A Subpopulation of Stromal Cells Controls Cancer Cell Homing to the Bone Marrow

In the bone marrow, hematopoietic stem cells need to maintain their stemness but also provide cells that differentiate to the various hematopoietic lineages (1). Epithelial cancer cells are detected in the bone marrow, and represent the so-called disseminated tumor cells (DTC; reviewed in ref. 2). These cells can similarly stay dormant or proliferate and form metastases. This raises the possibility that hematopoietic and cancer cells share common niches. Because two prototypes of epithelial cancers, namely breast and prostate are relatively common and associated with bone metastases much effort was spent on characterization of these niches (3). Increasing the number of niches, defined by the number of hematopoietic stem cells, led to augmented homing of cancer cells to the bone marrow (4, 5). In addition, cancer cells seem to compete with the stem cells for their niches, and use similar mechanisms as the hematopoietic stem cells to home to the bone marrow (5).

In the niche itself, both hematopoietic stem cells and cancer cells usually remain in a quiescent G₀ state and cancer cells may also exhibit a G₀ growth arrest after homing to the niche (6). This cellular dormancy protects the cancer cells from immune surveillance and chemotherapeutics, as the majority of chemotherapeutics target rapidly dividing cells (7). Therefore, some quiescent tumor cells can survive within the niches during and after treatment, where they later form local or even distant metastases, leading to the reported use of persistent cancer cells in the bone marrow as an independent predictor of recurrence in patients with prostate or breast cancer (8, 9).

The composition of the niche remains the subject of intense discussions. Osteoblasts in culture support survival, expansion, and differentiation of hematopoietic stem cells (10). In addition, transplanted hematopoietic stem cells engraft close to the endosteal surface (11), and cancer cells localize close to osteoblast-rich bone (12). Induced osteoblast deficiency led to a diminished number of hematopoietic stem cells (13), while stimulating the osteoblasts with intermittent PTH increased the number of cancer cells in the bone marrow (5).

The other major components of the niche are the vascular cells. Hematopoietic stem cells localize close to the sinusoid (14). Similarly to the localization of stem and cancer cells in proximity to osteoblasts, both leukemia and breast cancer cells engraft close to the endothelium in the bone marrow (15). Finally, some investigators have shown localization of stem cells to both osteoblasts and vascular cells (16). The question therefore still remains whether the osteoblastic and vascular niches are two distinct niches or just specified compartments with distinct features of one hematopoietic stem cell or premetastatic niche.

Besides osteoblasts and endothelial cells, mesenchymal stromal cells (MSC) may also contribute to the stem cell niche and support hematopoietic stem cells (17). Historically, MSCs were identified by their strong adhesion on plastic surfaces, clonal expansion in vitro (colony-forming-unit fibroblast, CFU-f), ability to differentiate into various cells of the mesenchymal lineage in vitro, and ability to reconstitute the hematopoietic microenvironment in vivo when transplanted subcutaneously (18). A benefit in supporting hematopoietic stem cells was also achieved when MSCs were cotransplanted with hematopoietic stem cells (19). Yet the phenotypical characterization of MSCs including the surface markers to be used remains controversial. Initial excitement and definition of several markers was quickly followed by the realization that the cells change their characteristics in culture (20). Some markers however are consistent with vascular stem cells or cells of vascular origin, for example CD45⁻CD146⁺ (21) or CD45⁻Ter119⁻Sca1⁺CD140a⁺ (22). Despite these advances much remains to be understood about the function of these cells and their characteristics.

The aim of this study was to investigate a possible role of MSCs in the premetastatic stem cell niche of the bone marrow and the consequences of their modification in cancer cell homing.

CD1-Foxn1^nu mice were obtained from Charles-River-Laboratories. Their T cells lack thymus schooling and are therefore nonfunctional. These mice are suited for xenotransplantation. Mice possessing the Mx1 promotor driving the cre recombinase and homozygous for the floxed β1 integrin gene (Mx cre β1^fl/fl) were used (23–25).

