Hematopoietic Age at Onset of Triple-Negative Breast Cancer Dictates Disease Aggressiveness and Progression

Aging is associated with increased breast cancer risk (1), whereas young age at diagnosis is associated with higher recurrence rates (2). Triple-negative breast cancer (TNBC) accounts for about 15% of breast cancer cases and carries a poor prognosis due to lack of targeted therapies and limited success of chemotherapy (3). Older women with TNBC have better outcome than younger women for unknown reasons (4). Nevertheless, older patients tend to forego treatment due to co-morbidities, resulting in a poor outcome for these patients (5). Although women older than 60 years represent nearly half of new breast cancer cases (6), only 10% of patients enrolled in clinical trials are older than 65 years (7), thus complicating our understanding. Treatment of patients with TNBC therefore presents age-specific challenges that might be overcome if more were understood about the disease.

The prevailing view that accumulation of DNA damage over time translates into age-associated increased cancer risk (8) does not explain the fact that TNBC incidence is unchanged with age (4). It is likely that age-related changes to the non-neoplastic cells within the tumor microenvironment (TME) also influence disease progression. Indeed, aging has multiple effects on tissue homeostasis that have either pro- or antitumorigenic consequences (9–13). Nevertheless, very little is known about the aging TNBC microenvironment.

The bone marrow is an important source of hematopoietic cells that comprise the TME (14) and one of the major organs affected by systemic signaling in cancer (15). It is well established that hematopoiesis changes with age (16, 17), leading to altered effector cell production and lineage potential (18). Given the importance of bone marrow cells (BMC) to tumor progression (19), surprisingly little is known about how aging affects tumor-infiltrating BMCs.

Here, we investigated how age affects the composition and function of protumorigenic BMCs and their impact on TNBC progression. Consistent with clinical observations, we found that TNBC was more aggressive in young mice than in old mice. We identified age- and tumor-dependent molecular and functional changes to hematopoietic cells that impacted TNBC progression. Importantly, BMCs from young mice with TNBC were sufficient to activate a supportive TME and stimulate tumor progression in old mice. The results presented here indicate that hematopoietic age is an important determinant of TNBC progression.

HMLER-HR and BPLER transformed human mammary epithelial tumor cells were generated and maintained as described (20, 21). MCF7Ras cells (22) and GFP+ immortalized human mammary fibroblasts (hMF) were a generous gift from Dr. Robert Weinberg (Whitehead Institute, Cambridge, MA) and maintained in DMEM with 10% FCS (Hyclone) and penicillin/streptomycin (Sigma). All cell lines were validated as mycoplasma-negative and all appropriate human cell lines authenticated at the Molecular Diagnostics Laboratory at Dana Farber Cancer Center (Boston, MA).

Female NCR-NU were purchased from Taconic and FVB/NJ mice were purchased from Taconic and Jackson Labs (stock no. 001800). All experiments were conducted in accordance with regulations of the Children's Hospital Institutional Animal Care and Use Committee (protocol 12-11-2308R) and MIT Committee on Animal Care protocol (1005-076-08). Tumor cells were prepared in 20% Matrigel (BD Biosciences) and injected subcutaneously into mice (see Supplementary Information for cell numbers). Tumors were measured using calipers with volume calculated as 4/3 × π ×(radius³).

Granulin-dependent cancer-associated fibroblasts (CAF) and granulin-independent CAFs were generated as published (23, 24), respectively. For gene expression analyses, total RNA was isolated from CAFs using standard procedures (Qiagen) and hybridized to Affymetrix HG-U133A plus 2 arrays (samples run in triplicate). See Supplementary Information for details. Full data sets were deposited online: GEO GSE75333.

BMCs were isolated as reported (25) and suspended in sterile PBS containing 2% FCS and 0.01% NaN₃. FcγR II/III receptors were blocked using anti-CD16/32, and cells labeled with 7-aminoctinomycin D (7-AAD; BioLegend) and appropriate antibodies for 30 minutes at 4°C. Sca1+/cKit− cells were analyzed and/or isolated from total bone marrow using Canto II or FACSAria IIu/FACSDiva (BD Biosciences). Analyses were performed using FlowJo (TreeStar). Antibody details are listed in Supplementary Table S3.

