Bone metastasis from prostate cancer can occur years after prostatectomy, due to reactivation of dormant disseminated tumor cells (DTC) in the bone, yet the mechanism by which DTCs are initially induced into a dormant state in the bone remains to be elucidated. We show here that the bone microenvironment confers dormancy to C4-2B4 prostate cancer cells, as they become dormant when injected into mouse femurs but not under the skin. Live-cell imaging of dormant cells at the single-cell level revealed that conditioned medium from differentiated, but not undifferentiated, osteoblasts induced C4-2B4 cellular quiescence, suggesting that differentiated osteoblasts present locally around the tumor cells in the bone conferred dormancy to prostate cancer cells. Gene array analyses identified GDF10 and TGFβ2 among osteoblast-secreted proteins that induced quiescence of C4-2B4, C4-2b, and PC3-mm2, but not 22RV1 or BPH-1 cells, indicating prostate cancer tumor cells differ in their dormancy response. TGFβ2 and GDF10 induced dormancy through TGFβRIII to activate phospho-p38MAPK, which phosphorylates retinoblastoma (RB) at the novel N-terminal S249/T252 sites to block prostate cancer cell proliferation. Consistently, expression of dominant-negative p38MAPK in C4-2b and C4-2B4 prostate cancer cell lines abolished tumor cell dormancy both in vitro and in vivo. Lower TGFβRIII expression in patients with prostate cancer correlated with increased metastatic potential and decreased survival rates. Together, our results identify a dormancy mechanism by which DTCs are induced into a dormant state through TGFβRIII–p38MAPK–pS249/pT252–RB signaling and offer a rationale for developing strategies to prevent prostate cancer recurrence in the bone.

Significance: These findings provide mechanistic insights into the dormancy of metastatic prostate cancer in the bone and offer a rationale for developing strategies to prevent prostate cancer recurrence in the bone. Cancer Res; 78(11); 2911–24. ©2018 AACR.

Prostate cancer can be effectively treated by surgery when the disease is localized. However, bone metastasis can occur decades after prostatectomy in some patients (1). Recurrence is likely due to reactivation of disseminated tumor cells (DTC) that had been dormant at the metastatic site in bone. Because most therapies mainly target proliferating tumor cells, tumor dormancy is one mechanism for tumor cells to relapse at a later time. How DTCs initially become dormant in bone remains largely unknown.

Metastasis is attributed to the interaction of tumor cells (seed) with the microenvironment of the metastasized organs (soil). Signaling in the bone microenvironment may regulate the switch between dormancy and proliferation of DTCs. Tumor dormancy may be mediated through different cell types present in the bone microenvironment. However, prostate cancer preferentially metastasizes to bone and has unique interactions with osteoblasts (2). The differentiation state of osteoblasts in the bone microenvironment may influence the switch between dormancy versus proliferation in DTCs. At present, little is known about prostate tumor dormancy in bone as this disease state is not observed in clinical settings until the tumors have exited dormancy and progressed into overt bone metastasis. Understanding the mechanism(s) leading to dormancy of metastatic prostate cancer in bone will provide a rationale for developing strategies to prevent prostate cancer recurrence in bone.

In this study, we used intrabone and intracardiac injection models and established that the bone microenvironment can induce prostate tumor dormancy. We found that secreted factors from differentiated osteoblasts can induce cellular quiescence in some prostate cancer cell lines, but not in others, suggesting heterogeneity in dormancy response of prostate tumors. We further identified that osteoblast-secreted factors, including GDF10 and TGFβ2, induce tumor dormancy through activation of the TGFβRIII–p38MAPK–phospho(S249/T252)–retinoblastoma (RB) signaling pathway and that inhibition of p38MAPK activation abrogates prostate tumor dormancy in bone.

Cell lines, antibodies, and reagents

Cell lines.

The following cell lines were used: Human prostate cancer C4-2B4 (gift from Robert Sikes, University of Delaware, Newark, DE, 2006; refs. 3, 4), C4-2b (gift from Leland Chung, Cedars-Sinai Medical Center, Los Angeles, CA, 2004; ref. 5), PC3-mm2 (gift from Isaiah J. Fidler, MD Anderson Cancer Center, Houston, TX, 2005), 22Rv1 (ATCC, 2015; refs. 6, 7), and BPH-1 (gift from Simon Hayward, Vanderbilt University, Nashville, TN, 2015); mouse osteoblast precursor MC3T3 (ATCC, 2006); human lung carcinoma A549 (ATCC, 2013). The identity of all cell lines was verified by polymorphic short tandem repeat loci (STR) profiling, and all cell lines are mycoplasma free.

Recombinant human proteins.

The following recombinant human proteins were used: TGFβ2, TGFβ1, and GDF10 (BMP-3b; R&D Systems).

Antibodies.

The following antibodies were used: phospho-p38MAPK, p38MAPK, RB, Smad2/3, phospho-Smad1/5, p27Kip1, GAPDH, β-actin (Cell Signaling Technology); phospho(S249/T252)-RB (8); Ki-67 (Dako); TGFβ2 and TGFβRIII (ProteinTech); GDF10, p21, and TGFβRIII (R&D Systems); and β-tubulin, α-tubulin (GeneTex).

Reagents.

The following reagents were used: mouse TGFβ2 (Genorise) and osteocalcin (Alfa Aesar) ELISA kits; p38MAPK inhibitor SB202190 (InvivoGen); and Vybrant DiO lipophilic membrane dye (Invitrogen).

Subcutaneous, intrafemoral, and intracardiac injection of prostate cancer cells

Luciferase-expressing prostate cancer cells were injected under the skin, into the femurs (9), or into the left ventricle (10) of male SCID mice. Tumor growth was monitored using bioluminescence imaging. Animal studies were performed in accordance with as well as approved by an Institutional Animal Care and Use Committee.

IHC

Tumor-bearing femurs were fixed in formaldehyde and decalcified with formic acid before being embedded in paraffin. IHC was performed as described previously (11).

Differentiation of primary mouse osteoblasts

Primary mouse osteoblasts (PMO) were isolated from calvariae of 2- to 5-day-old pups, cultured to confluence, and the medium switched to differentiation medium, which contains ascorbic acid and β-glycerol phosphate. Osteoblast conditioned medium (OB-CM) was concentrated and medium exchanged by Centricon (Millipore) centrifugation.

Gene array analysis

RNAs prepared from undifferentiated and differentiated PMOs were subjected to whole-genome microarray analysis (4 × 44K, Agilent Technologies; Arraystar, Inc). Array data are deposited in NCBI GEO under accession number GSE90127.

PCR

RNAs from PMOs were analyzed by qRT-PCR using mouse primers (Supplementary Table S1). Total DNA was prepared from mouse hind legs and used for real-time PCR using primers for human Alu-repetitive sequences (Supplementary Table S1). The number of tumor cells in bone was calculated on the basis of Alu PCR of a serial dilution of DNA from C4-2B4 cells.

Live-cell time-lapse imaging

Prostate cancer cells were plated in Hi-Q4 dishes (Ibidi) and cultured in RPMI1650 containing 1:20 dilution of days 0, 6, 24, or 30 OB-CM or in RPMI1640 containing 0.1% FBS with TGFβ2, GDF10, or TGFβ1. Images were acquired every 20 minutes for 72 hours in a BioStation (Nikon). Grid-500 glass-bottom dishes (Ibidi) were used for live-cell monitoring in more fields. Data were compiled using NIS-Elements (Nikon) software.

