The MNK1/2–eIF4E Axis Supports Immune Suppression and Metastasis in Postpartum Breast Cancer

Postpartum breast cancer (PPBC) is defined as breast cancer diagnosed within 10 years of parturition (1). Given its highly metastatic nature (1, 2), the patient prognosis is poor. The physiologic process of mammary gland (MG) involution, the remodeling of the breast tissue back to its prepregnant state, has been hypothesized to cause premalignant epithelial cells to adopt invasive properties (2). Involution, akin to the process of wound healing, is accompanied by an orchestrated immune cell infiltration into the mammary gland (3). Data from murine PPBC models suggest that interactions between innate and adaptive immune cells as well as involution-activated fibroblasts are fundamental in establishing a suppressed microenvironment that is favorable for metastatic spread.

Cancer cell invasiveness is regulated by growth factors, cytokines, and chemokines produced by tumor cells and associated stromal cells within the tumor microenvironment (TME). Recently, the IL33/ST2 signaling axis has come to the forefront as an important mediator of metastasis. IL33 is an alarmin cytokine of the IL1 family, is involved in inflammation, tissue homeostasis, and tumor progression, and signals via binding to the ST2 receptor (4). Although there are many cellular sources of IL33, its secretion by cancer-associated fibroblasts has been shown to promote the epithelial-to-mesenchymal like transition and tumor cell invasion (5, 6). IL33 can also promote tumor progression and immune suppression via activation of immune cells such as CD11b⁺Gr1⁺ myeloid-derived suppressor cells (MDSC; ref. 7) and innate lymphoid cells type 2 (ILC2; ref. 8). Once activated, these cells serve as potent inhibitors of cytotoxic T-cell tumor infiltration and antitumor function (9, 10).

Regulation of gene expression at the level of mRNA translation initiation is becoming increasingly studied in the field of oncoimmunology. Indeed, dysregulation of translational control is a prominent feature of many cancers (11). For example, elevated levels of the eukaryotic translation initiator factor 4E (eIF4E), which binds to the 7-methylguanosine cap at the 5′ end of the mRNA, are associated with malignancy and poor prognosis in several cancer types (12). eIF4E can be phosphorylated at serine 209 (S209) by MAP kinase–interacting serine/threonine-protein kinases 1 and 2 (MNK1/2), and this posttranslational modification is essential for its proinvasive effects (13). Increased MNK1/2 activity has been associated with therapeutic resistance, tumorigenesis, invasion, and metastasis (14). We and others have previously shown that phosphorylation of eIF4E leads to the translational upregulation of mRNAs, such as Myc, Mcl1, Mmp3 and Snai1, that support tumor cell survival and a proinvasive phenotype (13, 15). In the context of the TME, phospho-eIF4E has recently been reported to reinforce the survival of prometastatic neutrophils (16) and regulate the protumor functions of bone marrow–derived macrophages (17). In murine models of melanoma, we showed that phospho-eIF4E deficiency associates with decreased PD-L1 expression on dendritic cells and MDSCs in the TME (18). However, there remain large gaps in our understanding of how the regulation of eIF4E phosphorylation impacts the behavior of other immune and nonimmune stromal cells found within the breast TME.

Here, we show that host phospho-eIF4E has a pleiotropic effect in the TME of an animal model of PPBC, regulating the functions of fibroblasts and ILC2, two cell types important for the metastatic process. The altered functionality of fibroblasts, in turn, differentially affects tumor cells, to support the immune evasion and metastasis of PPBC tumors. We show that immune composition of both the primary tumor and metastatic niche is altered in the phospho-eIF4E–deficient animals. In a pioneering approach, using imaging mass cytometry on a cohort of human PPBC and non-PPBC tumors, we show that the human PPBC TME is characterized by markers of immune dysfunction. Finally, we provide evidence for a potential therapeutic intervention in PPBC, by showing that the combination of the MNK1/2 inhibitor SEL201 and PD-1 blockade decreases lung metastasis in a murine model of PPBC.

Wild-type (WT) BALB/c and C57BL/6N mice were purchased from Charles River Laboratory. eIF4E^S209A/S209A BALB/c and eIF4E^S209A/S209A C57BL/6N mice were gifts from Dr. Nahum Sonenberg at McGill University (Montréal, Canada) and have been described previously (13). PPBC models were set up as reported previously (19). Briefly, 5- or 6-week-old WT or eIF4E^S209A/S209A female mice were mated with male mice. Pregnant mice were monitored until pups were born and allowed to lactate for 11 to 14 days. The pups were removed from the dams causing the dams to undergo forced weaning-induced mammary gland involution. On involution day 1, that is 24 hours after forced weaning, 200,000 66cl4 cells were injected into the inguinal mammary gland of BALB/c mice and tumors were allowed to grow for 14 (early metastasis) or 33 days (full metastasis; Fig. 1A). E0771 cells (200,000) were injected into the mammary gland of C57BL/6N mice for 26 days. For drug treatment experiments, animals were treated at indicated time points, starting at 4 days after tumor cell injection. SEL201 (Ryvu Therapeutics) was dissolved in DMSO and then diluted in N-Methylpyrrolidone (NMP, Thermo Fisher Scientific) and Captisol (ligand) for administration by oral gavage at 75 mg/kg bodyweight per mouse per day, 5 days per week (with 2 days off) for 3 weeks (a total of 15 doses). The anti-mouse PD-1 mAb and IgG isotype control (BioCell) were diluted in PBS and administrated through intraperitoneal injection at 10 mg/kg bodyweight per mouse per day, once per week for 3 weeks (a total of 3 doses). Animal experiments were conducted following protocols approved by McGill University Animal Care and Use Committee.

