Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Nov 12;34(5):823-839.e7.
doi: 10.1016/j.ccell.2018.10.002.

The Osteogenic Niche Is a Calcium Reservoir of Bone Micrometastases and Confers Unexpected Therapeutic Vulnerability

Affiliations
Free PMC article

The Osteogenic Niche Is a Calcium Reservoir of Bone Micrometastases and Confers Unexpected Therapeutic Vulnerability

Hai Wang et al. Cancer Cell. .
Free PMC article

Abstract

The fate of disseminated tumor cells is largely determined by microenvironment (ME) niche. The osteogenic niche promotes cancer cell proliferation and bone metastasis progression. We investigated the underlying mechanisms using pre-clinical models and analyses of clinical data. We discovered that the osteogenic niche serves as a calcium (Ca) reservoir for cancer cells through gap junctions. Cancer cells cannot efficiently absorb Ca from ME, but depend on osteogenic cells to increase intracellular Ca concentration. The Ca signaling, together with previously identified mammalian target of rapamycin signaling, promotes bone metastasis progression. Interestingly, effective inhibition of these pathways can be achieved by danusertib, or a combination of everolimus and arsenic trioxide, which provide possibilities of eliminating bone micrometastases using clinically established drugs.

Keywords: bone metastasis; breast cancer; calcium signaling; drug discovery or repositioning; gap junctions; microenvironment; micrometastasis; prostate cancer; the osteogenic niche; therapeutic responses.