Immunofluorescence imaging

Following live-cell imaging, cells were fixed and permeabilized, coincubated with anti-Ki67 and anti-p27, and reimaged on the BioStation.

Proximity ligation assay

Proximity ligation assay (PLA) was performed using Duolink PLA In Situ Green Starter Kit (Mouse DUO92004/Rabbit DUO92004, Sigma). Primary antibodies were anti-phospho-p38MAPK (28B10) and anti-phospho-(S249/T252)-RB (8). Images were acquired using FluoView 1000 IX2 confocal microscopy (Olympus).

Generation of C4-2B4 cells with knockdown of TGFβRIII

TGFβRIII was knocked down by RNA interference via lentivirus-expressing shRNAs in pGIPZ. Clones C4-2B4-pGIPZ-sh-TβRIII #2 and #3 were generated. Antisense sequences for sh-TβRIII Clone #2: 5′-ATAGCTCCATGTTGAAGGT-3′ (NM_001195683) and Clone #3: 5′-ATAGTAGACCACACCATCA-3′ (MN_003243).

Generation of C4-2B4 and C4-2b cells with dominant-negative p38MAPK

C4-2B4 and C4-2b cells were transduced with retroviral-expressing p38α dominant-negative MAPK (p38DN), containing mutations in the activation loop between the two kinase domains, from Thr180-Gly-Tyr182 to Ala180-Gly-Phe182 (12), in a pBMN-I-GFP vector.

Human prostate cancer datasets

The Kaplan–Meier method with log-rank test was used to evaluate overall disease-specific survival curves for patients with prostate cancer from the Nakagawa dataset (13). Patients from the Taylor dataset (14) were used to evaluate prostate cancer metastasis. Patients from the Lapointe dataset (15) were used to evaluate prostate cancer metastatic progression. For computing TGFBR3 gene score based on expression profiling data from human prostate cancer tumors, TGFBR3 gene was first z-normalized to SD from the median across the primary tumor samples. The average of the z-normalized values was used to represent the score for each sample profile.

Statistical analysis

Data quantification was performed by using the Student t test and expressed as mean ± SEM. P values of <0.05 were considered statistically significant.

C4-2B4 becomes dormant when injected into bone but not subcutaneous sites

The LNCaP subline C4-2B4 cells (3, 4) labeled with luciferase and red fluorescence protein Tomato (C4-2B4-LT; ref. 11) were injected subcutaneously or intrafemorally into SCID mice. While C4-2B4-LT grew exponentially under the skin, C4-2B4-LT grew slowly in mouse femurs, as quantified by bioluminescence over 5 weeks (Fig. 1A) and by histologic analysis (Fig. 1B). We examined whether the limited tumor growth in bone might be due to increased cellular dormancy. Dormant cells have been characterized as nonproliferating, slow-cycling (16, 17), and Ki-67 negative (18). C4-2B4-LT cells showed strong Ki-67 staining in subcutaneous tumors, but weak, diffuse staining in femoral tumors (Fig. 1B). Furthermore, we examined the expression of G0–G1 markers, the p27 and p21 cell-cycle inhibitors (19, 20), in the tumors, and found p27 and p21 staining in C4-2B4 femoral tumors, but little to no staining in subcutaneous tumors (Fig. 1C). These results suggest that the bone microenvironment provides factors that induce dormancy of C4-2B4 cells.

Figure 1.

Osteoblasts in bone microenvironment confer dormancy on C4-2B4 tumor. A, C4-2B4-LT cells (1 × 106) were injected subcutaneously (subcu; n = 10) or into the femur of SCID mice (n = 10). Tumor growth was monitored by bioluminescence. Consecutive tissue sections were stained as follows: hematoxylin and eosin (H&E) stained or immunostained with anti-Ki-67 (×200; B); anti-p27 or anti-p21 (×400; C). Box, enlarged; T, tumor; B, bone. D, PMOs during differentiation in culture. Cell lysates were used for alkaline phosphatase activity, conditioned media for osteocalcin ELISA, and RNAs for qRT-PCR of mDMP1 and mSOST-1. E, Live-cell imaging (h:min) of C4-2B4 cells treated for 72 hours with day 0, 6, 24, and 30 OB-CM. Colored asterisks, day 0 progenies. Round cells are mitotic. Cells a–d, quiescent C4-2B4 cells in day 6, 24, or 30 OB-CM. Following time-lapse, cells were immunostained for Ki-67. Cell outlines are traced. All bars, 10 μm. F, Quantification of OB-CM–treated quiescent C4-2B4 cells that did not divide over 55–72 hours relative to total cells examined. Representative of two independent experiments.

Figure 1.

Osteoblasts in bone microenvironment confer dormancy on C4-2B4 tumor. A, C4-2B4-LT cells (1 × 106) were injected subcutaneously (subcu; n = 10) or into the femur of SCID mice (n = 10). Tumor growth was monitored by bioluminescence. Consecutive tissue sections were stained as follows: hematoxylin and eosin (H&E) stained or immunostained with anti-Ki-67 (×200; B); anti-p27 or anti-p21 (×400; C). Box, enlarged; T, tumor; B, bone. D, PMOs during differentiation in culture. Cell lysates were used for alkaline phosphatase activity, conditioned media for osteocalcin ELISA, and RNAs for qRT-PCR of mDMP1 and mSOST-1. E, Live-cell imaging (h:min) of C4-2B4 cells treated for 72 hours with day 0, 6, 24, and 30 OB-CM. Colored asterisks, day 0 progenies. Round cells are mitotic. Cells a–d, quiescent C4-2B4 cells in day 6, 24, or 30 OB-CM. Following time-lapse, cells were immunostained for Ki-67. Cell outlines are traced. All bars, 10 μm. F, Quantification of OB-CM–treated quiescent C4-2B4 cells that did not divide over 55–72 hours relative to total cells examined. Representative of two independent experiments.

Close modal

Conditioned media from differentiated mouse osteoblasts lead to quiescence of C4-2B4 cells

We hypothesized that some dormancy-inducing factors in the bone microenvironment are secreted from osteoblasts. Primary mouse osteoblasts (PMO) were cultured in differentiation media for 30 days. Osteoblast differentiation markers alkaline phosphatase and osteocalcin increased at 18 and 21 days, respectively (Fig. 1D; ref. 21). Increases in dentin matrix protein (mDMP1) and sclerostin (mSOST-1), two osteocyte markers (21), were also detected at days 24 and 30, respectively (Fig. 1D).

We prepared conditioned media from these differentiating osteoblasts (OB-CM) and examined C4-2B4 cellular response by live-cell imaging. When treated with nondifferentiated day 0 OB-CM for 72 hours, the majority of C4-2B4 cells underwent several cell divisions with approximately 10% of cells nonproliferating or slow-cycling (Fig. 1E and F). When treated with day 6, 24, or 30 OB-CM, approximately 20%–35% of cells were nonproliferating (Fig. 1E and F), with little cell death (∼1%) observed. Quiescent cells exhibited a low level of movement with cell shape changes observed during time-lapse (Fig. 1E, cells a–d). While proliferating cells were Ki-67 positive, OB-CM–treated nonproliferating C4-2B4 cells were Ki-67 negative (Fig. 1E, day 0 vs. days 6–30). These observations suggest that factors in OB-CM from differentiated osteoblasts are capable of inducing cellular quiescence of C4-2B4 cells.

