Letter to the Editor: The suggestion for additional clinical scores in decoding tumor heterogeneity in HCC.

Digital analysis of pathology whole-slide images is fast becoming a game changer in cancer diagnosis and treatment. Specifically, deep learning methods have shown great potential to support pathology analysis, with recent studies identifying molecular traits that were not previously recognized in pathology H&E whole-slide images. Simultaneous to these developments, it is becoming increasingly evident that tumor heterogeneity is an important determinant of cancer prognosis and susceptibility to treatment, and should therefore play a role in the evolving practices of matching treatment protocols to patients. State of the art diagnostic procedures, however, do not provide automated methods for characterizing and/or quantifying tumor heterogeneity, certainly not in a spatial context. Further, existing methods for analyzing pathology whole-slide images from bulk measurements require many training samples and complex pipelines. Our work addresses these two challenges. First, we train deep learning models to spatially resolve bulk mRNA and miRNA expression levels on pathology whole-slide images (WSIs). Our models reach up to 0.95 AUC on held-out test sets from two cancer cohorts using a simple training pipeline and a small number of training samples. Using the inferred gene expression levels, we further develop a method to spatially characterize tumor heterogeneity. Specifically, we produce tumor molecular cartographies and heterogeneity maps of WSIs and formulate a heterogeneity index (HTI) that quantifies the level of heterogeneity within these maps. Applying our methods to breast and lung cancer slides, we show a significant statistical link between heterogeneity and survival. Our methods potentially open a new and accessible approach to investigating tumor heterogeneity and other spatial molecular properties and their link to clinical characteristics, including treatment susceptibility and survival.

Objectives/Goals: Our aim was to identify how the epithelial–mesenchymal transition shields heterogeneous breast tumors against immune attack. Additionally, we endeavored to understand whether our findings were conserved in canine mammary tumors as a translational model for human breast tumors. Methods/Study Population: To understand interactions between quasi-mesenchymal (qM) tumor cells, epithelial (E) tumor cells, and immune cells within heterogeneous breast tumors, we utilized a preclinical mouse model established in our lab. In this system, we can precisely control the proportions of E and qM tumor cells within tumors and study what immune cells infiltrate these tumors in response, using flow cytometry and immunofluorescent staining. Using this model, we have also established cell lines to study E and qM tumor cells in vitro. Finally, we used immunohistochemistry to label immune cells in canine mammary tumors and quantified the presence of these cells in relation to the expression of epithelial and mesenchymal cellular markers. Results/Anticipated Results: We observed that immune suppression within heterogeneous mammary tumors is driven by local, rather than systemic, effects of quasi-mesenchymal (qM) tumor cells. The presence of systemic qM-derived factors does not alter immune cell infiltration nor sensitivity to immunotherapy of epithelial (E) tumors. Furthermore, I found that the local activity of qM-derived factors within heterogeneous tumors induces immune-suppressive changes in surrounding E cells, which protects them against immune attack. Finally, I found that canine mammary tumors with higher proportions of qM tumor cells assemble an immune-suppressive tumor microenvironment, highlighting the translational potential of our findings. Discussion/Significance of Impact: We identified that the epithelial–mesenchymal transition induces immune-suppressive changes in heterogeneous tumors. These findings may reveal novel therapeutic targets for treatment of refractory tumors. Our findings in canine tumors suggest that these mechanisms are conserved across species.

