma [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 2607.

Abstract 4984: Where do the T lymphocytes go in prostate cancer: Correlation with CD103, CD39 expression and clinical pathological parameters

Abstract Introduction: Many solid tumors such as prostate cancer have shown limited response to immunotherapy treatment. Whether this is due to absence of tumor specific antigens or poor immune infiltrate, the exact mechanism is not well understood. Our study evaluates the presence and spatial distribution in benign versus malignant tissue of CD4+ and CD8+ cells in whole mounted radical prostatectomies samples and characterized the subset of intraepithelial CD103+ and CD39+ cells, which have been previously shown to harbor more cancer specific T-cells. Materials and Methods: Sixteen cases of whole mounted radical prostatectomy samples were used to asses CD4, CD8, CD103 (all abcam) and CD39 (Sigma) expression by IHC. Location of positive cells within epithelium and stroma in tumor and benign prostatic tissue was evaluated. The results were confirmed using QuPath Software, for positive detection of immune cells by color deconvolution and threshold adjustments. Results: The CD4 and CD8 staining demonstrated that most of the immune cells were present in the stroma but not within the malignant glands. The whole prostate sections (mean size 51.4 × 44mm) generated around 1650 ROI's per/case. All the ROI's were included in the analysis. CD4+ and CD8+ were evaluated in whole gland vs tumor area. No significant difference = of CD4+ and CD8+ cell density was seen between whole prostate area and tumor areas (CD4+ 2.9 vs 1.9%) and (CD8+ 1.3 vs 1.2%). The density of CD4+ and CD8+ cells in tumors vs benign gland was approximately the same (means 0.93 for both). For the CD103+ analysis, the positive cells were rarely found in the tumor epithelium in comparison with the benign glands (mean 1.09 vs 5.41; representing 2 vs 9.3% of the total lymphocytes). The CD39 stain showed positive cells mostly at the benign epithelium. No correlation was found between our results and Gleason grade, stage or prognosis. Conclusion: Our study suggests the possibility that in solid cancers such as prostate cancer, immune cells may not be able to reach appropriately the malignant epithelium with only a small subset of T cells reaching their target. This may have important clinical implication for the success of the immune therapy in this group of patients. Further studies are being performed to deeply characterized the tumor infiltrating cell population. Citation Format: Beatriz Walter, Billel Gasmi, Xu Naizen, Vladimir Valera, Peter Pinto, Maria J. Merino. Where do the T lymphocytes go in prostate cancer: Correlation with CD103, CD39 expression and clinical pathological parameters [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4984.

Computerised scoring protocol for identification and quantification of different immune cell populations in breast tumour regions by the use of QuPath software.

As important prognostic and predictive information can be obtained from the composition, functionality and spatial arrangement of different immune cell subtypes, this study aims at characterizing the immune infiltrate in breast tumours. Tumour-infiltrating lymphocytes (TILs) in 62 patients with luminal B-like breast cancer were characterised by immunohistochemical staining with standard markers, and were subsequently classified and quantified by the use of QuPath software. In different delineated tumour regions, the proportion and density of CD3+ , CD4+ , CD5+ , CD8+ , CD20+ and FOXP3+ cells were assessed. The results of the software analysis were compared with those of manual counting for CD8 and CD20 staining. The QuPath scoring protocol slightly overestimated positive, negative and total lymphocyte counts and density, while minimally underestimating the proportion of positively stained lymphocytes. However, for density and proportion, no real differences from manual counting were observed. For all markers, the density of positively stained immune cells was higher in the invasive front than in the tumour centre, pointing to an accumulation of immune cells near the tumour boundaries. When we looked at the proportion of immunohistochemically positive immune cells, we observed enrichment of CD5 (P=0.025) and CD20 (P<0.001) at the periphery, and FOXP3 enrichment in the centre (P<0.001). The QuPath scoring protocol can adequately identify positively stained immune cells in breast tumours, and allows the evaluation of differences in immune cell proportion and density within different tumour regions. The entire tumour section can be quantitatively assessed quite rapidly, which is a major advantage over manual counting.

Using digital-image analysis of tumor-infiltrating lymphocytes to predict survival outcomes in primary melanoma.

