A single-cell transcriptomic study reveals immune suppressive cancer cell-immune cell interactions in the triple negative canine breast cancers

doi:10.21203/rs.3.rs-3246929/v1

Download PDF

Article

A single-cell transcriptomic study reveals immune suppressive cancer cell-immune cell interactions in the triple negative canine breast cancers

https://doi.org/10.21203/rs.3.rs-3246929/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Clinical trials show promising outcomes for dogs with advanced solid tumors following treatment with immune checkpoint inhibitors (ICIs). Triple-negative breast cancer (TNBC) is very aggressive with very low response rates to ICIs. No study defines how canine TNBC interacts with the immune system within the tumor microenvironment, which is investigated in this study at the single cell level. Single cell RNA sequencing (scRNA-seq) datasets, including 6 groups of 30 dogs, were subject to integrated bioinformatic analysis. Immune modulatory TNBC subsets were identified by functional enrichment with immune-suppressive gene sets, including anti-inflammatory and M2-like macrophages. Key genes and immune-suppressive signaling pathways for TNBC included angiogenesis and leukocyte chemotaxis. Interactome analysis identified significant interactions between distinct subsets of cancer cells and effector T cells, suggesting T cell suppression. This is the first study to define immune-suppressive cancer cell subsets at the single-cell level, revealing potential mechanisms by which TNBC induces immune evasion in dogs.

Biological sciences/Cancer/Cancer microenvironment

Biological sciences/Immunology/Immune evasion

scRNA-seq

dog

immune checkpoint

triple-negative breast cancer

interactome

Immunotherapy has revolutionized cancer therapy and changed the paradigm of cancer treatment in many types of human cancers ¹. Among different types of immunotherapies, ICIs are the most prescribed cancer immunotherapy ¹. The major action of ICIs is to either rejuvenate pre-existing exhausted CD8⁺ T cells or replace old CD8⁺ T cell clones with new clones from blood or adjacent normal tissues ². ICIs have shown objective response rates of 15–45% depending on the type of cancer ³. ICIs have demonstrated effective and durable responses and become the standard of care for various malignancies ⁴.

Recently, clinical trials using dogs have shown 7.7% ⁵ and 9.5% ^6,7 of complete remission against advanced oral malignant melanoma after administration of caninized anti-PD1 and anti-PD-L1 antibodies. In addition, anti-human CCR4 antibodies targeting tumor-infiltrating regulatory T cells elicited 30% ⁸ and 71% ⁹ partial remission rates in canine models of prostate and bladder cancers, respectively. ICIs in canines are durable, although there is one report of an adverse-event-related fatality ⁷. Multiple mechanisms are assumed to be involved in the immune suppressive tumor microenvironment in dogs, including infiltration of immune suppressive cells ^10,11 and expression of immune checkpoints that suppress T cell activity ^{5–7,12–14}. Different types of immune modulatory therapies applied to dogs have shown that the tumor immune environment (TiME) among the cancer-immune cycle is the rate-limiting step for immune evasion and effective cancer immunotherapy ^{5–7,12,15,16}. However, little information is available regarding the composition of the tumor microenvironment, its interaction with the tumor in the context of its microenvironment, and tumor antigens in dogs ¹⁷. Meanwhile, there is limited data on the efficacy of adoptive T-cell therapy in dogs ^16,18.

Mammary gland tumors (MGT) are the most frequently diagnosed tumors in female dogs. Among MGT subtypes, TNBCs are characterized by HER2-, ER- and PR-negative with or without basal-like (cytokeratin 5/6-positive) type ¹⁹. TNBCs are heterogeneous ²⁰ and represent a promising model system for comparative immunotherapy research in both canine and humans ²¹.

scRNA-seq has just begun to unlock the secrets of veterinary diseases. Currently, there is no scRNA-seq analysis to study tumor-infiltrating immune cells in dogs. In this study, we utilized scRNA-seq studies to attest to our hypothesis that an immune modulatory subset of TNBCs is responsible for immune suppression via the interaction with effector T cells. As the first step toward cancer immunotherapy, we defined immune suppressive subsets of TNBC at the molecular level and characterized the crosstalk between cancer cells and effector CD4⁺ and CD8⁺ T cells. Our data indicate potential mechanisms by which TNBCs shape the immune suppressive tumor immune microenvironment (TiME). Our integrated scRNA-seq analysis lays out the groundwork for methodology development to study complex cell-cell interactions in the TiME of different cancer types or other immune-related syndromes in dogs.

scRNA-seq datasets

scRNA-seq datasets in dogs are available in Gene Expression Omnibus (GEO) database and published papers, including a total of 30 scRNA-seq datasets: 10 datasets of peripheral blood mononuclear cells (PBMC) from dogs with or without atopic dermatitis (GSE144730) ²², 3 datasets of peripheral blood T cell receptor (TCR) αβ T cells (PBT) from healthy dogs (GSE218355) ²³, 4 datasets of primary triple-negative breast cancer cells (TNBC) with or without in vitro vaccinia virus infection (GSE142184) ²⁴, a dataset of nuclei from lung tissue from a healthy dog (GSE183300) ²⁵, 4 datasets of immune cells from bronchoalveolar lavage (BAL) (E-MTAB-9265) from dogs ²⁶, and 8 datasets of immune cells from BAL from dogs with or without idiopathic pulmonary fibrosis (IPF) (E-MTAB-9263) ²⁷. These datasets were integrated and subject to bioinformatic analysis (Fig. 1). Each individual dataset was confirmed to have canine official gene symbols that are associated human homologs. All studies here used CanFam3.1 (Canis lupus familiaris genome assembly) for alignment of reads to the reference genome, except for a dataset of nuclei from lung tissue that did not specify the canine genome assembly.

