Method for selecting patient group predicted to respond to immunotherapy cancer treatment

Abstract

The present invention relates to a method for selecting a patient group predicted to respond to immunotherapy cancer treatment. According to the method of the present invention, it is possible to distinguish, in advance, between a patient group that is likely to respond to anti-TIGIT and / or anti-IL-6 immunotherapeutic agents and a patient group that is unlikely to respond to same, enabling effective treatment to be performed, and thus the method is expected to be widely applicable in the medical field.

Description

Methods for selecting patient groups expected to benefit from immunotherapy

[0001] The present invention relates to a method for selecting a patient group expected to be responsive to immunotherapy, specifically, anti-TIGIT and / or anti-IL-6 immunotherapy.

[0002] Traditional cancer treatments have focused on removing as many cancer cells as possible from the patient through surgery, radiation therapy, and chemotherapy. However, surgery and radiation therapy are effective only when they completely remove cancer cells in relatively early stages, before metastasis has occurred. This means that even if cancerous tissue is removed surgically, the risk of recurrence is high if a small number of cancer cells have spread to other parts of the body. Furthermore, while chemotherapy can be widely used in cancer treatment, it has a low cure rate for most solid tumors and has the disadvantage of killing rapidly dividing normal cells, leading to various side effects. To address these issues, immunotherapy has recently emerged.

[0003] Cancer immunotherapy is a treatment method that activates the body's immune system to attack cancer cells. It fundamentally differs from conventional cancer treatment in that it utilizes the body's natural power to fight cancer. Cancer immunotherapy addresses the shortcomings of existing cancer treatments. While first-generation chemotherapy drugs directly attack cancer cells and second-generation targeted chemotherapy drugs target cancer-related genes, third-generation chemotherapy drugs, called immunotherapy drugs, strengthen the immune system to treat cancer. Immunotherapy drugs include immune checkpoint inhibitors, immunotherapy drugs, and immunotherapy drugs, but the most representative of these are immune checkpoint inhibitors, which target PD-1, PD-L1, CTLA-4, and TIGIT. Among these, Pembrolizumab (Keytruda) and Nivolumab (Opdivo) are PD-1 inhibitors, Atezolizumab (Tecentriq) and Durvalumab (Imfinzi) are PD-L1 inhibitors, and Ipilimumab (Yervoy) is CTLA-4 inhibitor. However, research on drugs targeting TIGIT and patient targeting related to TIGIT is still insufficient.

[0004] Immunotherapy for cancer can be highly responsive to specific immunotherapies, with some patients responding significantly differently. Choosing the wrong immunotherapy for a given patient can result in serious side effects and significant time and cost losses. Therefore, a method to preemptively distinguish between patients who will benefit from specific immunotherapies and those who will not is essential. Given that TIGIT presents a relatively unknown area compared to other immune checkpoint inhibitors, the development of a method to preemptively identify patients likely to benefit from anti-TIGIT immunotherapy is urgently needed.

[0005] Accordingly, the present invention has been conceived to solve the above problems, and relates to a method for selecting a patient group expected to benefit from immunotherapy, specifically anti-TIGIT immunotherapy, or the closely related anti-IL-6 immunotherapy. According to the method of the present invention, it is possible to distinguish in advance a patient group expected to benefit from anti-TIGIT and / or anti-IL-6 immunotherapy from a patient group expected to be ineffective, and thus it is expected to be widely utilized in the medical field.

[0006] The present invention has been devised to solve the above-mentioned problems in the conventional technology, and relates to a method for selecting a patient group expected to be responsive to immunotherapy, specifically, anti-TIGIT and / or anti-IL-6 immunotherapy.

[0007] However, the technical problems to be solved by the present invention are not limited to the problems mentioned above, and other problems not mentioned can be clearly understood by those skilled in the art from the description below.

[0008] Hereinafter, various embodiments described herein will be described with reference to the drawings. In the following description, various specific details, such as specific configurations, compositions, and processes, are set forth to provide a thorough understanding of the present invention. However, certain embodiments may be practiced without one or more of these specific details, or in conjunction with other known methods and configurations. In other instances, well-known processes and manufacturing techniques have not been described in specific detail so as not to unnecessarily obscure the present invention. Reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, configuration, composition, or characteristic described in connection with the embodiment is included in one or more embodiments of the present invention. Thus, the appearances of "in one embodiment" or "an embodiment" in various places throughout this specification do not necessarily refer to the same embodiment of the present invention. Additionally, the particular features, configurations, compositions, or characteristics may be combined in any suitable manner in one or more embodiments.

[0009] Unless otherwise specifically defined in the specification, all scientific and technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs.

[0010] Throughout the specification, whenever a part is said to "include" a component, this does not mean that it excludes other components, but rather that it may include other components, unless otherwise stated.

[0011] In one embodiment of the present invention, AKR1C1 is isolated from a biological sample from a cancer patient. + Inflammatory fibroblasts (AKR1C1 +Provided is a method for selecting a patient group predicted to have a good response to immunotherapy, including a step of confirming the presence of inflammatory fibroblasts or mesothelium-derived fibroblasts.

[0012] In another aspect of the present invention, AKR1C1 is isolated from a biological sample from a cancer patient. + Inflammatory fibroblasts (AKR1C1 + Provided is a method for predicting the effectiveness of immunotherapy in a cancer patient, comprising a step of determining the presence of inflammatory fibroblasts or mesothelium-derived fibroblasts.

[0013] Hereinafter, in the specification of the present invention, cancer immunotherapy refers to an anticancer drug that activates the body's immune system to help attack cancer cells. As a treatment that complements the shortcomings of existing cancer treatments, if first-generation chemotherapy drugs directly attack cancer cells and second-generation targeted anticancer drugs function to attack cancer-related genes, then third-generation cancer immunotherapy drugs strengthen the immune system to treat cancer. Cancer immunotherapy drugs include immune checkpoint inhibitors, immune cell therapy drugs, and immune virus therapy drugs, but the most representative of them is the immune checkpoint inhibitor, which is a drug that targets PD-1, PD-L1, CTLA-4, and TIGIT. Among these, Pembrolizumab (Keytruda) and Nivolumab (Opdivo) are PD-1 inhibitors, Atezolizumab (Tecentriq) and Durvalumab (Imfinzi) are PD-L1 inhibitors, and Ipilimumab (Yervoy) is CTLA-4 inhibitor. However, research on drugs targeting TIGIT (T-cell immunoreceptor with immunoglobulin and ITIM domain) and patient targeting related to TIGIT is still insufficient.

