Arm-level SCNAs are associated with survival of patients with TNBC

We first assessed the association between arm-level SCNAs and overall survival of patients with TNBC to identify novel putative drivers of this aggressive form of BC. To achieve this, we curated a list of TNBC patients from the TCGA and identified 152 patients that were negative for ER/PR by immunohistochemistry (IHC), and negative for HER2 via fluorescence in situ hybridization and IHC (Fig. 1). We then performed a log-rank test to identify arm-level losses or gains significantly associated with survival. Out of all arm-level SCNAs, 21q gain and 7p gain were frequent events that significantly correlated with poor overall survival (Fig. 2A). In addition, 11q gain, 18p loss, 11q loss, and 20q loss were significantly correlated with survival, although they were much less frequent (Fig. 2A, Supplementary Table 1). Given that known oncogenes, such as EGFR, are located on the short arm of chromosome 7 [12,13,14,15], it was not surprising to find 7p gain as the most frequent SCNA that significantly associated with poor overall survival in our patient cohort (Fig. 2A, Supplementary Table S1). Interestingly, we found 21q gain was significantly associated with reduced overall survival (Fig. 2B). 21q gain has been identified to be more frequent in TNBC compared to other BC subtypes, but its association with survival has not been characterized in detail [16, 17]. To assess whether the link between 21q gain and survival in TNBC was merely a result of widespread arm-level copy number alterations (akin to the effects of chromosomal instability), we divided TNBC patients into two groups based on the median value of the total arm-level SCNAs present in each tumor. We then performed a Kaplan–Meier survival test and observed no significant differences in overall survival (Fig. 2C), suggesting that the prognostic value of 21q gain is more specific and predictive than simply the total quantity of arm-level SCNAs.

Fig. 1: Schematic representation of the workflow employed in this study.

Created with BioRender.com.

Fig. 2: 21q gain and RIPK4 expression correlate with poor survival in TNBC.

A Average frequency and survival association of chromosome arm-level gains and losses in patients with TNBC. B Kaplan–Meier survival curves for TNBC patients with 21q gain versus wild-type (WT). C Kaplan–Meier survival curve of patients with TNBC based on the median sum of all arm-level SCNAs. D Genes on 21q that are most significantly correlated with survival in TNBC. E Survival correlation of genes on 21q across 36 cancer types from PRECOG. Red represents a negative survival correlation while green represents a positive survival correlation. F Z-scores of the association between RIPK4 expression and survival of patients across 36 cancer types. G Quantification of RIPK4 expression in TNBC tumors compared to other BC subtypes. H Quantification of RIPK4 expression in tumors with a 21q gain compared to tumors that are 21q WT. I Selected relevant pathways from Ingenuity Pathway Analysis from primary 4T1 cell lines or human TNBC cell lines. The size of the dots represents the –log(p value), while the color represents the magnitude of the z-score. The numbers next to the dot represent the rank of the pathway when ordered based on z-score significance. Non-parametric Mann–Whitney t tests were performed for all bar graphs. P ≤ 0.05*, P ≤ 0.01**, P ≤ 0.001***, P ≤ 0.0001****.

We next assessed the putative importance of specific genes located on 21q. We found that DONSON, RIPPLY3, SLC19A1, PTTG1GIP, ERG, ATP5PF, MAP3K7CL, and RIPK4 were the only genes located on 21q whose elevated expression was significantly correlated with overall survival (Fig. 2D). DONSON was the most significantly correlated with survival, a gene well documented to promote BC progression [18]. To further investigate the survival association of these genes, we used the PREdiction of Clinical Outcomes from Genomic Profiles (PRECOG), an integrated cancer gene expression and clinical outcome dataset encompassing 166 independent cancer expression datasets for approximately 18,000 patients diagnosed with 39 distinct malignancies [19]. The most intriguing gene identified through this analysis was RIPK4 which, to our knowledge, has not previously been linked to BC progression (Fig. 2E). Indeed, it was also most significantly correlated with poor survival outcome in BC over any of the other 38 cancer types (P = 0.000000003 in BC; Fig. 2F). Further, we also found that RIPK4 expression was significantly higher in TNBC tumors compared to other subtypes (Fig. 2G) and that TNBC tumors harboring a 21q gain express RIPK4 at higher levels than tumors that are 21q WT (Fig. 2H).