scRNA-seq data integration and analysis

Seurat objects of all 30 samples were merged and integrated into one master object, as previously described ^28,29. R toolkit Seurat (v. 4.3.0) was used for the data processing, generating the Seurat object as an input file on RStudio (v. 4.2) for subsequent bioinformatic processes. Briefly, low-quality cells with either unique feature counts of less than 200 or over 5,000 or mitochondrial counts of more than 10% were filtered out. Samples were normalized with the default setting. Preparation for integration used 3,000 anchor features. Principal component analysis (PCA) was used for linear dimensional reduction. Principal components 1 through 30 were utilized for further dimensional reduction, which was based on the most significant principal component (P < 1E-5) from the Jackstraw substitution test algorithm and the ranking of principal components based on the percentage of variance. The t-distributed stochastic neighbor embedding (t-SNE) was used for graph-based clustering with a resolution of 2.7.

scDblFinder (v. 1.4.0) R package was used to remove potential doublets, as previously described ²⁸. Doublet prediction was run on each study group to remove batch effects. Following singlet selection, single-cell clusters were identified and labeled based on the investigation of markers derived from the original studies where the dataset was driven, canonical markers for lineage, markers for rare and unique populations from previous publications, or SingleR (v. 1.8.1)-based unbiased cell type recognition ³⁰. The celldex package (v. 1.6.0) was used to leverage reference signatures of pure cell types to infer the cell of origin of every single cell.

Differential gene expression analysis

The likelihood-ratio test was used to find the differential expression for a single cluster, compared to all other cells. To identify cluster markers, the “FindAllMarkers” function was used in the Seurat package with the absolute log₂-fold change threshold > 0.25 and minimum percentage of cells where the gene is detected in a specific cluster > 25%. To identify cluster differentially expressed genes (DEGs) for all clusters across groups, the “FindMarkers” function was used with the absolute log₂-fold change threshold > 0.36 and P value < 0.05.

Gene set enrichment analysis (GSEA) and Gene Ontology analysis

Single-cell GSEA was performed using the escape R package (v. 1.6.0), as previously described ²⁸. Gene sets were derived from the Hallmark library of the Molecular Signature Database. Canine gene sets associated with cancer types were derived from a previous publication ^31–34. DEGs were also subjected to either Gene Ontology (GO) enrichment analysis using PANTHER annotation datasets with a species of Canis lupus familiaris (http://geneontology.org/) or ShinyGO (v. 0.77), a graphical gene-set enrichment tool with a species of dog. GO or Kyoto Encyclopedia of Genes and Genomes (KEGG) results were filtered based on the criteria of P value < 0.05 and false discovery rate (FDR) < 0.05. DittoSeq (v. 1.4.4) and pheatmap (v. 1.0.12) R packages were used to visualize gene sets defining specific molecular and biological pathways.

Cell cycle analysis

Cell cycle assignment was performed by using “CellCycleScoring” function and calling “cc.genes.updated.2019” in Seurat.

Cell-to-cell interaction analysis

CellChat R package (v. 1.4.0) was used to quantitatively infer intercellular communication networks from scRNA-seq data ³⁵. Single cells derived from PBMC, αβ T, and TNBC groups were subject to interactome analysis. PTPRC⁻ non-immune cells, potential doublets, and clusters that were simultaneously assigned to TNBC and immune cell groups, were not included in the interaction analysis. To find potential ligand-receptor pair, the “netVisual_bubble” function was used with a threshold of P value < 0.01.

Data availability

The raw data from scRNA-seq in this study was publicly available as described above with the accession numbers (GSE144730, GSE218355, GSE142184, GSE183300, E-MTAB-9265, and E-MTAB-9263).

Statistical methods

Default statistical methods available within the Seurat package were used in this study. Non-parametric Wilcoxon rank-sum test was used to compare the significance of two-sample differential expression in the “FindAllMarkers” function. A two-tailed unpaired Student’s t-test available within the ggpubr R package (v. 0.4.0) was used for statistical tests for the distribution of genes on count-level mRNA data.

Standard pre-processing and quality control of the integrated scRNA-seq data. The study workflow is presented in Fig. 1A. The standard pre-processing and rigorous quality control of scRNA-seq data integrated by studies are available (Supplementary Fig. 1). The number of genes, percentage of reads that map to the mitochondrial genome, and percentage of canine Ensemble genes detected in each study are shown (Supplementary Fig. 1A). Mitochondrial genes were not identified in BAL and lung groups (Supplementary Fig. 1B). Representative top 10 highest variable features among a total of 3,000 features that exhibit high cell-to-cell variation in the integrated scRNA-seq dataset are shown (Supplementary Fig. 1B). Principal components of 20 showing strong enrichment of features with low P values were selected based on the JackStraw (Supplementary Fig. 1C) and Elbow (Supplementary Fig. 1D) plots. Following potential doublet exclusion (Supplementary Fig. 1E), 69,035 single cells were obtained from PBMC (GSE144730, n = 20,078), peripheral blood TCR αβ T cells (GSE218355, n = 19,796), lung (GSE183300, n = 3,694), BAL (E-MTAB-9265, n = 4,240), BAL (E-MTAB-9623, n = 16,171), and TNBC (GSE142184, n = 5,056) (Supplementary Fig. 1F).