Animal studies were approved by the appropriate office for animal welfare and performed according to its guidelines (Regierungspraesidium Karlsruhe/Germany). The protocols carry the following numbers: G-139/09, G-44/12, G-73/13, G-102/14, G-186/14, G-276/14, G-6/15; G-110/15, G-180/15, G-275/15, G-6/16, G-1/17, G-111/17, G-135/17, G-139/17).

Human PTH 1-34 (MPI of Biochemistry, Martinsried) was administered for 4 days at 400 μg/kg/d subcutaneously and ZA (Hexal) intraperitoneally on days 1 and 3 of PTH treatment at 100 μg/kg. Animals received pravastatin (Hexal) for 4 days orally at 100 mg/kg/d, clodronate (Hexal) for 4 days at 50 mg/kg/d intraperitoneally, and G-CSF at 100 μL (6 × 10⁵ units) subcutaneously daily for 5 days.

MDA-MB-231B/luc+ or PC-3M-Pro4/luc+ were cultured in DMEM/10% FCS with 800 and 500 μg/mL geneticine respectively. These cells were obtained from M. Cecchini (26) in 2005 and labelled P1. They were used in passages ranging between 110 and 130, and applied within 4 weeks after thawing. Mycoplasma contamination was never detected in these lines. Cells were counted using an automated cell counter (CASY-TT, Innovatis).

Intracardiac injection was performed as described and 100,000 cells injected (27). CD34⁺ cells were injected at 500,000/mouse.

Intratibial injections were performed as described and 50,000 cells injected (28). Tumor cells were either injected alone or in combination with stromal cell populations at 1:2 ratio.

A total of 5 to 10 mL bone marrow were aspirated from the anterior iliac crests and tumor cell detection was performed based on the recommendations published by the German Consensus group of Senology.

To quantify MSCs in murine and human bone marrow, CFU-f were cultured in vitro. A total of 1 × 10⁶ cells were cultured in duplicates in six-well-plates in αMEM medium (Gibco) supplemented with heparin (1 unit/mL, HEXAL), 1% penicillin–streptomycin (Gibco) and 10% platelet lysate (IKT, Ulm, Germany—for production protocol, see ref. 29). CFU-f colonies with more than 50 cells were counted after 14 days using crystal violet staining.

For the enumeration of hematopoietic colonies, the MethoCult GF M3434 – Assay (Stemcell Technologies) was used according to the supplier's protocol.

Histomorphometry was performed as described (30). Sections were also used for the analysis of bone marrow vascularity.

Evaluation of homing of tumor cells was performed as published (27) using the resistance gene geneticine introduced in the MDA and PC3 cell line. CD34⁺ stem cells were detected using human alu primers and probe: forward: 5′CATGGTGAAACCCCGTCTCTA3′; reverse: 5′GCCTCAGCCTCCCGAGTAG3′; probe: 5′ATTAGCCGGGCGTGGTGGCG3′. Results were quantified using standard curves and normalized to mouse bone marrow cells using probe #64 and primers for β-actin (Universal probe library; Roche).

Bioluminescence imaging was performed as described (10, 28).

For MSCs, bone marrow was depleted of CD45⁺ cells. In a second step, Sca1⁺ cells were isolated using Dynabeads (Thermo Fisher), stained with CFSE (1:100; Biolegend) and injected in tail veins. As control, CD45⁺ cells were depleted of Sca1⁺ cells. The hepatocarcinoma cell line Huh-7 and cancer-associated fibroblasts (CAF) isolated from tumors depleted of murine immune cells (muCD45⁻, 30-F11) and human cancer cells (β2-microglobulin, B2M-01) using beads were also used. Mice received tumor cells after 24 hours and were euthanized after 48 hours.

All studies involving human material were approved by the appropriate committees and performed in accordance with the declaration of Helsinki. CD34⁺ cells were obtained from Cytotech and the protocol was approved by the human investigation committee/University of Heidelberg (S-645/15). Frozen bone marrow aspirates from patients with prostate or breast cancer were obtained at the University Hospital of Tuebingen or Dresden respectively, and approved by the appropriate human investigation committees (130/2016BO2, EK 237082012) after obtaining informed consent.