[0014] The above TIGIT is one of the immune checkpoint receptors, such as PD-1 or CTLA-4, and is expressed on T cells. It is regulated by ligands such as PVR (CD155) or Nectin2 (CD112), which are expressed on cancer cells, fibroblasts, and antigen-presenting cells, and plays a role in suppressing T cell activity.

[0015] As third-generation anticancer drugs, immunotherapy activates the body's immune system to fight cancer cells. Therefore, it generally has fewer side effects than cytotoxic chemotherapy, overcomes the limitations of existing drugs for metastatic cancer, and addresses resistance issues. However, the biggest problem with immunotherapy is its low response rate. In cancer treatment, the "response rate" refers to the percentage of patients whose cancer shrinks or disappears as a result of treatment. Monotherapy with immune checkpoint inhibitors, among immunotherapy drugs, typically has a response rate of around 20%, although this varies depending on the cancer type. While melanoma, which responds very well to treatment, can show a high response rate of around 40%, lung cancer can show a response rate of around 20%, and stomach or bile duct cancer can show a lower response rate of around 10%. Furthermore, in a small number of patients, immunotherapy treatment has been reported to accelerate tumor growth, a phenomenon known as "rapid progression," rather than suppressing cancer progression. It is understood that the very different treatment response rates depending on the individual are because the immune response may occur differently depending on the tumor microenvironment in which the cancer cells grow, even for the same type of cancer cell.

[0016] If immunotherapy fails to respond to cancer patients, or if tumor growth accelerates, the burden of high treatment costs is significant. Furthermore, the golden window for effective treatment, when appropriate treatment could be achieved, is missed, making treatment more difficult. Therefore, there is an urgent and critical need for technologies that can predict individual patient responsiveness to specific immunotherapy agents before administration.

[0017] According to the present invention, by implementing the AND gating algorithm, the characteristic genes that are repeatedly up- or down-regulated in tumors compared to normal tissues in the major cell types that constitute the TME of various organs were systematically characterized, and as a result, CD8 of pancreatic tumor tissues + T cells did not show upregulation of PDCD1 and LAG3, which may explain the current inapplicability of immune checkpoint inhibitors in pancreatic cancer (PAAD) unlike other cancer types. In addition, when the expression levels of immune checkpoint protein-related ligands in cancer cells themselves were compared by cancer type, pancreatic cancer showed a uniformly high expression of PVR, Nectin2, and Nectin4, which are ligands for TIGIT. This suggests that specific CAFs (AKR1C1) are present in pancreatic cancer, for which immune checkpoint inhibitors are currently not applicable. + This suggests that inflammatory fibroblasts and / or mesothelial cell-derived fibroblasts may be a clue for anti-TIGIT immunotherapy.

[0018] Accordingly, the method for selecting a patient group predicted to have a good effect on the immunotherapy treatment of the present invention is to use AKR1C1 in a biological sample isolated from a cancer patient. + Inflammatory fibroblasts (AKR1C1 + To determine the presence of inflammatory fibroblasts, or mesothelium-derived fibroblasts, AKR1C1 was added to the sample. + If there are more inflammatory fibroblasts or mesothelial cell-derived fibroblasts than in the control group, this may be a method for predicting the cancer patient as a group of patients who will respond well to immunotherapy.

[0019] In addition, the method of predicting the effect of immunotherapy on cancer patients of the present invention is to detect AKR1C1 in a biological sample isolated from a cancer patient. + Inflammatory fibroblasts (AKR1C1 +To determine the presence of inflammatory fibroblasts, or mesothelium-derived fibroblasts, AKR1C1 was added to the sample. + This may be a method for predicting that the immune anticancer treatment effect of the cancer patient will be good if the number of inflammatory fibroblasts or mesothelial cell-derived fibroblasts is greater than that of the control group.

[0020] In the specification of the present invention, the immuno-oncology agent may be an immune checkpoint inhibitor, and the immune checkpoint inhibitor may include at least one selected from the group consisting of a CTLA-4 (Cytotoxic T-lymphocyte-associated antigen-4) inhibitor, a PD-1 (Programmed cell death protein 1) inhibitor, a PD-L1 (Programmed death-ligand 1) inhibitor, a KIR (Killer-cell immunoglobulin-like receptor) inhibitor, a LAG3 (Lymphocyte Activation Gene-3) inhibitor, a CD137 inhibitor, an OX40 inhibitor, a CD47 inhibitor, a CD276 inhibitor, a CD27 inhibitor, a GITR (Glucocorticoid-induced tumor necrosis factor receptor-related protein) inhibitor, a TIGIT (T-cell innunoreceptor with immunoglobulin and ITIM domain) inhibitor, and an IL-6 (interleukin-6) inhibitor, preferably a TIGIT inhibitor, or an IL-6 It is an inhibitor, but is not limited to it.

[0021] In the specification of the present invention, the cancer may be at least one selected from the group consisting of breast cancer, cervical cancer, glioma, brain cancer, melanoma, lung cancer, bladder cancer, prostate cancer, leukemia, kidney cancer, liver cancer, colon cancer, pancreatic cancer, stomach cancer, gallbladder cancer, ovarian cancer, lymphoma, osteosarcoma, uterine cancer, oral cancer, bronchial cancer, nasopharyngeal cancer, laryngeal cancer, skin cancer, blood cancer, thyroid cancer, parathyroid cancer, ureteral cancer, adenocarcinoma, and thymic cancer, and preferably may be breast cancer or pancreatic cancer, and more preferably may be triple-negative breast cancer or pancreatic cancer, but is not limited thereto.

[0022] In addition, in the specification of the present invention, the biological sample isolated from the cancer patient is cancer tissue, or tissue containing cancer cells, cell, cell extract, whole blood, leukocytes, peripheral blood mononuclear cells, buffy coat, plasma, serum, sputum, tears, mucus, nasal washes, nasal aspirate, breath, urine, semen, saliva, peritoneal washings, ascites, cystic fluid, meningeal fluid, amniotic fluid, glandular fluid, pancreatic fluid, lymph fluid, pleural fluid, nipple aspirate, bronchial It may be at least one selected from the group consisting of bronchial aspirate, synovial fluid, joint aspirate, organ secretions, and cerebrospinal fluid, and more preferably, it may include a tumor microenvironment.