Identification of the major single-cell clusters by scRNA-seq

Following scRNA-seq data integration, a total of 45 clusters were identified in our master Seurat object (Fig. 1B). We assessed the clustering performance by identifying the major cell types and found a clear separation of CD45⁺ (or PTPRC⁺) leukocytes, CD4⁺ T, CD8⁺ T (or CD8A⁺), myeloid (LYZ⁺), epithelial cells (EPCAM⁺ and/or COL1A2⁺), and lung (BMP1⁺) populations (Figs. 1C-1D). Out of the top 20 DEGs to define the major immune and non-immune subsets, representative genes are presented on the heatmap (Fig. 1E). Functionally unique immune subsets from PBMC, αβ T, and BAL groups were further defined by various canonical markers and genes derived from previous studies (Fig. 1F and Supplementary Fig. 2). Unbiased cell type recognition was also used to define distinct clusters, supporting our clustering performance (Supplementary Fig. 3).

A total of 16 clusters of CD3⁺ T cell were identified, which mainly consisted of single-positive CD4⁺ or CD8⁺ cells, as well as a few double-positive (clusters 11 and 25) and double-negative (clusters 7 and 22, largely derived from BAL groups) subpopulations (Supplementary Fig. 2A). CD4⁺ T cells were further classified into 5 transcriptionally unique subpopulations, including LEF1⁺SELL⁺CCR7⁺ naïve (clusters 0 and 2), CXCR3^high pre-effector (cluster 8), GATA3⁺ Th2-like pre-effector (cluster 9), CCR4^highCXCR4^high effector memory (cluster 17), FOXP3⁺ regulatory T cells (Treg) (cluster 31), and uncharacterized (cluster 38). Pre-effector CD4⁺ T cells were defined by the increasing tendency of HOPX expression – a marker for pre-effector T cells destined for subsequent effector differentiation ³⁶. A total of 5 CD8⁺ T clusters were identified. Genes encoding killer cell lectin-like receptors, such as KLRG1 and KLRB1, and cytotoxicity, such as GZMA, PRF1, and NCR3, were highly expressed in clusters 14 and 26, suggesting cytotoxic CD8⁺ T cells (Supplementary Fig. 2A). In addition, cluster 26 specifically expressed FCER1G, a marker of an innate-like phenotype ³⁷. Cluster 24 resembled an exhausted progenitor phenotype based on CD7, TOX, and to some extent TCF7, but not CD27 expression. Cluster 35 showed high expression of genes associated with T cell activation, such as CD69, JUND, and KLF6, suggesting activated CD8⁺ T cells. Cluster 30 expressed HOPX, but less expression of cytotoxicity-related genes, thus considered pre-effector type. Cluster 18 was characterized by high CCL4, CCL5, and S100A4 expression, suggesting a migratory phenotype. Meanwhile, only a very small number of immune cells, such as gamma delta T cells, and natural killer (NK) cells, were identified based on unbiased cell type annotation, but not assigned to a single independent cluster (Supplementary Fig. 3).

Myeloid cells that were defined by LYZ expression (Supplementary Fig. 2B) were distinguishably separated on the tSNE plot by the expression of CSF3R and CD163 – the classic M1 and M2 macrophage polarization markers, respectively (Fig. 1G). Myeloid clusters expressing CSF3R were ITGB2^high monocytes (clusters 3, 6, 10, 12, 16, and 20), mainly derived from peripheral blood (Supplementary Fig. 3B). Myeloid clusters expressing CD163 were mainly CD68^high macrophages (clusters 1, 4, 19, 21, and 39), largely derived from BAL. Cluster 20 was further defined by IFN-related monocytes based on the high expression of genes associated with interferon signaling pathways, such as ISG15, MX1, and MX2. We also identified CD83⁺CD86⁺ITGAX⁺ dendritic cells (DC) (cluster 23), FSCN1⁺ DC (cluster 41), and CD177⁺ neutrophils (cluster 40).

B cells (cluster 34) and plasmablasts (clusters 28 and 33) specifically expressed MS4A1 and IRF4, respectively (Supplementary Figs. 2A, 3A-3B). Single cells derived from lung were not immune cells but highly expressed BMP1 gene (Fig. 1D, Supplementary Fig. 2C). Single cells from the TNBC group showed specific expression of epithelial cell markers, such as COL1A2, KRT14, CA2, and SPP1 (Supplementary Fig. 2C). Meanwhile, some immune and TNBC cells were assigned to the same clusters (29, 36, and 43), perhaps due to a similar global structure of RNA expression across single cells. Taken together, our integrated analysis successfully identified major distinct subsets of immune and TNBC cells at the single-cell level in dogs.

Subsets of cancer cells have a distinct immune-suppressive phenotype

Clinical trials have revealed the existence of an immune suppressive TiME within multiple types of canine cancers ^8,9,38,39. Whether canine TNBCs have an immune suppressive TiME is still unknown. We performed GSEA using various gene sets associated with immune-associated pathways (Fig. 2). Overall, we identified two distinct patterns of gene set enrichment. First, PBMC was preferentially enriched with inflammatory gene sets, such as interferon signaling, M1 macrophage, proinflammatory, and leukocyte-mediated immunity (Fig. 2A, red boxed). Especially, αβ T cells within the PBMC showed marked enrichment of T cell-specific gene signatures, such as Treg and terminal T cell differentiation. Second, there was preferential enrichment of immune suppressive gene sets within TNBCs, such as those related to TGF-β, TNF-α, anergy, anti-inflammatory, M2 macrophage, and exhaustion (Fig. 2A, blue boxed). Among them, the enrichment pattern of anti-inflammatory and M2 macrophage was particularly specific to TNBC, compared to other groups (Fig. 2B). TNBCs were also enriched with gene sets associated with alternative metabolic pathways, oxidative stress, and importantly canine gene sets (Fig. 2C), showing the reliability of GSEA implemented in this study. Consistent with previous scRNA-seq findings in dogs ²⁷, BAL affected by idiopathic pulmonary fibrosis contained single-cell clusters that were highly enriched with the M2 macrophage gene set (Figs. 2A-2B). Meanwhile, lung cells showed only sporadic enrichment patterns of a few gene sets, such as TGF-β and anergy (Figs. 2A-2B). There was no noticeable difference in the enrichment pattern between healthy and atopic dermatitis conditions of PBMC.