RNA analysis was performed as described (10, 28) and probes for HPRT #95 and IL1β #42 with primers as suggested by Roche universal probe library were used.

SDS-PAGE (10%) was performed and PTHR1 was detected using a rabbit polyclonal antibody (1:100; Biolegend, #906401). GAPDH was used as a loading control (1:10,000; Sigma, #SAB2100894).

Murine bone marrow was flushed from femur with DPBS and red cells lysed. Human biopsies were thawed on ice and washed with cold αMEM medium (Gibco). Cells were stained with the appropriate antibodies for 30 minutes at 4°C. The used antibodies are as follows (unless stated otherwise, obtained from Biolegend; clone in brackets): Annexin V; CD271 (polyclonal); CD44 (IM7); CD140b (APB5); CD144 (TEA1/31; Beckman;); CD146 (P1H12); CD235a (HI264); CD271 (ME20.4); CD31 (WM-59); CD349 (W3C4E11); CD45 (HI30); CD56 (5.1H11); CD73 (AD2); FAP (427819; R&D); MSCA1 (W8B2); Stro1 (Stro-1); CD11b (M1/70); CD16/32(93); CD29 (HMb1-1); CD31 (MEC13.3); CD34 (RAM34; BD Pharmingen); CD45 (30-F11); CD45R (RA3-6B2); CD5 (53-7.3); CD127 (A7R34); CD140a (APA5); CD144 (BV13); CD146 (ME-9F1); α-SMA (1A4; Sigma); ckit (2B8); Gr1 (RB6-8C5); Leptin receptor (polyclonal; R&D); Sca1 (E13-161.7); Ter-119.

For analysis of apoptosis, cells were stained with Annexin-V/PI for 15 minutes, fixed in 1% PFA for 10 minutes on ice and digested with RNAse (50 μg/mL, Fermentas) for 15 minutes at 37°C. Staining for flow cytometry markers followed. Flow cytometry was performed using LSR-2 (BD-Biosciences), and the BD FACSDiva Software.

HSPCs, more committed progenitor cells, and MSCs were isolated by flow cytometry (BD FACSAria). Bone marrow was depleted of lineage cells (B220, CD5, Ter119, Gr1, and CD11b) using Protein G Dynabeads and stained with ckit and Sca1. HSPCs are defined as c-kit⁺ Sca-1⁺ (LSK) and more committed progenitors as c-kit⁺ Sca-1⁻ (LKS-). MSCs were stained for CD45, Ter119, CD31, CD146, and CD44 and subjected to sorting.

Cells sorted and seeded at 2 × 10⁵ cells/well in 24-well-plates were treated with PTH (2.5–80 μmol/L) and ZA (2.5–10 μmol/L). Medium with PTH and ZA was removed after 1 hour and replaced with αMEM medium (Gibco) without any supplement. Apoptosis was assessed after 24 hours.

Analyses were performed using SPSS (V20). Analysis of variance and repeated measures analysis of variance tests were used as appropriate. If global probability values were smaller than 5%, subsequent comparisons between selected group pairs were then performed using Student t, Mann–Whitney, or Wilcoxon paired tests as appropriate. Survival was evaluated using the Kaplan–Meier method. Correlations were determined using the Pearson correlation coefficient. Stepwise multilinear regression models were calculated to find the best flow cytometry marker combination in vivo. Results are expressed as mean ± SEM.

To analyze the stem cell niche, osteoblasts and pericytes were stimulated with high-dose parathyroid hormone (PTH), the receptor of which is expressed on both cell types (31, 32). Because PTH also indirectly stimulates the osteoclasts, leading to bone resorption and release of the growth factors stored in the bone matrix, we aimed to prevent this confounding effect by inhibiting the osteoclasts using zoledronic acid (ZA; ref. 33).

Wild-type mice were treated for 4 days with daily injections of PTH at 400 μg/kg/d, a dose that prevents osteoblast apoptosis. This dose is higher and the duration is shorter than used to increase the number of hematopoietic stem cells (31, 34). Mice also received ZA at 100 μg/kg once on day 1 and once on day 3. This dose prevents bone loss in cancer (33). On the fifth day, 10⁵ human cancer cells were injected intracardially and the number of cancer cells in the bone marrow after 24 hours was evaluated using qPCR to detect the construct transfected in the human cells. This number was adjusted to the number of nucleated murine cells in the bone marrow.

