Analysis of Immune Intratumor Heterogeneity Highlights Immunoregulatory and Coinhibitory Lymphocytes as Hallmarks of Recurrence in Stage I Non-Small Cell Lung Cancer.

Our understanding of the molecular mechanisms underlying postsurgical recurrence of non-small cell lung cancer (NSCLC) is rudimentary. Molecular and T cell repertoire intratumor heterogeneity (ITH) have been reported to be associated with postsurgical relapse; however, how ITH at the cellular level impacts survival is largely unknown. Here we report the analysis of 2880 multispectral images representing 14.2% to 27% of tumor areas from 33 patients with stage I NSCLC, including 17 cases (relapsed within 3 years after surgery) and 16 controls (without recurrence ≥5 years after surgery) using multiplex immunofluorescence. Spatial analysis was conducted to quantify the minimum distance between different cell types and immune cell infiltration around malignant cells. Immune ITH was defined as the variance of immune cells from 3 intratumor regions. We found that tumors from patients having relapsed display different immune biology compared with nonrecurrent tumors, with a higher percentage of tumor cells and macrophages expressing PD-L1 (P =.031 and P =.024, respectively), along with an increase in regulatory T cells (Treg) (P =.018), antigen-experienced T cells (P =.025), and effector-memory T cells (P =.041). Spatial analysis revealed that a higher level of infiltration of PD-L1+ macrophages (CD68+PD-L1+) or antigen-experienced cytotoxic T cells (CD3+CD8+PD-1+) in the tumor was associated with poor overall survival (P =.021 and P =.006, respectively). A higher degree of Treg ITH was associated with inferior recurrence-free survival regardless of tumor mutational burden (P =.022), neoantigen burden (P =.021), genomic ITH (P =.012) and T cell repertoire ITH (P =.001). Using multiregion multiplex immunofluorescence, we characterized ITH at the immune cell level along with whole exome and T cell repertoire sequencing from the same tumor regions. This approach highlights the role of immunoregulatory and coinhibitory signals as well as their spatial distribution and ITH that define the hallmarks of tumor relapse of stage I NSCLC.

Heterogeneity in intrahepatic macrophage populations and druggable target expression in patients with steatotic liver disease-related fibrosis.

Background & Aims: Clinical trials for reducing fibrosis in steatotic liver disease (SLD) have targeted macrophages with variable results. We evaluated intrahepatic macrophages in patients with SLD to determine if activity scores or fibrosis stages influenced phenotypes and expression of druggable targets, such as CCR2 and galectin-3. Methods: Liver biopsies from controls or patients with minimal or advanced fibrosis were subject to gene expression analysis using nCounter to determine differences in macrophage-related genes (n = 30). To investigate variability among individual patients, we compared additional biopsies by staining them with multiplex antibody panels (CD68/CD14/CD16/CD163/Mac387 or CD163/CCR2/galectin-3/Mac387) followed by spectral imaging and spatial analysis. Algorithms that utilize deep learning/artificial intelligence were applied to create cell cluster plots, phenotype profile maps, and to determine levels of protein expression (n = 34). Results: Several genes known to be pro-fibrotic (e.g. CD206, TREM2, CD163, and ARG1) showed either no significant differences or significantly decreased with advanced fibrosis. Although marked variability in gene expression was observed in individual patients with cirrhosis, several druggable targets and their ligands (e.g. CCR2, CCR5, CCL2, CCL5, and LGALS3) were significantly increased when compared to patients with minimal fibrosis. Antibody panels identified populations that were significantly increased (e.g. Mac387+), decreased (e.g. CD14+), or enriched (e.g. interactions of Mac387) in patients that had progression of disease or advanced fibrosis. Despite heterogeneity in patients with SLD, several macrophage phenotypes and druggable targets showed a positive correlation with increasing NAFLD activity scores and fibrosis stages. Conclusions: Patients with SLD have markedly varied macrophage- and druggable target-related gene and protein expression in their livers. Several patients had relatively high expression, while others were like controls. Overall, patients with more advanced disease had significantly higher expression of CCR2 and galectin-3 at both the gene and protein levels. Impact and implications: Appreciating individual differences within the hepatic microenvironment of patients with SLD may be paramount to developing effective treatments. These results may explain why such a small percentage of patients have responded to macrophage-targeting therapies and provide additional support for precision medicine-guided treatment of chronic liver diseases.