Taken together, GSEA confirmed that TNBC contributes to immune suppressive TiME in dogs, providing a rationale for further analysis with a focus on identifying specific TBNC clusters that might have led to the distinct immune-suppressive phenotype.

Identification and characterization of cancer cell subsets within TNBCs

To scrutinize functionally unique cancer cell subpopulations, we performed sub-clustering of all cancer cells and defined a total of 11 sub-clusters (Fig. 3A). Each cluster was clearly separated in the tSNE plot, demonstrating prominent intratumoral heterogeneity and distinct global structure of transcriptomes across clusters (Supplementary Fig. 4). Representative markers used to define each cluster are available (Fig. 3B, Supplementary Figs. 4A-4B, and Supplementary Table 1). Genes encoding oncolytic vaccinia viral proteins, such as A12L and A7L, were specifically expressed in cluster 4, indicative of viral infection. Expression of genes associated with immune suppression, such as SPP1 and HMGA1, was mainly identified in clusters 2, 3, 5, 6, and 7. Clusters 0 and 1 were characterized by specific expressions of SFRP2 and COL2A1, respectively. DDIT4 – which has been reported to confer resistance canine TNBC against viral infection ²⁴ – was expressed in clusters 0, 1, 8, and 9.

We next performed functional enrichment analysis to investigate cancer cell subsets that can potentially contribute to immune suppression. Interestingly, GSEA identified clusters 2, 3, 5, 6, and 7 that were preferentially enriched with gene sets associated with immune suppression, such as anti-inflammatory, M2 macrophage, anergy, IL-18, TNF-α, and TGF-β (Figs. 3C, blue boxed, 3D, and Supplementary Fig. 4C). Interestingly, enrichment with an anti-inflammatory signature tended to increase in virus-infected clusters relative to non-viral infected cancer cell clusters (Fig. 3D). Among them, clusters 2 and 6 showed simultaneous enrichment of anti-inflammatory and M2 macrophage gene sets, indicative of potential immune-suppressive cancer cells (Fig. 3E). KEGG analysis using cluster markers revealed that cluster 2 was significantly associated with extracellular matrix-receptor interaction (Supplementary Fig. 4D). Cluster 6 was significantly associated with blood vessel development and cell migration (Supplementary Fig. 4D). These enrichment patterns were not associated with cell cycle alteration (Fig. 3E). Viral infection tended to arrest cell cycle progression (Supplementary Fig. 4E). Based on our GSEA and expression analyses, we classified cancer cells (Fig. 3G) into virus-susceptible, virus-resistant, and immune-modulatory – which was originally defined as bystander cancer cells in a previous study ²⁴ (Fig. 3F). During virus infection, bystander cancer cells can change their behavior, potentially favoring immune evasion ⁴⁰. The increasing tendency of anti-inflammatory enrichment in virus-infected, particularly immune-modulatory cancer cells, might imply genes that can play a key role in shaping an immune suppressive TiME. Among genes belonging to the immune suppressive signatures, we identified 40 DEGs across cancer cell subsets related to different viral infectious status (Fig. 3H, Supplementary Table 2, and Supplementary Fig. 4F). Overall, these genes were significantly enriched with many biological processes, particularly angiogenesis and leukocyte chemotaxis (Fig. 3I, Supplementary Figs. 4F-4G, and Supplementary Table 3). In support of this, VEGFA that belonged to these GO terms, significantly increased in virus-infected cancer cell clusters, compared to non-infected clusters (Fig. 3J and Supplementary Fig. 4G). Interestingly, VEGFC, which is associated with immune suppression during canine mammary cancer development ⁴¹, was also significantly upregulated in a subset of cancer cell population (Supplementary Fig. 4G). Other potential immune modulatory candidate genes, such as HSD11B1, LIF, PTGS2, GADD45B, and JAG1, were present (Fig. 3J).

Identification and characterization of cell-to-cell interaction between cancer cells and PBMCs

Tumor-immune cell interaction – a hallmark of cancer immunology – plays a critical role in T cell exhaustion within the TiME, leading to ineffective cancer immunotherapy. As a potential mechanism by which cancer cells induce the exhaustion stage of tumor-infiltrating T cells, we hypothesize that peripheral T cells interact with cancer cells via distinct ligand-receptor pairs. To decipher coordinated tumor-immune interactions, single cell clusters from cancer cells and PBMC groups were subject to CellChat analysis (Supplementary Fig. 5A). While analyzing the interactome, we found two major patterns of cell-cell interaction. First, there were distinct subsets of cancer cells (clusters 13, 27, and 32) that strongly sent signals to other cells in a paracrine fashion or to themselves via an autocrine pathway (Figs. 4A-4B, blue boxed). Second, effector CD4⁺ (cluster 8 pre-effector CXCR3^high, cluster 17 effector memory, and cluster 31 Treg) and CD8⁺ (cluster 14 cytotoxic and cluster 26 innate-like cytotoxic) T cells were able to receive signals (Figs. 4A-4B, red boxed, and Supplementary Fig. 5B). We next focused on these clusters to identify significant signaling pathways and ligand and receptor pairs (Fig. 4C).