RIPK4 encodes receptor-interacting protein kinase 4 enzyme (RIPK4) and is part of a family of 4 genes, RIPK1-4. RIPK1 and RIPK3 function as regulators of necroptosis [20], while RIPK2, most structurally similar to RIPK4, is involved in the NOD2 activation pathway, a key part of innate immune signaling [21]. RIPK4 has not been studied in the context of TNBC, but it has been shown to promote metastatic behavior of cells through epithelial to mesenchymal transition in a cell-intrinsic manner and can activate the NF-κB pathway [22,23,24].

To assess how RIPK4 expression may affect tumor cells, we leveraged the publicly available RNA sequencing data from the Cancer Cell Line Encyclopedia [25] and divided 47 breast cancer cell lines based on RIPK4 expression (Supplementary Fig. 1A, Supplementary Table 2). The top pathways predicted to be activated in RIPK4high cell lines were interferon alpha/beta signaling, pathogen-induced cytokine storm signaling, and interferon-gamma signaling. This is consistent with the literature demonstrating that RIPK4 is known to influence cytokine production and modulate interferon regulatory factor 6 (IRF6) [26, 27]. Importantly, we observed these pathways were among the top activated pathways when we performed RNA-sequencing on 4T1-primary cells, a syngeneic murine TNBC cell line in which we overexpressed Ripk4 by stably infecting these cells with a Ripk4 open reading frame (ORF- Fig. 2I, Supplementary Fig. 1B).

To explore which cells specifically express RIPK4 within the tumor microenvironment, we leveraged single-cell RNA sequencing data from the Human Protein Atlas (HPA) and found that the expression of RIPK4 was almost exclusively limited to breast glandular cells within normal breast tissue, with almost no expression in leukocytes [28] (Supplementary Fig. 1C, D). In contrast, other genes we discovered that correlate with poor survival, such as DONSON, were highly expressed by both breast tissue and infiltrating immune cells (Supplementary Fig. 1E), suggesting that DONSON expression assessed using bulk RNAseq (as in the TCGA) may be confounded by changes in the proportion of tumor-infiltrating cells rather than genomic changes within the tumor cells themselves. Taken together, these data implicate RIPK4 as a putative driver of TNBC progression through cancer cell-intrinsic expression of RIPK4.

Ripk4 knockdown drives aggressive TNBC phenotypes in vitro

Since mortality of BC patients is linked to metastatic disease and the lungs represent a common metastatic site, we focused our attention on the role of RIPK4 on TNBC dissemination. We introduced a small-hairpin RNA (shRNA) against Ripk4 by lentiviral transduction in 4T1-LuM, a murine lung-metastatic TNBC cell line that is syngeneic in immunocompetent Balb/c mice. Real-time quantitative PCR confirmed that Ripk4 expression was reduced by 87% relative to scrambled control (shSCR – Supplementary Fig. 2A). We then explored the mechanisms by which RIPK4 promotes TNBC metastasis by investigating its contributions to specific metastatic phenotypes. We performed a series of in vitro assays aimed at determining changes in proliferation, migration, and invasion of TNBC cells with Ripk4 perturbation. We found Ripk4 knockdown (KD) did not result in significant changes in the proliferation (Fig. 3A). Surprisingly, Ripk4 KD enhanced migration and invasion in vitro using transwell assays (Fig. 3B, C; Supplementary Fig. 2B, C). To gain further mechanistic understanding into the effect of Ripk4 KD in our model, we first imaged these cells using holotomography, which revealed no observable changes in cell morphology following genetic perturbation (Supplementary Fig. 2D). Given these surprising results, we wanted to examine the genetic effect of Ripk4 KD in these cells by performing RNA sequencing followed by Ingenuity Pathway Analysis. PAK Signaling and Rho GTPase Cycle, which are involved in the regulation of a wide variety of cell processes, such as cell survival and adhesion, were top pathways predicted to be upregulated in Ripk4 KD cells compared to control [29, 30]. Consistent with our previous pathways analysis (Fig. 2I), interferon signaling was among the top pathways predicted to be downregulated with Ripk4 KD (Supplementary Fig. 2E). These findings suggest that the functions of RIPK4 are likely dependent on the structural and compositional complexity of a three-dimensional tumor immune microenvironment.