Formalin-fixed, paraffin-embedded tissues were sectioned onto ProbeOn Plus microscope slides (Fisher Scientific) and immunohistochemistry performed as described (25). Images were captured with identical exposure and gain using a Nikon Eclipse Ni microscope and quantified using CellProfiler as previously described (23). Antibody details are listed in Supplementary Table S3.

Total RNA was isolated from BMCs with TRIzol (Invitrogen), purified using RNeasy Mini Kit (Qiagen) and 100 ng mRNA reverse transcribed (ProtoScript AMV First Strand cDNA Synthesis Kit; New England Biolabs). TaqMan Master Mix (Applied Biosystems) was used for real-time PCR on an ABI-7900HT quantitative-PCR system (Applied Biosystems). Samples were analyzed in triplicate for a minimum of 3 animals per group. Values were normalized to GAPDH expression and analyses performed using the ΔΔC_t method. Primer details are listed in Supplementary Table S4.

Whole blood was collected from mice in EDTA-coated tubes (VWR) and centrifuged at 1.5 × g for 8 minutes to isolate plasma. Osteopontin concentrations were determined by ELISA according to manufacturer's instructions (R&D) and analyzed using an ABI-7900HT plate reader (Applied Biosciences).

RNA was obtained using RNeasy Plus Micro Kit (Qiagen); 500 pg mRNA was used to generate biotinylated cRNA according to the Nugen Ovation Pico WTA system. Following fragmentation, 5 μg of cRNA was hybridized to the Affymetrix GeneChip Mouse Gene 2.0 ST microarrays and chips scanned using the GeneChip Scanner 3000 7G. See Supplementary Information and GEO GSE74120 for microarray analyses.

Data are represented as mean ± SEM and analyzed by one-way ANOVA and/or Student t test, with Kaplan–Meier curves for survival analysis using GraphPad Prism 6.0, unless otherwise stated. P < 0.05 was considered statistically significant.

To investigate the effects of aging on host factors that support cancer progression, we established models that mimic TNBC progression with age. We used human TNBC cells, termed BPLER, previously shown to form aggressively growing mammary tumors in young mice (21, 26). When injected into young (8–10 weeks) and old (>10 months) nude mice, BPLER tumors showed delayed onset (∼16 days) and slower growth kinetics in the old cohort (Fig. 1A). At the experimental endpoint (45 days), tumors from both cohorts formed with 100% incidence; however, the tumors from old mice were about 4-fold smaller (Fig. 1A).

Histopathologically, TNBC tumors from young mice resembled advanced adenocarcinoma. About 80% of tumors from old mice displayed areas of edema and necrosis, whereas only 20% of those from young mice had visible necrotic areas (Fig. 1B). In addition, the tumors from young mice had a higher mitotic index (P = 0.001; Fig. 1C).

We next assessed the TME by immunohistochemistry and image analysis, excluding necrotic areas. Staining for mouse endothelial cell antigen (MECA32) was significantly lower (P = 0.03) in the tumors from old mice (Fig. 1D). Tumor vascularization is facilitated by CAFs (22), which are α-smooth muscle actin (αSMA)-positive and associated with poor prognosis (27). Positive staining for αSMA was more than 50% lower (P < 0.001) in the tumors from old mice than in those from young mice (Fig. 1E).

Various factors expressed by cells within the tumor microenvironment are known to activate tumor-supportive myofibroblasts (14, 28). Specifically, the hematopoietic cell–derived secreted growth factor, granulin, induces formation of αSMA-rich stroma and supports outgrowth of indolent tumors in mouse models of TNBC (23). Therefore, we examined tumors for infiltration of granulin-positive cells and found that tumors from old mice had recruited significantly fewer (P = 0.008) granulin+ cells than those from young mice (Fig. 1F).