TGFβ2 and GDF10 are upregulated during osteoblast differentiation

To identify the dormancy factors from differentiated osteoblasts, we performed a gene array analysis using RNAs prepared from undifferentiated (day 0) versus differentiated (day 30) osteoblasts. The levels of 1844 and 1873 mRNAs were upregulated and downregulated with >2-fold change, respectively (see GSE90127). A total of 194 genes encoding secretory proteins, including chemokines, cytokines, and peptide growth factors, were upregulated in differentiated PMOs (Supplementary Table S2). Selected factors were further categorized into three groups based on P values (Fig. 2A; Supplementary Table S3). The P values were determined from duplicate samples used for the gene array analysis. Some of the factors have high fold of differences, but less consistent fold changes in duplicate samples, and this has resulted in less significance in their P values. In group 1, increases in CXCL12, GDF10, TGFβ2, CCL4, and CCL12 levels were highly significant, with P < 0.001 and false discovery rate (FDR) <0.15. Notably, the expression of TGFβ2, a factor shown to induce dormancy in various cancers (16–18, 22), was increased 2.5-fold with P = 0.0008 (FDR < 0.14) in differentiated osteoblasts. GDF10 (BMP-3b) is a member of the TGFβ family and its role in dormancy is unknown.

Figure 2.

Gene array analyses of osteoblasts before and after culturing in differentiation medium for 30 days. A, RNAs from undifferentiated (day 0) or differentiated (day 30) osteoblasts were used in a whole-genome microarray analysis. Forty-two transcripts encoding secreted factors upregulated in day 30 osteoblasts were subdivided into three groups, based on P values and FDR. Fold increase is shown on right. B, RNA levels of select TGFβ and GDF family genes during osteoblast differentiation. C, TGFβ2 or GDF10 expression by qRT-PCR (top) or Western blot (bottom). Undifferentiated osteoblast precursor MC3T3-E1 cells were used as a control. D, TGFβ1 RNA levels. *, P < 0.01 by t test. E, TGFβ2 and GDF10 were two of many factors upregulated and secreted by osteoblasts over 30 days of differentiation in culture.

Figure 2.

Gene array analyses of osteoblasts before and after culturing in differentiation medium for 30 days. A, RNAs from undifferentiated (day 0) or differentiated (day 30) osteoblasts were used in a whole-genome microarray analysis. Forty-two transcripts encoding secreted factors upregulated in day 30 osteoblasts were subdivided into three groups, based on P values and FDR. Fold increase is shown on right. B, RNA levels of select TGFβ and GDF family genes during osteoblast differentiation. C, TGFβ2 or GDF10 expression by qRT-PCR (top) or Western blot (bottom). Undifferentiated osteoblast precursor MC3T3-E1 cells were used as a control. D, TGFβ1 RNA levels. *, P < 0.01 by t test. E, TGFβ2 and GDF10 were two of many factors upregulated and secreted by osteoblasts over 30 days of differentiation in culture.

Close modal

We next confirmed the expression of TGFβ and GDF family proteins in day 0 versus day 27 osteoblasts. Among TGFβ family, expression of TGFβ1 and TGFβ2 was increased by 17- and 26-fold, respectively, at day 27 of osteoblast differentiation (Fig. 2B). Among GDF family, GDF10 expression increased 8-fold, while GDF2 and GDF15 increased 2- to 3-fold (Fig. 2B). Specifically, TGFβ2 mRNA levels increased aproximately 20-fold over 30 days of osteoblast differentiation, but were undetectable in control undifferentiated osteoblast precursor MC3T3-E1 cells (Fig. 2C, top). Immunoblot of the corresponding OB-CM confirmed an increase in secreted TGFβ2 during osteoblast differentiation (Fig. 2C, bottom). GDF10 mRNA and protein also increased with similar kinetics as TGFβ2 (Fig. 2C). In contrast, TGFβ1 exhibited a different pattern of gene expression, increasing at day 6 and remaining elevated to day 30 (Fig. 2D). Thus, differentiated osteoblasts secrete factors such as TGFβ2 and GDF10 (Fig. 2E) that may play a role in inducing cellular quiescence of prostate cancer cells.

TGFβ2 and GDF10 induce quiescence of C4-2B4 cells

Dormant cells are defined as “nonproliferating or slow-cycling cells, negative/low for proliferation markers and positive/high for CDK inhibitor expression” (16, 23). Using these characteristics, we examined the potential dormancy-inducing activity of TGFβ2 and GDF10 on C4-2B4 cells using live-cell imaging over 72 hours. TGFβ2 at 50 ng/mL (1 nmol/L) and as low as 10 ng/mL (200 pmol/L) significantly increased the level of quiescent C4-2B4 cells relative to untreated cells (Fig. 3A). TGFβ2-treated C4-2B4 cells that did not divide were negative for the proliferation marker Ki-67 (Fig. 3B, cells a–d) but stained positive for the CDK inhibitor p27 in the nucleus, suggesting that TGFβ2 induces cell-cycle arrest leading to cellular dormancy.

Figure 3.

TGFβ2 and GDF10 induce prostate cancer cellular quiescence in vitro. A, C4-2B4 cells treated with TGFβ2, GDF10, or TGFβ1 for 72 hours were analyzed by live-cell imaging. Quiescent cells that did not divide in 60–72 hours relative to total cells counted were quantified (mean ± SEM). n cells were analyzed in multiple experiments: TGFβ2 (N = 5–8), GDF10 (N = 4), and TGFβ1 (N = 5), except for TGFβ2 or GDF10 dose response (N = 1). P values were by t test. Live-cell images (h:min) of cells treated with TGFβ2 (B) or GDF10 (C), followed by coimmunostaining for Ki-67 and p27. Colored asterisks, control cell progenies. Red box, enlarged. Cells a–d, quiescent cells following TGFβ2 or GDF10 treatment. All bars, 10 μm. D, Cellular quiescence of several other prostate cancer cells treated and analyzed as in A in multiple experiments: C4-2b (N = 3–5), PC3-mm2 (N = 2–3), 22Rv1 (N = 2–3), BPH-1 (N = 2–3).

Figure 3.

TGFβ2 and GDF10 induce prostate cancer cellular quiescence in vitro. A, C4-2B4 cells treated with TGFβ2, GDF10, or TGFβ1 for 72 hours were analyzed by live-cell imaging. Quiescent cells that did not divide in 60–72 hours relative to total cells counted were quantified (mean ± SEM). n cells were analyzed in multiple experiments: TGFβ2 (N = 5–8), GDF10 (N = 4), and TGFβ1 (N = 5), except for TGFβ2 or GDF10 dose response (N = 1). P values were by t test. Live-cell images (h:min) of cells treated with TGFβ2 (B) or GDF10 (C), followed by coimmunostaining for Ki-67 and p27. Colored asterisks, control cell progenies. Red box, enlarged. Cells a–d, quiescent cells following TGFβ2 or GDF10 treatment. All bars, 10 μm. D, Cellular quiescence of several other prostate cancer cells treated and analyzed as in A in multiple experiments: C4-2b (N = 3–5), PC3-mm2 (N = 2–3), 22Rv1 (N = 2–3), BPH-1 (N = 2–3).