[0023] In the present invention, the tumor microenvironment (TME) encompasses the constituent cell populations, such as vascular cells, immune cells, and stromal cells, present within a tumor, and their environments. It is the overall environment in which cancer cells proliferate and evolve. It is known to directly and indirectly influence tumor growth and progression. The TME forms a physical and signaling barrier around the tumor, hindering the penetration of immune cells and drugs. Consequently, it causes non-responsiveness and drug resistance to anticancer agents, such as targeted therapies and immunotherapies, and causes cancer growth and metastasis.

[0024] In the specification of the present invention, the above AKR1C1 + Inflammatory fibroblasts (AKR1C1 + Inflammatory fibroblasts (CFFs), or mesothelium-derived fibroblasts (MSFs), are a type of cancer-associated fibroblasts (CAFs). CAFs are fibroblasts that reside around cancer cells, i.e., within the tumor-like membrane (TME). They are known to promote cancer cell invasion and metastasis, block drug effects, increase angiogenesis in cancer tissue, and secrete various cytokines, promoting cancer progression and metastasis. They also bind to the immune protein immunoglobulin A (IgA) to suppress the immune response.

[0025] Recently, it has been suggested that CAFs may act differently in different cancer types, affecting cancer cell growth, metastasis, and immune responses. Therefore, it is necessary to understand the function of CAFs through research on each cancer type and develop customized diagnosis and treatment strategies based on this. Accordingly, the inventors of the present invention tracked CAFs that specifically promote cancer development in breast and pancreatic cancers and identified AKR1C1. + Inflammatory fibroblasts (AKR1C1 +inflammatory fibroblasts), and / or mesothelium-derived fibroblasts.

[0026] In the present invention, the AKR1C1 + Inflammatory fibroblasts and / or mesothelial cell-derived fibroblasts may exist in the tumor microenvironment, may promote cancer development, and may actively express TIGIT (T-cell innunoreceptor with immunoglobulin and ITIM domain) and / or IL-6 (interleukin-6), wherein TIGIT or IL-6 may function as a type of immune checkpoint inhibitor. Therefore, AKR1C1 in biological samples isolated from cancer patients + Inflammatory fibroblasts (AKR1C1 + The presence of inflammatory fibroblasts or mesothelium-derived fibroblasts can be used to predict the responsiveness of patients with cancer to TIGIT or IL-6-related immunotherapy.

[0027] According to the method of the present invention, it is possible to distinguish in advance a group of patients for whom anti-TIGIT and / or anti-IL-6 immunotherapy is expected to be effective and a group of patients for whom anti-TIGIT and / or anti-IL-6 immunotherapy is expected to be ineffective, and effective treatment can be provided based on this distinction.

[0028] Figure 1 presents the results of characterizing genes that are repeatedly up- or down-regulated in tumors compared to normal tissues in the major cell types that make up the TME of various organs.

[0029] Figure 2 shows the results of gene ontology (GO) analysis in the major cell types that make up the TME of various organs.

[0030] Figure 3 shows the results of comparing the ligand expression levels in cancer cells themselves by cancer type.

[0031] Figure 4 shows the results of identifying several fibroblast subtypes exhibiting immune-related gene expression.

[0032] Figure 5 is AKR1C1 + We present the results of spatial transcriptomic analysis to investigate the co-localization patterns of inflammatory fibroblasts.

[0033] Figure 6 shows the results of deconvolution of the bulk transcriptome of samples treated with immune checkpoint inhibitors in a cohort of various cancer types.

[0034] Figure 7 shows the expression patterns of immune checkpoint inhibitor-related ligands in fibroblasts classified in detail from single-cell transcriptome data of breast and pancreatic cancer.

[0035] Figure 8 shows the expression patterns of immune checkpoint inhibitor-related ligands in independent patient data of breast cancer and pancreatic cancer.

[0036] Figure 9: AKR1C1 by breast cancer subtype + It shows the activity pattern of inflammatory fibroblasts or mesothelial cell-derived fibroblasts.

[0037] Figure 10 shows the expression pattern of PVR or IL-6 by breast cancer subtype.

[0038] Figure 11 AKR1C1 in breast cancer TME + Inflammatory fibroblasts, cytotoxic T cells, and CD8 + Indicates the relevance of T cells.

[0039] Figure 12 shows the expression pattern of PVR or TIGIT in mesothelial cell-derived fibroblasts of pancreatic cancer TME.

[0040] Figure 13: AKR1C1 in pancreatic cancer TME +The expression pattern of IL-6 in inflammatory fibroblasts or mesothelial cell-derived fibroblasts is shown.

[0041] Figure 14 AKR1C1 in pancreatic cancer TME + Inflammatory fibroblasts or mesothelial cell-derived fibroblasts and cytotoxic T cells and CD8 + Indicates the relevance of T cells.

[0042] The expression patterns of each ligand were examined in breast and pancreatic cancer cells in fibroblasts classified in detail from single-cell transcriptome data. As a result, PVR, which was underexpressed in breast cancer cells, was found to be significantly higher than AKR1C1. + It was observed that PVR was specifically and strongly expressed in inflammatory fibroblasts. In pancreatic cancer, PVR was highly expressed not only in cancer cells but also in mesothelial cell-derived fibroblasts.

[0043] To validate the results obtained from the single-cell transcriptome data at the protein level in independent patient data, transcriptome and proteome data for pancreatic cancer and breast cancer, respectively, were obtained from The Cancer Genome Atlas (TCGA) and the Clinical Proteomic Tumor Analysis Consortium (CPTAC), and the expression levels of each ligand were confirmed. As a result, it was confirmed that PVR, Nectin2, and Nectin4 were all expressed significantly higher than PD-L1 at the protein level in both breast and pancreatic cancers.

[0044] Meanwhile, when breast cancer was divided into subtypes according to the PAM50 classification system, AKR1C1 was found in Basal (TNBC) breast cancer. + We confirmed that inflammatory fibroblasts showed the strongest activity. Mesothelial cell-derived fibroblasts also showed somewhat stronger activity in basal (TNBC). In the single-cell transcriptome data above, AKR1C1 +Since PVR and IL-6 were strongly expressed in inflammatory fibroblasts, the expected pattern was confirmed in the validation data when PVR and IL-6 were highly expressed in basal (TNBC) breast cancer. In addition, AKR1C1 + The higher the activity of inflammatory fibroblasts, the more cytotoxic T cells and CD8 cells infiltrated into cancer tissue. + Lower T cell counts have also been observed in breast cancer.