After extensive interactome analysis, we ended up identifying key ligand and receptor pairs, which were mainly derived from secretory (SPP1), cell-cell intact (APP), and extracellular matrix-receptor (FN1 and Collagen) pathways (Fig. 4D and Supplementary Fig. 5C). Interestingly, T cell receptors CD44 and CD74 showed high communication probability, contributing most to the outgoing signaling of the representative ligands from cancer cells (Supplementary Fig. 5D). The ligand and receptor genes showed high expression levels in the interacting cancer cells and T clusters (Fig. 4E). Finally, ligands from cancer cells were also expressed in immune modulatory cancer cell subpopulations (clusters 2, 3, 5, 6, and 7) (Fig. 4F), suggesting potential immune suppressive mechanisms mediated by cancer cell-T interaction.

In the present study, we took advantage of utilizing open-source canine scRNA-seq datasets, including primary TNBC and peripheral immune cells, and investigated the mechanism by which TNBC induces immune suppression in dogs. Our integrated scRNA-seq analysis for the first time reveals that immune suppressive subsets of cancer cells, which were identified to preferentially interact with effector types of CD4⁺ and CD8⁺ T cells in dogs. Potential mechanisms by which cancer cells shape the immune suppressive TiME are suggested to regulate angiogenesis and immune cell infiltration.

Leveraging the integrated scRNA-seq analysis, we were able to specify transcriptionally distinct cancer cell subsets with potentially different cancer immunity and their interaction with T cells, which cannot be discovered or addressed by bulk RNA-seq ⁴². In the present study, immune modulatory cancer cell subsets were defined by unique cluster markers, such as HMGA1, SPP1, and WNT5A. Interestingly, these genes have been closely involved in the development and potential immune evasion of canine ^43,44 and human TNBC ^45–48. For example, HMGA1 – a downstream gene of PD-L1 that regulates cancer immunity in human TNBC ⁴⁹ – was exclusively expressed in immune-modulatory TNBC subsets. Wnt singling triggered by WNT5A promotes the immune escape of TNBC ⁵⁰. Likewise, SPP1 expressed on malignant cells contributes to the suppression of T cells by CD44-mediated binding ⁵¹. Thus, we suggest that HMGA1⁺SPP1⁺ or WNT5A⁺ TNBC cells might elicit immune evasion and thus be a potential target for anticancer immunity. Functional enrichment of immune suppressive gene sets in these TNBC subpopulations also supports our notion.

Immunosuppressive pathways could play a prominent role in the resistance of tumor cells to oncolytic viral infection ⁴⁰. Accordingly, we leveraged oncolytic viral infection as an anti-inflammatory factor and identified candidate genes for immune evasion and mechanisms by which TNBC escapes immune surveillance by the host. GO analysis using DEGs in cells infected by the virus identified two representative signaling pathways: angiogenesis and leukocyte chemotaxis. Angiogenesis is a hallmark of canine mammary gland tumorigenesis, and VEGF signaling is critical for the pathophysiology of canine TNBC ⁵². In addition, angiogenesis is significantly correlated with immune suppression in dogs ⁴¹. In this study, immune modulatory TNBC subsets affected by viral infection significantly upregulated VEGFA and VEGFC. Angiogenesis induces infiltration of various types of immune suppressive cells to MGT in dogs ⁵³. Indeed, VEGFC released by canine MGT contributes to immune suppression via the recruitment of Treg and myeloid-derived suppressive cells (MDSCs) ⁴¹. Infiltration of Tregs ⁵⁴ and tumor-associated macrophages (TAMs) ⁵⁵ is promoted by VEGF signaling in canine MGT. The recruited immune suppressive cells have been strongly suggested to inhibit anti-cancer T cell activity, leading to poor prognosis in canine MGT and TNBC ^41,53,56. Thus, based on previous findings and our results, we postulate that TNBC mainly elicits cancer immunity via angiogenesis and VEGF-mediated immune cell infiltration. Future work is warranted to elucidate the mechanism by which canine TNBC modulates cancer immunity by regulating other candidate genes. For example, in this study cancer cell themselves within TNBC were strongly inferred to specifically interact each other, suggesting the potential autocrine and paracrine communications. Extrapolating from this, we propose that the candidate genes might constitute to a positive feedback loop to amplify anti-inflammatory singling pathways of cancer cells. Indeed, oncolytic virus-regulated candidate genes, such as HSD11B1, LIF, PTGS2, GADD45B, and JAG1, have been involved in autocrine and/or paracrine signaling in TiME of human TNBC ^57–59. Accordingly, exploiting therapeutic strategies for abrogating the intertumoral feedback loop, e.g., novel therapeutic modalities, such as proteolysis-targeting chimera ⁶⁰, might be effective to normalize TiME.

Our interactome analysis reveals that cancer cells directly modulate T cell activity, potentially favoring immune suppression. In this study, we suggest that SPP1⁺, FN1⁺, or COL1A2⁺ cancer cells might be a key subset for T cell suppression in which CD44 is likely to be the immune checkpoint in dogs. Indeed, SPP1-CD44 interaction has been demonstrated to suppress effector T cell activity infiltrated into multiple types of cancers ^51,61. Binding of CD44⁺ tumor-infiltrating T cells to type I collagen induces a more aggressive phenotype of malignant melanoma ⁶². In addition, FN1⁺ TNBC cells are positively associated with CD8⁺ T cell infiltration and immune suppression ⁶³. Although little information regarding interaction is available in dogs, substantial studies have supported the anti-inflammatory role of SPP1, FN1, and COL1A2 in canine MGT and TNBC ^42,43,64. Meanwhile, it might be also interesting to investigate the impact of TNBC-Tregs interaction. It might be associated with tumor-mediated direct induction of Tregs ⁶⁵, supporting CD4⁺ T cell-mediated poor prognosis of canine mammary carcinoma ⁶⁶ and TNBC ⁵⁶. Future studies are warranted to demonstrate the clinical relevance of the binding of TNBC ligands to CD44⁺ T cells on the immune suppression of canine TNBC.