The incidence of luminal breast cancer (LBC) is higher in older women, whereas the incidence of TNBC does not correlate with age; nevertheless, both subtypes are more aggressive in younger women (4, 6). Therefore, we also tested a model of MCF7Ras LBC. Consistent with clinical observations, LBC tumors grew more robustly in young mice (Fig. 1G and H), were significantly more proliferative (P < 0.001), and had higher microvessel density (P = 0.016) than their old counterparts (Fig. 1I and J). Unlike TNBC, LBC tumors had increased SMA+ coverage with age (P < 0.0001), and granulin expression was negligible in both young and old mice (Fig. 1K and L).

Collectively, these results indicated that age significantly influences tumor latency, growth kinetics, histopathology, and microenvironment in these models of TNBC and LBC. Importantly, these findings were consistent with clinical observations that granulin expression is associated with TNBC but not LBC (23) and suggested that granulin expression may be a function of age in the context of TNBC.

To gain insights into recurrence rate differences that are commonly observed between young and older TNBC patients (2), we employed a model of TNBC progression that mimics early phases of metastasis when patients harbor indolent disseminated tumor cells in the periphery at the time of primary diagnosis. We previously reported that HMLER-HR cells (weakly tumorigenic breast cancer cells oncotype-matched to BPLER cells; ref. 29), which otherwise do not form lung or subcutaneous tumors in cancer-free mice, form adenocarcinomas when implanted into mice bearing distant BPLER tumors (25). Given the impact of host age on tumor growth and disease histopathology (Fig. 1), we hypothesized that the ability of BPLER tumors to systemically instigate outgrowth of distant tumors might be impaired in old mice.

To test our hypothesis, we injected HMLER-HR cells (“distant tumors”) contralaterally to BPLER tumors (“primary tumors”) in either young or old mice (Fig. 2A). We harvested tissues at different time points (28–92 days in young; 45–98 days in old) to obtain primary tumors of equivalent average size in both young (96.5 ± 18.1 mm³) and old (114.9 ± 13.9 mm³) cohorts (Fig. 2B, Supplementary Fig. S1A).

Distant HMLER-HR tumors in young mice formed with 100% incidence (8 of 8) when injected into TNBC-bearing hosts (Fig. 2C). However, despite similar primary tumor size, only 20% of the distant tumors (2 of 10) grew in the old cohort (Fig. 2C). The average size of the resulting distant tumors at the various experimental end points (Supplementary Fig. S1A) was significantly smaller in the old mice (∼3.8 mm³) than in the young mice (∼68 mm³; Fig. 2C). We made similar observations when we conducted an experiment over the course of 70 days (Supplementary Fig. S2A–S2C).

Systemic instigation of distant tumors is dependent on secretion of the cytokine osteopontin (OPN) into the circulation by TNBC tumors (25). Indeed, elevated OPN plasma levels are correlated with poor prognosis (30). To determine whether size-matched TNBC tumors had equal potential to systemically instigate growth of distant tumors in young and old mice, we measured circulating levels of human and mouse OPN (hOPN and mOPN, respectively). Plasma levels of host mOPN were not significantly different between cancer-free and tumor-bearing young or old mice (Supplementary Fig. S1B). As expected, plasma levels of hOPN were negligible in young cancer-free mice (Fig. 2D). Importantly, there were no differences in the plasma levels of tumor-secreted hOPN between young (14.3 ± 1.3 ng/mL) and old (16.5 ± 1.8 ng/mL) tumor-bearing mice (Fig. 2D), suggesting that the primary TNBC tumors should have had equal instigating potential in young and old mice.

We next analyzed primary and distant tumors for evidence of systemically acting factors that are known to impact the microenvironment during TNBC progression (15). Hence, we first stained tumors for granulin, which had appeared to be affected by age in our earlier experiments (Fig. 1F). While granulin-positive staining was abundant in both primary and distant tumors from young mice, these levels were about 75% lower in the primary tumors and about 60% lower in the distant tumors of the old mice (Fig. 2E and F).

Since granulin induces formation of reactive stroma (23), we stained tumor sections for αSMA. Unlike the primary BPLER tumors observed in our earlier experiments (Fig. 1), αSMA-positive staining in these size-matched primary tumors was only modestly reduced in the old mice (Fig. 2G and H). In the distant tumors, we observed a subtle yet significant increase in αSMA staining in the tumors from old mice relative to those from young mice (Fig. 2G and H).