[0045] In the case of pancreatic cancer, PVR was expressed in both mesothelial cell-derived fibroblasts and cancer cells in the single-cell transcriptome data. On the other hand, in the validation data, PVR was not specifically high in the patient group with strong activity of mesothelial cell-derived fibroblasts. This is presumably because the validation data is not expression data at the single-cell level and therefore appears mixed with the expression level in cancer cells. However, TIGIT, a receptor for PVR, was specifically expressed significantly higher in proportion to the high activity of mesothelial cell-derived fibroblasts. In the single-cell transcriptome data, in the case of IL-6, AKR1C1 + It was strongly expressed in inflammatory fibroblasts and at a significant level in mesothelial cell-derived fibroblasts, which was confirmed in the validation data. In addition, AKR1C1 was expressed in pancreatic cancer. + The higher the activity of inflammatory fibroblasts and mesothelial cell-derived fibroblasts, the more cytotoxic T cells and CD8 cells infiltrated into cancer tissue. + It was observed that the amount of T cells was low. In the case of IL-6 blockade, there is a study result that it can reduce the side effects that usually occur when performing immune checkpoint inhibitor treatment while maintaining the therapeutic effect (Cancer Cell 40:509 (2022)). Therefore, the above results indicate that AKR1C1 in pancreatic cancer +This may provide a clue that blocking IL-6 through a method of suppressing inflammatory fibroblasts and / or mesothelial cell-derived fibroblasts may enhance the therapeutic effect of immune checkpoint inhibitors.

[0046] Hereinafter, the present invention will be described in detail with reference to the following examples. However, the following examples are merely illustrative of the present invention, and the content of the present invention is not limited to the following examples.

[0047] Example

[0048] [Implementation Method]

[0049] 1. Immunotherapy cohort data

[0050] The transcriptome data of the immunotherapy cohort used in the present invention are from Kim et al. (Nat. Genet. 55, 221-231 (2023))(n=335), Van Allen et al. (Science350, 207-211 (2015))(n=75), Gide et al. (Cancer Cell35, 238-255.e6 (2019))(n=73), Riaz et al. (Cell171, 934-949.e16 (2017))(n=46), Hugo et al. (Cellvol. 165 35-44 (2016))(n=25), Mariathasan et al. (Nature554, 544-548 (2018))(n=347), McDermott et al. (Nat. Med.24, 749-757 (2018))(n = 165), Miao et al. (Bioinformatics30, 2114-2120 (2014))(n = 33).

[0051] 2. Single-cell data inclusion criteria and data collection

[0052] For data selection, we first searched relevant 10x scRNA-seq datasets to homogenize and minimize batching issues caused by various chemicals. Data sets were searched and downloaded from PubMed, Google Scholar, Gene Expression Omnibus, Single Cell Portal (https: / singlecell.broadinstitute.org / single_cell), COVID-19 Cell Atlas (https: / www.covid19cellatlas.org / ), and Curated Cancer Cell Atlas (https: / www.weizmann.ac.il / sites / 3CA / ).

[0053] Studies generated from the 10x-genome reagent kit and included cancer, precancerous, benign tumor, and normal samples, and among the normal control samples, non-malignant tissues derived from cancer patients (annotated as adjacent normal) and tissues from healthy normal individuals (annotated as normal) were collected separately. Cells labeled (e.g., CD45 + Studies that included only body fluid samples (e.g., ascites, cerebrospinal fluid, or PBMCs), cell line cultures, mouse studies, and studies generated from nuclei-seq were excluded. Consequently,

[0054] We obtained single-cell transcriptome data for 1,070 cancer tissues and 493 normal tissues for more than 30 cancer types, and classified various cell types (cancer cells, immune cells, fibroblasts, etc.) in detail to confirm gene expression in each.

[0055] 3. Analysis of single-cell RNA sequencing data

[0056] The gene columns in each dataset were realigned to the GRCh38 human reference genome (official Cell Ranger reference, version 2020-A). Cells with UMI counts less than 2000 and 500 detected genes in each dataset were considered empty droplets and removed from the dataset. Cells with more than 7000 detected genes were also considered potential doublets and removed from the dataset. The cell-gene count matrix was loaded and analyzed using the Scanpy (v. 1.8.2) Python package, and clustering, annotation, and downstream analysis were performed using tools from the Scanpy package and some custom code. Scrublet was used for doublet detection. Additionally, to reduce the computational burden and accelerate downstream analyses, we selected cell subsets for each dataset using geometric sketches that reflect transcriptional diversity and preserve rare cell types.

[0057] 4. Cell type annotation and batch editing

[0058] In this invention, we merged all tumor-normal scRNA-seq data and divided the dataset to reduce computational burden, visualize cells, and annotate them. We used BBKNN as a batch-effect correction algorithm to generate a connected graph structure, obtained UMAP at a global scale, and then annotated the data based on cell type-specific marker genes. We then examined and refined the key cell type annotations for each dataset.

[0059] 5. Copy number variation inference for malignant cell identification

[0060] Single-cell transcriptome-based large-scale copy number variations (CNVs) of malignant cells were inferred using inferCNVpy (available at https: / / github.com / icbi-lab / infercnvpy) with the default window size and gencode v29 as the genomic location reference. In this study, infercnvpy.tl.infercnv was used to infer CNVs, and normal immune cells or fibroblasts were selected as reference normal cells depending on each cancer type. Cells were visualized in CNV UMAP (infercnvpy.tl.umap) based on dimensionality reduction (infercnvpy.tl.pca) and clustering based on CNV profiles (infercnvpy.tl.leiden), and CNV scores were calculated using infercnvpy.tl.cnv_score. Cells were considered malignant if they formed separate clusters, a known characteristic of malignant cells, or had a high CNV score compared to known normal cell types (normal epithelial cells, fibroblasts, or immune cells, depending on the cancer type).

[0061] 6. AND gating algorithm for differential expression of genes to identify characteristic features of genes.

[0062] In this study, we applied the AND gating algorithm to extract tumor-enriched or immunotherapy-advantageous gene signatures for each cell type in various cancers. Cells from specific organs were divided into subsets and differential expression analysis was performed to identify genes that were upregulated in the cell type of interest compared to other cell types (log2 fold change > 0), or highly expressed in tumor tissue (or immunotherapy responders) compared to normal tissue (or immunotherapy non-responders; log2 fold change > 0.5 and adjusted p-value < 0.05). Genes that met both criteria were retained to generate tumor-specific / immunotherapy-advantageous gene signatures. P-values ​​were calculated using a two-tailed t-test on the log-normalized gene matrix and adjusted using the Benjamini-Hochberg method (Python packages scipy.stats v. 1.10.0 and statsmodels.stats v. 0.13.5). After obtaining the gene signature derived from AND-gating for each cell type and organ, the gene signatures from multiple organs were integrated to identify characteristic gene signatures for each cell type.