The present study has limitations. Despite the well-established integrating methodology provided by Seurat, potential institutional or batch effects across scRNA-seq datasets might have occurred during analysis. Currently, a functionally validated canine gene set database is absent. Gene sets that consist of canine official gene symbol that has the associated human homolog are subjected to functional enrichment analysis to predict immunological functions. In other words, most canine Ensemble genes cannot be included in the bioinformatic analysis in this study. Although humans and dogs share a high degree of homology with the corresponding human sequences and orthologous genes, especially showing well-conserved interspecies immunological functions ⁶⁷, a more accurate assessment of immune-related functions should be made by using canine gene sets. Genomic annotation of the newly released dog reference genome CanFam6 will provide a more robust and high-resolution transcriptomic analysis, compared to CanFam3.1 ⁶⁸. We did not analyze tumor-infiltrating immune cells, but peripheral ones. Although PBMCs are shown to infiltrate tumor sites, interact with tumor cells, and become exhausted by the immune-tumor interaction, future studies need to prove it in dogs, which is currently an ongoing experiment of our group to identify novel immune checkpoints.

The present study reveals that triple-negative canine breast cancer shapes immune-suppressive tumor microenvironment which is mediated by immune-TNBC interaction mainly affected by exhausted CD44⁺ effector CD4⁺ T cells.

Acknowledgment

We appreciate the University of Florida High-Performance Computing Center for performing HiPerGator 3.0 supercomputer, which serves as the primary working space for integrating multiple scRNA-seq datasets in this project. We thank Weizhou Zhang for sponsoring Myung-Chul to operate HiPerGator.

Conflict of interest

The authors declare no conflicts of interest.

Funding

This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (#RS-2023-00241779).

Author’s contribution

Conception and design: M.K.

Development of methodology: M.K.

Data acquisition: M.K., N.B.

Bioinformatics: M.K., N.B.

Analysis and interpretation of data: M.K., W.S., R.K., N.B., W.Z.

Writing, review, and/or revision of the manuscript: M.K., R.K., W.S., N.B., W.Z.

Study supervision: M.K.