SEL201 was a generous gift from Dr. Tomasz Rzymski at Ryvu Therapeutics. The 66cl4 and MDA-MB-231 cell lines were kind gifts from Dr. Josie Ursini-Siegel at McGill University. The E0771 cell line was purchased from CH3 BioSystems. All cell lines used are routinely (every 3 months) tested for Mycoplasma using the e-Myco PLUS Mycoplasma PCR Detection Kit (LiliF Diagnostics). Cells were injected into animals no later than 4 to 5 passages after thawing. 66cl4 was cultured in RPMI with 10% FBS and antibiotics (1× penicillin/streptomycin, Wisent). E0771 was cultured in RPMI supplemented with 10 mmol/L HEPES, 10% FBS, and antibiotics. WT and eIF4E^S209A/S209A (referred to as eIF4E^S209A) primary mammary gland fibroblasts were obtained by digesting minced mammary glands pooled from 3 to 4 donor mice in 1 mg/mL Collagenase IV in DMEM Advanced F12 for 1 hour at 37°C, passing the suspension through a 70-μm cell strainer, and centrifuging at 300 × g for 10 minutes. The pelleted cell suspension including fibroblasts was plated in DMEM supplemented with 10% FBS and antibiotics. To enrich for fibroblasts, culture medium was changed 30 minutes after plating, by removing old media and nonadherent cells and adding fresh media. Cells were expanded for 8 to 9 days. Conditioned media was prepared by thoroughly washing away culture media and culturing the fibroblasts in serum-free DMEM F12 for 48 hours. Presence of secreted IL33 in the conditioned media was visualized on a Proteome Profiler Mouse XL Cytokine Array (R&D Systems). The concentration of IL33 secreted in the conditioned medium was measured on a V-PLEX Mouse Cytokine 19-Plex Kit (Meso Scale Diagnostics) and normalized to total protein input measured by Nanodrop. WT and eIF4E^S209A/S209A (termed eIF4E^S209A) mouse embryonic fibroblasts (MEF) have been described previously (13, 15). Cancer-associated fibroblasts (CAF) derived from patients with breast cancer were obtained in collaboration with Dr. Mark Basik at McGill University as described previously (20). The collection and use of human tissues was approved by the Institutional Review Board, JGH (no. 05–006), which is in accordance with the Declaration of Helsinki and the Belmont Report. CAFs and MDA-MB-231 cells were cultured with DMEM supplemented with 10% FBS and antibiotics (1x penicillin/streptomycin, Wisent).

Tumor-bearing PPBC mice were sacrificed by cardiac puncture at day 14 of tumor growth and equal volumes/animal of peripheral blood were collected into EDTA-coated tubes. After treatment with red blood cell lysis buffer (Sigma-Aldrich) to selectively deplete erythrocytes and centrifugation, the cell pellet was resuspended in RPMI containing 5 mg/L 6-thioguanine (Sigma-Aldrich), plated in one well/animal of a 6-well plate and cultured at standard tissue culture conditions (37°C 5% CO₂). After 1 week, surviving 6-thioguanine resistant cells (circulating 66cl4 cells) were counted using a brightfield microscope under a ×10 magnification. For each animal, cell numbers in 4 individual fields of view were counted and the sum shown as tumor cells per 500 μL collected blood.

Lungs and tumors were resected from sacrificed animals at indicated time points and digested into a single-cell suspension via mechanical mincing into small pieces and incubation with collagenase IV (1 mg/mL) in RPMI medium (Gibco) for 1 hour at 37°C. After treatment with red blood cell lysis buffer to selectively deplete erythrocytes, cells were counted, blocked with anti-CD16/32 and stained with indicated antibodies (Supplementary Table S1). Flow cytometry data were acquired on a BD LSRFortessa flow cytometer (BD Biosciences) and analyzed with FlowJo software, version 10.7.1 (BD Biosciences). The gating strategies used for analysis are found in Supplementary Fig. S2.

ILC2Ps, the progenitors of ILC2s, were isolated from bone marrow and expanded as reported previously (21). AlamarBlue Cell Viability assay was performed according to the manufacturer's instructions (Thermo Fisher Scientific), and the absorbance was subsequently measured for cell division with excitation and emission wavelengths at 560 and 590 nm, respectively, as reported previously (21). IL5 and IL13 secretion by ILC2s was quantified by ELISA as reported previously (21). For Annexin V and live dead staining, cells were seeded at 10,000 cells/well in complete media and cytokine-starved for 3 hours prior to cytokine stimulation and drug inhibition. This was followed by addition of 0.5, 2.5, or 5 μmol/L SEL201, or DMSO with respective cytokines (10 ng/mL). After 5 days, cells were stained with Annexin V Apoptosis Detection Kit (eBioscience) according to the manufacturer's protocols and eFluor 780 Fixable Viability Dye (eBioscience). Data were acquired using a BD LSRFortessa flow cytometer.