Interestingly, the distant tumor indolence observed in old mice occurred specifically in the context of TNBC. HMLER-HR cells injected contralaterally to a skin wound formed tumors efficiently in both young (6 of 8 injections) and old (4 of 8 injections) mice, with no significant difference in average tumor volume observed after 28 days (Supplementary Fig. S2D and S2E). Likewise, systemic instigation of distant tumors was more effective in old mice with LBC than in those with TNBC (Supplementary Fig. S2F). As we had observed in our earlier experiments (Fig. 1), distant tumor progression in old mice with LBC appeared to occur in a granulin-independent manner (Supplementary Fig. S2G). Thus, when provided with protumorigenic systemic conditions such as those stimulated by LBC or wound healing, distant HMLER-HR tumors can grow efficiently in old mice.

Collectively, these results indicated that the ability of TNBC primary tumors to drive outgrowth of distant tumors was impaired in old mice. Moreover, histopathologic observations suggested phenotypic differences in stromal components not only between young and old mice with TNBC but also between different breast cancer subtypes.

The fact that the distant tumor microenvironment from old hosts with TNBC was abundant with myofibroblasts yet displayed significantly lower granulin positivity suggested that stromal desmoplasia had occurred in a granulin-independent fashion in this cohort. Indeed, CAFs are heterogeneous and arise via different mechanisms (27, 31).

We previously reported a gene expression signature of granulin-treated normal human mammary fibroblasts (hMF) that is similar to CAFs isolated from patient breast adenocarcinomas; these expression signatures were generated by comparing normal to granulin-treated hMFs (23). Here, we wished to know whether gene expression signatures differed between granulin-dependent and granulin-independent CAFs. We prepared granulin-dependent CAFs by treating hMFs with granulin, and granulin-independent CAFs by exposing the same hMFs to MCF7Ras LBC tumor cells (Fig. 3A; ref. 24). Importantly, both types of CAFs express αSMA and support primary tumor growth (22–24).

Gene expression profiling revealed significant differences between granulin-dependent and granulin-independent CAFs (GEO GSE75333). Most notably, the granulin-dependent CAFs expressed 29.1-fold higher levels of collagen-IV (Col-IV alpha-5 isoform), 11.8-fold higher levels of platelet-derived growth factor receptor (PDGFR) A, and 11.4-fold higher levels of PDGFRB than their counterpart granulin-independent CAFs (Fig. 3B). Immunohistochemical analysis confirmed increased Col IV and PDGFR in tumors from young relative to old mice (Fig. 3C). Moreover, stromal cells within tumors from young mice were highly proliferative, as measured by Ki67 staining, compared with those from old mice (Fig. 3D).

These findings were consistent with the notion that CAFs present in TNBC tumors from young mice had formed in a granulin-dependent fashion, whereas those observed in tumors from old mice likely formed in a granulin-independent manner.

We further investigated whether differential recruitment of granulin+ bone marrow–derived cells (BMDC) could explain differences in reactive stroma between young and old TNBC tumors. A subpopulation of Sca1+/cKit− hematopoietic progenitor cells expresses high levels of granulin and has been reported to infiltrate various tumors following mobilization from the bone marrow (32, 33). In particular, these BMDCs promote formation of desmoplasia and outgrowth of indolent tumors in mice with TNBC (23).

To determine if Sca1+/cKit− BMDCs recruitment into tumors was age-dependent, we first stained distant tumors for Sca1 and granulin as hallmarks of these specialized BMDCs (Supplementary Fig. S3A). While hematopoietic CD45+ cells were efficiently recruited to the distant tumors that formed in both young and old mice with TNBC (Supplementary Fig. S3A), the numbers of Sca1+ cells were considerably higher in young mice than in old mice (Fig. 4A). Specifically, after normalizing for the number of Large-T antigen (LgT+) tumor cells, Sca1+ cells were about 9-fold more abundant in these tumors (Fig. 4A). Likewise, granulin-positive cells were 3-fold more abundant in the tumors from young mice (Fig. 4A and Supplementary Fig. S3A).