Waldman, A. D., Fritz, J. M. & Lenardo, M. J. A guide to cancer immunotherapy: from T cell basic science to clinical practice. Nat. Rev. Immunol. 20, 651–668 (2020).
Kim, M.-C. et al. Updates on immunotherapy and immune landscape in renal clear cell carcinoma. Cancers (Basel) 13, (2021).
Galluzzi, L., Humeau, J., Buqué, A., Zitvogel, L. & Kroemer, G. Immunostimulation with chemotherapy in the era of immune checkpoint inhibitors. Nat. Rev. Clin. Oncol. 17, 725–741 (2020).
Yang, Y. Cancer immunotherapy: harnessing the immune system to battle cancer. J. Clin. Invest. 125, 3335–3337 (2015).
Maekawa, N. et al. PD-L1 immunohistochemistry for canine cancers and clinical benefit of anti-PD-L1 antibody in dogs with pulmonary metastatic oral malignant melanoma. NPJ Precis. Oncol. 5, 10 (2021).
Igase, M. et al. Long-term survival of dogs with stage 4 oral malignant melanoma treated with anti-canine PD-1 therapeutic antibody: A follow-up case report. Vet. Comp. Oncol. 20, 901–905 (2022).
Igase, M. et al. A pilot clinical study of the therapeutic antibody against canine PD-1 for advanced spontaneous cancers in dogs. Sci. Rep. 10, 18311 (2020).
Maeda, S. et al. Anti-CCR4 treatment depletes regulatory T cells and leads to clinical activity in a canine model of advanced prostate cancer. J. Immunother. Cancer 10, (2022).
Maeda, S., Murakami, K., Inoue, A., Yonezawa, T. & Matsuki, N. CCR4 blockade depletes regulatory T cells and prolongs survival in a canine model of bladder cancer. Cancer Immunol. Res. 7, 1175–1187 (2019).
Kim, J. H. et al. Correlation of Foxp3 positive regulatory T cells with prognostic factors in canine mammary carcinomas. Vet. J. 193, 222–227 (2012).
Gulay, K. C. M. et al. Hemangiosarcoma cells induce M2 polarization and PD-L1 expression in macrophages. Sci. Rep. 12, 2124 (2022).
Maekawa, N. et al. A canine chimeric monoclonal antibody targeting PD-L1 and its clinical efficacy in canine oral malignant melanoma or undifferentiated sarcoma. Sci. Rep. 7, 8951 (2017).
Maekawa, N. et al. Immunohistochemical Analysis of PD-L1 Expression in Canine Malignant Cancers and PD-1 Expression on Lymphocytes in Canine Oral Melanoma. PLoS ONE 11, e0157176 (2016).
Cascio, M. J. et al. Canine osteosarcoma checkpoint expression correlates with metastasis and T-cell infiltrate. Vet. Immunol. Immunopathol. 232, 110169 (2021).
Rebhun, R. B. et al. Inhaled recombinant human IL-15 in dogs with naturally occurring pulmonary metastases from osteosarcoma or melanoma: a phase 1 study of clinical activity and correlates of response. J. Immunother. Cancer 10, (2022).
Panjwani, M. K. et al. Establishing a model system for evaluating CAR T cell therapy using dogs with spontaneous diffuse large B cell lymphoma. Oncoimmunology 9, 1676615 (2020).
Klingemann, H. Immunotherapy for dogs: still running behind humans. Front. Immunol. 12, 665784 (2021).
Panjwani, M. K. et al. Feasibility and Safety of RNA-transfected CD20-specific Chimeric Antigen Receptor T Cells in Dogs with Spontaneous B Cell Lymphoma. Mol. Ther. 24, 1602–1614 (2016).
Abadie, J. et al. Canine invasive mammary carcinomas as models of human breast cancer. Part 2: immunophenotypes and prognostic significance. Breast Cancer Res. Treat. 167, 459–468 (2018).
Im, K. S. et al. Analysis of a new histological and molecular-based classification of canine mammary neoplasia. Vet. Pathol. 51, 549–559 (2014).
Machado, M. C. A. et al. Triple-negative mammary carcinoma in two male dogs. J. Vet. Diagn. Invest. 32, 94–98 (2020).
Früh, S. P. et al. Elevated circulating Th2 but not group 2 innate lymphoid cell responses characterize canine atopic dermatitis. Vet. Immunol. Immunopathol. 221, 110015 (2020).
Eschke, M., Moore, P. F., Chang, H., Alber, G. & Keller, S. M. Canine peripheral blood TCRαβ T cell atlas: Identification of diverse subsets including CD8A + MAIT-like cells by combined single-cell transcriptome and V(D)J repertoire analysis. Front. Immunol. 14, 1123366 (2023).
Cambien, B. et al. Identification of oncolytic vaccinia restriction factors in canine high-grade mammary tumor cells using single-cell transcriptomics. PLoS Pathog. 16, e1008660 (2020).
Chen, D. et al. Single cell atlas for 11 non-model mammals, reptiles and birds. Nat. Commun. 12, 7083 (2021).
Fastrès, A. et al. Characterization of the bronchoalveolar lavage fluid by single cell gene expression analysis in healthy dogs: A promising technique. Front. Immunol. 11, 1707 (2020).
Fastrès, A. et al. Identification of Pro-Fibrotic Macrophage Populations by Single-Cell Transcriptomic Analysis in West Highland White Terriers Affected With Canine Idiopathic Pulmonary Fibrosis. Front. Immunol. 11, 611749 (2020).
Borcherding, N. et al. Mapping the immune environment in clear cell renal carcinoma by single-cell genomics. Commun. Biol. 4, 122 (2021).
Kim, M.-C. et al. CD177 modulates the function and homeostasis of tumor-infiltrating regulatory T cells. Nat. Commun. 12, 5764 (2021).
Aran, D. et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat. Immunol. 20, 163–172 (2019).
Sheet, S., Krishnamoorthy, S., Cha, J., Choi, S. & Choi, B.-H. Identification of Candidate Genes and Pathways Associated with Obesity-Related Traits in Canines via Gene-Set Enrichment and Pathway-Based GWAS Analysis. Animals (Basel) 10, (2020).
Lee, K.-H., Park, H.-M., Son, K.-H., Shin, T.-J. & Cho, J.-Y. Transcriptome signatures of canine mammary gland tumors and its comparison to human breast cancers. Cancers (Basel) 10, (2018).
Tawa, G. J. et al. Transcriptomic profiling in canines and humans reveals cancer specific gene modules and biological mechanisms common to both species. PLoS Comput. Biol. 17, e1009450 (2021).
Graim, K. et al. Modeling molecular development of breast cancer in canine mammary tumors. Genome Res. 31, 337–347 (2020).
Jin, S. et al. Inference and analysis of cell-cell communication using CellChat. Nat. Commun. 12, 1088 (2021).
Bourque, J., Kousnetsov, R. & Hawiger, D. Roles of Hopx in the differentiation and functions of immune cells. Eur. J. Cell Biol. 101, 151242 (2022).
Morrish, E. & Ruland, J. Cytotoxic FCER1G + innate-like T cells: new potential for tumour immunotherapy. Signal Transduct. Target. Ther. 7, 204 (2022).
Maeda, S., Ohno, K., Fujiwara-Igarashi, A., Uchida, K. & Tsujimoto, H. Changes in Foxp3-Positive Regulatory T Cell Number in the Intestine of Dogs With Idiopathic Inflammatory Bowel Disease and Intestinal Lymphoma. Vet. Pathol. 53, 102–112 (2016).