Close modal

A role of GDF10 in tumor dormancy has never been examined. GDF10 significantly increased the level of quiescent C4-2B4 cells (Fig. 3A). GDF10-treated quiescent cells were Ki-67 negative but showed strong p27 positivity in the nucleus (Fig. 3C, cells a–d), suggesting that GDF10 is a novel dormancy-inducing factor for C4-2B4 cells. TGFβ1 did not induce significant cellular quiescence in C4-2B4 cells (Fig. 3A), in agreement with reports that TGFβ1 does not affect dormancy in HNSCC HEp3 (16) and breast cancer cells (18).

TGFβ2 and GDF10 induce long-term dormancy in C4-2B4 cells

While most of the studies on cellular dormancy have been performed using a 48-hour time frame (16, 17, 24), we next measured cellular dormancy over a much longer period, that is, over 7 days. C4-2B4 cells were labeled with the lipophilic membrane dye DiO (green), seeded in Grid-500 dishes to allow for cell monitoring in more fields, and followed for 7 days. The intensity of the photostable fluorescent DiO dye was diluted out by cell division in proliferating cells (Supplementary Fig. S1A) but not in nonproliferating cells (Supplementary Fig. S1B). We found that there was a higher number of DiO-positive cells with TGFβ2 (13.1% ± 2.3%) or GDF10 (11.1 ± 3.0%) treatment compared with untreated control cells (3.4% ± 0.8 %) over this extended period of time (Supplementary Fig. S2A). Furthermore, these long-term DiO-positive cells also stained positive for p27 (Supplementary Fig. S2B, cells a and b). We followed some of the DiO-labeled quiescent prostate cancer cells over 2–3 weeks (Supplementary Fig. S2C and S2D) and did not observe cell proliferation, suggesting that these cells represent long-term “slow-cycling or nonproliferating” dormant cells. These results suggest that TGFβ2 or GDF10 can induce long-term cellular dormancy in C4-2B4 cells. Moreover, removal of TGFβ2 or GDF10 from long-term dormant cells was unable to release cells from dormancy, suggesting that the simple removal of a factor may not be sufficient for dormant prostate cancer cells to exit dormancy and resume cell proliferation. However, we cannot exclude the possibility that this is due to limitations of in vitro tissue culture conditions.

TGFβ2 and GDF10 induce quiescence in other prostate cancer cells

We found that both TGFβ2 and GDF10 also increased cellular quiescence in C4-2b cells (5) and PC3-mm2 cells (25), but not 22Rv1 (7) and BPH-1 (26) cells (Fig. 3D). Although both C4-2b (5) and C4-2B4 (3, 4) are LNCaP sublines, we found by live-cell imaging that they have slightly different cell morphology and properties under the same culture conditions. C4-2b cells tend to form aggregates and are more mobile in the culture dish, while C4-2B4 cells are more dispersed, flattened, and less mobile (Supplementary Fig. S3A). These results show a differential response of prostate cancer cell lines to dormancy induction by TGFβ2 and GDF10.

Knockdown of TGFβRIII in C4-2B4 cells prevents dormancy induction

TGFβ2 or GDF10 likely induces dormancy response through membrane receptors. TGFβ2 receptor is composed of TGFβRI, TGFβRII, and TGFβRIII (or betaglycan). TGFβRIII is particularly important as TGFβ2 cannot bind TGFβRII independently without TGFβRIII as a coreceptor (27). GDF10 receptor is reported to be TGFβR1 and TGFβRII in several cell lines (28–30). Whether TGFβRIII is required for GDF10 to function is unknown. We knocked down TGFβRIII, a proteoglycan with a 110-kDa core protein (31), which migrates as a diffuse band between 110-kDa and 200-kDa (Fig. 4A). Two clones, C4-2B4-shTβRIII-#2 and #3, both showing approximately 60% decrease in TGFβRIII, lost dormancy response to TGFβ2 or GDF10, while sh-Vector control cells remained responsive (Fig. 4B), suggesting that TGFβRIII regulates both TGFβ2 and GDF10 signaling in dormancy induction in C4-2B4 cells.

Figure 4.

TGFβRIII receptor and activation of p38MAPK signaling mediate TGFβ2 and GDF10 dormancy-inducing effects in prostate cancer cells. A, TGFβRIII levels in knockdown clones C4-2B4-shTβRIII-#2 and #3 were analyzed by immunoprecipitation, followed by immunoblotting, quantified against pGIPZ-sh-Vector cells, and signals normalized against actin controls. Scale bar, TGFβRIII 100-kDa core protein and glycosylated forms. B, Cells in A were treated with 50 ng/mL TGFβ2 or GDF10 for 72 hours. Cellular quiescence was determined by live-cell imaging. n cells were monitored in N = 3–5 experiments, except for shTβRIII-#3 cells (N = 2). P values are by t test. C, C4-2B4 cells were treated with TGFβ2, immunostained for p38MAPK, and counterstained with DAPI. Images were captured by confocal microscopy (left). Box, enlarged. Right, phospho-p38MAPK (p-p38) immunostaining following TGFβ2 or GDF10 time course. C4-2b and PC3-mm2 cells (D) and C4-2B4-sh-Vector, C4-2B4-shTβRIII-#2, and C4-2B4-shTβRIII-#3 cells (E) were immunostained for p-p38 after treatment with TGFβ2 or GDF10 for 60 minutes. All bars, 10 μm.

Figure 4.

TGFβRIII receptor and activation of p38MAPK signaling mediate TGFβ2 and GDF10 dormancy-inducing effects in prostate cancer cells. A, TGFβRIII levels in knockdown clones C4-2B4-shTβRIII-#2 and #3 were analyzed by immunoprecipitation, followed by immunoblotting, quantified against pGIPZ-sh-Vector cells, and signals normalized against actin controls. Scale bar, TGFβRIII 100-kDa core protein and glycosylated forms. B, Cells in A were treated with 50 ng/mL TGFβ2 or GDF10 for 72 hours. Cellular quiescence was determined by live-cell imaging. n cells were monitored in N = 3–5 experiments, except for shTβRIII-#3 cells (N = 2). P values are by t test. C, C4-2B4 cells were treated with TGFβ2, immunostained for p38MAPK, and counterstained with DAPI. Images were captured by confocal microscopy (left). Box, enlarged. Right, phospho-p38MAPK (p-p38) immunostaining following TGFβ2 or GDF10 time course. C4-2b and PC3-mm2 cells (D) and C4-2B4-sh-Vector, C4-2B4-shTβRIII-#2, and C4-2B4-shTβRIII-#3 cells (E) were immunostained for p-p38 after treatment with TGFβ2 or GDF10 for 60 minutes. All bars, 10 μm.