66cl4 and E0771 cells were seeded at one (migration and invasion assay) or two (coculture assay) million cells per 10-cm dish on day 1 in full media, then starved overnight by switching to serum-free media on day 2. For the coculture assay, 200,000 WT or eIF4E^S209A MEFs were seeded into 12-well companion plates on day 2. On day 3, transwells (Corning) were coated with Collagen I (20 μg/mL) as reported previously (22). A total of 200,000 (migration and invasion) or 50,000 (coculture assay) tumor cells were seeded into the transwells and were allowed to migrate and invade for 16 hours (migration and invasion) or 48 hours (coculture). Migrated cells were fixed with 5% glutaraldehyde (Sigma) and stained with 0.5% crystal violet (Sigma) as reported previously (22). Stained cells were then counted and quantified. WT and eIF4E^S209A fibroblasts were harvested for WB or qPCR

For experiments with patient-derived CAFs, MDA-MB-231 cells were seeded at 3 × 10⁶ cells per 10-cm dish on day 1 in full media, then switched to serum-free media on day 2 and starved overnight. Patient-derived CAFs (50,000) were seeded into 6-well companion plates on day 2. On day 3, transwells were coated with Collagen I (20 μg/mL). MDA-MB-231 cells (200,000) were seeded into the transwells and were allowed to migrate and invade towards CAFs for 48 hours. Migrated cells were fixed, stained, and quantified as described above. CAFs were harvested for Western blotting.

IHC and hematoxylin and eosin staining were performed as described previously (22). Briefly, tumor and lung sections were stained for IL33, phospho-eIF4E, and CD8, and counterstained with 20% Harris-modified hematoxylin (Thermo Fisher Scientific). Antibody information is listed in Supplementary Table S2. Slides were scanned and assessed using Spectrum (Aperio Technologies). All animal and patient IHC samples were quantified by QuPath software.

Immunofluorescence (IF) staining was performed as described previously (23). Briefly, cells or tissues were stained for the indicated proteins, and nucleus were labeled with DAPI. Primary and secondary antibodies are listed in Supplementary Table S2. Slides were scanned with an axioscan Z1 slide scanner microscope (Zeiss) using a 20×/0.75NA objective. Images were analyzed using Zen blue software (Zeiss) and Qupath.

Cells were lysed with RIPA buffer (150 mmol/L Tris-HCl, pH 7, 150 mmol/L NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS) supplemented with protease and phosphatase inhibitors (Roche) as described previously (22, 24). Equal amounts of protein were loaded and separated on 10% SDS-PAGE gels. Antibodies were used to detect the indicated proteins. GAPDH was probed to confirm equal protein loading. Antibody information is listed in Supplementary Table S2.

RNA was prepared using E.Z.N.A. Total RNA Isolation Kit (OMEGA Bio-Tek). cDNA was prepared from 1 μg of total RNA, using iScript cDNA Synthesis Kit (Bio-Rad). Target genes were quantified using the Applied Biosystems 7500 Fast Real-Time PCR System with SYBR Green. Primers are listed in Supplementary Table S3.

ST2 knockdown in 66cl4 cells was performed as described previously (25). Briefly, scramble siRNA (AllStars Negative Control siRNA, Qiagen) or ST2 siRNA (IDT, sequences listed in Supplementary Table S4) were introduced into 66cl4 cells with the aid of lipofectamine RNA iMax reagents (Invitrogen) following the manufacturer's instructions.

Mammary glands from BALB/c eiF4E^WT and BALB/c eIF4E^S209A at day 8 after pregnancy (LD8) and day 2, 3, 4, 5, and 6 after weaning (ID2, ID3, ID4, ID5, and ID6, respectively) were collected, fixed in Carnoy's fixative, defatted one time with xylene, and stained in carmine red solution overnight. Tissues were dehydrated in increasing concentrations of ethanol (70%, 95%, and 100%, respectively), then cleared in xylene, and mounted with Permount. Pictures were taken on a surgical microscope at 25×. The average percentage of the area occupied by adipocytes compared with epithelial cells of 5 random fields at the indicated time points was quantified using ImageJ.

The study was approved by the ethics committee and in compliance with institutional review board approval from the two participating institutions: UZ/KU Leuven (Belgium) and Jewish General Hospital (Canada). Written informed consent was obtained from all patients (nulliparous breast cancer samples, breast cancer samples from patients diagnosed during pregnancy or diagnosed postpartum), and the study conducted in accordance with the Declaration of Helsinki. Human breast cancer samples were arrayed onto a slide, stained using the panel of antibodies listed in Supplementary Table S5, and processed with the Hyperion Imaging System (Fluidigm) by the Single Cell Imaging and Mass Cytometry Analysis Platform (SCIMAP) of the Goodman Cancer Research Centre, McGill University, according to their guidelines. Areas of dimension 1,000 × 1,000 μm were acquired for 23 sample cores. The resulting data files were stored in MCD binary format.

Software (GraphPad) was used to determine statistical significance of differences. Normality of data was evaluated by using the Shapiro–Wilk test. Normal data were interpreted using unpaired Student t test, one-way ANOVA followed by the Tukey post hoc test for multiple comparisons or two-way ANOVA followed by the Tukey post hoc test for multiple comparisons. Nonnormal data were interpreted using Mann–Whitney test. P values <0.05 were considered statistically significant. The details of statistical analysis for each experiment are listed in Supplementary Tables S6 and S7.