Fig. 3: Ripk4 KD decreases lung metastases in vivo.

A Growth kinetics of Ripk4 knockdown cells in vitro. B, C Bar plots illustrating the migration and invasion potential of cells with Ripk4 knockdown in vitro. Graphs show fold-change relative to the average of controls. D Percent tumor area of lungs from mice injected intravenously (i.v.) with either Ripk4 KD cells or the non-target control (shSCR) 14 days post-injection. E Representative images of lungs harvested and stained with H&E from mice in (D). F Kaplan–Meier survival curves of mice injected i.v. G Left panel: Gross anatomy of the lungs of mice injected i.v. with engineered 4T1-LuM lines at endpoint. Right panel: Dotted lines encircle tumors from representative images of lungs in the left panel. H Number of CFSE+ cells per µl of sample acquired able to extravasate into the lung parenchyma within 48 h. I CFSE+ cells able to extravasate to the lung parenchyma within 48 h as a percentage of live cells. J Percentage of CFSE+ tumor cells positive for cleaved caspase-3. K Representative images used for the quantification of cleaved caspase-3+ cells. Non-parametric Mann–Whitney t tests were performed for all bar graphs, except for (B) where a parametric t test was performed. P ≤ 0.05*, P ≤ 0.01**, P ≤ 0.001***, P ≤ 0.0001****.

Ripk4 promotes lung metastasis of murine TNBC

To assess the metastatic capabilities of 4T1-LuM cells upon Ripk4 KD, we performed intravenous injection into WT Balb/c mice and assessed experimental lung metastasis after 14 days. Histological analysis revealed that mice injected with Ripk4 KD cells had significantly fewer lung metastases compared to those injected with shSCR cells (Fig. 3D, E). We then performed a survival experiment to measure the protective effect of Ripk4 KD. We found that mice injected intravenously with Ripk4 KD cells had a significant survival advantage over mice injected with the control cells (Fig. 3F). However, despite the noticeable reduction in tumor burden at day 14 in mice with Ripk4 KD, the tumors that persisted continued to grow, leading to the mice eventually succumbing to the tumor burden in their lungs (Fig. 3G). These data functionally confirm our computational predictions linking RIPK4 with survival in patients, by demonstrating that Ripk4 expression within tumor cells is sufficient to promote TNBC metastatic seeding and outgrowth in preclinical models.