Close modal

TGFβ2 and GDF10 stimulate nuclear translocation of p-p38MAPK in prostate cancer cells

Increases in cellular quiescence by TGFβ2 or GDF10 may be due to a switch from a mitogenic to a dormancy pathway, for example, from ERK to p38MAPK (p38) signaling (32–34). While TGFβ family proteins activate the canonical Smad pathway (35, 36), they also signal through noncanonical (p38MAPK, ERK1/2, PI3K/Akt) pathways (37, 38). We found that p38MAPK was localized in the cytoplasm of C4-2B4 cells, but became enriched in the nucleus in response to TGFβ2 (Fig. 4C, left), indicating p38MAPK activation. p38MAPK activation is accompanied by dual phosphorylation of pThr180/pTyr182 (p-p38) (12) and p-p38 translocation into the nucleus. Both TGFβ2 and GDF10 stimulated p-p38 nuclear translocation within 30 minutes up to 3-hour stimulation of C4-2B4 cells (Fig. 4C, right); this effect was strongly inhibited by the p38MAPK inhibitor SB202190 (Supplementary Fig. S3B). TGFβ2 and GDF10 also stimulated p-p38 nuclear translocation in C4-2b, PC3-mm2 (Fig. 4D) and C4-2B4-sh-Vector cells (Fig. 4E), but not C4-2B4-shTβRIII-#2 and #3 cells. In contrast, little activation of Smad signaling was observed (Supplementary Fig. S3C–S3F). These results suggest that activation of a TGFβRIII–p38MAPK signaling pathway mediates cellular quiescence in prostate cancer cells.

N-terminal pS249/pT252 phosphorylation of RB by p38MAPK during dormancy induction in prostate cancer cells

Nuclear p-p38 may increase cellular quiescence by modulating the activity of cell-cycle regulatory proteins. p-p38 has recently been shown to phosphorylate the novel N-terminal Ser249/Thr252 sites in the tumor suppressor RB protein (8). p38-mediated phosphorylation of Ser249/Thr252-RB renders RB insensitive to cyclin-dependent kinase regulation, resulting in a delay in G1 cell-cycle progression (8). Both TGFβ2 and GDF10 increased pS249/pT252-RB in the nucleus within 30 minutes of stimulation, a timing that closely followed p-p38 nuclear translocation (Fig. 5A). To detect potential physical association of endogenous pS249/pT252-RB with p-p38, we performed a PLA (39, 40). We found a significant increase in PLA signals, visualized as individual fluorescent spots, in the nucleus of TGFβ2- or GDF10-treated cells relative to control cells (Fig. 5B, see enlarged), supporting an interaction between p-p38 and its substrate pS249/pT252-RB. As pS249/pT252-RB exhibits enhanced RB's repressor activity (8), these results suggest that p38MAPK–pS249/pT252–RB signaling blocks cell-cycle progression, leading to cellular quiescence.

Figure 5.

Phospho-p38MAPK and novel N-terminal phospho-S249/T252-RB colocalize in the nucleus of TGFβ2- or GDF10-treated C4-2B4 cells. A, C4-2B4 cells were treated with 50 ng/mL TGFβ2 or GDF10 and coimmunostained for p-p38 and phospho-S249/T252-RB (8). B, Cells treated as in A for 180 minutes were coimmunostained as in A, followed with PLA. Cells incubated with either primary antibodies alone or no antibodies served as negative controls. Boxes, enlarged. All bars, 10 μm. PLA spots/nucleus were quantified in n nuclei. P values are by t test. #, no spots detected.

Figure 5.

Phospho-p38MAPK and novel N-terminal phospho-S249/T252-RB colocalize in the nucleus of TGFβ2- or GDF10-treated C4-2B4 cells. A, C4-2B4 cells were treated with 50 ng/mL TGFβ2 or GDF10 and coimmunostained for p-p38 and phospho-S249/T252-RB (8). B, Cells treated as in A for 180 minutes were coimmunostained as in A, followed with PLA. Cells incubated with either primary antibodies alone or no antibodies served as negative controls. Boxes, enlarged. All bars, 10 μm. PLA spots/nucleus were quantified in n nuclei. P values are by t test. #, no spots detected.

Close modal

Expression of dominant-negative p38MAPK in prostate cancer cells prevents dormancy induction in vitro

Next, we stably transduced C4-2B4 cells with p38α dominant-negative MAPK (p38DN; ref. 12). Overexpression of p38DN did not affect the basal level of p-p38 expression (Fig. 6A). However, p38DN overexpression prevented nuclear translocation of p-p38 in C4-2B4-p38DN cells (Fig. 6B) and blocked dormancy induction in response to TGFβ2 and GDF10 (Fig. 6C). In contrast, robust nuclear translocation of p-p38 (Fig. 6B) and dormancy induction (Fig. 6C) was observed in C4-2B4-Vector control cells in response to TGFβ2 or GDF10. Similarly, overexpressing p38DN in C4-2b cells (Fig. 6D) blocked nuclear translocation of p-p38 (Fig. 6E) and prevented dormancy induction in C4-2b-p38DN cells (Fig. 6F) in response to either TGFβ2 or GDF10, while dormancy induction was not affected in C4-2b-Vector control cells. Thus, activation of p38MAPK signaling is required for TGFβ2- and GDF10-mediated dormancy induction in prostate cancer cells.

Figure 6.

Dominant-negative p38MAPK prevents dormancy induction by TGFβ2 or GDF10 in prostate cancer cells in vitro. A, C4-2B4 cells stably expressing empty vector or p38DN were examined by Western blot analysis for p-p38, followed by p38MAPK. β-tubulin, loading control. B, Cells in A were treated with 50 ng/mL TGFβ2 or GDF10 for 60 minutes and immunostained for p-p38. C, Cells in A were treated with TGFβ2 or GDF10 for 72 hours. Cellular quiescence was determined by live-cell imaging. n cells were monitored in multiple experiments: C4-2B4-Vector (N = 5–7), C4-2B4-p38DN (N = 4–5). P values are by t test. D, C4-2b-Vector or C4-2b-p38DN cells were generated and examined by Western blot analysis as in A. E, Cells in D were treated as in B and immunostained for p-p38. All bars, 10 μm. F, Cellular quiescence in C4-2b-Vector or C4-2b-p38DN cells after TGFβ2 or GDF10 incubation was determined as in C. n cells were followed in multiple experiments: C4-2b-Vector (N = 4–5) and C4-2b-p38DN (N = 3–5).

Figure 6.

Dominant-negative p38MAPK prevents dormancy induction by TGFβ2 or GDF10 in prostate cancer cells in vitro. A, C4-2B4 cells stably expressing empty vector or p38DN were examined by Western blot analysis for p-p38, followed by p38MAPK. β-tubulin, loading control. B, Cells in A were treated with 50 ng/mL TGFβ2 or GDF10 for 60 minutes and immunostained for p-p38. C, Cells in A were treated with TGFβ2 or GDF10 for 72 hours. Cellular quiescence was determined by live-cell imaging. n cells were monitored in multiple experiments: C4-2B4-Vector (N = 5–7), C4-2B4-p38DN (N = 4–5). P values are by t test. D, C4-2b-Vector or C4-2b-p38DN cells were generated and examined by Western blot analysis as in A. E, Cells in D were treated as in B and immunostained for p-p38. All bars, 10 μm. F, Cellular quiescence in C4-2b-Vector or C4-2b-p38DN cells after TGFβ2 or GDF10 incubation was determined as in C. n cells were followed in multiple experiments: C4-2b-Vector (N = 4–5) and C4-2b-p38DN (N = 3–5).