SPARTIN: a Bayesian method for the quantification and characterization of cell type interactions in spatial pathology data.

Introduction: The acquisition of high-resolution digital pathology imaging data has sparked the development of methods to extract context-specific features from such complex data. In the context of cancer, this has led to increased exploration of the tumor microenvironment with respect to the presence and spatial composition of immune cells. Spatial statistical modeling of the immune microenvironment may yield insights into the role played by the immune system in the natural development of cancer as well as downstream therapeutic interventions. Methods: In this paper, we present SPatial Analysis of paRtitioned Tumor-Immune imagiNg (SPARTIN), a Bayesian method for the spatial quantification of immune cell infiltration from pathology images. SPARTIN uses Bayesian point processes to characterize a novel measure of local tumor-immune cell interaction, Cell Type Interaction Probability (CTIP). CTIP allows rigorous incorporation of uncertainty and is highly interpretable, both within and across biopsies, and can be used to assess associations with genomic and clinical features. Results: Through simulations, we show SPARTIN can accurately distinguish various patterns of cellular interactions as compared to existing methods. Using SPARTIN, we characterized the local spatial immune cell infiltration within and across 335 melanoma biopsies and evaluated their association with genomic, phenotypic, and clinical outcomes. We found that CTIP was significantly (negatively) associated with deconvolved immune cell prevalence scores including CD8+ T-Cells and Natural Killer cells. Furthermore, average CTIP scores differed significantly across previously established transcriptomic classes and significantly associated with survival outcomes. Discussion: SPARTIN provides a general framework for investigating spatial cellular interactions in high-resolution digital histopathology imaging data and its associations with patient level characteristics. The results of our analysis have potential implications relevant to both treatment and prognosis in the context of Skin Cutaneous Melanoma. The R-package for SPARTIN is available at https://github.com/bayesrx/SPARTIN along with a visualization tool for the images and results at: https://nateosher.github.io/SPARTIN.

Low-parameter supervised learning models can discriminate pseudoprogression and true progression in non-perfusion-based MRI.

Discrimination of pseudoprogression and true progression is one challenge to the treatment of malignant gliomas. Although some techniques such as circulating tumor DNA (ctDNA) and perfusion-weighted imaging (PWI) demonstrate promise in distinguishing PsP from TP, we investigate robust and replicable alternatives to distinguish the two entities based on more widely-available media. In this study, we use low-parametric supervised learning techniques based on geographically-weighted regression (GWR) to investigate the utility of both conventional MRI sequences as well as a diffusion-weighted sequence (apparent diffusion coefficient or ADC) in the discrimination of PsP v TP. GWR applied to MRI modality pairs is a unique approach for small sample sizes and is a novel approach in this arena. From our analysis, all modality pairs involving ADC maps, and those involving post-contrast T1 regressed onto T2 showed potential promise. This work on ADC data adds to a growing body of research suggesting the predictive benefits of ADC, and suggests further research on the relationships between post-contrast T1 and T2.Clinical relevance- Few studies have investigated predictive potential of conventional MRI and ADC to detect PsP. Our study adds to the growing research on the topic and presents a new perspective to research by exploiting the utility of ADC in PsP v TP distinction. In addition, our GWR methodology for low-parametric supervised computer vision models demonstrates a unique approach for image processing of small sample sizes.

Patients with fibrosis from non-alcoholic steatohepatitis have heterogeneous intrahepatic macrophages and therapeutic targets.