We have previously reported that the absence of phospho-eIF4E in both tumor and stromal cells is sufficient to reduce lung metastasis in the PyMT transgenic model of breast cancer (15). To dissect the importance of stromal phospho-eIF4E specifically in PPBC, we investigated whether stromal phospho-eIF4E deficiency is sufficient to block metastasis in a preclinical mouse model of this disease. Using the involuting mammary gland as an experimental platform to model PPBC metastasis, 66cl4 murine breast cancer cells were injected into the inguinal mammary glands of WT or eIF4E^S209A/S209A (phospho-eIF4E null, henceforth termed eIF4E^S209A) BALB/c mice one day following weaning-induced involution (Fig. 1A). Consistent with previously published data, tumor cells injected into the involuting mammary gland are more metastatic, compared with the same tumor cells injected into virgin mammary glands of age-matched mice (Supplementary Fig. S1A). In the PPBC model, we did not observe a difference in primary tumor outgrowth between WT and eIF4E^S209A PPBC mice, as both tumor initiation, growth, and weight at endpoint were similar (Fig. 1B; Supplementary Fig. S1B and S1C). Strikingly, we observed a significant decrease in lung metastases in eIF4E^S209A PPBC mice, that is, mice devoid of phosphorylated eIF4E, compared with their WT PPBC counterparts (Fig. 1C). Moreover, there were fewer tumor cells detected in the circulation of eIF4E^S209A PPBC mice on day 14 of tumor growth (Fig. 1D), suggesting a defect in tumor extravasation. Ki67 staining of lung metastases showed no difference in the percentage of Ki67-positive cells in the WT and eIF4E^S209A lungs (Fig. 1E), indicating similar tumor cell proliferation at the pulmonary metastatic site. Similar results were obtained when E0771 murine breast cancer cells, syngeneic to C57BL/6 mice, were injected into the involuting mammary glands of WT or eIF4E^S209A (Fig. 1F and G; Supplementary Fig. S1D–S1G).

Myeloid cells expressing CD11b (marker for myeloid cells) and Gr1 (granulocyte marker present in both Ly6G and Ly6C molecules) are known to increase at premetastatic niches and support metastasis (16, 26). Given the differences in the metastatic burden in the lung, but not primary tumor outgrowth, observed between WT and eIF4E^S209A PPBC mice, we focused on whether phospho-eIF4E deficiency at the lung metastatic site altered the infiltration of CD11b⁺, Ly6G⁺, Ly6C⁺ myeloid cell populations. We employed multicolor flow cytometry to immune phenotype the tumor-bearing lungs and discovered a significant reduction in granulocytes (CD45⁺CD11b⁺Ly6G⁺Ly6C^lo) and elevation of CD8⁺ T cells in the lungs of eIF4E^S209A PPBC mice at the experimental endpoint of the tumor model, 33 days after tumor injection (Fig. 1H; Supplementary Fig. S2 for gating strategies). However, the pulmonary levels of monocytic cells (CD45⁺CD11b⁺Ly6G⁻Ly6C^hi) were similar between WT and eIF4E^S209A PPBC mice (Fig. 1H). These differences in immune cell infiltrates arise as a consequence of tumor development, as similar changes were not detected in the lungs of age-matched naïve animals (Supplementary Fig. S1H). In conclusion, during tumor progression, phospho-eIF4E deficiency in the host reduces pulmonary recruitment of myeloid cells that support metastasis, increases the presence of CD8⁺ T cells, and impairs lung metastasis.

Mammary gland involution is characterized by the elimination of milk-secreting mammary epithelia and the repopulation of adipocytes. We next addressed whether the reduced metastatic burden in the lungs of eIF4E^S209A PPBC mice was due to a defect in their ability to undergo the physiologic process of mammary gland involution. We quantified the ratio of adipocytes over epithelial cells at lactation day 8, involution days 2, 4, and 6 in WT and eIF4E^S209A mice. The adipocyte/epithelium ratio increases in a similar pattern over the course of WT and eIF4E^S209A mammary gland involution (Supplementary Fig. S1I), and WT and eIF4E^S209A show similar gross morphology during involution (Supplementary Fig. S1J). Phosphorylation of STAT3 is known to be induced and required for mammary gland involution (27); thus, we also examined the levels of phospho-STAT3 in the WT and eIF4E^S209A mice, but found no difference in STAT3 phosphorylation (Supplementary Fig. S1K). Together, these results suggest that mice devoid of phospho-eIF4E undergo the physiologic process of involution similarly to their WT counterparts. Thus, the reduced metastasis observed in phospho-eIF4E null PPBC mice is not the result of overt defects in mammary gland involution.

Tumor cell invasion and metastasis are regulated to a large degree by molecular signals that can originate within the primary TME. We hypothesized that the reduction in metastatic colonies observed in the lungs of the eIF4E^S209A PPBC mice is reflective of a differential expression of such signals in the TME of phospho-eIF4E–deficient hosts. Recently, the IL33/ST2 signaling axis has been implicated as a potent modulator of the TME, regulating the recruitment and activation of immune cells as well as tumor cell invasiveness (28–31). We performed IHC on the 66cl4-derived primary tumors that were grown for 2 weeks either in WT or eIF4E^S209A PPBC mice, and found that IL33 levels are lower in the tumors grown in eIF4E^S209A PPBC mice, compared with those tumors derived from WT PPBC mice (Fig. 2A). Moreover, the expression of phospho-eIF4E correlates with IL33 expression in WT PPBC tumors (Fig. 2B).

Next, we sought to determine the cellular components in the PPBC tumors that produce IL33. Fibroblasts become activated during mammary gland involution, and they support PPBC invasion and metastasis, in part, via their active secretome (32). We thus sought to determine whether fibroblasts were a major source of IL33 in our PPBC model. We isolated primary fibroblasts from the mammary glands of WT and eIF4E^S209A mice, expanded them ex vivo, and analyzed their secretome. Mammary fibroblasts derived from eIF4E^S209A mice were found to secrete lower levels of IL33 compared with WT cells (Fig. 2C and D). We also exploited MEFs derived from WT or eIF4E^S209A mice as an additional tool to confirm that the phosphorylation of eIF4E was indeed required for the regulation of IL33 protein expression (Fig. 2E).