Only the combination of PTH/ZA resulted in a doubling of homing of cancer cells to the bone marrow but not to other organs (Fig. 1A; Supplementary Fig. S1A). Neither the number of bone marrow blood vessels nor binding of cancer cells to endothelial or sinusoidal cells were affected (Supplementary Fig. S1B and S1C). The increase in homing was confirmed by using two different cancer cell lines (Breast cancer: MDA-MB-231 and prostate cancer: PC3), as well as using nonmalignant cells, namely human CD34⁺ stem cells, which were detected using the human alu sequence. Neither PTH nor ZA alone increased homing, however.

Enhanced homing was not linked to an increase in the number of osteoblasts, which were elevated only in the PTH-treated group, but diminished with the use of the combination of PTH/ZA or ZA alone, presumably due to ZA effects on osteoblasts (35). The change in homing could not be attributed to altered osteoclast numbers (Fig. 1B), or a change in hematopoietic stem and progenitor cells (HSPC), because neither the percentage of HSPCs nor the number of colony forming units that reflect early hematopoietic progenitors, cultured over 2 weeks in specialized media, called CFU-GEMM, or other types of hematopoietic CFUs were affected (Fig. 1C and D; Supplementary Fig. S1D). Instead, both the percentage and the absolute number of nonhematopoietic cells were diminished with combined PTH/ZA treatment. This was confirmed by quantifying the number of fibroblastic CFU-f (Fig. 1E and F).

In summary, the combination of PTH/ZA increased homing of cells to the bone marrow and concomitantly decreased nonhematopoietic cells, presumably MSCs.

We next aimed to further define the population changed in the presence of increased homing. To do this, we repeated the experiments of PTH/ZA treatments followed by evaluation of homing. Simultaneously, the bone marrow was stained with a variety of markers expressed on MSCs selected based on the literature and summarized in Supplementary Table S1A. These were Sca1, CD29, CD31, CD44, α-SMA, leptin receptor, CD146, CD140a, CD271, and PTH receptor type I. The cells were evaluated after exclusion of hematopoietic cells (CD45⁻Ter119⁻).

The expression of some, but not all markers was decreased on nonhematopoietic cells of PTH/ZA-treated mice (Supplementary Fig. S2A and S2B). Of those affected by PTH/ZA, we excluded leptin receptor (low expression), αSMA (intracellular staining required), and CD29 (expressed in more than 80% of bone marrow cells and for reasons shown later). Statistical evaluation of the remaining markers using multiple regression analysis (Supplementary Table S1B) confirmed a correlation between the number of cancer cells in the bone marrow and the subpopulation identified as CD45⁻Ter119⁻Sca1⁺CD31⁺CD146⁺CD44⁻. The obtained R² value of 0.979 was significant (Fig. 1G; Supplementary Table S1B; Supplementary Fig. S3A). Thus, the identified MSC subpopulation expresses a stem cell (Sca1), an endothelial cell (CD31), and a pericyte marker (CD146). In support of these findings, the number of CFU-fs also correlated with this MSC subpopulation (Supplementary Fig. S3B). The decrease in this MSC subpopulation was further confirmed in mice treated with PTH/ZA but not subjected to cancer cell injections (Supplementary Fig. S3C).

Consequently, we identified a subpopulation of MSCs that negatively correlate with the number of cancer cells that home to the bone marrow.

The combination PTH/ZA decreased a subpopulation of MSCs in the bone marrow. ZA increases cell apoptosis by inhibiting prenylation of proteins that depend on geranylgeranylation for their function (36). Because ZA is quickly incorporated within the bone matrix, it is released during bone resorption and ingested by the osteoclasts, leading to their apoptosis. We therefore evaluated whether inhibition of prenylation was responsible for our findings. The combination of PTH with a statin diminished MSCs and increased cancer cell homing (Fig. 2A). In contrast, inducing apoptosis in osteoclasts without affecting prenylation as occurs with clodronate administration neither affected MSCs nor homing (Fig. 2B; ref. 36). Thus, prenylation inhibition mediates the decrease in MSCs, without osteoclast involvement.