Background: Bladder Cancer is a highly prevalent cancer with staggering mortality of about 20%. Metastatic BLCA (mBLCA) with variant histologies are aggressive and have a poor prognosis. The extent and impact of tumor heterogeneity in these variant subtypes are not fully understood. This work explores the intricate characterization of intra-patient and inter-patient tumor heterogeneity through genomic and transcriptomic analyses, drawing insights from a unique rapid autopsy cohort comprising 20 patients. The resulting revelations provide critical perspectives on heterogeneity and clonal evolution in mBLCA, underscoring the essential role of understanding these dynamics for the development of personalized treatment strategies. Approach: Our research focuses on the genetic and molecular characteristics of these tumors, including the vast landscape of tumor heterogeneity, clonal evolution, genomic alterations, mutational signatures, and structural variants. Within the tumor microenvironment, we identified diverse cell populations with distinct genetic profiles. A detailed examination of mutational signatures sheds light on the specific processes underlying genetic alterations, while the investigation of structural variants elucidates their consequential role in cancer development and progression. Advanced bioinformatics tools are deployed to process and interpret the vast genomic and transcriptomic datasets, facilitating the identification of clonal populations, their genetic features, and the trajectory of clonal evolution. Furthermore, statistical models are employed to discern correlations between mutational signatures, treatment responses, and clinical outcomes. Summary: The study unveils substantial inter-patient heterogeneity and discernible differences in mutational signatures, revealing intriguing correlations with treatment responses. These variations are intricately connected to underlying genomic alterations, copy number alterations, and gene expression patterns, offer a detailed perspective on the molecular landscape of mBLCA. Conclusion: This multi-omics exploration stands as the most extensive study of its kind, leveraging rapid autopsy samples of metastatic variant histology BLCA. The research provides unprecedented resolution of genomic alterations and intra- and inter-patient tumor heterogeneity. The exploration of clonal evolution enhances our appreciation of the temporal dynamics of tumor progression, establishing a robust foundation for future therapeutic interventions precisely targeting evolving tumor clones. As we advance into an era of personalized medicine, these findings pave the way for tailored therapeutic strategies in the challenging landscape of metastatic variant histology Bladder Cancer. Citation Format: Pushpa Itagi, Sonali Arora, Thomas Persse, Michael Yang, Patricia Galipeau, Yixin Lin, Funda Vakar-Lopez, John K. Lee, Petros Grivas, Robert B. Montgomery, Jonathan L. Wright, Hung-Ming Lam, Andrew Hsieh, Gavin Ha. Elucidating tumor evolution and heterogeneity in metastatic bladder cancer from rapid autopsies [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 1610.

Objective. We propose a criterion of biological effectiveness of nonuniform hypoxia-targeted dose distributions in heterogeneous hypoxic tumors based on equivalent uniform aerobic dose (EUAD). We demonstrate the utility of this criterion by applying it to the model problems in radiotherapy for tumors with different levels of oxygen enhancement ratio (OER) and different degrees of dose nonuniformity.Approach. The EUAD is defined as the uniform dose that, under well-oxygenated conditions, produces equal integrated survival of clonogenic cells in radiotherapy for heterogeneous hypoxic tumors with a non-uniform dose distribution. We define the dose nonuniformity effectiveness (DNE) in heterogeneous tumors as the ratio of the EUAD(DN) for a non-uniform distributionDNand the reference EUAD(DU) for the uniform dose distributionDUwith equal integral tumor dose. The DNE concept is illustrated in a radiotherapy model problem for non-small cell lung cancer treated with hypoxia targeted dose escalation. A two-level cell population tumor model was used to consider the hypoxic and oxygenated tumor cells.Results. Theoretical analysis of the DNE shows that the entire region of the OER can be separated in two regions by a threshold OERth: (1) OER > OERthwhere DNE > 1 indicating higher effectiveness of nonuniform dose distributions and (2) OER < OERthwhere DNE < 1 indicating higher effectiveness of uniform dose distributions. Our simulations show that the value of the threshold OERthin radiotherapy with conventional fractionation is significant in the range of about 1.2-1.6 depending on selected radiotherapy parameters. In general, the OERthincreases with reoxygenation rate, relative hypoxic volume and dose escalation factor. The threshold value of OERthis smaller of about 1.1 for hypofractionated radiotherapy.Significance. The analysis of dose distributions using the DNE shows that the uniform dose distributions may improve biological cell killing effect in heterogeneous tumors with intermediate oxygen levels compared to targeted nonuniform dose distribution.