10066 Background: Inclusion of tumor-infiltrating lymphocytes (TIL) into AJCC staging criteria has been proposed due to evidence suggesting its prognostic significance. However, subjective inter-observer discordance prevents adoption of semi-quantitative TIL grading (e.g. absent, non-brisk, brisk) into clinical practice. We hypothesize that digital-image analysis (DIA) of TIL can provide a standardized, quantitative scoring system that more accurately predicts survival compared to currently used semi-quantitative grading methods. Methods: Clinical data and tumor specimens were analyzed from prospectively enrolled primary melanoma patients in the New York University Interdisciplinary Melanoma Cooperative Group with median follow-up of 5 years. H&amp;E-stained slides were digitized using an Aperio ScanScope at 20X magnification. QuPath software was used for automated TIL quantification. Cox regression analysis was used to assess the improved prognostic value of TIL on recurrence-free (RFS) and overall survival (OS). Patients were separated into high- and low-TIL groups using a score threshold determined by the Youden Index. Results: 453 patients (18% stage I, 42% stage II, 40% stage III) were scored using automated TIL assessment and scores were significantly correlated with better RFS and OS per 10% increase in TIL (stage adjusted hazard ratio [aHR] = 0.92 [0.84-1.00] for RFS and aHR = 0.90 [0.83-0.99] for OS). A model combining TIL score with stage increased prognostic ability for both RFS (0.68 to 0.70, P = 0.02) and OS (0.62 to 0.64, P = 0.01), as assessed by concordance indices (C-index). Kaplan-Meier curves of high- ( &gt; 16.6%) versus low-TIL (≤16.6%) patients showed clear separation in RFS and OS (median RFS = 155 vs 48 months, P &lt; 0.001; median OS = 155 vs 89 months, P = 0.002). For comparison, a subset of the cohort (n = 250) was semi-quantitatively graded (absent, non-brisk, brisk) by an attending melanoma pathologist; however, this did not significantly differentiate RFS between groups (P &gt; 0.05). Conclusions: A standardized, quantitative TIL scoring system significantly improved prediction of RFS and OS in primary melanoma patients compared with semi-quantitative TIL grading. Incorporation of quantitative TIL scoring into prognostic algorithms, such as AJCC criteria, should be considered.

Internal Jugular Vein Compression Collar Mitigates Histopathological Alterations after Closed Head Rotational Head Impact in Swine: A Pilot Study

Recently, there has been increased concern about microstructural brain changes after head trauma. Clinical studies have investigated a neck collar that applies gentle bilateral jugular vein compression, designed to increase intracranial blood volume and brain stiffness during head trauma, which neuroimaging has shown to result in a reduction in brain microstructural alterations after a season of American football and soccer. Here, we utilized a swine model of mild traumatic brain injury to investigate the effects of internal jugular vein (IJV) compression on histopathological outcomes after injury. Animals were randomized to collar treatment (n = 8) or non-collar treatment (n = 6), anesthetized and suspended such that the head was supported by breakable tape. A custom-built device was used to impact the head, thus allowing the head to break the tape and rotate along the sagittal plane. Accelerometer data were collected for each group. Sham injured animals (n = 2) were exposed to anesthesia only. Following single head trauma, animals were euthanized and brains collected for histology. Whole slide immunohistochemistry was analyzed using Qupath software. There was no difference in linear or rotational acceleration between injured collar and non-collar animals (p > 0.05). Injured animals demonstrated higher levels of the phosphorylated tau epitope AT8 (p < 0.05) and the inflammatory microglial marker IBA1 (p < 0.05) across the entire brain, but the effect of injury was markedly reduced by collar treatment (p < 0.05) The current results indicate that internal jugular venous compression protects against histopathological alterations related to closed head trauma exposure.

Abstract A46: CD73 inhibition enhances the effect of anti-PD-1 therapy on KRAS-mutated pancreatic cancer model