We next evaluated apoptosis in bone marrow cells. Both PTH/ZA und ZA increased apoptosis in total bone marrow, CD45⁺, and CD31⁺ cells (Fig. 2C; Supplementary Fig. S3D). In nonhematopoietic cells (CD45⁻Ter119⁻), CD146⁺ cells or the subpopulation of MSCs with the best correlation with homing, apoptosis only increased in mice that received the combination PTH/ZA (Fig. 2C; Supplementary Fig. S3D). This suggests that stromal cells and the subpopulation of MSCs is only susceptible to apoptosis induced by ZA in the presence of PTH. Indeed, PTH receptor I was detected on MSCs but not on hematopoietic stem cells (LSK: lin⁻ Sca1⁺ ckit⁺) or early hematopoietic progenitors (LKS⁻: lin⁻ ckit⁺ Sca1⁻), whereas some expression was detected in CD45⁺ cells (Fig. 2D). We therefore cultured sorted stromal CD45⁻Sca1⁺ and hematopoietic CD45⁺Sca1⁻ cells in vitro with a ZA concentration that does not induce apoptosis by itself in CD45⁻ cells (2.5 μmol/L). Addition of PTH led to higher apoptosis in the stromal CD45⁻ cells, but did not further enhance ZA-induced apoptosis in the hematopoietic CD45⁺ cells (Fig. 2E).

It has been reported that the sensitivity of smooth muscle cells to the proapoptotic effect of prenylation inhibitors increases in the presence of PTH due to upregulation of IL1β (37). We therefore evaluated IL1β mRNA expression and confirmed its increase in the bone marrow of PTH/ZA-treated mice (Fig. 2F).

Thus, the combination of PTH and ZA increases apoptosis and leads to diminished numbers of stromal cells.

We next asked whether the increase in homing will result in increased cancer growth in PTH/ZA-treated mice.

Evaluation of bone marrow 1 week after cancer cell injection confirmed the persistence of elevated cancer cells in the bone marrow of mice treated with PTH/ZA (Fig. 3A). As would be expected by enhanced homing in PTH/ZA-treated mice, early cancer growth presented as the number of lesions as well as total tumor burden was increased at 2 weeks (Fig. 3B). After 6 weeks growth was no longer increased, in line with published reports on the ability of ZA to inhibit bone metastasis growth, however (26). Indeed, no differences could be detected between PTH/ZA or ZA alone, except for the number of lesions, which remained higher in PTH/ZA-treated mice in line with a larger number of microfoci at the start of cancer development (Fig. 3C). Although survival improved with treatment with ZA alone, it was no longer significantly better with the addition of PTH to ZA (Fig. 3D). This shows that PTH/ZA increases homing and early cancer growth in the bone marrow.

Published reports suggest that cancer cells and hematopoietic stem cells compete for the same niches (5). Furthermore, G-CSF mobilizes hematopoietic stem cells out of the niches (38). We therefore administered G-CSF to mice and confirmed a decrease in HSPCs in bone marrow (Fig. 4A; Supplementary Fig. S4A). In contrast, G-CSF increased both the percentage and absolute number of MSCs (Fig. 4A; Supplementary Fig. S4A). Instead of the enhanced homing that would have been seen if cancer cells localized to the niches emptied from HSPCs, homing was diminished (Fig. 4A). Furthermore, administration of G-CSF in combination with PTH/ZA failed to diminish HSPC numbers in the bone marrow further, but diminished MSCs, leading to increased homing (the dotted line represents untreated CT, Fig. 4B; Supplementary Fig. S4B). Thus, PTH/ZA enhancement of homing is independent of a change in HSPCs.