Background and Aims: In clinical trials for reducing fibrosis in NASH patients, therapeutics that target macrophages have had variable results. We evaluated intrahepatic macrophages in patients with non-alcoholic steatohepatitis to determine if fibrosis influenced phenotypes and expression of CCR2 and Galectin-3. Approach & Results: We used nCounter to analyze liver biopsies from well-matched patients with minimal (n=12) or advanced (n=12) fibrosis to determine which macrophage-related genes would be significantly different. Known therapy targets (e.g., CCR2 and Galectin-3) were significantly increased in patients with cirrhosis.However, several genes (e.g., CD68, CD16, and CD14) did not show significant differences, and CD163, a marker of pro-fibrotic macrophages was significantly decreased with cirrhosis. Next, we analyzed patients with minimal (n=6) or advanced fibrosis (n=5) using approaches that preserved hepatic architecture by multiplex-staining with anti-CD68, Mac387, CD163, CD14, and CD16. Spectral data were analyzed using deep learning/artificial intelligence to determine percentages and spatial relationships. This approach showed patients with advanced fibrosis had increased CD68+, CD16+, Mac387+, CD163+, and CD16+CD163+ populations. Interaction of CD68+ and Mac387+ populations was significantly increased in patients with cirrhosis and enrichment of these same phenotypes in individuals with minimal fibrosis correlated with poor outcomes. Evaluation of a final set of patients (n=4) also showed heterogenous expression of CD163, CCR2, Galectin-3, and Mac387, and significant differences were not dependent on fibrosis stage or NAFLD activity. Conclusions: Approaches that leave hepatic architecture intact, like multispectral imaging, may be paramount to developing effective treatments for NASH. In addition, understanding individual differences in patients may be required for optimal responses to macrophage-targeting therapies.

Unique characteristics of the tumor immune microenvironment in young patients with metastatic colorectal cancer.

Introduction: Metastatic colorectal cancer (mCRC) remains a common and highly morbid disease, with a recent increase in incidence in patients younger than 50 years. There is an acute need to better understand differences in tumor biology, molecular characteristics, and other age-related differences in the tumor microenvironment (TME). Methods: 111 patients undergoing curative-intent resection of colorectal liver metastases were stratified by age into those <50 years or >65 years old, and tumors were subjected to multiplex fluorescent immunohistochemistry (mfIHC) to characterize immune infiltration and cellular engagement. Results: There was no difference in infiltration or proportion of immune cells based upon age, but the younger cohort had a higher proportion of programmed death-ligand 1 (PD-L1)+ expressing antigen presenting cells (APCs) and demonstrated decreased intercellular distance and increased cellular engagement between tumor cells (TCs) and cytotoxic T lymphocytes (CTLs), and between TCs and APCs. These trends were independent of microsatellite instability in tumors. Discussion: Age-related differences in PD-L1 expression and cellular engagement in the tumor microenvironment of patients with mCRC, findings which were unrelated to microsatellite status, suggest a more active immune microenvironment in younger patients that may offer an opportunity for therapeutic intervention with immune based therapy.

GaWRDenMap: a quantitative framework to study the local variation in cell-cell interactions in pancreatic disease subtypes.

Spatial pattern modelling concepts are being increasingly used in capturing disease heterogeneity. Quantification of heterogeneity in the tumor microenvironment is extremely important in pancreatic ductal adenocarcinoma (PDAC), which has been shown to co-occur with other pancreatic diseases and neoplasms with certain attributes that make visual discrimination difficult. In this paper, we propose the GaWRDenMap framework, that utilizes the concepts of geographically weighted regression (GWR) and a density function-based classification model, and apply it to a cohort of multiplex immunofluorescence images from patients belonging to six different pancreatic diseases. We used an internal cohort of 228 patients comprised of 34 Chronic Pancreatitis (CP), 71 PDAC, 70 intraductal papillary mucinous neoplasm (IPMN), 16 mucinous cystic neoplasm (MCN), 29 pancreatic intraductal neoplasia (PanIN) and 8 IPMN-associated PDAC patients. We utilized GWR to model the relationship between epithelial cells and immune cells on a spatial grid. The GWR model estimates were used to generate density signatures which were used in subsequent pairwise classification models to distinguish between any two pairs of disease groups. Image-level, as well as subject-level analysis, were performed. When applied to this dataset, our classification model showed significant discrimination ability in multiple pairwise comparisons, in comparison to commonly used abundance-based metrics, like the Morisita-Horn index. The model was able to best discriminate between CP and PDAC at both the subject- and image-levels. It was also able to reasonably discriminate between PDAC and IPMN. These results point to a potential difference in the spatial arrangement of epithelial and immune cells between CP, PDAC and IPMN, that could be of high diagnostic significance. Further validation on a more comprehensive dataset would be warranted.