Figures

Figure 1.
Figure 1.. Ca signaling is enriched in bone metastasis compared to metastases in other organs
(A) An overview of bioinformatics analysis leading to Ca signaling. GSVA: Gene Set variation analysis. FDR: False discovery rate. Both p values and FDR were computed by the Statistics of Microarray (SAM) algorithm. MSigDB C3 transcription factor targets (TFT) database was analyzed. (B-D) Boxplots show the expression of a NFAT signature (B), an MEF2 signature (C), and MECP2 (D) across human BCa metastases in different organs (GSE14020). E. Similar to B-D, except that human PCa metastases (GSE77930) is examined. (B-E): p values by one-way ANOVA. For all boxplots: line inside the box: median value; top/bottom of the box: 75th/25th percentile; upper/lower whiskers: largest/smallest values no further than 1.5 × inter-quartile range from the hinge; data points beyond whiskers: possible outliers. Same for the boxplots in other figures. See also Figure S1.
Figure 2.
Figure 2.. Experimental models recapitulate the enhancement of Ca signaling in cancer cells interacting with the bone ME.
(A) Representative IF staining of MeCP2 in orthotopic tumors and bone lesions. Keratin 8 (K8) marks cancer cell cytoplasm; DAPI marks nucleus; translocalization is highlighted by single channel images; Scale bar, 25 µm. (B) Boxplots show the expression of the NFAT signature, the MEF2 signature and MECP2 in in vivo bone lesions (IVBL, n=8) or orthotopic tumors (Orth, n=7) derived from MCF7 cells. (C) IF staining of NFAT is shown in mammospheres of MCF7 cells alone or heterotypic organoids of admixed MSCs and MCF7 cells to mimic bone ME. Scale bar, 25 µm. (D) Western blot of indicated (phospho-)proteins in 3D cultures of cancer cells or OBs either alone or admixed. The slight shift between phosphor-NFAT (pNFAT) or non-phosphor-NFAT (npNFAT) is highlighted by altering the display width/length ratio of the gel shown as an additional row. (E) Similar to (C) except that MeCP2 staining is shown. (F) Western blot shows total (TP) or S421-phosphorylated MeCP2 in MCF7 cells with or without MSCs in 3D co-cultures. Firefly luciferase (Fluc) was used as loading control for cancer cells. Samples of (C-F) were harvested after mono- or co-culture for 48 hr. (G) A schematic illustration of BICA platform. (H) Bioluminescence intensity (BL) in BICA of MCF7 cells expressing empty vector, dominant negative CaN (DN-CaN), dominant negative CaMKII (DN-CaMKII), and both (DN-CaN+DN-CaMKII). n=12 bone fragment for each group. (I) Similar to (H) except that prostate cancer cell lines C4-2 and 22Rv1 were used. (J) Bone colonization curves of IIA-injected MCF7 cells, either expressing empty vector or DN-CaN+DN-CaMKII. n=6 and 7 for the two groups, respectively. (K) Represented BL pictures of mice in the vector and DN-CaN+DN-CaMKII (DN) groups. (L) Quantitation of ex vivo BL signals emitted from extracted hind limb bones. Representative pictures are shown. (M) Representative IHC staining of bone lesions derived from (J). Tumor lesions were indicated by K8 (brown) staining. Scale bar, 100 µm. p values by one-way ANOVA (B), LSD (H), Student’s t tests (I, and L) or repeated measures ANOVA test (J). Error bars = S.D.(H, I, L) or S.E.M.(J) See also Figure S2.
Figure 3.
Figure 3.. Ca2+ flows from osteogenic cells to cancer cells.
(A) A schema of MCF7-OB co-culture assay with GCaMP indicating alteration of [Ca2+] in cancer cells (left) and representative intensity curves of GCaMP under the indicated four conditions (right). (B) GCaMP intensity in MCF7 cells (alone or co-cultured with indicated cell types) as a function of different cell types in co-cultures and medium [Ca2+]. Bars show GCaMP intensity in MCF7 cells in medium with [Ca2+]=0 mM (solid) or [Ca2+]=2 mM (hashed). MSC: mesenchymal stem cells. (C) Fold change (FC) of MCF7 cell quantity in 3D co-cultures with indicated cell types as compared to MCF7 alone. MCF7 cells were quantitated by BL. Error bars = S.D. (n=3 technical replicates). (D) A scatter plot shows the correlation between increase of [Ca2+] as indicated by changes of median GCaMP intensity (MGI) and the FC of MCF7 cells in co-cultures. In (B-D): RAW: RAW264.7 cells, osteoclast precursors. dU937: differentiated U937 cells, osteoclasts. HMF: human mammary fibroblast cells. 3T3: NIH3T3 cells, mouse embryo fibroblasts. Pearson correlation coefficient and p value (by Student’s t test) are shown. See also Figure S3.
Figure 4.
Figure 4.. GJ mediates the Ca2+ flow from osteogenic cells to cancer cells.