As BMCs can be rendered protumorigenic in the marrow prior to their mobilization into the circulation and recruitment to tumors (19, 25), we analyzed BMCs of cancer-free and TNBC tumor–bearing young and old mice. We found that the numbers of Sca1+/cKit− cells in the bone marrow were increased about 2-fold with age (P = 0.018; Fig. 4B and C). These findings were consistent with reports that Lin−/Sca1+/cKit− progenitor cells accumulate in the marrow of elderly humans and mice (34). Interestingly, TNBC did not significantly affect these numbers in either young or old mice (Fig. 4C).

Both age- and tumor-dependent associations became apparent when we assessed granulin expression in bone marrow Sca1+/cKit− cells. First, granulin expression was increased about 2-fold in Sca1+/cKit− BMCs from young tumor-bearing mice compared with young cancer-free mice, consistent with our previous report (P = 0.045, Fig. 4D; ref. 23). Second, baseline levels of granulin in Sca1+/cKit− BMCs of cancer-free old mice were below half that of their young counterparts (P = 0.017; Fig. 4D). Interestingly, the levels of granulin did not increase upon tumor development in old mice as they had in the young cohort (Fig. 4D).

We observed similar age-dependent trends in numbers and granulin expression in Sca1+/cKit− BMCs in an immunocompetent Her2/Neu breast cancer mouse model (Supplementary Fig. S3B and S3C). However, unlike these mice or the mice with TNBC, granulin expression in Sca1+/cKit− BMCs was not different between young and old mice with LBC tumors (Supplementary Fig. S3D).

We also assessed whether the potential for Sca1+/cKit− BMC recruitment was impaired in old mice with TNBC, thus explaining why fewer of these cells were observed in tumors from old mice despite their enhanced numbers in the marrow. The colony stimulating factor 1 (CSF-1) and its receptor CSF-1R axis is a predominant mechanism of hematopoietic cell mobilization and recruitment under normal and pathophysiologic conditions (35–37). CSF-1R expression on the Sca1+/cKit− BMCs from old cancer-free mice was significantly reduced (P = 0.014) relative to those from the young cancer-free mice (Fig. 4E). Significant increases in CSF-1R expression were observed in both young and old tumor-bearing mice when compared with their respective cancer-free counterparts (Fig. 4E). Nevertheless, these levels in tumor-bearing young mice significantly exceeded those of old tumor-bearing mice (P = 0.025; Fig. 4E). No differences in Sca1+/cKit− BMC expression of CSF-1R were observed between young and old mice bearing LBC tumors (Supplementary Fig. S3E).

Collectively, these results indicated that despite increased numbers of Sca1+/cKit− cells in the marrow of old cancer-free mice, these BMCs failed to respond to TNBC breast tumors that normally induce their expression of granulin and CSF-1R in young mice.

Our results suggested that the tumor-promoting function of Sca1+/cKit− BMCs might be impaired in old mice. As noted earlier, Sca1+/cKit− BMCs become protumorigenic in the marrow prior to their mobilization; hence, BMCs can be tested for their tumor support function using in vivo assays (15). Therefore, we harvested BMCs from young and old mice bearing size-matched BPLER tumors (68.3 ± 18.3 mg for young and 75.0 ± 35.3 mg for old mice; P = n.s.), immediately admixed donor BMCs with indolent HMLER-HR cells and injected the mixtures subcutaneously into either young or old recipient mice (Fig. 5A).

As expected, BMCs from young BPLER-bearing mice promoted HMLER-HR tumor growth in young recipient mice whereby 80% of the mice had formed tumors by the experimental end point (90 days; cohort 3, Fig. 5B). These tumors appeared invasive, proliferative, and displayed collagen-rich stroma (Fig. 5C and Supplementary Fig. S4A). In contrast, BMCs from old TNBC-bearing mice were significantly less efficient at promoting tumor growth, with only 20% of the old recipient mice (cohort 1) and 25% of the young recipient mice having formed tumors by the experimental end point (Fig. 5B). Tumors from both of these cohorts had only moderate mitotic index and little evidence of reactive stroma (Fig. 5C and Supplementary Fig. S4A).