Maeda, S. et al. Foxp3 + regulatory T cells associated with CCL17/CCR4 expression in carcinomas of dogs. Vet. Pathol. 57, 497–506 (2020).
Nguyen, T. T. et al. Reshaping the tumor microenvironment with oncolytic viruses, positive regulation of the immune synapse, and blockade of the immunosuppressive oncometabolic circuitry. J. Immunother. Cancer 10, (2022).
Mucha, J. et al. Immunosuppression in dogs during mammary cancer development. Vet. Pathol. 53, 1147–1153 (2016).
Hussain, S. et al. Gene expression profiling of spontaneously occurring canine mammary tumours: Insight into gene networks and pathways linked to cancer pathogenesis. PLoS ONE 13, e0208656 (2018).
Mohammed, S. I. et al. Ductal carcinoma in situ progression in dog model of breast cancer. Cancers (Basel) 12, (2020).
Beetch, M. et al. DNA methylation landscape of triple-negative ductal carcinoma in situ (DCIS) progressing to the invasive stage in canine breast cancer. Sci. Rep. 10, 2415 (2020).
Wang, X. et al. Increased expression of osteopontin in patients with triple-negative breast cancer. Eur. J. Clin. Invest. 38, 438–446 (2008).
Bourdakou, M. M., Athanasiadis, E. I. & Spyrou, G. M. Discovering gene re-ranking efficiency and conserved gene-gene relationships derived from gene co-expression network analysis on breast cancer data. Sci. Rep. 6, 20518 (2016).
Zanin, R. et al. HMGA1 promotes breast cancer angiogenesis supporting the stability, nuclear localization and transcriptional activity of FOXM1. J. Exp. Clin. Cancer Res. 38, 313 (2019).
Chantada-Vázquez, M. D. P. et al. Protein Corona Gold Nanoparticles Fingerprinting Reveals a Profile of Blood Coagulation Proteins in the Serum of HER2-Overexpressing Breast Cancer Patients. Int. J. Mol. Sci. 21, (2020).
Chang, X. et al. Targeting HMGA1 contributes to immunotherapy in aggressive breast cancer while suppressing EMT. Biochem. Pharmacol. 212, 115582 (2023).
Merikhian, P., Eisavand, M. R. & Farahmand, L. Triple-negative breast cancer: understanding Wnt signaling in drug resistance. Cancer Cell Int. 21, 419 (2021).
Shurin, M. R. Osteopontin controls immunosuppression in the tumor microenvironment. J. Clin. Invest. 128, 5209–5212 (2018).
Amirkhani Namagerdi, A. et al. Triple-Negative Breast Cancer Comparison With Canine Mammary Tumors From Light Microscopy to Molecular Pathology. Front. Oncol. 10, 563779 (2020).
Muscatello, L. V. et al. Standardized approach for evaluating tumor infiltrating lymphocytes in canine mammary carcinoma: Spatial distribution and score as relevant features of tumor malignancy. Vet. J. 283–284, 105833 (2022).
Carvalho, M. I. et al. Intratumoral FoxP3 expression is associated with angiogenesis and prognosis in malignant canine mammary tumors. Vet. Immunol. Immunopathol. 178, 1–9 (2016).
Raposo, T. P. et al. Tumour-associated macrophages are associated with vascular endothelial growth factor expression in canine mammary tumours. Vet. Comp. Oncol. 13, 464–474 (2015).
Franzoni, M. S. et al. Tumor-infiltrating CD4 + and CD8 + lymphocytes and macrophages are associated with prognostic factors in triple-negative canine mammary complex type carcinoma. Res. Vet. Sci. 126, 29–36 (2019).
Christianson, J., Oxford, J. T. & Jorcyk, C. L. Emerging perspectives on leukemia inhibitory factor and its receptor in cancer. Front. Oncol. 11, 693724 (2021).
Somasundaram, V. et al. Systemic Nos2 Depletion and Cox inhibition limits TNBC disease progression and alters lymphoid cell spatial orientation and density. Redox Biol. 58, 102529 (2022).
Meng, J. et al. Tumor-derived Jagged1 promotes cancer progression through immune evasion. Cell Rep. 38, 110492 (2022).
Kolb, R. et al. Proteolysis-targeting chimera against BCL-XL destroys tumor-infiltrating regulatory T cells. Nat. Commun. 12, 1281 (2021).
Klement, J. D. et al. An osteopontin/CD44 immune checkpoint controls CD8 + T cell activation and tumor immune evasion. J. Clin. Invest. 128, 5549–5560 (2018).
Weimann, T. K., Wagner, C., Goos, M. & Wagner, S. N. CD44 variant isoform v10 is expressed on tumor-infiltrating lymphocytes and mediates hyaluronan-independent heterotypic cell-cell adhesion to melanoma cells. Exp. Dermatol. 12, 204–212 (2003).
Zhang, X.-X., Luo, J.-H. & Wu, L.-Q. FN1 overexpression is correlated with unfavorable prognosis and immune infiltrates in breast cancer. Front. Genet. 13, 913659 (2022).
Monteiro, L. N. et al. Osteopontin expression and its relationship with prognostic biomarkers in canine mammary carcinomas. Pesq. Vet. Bras. 40, 210–219 (2020).
Oliveira, G. et al. Landscape of helper and regulatory antitumour CD4 + T cells in melanoma. Nature 605, 532–538 (2022).
Estrela-Lima, A. et al. Immunophenotypic features of tumor infiltrating lymphocytes from mammary carcinomas in female dogs associated with prognostic factors and survival rates. BMC Cancer 10, 256 (2010).
Wang, C. et al. A novel canine reference genome resolves genomic architecture and uncovers transcript complexity. Commun. Biol. 4, 185 (2021).
Jagannathan, V. et al. Dog10K_Boxer_Tasha_1.0: A Long-Read Assembly of the Dog Reference Genome. Genes (Basel) 12, (2021).

No competing interests reported.

SupplementaryTable1.xlsx
Supplementary Table1 . List of top 20 cluster markers defining TNBC subpopulations.
SupplementrayTable2.xlsx
Supplementary Table 2. List of DEGs defining immune suppressive TNBC subsets and their enrichment gene sets.
SupplementaryTable3.xlsx
Supplementary Table 3. A list of GO biological processes using DEGs is listed in Supplementary Table 2.
SupplementaryFiguresandFigureLegend.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

A single-cell transcriptomic study reveals immune suppressive cancer cell-immune cell interactions in the triple negative canine breast cancers

Status:

Version 1

Abstract

Figures

Introduction

Materials and Methods

scRNA-seq datasets

scRNA-seq data integration and analysis

Differential gene expression analysis

Gene set enrichment analysis (GSEA) and Gene Ontology analysis

Cell cycle analysis

Cell-to-cell interaction analysis

Data availability

Statistical methods

Results

Identification of the major single-cell clusters by scRNA-seq

Subsets of cancer cells have a distinct immune-suppressive phenotype

Identification and characterization of cancer cell subsets within TNBCs

Identification and characterization of cell-to-cell interaction between cancer cells and PBMCs

Discussion

Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1