Finally, deleting β1 integrin, which is expressed on MSCs, using the Mx promoter to drive cre expression in mice homozygous for a floxed β1 integrin gene succeeded in total bone marrow and stromal cells (which represent less than 15% of total bone marrow; Fig. 4C). This did not affect HSPCs, but increased MSCs and CFU-fs. Consequently, cancer cell homing was diminished (Fig. 4D; Supplementary Fig. S4C).

Taken together, these data support an inverse relationship between the subpopulation of MSCs and homing of cancer cells to the bone marrow in various models.

To establish causality between MSCs and cancer cell homing, we increased MSCs in the bone marrow by adoptive transfer. Wild-type bone marrow depleted of CD45⁺ cells was exposed to Sca1-antibody-coated magnetic beads to isolate Sca1⁺ cells from the negative fraction. Purity of the obtained populations was confirmed by flow cytometry (Fig. 5A). These cells were stained with CFSE in order to track them in the bone marrow. Wild-type mice were injected with increasing numbers of CD45⁻Sca1⁺ prestained cells (0.1–5 × 10⁶). This resulted in a stepwise increase in CFSE⁺ cells and MSCs in the bone marrow after 24 hours and also led to higher numbers of CFU-fs after culture of the bone marrow for 2 weeks (Fig. 5B).

We then evaluated the effect of administration of bone marrow CD45⁻Sca1⁺ cells on homing of cancer cells in comparison to 4 controls: untreated mice (CT), mice injected with hematopoietic CD45⁺ cells, fibroblasts isolated from tumors (after depletion of immune cells and cancer cells, CAFs) or a cancer cell line that does not home to bone marrow (Huh-7; ref. 27). The number of cancer cells in bone marrow 24 hours after intracardiac administration (or 48 hours after cell transfer) was diminished when either hematopoietic CD45⁺ or stromal CD45⁻Sca1⁺ cells were administered (Fig. 5C), suggesting that these two cell types share niches with cancer cells in line with our suggestion regarding MSCs and published reports on hematopoietic stem cell niches (5).

Because our data suggest a beneficial effect of bone marrow fibroblasts in preventing cancer cell homing, while a growth promoting role for cancer has been attributed to fibroblasts in the tumors (39), we evaluated whether bone marrow CD45⁻Sca1⁺ cells modified local tumor growth, and compared these to CAFs. Both bone marrow CD45⁻Sca1⁺ cells and CAFs were mixed in a ratio of 2:1 with cancer cells, and injected into the tibia to induce cancer lesions. As shown in Figure 5D, bone marrow CD45⁻Sca1⁺ cells diminished cancer growth. These data are in line with a weak inhibitory effect of bone marrow stroma on cancer growth.

To determine whether a similar relationship is found in humans, various MSC markers were first evaluated using cells isolated based on plastic adherence of total bone marrow aspirates in passages 0 and 1. The following markers were selected based on the literature (Supplementary Table S2A), and their expression confirmed: CD31, CD44, CD56, CD73, CD140b, CD144, CD146, CD271, CD349, FAP, MSCA1 and Stro1 (Supplementary Fig. S4D and S4E). The expression of these markers was then analysed in bone marrow aspirates obtained at the time of first operation after cancer diagnosis and frozen until analysis in two cohorts of patients with typical bone-metastasizing tumors (Supplementary Table S2B).

The first group consisted of 36 prostate cancer patients: 15 patients with no evidence of cytokeratin-stained cells were compared to 21 patients with cytokeratin⁺ cells in bone marrow aspirates (Supplementary Table S3A and Supplementary Fig. S5A). We detected an inverse relationship similar to that seen in mice but the panel differed and included CD271⁺CD31⁺ cells (Fig. 6A and B; Supplementary Table S2B). This was confirmed by quantifying CFU-fs in bone marrow cultures (Fig. 6C).