The histologic heterogeneity of childhood supratentorial neuroglial tumors, when quantified, identifies relatively homogeneous subgroups for prognostic purposes and for assignment in clinical trials. Our sample consisted of supratentorial tumors in the Childhood Brain Tumor Consortium. The data consist of reliably identified histologic features and demographic, clinical, operative, and survival information. Factor analysis was used to identify uncorrelated "factors," each represented by a different combination of histologic features in 703 tumors. The defining histologic features were used to label each factor. The heterogeneity of each tumor was summarized using the factor scores for each factor. We compared the survival estimates of subgroups of tumors within common diagnostic classes. We identified five uncorrelated quantitative factors that accounted for much of the histologic variation. Our factor labels were Jumbo, Fibrillary, Proliferative, Spongy, an Oligodendroglial. Two thirds of tumors had high scores on two or more factors, indicating a high degree of heterogeneity among these tumors. Eighty-four percent of supratentorial tumors were accounted for by 19 nonoverlapping relatively homogeneous histologic groups. The five quantitative factors complement standard qualitative taxonomies by summarizing more completely the histologic feature aspects of a tumor than by diagnosis alone and quantify the histologic heterogeneity of individual tumors. Histologically homogeneous groups of tumors are essential for clinical trials, biologic research, and prognostic models.

Oxygen heterogeneity in solid tumors is recognized as a limiting factor for therapeutic efficacy. This heterogeneity arises from the abnormal vascular structure of the tumor, but the precise mechanisms linking abnormal structure and compromised oxygen transport are only partially understood. In this paper, we investigate the role that red blood cell (RBC) transport plays in establishing oxygen heterogeneity in tumor tissue. We focus on heterogeneity driven by network effects, which are challenging to observe experimentally due to the reduced fields of view typically considered. Motivated by our findings of abnormal vascular patterns linked to deviations from current RBC transport theory, we calculated average vessel lengths [Formula: see text] and diameters [Formula: see text] from tumor allografts of three cancer cell lines and observed a substantial reduction in the ratio [Formula: see text] compared to physiological conditions. Mathematical modeling reveals that small values of the ratio λ (i.e., [Formula: see text]) can bias hematocrit distribution in tumor vascular networks and drive heterogeneous oxygenation of tumor tissue. Finally, we show an increase in the value of λ in tumor vascular networks following treatment with the antiangiogenic cancer agent DC101. Based on our findings, we propose λ as an effective way of monitoring the efficacy of antiangiogenic agents and as a proxy measure of perfusion and oxygenation in tumor tissue undergoing antiangiogenic treatment.

The objective of this study is to evaluate the ability of dynamic contrast enhanced computed tomography (DCE-CT) to assess intratumor physiological heterogeneity in tumors with different angiogenic phenotypes. DCE-CT imaging was performed on athymic nude mice bearing xenograft wild type (MCF-7(neo)) and VEGF-transfected (MCF-7(VEGF)) tumors by using a clinical multislice CT, and compared to skeletal muscle. Parametrical maps of tumor physiology--perfusion (F), permeability-surface area (PS), fractional intravascular plasma (f(p)), and interstitial space (f(is))--were obtained by fitting the time-dependent contrast-enhanced curves to a two-compartmental kinetic model for each voxel (0.3 x 0.3 x 0.75 mm(3)). Mean physiological measurements were compared with (positron emission tomography (PET) imaging, and the spatial distribution of tumor vasculature was compared with histology. No statistically significant difference was found in mean physiological values of F, PS, and f(p) in MCF-7(neo) and muscle, while f(is) of MCF-7(neo) was a factor of two higher ( p < 0.04). MCF-7(neo) tumors also showed a radial heterogeneity with significant higher physiological values in periphery than those in middle and core regions ( p < 0.01 for all physiological parameters). MCF-7(VEGF) tumors demonstrated significant increases in all physiological parameters compared with MCF-7(neo) tumors, and a distinct saccular heterogeneous pattern compared with MCF-7(neo) and muscle. Both PET imaging and histological results showed good correlation with the above results for this same mouse model. No statistically significant difference was found in the mean perfusion and intravascular volume measured by PET imaging and DCE-CT. Increases in cross-sectional area of blood vessels ( p < 0.002) were observed in MCF-7(VEGF) tumors than MCF-7(neo), and their spatial distribution correlated well with the spatial distribution of f(p) obtained by DCE-CT. The results of this study demonstrated the feasibility of DCE-CT in quantification of spatial heterogeneity in tumor physiology in small animal models. Monitoring variations in the tumor environment using DCE-CT offers an in vivo tool for the evaluation and optimization of new therapeutic strategies.