Towards the characterization of the tumor microenvironment through dictionary learning-based interpretable classification of multiplexed immunofluorescence images.

Objective.Histology image analysis is a crucial diagnostic step in staging and treatment planning, especially for cancerous lesions. With the increasing adoption of computational methods for image analysis, significant strides are being made to improve the performance metrics of image segmentation and classification frameworks. However, many developed frameworks effectively function as black boxes, granting minimal context to the decision-making process. Thus, there is a need to develop methods that offer reasonable discriminatory power and a biologically-informed intuition to the decision-making process.Approach.In this study, we utilized and modified a discriminative feature-based dictionary learning (DFDL) paradigm to generate a classification framework that allows for discrimination between two distinct clinical histologies. This framework allows us (i) to discriminate between 2 clinically distinct diseases or histologies and (ii) provides interpretable group-specific representative dictionary image patches, or 'atoms', generated during classifier training. This implementation is performed on multiplexed immunofluorescence images from two separate patient cohorts- a pancreatic cohort consisting of cancerous and non-cancerous tissues and a metastatic non-small cell lung cancer (mNSCLC) cohort of responders and non-responders to an immunotherapeutic treatment regimen. The analysis was done at both the image-level and subject-level. Five cell types were selected, namely, epithelial cells, cytotoxic lymphocytes, antigen presenting cells, HelperT cells, and T-regulatory cells, as our phenotypes of interest.Results.We showed that DFDL had significant discriminant capabilities for both the pancreatic pathologies cohort (subject-level AUC-0.8878) and the mNSCLC immunotherapy response cohort (subject-level AUC-0.7221). The secondary analysis also showed that more than 50% of the obtained dictionary atoms from the classifier contained biologically relevant information.Significance.Our method shows that the generated dictionary features can help distinguish patients presenting two different histologies with strong sensitivity and specificity metrics. These features allow for an additional layer of model interpretability, a highly desirable element in clinical applications for identifying novel biological phenomena.

Cellular engagement and interaction in the tumor microenvironment predict non-response to PD-1/PD-L1 inhibitors in metastatic non-small cell lung cancer.

Immune checkpoint inhibitors (ICI) with anti-PD-1/PD-L1 agents have improved the survival of patients with metastatic non-small cell lung cancer (mNSCLC). Tumor PD-L1 expression is an imperfect biomarker as it does not capture the complex interactions between constituents of the tumor microenvironment (TME). Using multiplex fluorescent immunohistochemistry (mfIHC), we modeled the TME to study the influence of cellular distribution and engagement on response to ICI in mNSCLC. We performed mfIHC on pretreatment tissue from patients with mNSCLC who received ICI. We used primary antibodies against CD3, CD8, CD163, PD-L1, pancytokeratin, and FOXP3; simple and complex phenotyping as well as spatial analyses was performed. We analyzed 68 distinct samples from 52 patients with mNSCLC. Patients were 39-79 years old (median 67); 44% were male and 75% had adenocarcinoma histology. The most used ICI was atezolizumab (48%). The percentage of PD-L1 positive epithelial tumor cells (EC), degree of cytotoxic T lymphocyte (CTL) engagement with EC, and degree of CTL engagement with helper T lymphocytes (HTL) were significantly lower in non-responders versus responders (p = 0.0163, p = 0.0026 and p = 0.0006, respectively). The combination of these 3 characteristics generated the best sensitivity and specificity to predict non-response to ICI and was also associated with shortened overall survival (p = 0.0271). The combination of low CTL engagement with EC and HTL along with low expression of EC PD-L1 represents a state of impaired endogenous immune reactivity. Together, they more precisely identified non-responders to ICI compared to PD-L1 alone and illustrate the importance of cellular interactions in the TME.

