Targeting Axl favors an antitumorigenic microenvironment that enhances immunotherapy responses by decreasing Hif-1α levels

Axl Contributes to the Deregulation of the TME in a Model of HER2⁺ Breast Cancer.

Axl germline knockout (KO) mice were crossed with the HER2⁺ breast cancer model that expresses an activated form of the HER2 rat ortholog in the mammary gland under the control of mouse mammary tumor virus (MMTV; NeuNDL2-5). This Neu⁺ mouse model develops breast tumors that spontaneously metastasize to the lungs, and we previously reported that Neu⁺:Axl^−/− are protected from this metastatic progression without affecting primary tumor growth (20). Transcriptomic analyses of the Axl-null tumors 5 wks after tumor onset suggested a dampening of “hypoxia” and “angiogenesis” and an enhancement of “activation of immune response” by gene set enrichment analyses (GSEA) (Fig. 1 A–C and SI Appendix, Table S1). We hypothesized that the Axl-deficient tumors exhibit an antitumorigenic TME that can contribute to the poor capacity of those tumors to metastasize.

To address this, we collected the primary tumors 5 wks after the tumor onset to characterize their TME by immunostaining. Neu⁺:Axl^−/− tumors showed lower hypoxia levels as indicated by the decrease of the mean fluorescence intensity of two independent hypoxia markers, pimonidazole and GLUT-1 (Fig. 1D and SI Appendix, Fig. S1A). We next investigated the structural and functional changes in blood vessels and found that Neu⁺:Axl^−/− tumor blood vessels covered a larger area and had increased diameters when compared to control tumors (Fig. 1E and SI Appendix, Fig. S1B). They were less permeable, as suggested by a reduction of extravascular fibrinogen leakiness, and were better perfused, as demonstrated by intravenous injection of labeled Lectin (Fig. 1 F and G). The vessels also presented an increased coverage by αSMA⁺ pericytes and Laminin⁺ basement membrane (BM) (Fig. 1 H and I). Thus, the decrease in intratumoral hypoxia and the vessel normalization observed could contribute to the reduction of metastasis observed in Neu⁺:Axl^−/− mice.

Moreover, Neu⁺:Axl^−/− tumors showed a differential immune cells profile compared to control tumors, as assessed by a flow cytometry analysis of a panel of lymphoid and myeloid markers (Fig. 1J). Interestingly, macrophages were less polarized toward an immune-suppressive phenotype (CD206⁺) and also presented an increase in the expression of MHC class II (I-A/I-E), consistent with a proinflammatory activation state in the Neu⁺:Axl^−/− tumors (Fig. 1J). While CD3⁺ lymphocytes were recruited into the tumor core, as shown by staining of tumor sections, no significant increase of this infiltration in Neu⁺:Axl^−/− tumors was observed (SI Appendix, Fig. S1C). Nevertheless, the difference of infiltration suggested a possible enhancement of antitumor response with an increase of natural killer (NK) cell activation (as measured by higher CD44 expression) and a significant reduction in the number of regulatory T cells (Treg) in Neu⁺:Axl^−/− tumors (Fig. 1J). Collectively, these results suggest that deleting Axl reduced the intratumoral hypoxia, normalized the blood vessels, and mediated a differential immune cells’ representation in tumors.

Conditional Deletion of Axl in the Mammary Epithelial Cells Generates an Antitumorigenic Microenvironment and Reduces the Metastatic Burden.