IL33 has been shown to directly impact invasion and metastasis via binding its receptor ST2, which is encoded by interleukin 1 receptor-like 1 (Il1rl1) on tumor cells (29, 33). Hence, we sought to determine whether fibroblast-derived IL33 positively supports breast tumor cell invasion. We used a coculture model system to study interactions between fibroblasts and the 66cl4 and E0771 breast cancer cells used in our in vivo PPBC models (Fig. 2F). When 66cl4 or E0771 breast cancer cells were cocultured with WT or eIF4E^S209A fibroblasts, both breast cancer cell lines displayed a decreased propensity to invade through a collagen I matrix in the presence of the eIF4E^S209A fibroblasts, as compared with WT fibroblasts (Fig. 2G; Supplementary Fig. S3A). We also observed a robust increase in the expression of fibroblast-derived IL33 mRNA and protein when we cultured breast cancer cells in the presence of fibroblasts; however, eIF4E^S209A fibroblasts still express significantly less IL33 mRNA and protein, as compared with their WT counterparts (Fig. 2H and I; Supplementary Fig. S3B and S3C).

As we observed that breast cancer cells display an increased propensity to invade toward WT fibroblasts compared with eIF4E^S209A fibroblasts, we next examined whether we could pharmacologically inhibit this process using the MNK1/2 inhibitor SEL201 (34). WT fibroblasts were treated with SEL201, and cocultured with either 66cl4 or E0771 cells. Concomitant with repressed phospho-eIF4E expression in fibroblasts, SEL201 treatment decreased IL33 protein levels in WT fibroblasts (Fig. 2J; Supplementary S3D). Moreover, the invasion of 66cl4 and E0771 cells was less robust when cocultured with SEL201-treated fibroblasts (Fig. 2K; Supplementary Fig. S3E).

Finally, we addressed the clinical relevance of our findings by coculturing patient-derived CAFs with MDA-MB-231 human breast cancer cells. We obtained primary CAFs that were isolated from the freshly resected human breast tumors of four patients. Primary CAFs were treated with vehicle or SEL201, and subsequently cocultured with MDA-MB-231. Similar to our findings in the murine fibroblasts, SEL201 repressed phospho-eIF4E and IL33 expression in patient-derived CAFs, and MDA-MB-231 invaded less robustly in the presence of SEL201-treated CAFs (Fig. 2L and M). Collectively, our data show that IL33 expression in fibroblasts is regulated by the MNK1/2–eIF4E axis.

Having shown the important role of fibroblast-derived IL33 in supporting breast cancer cell invasion, as well as the reduction of IL33 in the TME of eIF4E^S209A PPBC tumors, we next investigated whether fibroblast-derived IL33 acts via the IL33 receptor, ST2, expressed on 66cl4 cells, to promote breast cancer invasion. By ablating the expression of Il1rl1 using siRNA in 66cl4 cells, we observed an impaired ability of the ST2-deficient tumor cells to invade in the presence of WT fibroblasts (Fig. 3A; Supplementary Fig. S4A). Such data indicate that fibroblast-derived IL33 signals in a paracrine fashion to ST2-expressing breast cancer cells to augment tumor cell invasion. Therefore, we chose to further dissect how IL33 signals downstream of ST2 in breast tumor cells, focusing on the p38 and ERK1/2 MAPK signaling proteins, which lie immediately upstream of MNK1/2 activation (14). Stimulation of 66cl4 cells with recombinant murine IL33 (rIL33) resulted in increased phosphorylation of both p38 MAPK and eIF4E, but not phosphorylation of ERK1/2 (Fig. 3B; Supplementary Fig. S4B), and promoted 66cl4 (Fig. 3C) and E0771 invasion (Supplementary Fig. S4C). In addition, we hypothesized that IL33 might stimulate the expression of proinflammatory and protumorigenic cytokines/chemokines in tumor cells, which may further remodel the TME to favor invasion. rIL33 stimulated the mRNA expression of Cxcl1, Ccl17, Csf2 (which encodes GMCSF), and Il6, without significantly affecting Il4 and Cxcl2 levels (Fig. 3D). Finally, a main immune cell type whose expansion is reliant on IL33/ST2 signaling, and has been shown to play an important role in tumor immunity, are ILC2 cells (8). Using ex vivo expanded ILC2s from the bone marrow of BALB/c mice, we examined their cellular proliferation as well as their ability to secrete IL5 and IL13 in response to the costimulation with rIL7 plus rIL33 in the presence and absence of SEL201 (21). SEL201 treatment of ILC2s reduced their secretion of IL5 and IL13, in response to combined rIL7 and rIL33 (Fig. 3E). Although we observed minimal effects of SEL201 on the proliferation of ILC2 cells (Fig. 3F), their viability was significantly inhibited (Supplementary Fig. S4D). Together, these results suggest that blocking the phosphorylation of eIF4E in ILC2 cells ultimately negatively impacts their secretion of IL5 and IL13, which are important for the recruitment of CD11b⁺Gr1⁺ cells (7–10, 35–38). In toto, given the reported functions of CXCL-1, CCL-17, GMCSF, IL6, IL5, and IL13 in tumor immune evasion (36, 37, 39–47), our data provide evidence that IL33 may serve to create an immunosuppressive PPBC TME to facilitate metastasis in a MNK1/2–phospho-eIF4E–dependent way.