MHC Class II Expression Influences the Composition and Distribution of Immune Cells in the Metastatic Colorectal Cancer Microenvironment.

Despite advances in therapy over the past decades, metastatic colorectal cancer (mCRC) remains a highly morbid disease. While the impact of MHC-I on immune infiltration in mCRC has been well studied, data on the consequences of MHC-II loss are lacking. Multiplex fluorescent immunohistochemistry (mfIHC) was performed on 149 patients undergoing curative intent resection for mCRC and stratified into high and low human leukocyte antigen isotype DR (HLA-DR) expressing tumors. Intratumoral HLA-DR expression was found in stromal bands, and its expression level was associated with different infiltrating immune cell makeup and distribution. Low HLA-DR expression was associated with increased intercellular distances and decreased population mixing of T helper cells and antigen-presenting cells (APC), suggestive of decreased interactions. This was associated with less co-localization of tumor cells and cytotoxic T lymphocytes (CTLs), which tended to be in a less activated state as determined by Ki67 and granzyme B expression. These findings suggest that low HLA-DR in the tumor microenvironment of mCRC may reflect a state of poor helper T-cell interactions with APCs and CTL-mediated anti-tumor activity. Efforts to restore/enhance MHC-II presentation may be a useful strategy to enhance checkpoint inhibition therapy in the future.

Central nervous system immune interactome is a function of cancer lineage, tumor microenvironment, and STAT3 expression.

BACKGROUNDImmune cell profiling of primary and metastatic CNS tumors has been focused on the tumor, not the tumor microenvironment (TME), or has been analyzed via biopsies.METHODSEn bloc resections of gliomas (n = 10) and lung metastases (n = 10) were analyzed via tissue segmentation and high-dimension Opal 7-color multiplex imaging. Single-cell RNA analyses were used to infer immune cell functionality.RESULTSWithin gliomas, T cells were localized in the infiltrating edge and perivascular space of tumors, while residing mostly in the stroma of metastatic tumors. CD163+ macrophages were evident throughout the TME of metastatic tumors, whereas in gliomas, CD68+, CD11c+CD68+, and CD11c+CD68+CD163+ cell subtypes were commonly observed. In lung metastases, T cells interacted with CD163+ macrophages as dyads and clusters at the brain-tumor interface and within the tumor itself and as clusters within the necrotic core. In contrast, gliomas typically lacked dyad and cluster interactions, except for T cell CD68+ cell dyads within the tumor. Analysis of transcriptomic data in glioblastomas revealed that innate immune cells expressed both proinflammatory and immunosuppressive gene signatures.CONCLUSIONOur results show that immunosuppressive macrophages are abundant within the TME and that the immune cell interactome between cancer lineages is distinct. Further, these data provide information for evaluating the role of different immune cell populations in brain tumor growth and therapeutic responses.FUNDINGThis study was supported by the NIH (NS120547), a Developmental research project award (P50CA221747), ReMission Alliance, institutional funding from Northwestern University and the Lurie Comprehensive Cancer Center, and gifts from the Mosky family and Perry McKay. Performed in the Flow Cytometry & Cellular Imaging Core Facility at MD Anderson Cancer Center, this study received support in part from the NIH (CA016672) and the National Cancer Institute (NCI) Research Specialist award 1 (R50 CA243707). Additional support was provided by CCSG Bioinformatics Shared Resource 5 (P30 CA046592), a gift from Agilent Technologies, a Research Scholar Grant from the American Cancer Society (RSG-16-005-01), a Precision Health Investigator Award from University of Michigan (U-M) Precision Health, the NCI (R37-CA214955), startup institutional research funds from U-M, and a Biomedical Informatics & Data Science Training Grant (T32GM141746).

CGAT: Cell Graph ATtention Network for Grading of Pancreatic Disease Histology Images.