Strikingly, BMCs from young TNBC-bearing mice rescued tumor growth in the old recipient mice (cohort 4, Fig. 5B). In old mice, tumor incidence was 50% within the first 2 weeks of implantation and had reached 90% by the experimental end point (Fig. 5B). Final tumor mass was equivalent to that of the young mice (∼3-fold greater than those mixed with old BMCs) and these tumors appeared much like those from young mice with respect to invasion, proliferation, and reactive stroma (Fig. 5C and Supplementary Fig. S4A).

Granulin+ cells were abundant in the tumors from both young and old recipient mice (cohorts 3 and 4; Fig. 5D). These numbers were significantly reduced in tumors that formed after admixture with BMCs from old donor hosts in both young and old recipients (cohorts 1 and 2; Fig. 5D), consistent with our finding that granulin expression is not induced in the bone marrow of old mice with TNBC.

Most noticeably, αSMA-positive cells were abundant in the tumors that arose after admixture of young BMCs and implantation into young recipient mice (cohorts 3; Fig. 5E). When BMCs from old tumor-bearing mice were used for admixture, staining for αSMA was decreased about 3- to 5-fold in both young and old recipients (cohorts 1 and 2) relative to when young BMCs were used (cohort 3; Fig. 5E). Strikingly, BMCs from young TNBC-bearing mice promoted formation of αSMA-enriched stroma in old recipient mice (cohort 4) to a level similar to that of the young mice (cohort 3; Fig. 5E). We confirmed that tumor-promoting activity and appearance of αSMA+ stromal cells depend on the presence of Sca1+ cells in the donor bone marrow when we tested the protumorigenic function of other BMC subpopulations from young mice bearing BPLER tumors (Supplementary Fig. S5).

Similar results were obtained using another human tumor xenograft, MDA-MB-435, whereby BMCs from young tumor-bearing donor mice rescued HMLER-HR tumor growth in old recipients (Supplementary Fig. S4B). Hence, the fact that young BMCs promoted tumor growth, even in old hosts, suggested that BMCs are critical for TNBC progression and that hematopoietic age supersedes other aspects of physiologic aging during TNBC progression from a state of indolence to one of aggressive growth.

We next analyzed gene expression profiles corresponding to tumorigenic potential of the Sca1+/cKit− BMCs harvested from young and old cancer-free or TNBC tumor-bearing mice. Gene set enrichment analyses (GSEA) revealed biologic processes and signaling pathways that were age-dependent, TNBC-dependent, or both (Fig. 6; Supplementary Tables S1 and S2; GEO GSE74120).

First, we compared gene expression profiles of Sca1+/cKit− cells that are poised to respond to TNBC (i.e., from young cancer-free mice) to those that do not respond (i.e., from old cancer-free mice). We observed overall decreases in gene expression with age (Supplementary Fig. S6) and through GSEA found that many of these genes regulate known age-related pathways such as oxidative phosphorylation, metabolism, and DNA repair (Fig. 6 and Supplementary Table S2). Expression of angiogenesis-related genes was also reduced with age, perhaps revealing a previously unknown function of these BMCs and explaining reduced vascularization observed in old mice (Fig. 1).

Second, tumor-dependent differences in gene expression were observed in Sca1+/cKit− BMCs from tumor-bearing mice compared with cancer-free controls. Common pathways that decreased in both young and old tumor-bearing cohorts were mostly related to immunosurveillance and DNA repair (Fig. 6 and Supplementary Table S1). However, we also observed alterations to pathways unique to young or old mice (Fig. 6), suggesting that host age governs the manner in which Sca1+/cKit− cells respond to systemic signals from the tumor—in young mice, such alterations rendered BMCs tumor-supportive while in old mice, they did not. In support of this notion, while there were more than 40 differentially expressed genes in Sca1+/cKit− cells comparing tumor-bearing young to tumor-bearing old mice (Supplementary Fig. S7), GSEA did not reveal any uniquely tumor-dependent influences on the old BMC population (Fig. 6 and Supplementary Tables S1 and S2).