We next asked whether the effects of Axl KO were driven by Axl deletion in the tumor cells intrinsically or by its deletion in cells present in the microenvironment such as immune or endothelial cells. We generated primary tumor cell lines from Neu⁺:Axl^+/+ and Neu⁺:Axl^−/− hosts and transplanted them into syngeneic wild-type FVB mice. Notably, RNA sequencing of Neu⁺:Axl^−/− tumor grafts recapitulated the gene expression signatures linked to the TME deregulation found initially in the tumors from Neu⁺:Axl^−/− mice (SI Appendix, Fig. S2 A and B and Table S2). Neu⁺:Axl^−/− tumor grafts also presented less hypoxia and produced blood vessels with a decreased permeability, increased perfusion, and increased pericytes and BM coverage (SI Appendix, Fig. S2 C–I). Thus, the normalized blood vessel phenotype could be reproduced when Neu⁺:Axl^−/− cells were grafted in a wild-type host, suggesting that this deregulation is a consequence of losing Axl in the cancer cells themselves. Furthermore, when investigating the immune profile of Neu⁺:Axl^−/− tumor grafts by flow cytometry, we found that macrophages were polarized toward a proinflammatory activation state (CD206⁻, I-A/I-E⁺), and there was a reduction in neutrophils compared to the controls (SI Appendix, Fig. S2J). These myeloid cells are known to regulate the TME by producing various inflammatory cytokines and proangiogenic factors, suggesting that Axl-expressing cancer cells could educate immune cells to contribute to the TME remodeling (26, 27). Thus, this experiment indicated that the TME changes observed in Neu⁺:Axl^−/− mice have been orchestrated by the cancer cells.

To confirm the cancer cell–autonomous role of Axl in the metastatic progression and the deregulation of the TME, Axl conditional KO mice were crossed with MMTV-NIC mice (MMTV-NeuNDL2-5-IRES-Cre) to obtain a specific deletion in the mammary epithelial cells (SI Appendix, Fig. S3A). Weekly monitoring of cohorts of NIC⁺:Axl^wt/wt, NIC⁺:Axl^wt/fx, and NIC⁺:Axl^fx/fx mice by palpation suggested no difference in tumor initiation (Fig. 2A). Moreover, there was no difference in the number of mammary intraepithelial neoplastic lesions 5 wks after tumor onset (SI Appendix, Fig. S3B). The tumor and metastatic burdens were also analyzed 5 wks after the appearance of the first mass. No difference in the total tumor weight per mouse was observed, and there were no significant changes in the proliferative and apoptotic status in the tumors (Ki67 or TUNEL, respectively) (Fig. 2B and SI Appendix, Fig. S3 C and D). Importantly, the loss of Axl altered the lung metastatic frequency and burden, as we observed in total Axl KO mice (20), suggesting a strong cancer cell–autonomous role in the phenotype (Fig. 2C and SI Appendix, Fig. S3E). Furthermore, the number of circulating tumor cells in the blood of the mice was reduced in NIC⁺:Axl^wt/flx and NIC⁺:Axl^flx/flx animals, suggesting that the cancer cells may be poorly invasive, as we have shown in ref. 20, and/or that the TME is unfavorable for intravasation (Fig. 2D). Thus, we decided to investigate whether the TME could also be rewired toward a tumor-suppressive phenotype in those tumors and consequently contribute to limit metastasis.

NIC⁺:Axl^fx/fx tumors displayed reduced hypoxia levels, and the blood vessels looked functionally normalized (Fig. 2 E–I and SI Appendix, Fig. S3 F and G). Similarly, vessels in NIC⁺:Axl^fx/fx tumors were less permeable, more perfused, and they presented a better pericyte and BM coverage as shown by immunostaining (Fig. 2 E–I). Furthermore, we investigated immune cell infiltration by flow cytometry to assess whether Axl deletion in the tumor cells could also generate a better immune response. Interestingly, macrophages and neutrophils were less present in NIC⁺:Axl^fx/fx tumors, and again, those macrophages showed a proinflammatory activation state with less CD206 and more I-A/I-E expression (Fig. 2J). Moreover, NIC⁺:Axl^fx/fx tumor cells presented more CD45⁺ immune cells with an increase of cytotoxic CD8 T cells and CD4 T cells (Fig. 2J). Globally, there was an increase in the recruitment of CD3+ lymphocytes in the core of NIC⁺:Axl^fx/fx tumors compared to the control (SI Appendix, Fig. S3H). Collectively, these data suggest that Axl expression in cancer cells contributes to deregulating the microenvironment to promote metastatic progression.

AXL Is Required to Sustain HIF-1α Levels under Hypoxic Conditions.