Early detection of Pancreatic Ductal Adenocarcinoma (PDAC), one of the most aggressive malignancies of the pancreas, is crucial to avoid metastatic spread to other body regions. Detection of pancreatic cancer is typically carried out by assessing the distribution and arrangement of tumor and immune cells in histology images. This is further complicated due to morphological similarities with chronic pancreatitis (CP), and the co-occurrence of precursor lesions in the same tissue. Most of the current automated methods for grading pancreatic cancers rely on extensive feature engineering involving accurate identification of cell features or utilising single number spatially informed indices for grading purposes. Moreover, sophisticated methods involving black-box approaches, such as neural networks, do not offer insights into the model's ability to accurately identify the correct disease grade. In this paper, we develop a novel cell-graph based Cell-Graph Attention (CGAT) network for the precise classification of pancreatic cancer and its precursors from multiplexed immunofluorescence histology images into the six different types of pancreatic diseases. The issue of class imbalance is addressed through bootstrapping multiple CGAT-nets, while the self-attention mechanism facilitates visualization of cell-cell features that are likely responsible for the predictive capabilities of the model. It is also shown that the model significantly outperforms the decision tree classifiers built using spatially informed metric, such as the Morisita-Horn (MH) indices.

Multispectral Imaging Enables Characterization of Intrahepatic Macrophages in Patients With Chronic Liver Disease.

Intrahepatic macrophages influence the composition of the microenvironment, host immune response to liver injury, and development of fibrosis. Compared with stellate cells, the role of macrophages in the development of fibrosis remains unclear. Multispectral imaging allows detection of multiple markers in situ in human formalin-fixed, paraffin-embedded tissue. This cutting-edge technology is ideal for analyzing human liver tissues, as it allows spectral unmixing of fluorophore signals, subtraction of auto-fluorescence, and preservation of hepatic architecture. We analyzed five different antibodies commonly observed on macrophage populations (CD68, MAC387, CD163, CD14, and CD16). After optimization of the monoplex stains and development of a Spectral Library, we combined all of the antibodies into a multiplex protocol and used them to stain biopsies collected from representative patients with chronic liver diseases, including chronic hepatitis C, nonalcoholic steatohepatitis, and autoimmune hepatitis. Various imaging modalities were tested, including cell phenotyping, tissue segmentation, t-distributed stochastic neighbor embedding plots, and phenotype matrices that facilitated comparison and visualization of the identified macrophage and other cellular profiles. We then tested the feasibility of this platform to analyze numerous regions of interest from liver biopsies with multiple patients per group, using batch analysis algorithms. Five populations showed significant differences between patients positive for hepatitis C virus with advanced fibrosis when compared with controls. Three of these were significantly increased in patients with advanced fibrosis when compared to controls, and these included CD163+CD16+, CD68+, and CD68+MAC387+. Conclusion: Spectral imaging microscopy is a powerful tool that enables in situ analysis of macrophages and other cells in human liver biopsies and may lead to more personalized therapeutic approaches in the future.

Tumor core biopsies adequately represent immune microenvironment of high-grade serous carcinoma.

The prognostic and therapeutic value of the tumor microenvironment (TME) in various cancer types is of major interest. Characterization of the TME often relies on a small representative tissue sample. However, the adequacy of such a sample for assessing components of the TME is not yet known. Here, we used immunohistochemical (IHC) staining and 7-color multiplex staining to evaluate CD8 (cluster of differentiation 8), CD68, PD-L1 (programmed death-ligand 1), CD34, FAP (fibroblast activation protein), and cytokeratin in 220 tissue cores from 26 high-grade serous ovarian cancer samples. Comparisons were drawn between a larger tumor specimen and smaller core biopsies based on number and location (central tumor vs. peripheral tumor) of biopsies. Our analysis found that the correlation between marker-specific cell subsets in larger tumor versus smaller core was stronger with two core biopsies and was not further strengthened with additional biopsies. Moreover, this correlation was consistently strong regardless of whether the biopsy was taken at the center or at the periphery of the original tumor sample. These findings could have a substantial impact on longitudinal assessment for detection of biomarkers in clinical trials.