The expression of phospho-eIF4E positively correlates with IL33 protein level in 66cl4 tumors grown in WT PPBC mice (Fig. 2B). To interrogate whether these observations are clinically relevant, we used IF staining to evaluate the expression of phospho-eIF4E and IL33 in a cohort of PPBC patient samples. Consistent with our PPBC murine data, phospho-eIF4E expression correlates with IL33 levels in human PPBC tumors (Fig. 4A). To verify the broader implications of our observations, we examined The Cancer Genome Atlas (TCGA) data using UCSC Xena (http://xenabrowser.net/) for the relationship between MNK1 and IL33 expression. We found that MKNK1 and IL33 mRNA levels significantly correlate with one another (Fig. 4B).

The limited success of immune targeted therapy in breast cancer, relative to other malignancies, has been attributed in part to the heterogeneity of the breast TME. The TME of human PPBC has not been well defined, and we used CyTOF imaging mass cytometry (IMC) to simultaneously quantify the expression of 26 proteins within the TME of nulliparous breast cancer, breast cancer diagnosed during pregnancy (PrBC), and PPBC (Supplementary Table S8; Supplementary Fig. S5A). We report that PrBC and PPBC differ from breast cancer primarily in the relative proportion of tumor cells and immune cells, as well as their level of activation of the MNK1/2–eIF4E axis. Phospho-eIF4E, eIF4E, and MNK1 were detectable in tumor cells, immune cells, fibroblasts, pericytes, and endothelial cells (Fig. 4C; Supplementary Fig. S5B–S5D). In particular, the tumor cell population in PPBC and PrBC showed a significantly increased level of phospho-eIF4E expression, compared with the tumor cells represented in breast cancer (Fig. 4D). Furthermore, PD-L1 was expressed in tumor cells regardless of the cancer subtype, with PD-L1 expression being significantly increased in PPBC tumor cells compared with PrBC tumor cells (Fig. 4D). Moreover, when we examined tumor cells with the highest expression of phospho-eIF4E (99th percentile), the expression of PD-L1 was most abundant in PPBC, compared with breast cancer and PrBC (Supplementary Fig. S5E). CD8 is an important tumor immune biomarker, so we next investigated the proportion of CD8⁺ T cells in the three patient cohorts. We observed a significant increase in CD8⁺ T cells in the PPBC samples, and the phosphorylation of eIF4E is significantly increased in the CD8⁺ T cells present in PPBC, compared with breast cancer or PrBC samples (Fig. 4E). Further characterization of the phenotype of the CD8⁺ T cells showed that the coexpression of HLA-DR and PD-1 was significantly higher in PPBC samples, compared with CD8⁺ T cells present in PrBC or breast cancer (Fig. 4F). As HLA-DR is known to be expressed on activated T cells, and PD-1 is an exhaustion marker, our data suggest that the CD8⁺ T cells present in PPBC express an activated dysfunctional phenotype (48, 49), although we note that we cannot test the functionality of HLA-DR⁺PD-1⁺ T cells identified in the archived PPBC samples by IMC. These human data did, however, prompt us to profile the TME in the PPBC murine model early in tumorigenesis (i.e., day 14 after tumor cell injection). We observed that the recruitment of various immune cell subsets in the TME of the PPBC mice was not affected by phospho-eIF4E deficiency (Supplementary Fig. S5F–S5H). However, immune phenotyping using multicolor flow cytometry revealed that phospho-eIF4E competent WT PPBC mice contained fewer IFNγ- and CD107a-expressing tumor-infiltrating T cells (both activation markers), compared with phospho-eIF4E–deficient mice (Fig. 4G). In addition, the expression level of the T-cell costimulatory molecule CD86 was found to be lower on dendritic cells in WT PPBC mice compared with phospho-eIF4E null PPBC mice (Supplementary Fig. S5I). Thus, immune characterization of the TME in both human and murine PPBC tumors suggests an association between phospho-eIF4E activity in the TME and T-cell phenotypes.

The efficacy of immune checkpoint inhibitors in metastatic breast cancer, including PPBC, would likely be improved by overcoming tumor immune escape. We observed that the lungs of tumor-bearing phospho-eIF4E–deficient animals were infiltrated by higher numbers of cytotoxic CD8⁺ cells at day 14 of tumor growth, prior to overt metastasis (Fig. 5A; Supplementary Fig. S6A). Proportions of CD11b⁺Ly6C⁺ and CD11b⁺Ly6G⁺, as well as total CD3⁺ cells were unchanged at this time point (Supplementary Fig. S6B). Within the CD8⁺ population, in turn, more cells were positive for the IFNγ, a marker of T-cell activation, at day 14 of tumor growth (Fig. 5B), an effect that is not observed in the lungs of non–tumor-bearing animals (Supplementary Fig. S6C). These data suggest that phospho-eIF4E can contribute to PPBC immune evasion during the establishment of the metastatic niche. We thus hypothesized that SEL201 might sensitize PPBC mice to the antimetastatic effects of PD-1 blockade. To test this hypothesis, WT PPBC mice were administered vehicle, SEL201, anti–PD-1 antibody, or the combination of SEL201 plus anti–PD-1 antibody (Supplementary Fig. S6D). These drug treatments did not significantly affect tumor growth (Fig. 5C; Supplementary Fig. S6E). Intriguingly, SEL201 plus anti–PD-1 blockade decreased PPBC lung metastasis, while SEL201 or anti–PD-1 alone did not show any significant antimetastatic effects (Fig. 5D). However, SEL201 treatment alone resulted in a significant decrease in phospho-eIF4E expression, suggesting efficient target engagement by the MNK1/2 inhibitor, and a robust increase in CD8⁺ cells in the lung metastases (Fig. 5E; Supplementary Fig. S6F), which was not enhanced by addition of anti–PD-1. Administration of SEL201 or anti–PD-1 did not have any overt systemic toxicity (Supplementary Fig. S6G), consistent with our previous work (22, 50). Thus, targeting the MNK1/2–eIF4E axis might have therapeutic benefit for augmenting the efficacy of immunotherapy in women diagnosed with PPBC.