The second cohort consisted of 60 breast cancer patients; 30 without evidence of cytokeratin-stained cells and 30 harboring cytokeratin⁺ cells in the bone marrow (Supplementary Table S3B; Supplementary Fig. S5B). Because of the suggestion that CD146⁺, CD140b⁺, or CD144⁺ in single stained cells might be related to the absence of cancer cells in the bone marrow (Fig. 6D, not changed in the prostate cohort as shown in Supplementary Fig. S5A), we expanded the combination of markers in the breast cancer samples. The association between higher MSCs and CFU-fs and absence of cytokeratin-stained cells was confirmed (Fig. 6E and F), and was even more pronounced after inclusion of the three markers CD146, CD140b, and CD144 (Fig. 6E; Supplementary Table S2B). Thus, our MSC population expresses markers of endothelial cells (CD31, CD144, CD146) and pericytes (CD146, CD140b).

Some breast cancer patients received neoadjuvant chemotherapy by the time of surgery. Separating those who did from those who did not receive neodjuvant chemotherapy revealed that neoadjuvant chemotherapy did not influence the relationship between MSCs and the cytokeratin⁺ cells in bone marrow. The MSC subpopulation was decreased in samples containing cytokeratin⁺ cells from both patient groups (Supplementary Fig. S5C). Interestingly, the only two patients from the breast cancer cohort that developed metastatic visceral and bone lesions (marked with red circles in Fig. 6E) had no evidence of cytokeratin staining in bone marrow aspirates but had the lowest number of MSCs in this group using the expanded panel.

Based on these data, we conclude that cancer cell homing to the bone marrow is modified by MSCs expressing endothelial and pericytic markers in patients with prostate and breast cancer.

The principal finding of this study is that a subpopulation of MSCs prevents the homing of tumor cells to the bone marrow in mice. A related subpopulation is linked to the absence of disseminated cancer cells in the bone marrow of patients with prostate or breast cancer modifying the prognosis of the disease.

Even though MSCs are accepted as part of the hematopoietic stem cell niche in the bone marrow, further characterization of these cells has been difficult. Pending the development of a consensus, no specific markers can be viewed as better or worse (17). A correlation between various MSCs and CFU-fs has been shown (40). We therefore evaluated CFU-fs and were able to confirm that the decrease in the MSC subpopulations led to a drop in the number of CFU-fs in all mouse models as well as in the human samples.

Culture itself changes the expression of MSC markers and the behavior of the stromal cells, which then no longer necessarily represent the in vivo population (41). Therefore, in vivo characterization of MSCs is more revealing. The subpopulation we identified in vivo expresses an interesting combination of surface markers. The first marker Sca1 stands for stem cell antigen-1 is related to Ly6 expressed on leukocytes and is used for detection of HSPCs in combination with c-kit (42). The absence of CD44, which is the adhesive hyaluronic acid receptor mostly found on leukocytes speaks for a nonhematopoietic function of this cell population (43). In the human cells, CD271, also called low-affinity nerve growth factor receptor, was expressed. It has been linked to pluripotency of MSCs (44). Thus, both Sca1 and CD271 convey an undifferentiated state of the MSCs. The endothelial markers CD31 also called PECAM-1 in mice and humans and CD144 also called VE-cadherin in human samples suggest a vascular role of the MSCs. The remaining markers are mainly detected on pericytes, namely CD146 (also called MCAM, less than 5% of endothelial cells in bone marrow express it) and CD140b (the PDGF receptor β; refs. 21, 45–47). Some overlap between endothelial and pericytic markers, however has been reported (47) as was a mesenchymal to endothelial transition (48). In summary, the population we identified has stem cell characteristics and carries markers characteristic of vascular cells. This suggests that cells with vascular and possibly pericytic characteristics constitute part of the elusive niche to which cancer cells home. It is revealing in this context that CD146⁺ cells in the bone marrow seemed to represent part of the hematopoietic niche (21).