We found that removing Axl in the cancer cells in vivo reduced the circulating tumor cells and metastasis. This could be explained by the reduction of hypoxia found in the NIC⁺:Axl^−/− tumors that could lead to a decrease of tumor cell invasion, normalized blood vessels, and increased immunosurveillance. Furthermore, we previously reported the interaction and crosstalk between AXL and HER2 to promote metastasis (20). This led us to explore their roles in hypoxic stress using tumor cell lines derived from the NIC⁺:Axl^+/+ and NIC⁺:Axl^−/− mouse model, termed NIC WT (wild type) and Axl KO (knockout). Various receptor tyrosine kinases are known to regulate HIF complex levels and activity via different downstream pathways including PI3K/AKT, JAK/STAT3, and MAPK (28). For instance, HER2 regulates the synthesis of HIF-1α subunits by directly activating the PI3K/AKT/FRAP pathway (29).

Under hypoxic conditions (1% O₂ for 24h), Axl KO cells showed drastically reduced Her2 and Hif-1α protein levels when compared to control cells, while Hif-2α was not affected (Fig. 3A). Proteasome inhibition with MG132 rescued Her2 expression, suggesting a role for Axl in Her2 protein stability during hypoxia (SI Appendix, Fig. S4A). Therefore, Axl may modulate Hif-1α levels by regulating Her2 stability or by directly influencing hypoxic signaling pathways. HIF-1α subunit induction by hypoxia was also abrogated upon AXL depletion using small interfering RNA (siRNA) in human MCF10A cells stably expressing HER2, validating in another system the results observed in the mouse-derived cells (SI Appendix, Fig. S4B). Furthermore, in human HER2⁺ breast cancer cells BT474 that do not express AXL, Axl overexpression potentiated HIF-1α induction and stabilized HER2 under hypoxia (Fig. 3B).

To demonstrate that Axl and Her2 activities are required for this effect, NIC WT cells were treated with Lapatinib (HER2 inhibitor) or R428 (AXL inhibitor). Single Her2 or Axl inhibition significantly decreased Hif-1α expression under hypoxia indicating that Her2 and Axl activity regulates Hif-1α levels under hypoxia in these cells. The combination of both inhibitors caused an additional reduction in Hif-1α levels, suggesting that Her2 and Axl can contribute independently to regulate Hif-1α expression during hypoxia (Fig. 3C). Furthermore, R428 decreased Her2 levels under hypoxia, confirming the role of Axl kinase activity on Her2 stability in hypoxic conditions (Fig. 3C). As expected, Her2 and Axl inhibition reduced pAkt levels, indicating that the Pi3k/Akt pathway could be implicated in the signaling of these receptors under hypoxia. We found that Pi3k inhibition with LY294002 also reduced Hif-1α levels in NIC WT cells (Fig. 3D). Thus, our results show that both Axl and Her2 can regulate Hif-1α levels under hypoxia potentially via activating the Pi3k/Akt pathway. Similar results were obtained in MCF10A-HER2 cells (SI Appendix, Fig. S4 C and D). To further dissect the contribution of AXL independently of HER2 signaling, triple-negative breast cancer cells expressing high levels of AXL, Hs578T, were used. The reduction of AXL levels using an siRNA decreased HIF-1α levels in hypoxia supporting a role for AXL independently of HER2 (Fig. 3E). Furthermore, ectopic expression of Axl in Axl KO cells rescued Her2 and Hif-1α expression (Fig. 3F). However, Hif-1α up-regulation was not completely lost by Lapatinib treatment, suggesting that Axl partially contributes to this up-regulation by stabilizing Her2 (Fig. 3F). Furthermore, LY294002 completely abrogated Axl-dependent HIF-1α up-regulation, indicating that Pi3k/Akt acts downstream of Axl to increase Hif-1α levels under hypoxic conditions (Fig. 3F). Remarkably, AXL expression in breast cancer patient samples from the TCGA-BRCA cohort correlated with HIF-1α gene expression independently of the subtype (SI Appendix, Fig. S4E). Taken together, these results indicate that Axl is required to ensure hypoxic Hif-1α expression by stabilizing Her2 and promoting Pi3k/Akt signaling. Hence, Axl can promote the deregulation of the TME and metastatic progression by regulating the Hif-1 complex during hypoxia.