Metastasis associated with PPBC and mortality due to lack of effective treatment strategies necessitate a fuller understanding of this disease (51, 52). Recent breakthroughs in immune checkpoint blockade therapies have stimulated research to better understand the TME of breast cancer, aiming to discover possible approaches to sensitize metastatic breast cancer to immunotherapies. Here, we demonstrated the central role of stromal phospho-eIF4E in promoting protumorigenic immunity in a model of metastatic PPBC (graphical abstract). Collectively, our work highlights the important role of the IL33–MNK1/2–eIF4E axis in PPBC invasion and metastasis by impacting multiple cellular compartments in the TME. IL33 might hold potential as a therapeutically targetable cytokine in PPBC, and perhaps more broadly in breast cancer.

The regulation of IL33 by the MNK1/2–eIF4E axis is potentially relevant for immune cell function, as IL33 is known to reinforce protumorigenic inflammation (53). Within the immune cell compartment, IL33 has important effects on numerous cell types such as eosinophils, mast cells and ILC2s (4, 31, 54). IL33 is essential for the polarization of ILC2s together with IL7, IL25, and TSLP (55). Our understanding of the roles ILC2 plays in the context of tumor biology is still rudimentary, although recent studies have described their tumor-promoting and antitumor roles in several cancer types, including breast cancer (9, 10, 56, 57). In our study, we showed that the pharmacologic inhibition of MNK1/2 blocked the IL33-induced expression of ex vivo expanded ILC2-derived cytokines IL5 and IL13. We have yet to test whether hosts devoid of phospho-eIF4E, which presented with less granulocytic CD45⁺CD11b⁺Ly6G⁺Ly6C^lo cells in the lungs, is due to suppression of the IL33–ILC2–MDSC axis (10). In addition, our study has expanded the repertoire of IL33-induced proinflammatory cytokines and chemokines (i.e., Cxcl1, Ccl17, Il6, and Csf2) produced by tumor cells. The significance of these four factors in tumor immune evasion has been supported by multiple previous reports. For example, overexpression of CXCL-1 and its receptor CXCR-2, as well as elevated circulating IL6 levels, are all correlated to breast cancer metastasis and poor survival rate (58, 59), and CXCL-1, IL6, and GMCSF are all potent mediators for the recruitment and expansion of MDSCs and M2-like macrophages (41–43, 45). CCL-17, an important ligand for CCR-4, has also been demonstrated to elicit Th2 and T_reg-mediated cancer immune evasion (39, 44). Taken together, we show that IL33 acts directly on ILC2 and breast tumor cells to induce the expression of selected immunosuppressive chemokines and cytokines.

In addition to its impact on immune cells, IL33 has also been reported as a protumorigenic and proinvasive cytokine (28–30). Elevation of IL33 was observed in the serum of patients with breast cancer (60, 61). Moreover, the levels of matrix metallopeptidase 11 (MMP11), a proinvasive enzyme responsible for tissue remodeling, are directly correlated with IL33 levels in patients with breast cancer (61), supporting a possible proinvasive function of IL33. In this context, we have demonstrated that tumor-bearing phospho-eIF4E–deficient mice have fewer tumor cells in circulation. Disrupting IL33/ST2 signaling in vitro diminished the ability of cancer cells to invade. Thus, our data highlight the cross-talk between cancer cells and fibroblasts, whereby cancer cells educate fibroblasts to secrete more IL33, thus allowing breast cancer cells to gain more invasive properties and implicate the MNK1/2–phospho–eIF4E axis as the driver of this cross-talk.

Interestingly, in the preparation of this manuscript, a study was published linking IL33 derived from fibroblasts at the metastatic site to type II immunity, leading to a fecund microenvironment for cancer cell colonization (31). Their study showed that targeting IL33 can reduce metastasis in the 4T1 murine model of breast cancer, which is consistent with the results presented herein. It should be noted that in our study, we report a potentially important role of fibroblast-produced IL33 at the primary tumor site to facilitate tumor cell extravasation. Moreover, IL33 may give rise to systemic changes in cytokine production, directly or indirectly causing downstream effects on immune cell recruitment to the lungs. As we see increased expression of type I immunity activation markers in the T cells at both the primary site and lung in the phospho-eIF4E–deficient animals, our overall conclusions regarding protumor IL33 strengthen the potential of MNK1/2 inhibitors, IL33-targeting agents, and immunotherapy in clinical applications.