[0063] 7. Biological annotation of characteristic gene signatures in the tumor-normal ecosystem.

[0064] In this study, Enrichr was used to annotate the biological functions of tumor-specific hallmark gene signatures across various cell types. GO terms from MsigDB Hallmark 2020, GO Biological Process 2023, and GO Molecular Function 2023 were used, and only terms with an adjusted p-value less than 0.05 were considered significant.

[0065] 8. NMF preprocessing and visualization

[0066] After cell type annotation, NMF was performed separately for each individual tissue, taking into account each cell type category and tissue origin, to generate cell states that contribute to the heterogeneity within each individual. Starting from a log-normalized centered expression matrix of all genes, negative values ​​were set to zero. The sklearn.decomposition.NMF method was applied with default parameters implemented in the scikit-learn Python package v1.0.2. Considering that NMF requires a K parameter that influences the results, we ran NMF with different values ​​(K=5, 6, 7, 8, and 9), generating 35 modules for each individual. Next, we clustered and visualized the NMF modules graphically. First, all modules derived from each cell type were max-normalized and merged to anndata objects. After highly variable gene selection, low-quality modules (modules with NMF weights less than 10–20 or greater than 150–170, depending on the cell type) were removed. Dimensionality reduction was performed using principal components and UMAP visualization, and small module clusters with fewer than 150 modules were removed. Leiden clustering was then performed to derive a list of the top 50 genes for each cluster. Clusters enriched in ribosomal protein genes or mitochondrial encoding genes, composed of NMF modules from a single study, and suspected of reflecting a soup effect based on high similarity to expression profiles in doublet cells or other cell types were removed.

[0067] 9. Automated removal of doublet or soup effect clusters.

[0068] To identify and remove NMF module clusters (cell states) exhibiting doublet cells or soup-effect effects, we developed an algorithm to automate the detection of doublet or soup-effect clusters. Specifically, we identified two organs (only one organ if the difference in the number of NMF modules between the two dominant organs was greater than twofold) that comprised the majority of NMF clusters of interest and subdivided these organs from a geometrically sketched anndata of the tumor-normal meta-atlas. We then scored each cell type category in the subdivided anndata using the sc.tl.score function in the Scanpy package, using the top 50 genes derived from these clusters. Clusters were defined as doublet or soup-effect clusters and subsequently removed if their scores were greater than 0.2 in other cell types (e.g., T cells) compared to the cell type of interest (e.g., mesenchymal cells). To prevent the removal of EMT states, clusters with higher mesenchymal scores were excluded from the epithelial cell state.

[0069] 10. Defining and annotating cell states

[0070] After visualizing the NMF modules and clustering the states, we identified the top 50 genes with weighted averages for each state. For genes that overlapped between cell states, we orthogonally assigned the gene to the state with the higher NMF-weighted average. Azimuth (https: / / azimuth.hubmapconsortium.org / ), The Human Protein Atlas (https: / / www.proteinatlas.org / ), and Enrichr71 were used as primary references for annotating cell states. Furthermore, to validate cell states, we compared gene signatures obtained from other studies using Pearson correlation.

[0071] 11. Building a reference component using cell states

[0072] To assess the correspondence between cell states and cell subtypes, cells were projected using cell-type-specific cell state profiles as reference components. For each cell state category, orthogonal genes with weighted averages were identified for each state (see Cell State Definition and Annotation). Subsequently, cell cycle and cell state-derived features derived from surrounding RNA or doublets were removed, and genes constituting the remaining states were selected as variable genes from the log-normalized scRNA-seq dataset. The reference components were constructed by performing matrix multiplication between the scRNA-seq anndata and the cell state-weighted average (RCA = anndata. X.dot(cell state-weighted average)). These reference components were used to replace the principal components, followed by BBKNN using the dataset as the batch key. Final cell type annotations were then created based on the cell type-specific marker genes.

[0073] 12. Cell status score distribution measurement and concordance analysis

[0074] To determine which cell states are enriched in the cells of each individual, we used the sc.tl.score_genes function to score each individual by their orthologous genes for cell states, resulting in a cell state score. For the eight cancer types that comprise the majority of the pan-cancer atlas (BRCA, CRC, HCC, HNSC, LC, OV, PAAD, and RCC), we calculated the average score for each individual and performed a Pearson correlation between cell states to measure concordance. For each concordance between cell states, we calculated adjacency using the WGCNA package (v. 1.71) and plotted a Circos plot using the circlize package (v. 0.4.15). The thickness of the line in the Circos plot corresponds to the adjacency between cell states.

[0075] 13. Ratio of observed and expected cell states

[0076] To quantify the tissue or organ preference of a cell state, we calculated the ratio of the observed to expected (Ro / e). To quantify tissue Ro / e, we created a 3 x 2 contingency table by counting the occurrence of tissue origin (i.e., normal, adjacent normal, and tumor) of NMF modules in the cell state of interest and other cell states. To simultaneously consider tissue and organ origin when calculating Ro / e, we first extracted the NMF modules of the cell states, determined the proportion of organ origin within these modules, and then filtered out those derived from organs that comprised less than 3% of the total. We then created a contingency table by counting the organ origin occurrence of NMF modules in the cell state of interest and other cell states for each tissue origin. The expected counts were derived using chi-square analysis, and Ro / e was calculated using Equation 1 below.

[0077] [Mathematical Formula 1]

[0078]

[0079] If Ro / e > 0 or Ro / e < 0, the cell state was considered to be enriched or depleted in the specific tissue / organ.

[0080] 14. Ligand-receptor interaction analysis

[0081] AKR1C1 + To understand the functional properties of inflammatory fibroblasts and mesothelial cell-derived fibroblasts, we used cell-cell interaction inference tools such as CellPhoneDB to identify AKR1C1 + We investigated potential cellular interactions between inflammatory fibroblasts and / or mesothelial-derived fibroblasts and other cell types, focusing on gene expression programs specific to these two fibroblasts. The strength of the interaction was calculated by multiplying the normalized expression values ​​of the ligand and receptor for each cell-cell pair.