Abstract Background: CD73 catalyzes the extracellular generation of adenosine (ADO) from adenosine monophosphate (AMP). High levels of ADO, found in the tumor microenvironment (TME), have been shown to suppress immune responses and curtail T-cell activation in the presence of anti-PD-1/PD-L1 blocking antibodies. Oncogenic drivers resulting from activating mutations, such as KRAS, BRAF, and EGFR mutations, are commonly treated with tyrosine kinase inhibitors that result in robust but nondurable responses. We show here that KRAS mutations, of which 60% were derived from pancreatic adenocarcinoma (PDAC) samples, significantly upregulated CD73 expression and resulted in worsening prognosis. In a murine model of pancreatic cancer bearing KRASG12C mutation, coadministration of CD73 inhibitor with anti-PD-1 in established tumors resulted in significant tumor growth retardation, comparable to KRAS inhibitor alone. These data support the rationale for the clinical development of CD73 inhibitors in pancreatic cancer. Methods: Linear models were used to evaluate the ability of 299 pan-cancer consensus oncogenic drivers to predict CD73 expression independent of tumor type in the TCGA dataset. Immunohistochemistry (IHC) on formalin-fixed, paraffin-embedded (FFPE) samples was performed using CD73 (Cell Signaling, D7F9A) primary antibody and detected using anti-rabbit HRP. CD73 expressing cells were detected by DAB chromogen and quantified using QuPath Software. CD73 activity in fresh frozen tissues were detected using the Wachstein-Meisel method. C57BL/6J mice bearing established KP4662-G12C tumors (&amp;gt;150 mm3) were treated with A0001421 (CD73i), 30 mg/kg; once a day; s.c. and anti-PD1 (Clone RMP 1-14): 10 mg/kg; twice a week; i.p. Mice were treated for two weeks starting from day 10 post-tumor. Treatment efficacy was monitored using micro-computed tomography (Genisys PET/CT G8 (Sofie Biosciences). All small-molecule inhibitors were synthesized by Arcus Biosciences, Inc. Results: Out of the 299 oncogenes, alterations in KRAS, BRAF, and EGFR were the top 3 oncogenes upregulating CD73 expression. Increase in CD73 expression in KRAS-mutated tumors was further confirmed by IHC. Data extrapolated from Koyama et al. suggest that PD-1 nonresponsive mice express higher levels of adenosine pathway genes, including CD73 and CD39. Coadministration of CD73 inhibitor with anti-PD-1 was superior to anti-PD-1 alone in limiting tumor growth of an aggressive model of pancreatic cancer. Global changes in immune contexture and TME were also observed, consistent with the immune modulatory effects of inhibiting CD73 activity. Conclusion: CD73 is a highly efficient ecto-enzyme that catalyzes the hydrolysis of AMP to immune-suppressive ADO and is associated with worsening prognosis across multiple malignancies. AB680, a potent and selective CD73 inhibitor with excellent safely and pharmacokinetic profile, is currently undergoing clinical evaluation in 1L metastatic PDAC in combination with AB122 and chemotherapy. Citation Format: Thuc M. Le, Akshata Udyavar, Woosuk Kim, Arthur E. Cho, Luyi Li, Daniel DiRenzo, Brandon Rosen, Matthew J. Walters, Joanne B.L. Tan, Caius G. Radu. CD73 inhibition enhances the effect of anti-PD-1 therapy on KRAS-mutated pancreatic cancer model [abstract]. In: Proceedings of the AACR Special Conference on Tumor Immunology and Immunotherapy; 2019 Nov 17-20; Boston, MA. Philadelphia (PA): AACR; Cancer Immunol Res 2020;8(3 Suppl):Abstract nr A46.

QuPath Automated Analysis of Optic Nerve Degeneration in Brown Norway Rats.

PurposeA novel application of QuPath open-source digital analysis software is used to provide in-depth morphological analysis of progressive optic nerve (ON) degeneration in rats.MethodsQuPath software was adapted to assess axon and gliotic morphology in toluidine blue-stained, Brown Norway rat ON light micrographs. QuPath axon numbers, density, size distributions, and gliotic areas were obtained from test images and ON cross-sections separated by damage grade. QuPath results were compared with manual counting, AxonJ, and electron microscopy axon estimates.ResultsQuPath-derived axon number, density, and diameter decreased with increasing ON damage. Axon density negatively correlated with gliotic areas in test images (R2 = 0.759; P < 0.0001; N = 40) and in ON cross-sections (R2 = 0.803; P < 0.0004; N = 10). Although axon losses occurred across most axon diameters, large axons were more susceptible to degeneration. The exception was swollen axons > 2 µm, which increased in moderately but not severely damaged images. QuPath axon counts correlated strongly with manual counts of test images (R2 = 0.956; P < 0.0001). QuPath outperformed AxonJ on test images and total ON axon counts. Compared to electron microscopy analysis, QuPath undercounted ON axons; however, correlation between the methods was robust (R2 = 0.797; P < 0.001; N = 10).ConclusionsQuPath analysis reliably identified axon loss, axon morphology changes, and gliotic expansion that occurred in degenerating ONs.Translational RelevanceQuPath is a valuable tool for rapid, automated, analysis of healthy and degenerating ONs. Reproducible preclinical studies for new glaucoma treatments depend on unbiased in-depth analysis of ON pathology. This was provided by the QuPath approach.