Close modal

Expression of dominant-negative p38MAPK in prostate cancer cells abrogates dormancy in bone

We further injected C4-2B4-p38DN cells intrafemorally to examine their response to the dormancy-inducing activity in the bone microenvironment. C4-2B4-p38DN cells showed an increase in tumor growth in femur over 6 weeks, while C4-2B4-Vector control cells showed an overall dormant phenotype (Fig. 7A). Close monitoring of individual mouse injected with C4-2B4-Vector cells showed fluctuations in tumor growth (Supplementary Fig. S4A), likely reflecting the failure of prostate cancer micrometastases to expand in the dormancy-inducing microenvironment of the bone. Consistently, femoral tumors from C4-2B4-Vector cells revealed areas of cell death (Fig. 7B, i, asterisk), resulting in fewer viable cells, and weak and diffuse Ki-67 staining (Fig. 7B, ii). In contrast, femoral tumors from C4-2B4-p38DN cells showed an abundance of mitotic figures (Fig. 7B, iii, arrowheads; enlarged) and strong Ki-67 positivity (Fig. 7B, iv). Quantification confirmed significant increases in the number of nuclei that have moderate to strong Ki-67 positivity in the C4-2B4-p38DN tumors compared with C4-2B4-Vector tumors (Fig. 7C). Similar growth kinetics (Supplementary Fig. S4B and S4C) and increases in moderate to strong Ki-67 positive nuclei were also observed in C4-2b-p38DN tumors compared with C4-2b-Vector tumors (Supplementary Fig. S4D).

Figure 7.

Dominant-negative p38MAPK prevents dormancy in vivo and lower TGFβRIII expression is associated with poor survival rates in human prostate cancer. A, Intrabone injection. C4-2B4-Vector or C4-2B4-p38DN cells (1 × 106) were injected into the femur of SCID mice (n = 5 each). Tumor growth was monitored by bioluminescence and expressed as fold increase relative to week 1. B, Consecutive sections of C4-2B4-Vector (i and ii)- or C4-2B4-p38DN (iii and iv)-derived femoral tumors in A were analyzed by hematoxylin and eosin (H&E) and Ki-67. Boxes, enlarged mitotic cells: iii, metaphase (left), prometaphase (right); iv, cells separating in telophase. T, tumor; B, bone; *, dead cells; arrowheads, mitotic cells. ×200. C, Quantification of Ki-67 levels in nuclei of C4-2B4 tumors using the Aperio ImageScope software. D, Intracardiac injection. Left, C4-2B4 cells (1 × 106) were injected into the left ventricle of SCID mice. At the indicated times, total femur and tibia DNA was harvested from hind legs and subjected to quantitative PCR for human Alu repetitive sequence using 20 ng DNA. DNA from C4-2B4 was used to generate a standard curve for calculating cell numbers/20 ng DNA. The total cell numbers/leg was then calculated on the basis of total DNA per mouse leg. Right, C4-2B4-Vector cells or C4-2B4-p38DN cells (1 × 106) were injected into the left ventricle of SCID mice (n = 3 each). At 19 weeks postinjection, Alu PCR was performed. P values are by t test. NS, not significant. E, Kaplan–Meier plots for correlations of TGFBR3 expression with disease-specific survival of patients with prostate cancer (Nakagawa dataset; ref. 13). Groups with highest (“top third”), lowest (“bottom third”), and intermediate (“middle third”) TGFBR3 mRNA levels were compared. P values were determined by univariate Cox and stratified log-rank test. n, number of patients. F, Box plots show correlations of TGFBR3 expression with prostate cancer metastatic potential (Taylor dataset; ref. 14) and prostate cancer metastatic progression (Lapointe dataset; ref. 15). P values are by t test. G, Model for prostate cancer tumor dormancy in bone. See text.

Figure 7.

Dominant-negative p38MAPK prevents dormancy in vivo and lower TGFβRIII expression is associated with poor survival rates in human prostate cancer. A, Intrabone injection. C4-2B4-Vector or C4-2B4-p38DN cells (1 × 106) were injected into the femur of SCID mice (n = 5 each). Tumor growth was monitored by bioluminescence and expressed as fold increase relative to week 1. B, Consecutive sections of C4-2B4-Vector (i and ii)- or C4-2B4-p38DN (iii and iv)-derived femoral tumors in A were analyzed by hematoxylin and eosin (H&E) and Ki-67. Boxes, enlarged mitotic cells: iii, metaphase (left), prometaphase (right); iv, cells separating in telophase. T, tumor; B, bone; *, dead cells; arrowheads, mitotic cells. ×200. C, Quantification of Ki-67 levels in nuclei of C4-2B4 tumors using the Aperio ImageScope software. D, Intracardiac injection. Left, C4-2B4 cells (1 × 106) were injected into the left ventricle of SCID mice. At the indicated times, total femur and tibia DNA was harvested from hind legs and subjected to quantitative PCR for human Alu repetitive sequence using 20 ng DNA. DNA from C4-2B4 was used to generate a standard curve for calculating cell numbers/20 ng DNA. The total cell numbers/leg was then calculated on the basis of total DNA per mouse leg. Right, C4-2B4-Vector cells or C4-2B4-p38DN cells (1 × 106) were injected into the left ventricle of SCID mice (n = 3 each). At 19 weeks postinjection, Alu PCR was performed. P values are by t test. NS, not significant. E, Kaplan–Meier plots for correlations of TGFBR3 expression with disease-specific survival of patients with prostate cancer (Nakagawa dataset; ref. 13). Groups with highest (“top third”), lowest (“bottom third”), and intermediate (“middle third”) TGFBR3 mRNA levels were compared. P values were determined by univariate Cox and stratified log-rank test. n, number of patients. F, Box plots show correlations of TGFBR3 expression with prostate cancer metastatic potential (Taylor dataset; ref. 14) and prostate cancer metastatic progression (Lapointe dataset; ref. 15). P values are by t test. G, Model for prostate cancer tumor dormancy in bone. See text.

Close modal

Because DTCs arrive in bone through the circulation, we also delivered tumor cells by intracardiac injection of prostate cancer cells. We did not detect overt tumor growth in bone over 19 weeks by bioluminescence imaging. However, we detected human Alu repetitive sequence signals at the single-cell level in total femur and tibia DNA prepared from mouse hind legs, as determined by real-time PCR using a serial dilution of DNA prepared from C4-2B4 cells (Fig. 7D). These results showed that C4-2B4 cells could disseminate to bone through intracardiac injection. At 1 week postinjection, the number of cells was around 350 ± 66 cells/leg (Fig. 7D, left), that is, about 0.035% of the tumor cells per one million cells injected. These observations indicated that C4-2B4 cells have disseminated to bone and are consistent with clinical observations that few DTCs (<0.1%) survive to be detected in distant organs (41). At 5 weeks postinjection, the number of cells was about 7,699 ± 4,270 cells/leg. At 19 weeks postinjection, the number of cells was about 14,636 ± 8,147 cells/leg. Although there was an increase in DTCs between weeks 1 and 5, it did not reach statistical significance (Fig. 7D, left). The level of DTCs detected between weeks 5 and 19 also did not reach statistical significance. However, there was a trend of continued increase in cell numbers over 19 weeks, suggesting that these DTCs were dormant or very slow-cycling in bone. These results are consistent with the studies by Lawson and colleagues (42), which showed that by intravital microscopy over 3 weeks some of the disseminated multiple myeloma cells were dormant/slow-cycling while others continued to proliferate in the bone marrow.