[0082] 15. Survival analysis using bulk transcriptomes

[0083] To assess the prognostic value of cell status and hallmark features in each cancer type, survival analysis was performed using TCGA RNA-seq data. Upper-quartile normalized FPKM data were collected across 28 cancer types at UCSC Xena. TCGA clinical data (OS) were obtained from the TCGA Pan-Cancer clinical data resource. Enrichment of cell status and hallmark features was calculated for each TCGA primary cancer sample using the single-sample gene set enrichment analysis (ssGSEA) function implemented in the Corto package (v. 1.1.10). Patients were grouped into depleted and enriched groups based on the mean cell status score of the analyzed samples. Kaplan-Meier curves were plotted using the ggsurvplot function, and statistical significance was quantified using the log-rank test, with multiple testing corrected using the Benjamini-Hochberg method. To assess the prognostic significance of cell status in the relevant organ, the Ro / e filtering threshold for each cell type was determined using Equation 2 below.

[0084] [Equation 2]

[0085]

[0086] Cell states were then identified as rare within an organ if the tumor-derived Ro / e value did not exceed a filtering threshold. This prevents deconvolution of rare cell states that are irrelevant to survival analysis of a specific organ.

[0087] 16. Building a network using cell states

[0088] We constructed an undirected network using cell states to visualize co-occurrence patterns. Each tissue network was constructed using cell states identified from the corresponding tissue origin. After calculating the co-occurrence of cell states, we calculated an adjacency matrix using the WGCNA package (v. 1.71). The adjacency values ​​of cell state pairs with a p-value greater than 0.05 were set to 0 to minimize false positives. The adjacency matrix was then imported into gephi (v. 0.10.1) to construct a connected network. Community detection was performed with default parameters, nodes were colored by modularity class, and nodes were scaled by average weight. ForceAtlas2 was selected for graph embedding.

[0089] 17. Collection and processing of transcriptome data from the immunotherapy cohort.

[0090] We collected large-scale transcriptomes from cohorts (eight cohorts across four cancer types) that received immunotherapy. Transcriptome data for the cohorts were collected from Kim et al. (n = 335), Van Allen et al. (n = 75), Gide et al. (n = 73), Riaz et al. (n = 46), Hugo et al. (n = 25), Mariathasan et al. (n = 347), McDermott et al. (n = 165), and Miao et al. (n = 33). Raw FASTQ files were obtained from all cohorts and processed using an integrated pipeline. First, adapter sequences in FASTQ files were trimmed with Trimmomatic (v. 0.39), and SortMeRNA (v. 2.1b) was used to filter rRNA. The filtered reads were aligned to the hg38 reference genome with STAR aligner (v. 2.7.6a) in two-pass default mode with gencode annotation (v. 35). The aligned reads were aligned with samtools (v. 1.7) and then read counts were calculated with HTSeq (v. 0.12.4). Read counts were normalized to TPM values ​​to quantify gene expression.

[0091] 18. Immunotherapy Cohort Data Analysis

[0092] We performed ssGSEA using Gseapy (v. 0.10.8) to score cell status per sample, and used the normalized enrichment scores for analysis. For the pan-cancer analysis, only samples with both response and survival data were included. Patients with a durable clinical benefit (complete response, partial response, stable disease with PFS > 6 months or OS > 1 year) were classified as responders, while other patients were classified as non-responders. We then performed a meta-analysis to examine the association between clinical response to immunotherapy and cell status across multiple cohorts. First, the scaled signature score was fitted to clinical response in each cohort using logistic regression. The calculated estimates and standard errors were pooled across cohorts using the metagen function in the meta package. Finally, a random-effects model was established to estimate the effect of cell status, accounting for heterogeneity between studies. The overall estimates, standard errors, and p-values ​​were obtained from the random-effects model.

[0093] 19. Precancerous spatial transcriptome analysis

[0094] Spatial transcriptomic analysis of 137 cancer datasets across 11 cancer types was performed using cell2location94 with default parameters to quantify the spatial distribution of cell types. For each spatial transcriptomic cancer type, we used the corresponding cancer scRNA-seq dataset from the Pan-Cancer Single-Cell Atlas as a reference. We then quantified spatial colocalization patterns using spot-wise Pearson correlations, along with estimated cell type abundances, similar to published data. A high positive Pearson correlation indicates similar spatial distributions between two cell types, whereas a negative Pearson correlation suggests distinct spatial distributions between the two cell types.

[0095] [Results]

[0096] 1. Identification of universal signature gene signatures in the tumor-normal ecosystem.

[0097] By implementing an AND gating algorithm, we systematically characterized the characteristic genes that are repeatedly up- or down-regulated in tumors compared to normal tissues in the major cell types that constitute the tumor microenvironment (TME) of various organs. CD8 + For T cells, co-stimulatory molecule (CD27) and immune checkpoint or exhaustion markers such as CXCL13, PDCD1, TIGIT, CTLA4, LAG3, and TNFRSF9 were generally increased in tumors, whereas IL7R, PTGER2, and PTGER4 were increased in normal tissues (Fig. 1). Of note, CD8 + T cells did not show upregulation of PDCD1 and LAG3, which may explain the current inapplicability of immune checkpoint inhibitors in pancreatic cancer (PAAD) in contrast to other cancer types. Similarly, tumor-associated NK cells were marked by ZNF683 and KRT81. Tumor-infiltrating Treg upregulate genes with regulatory functions, such as RBPJ, CXCR3, and ZBED2, whereas normal tissue Treg upregulate CCR7 and CXCR5, indicating distinct mechanisms for immune cell recruitment and infiltration. Notably, tumor-infiltrating macrophages universally expressed immune checkpoint (IL4I1), M2 polarization-related (SPP1), and inflammatory genes (CCL7, ADAMDEC1, and SLAMF9), whereas tumor-infiltrating dendritic cells showed increased expression of CCL19 and LAMP3, which are associated with inflammatory and migratory functions (Figure 1). Gene ontology (GO) analysis revealed that tumor-infiltrating macrophages, dendritic cells, and CD8 +Genes upregulated in T cells were found to be enriched in relevant functions and pathways, including defense responses to viruses, responses to type II interferons, inflammatory responses, chemotaxis of lymphocytes, and cytokine-mediated signaling pathways. In non-immune cell types, cancer cells universally expressed GO terms associated with protein serine / threonine kinase activity (PRKCA, GSK3B, and CAMKK2), glycolysis (PLOD1, EGLN3, and P4HA1), mTORC1 signaling (SLC2A1, GMPS, and PDK1), and positive regulation of cell cycle processes (E2F7, E2F8, and KIF23) (Figure 2). Cancer-associated fibroblasts (CAFs) expressed well-known markers, including FAP, COL1A1, COL10A1, MMP11, and CTHRC1, as well as other genes, such as INHBA, SLC12A8, F2R, and COL12A1, in various organs, while tumor endothelial cells upregulated angiogenesis-related genes, including CHST1, FOLH1, and MMP15 (Figure 1). Both tumor-associated fibroblasts and endothelial cells were enriched for aspects related to extracellular matrix organization, cell migration regulation, cell-matrix adhesion, and filamin binding (Figure 2). Overall, these results illustrate the characteristic dysregulated nature of all TME components.