Background: Pancreatic cancer is a lethal malignancy characterized by complex intratumoral metabolic reprogramming and intercellular nutrient sharing between cells in the tumor microenvironment (TME) that promote pancreatic cancer progression. However, this crosstalk, as well as regional variation in perfusion and oxygenation, can lead to metabolic heterogeneity that has not been appreciated by metabolomics of whole tumors. Here we quantify amino acids and tricarboxylic acid cycle (TCA) intermediates using a novel methodology that allows us to portray global tumor metabolite heterogeneity in a tumor. Methods: Human PaTu-8902 or murine HY19636 (from female KPC mice p48-Cre+, KRASLSL-G12D/+, Trp53lox/+) pancreatic cancer cell lines were orthotopically injected into pancreata of NCr nude mice (n=3) or C57BL/6 mice (n=2). Mice were euthanized after 3-5 weeks and tumors were harvested. Tumor slices were further sectioned into 1mm x 1mm x 1mm cubes using a custom-made multisectional slicing device and each cube location was recorded. Each cube was extracted using methanol, water, and chloroform with labelled amino acid standards, derivitized, and resolved using gas chromatography-mass spectrometry (DB-35MS column with Agilent 7890B gas chromatograph coupled to a single quadrupole 5977B mass spectrometer). 22 metabolites (15 amino acids, 5 TCA intermediates, lactate, and pyruvate) were identified by unique fragments and retention time compared to known standards. Peaks were picked using OpenChrom and analyzed using MATLAB. Data was analyzed using Graphpad Prism. Principal Component Analysis (PCA) was visualized using Python on a Jupyter notebook. Results: Both orthotopic human and murine pancreatic tumors demonstrated striking levels of intratumoral metabolite heterogeneity. Glycine, glutamine, and proline were the amino acids with the highest coefficient of variance, while leucine, isoleucine, and serine had the lowest coefficient of variance. α-ketoglutarate and succinate were the TCA intermediates with highest coefficient of variance. Lactate had the lowest coefficient of variance among all examined metabolites. Spatial mapping of each metabolite demonstrated distinct regions with varying abundance levels of metabolites. PCA demonstrated 75% of variance was carried by PC1 and 10% carried by PC2. Conclusions: This study reveals insights into the degree of intratumoral heterogeneity present in pancreatic tumors that illustrate the difficulty of in vivo metabolomics analysis and suggest that high-resolution (single cell) metabolomics techniques will be critical to study metabolism in the complex TME. Citation Format: Peter Yu, Robert Banh, Stephen Martis, Douglas Biancur, Albert Sohn, Elaine Lin, Keisuke Yamamoto, Benjamin Greenbaum, Alec Kimmelman. Topographical investigation of metabolites in excised squares (TIMES2): Mapping in vivo of metabolic heterogeneity in pancreatic tumors [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Pancreatic Cancer; 2023 Sep 27-30; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(2 Suppl):Abstract nr C054.

The histologic heterogeneity of childhood supratentorial neuroglial tumors, when quantified, identifies relatively homogeneous subgroups for prognostic purposes and for assignment in clinical trials. Our sample consisted of supratentorial tumors in the Childhood Brain Tumor Consortium. The data consist of reliably identified histologic features and demographic, clinical, operative, and survival information. Factor analysis was used to identify uncorrelated “factors,” each represented by a different combination of histologic features in 703 tumors. The defining histologic features were used to label each factor. The heterogeneity of each tumor was summarized using the factor scores for each factor. We compared the survival estimates of subgroups of tumors within common diagnostic classes. We identified five uncorrelated quantitative factors that accounted for much of the histologic variation. Our factor labels were Jumbo, Fibrillary, Proliferative, Spongy, and Oligodendroglial. Two thirds of tumors had high scores on two or more factors, indicating a high degree of heterogeneity among these tumors. Eighty-four percent of supratentorial tumors were accounted for by 19 nonoverlapping relatively homogeneous histologic groups. The five quantitative factors complement standard qualitative taxonomies by summarizing more completely the histologic feature aspects of a tumor than by diagnosis alone and quantify the histologic heterogeneity of individual tumors. Histologically homogeneous groups of tumors are essential for clinical trials, biologic research, and prognostic models.