A possible explanation for enhanced homing in PTH/ZA-treated mice could have been increased porosity in the vascular barrier attributable to augmented apoptosis in the bone marrow of ZA-treated mice (49). However, boosted homing/infiltration of cancer cells in PTH/ZA-treated mice was only detected in the bone marrow and not in other organs including the liver and spleen, which also contain sinusoids (Supplementary Fig. S1A). Another explanation would be through loss of HSPCs, because tumor cells and hematopoietic stem cells compete for the same niche, and cancer cells might be even able to dislodge the hematopoietic stem cells out of their niches to make these their new home (5). The combination of PTH/ZA diminished the MSC subpopulation we identified by increasing apoptosis. Because HSPCs do not express the PTH receptor type I, however, the difference in homing of cancer cells cannot be directly attributed to a pharmacologically induced apoptosis in HSPCs and hence emptying the niches. This was confirmed by our data emptying the hematopoietic stem cell niches using G-CSF and showing a decrease in homing instead of enhanced homing. Furthermore, in the genetic Mx-β1^fl/fl model no alterations in the hematopoietic stem cells were detected. Instead, the change in cancer cell homing was opposite to the change in MSCs. Thus, in the models presented, we found that the number of MSCs correlated with the number of cancer cells that home to the bone marrow, whereas the number of HSPCs did not. We therefore propose that the subpopulation of MSCs we identified is part of the premetastatic niche.

The inverse relationship between MSCs and cancer cells could have been a mere association. The adoptive transfer experiments however confirm that less cancer cells can home to the bone marrow whenever the number of cells in the bone marrow is increased. It is tempting to speculate that decreases in various populations (MSCs or hematopoietic stem cells) leads to the formation of “holes” in the bone marrow that can be invaded by circulating cancer cells, which, depending on their proximity to specific bone marrow cell types can either stay dormant or proliferate. This raises an interesting possibility. Namely to attempt to increase the number of MSCs or hematopoietic cells in order to diminish cancer cell homing.

In patients, cells showing staining for cytokeratin, which is only expressed in epithelial cells (and hence in most types of prostate and breast cancer cells) are used as an indicator for the presence of cancer in the bone marrow. Therefore, detection of cancer cells in the bone marrow was evaluated in relationship to prognosis and found to be a predictor of relapse and/or poor prognosis (8, 9). Our findings in patients do not address the possibility that some tumors might release low numbers of cancer cells, resulting in diminished cancer cell homing to the bone marrow irrespective of the number of MSCs. Furthermore, cancer cells occasionally undergo epithelial–mesenchymal transition and hence no longer express cytokeratins making them undetectable in the bone marrow (50). Nevertheless, our data show a relationship between MSCs and cytokeratin-stained cells in the bone marrow in humans similar to the relationship between MSCs and cancer cell homing in mice. Whether MSC subpopulations might provide a reliable method for the prediction of metastases will have to be determined prospectively in larger cohorts of patients.

The new findings of this study highlight the role of a subpopulation of MSCs in early homing of tumor cells in mice and humans and offers new directions in early diagnosis of bone metastasis and in modifying the development of metastases.

F. Jakob reports receiving a commercial research grant from Novartis and has received speakers bureau honoraria from Amgen, Novartis, Lilly, and Alexion. T. Todenhöfer is a consultant/advisory board member for Amgen and Astellas. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Rossnagl, F. Jakob, S. Schott, P. Wimberger, I.A. Nakchbandi

Development of methodology: C. Groth, F. Jakob, I.A. Nakchbandi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Ghura, C. Groth, E. Altrock, S. Schott, P. Wimberger, T. Link, J.D. Kuhlmann, A. Stenzl, J. Hennenlotter, T. Todenhöfer, K. Bieback, I.A. Nakchbandi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Rossnagl, H. Ghura, C. Groth, F. Jakob, S. Schott, P. Wimberger, J.D. Kuhlmann, I.A. Nakchbandi

Writing, review, and/or revision of the manuscript: S. Rossnagl, F. Jakob, S. Schott, P. Wimberger, J.D. Kuhlmann, A. Stenzl, T. Todenhöfer, K. Bieback, I.A. Nakchbandi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Groth, P. Wimberger, A. Stenzl, J. Hennenlotter

Study supervision: P. Wimberger, A. Stenzl, I.A. Nakchbandi

Other (conception and design for characterization of mesenchymal stromal cells): M. Rojewski

We thank R. Fässler for his input. I.A. Nakchbandi (Max-Planck Society: M.KF.A.BIOC0001; German Research Council-DFG-: NA400/5, NA400/7, NA400/9); H. Ghura (Doctoral fellowship from the German Academic Exchange Service-DAAD).

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.