(A) A schema of experiment design and key reagents and molecules (left) and the alteration of median GCaMP intensity by co-cultured OBs as a function of treatment of various inhibitors (right). The co-culture was performed in Ca2+-free medium. GJ: GJ. VSCC: voltage-sensitive Ca channel; SOCE: store-operated Ca entry; ROCE: receptor-operated Ca entry; CBX: carbenoxolone; MEFL: mefloquine; NIFE: nifedipine; SKF: SKF96365; Veh: vehicle control; MGI: Median GCaMP intensity. (B) Fold change (FC) of MCF7 quantity, quantitated by BL, in the same experiment as (A). Error bars = S.D. (n=3 per group). p value by one-way ANOVA. (C) Changes of GCaMP intensity in MCF7 cells (left) and OBs (right) in monocultures, co-cultures, and co-cultures with 10 μM CBX. Data are based on two separate experiments: GCaMP+ MCF7 + GCaMP OB, and GCaMP+ OB + GCaMP MCF (D) Western blot of indicated (phosphor-)proteins in MCF7 cells with or without 10 µM CBX treatment after mono- or coculture for 4 hr. The gel-shift of NFAT1 is highlighted in additional rows. (E) Expression of NFAT target gene set, MEF2 target gene set, and MECP2 in fibroblasts, MSCs and myoepithelial cells when they are co-cultured with various cancer cell lines. GSVA scores were calculated for NFAT and MEF2 target gene sets and re-scaled linearly between 0-1. Analyses were based on GSE27120. p: by Student’s t tests between indicated groups. *p: by one-way ANOVA across all groups. (F and G) Calcein transfer through GJs in co-culture of MCF7 cells and OBs in the absence (F) and presence (G) of 10 µM CBX. Negative control (no calcein) was indicated by grey area. The direction of transfer and percent positive recipient cells are indicated. See also Figure S4.
Figure 5.
Figure 5.. Connexin 43 (Cx43) is the major constituent of GJ.
(A) IF co-staining of K8 and Cx43 (left) or K8 and ALP (OBs, right) in sequential sections of a microscopic bone lesion derived from IIA-injected MCF7 cells. Arrows in high magnification pictures indicate punctuated Cx43 staining. Scale bar, 10 µm. (B) Normalized expression of GJA1 in bone lesions versus orthotopic tumors (GSE110451). p value by Student’s t test. (C) A heatmap (left) showing the expression level difference of genes encoding connexin family proteins in five breast cancer cell lines and human mammary epithelial cells (HMEC) upon co-culture with OBs (GSE29036) and the paired Cx43 expression (right) between mono- and co-cultures of these cells. p value by paired t test. (D) The expression of GJA1 across BCa metastases in different organs (GSE14020, top) and in bone and brain metastases divided based on ER and Her2 statuses (bottom). Sample size of each group is shown in parentheses. p value by one-way ANOVA. (E) Similar to (D) except that human PCa metastases (GSE77930) is examined (F) Scatter plots showing the correlations between GJA1 expression and the NFAT signature, the MEF2 signature, and MECP2 expression in bone metastases of BCa and PCa. All values were linearly re-scaled to between 0 and 1. Pearson correlation coefficients and p values by Student’s t tests are shown. See also Figure S5.
Figure 6.
Figure 6.. Inhibition of Cx43 and GJs retarded bone colonization.
(A and B) The effect of erexpressing DN-Cx43 on MCF7 (A) and C4-2 (B) cell growth in BICA. n=12 bone fragment for each group. Representative BL images are shown on the right. p value by Student’s t test. (C) The effect of inducible condition knockout (KO) of Gja1 in Osx+ lineage cells on MCF7 cell growth in BICA. The experimental group, Osx1-GFP::Cre;Gja1f/f (Osx1-cre;Gja1f/f) animals with doxycycline withdrawn (KO induced), are compared to control groups, Osx1-GFP::Cre (Osx1-cre) and Osx1-GFP::Cre;Gja1f/f on doxycycline. n=16 for each group. Representative BL images are shown on the right. p value by one-way ANOVA. (D) A schematic illustration of experiments assessing the effects of inducible overexpression of WT-Cx43 or DN-Cx43 on bone colonization. (E) Growth curves of IIA-injected MCF7 cells with indicated genetic perturbation in hind limbs. Error bars = S.E.M. n=5 for each group. p value by repeated measures ANOVA test across three groups. Representative BL images are shown on the right. (F) Normalized, log2-transformed BL intensity of hind limb bones extracted from the three groups described in (D) and (E). p value by LSD test. (G) A schematic illustration of experiments assessing the effects of CBX/MEFL treatment on bone colonization of MCF7 cells. (H) Growth curves of IIA-injected MCF7 cells in hind limbs. CBX or MEFL treatment last for 4 weeks as indicated by the arrow. Error bars = S.E.M. n=5 for each group. p value by repeated measures ANOVA test across three groups. Representative BL images are shown on the right. Veh: vehicle. (I) BL intensity of hind limb bones extracted from the three groups described in (G) and (H). p value by LSD test. (J) A schematic illustration of experiments assessing the effects of CBX treatment on spontaneous bone metastasis of 4T1.2 cells. (K) Representative BL images of spontaneous metastases two weeks after orthotopic tumor resection. (L) BL intensity of hind limbs and lungs extracted from control and CBX-treated animals. n=7 for each group. p values by Student’s t tests. Error bars = S.D. except (E) and (H). See also Figure S6.