Ripk4 increases survival of cancer cells at distant metastatic sites

To gain deeper insights into how Ripk4 expression correlates with poor survival in vivo, we explored whether the expression of Ripk4 in cancer cells might facilitate extravasation to metastatic sites, such as the lungs. Intravenously injected CFSE-stained cancer cells were given 48 h to seed the lungs before harvesting. Flow cytometric analyses revealed no discernible difference in the number of cells capable of seeding the lungs (Fig. 3H, I). Considering the well-established role of the RIPK-family protein in cell survival, we investigated whether the observed outgrowth phenotype in the Ripk4 KD (Fig. 3D, E) group could be linked to increased cell death. Immunofluorescent analysis on the harvested lungs revealed a significant elevation in cleaved-caspase 3+ cancer cells in mice injected with the KD cells compared to controls (Fig. 3J, K). Using STRING v.12.0 [31], a database of known and predicted protein-protein interactions, we determined that RIPK4 is likely to bind to Protein kinase C delta type regulatory subunit (PRKCD). PRKCD is a serine/threonine-protein kinase that regulates apoptosis triggered by cytokine receptors [32]. These findings suggest that RIPK4 may be involved in sustaining cancer cell survival at distant metastatic sites. Moreover, examination of lungs from mice at endpoint revealed distinct macroscopic growth patterns aligning with differences in cancer cell survival (Fig. 3G). Notably, Ripk4 KD-injected mice exhibited fewer but larger tumors compared to the numerous smaller tumors found in the lungs of mice injected with control cells. This disparity further explains the survival advantage observed in mice injected with Ripk4 KD cells (Fig. 3C). The diminished number of surviving cells in the lung tissue takes more time to develop into expansive and dispersed tumors capable of causing mortality, in contrast to the control cells that, despite surviving extravasation, form smaller tumors covering the entire lungs at the time of death. In essence, the distinct tumor growth patterns shed light on the survival advantage conferred by Ripk4 knockdown and underscores that RIPK4 may be mediating the rate-limiting step of cancer metastasis, seeding at the secondary site.

The tumor immune landscape of metastatic lesions is modulated by Ripk4

Given that RIPK4 has been shown to play a role in the activation of the NF-κB pathway, and its expression negatively correlates with the predicted presence of T cells in human ovarian cancer [22, 33], we reasoned that RIPK4 may be modulating the tumor microenvironment. To comprehensively assess how RIPK4 shapes the immune landscape of TNBC, we performed spectral flow cytometry on our experimental lung metastasis assay using a 19-plex antibody panel. We first performed a broad characterization of major immune cell types and found a significant increase in lymphoid cells and a relative decrease in myeloid cells in lungs harboring Ripk4 KD metastases compared to shSCR controls (Fig. 4A; Supplementary Fig. 3A, B). We performed these assays 14 days post injection, when both groups had tumors in the lungs but with clearly distinct growth patterns (Fig. 3E). Within the total leukocyte pool, this increase in lymphocytes following Ripk4 KD was driven by an increase in CD4+ and CD8+ T cells, with no change observed in B cells (Fig. 4B–E). We also found a significant decrease in the percentage of eosinophils and non-tissue resident macrophages following Ripk4 KD compared to controls, while tissue-resident macrophages remained at similar levels between both groups (Supplementary Fig. 3C–E). Given that a hallmark feature of TNBC is an “immune-cold” microenvironment where lymphocyte infiltration into the tumor is limited [15, 34, 35], our data indicate that targeting RIPK4 may foster an increase abundance of lymphocytes at the tumor site. To further dissect specific populations within the microenvironment that were sensitive to Ripk4 alterations, we performed unsupervised clustering of our spectral flow cytometry data via PhenoGraph [36]. Partitioning high-dimensional data into clusters using this approach enables the identification of rare phenotypic subsets or functional states, thus revealing potential biomarkers or novel cell populations that are more likely to be overlooked using traditional gating approaches. Using PhenoGraph, we identified 29 clusters in total (Fig. 4F, H–J), including 9 neutrophil clusters (1, 12, 4, 11, 18, 29, 2, 9, 3), 3 CD8+ T cell clusters (26, 14, 16), 3 CD4+ T cell clusters (6, 21, 10), and 5 monocyte/macrophage clusters (9, 22/8, 15, 27). We also saw individual clusters for CD4−CD8− double-negative T cells (23), eosinophils (20), dendritic cells (28), and B cells (7), suggesting reduced functional heterogeneity compared to other major cell types; as well as 5 leukocyte clusters that were undefined by the markers in our panel.

Fig. 4: Ripk4 expression skews the tumor immune microenvironment towards a pro-tumorigenic state.