[0098] 2. Comparison of ligand expression levels by cancer type

[0099] The expression levels of ligands in cancer cells themselves were compared by cancer type. PD-L1, a ligand for PD-1, showed the highest expression in lung cancer cells, whereas PVR, Nectin2, and Nectin4, ligands for TIGIT, showed high expression in pancreatic and breast cancer cells (Fig. 3). Bladder cancer was the cancer type with the highest expression of Nectin4, but the small number of patients studied made it difficult to draw a definitive conclusion, and the expression of PVR and Nectin2 was low. On the other hand, pancreatic cancer showed high expression of all three ligands, and breast cancer showed the same high expression of Nectin4 and Nectin2 as pancreatic cancer, but the expression of PVR was low.

[0100] Recently, numerous studies have shown that cancer-associated fibroblasts (CAFs), particularly inflammatory CAFs (iCAFs), alter the immune environment of cancer tissues and inhibit T cell activation, thereby aiding the development and progression of cancer. Therefore, these results suggest that specific CAFs may provide a clue for anti-TIGIT immunotherapy in pancreatic cancer, where immune checkpoint inhibitors are currently ineffective.

[0101] 3. Characterization of fibroblasts by subtype

[0102] Fibroblasts are a highly heterogeneous population with diverse functions, including collagen deposition, angiogenesis, and cytokine secretion, and play a central role in shaping the TME. Fibroblasts promote inflammation and modulate the tissue microenvironment toward immunosuppression in the context of cancer, but the diversity of inflammatory fibroblasts has not been extensively explored in previous pan-cancer studies. Therefore, projecting a mesenchymal cell population to defined states, we identified several fibroblast subtypes exhibiting immune-related gene expression (Figure 4). Distinct patterns of interaction between fibroblast subtypes were identified. Accordingly, we hypothesized that different microenvironmental environments in each cancer type induce distinct phenotypes of specific fibroblasts. To investigate their colocalization patterns, we analyzed the spatial transcriptome. + Inflammatory fibroblasts include cancer cells, neutrophils, and CTSK + Significant colocalization with macrophages, DC1, and PRR-induced mo-DCs (Fig. 5).

[0103] 4. Determining cell status to predict immunotherapy according to cancer type

[0104] After identifying the diversity and dynamics of interferon-enriched and tumorigenic communities in tumors, adjacent normal tissues, and healthy normal tissues, we leveraged these cell states to deconvolve the bulk transcriptomes of samples treated with immune checkpoint inhibitors in a cohort of diverse cancers. The clinical benefits of checkpoint blockade were studied in the context of exhausted CD8 + T cells, mesenchymal-derived interferon, CXCL9 + Macrophage, CD160 + Intraepithelial lymphocytes, Tregs, DC1, ISG15 + Macrophage, XCL1 + / CD16 + NK cells, IFIT1 + Interferon signaling, Tfh, GCB, LAMP3 + DC, pDC, CD16 +Monocyte-derived macrophages, CCL19 + Fibroblasts and plasma cell precursor states were highlighted (Fig. 6). Fibroblasts, osteoblasts, mesothelial cell-derived fibroblasts, and CTSK + Macrophage cell status was associated with poor response to immunotherapy across the cohort, many of which belonged to the tumor-initiating community (Figure 6). These pro-tumorigenic component statuses (i.e., fibroblasts, osteoblasts, mesothelial cell-derived fibroblasts, and CTSK) + Given that macrophages negatively impact immunotherapy responses across various cancer types, alternative treatment strategies should be pursued for patients with a pro-tumor ecosystem.

[0105] 5. Transcriptome analysis of single fibroblasts in breast or pancreatic cancer.

[0106] The expression patterns of each ligand were examined in breast and pancreatic cancer cells in fibroblasts classified in detail from single-cell transcriptome data. As a result, PVR, which was underexpressed in breast cancer cells, was found to be significantly higher than AKR1C1. + It was observed that PVR was specifically and strongly expressed in inflammatory fibroblasts. In pancreatic cancer, PVR was highly expressed not only in cancer cells but also in mesothelial cell-derived fibroblasts (Fig. 7).

[0107] To validate the results obtained from the single-cell transcriptome data at the protein level in independent patient data, we obtained transcriptome and proteome data from pancreatic cancer or breast cancer from The Cancer Genome Atlas (TCGA) and Clinical Proteomic Tumor Analysis Consortium (CPTAC) and confirmed the expression levels of each ligand. As a result, it was confirmed that PVR, Nectin2, and Nectin4 were all expressed significantly higher than PD-L1 at the protein level in both breast and pancreatic cancers (Fig. 8).

[0108] Meanwhile, when breast cancer was divided into subtypes according to the PAM50 classification system, AKR1C1 was found in Basal (TNBC) breast cancer. + We confirmed that inflammatory fibroblasts showed the strongest activity. Mesothelial cell-derived fibroblasts also showed somewhat stronger activity in basal (TNBC) cells (Fig. 9). In the single-cell transcriptome data above, AKR1C1 + Since PVR and IL-6 were strongly expressed in inflammatory fibroblasts, the expected pattern was confirmed in the validation data when PVR and IL-6 were highly expressed in basal (TNBC) breast cancer (Fig. 10). In addition, AKR1C1 + The higher the activity of inflammatory fibroblasts, the more cytotoxic T cells and CD8 cells infiltrated into cancer tissue. + Lower levels of T cells were also observed in breast cancer (Fig. 11).