Axl Is Required for a Complete Hypoxic Response.

Since Axl KO cells reduced Hif-1α levels under hypoxia, Axl deletion could have an impact on the hypoxic transcriptome. To test this and to better understand the role of Axl in hypoxic adaptation required for TME remodeling and metastasis, we sequenced RNA isolated from NIC WT and Axl KO cells under 1% O₂. We conducted a gene ontology analysis of hypoxia–up-regulated genes that revealed that NIC WT cells showed an expected hypoxic response that includes the up-regulation of genes involved in migration, angiogenesis, and cytokine secretion. However, this response was reduced in Axl KO cells (Fig. 4A and SI Appendix, Fig. S5A). Overall, there were more up-regulated genes under hypoxia in NIC WT cells (Fig. 4B). The genes specifically up-regulated in Axl KO cells during hypoxia were linked to cell cycle, cellular metabolic process, cellular response to stress, and programmed cell death, suggesting a harder adaptation to hypoxia (SI Appendix, Fig. S5B). On the other hand, the Axl-dependent genes up-regulated under hypoxic conditions were principally associated with the processes mentioned in Fig. 4A but also specifically included the HIF-1α signaling pathway (Fig. 4C). Indeed, many HIF-1α target genes, mainly related to EMT, invasion, and immune evasion, were less expressed and/or differently modulated in the Axl KO compared to the WT cells (SI Appendix, Fig. S5C). This result suggests that Axl is essential for a complete hypoxic response. Therefore, these findings explain at least in part the changes in the TME and the reduced metastatic burden observed in our mouse models.

To investigate whether these observations can be clinically relevant, we analyzed the genes that correlate with AXL expression in breast cancer samples from METABRIC and TCGA breast cancer cohorts. In general, AXL correlated with the expression of genes linked to cell migration, inflammation, angiogenesis, and hypoxia, among others (SI Appendix, Fig. S5D). Interestingly, a significant proportion of the Axl-dependent genes induced under hypoxia and found in our RNA-seq analyses correlated with AXL expression in breast cancer patients (SI Appendix, Fig. S5E). These candidate genes were also related to processes linked to the TME and metastasis including response to O₂ levels, migration, angiogenesis, and inflammation, supporting a role for AXL in the hypoxic response and the TME deregulation in vivo (SI Appendix, Fig. S5F).

Axl Is Essential for Hypoxia-Induced Changes that Drive Metastasis and TME Remodeling.

Hypoxia can promote metastatic progression by increasing tumor cell EMT and invasion via the up-regulation of EMT transcription factors that are themselves HIF targets (10, 11). In our RNA-seq experiment, Snai2 and Twist1 were found to be up-regulated under hypoxia in an Axl-dependent manner and correlated with AXL expression in human breast cancer samples (SI Appendix, Fig. S6 A–C). This led us to hypothesize that Axl deletion could decrease the ability of cells to become more mesenchymal in a hypoxic environment. Accordingly, the increase in the levels of the mesenchymal marker Vimentin in NIC WT cells under hypoxic conditions was blunted in Axl KO cells, and this was rescued by re-expressing Axl (Fig. 5A). The same phenomenon was observed in human MCF10A-HER2 cells in which AXL levels were reduced by siRNA (SI Appendix, Fig. S6D). We next tested the role of AXL on hypoxia-induced cell motility. We found that hypoxia increased the invasion of NIC cells and MCF10A-HER2 cells in Boyden–Matrigel invasion assays (Fig. 5B and SI Appendix, Fig. S6E). In both cell lines, interfering with AXL by KO, siRNA, or R428 treatment reduced the basal invasion in normoxia and inhibited the promotion of invasion by hypoxia (Fig. 5B and SI Appendix, Fig. S6E). This was also reproduced in wound-healing assays in which AXL was required for hypoxia-induced cell migration (Fig. 5 C and D and SI Appendix, Fig. S6 F and G). These results demonstrate defects in hypoxia-induced EMT and invasion/migration when interfering with AXL. This further supports the notion that the ability of cancer cells to leave the primary site as circulating tumor cells and to form metastasis is impaired in NIC⁺:Axl^fx/fx mice.