We further used intracardiac injection to examine the effect of disrupting p38MAPK signaling on tumor dormancy in bone. Disrupting p38MAPK signaling by expression of p38DN in C4-2B4 cells led to a significant 3-fold increase in the number of DTCs to about 43,859 ± 9,815 cells/leg at 19 weeks postintracardiac injection (Fig. 7D, right). These observations are consistent with the involvement of p38MAPK in dormancy signaling.

Increases in tumor burden of C4-2B4-p38DN could be due to an increase in the proliferative activity or a decrease in dormant cells in the C4-2B4-p38DN tumor. We found that C4-2B4-p38DN cells showed a slightly higher proliferation rate (∼30%) than C4-2B4-Vector control cells in vitro (Supplementary Fig. S4E), but there was no difference in the percent of quiescent cells between C4-2B4-p38DN and C4-2B4-Vector cells (Fig. 6C, blue bars). Thus, the increase in tumor burden of C4-2B4-p38DN is likely due to an increase in the proliferative activity of C4-2B4-p38DN cells. Together, these results show that blocking p38MAPK activation via expression of dominant-negative p38MAPK overcame the effects of dormancy-inducing factors in the bone microenvironment and abrogated prostate cancer tumor dormancy response in bone.

Lower TGFβRIII predicts poor clinical outcomes

Because dormancy induction requires the expression of TGFβRIII on prostate cancer cells (Fig. 4B), prostate cancer cells expressing lower TGFβRIII might be less able to enter into dormancy when disseminated to bone. If so, men with prostate cancer that expressed lower TGFβRIII levels might have a shorter time between prostatectomy to the development of overt bone metastasis. Indeed, patients with prostate cancer with relatively lower levels of TGFBR3 expression have shorter disease-specific survival in the Nakagawa dataset (Fig. 7E; ref. 13). In addition, TGFBR3 levels are significantly lower in the metastasis (n = 19) than the primary (n = 131) prostate cancer tumor samples in the Taylor dataset (Fig. 7F; ref. 14). Furthermore, TGFBR3 levels significantly correlate with prostate cancer disease progression in the Lapointe dataset (15), in which primary prostate cancer (n = 59) expressed lower levels of TGFBR3 than normal prostate tissue (n = 39), and metastatic prostate cancer (n = 9) expressed the lowest levels of TGFBR3 (Fig. 7F). These analyses support that lower levels of TGFBR3 expression are associated with metastasis and poor clinical outcome of prostate cancer.

We examined how DTCs are initially induced into a dormant state in bone. Our findings elucidate that differentiated osteoblasts in the bone marrow are one cell type that regulates prostate cancer tumor dormancy in bone. Our studies support a model in which differentiated osteoblasts secrete dormancy-inducing factors, including TGFβ2 and GDF10, which activate TGFβRIII signaling in prostate cancer cells to stimulate p38MAPK phosphorylation and nuclear translocation. Nuclear p-p38 then phosphorylates the N-terminus of RB at pS249/pT252, leading to an increase in the CDK inhibitor p27 and blocking G1 cell-cycle progression (Fig. 7G). Our studies provide a mechanistic pathway for the clinical observation that bone is the major site of prostate cancer relapse. Understanding the mechanism underlying tumor dormancy provides the foundation for future development of strategies to prevent dormant tumors from emerging as overt bone metastasis.

Our studies identified factors secreted by differentiated osteoblasts that can regulate prostate tumor cell quiescence. GDF10 was first cloned from rat calvarial osteoblasts (42), and plays a role in osteoblast differentiation (43, 44). We elucidate for the first time that osteoblast-derived GDF10 induces dormancy of prostate cancer cells. Furthermore, the levels of GDF10 and TGFβ2 are higher in differentiated compared with undifferentiated osteoblasts, indicating that the status of osteoblast differentiation may affect tumor dormancy. Studies by Lawson and colleagues (45) also showed that osteoblasts or bone-lining cells on quiescent bone surface were the major cell type in the bone marrow that provided dormancy factors to myeloma cells. Thus, it is likely that GDF10 and TGFβ2 are secreted from the bone-lining cells.

Additional factors in the bone microenvironment are also likely involved in prostate cancer dormancy. In our osteoblast array analysis, we found GAS6, a factor reported to induce dormancy of prostate cancer cells (24, 46, 47). We also found BMP family proteins including BMP5 and BMP3, but not BMP4 or BMP7, in the microarray analysis (Fig. 2A). BMP7 was reported to be secreted by bone marrow stromal cells and was shown to induce dormancy of PC3 cells (17). Consistently, we did not detect BMP7 in either day 0 or day 30 osteoblasts (Fig. 2A). Thrombospondin-1 secreted by endothelial cells was also shown to sustain tumor cell quiescence (18). These observations suggest that many factors in the bone microenvironment affect dormancy.

We observed fluctuations of tumor growth in bone in our intrabone injection studies (Fig. 7A; Supplementary Figs. S4A–S4C). Lawson and colleagues (45) reported that they observed myeloma cells entering dormancy (dormancy switched on) when cells are located close to osteoblasts, but resumed proliferation (dormancy switched off) when cells are located farther away from osteoblasts in bone. This “on-and-off dormancy switch” has caused fluctuations in the dormancy status of myeloma cells. It is possible that a similar phenomenon might also occur in prostate cancer cell dormancy in bone. When DTCs disseminate to bone, DTCs first encounter the trabeculae, which is enriched with osteoblasts. It is likely that osteoblast-secreted factors control the rate of cell proliferation and induce dormancy of the tumor cells near to the bone surface. The changes in the position of DTCs relative to bone over time may be one mechanism for DTCs to be reactivated after initial dormancy. Proliferating DTCs may also be regulated by cell death and survival pathways in the bone microenvironment. Our previous studies showed that the bone microenvironment also provides survival signals to prostate cancer tumors, leading to therapy resistance (10). Thus, fluctuations of prostate cancer tumor growth in bone may be due to both dormancy signals and survival signals provided by the bone microenvironment. Such dynamic interactions between tumor and the bone may regulate the temporal development of bone metastasis.

We found that activation of p38MAPK contributes to TGFβ2- and GDF10-mediated prostate cancer cellular quiescence in vitro (Figs. 4 and 6) and prostate cancer tumor dormancy in vivo (Fig. 7). Similarly, Adam and colleagues (48) and Chery and colleagues (49) showed that p38MAPK pathway activation may direct tumor cells to transit into a dormant state. Interestingly, Gubern and colleagues (8) demonstrated that p38MAPK phosphorylates the N-terminus of RB at S249/T252, which prevents its inactivation by cyclin-dependent kinases and results in inhibition of proliferation of cancer cells. Our studies showed that phospho-p38MAPK colocalizes with pS249/pT252-RB (Fig. 5), suggesting that RB is one of the substrates of p38MAPK activated by TGFβ2 or GDF10 treatment. RB upregulates p27, a cell-cycle inhibitor and marker of quiescence, through p27 stabilization (50). Indeed, we found that dormant prostate cancer cells express p27, suggesting that TGFβ2 or GDF10 treatment inhibits cell-cycle progression at G1. Thus, our studies elucidate a p38MAPK–pS249/pT252–RB–p27 pathway as one mechanism by which prostate cancer DTCs enter dormancy in bone (Fig. 7G).