Background: Tumor ploidy and heterogeneity demonstrated to be pivotal in predicting immunotherapy response in cutaneous melanoma in several independent cohorts (Liu NatMed2019, Tarantino BioRXiv2022). Their estimation can guide more personalized and rational utilization of these immunotherapies. However, 1) the biology underpinning ploidy and heterogeneity is unknown; and 2) these findings were derived in patients using retrospective research tumor-normal paired whole exome sequencing which is not performed for clinical management. Methods: This study addresses this gap by employing deep learning models to predict these crucial markers using routinely available clinical assays, including H&amp;E images. Attention-based computer vision models enable the identification of key morphological features in H&amp;E slides. Segmentation of tumor tissue to perform automated masking, enhances ploidy inference.Moreover, biologically informed neural networks (P-Net, Elmarakeby Nature2021) uncover transcriptional and genomic features linked to ploidy and tumor heterogeneity. Our models are trained on publicly available data (e.g., Liu et al Nature Medicine 2019; TCGA SKCM) from melanoma cohorts and further validated in independent cohorts to ensure robustness. Results: We developed and validated automated tumor tissue masking, enabling the prediction of Whole Genome Doubling (WGD) from H&amp;E Slides with an AUC &amp;gt; 0.75. Attention-based models identify distinct Tumor Microenvironment (TME) structures predictive of high tumor heterogeneity. P-Net application revealed the Calmodulin pathway, previously associated with regulating proliferation in cancer and targeted with chemotherapy, as intricately linked to high tumor heterogeneity, providing valuable insights into underlying mechanisms. Conclusion: In conclusion, our study strategically harnesses and integrates existing datasets to rigorously test, refine, and validate hypotheses concerning the biological and therapeutic implications of genomic heterogeneity and ploidy. Notably, our predictive models with automated tumor masking demonstrate a remarkable AUC &amp;gt;0.75 for biomarkers traditionally challenging to derive from clinical assays. This breakthrough opens avenues for novel therapeutic strategies targeting genomic heterogeneity and ploidy, providing a transformative potential to enhance patient care and outcomes. Citation Format: Marc Glettig, Giuseppe Tarantino, Tyler Aprati, Haitham Elmarakeby, David Liu. Clinical inference and biological dissection of tumor ploidy and heterogeneity in cutaneous melanoma for immunotherapy response using deep learning [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 910.