Abstract P3-08-12: An open source, automated tumor infiltrating lymphocyte algorithm for prognosis in triple-negative breast cancer

Abstract Background: The presence of high TILs (tumor infiltrating lymphocytes) have been shown to be predictive of response to chemotherapy and is also a prognostic factor associated with better outcome in breast cancer, especially in early stage triple-negative (TNBC) and HER2-positive breast cancers. TIL assessment, while now more standardized due to the efforts of Salgado and the International TIL Working Group (https://www.tilsinbreastcancer.org/), are still a subjective test with variability in evaluation that has prevented broad adoption. Given the advances in application of artificial intelligence to pathology images, we believe the next step for TILs is to make them automated and objective and to identify a standardized and meaning TIL cut-point. The aim of this study is to build an open source, H&amp;E image-based automated TIL assessment algorithm for breast cancer that allows global standardization of TILs for prognostic value. Materials and methods: Using QuPath open source software, we first built a neural network classifier for image-based, automated assessment of TILs. It distinguishes tumor cells, lymphocytes, stromal cells and other cells on hematoxylin-eosin (H&amp;E) stained sections. We then defined “eTILs%” calculated as follows for the percentage of machine defined TILs: (TILs/TILs+Tumor cells) *100. A retrospective collection of 63 TNBC cases was used as the training set (Set A) and then tested for cell classifier accuracy and the optimal “eTILs%” cut point. The validation sets were a retrospective collection of 354 TNBC patients comprised of three independent validation subsets (Set B; N=87, Set C; N=183, and Set D; N=83) in both tissue microarray (TMA) and whole tissue section (WTS) format. Results: Using an optimal cut point (30%) derived from TNBC cohort training set A, patients with high eTILs% displayed an overall survival benefit (HR 0.4, p=0.0150). This algorithm was then applied in other three TNBC validation sets (Set B: HR=0.42, p=0.0033; Set C: HR=0.42, p=0.0127; Set D in TMA format: HR=0.39, p=0.0089). For Set D, we also tested WTS format which showed HR=0.23, (p=0.0155). The validation sets were combined to assess independence from clinical status in a multi-variable analysis where eTILs% was independently associated with improved overall survival (HR=0.35, p&amp;lt;0.0001). Conclusion: We have constructed a single institutional algorithm built from open source QuPath software that is a robust and independent prognostic factor. With further validation in tissues from other institutions and larger cohorts, this algorithm could be potentially useful as a broadly applicable, standardized and globally available method for assessment of TILs in triple negative breast cancer patients. Citation Format: Yalai Bai, Balazs Acs, Jon Zugazagoitia, Sandra Martinez-Morilla, Fahad Shabbir Ahmed, David L. Rimm. An open source, automated tumor infiltrating lymphocyte algorithm for prognosis in triple-negative breast cancer [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 P3-08-12.

Automated Quantification of Immunohistochemical Staining of Large Animal Brain Tissue Using QuPath Software

Large scale unbiased quantification of immunohistochemistry (IHC) is time consuming, expensive, and/or limited in scope. Heterogeneous tissue types such as brain tissue have presented a further challenge to the development of automated analysis, as differing cellular morphologies result in either limited applicability or require large amounts of training tissue for machine-learning methods. Here we present the use of QuPath, a free and open source software, to quantify whole-brain sections stained with the immunohistochemical markers IBA1 and AT8, for microglia and phosphorylated tau respectively. The pixel-based method of analysis herein allows for statistical comparison of global protein expression between brains and generates heat-maps of stain intensity, visualizing stain signal across whole sections and permitting more specific investigation of regions of interest. This method is fast, automated, unbiased, and easily replicable. We compared swine brains that had undergone a closed head traumatic brain injury with brains of sham animals, and found a global increase in both microglial signal expression and phosphorylated tau. We discuss the IHC methods necessary to utilize this analysis and provide detailed instruction on the use of QuPath in the pixel-based analysis of whole-slide images.