Figure 7.
Figure 7.. BICA-screening identified As2O3 and danusertib as candidate therapies to eliminate cancer cells interacting with osteoblasts.
(A) Schematic pipeline of multiplex drug tests using BICA. (B) Heatmaps showing the efficacies of As2O3 [100 nM] and danusertib [100 nM] on MCF7 and MDA-MB-361 cells in BICA. The color scale is based on proportion of proliferation (−1=100% cell number increase) and inhibition (1=100% cell number decrease). p value by LSD test. (C) Efficacies of As2O3 on indicated cancer cells as a function of dosage in 2D culture or BICA . BL intensity was measured on Week 2-3. Mean values of 6 bone pieces are shown for each dot. p value by repeated measures ANOVA test. (D) Efficacies of As2O3 [1.9 mg/kg] and danusertib [15 mg/kg] on spontaneous bone metastasis of 4T1.2 cells in syngeneic hosts. Treatment started immediately after orthotopic tumor resection. n=8 for each group. p value by one-way ANOVA test. Error bar = S.D. (E) Western blot shows the indicated (phosphor-)proteins under various (co-)cultures of MCF7 and MSCs, and treatment of 5 µM As2O3 or 100 nM danusertib. The gel-shift of NFAT is highlighted in an additional row. (F) IF staining of MeCP2 in MCF7 bone lesions treated with vehicle, danusertib or As2O3. K8 indicates cancer cells. Single-channel image of MeCP2 highlights cytoplasm or nucleus localization. Scale bar, 25 µm. (G) The effects of As2O3 or danusertib on [Ca2+] in MCF7-OB co-cultures as assessed by GCaMP intensity. The grey area indicates no-OB control. Alteration of median GCaMP intensity (∆MGI) was used to quantitate the difference and is shown as a bar graph on the right. (H) “Volcano plot” showing gene expression alteration induced by As2O3 treatment in MCF7 co-cultured with MSCs. GJA1 is indicated. (I) The effects of 5 µM As2O3 on relative GJA1 expression in cancer cells with or without co-culturing of murine MSCs, quantified by qPCR. GJA1 expression in mono-culture, vehicle treated MCF7 or MSC were normalized to 1. Log2-transformed fold change (FC) is plotted. Three biological replicates are included. p value by Student’s t tests. Error bar = S.D. Veh: vehicle; Danu: danusertib. See also Figure S7 and Table S1.
Figure 8.
Figure 8.. Adjuvant transient As2O3 treatment of reduced latent bone recurrence of ER+ breast cancer.
(A) Quantification (left) and representative BL images (middle) of in vivo bone colonization kinetics of IIA-injected MCF7 cells expressing empty vectors or CA-CaMKII treated with vehicles or As2O3 (n=6 for each group) and quantitation and representative BL images (right) of BL intensity (normalized, log2-transformed) of extracted bones. p value by repeated measures ANOVA (left) or LSD test (right). Error bars = S.E.M. (B) Growth curves of individual bone lesions in vehicle- and As2O3 groups (top) and Kaplan-Meier curves of the indicated groups (bottom). n= 10 for both groups. The arrow indicates duration of treatment. The end point is defined as the time point when hind limb signal intensity reaches half of the Day 0 value (Normalized bioL intensity > 0.5) and continues to increase at subsequent time points. p value by log-rank test. (C) Growth curves of individual bone lesions in vehicle-, everolimus-, and combined As2O3 and everolimus -treated groups (top) and Kaplan-Meier curves of different groups with indicated treatments (bottom). n= 8 for each group. The arrow, end point and p value are defined as in (B). (D) Top left: representative BL images showing high (H), median (M) and low (L) signals in each group. Top right: quantitation of BL intensity. Bottom: Representative IHC staining of bone lesions. Tumor lesions were indicated by K8 staining. Error bars = S.D. p value by LSD tests. Scale bar, 100 µm. (E) µCT scanning of hind limb bones of tumor-free animal and representative animals in different treatment groups in (C) and (D). Bone mineral density values are shown below the images. (F) Schematic model for interaction between osteogenic cells and cancer cells in bone colonization. FDA-approved drugs are indicated by purple font. Veh: vehicle; Eve: everolimus. See also Figure S8.

Comment in

Similar articles

See all similar articles

Cited by 11 articles

See all "Cited by" articles

Publication types

MeSH terms

Feedback