Cancer genomes are highly unstable with new genetic variations emerging even within a single metastatic site, making it difficult to track the causal changes that drive metastasis and treatment resistance. Here we present a two-pronged approach for analyzing the full spectrum of genetic variations present in cancer samples. The first approach allows for comprehensive and high-resolution characterization of a broad range of variant types on bulk tumor sample. While the second approach characterizes structural variation at the level of the single cell, allowing for the exploration of tumor clonality and heterogeneity. At the core of our approaches is a microfluidics platform that enables the production of hundreds of thousands to millions of partitioned barcoded reactions. This platform can partition high-molecular weight DNA or single cells. Together, these complementary approaches provide a more complete picture of the genomic variation and clonal structure present in a tumor. For bulk tumor analysis, we obtained high molecular weight DNA from known cancer cell lines and used the 10x Chromium Genome solution to produce Illumina-ready sequencing libraries. In this workflow, partitioning of a limited amount of genomic DNA allows for haplotype-level dilution of genome equivalents, which are then barcoded to create a novel data type referred to as “Linked-Reads”. These molecular barcodes are used to identify reads originating from the same input molecule providing long range information on highly accurate short reads. In addition to highly accurate SNP calling, this further enables identification of complex structural rearrangements in tumor genomes. To gain insight into tumor heterogeneity and clonal structure, we performed single cell DNA sequencing and analysis using 10x Chromium scDNA solution. This platform integrates single cell encapsulation, cell lysis and DNA barcoding into a streamlined workflow. Molecular barcodes are used to associate reads with individual cells allowing for copy number variant (CNV) detection. We applied our scDNA sequencing method to a variety of cancer cell lines revealing their clonal structure, as identified by CNVs, with the capability to identify as few as 10 cells in a sample size of one thousand cells. Using cluster analyses we were able to detect 100kb scale events and by aggregating reads in large clones we were able to confidently identify smaller CNV events down to tens of kilobases. Using whole genome bulk sequencing we identified more than 500 large structural variants in HCC1954, including balanced and unbalanced events. In this presentation, we will integrate this Linked-Read data with single cell genome analysis on the same samples, and compare the genetic variation revealed by these two approaches. We will further explore the power of combining these data types for a more complete picture of tumor genome dynamic Citation Format: Claudia Catalanotti, Sarah Garcia, Kamila Belhocine, Vijay Kumar, Zeljko Dzakula, Andrew Price, Shamoni Maheshwar, Yifeng Yin, Michael Schnall-Levin, Rajiv Bharadwaj, Sara Agee Le, Deanna M. Church. Characterizing genomic variation and tumor heterogeneity in cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 3400.

Despite advances in checkpoint blockade therapy or adoptive transfer of tumor-infiltrating lymphocytes (TILs), achieving cure remains challenging, but success of immunotherapy seems to depend on recognition of tumor-specific antigens (neoantigens). We aimed to determine how many neoantigens need to be recognized by how many different TCRs for eradication of solid tumors. Unmanipulated, naturally expressed (autochthonous) neoantigens were targeted with adoptively transferred TCR-engineered autologous T cells (TCR-therapy). Investigated were effects of TCR-engineered CD8+ T cells, TCR-engineered CD4+ T cells or a combination of both. The targeted tumors were established for at least three weeks and derived from primary autochthonous cancer cell cultures, resembling natural solid tumors and their heterogeneity as found in humans. CD8TCR-therapy was only effective against homogenous tumors and relapse was common in heterogeneous tumors, even when targeting multiple different autochthonous neoantigens. By contrast, a combination of CD8TCR-therapy with CD4TCR-therapy, each targeting an independent neoantigen, eradicated large and established solid tumors of natural heterogeneity. CD4TCR-therapy targeted a mutant neoantigen on tumor stroma while direct cancer cell recognition by CD8TCR-therapy was essential for cure. Thus, two cancer-specific TCRs can be essential and sufficient to eradicate heterogeneous solid tumors expressing unmanipulated, autochthonous neoantigens. Citation Format: Steven Wolf, Vasiliki Anastasopoulou, Kimberley Drousch, Markus Diehl, Boris Engels, Poh Yin Yew, Kazuma Kiyotani, Yusuke Nakamura, Karin Schreiber, Hans Schreiber, Matthias Leisegang. Adoptive transfer of one CD4TCR and one CD8TCR targeting autochthonous neoantigens are essential and sufficient for eradication of naturally heterogeneous solid tumors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB073.