Furthermore, cancer cells boost the production of different cytokines to remodel the TME and promote immune suppression under hypoxic conditions (10, 30). Interestingly, our RNA-seq data revealed that Axl removal affects the messenger RNA (mRNA) expression of some chemokines including C-C Motif Chemokine Ligand 2 (Ccl2), Colony-Stimulating Factor 1 (Csf1), C-X-C Motif Chemokine Ligand, and 2 (Cxcl1 and Cxcl2) under hypoxia (SI Appendix, Fig. S7A). A cytokine profiling by a multiplex array was done to assess the secretion by WT and Axl KO cells of a variety of chemokines (SI Appendix, Table S3). Indeed, Axl KO cells secreted lower amounts of Ccl2, Csf1, Cxcl1, and Cxcl2 when subjected to hypoxic conditions (Fig. 6A and SI Appendix, Fig. S7 B–D and Table S3). Interestingly, among these candidates, CCL2 expression correlated with AXL expression in human breast cancer, suggesting a possible clinical relevance of this finding (SI Appendix, Fig. S7E). CCL2 is a myeloid and lymphoid cell chemoattractant that promotes the recruitment of tumor-associated macrophages and affects their behavior in the TME (26, 31, 32). Tumor-associated macrophages are a major stromal component and can be reprogrammed by environmental signals into specialized subtypes including a spectrum of antitumoral (proinflammatory) or pro-tumoral (immune-suppressive or wound-healing) phenotypes (33–36). Importantly, we observed a reduction in macrophage numbers and polarization toward a protumoral phenotype in our Axl KO breast cancer mouse models (Fig. 1J and SI Appendix, Fig. S2J, and Fig. 2J). This infiltration was also validated by staining in the NIC⁺:Axl^wt/wt and NIC+:Axl^fx/fx tumors (Fig. 6B). Furthermore, these infiltrated macrophages present a decrease of proliferation in Axl-null tumors as demonstrated by fluorescence-activated cell sorting (FACS) analysis of staining against phospho-Histone 3 (Fig. 6C). Therefore, we tested in vitro the requirement of Axl in tumor cells for macrophage proliferation, invasion, and polarization during hypoxia by treating bone marrow–derived macrophages with conditioned media (CM) from normoxic and hypoxic NIC WT and Axl KO cells. First, we observed that CM from hypoxic Axl KO cells was less efficient to promote the proliferation of macrophages (Fig. 6D). The addition of a neutralizing antibody against Ccl2 in WT CM also moderately reduced their proliferation, suggesting a partial implication of Ccl2 in this phenotype (Fig. 6D). Furthermore, CM from hypoxic NIC WT cells increased macrophage invasion compared to normoxic CM (Fig. 6E). However, CM from hypoxic Axl KO cells failed to stimulate invasion to the same extent, suggesting that Axl deletion in tumor cells can also affect the invasion and recruitment of macrophages in the tumor (Fig. 6E). In this case, we revealed a role for Ccl2 as shown by a reduction in invasion when an anti-Ccl2 antibody was added to the hypoxic WT CM (Fig. 6E). Finally, we investigated the effect of the hypoxic NIC CM on the polarization of bone marrow–derived macrophages. First, the immunosuppressive macrophage marker CD206 was less expressed in macrophages treated with hypoxic CM from Axl KO cells compared to control, correlating with in vivo observations (Fig. 6F). This phenotype was also partially recapitulated by the inhibition of Ccl2 (Fig. 6F). Moreover, macrophages treated with hypoxic Axl KO CM secreted less factors known to promote TME deregulation and angiogenesis such as Vegfa, Cxcl1, and Cxcl2 (Fig. 6 G–I). These results suggest that Axl is required for the secretion of factors, such as Ccl2, leading to macrophage proliferation, invasion, and polarization when cancer cells are subjected to hypoxia. Consequently, this could affect their recruitment and reprogramming to promote angiogenesis and immune suppression to affect the TME and potentially the metastatic progression in our model of HER2⁺ breast cancer.