[0109] In the case of pancreatic cancer, PVR was expressed in both mesothelial cell-derived fibroblasts and cancer cells in the single-cell transcriptome data. On the other hand, in the validation data, PVR was not specifically high in the patient group with strong activity of mesothelial cell-derived fibroblasts. This is presumably because the validation data is not expression data at the single-cell level and therefore appears mixed with the expression level in cancer cells. However, TIGIT, a receptor for PVR, was specifically expressed significantly higher in proportion to the high activity of mesothelial cell-derived fibroblasts (Fig. 12). In the single-cell transcriptome data, in the case of IL-6, AKR1C1 + It was strongly expressed in inflammatory fibroblasts and at a significant level in mesothelial cell-derived fibroblasts, which was confirmed in the validation data (Fig. 13). In addition, AKR1C1 was expressed in pancreatic cancer. + The higher the activity of inflammatory fibroblasts and mesothelial cell-derived fibroblasts, the more cytotoxic T cells and CD8 cells infiltrated into cancer tissue.+ A low amount of T cells was observed (Fig. 14). There are research results showing that IL-6 blockade can reduce the side effects that usually occur when performing immune checkpoint inhibitor treatment while maintaining the therapeutic effect (Cancer Cell 40:509 (2022)). Therefore, the above results suggest that AKR1C1 in pancreatic cancer + This may provide a clue that blocking IL-6 through a method of suppressing inflammatory fibroblasts and / or mesothelial cell-derived fibroblasts may enhance the therapeutic effect of immune checkpoint inhibitors.

[0110] In addition, it has been revealed that IL-6 produced in the cancer microenvironment can interact with specific KRAS mutations and affect cancer progression (Cancer Cell 19:456 (2011)). Based on these results, pancreatic cancer and breast cancer can be selected as targets for anti-TIGIT and anti-IL-6 compared to other cancer types. In the case of breast cancer, a higher therapeutic effect can be expected, especially in the Basal (TNBC) subtype, and in the case of pancreatic cancer, AKR1C1 + A better therapeutic effect can be expected when the activity of inflammatory fibroblasts and mesothelial cell-derived fibroblasts is high.

[0111] While specific aspects of the present invention have been described in detail above, it should be apparent to those skilled in the art that these specific descriptions are merely preferred embodiments and do not limit the scope of the present invention. Therefore, the substantial scope of the present invention is defined by the appended claims and their equivalents.

[0112] The present invention relates to a method for screening a patient population expected to benefit from immunotherapy, specifically anti-TIGIT immunotherapy, or the closely related anti-IL-6 immunotherapy. The method of the present invention is expected to find significant application in the medical field, as it enables a priori distinction between patient populations predicted to benefit from anti-TIGIT and / or anti-IL-6 immunotherapy and those predicted to be ineffective.

Claims

1. AKR1C1 in biological samples isolated from cancer patients + Inflammatory fibroblasts (AKR1C1 + A method for selecting a patient group predicted to be responsive to immunotherapy, comprising a step of confirming the presence of inflammatory fibroblasts or mesothelium-derived fibroblasts.

2. In paragraph 1, A method wherein the above immunotherapy treatment is an immune checkpoint inhibitor treatment.

3. In paragraph 2, A method wherein the above immune checkpoint inhibitor comprises at least one selected from the group consisting of a CTLA-4 (Cytotoxic T-lymphocyte-associated antigen-4) inhibitor, a PD-1 (Programmed cell death protein 1) inhibitor, a PD-L1 (Programmed death-ligand 1) inhibitor, a KIR (Killer-cell immunoglobulin-like receptor) inhibitor, a LAG3 (Lymphocyte Activation Gene-3) inhibitor, a CD137 inhibitor, an OX40 inhibitor, a CD47 inhibitor, a CD276 inhibitor, a CD27 inhibitor, a GITR (Glucocorticoid-induced tumor necrosis factor receptor-related protein) inhibitor, a TIGIT (T-cell innunoreceptor with immunoglobulin and ITIM domain) inhibitor, and an IL-6 (interleukin-6) inhibitor.

4. In paragraph 3, A method wherein the above immune checkpoint inhibitor is a TIGIT inhibitor or an IL-6 inhibitor.

5. In paragraph 1, A method wherein the cancer is breast cancer or pancreatic cancer.

6. In paragraph 1, The above method, AKR1C1 in the above sample + A method for predicting a cancer patient as a group of patients likely to respond well to immunotherapy when the number of inflammatory fibroblasts or mesothelial cell-derived fibroblasts is greater than that of a control group.

7. In paragraph 6, A method wherein the above-mentioned immunotherapy agent is a TIGIT inhibitor or an IL-6 inhibitor.

8. In paragraph 1, A method wherein the biological sample comprises a tumor microenvironment.

9. AKR1C1 in biological samples isolated from cancer patients + Inflammatory fibroblasts (AKR1C1 + A method for predicting the effectiveness of immunotherapy in a cancer patient, comprising the step of determining the presence of inflammatory fibroblasts or mesothelium-derived fibroblasts.

10. In paragraph 9, A method wherein the above immunotherapy treatment is an immune checkpoint inhibitor treatment.

11. In paragraph 10, A method, wherein the immune checkpoint inhibitor comprises at least one selected from the group consisting of a CTLA-4 (Cytotoxic T-lymphocyte-associated antigen-4) inhibitor, a PD-1 (Programmed cell death protein 1) inhibitor, a PD-L1 (Programmed death-ligand 1) inhibitor, a KIR (Killer-cell immunoglobulin-like receptor) inhibitor, a LAG3 (Lymphocyte Activation Gene-3) inhibitor, a CD137 inhibitor, an OX40 inhibitor, a CD47 inhibitor, a CD276 inhibitor, a CD27 inhibitor, a GITR (Glucocorticoid-induced tumor necrosis factor receptor-related protein) inhibitor, a TIGIT (T-cell innunoreceptor with immunoglobulin and ITIM domain) inhibitor, and an IL-6 (interleukin-6) inhibitor.

12. In paragraph 11, A method wherein the above immune checkpoint inhibitor is a TIGIT inhibitor or an IL-6 inhibitor.

13. In paragraph 9, A method wherein the cancer is breast cancer or pancreatic cancer.

14. In paragraph 9, The above method, AKR1C1 in the above sample + A method for predicting that the immune anticancer treatment effect of a cancer patient will be good when the number of inflammatory fibroblasts or mesothelial cell-derived fibroblasts is greater than that of a control group.

15. In paragraph 14, A method wherein the above-mentioned immunotherapy agent is a TIGIT inhibitor or an IL-6 inhibitor.

16. In paragraph 9, A method wherein the biological sample comprises a tumor microenvironment.