Immune checkpoint blockades designed to release the brakes on exhausted cytotoxic T cells have largely improved the patient prognosis in several cancers (62), but are less effective to date in breast cancer. It is proposed that many breast cancers display failed or suboptimal T-cell priming. Specifically in PPBC, increase in PD-1 expression on T cells and efficacy of anti–PD-1 treatment in reversing involution-associated tumor growth was recently reported (63). In line with this finding, we observe increased expression of PD-1 on the tumor-infiltrating CD8⁺ cells in patients with PPBC. We propose, although we have not evaluated this claim experimentally, that in the TME of PPBC tumors, these cells present a dysfunctional immunosuppressive phenotype. In this context, it is interesting that these CD8⁺ cells also express elevated levels of phospho-eIF4E. Although the role of phospho-eIF4E in regulating functionality of specifically CD8⁺ has not yet been explored by the scientific community and is beyond the scope of the current study, emerging data suggest that the activity of MNK1/2–eIF4E axis in immune cells affects their function. For example, a recent article reported that the immunosuppressive phenotype of bone marrow–derived macrophages is governed by the MNK1/2–eIF4E axis and can be reversed by MNK2 inhibition, indirectly leading to increased CD8⁺ cell activation (17). Furthermore, a preclinical study of the MNK1/2 inhibitor tomivosertib (eFT508) in liver cancer has shown that this inhibitor enhances the activity of checkpoint inhibitors in a T-cell–dependent manner, leading to an antitumor immune response (64). In summary, we propose that MNK1/2 inhibitors may convert “cold” breast tumors to “hot” tumors, thus offering the opportunity for immune checkpoint blockade to become more effective in highly metastatic cancers such as PPBC. This study, as well as others showing effects of MNK1/2 inhibition on cells of the TME (16, 18), provide strong preclinical support to ongoing clinical testing of MNK1/2 inhibitors in breast cancer. Indeed, we are participating in a Stand Up To Cancer trial to test this question (ClinicalTrials.gov ID: NCT04261218).

M. Basik reports grants from Canadian Cancer Society during the conduct of the study, as well as grants from LabCorp and Pfizer outside the submitted work. J.H. Fritz reports grants from Canadian Institutes of Health Research (CIHR) during the conduct of the study. W.H. Miller Jr reports personal fees from Merck, Bristol Myers Squibb, Roche, Novartis, Amgen, GlaxoSmithKline, Mylan, and EMD Serono, and grants from Merck and Bristol Myers Squibb (to institution) outside the submitted work. No disclosures were reported by the other authors.

Q. Guo: Conceptualization, formal analysis, supervision, investigation, methodology, writing–original draft. M. Bartish: Formal analysis, supervision, investigation, methodology, writing–original draft, writing–review and editing. C. Gonçalves: Conceptualization, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. F. Huang: Investigation. J. Smith-Voudouris: Investigation. S.S. Krisna: Investigation, methodology. S.E.J. Preston: Investigation. A. Emond: Formal analysis, investigation, methodology. V.Z. Li: Investigation. C.U. Duerr: Investigation, methodology. Y. Gui: Resources. A. Cleret-Buhot: Resources, investigation, methodology. P. Thebault: Formal analysis, investigation. H. Lefrère: Investigation. L. Lenaerts: Resources, data curation. D. Plourde: Investigation. J. Su: Investigation. B.C. Mindt: Investigation. S.A. Hewgill: Investigation. T. Cotechini: Resources, investigation. C.C.T. Hindmarch: Investigation. W. Yang: Investigation. E. Khoury: Investigation. Y. Zhan: Investigation. V. Narykina: Investigation. Y. Wei: Resources. G. Floris: Resources, investigation. M. Basik: Resources, supervision. F. Amant: Resources, supervision. D.F. Quail: Resources. R. Lapointe: Resources, supervision. J.H. Fritz: Resources, formal analysis, supervision, investigation, methodology. S.V. del Rincon: Conceptualization, resources, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing. W.H. Miller Jr: Conceptualization, resources, supervision, funding acquisition, investigation, project administration, writing–review and editing.

The authors thank Christian Young (LDI Flow Cytometry Facility), Kathy Forner and Veronique Michaud (LDI Imaging and Phenotyping Core), Dr. Naciba Benlimame, and Lilian Canetti for experimental advice and technical support. They thank Drs. Pepper Schedin (Oregon Health & Science University) and Qiuchen Guo (Harvard Medical School) for helpful discussion and experimental design.

This research is funded by the Canadian Institutes of Health Research (grants MOP-142281 to W.H. Miller Jr; PJT-156269 to W.H. Miller Jr, S.V. del Rincon, and J.H. Fritz; and PJT-162260 to S.V. del Rincon and J.H. Fritz) and the Cancer Research Society and CURE Foundation (grant 20239 to W.H. Miller Jr). This work was also supported by a grant from McGill Interdisciplinary Initiative in Infection and Immunity (Mi4) to S.V. del Rincon. The research was further supported by the Rossy Cancer Network. F. Amant is senior researcher for the Research Fund Flanders (F.W.O.). Development of the MNK1/2 inhibitor SEL201 was performed by Ryvu Therapeutics. Q. Guo was financed by a Cole Foundation PhD fellowship, a McGill Integrated Cancer Research Training Program (MICRTP) graduate studentship, and a Ruth & Alexander Dworkin PhD fellowship. M. Bartish is supported by an international postdoc grant from the Swedish Research Council (VR). F. Huang was sponsored by MICRTP graduate studentships. S.S. Krisna and S.E.J. Preston are supported by FRQS doctorate fellowships. V.Z. Li was supported by an MICRTP undergraduate studentship. H. Lefrère was supported by a grant from the Flemish Cancer Society (3M150537). W. Yang was endowed with an MICRTP graduate studentship.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.