Axl Inhibition Generates an Antitumorigenic TME and Enhances Anti–PD-1 Immunotherapy Response.

Clinical trials so far suggest that achieving responses to immunotherapy in HER2⁺ cancers is challenging (3, 4), and a better understanding of the HER2⁺ TME may help to better classify possible responders. Indeed, there is a growing clinical need to define factors that modulate response to immunotherapies and specifically checkpoint inhibitors (3, 4, 37). The influence of hypoxia in the TME has been suggested to be one of those factors (38). As such, we chose to investigate the landscape of hypoxia among different molecular subtypes of breast cancer. Toward this goal, we performed a single-sample GSEA (ssGSEA). Hypoxia signatures scores, when applied to data from TCGA-BRCA, GSE58644, and METABRIC breast cancer samples (39–41), demonstrated that basal and HER2⁺ subtypes present higher levels of hypoxia when compared to other breast cancer subtypes (Fig. 7 A and B and SI Appendix, Fig. S8 A–D). Additionally, HER2⁺ patients whose tumors display higher hypoxia scores are associated with shorter overall and disease-free survival, suggesting that HER2⁺ tumors with evidence of elevated signatures of hypoxia are more aggressive (Fig. 7C and SI Appendix, Fig. S8 E–K). In our experiments, hypoxic response elements, identified to be regulated by Axl, EMT, and tumor cell–macrophage interactions, can promote immune evasion and have been established as roadblocks to efficient immunotherapy (26, 42). Moreover, vascular normalization, as observed in Neu⁺:Axl^−/− and NIC⁺:Axl^fx/fx tumors, was demonstrated to enhance cancer immunotherapy (16, 43). Collectively, these observations suggest AXL as a druggable candidate to generate an improved setting for immunotherapy. This could be promising for HER2⁺ patients that present prominent features of hypoxia.

Therefore, Neu⁺ mice were treated individually and in combination with anti–PD-1 and AXL small-molecule inhibitor R428, as illustrated in SI Appendix, Fig. S9A. We observed that R428 treatment efficiently modulated the TME similarly to what we observed in total and conditional KO models. Indeed, there was a reduction of intratumoral hypoxia and a normalization of the blood vessels characterized by decreased permeability, increased perfusion, and improved pericyte coverage as shown by immunostaining (Fig. 7 D–G). Moreover, flow cytometry experiments showed that upon Axl inhibition, macrophages presented a proinflammatory activation state characterized by a reduction of CD206 and an increase of MCH-II (I-A/I-E) expression, confirming the effect of Axl inhibition on macrophage phenotype (Fig. 7H). Furthermore, on analyzing the tumor and metastatic burden of treated mice, we determined that Neu⁺ mice were minimally responsive to anti–PD-1 therapy alone (Fig. 7 I–K). Only the combination of the anti–PD-1 immune checkpoint blockade and the Axl inhibitor was able to reduce the total tumor weight per mouse (Fig. 7I). Importantly, R428 treatment alone reduced the metastatic burden and frequency, as we reported previously (20), and the combination with anti–PD-1 potentiated this effect (Fig. 7 J and K). Interestingly, tumors treated with the combination of the two therapeutic agents presented an increase of CD8 T cells which, with the NK cells, were more activated (CD44⁺), possibly explaining the effect of the combined therapy (Fig. 7L). Indeed, an increase in CD3⁺-infiltrating cells was observed in the core of tumors cotreated with anti–PD-1 and R428 (SI Appendix, Fig. S9B). Furthermore, these tumors showed a significant increase in apoptosis but no difference in proliferation (SI Appendix, Fig. S9 C and D). These results suggest that AXL inhibition could be a powerful approach to generate an anti-tumorigenic microenvironment to improve immunotherapy and limit the metastatic spreading of breast cancer.