A Abundance of the myeloid and lymphoid compartments within the tumor immune microenvironment of mice injected i.v. B–E Relative abundance of B cells, T cells, CD4+ T cells, and CD8+ T cells, respectively, as a percentage of CD45+ cells. F UMAP depicting the 29 clusters obtained through the unsupervised clustering algorithm PhenoGraph. On the left is the non-target control and on the right is Ripk4 KD. Density plots of the UMAPs from each group can be found on the right. G UMAP depicting the supervised clustering of immune cell populations present within the tumor immune microenvironment of tumor-bearing lungs. Percent frequencies of the 29 clusters in the non-target control group (H) and Ripk4 KD group (I). J Heatmap representing the relative expression level of each marker across each cluster. Color boxes along the left y-axis match the supervised clustering cell type assignment. K Bioluminescence signals from mice injected i.v. with 4T1-LuM cells and treated with IgG, 13 days post-injection. L Bioluminescence signals from mice injected i.v. with 4T1-LuM cells and treated with anti-Ly6G, 13 days post-injection. Non-parametric Mann–Whitney t tests were performed for all bar graphs. P ≤ 0.05*, P ≤ 0.01**, P ≤ 0.001***, P ≤ 0.0001****.

Using this dataset, we quantified the frequencies of individual clusters. Consistent with our supervised gating results (Fig. 4D, E, G), we found increased frequencies of CD4+ and CD8+ T cell clusters in the Ripk4 KD group (Fig. 4H, I). Moreover, monocyte-derived macrophages (CD11b+ CD64+ F4/80+ CCR2+) were increased in the control group, while tissue-resident macrophages (CD11b- F4/80+ CD64+ Siglec-F+ CD11chi) remained unchanged (Fig. 4H–J). Interestingly, while neutrophils represented the largest leukocyte population in metastases (Fig. 4G–I), we did not see a significant difference in the frequency of total neutrophils between control and Ripk4 KD tumors (Supplementary Fig. 3F). Instead, the most striking changes were found amongst neutrophil clusters, where clusters 1, 12, 11, and 18 were increased in the Ripk4 KD metastases compared to shSCR control, and clusters 2, 19, and 3 were reduced (Fig. 4H, I). The most abundant neutrophil cluster in Ripk4 KD metastases (and the most abundant of all immune clusters) had highest levels of Ly6G and intermediate levels of both CXCR2 and Siglec-F (Fig. 4H–J), coinciding with reduced metastatic burden. Given that high Siglec-F in neutrophils imparts a pro-tumorigenic phenotype [37], the reduced expression of Siglec-F aligns with our observation of reduced metastatic burden in response to Ripk4 KD.

Neutrophils are highly heterogeneous in their ability to elicit either pro- or anti-tumorigenic effects on cancer metastasis [38]. Given the impact of RIPK4 on metastasis in our model, we further explored the functional link between tumor cell-RIPK4 and neutrophils. First, we explored whether RIPK4 regulates CXCR2 ligands using a cytokine array on tumor cell-conditioned media. Among the top proteins that were differentially produced, we found CXCL1 production by Ripk4 KD cells was upregulated compared to shSCR control cells (Supplementary Fig. 3G, H), potentially acting as a chemotactic factor. Next, we asked whether the efficacy of Ripk4 KD was dependent on neutrophils. To achieve this, neutrophils were depleted in our experimental lung metastasis assay using antibodies targeting Ly6G [39]. We found that the ability of Ripk4 KD to reduce metastasis was only observed in neutrophil-proficient IgG-treated mice (Fig. 4K, Supplementary Fig. 3I) and was lost in neutrophil-deficient mice treated with anti-Ly6G (Fig. 4L, Supplementary Fig. 3I). These findings indicate that the phenotype we observed with Ripk4 perturbation, is in part mediated through interaction with neutrophils within the tumor immune microenvironment.