Progression of Heterogeneous Breast Tumors

Ki-67 proliferative index (Ki-67) is a predictive and prognostic factor in breast cancer (BC). However, some international committees do not recommend its use in routine practice due to insufficient clinical evidence and lack of standardisation and assessment method reproducibility. Scoring of Ki-67 by digital pathology may contribute to overcome these drawbacks. We evaluated 136 core biopsies of BC patients and calculated the correlation of Ki-67 scored by two breast pathologists with two methods, eyeballing visual assessment (EB) on the microscope and digital image analysis (DI), both assessed from hot spot areas (HS) and the average between hot and cold spot areas (AVE). Good and higher correlation between pathologists was observed for HS using DI in comparison to EB (0.861 vs. 0.828). Correlation in HS with both methods was very similar in homogeneous tumours (0.869 vs. 0.866). Lower correlation was found in heterogeneous tumours if EB was used instead of DI (0.691 vs. 0.838). Good agreement with DI in AVE areas was observed in both homogenous and heterogeneous tumours (0.898 and 0.887). Concordance of tumour molecular profiles based on Ki-67 was better using DI in comparison to EB (Kappa index, 0.589 vs. 0675). Whereas EB and DI were alike in homogeneous tumour, DI improved agreement in heterogeneous tumours, particularly in AVE areas. Subgroup analysis for tumour grades also showed improvement of correlation by DI in AVE areas in all G1/G2/G3 groups. Digital pathology using AVE method can be useful for Ki-67 scoring in daily practice, especially in heterogeneous and G2 tumours, by a substantial improvement of agreement between observers and results accuracy.

Background: Clinical benefit of therapies matched with genetic profiling remains limited in patients with mCRC, which is partially attributed to tumor heterogeneity. Compared to molecular tumor heterogeneity using modern multi-omics approach, tumor heterogeneity in response to therapy has not been adequately studied. We aimed to characterize such tumor heterogeneity and evaluate its association with outcome in patients with mCRC. Methods: We used individual patient data from a clinical trial (NCT00364013) in Project Data Sphere who received first-line FOLFOX for mCRC and had ≥2 baseline metastatic target lesions. We used standard deviations of following tumor features in an individual patient to represent tumor heterogeneity: percentage change in size of a target lesion from baseline to first radiographic evaluation (H1), average speed of such percentage change in size as defined in H1 (H2), percentage change in size of a target lesion from baseline to the time of its best response to therapy (H3), and average speed of such percentage change in size as defined in H3 (H4). Cox proportional hazards models were used to evaluate association of tumor heterogeneity with survival. Results: Mean age of 328 patients included was 61 years. Median follow-up time was 18.2 months. 85% of them had liver metastasis; 15% had lung metastasis. All tumor heterogeneity variables in response to therapy (H1-4) were higher (p&amp;lt;0.001) in patients with disease progression compared to those without. In univariate analyses, higher tumor heterogeneity H1 (HR 7.2, 95% CI 2.3-22.6), H2 (HR 3.1, 95% CI 1.7-5.7), H3 (HR 6.4, 95% CI 2.1-19.3), and H4 (HR 4.1, 95% CI 2.4-7.1) were associated with shorter progression-free survival (PFS) (p&amp;lt;0.001). H4 high group (median as cutoff) had a shorter median PFS of 7.3 months compared to H4 low group (9.1 months, p=0.005). In multivariable analysis, higher H1 (HR 8.0, 95% CI 2.2-28.9, p=0.002), H2 (HR 3.9, 95% CI 2.0-7.6, p&amp;lt;0.001), H3 (HR 6.4, 95% CI 1.9-21.9, p=0.003), or H4 (HR 5.1, 95% CI 2.9-9.1, p&amp;lt;0.001) was each associated with a shorter PFS, after adjusted for baseline number of metastatic sites, number of target lesions, tumor size, age, albumin, ECOG, and lymph node metastasis. Similar findings were found when ratio of maximal change to baseline sum of tumor size due to therapy was adjusted for and when H1-4 were calculated from liver lesions only. Higher H1 (HR 3.6, 95% CI 1.2-11.2, p=0.03), H2 (HR 1.9, 95% CI 1.1-3.4, p=0.02), H3 (HR 5.2, 95% CI 1.8-14.8, p=0.002), or H4 (HR 2.5, 95% CI 1.5-4.0, p&amp;lt;0.001) was also each associated with a shorter overall survival in multivariable analysis. Conclusion: Higher tumor heterogeneity in response to chemotherapy within an individual patient was independently associated with early disease progression and shorter survival. It could provide prognostic value complementary to existing tumor heterogeneity variables in patients with mCRC. Citation Format: Junjia Liu, Hao Xie. Tumor heterogeneity in response to therapy and its association with outcome in patients with metastatic colorectal cancer (mCRC) [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 5255.