These analyses demonstrated that age determines the ability of Sca1+/cKit− BMCs to become protumorigenic in response to systemic signals imparted by TNBC. These findings also support the notion that expression profiling is useful for distinguishing BMCs that have tumor support function from those that do not.

Taken collectively, our results reveal the importance of age in governing the potential of bone marrow hematopoietic cells to promote TNBC progression and suggest that the microenvironment of TNBC is indeed different at the cellular and molecular levels with age (Fig. 7).

The incidence of LBC is higher in elderly patients than in young, whereas the incidence of TNBC does not significantly change with age (6, 38). Importantly, recurrence-free survival among all breast cancer subtypes, with the exception of Her2+ breast cancers, is lower in young patients than it is in older patients (4). In light of our findings, it is possible that TNBC is more reliant on age-dependent hematopoietic function than other breast cancer subtypes. In further support of this notion, we previously reported that BMCs from young mice bearing MCF7Ras LBC tumors (39) or PC3 prostate tumors (25) do not promote tumor growth.

Age-related gene expression profiles that correlate with protumorigenic potential or function of BMCs, as we have defined here, may be useful for screening bone marrow aspirates or blood samples. As such, it will be necessary to identify the homologous BMC population in human patients. In mice, the Sca1+/cKit− population is composed of different functional cell types, which all have human counterparts, including T- and B-cell precursors, nuocytes (33), a subset of CD25-expressing cells that expand with age (34), and plasmacytoid dendritic cells (pDC; ref. 33). While pDCs account for the vast majority of granulin expression in young cancer-free mice and humans (33), and pDCs have been implicated in tumor promotion (40), further examination is required to determine whether they play a role in TNBC progression with age. Moving forward, we will want to learn how tumor-promoting BMCs function in the context of the immune milieu. Indeed, age- and tumor-dependent changes to Sca1+/cKit− BMCs occurred in both nude and immunocompetent breast cancer models in our study.

Our analysis of a published stromal gene expression profile obtained from breast cancer patient tumors (GSE9014; ref. 41) revealed a negative correlation between age and stromal granulin for patients with TNBC (Pearson correlation r = −0.7; n = 7), with a weaker negative correlation across all breast cancers (Pearson correlation r = −0.17; n = 66). While this trend is provocative, small numbers of patients with TNBC prevent us from making solid conclusions about granulin+-infiltrating stromal cells in patient tumors. More information about stromal infiltrates and gene expression profiling of tumor-promoting hematopoietic BMCs with age are needed. The ability to detect tumor-supportive BMCs, using granulin expression as a biomarker for example, may thus help to identify both young and elderly patients most at risk for TNBC recurrence.

No potential conflicts of interest were disclosed.

Conception and design: T. Marsh, I. Wong, A. Barakat, S.S. McAllister

Development of methodology: T. Marsh, I. Wong, A. Barakat, Y. Qin, S.A. Stewart, S.S. McAllister

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Marsh, I. Wong, A. Barakat, Y. Qin, E. Alspach

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Marsh, I. Wong, J. Sceneay, Y. Qin, A. Sjodin, B. Nilsson, S.S. McAllister

Writing, review, and/or revision of the manuscript: T. Marsh, J. Sceneay, A. Barakat, S.A. Stewart, S.S. McAllister

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Marsh

Study supervision: T. Marsh, S.S. McAllister

The authors thank Mahnaz Paktinat, Boston Children's Hospital HSCI for flow cytometric assistance; Shannan Hui of the HSPH Bioinformatics Core, Harvard School of Public Health for computational analysis assistance; DF/HCC Research Pathology Core for tissue sectioning.

This work was supported by funds from Knut and Alice Wallengerg's Foundation grant 2012.0193 (B. Nilsson); NIH-NCI-R01CA151518 and American Cancer Society Research Scholar Award (S.A. Stewart); NIH-NCI-R21CA182614 and the Department of Defense CDMRP BCRP Era of Hope Scholar Award W81XWH-14-1-0191 (S.S. McAllister).

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.