Prostate cancer cells may exhibit different dormant phenotypes when disseminated to bone. Because paracrine factors signal through receptors, variations in receptor expression or the presence of tumor suppressor proteins such as RB in prostate cancer cells due to tumor heterogeneity may lead to differences in the extent of dormancy response in the same tumor microenvironment. Indeed, we found that levels of TGFβRIII in prostate cancer cells to be correlated with metastatic progression and poor clinical outcome of prostate cancer. In addition, the number of differentiated osteoblasts may affect the levels of dormancy-inducing factors. Furthermore, host immunity may have a significant impact on cancer dormancy. Studies by Wu and colleagues (3) showed that when C4-2 cells were injected intracardially, only 1 of 5 athymic mice developed bone metastasis, whereas in SCID/bg mice that are more immune-impaired than the athymic nude mice, 3 of 7 and 2 of 7 developed lymph node and bone metastasis, respectively. In light of these studies, it is possible that the levels of bone-derived dormancy-inducing factors may be regulated by host immunity in different mouse strains. Together, these different variables may contribute in part to the difference in the time-to-bone metastasis and the incidence of bone metastasis among patients with advanced prostate cancer (1).

We found that only a subset of available prostate cancer cell lines became dormant in in vitro or in vivo studies. In our in vitro studies, we observed that C4-2B4, C4-2b, and PC3-mm2, but not 22Rv1 or BPH-1 cells, showed dormancy response to TGFβ2 and GDF10 (Fig. 3D). We examined the expression of two molecules, that is, TGFβRIII and RB, along the TGFβRIII–p38MAPK–pS249/pT252–RB–p27 pathway. Interestingly, we found that these prostate cancer cell lines all express TGFβRIII and RB (Supplementary Fig. S5), suggesting that the differences in dormancy response in these prostate cancer cells may lie elsewhere along the TGFβRIII–p38MAPK–pS249/pT252–RB–p27 pathway, for example, phosphorylation and nuclear translocation of p38MAPK, identified in our study. Some of the prostate cancer cell lines may have intrinsic mechanisms that mediate reactivation from dormancy. We found that PC3-mm2 responded to TGFβ2 and GDF10 for dormancy induction (Fig. 3D), but they do not become dormant when injected into mouse femur. Thus, PC3-mm2 cells may have as-yet-unknown dormancy exit mechanisms that antagonize dormancy-inducing factors in the bone microenvironment. We have previously shown that PC3-mm2 cells express constitutively activated integrin β1 (10), which may lead to transition from quiescence to proliferation (19). Furthermore, PC3-mm2 cells induce osteolytic bone lesions that may alter the bone microenvironment favoring prostate cancer growth in bone. In breast cancer cells, VCAM1 expression was shown to lead to dormancy exit (51) and expression of Coco, a secreted BMP inhibitor, in MDA-MB-231 cells rendered these cells nondormant (52). Our studies suggest that entry into and exit out of cellular dormancy may involve distinct pathways.

The identification of dormancy-inducing factors and their signaling mechanisms that lead to DTC quiescence suggests possibilities of developing strategies to prevent dormant tumors from emergence as overt bone metastasis. These could include strategies that either enhance or prevent loss of dormancy-inducing factor production. Further studies on the molecular mechanisms regulating entry into and exit out of dormancy may reveal specific targets for therapeutic manipulation.

No potential conflicts of interest were disclosed.

Conception and design: L.-Y. Yu-Lee, Y.-C. Lee, G.E. Gallick, S.-H. Lin

Development of methodology: L.-Y. Yu-Lee, G. Yu, Y.-C. Lee, S.-C. Lin, J. Pan, B. Liu, J. Rodriguez-Canales, I.I. Wistuba, E. de Nadal, F. Posas, S.-H. Lin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.-Y. Yu-Lee, G. Yu, Y.-C. Lee, S.-C. Lin, J. Pan, T. Pan, K.-J. Yu, J. Rodriguez-Canales, P.A. Villalobos, G.E. Gallick, S.-H. Lin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.-Y. Yu-Lee, G. Yu, Y.-C. Lee, S.-C. Lin, J. Pan, K.-J. Yu, B. Liu, C.J. Creighton, I.I. Wistuba, G.E. Gallick, S.-H. Lin

Writing, review, and/or revision of the manuscript: L.-Y. Yu-Lee, E. de Nadal, F. Posas, G.E. Gallick, S.-H. Lin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.-Y. Yu-Lee, Y.-C. Lee, J. Pan, S.-H. Lin

Study supervision: L.-Y. Yu-Lee, S.-H. Lin

This work was supported by grants from the NIH R01CA174798 (to S.-H. Lin), NIH 5P50CA140388 (to C. Logothetis and S.-H. Lin), NIH P30CA16672 Core grant to MD Anderson Cancer Center (principal investigator: R. DePinho); Cancer Prevention Research Institute of Texas CPRIT RP150179 (to S.-H. Lin and L.-Y. Yu-Lee), CPRIT RP150282 (to G.E. Gallick and S.-H. Lin); and Sister Institute Network Fund at University of Texas MD Anderson Cancer Center (to S.-H. Lin).

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.

1.
Ahove
DA
,
Hoffman
KE
,
Hu
JC
,
Choueiri
TK
,
D'Amico
AV
,
Nguyen
PL
. 
Which patients with undetectable PSA levels 5 years after radical prostatectomy are still at risk of recurrence?–implications for a risk-adapted follow-up strategy
.
Urology
2010
;
76
:
1201
5
.
2.
Logothetis
C
,
Lin
SH
. 
Osteoblasts in prostate cancer metastasis to bone
.
Nat Rev Cancer
2005
;
5
:
21
8
.
3.
Wu
TT
,
Sikes
RA
,
Cui
Q
,
Thalmann
GN
,
Kao
C
,
Murphy
CF
, et al
Establishing human prostate cancer cell xenografts in bone: induction of osteoblastic reaction by prostate-specific antigen-producing tumors in athymic and SCID/bg mice using LNCaP and lineage-derived metastatic sublines
.
Int J Cancer
1998
;
77
:
887
94
.
4.
Thalmann
GN
,
Sikes
RA
,
Wu
TT
,
Degeorges
A
,
Chang
SM
,
Ozen
M
, et al
LNCaP progression model of human prostate cancer: androgen-independence and osseous metastasis
.
Prostate
2000
;
44
:
91
103
.
5.
Thalmann
GN
,
Anezinis
PE
,
Chang
S
,
Zhau
HE
,
Kim
E
,
Hopwood
VL
, et al
Androgen-independent cancer progression and bone metastasis in the LNCaP model of human prostate cancer
.
Cancer Res
1994
;
54
:
2577
81
.
6.
Wainstein
MA
,
He
F
,
Robinson
D
,
Kung
HJ
,
Schwartz
S
,
Giaconia
JM
, et al
CWR22: androgen-dependent xenograft model derived from a primary human prostatic carcinoma
.
Cancer Res
1994
;
54
:
6049
52
.
7.
Sramkoski
RM
,
Pretlow
TG
 II
,
Giaconia
JM
,
Pretlow
TP
,
Schwartz
S
,
Sy
MS
, et al
A new human prostate carcinoma cell line, 22Rv1
.
In Vitro Cell Dev Biol Anim
1999
;
35
:
403
9
.
8.
Gubern
A
,
Joaquin
M
,
Marques
M
,
Maseres
P
,
Garcia-Garcia
J
,
Amat
R
, et al
The N-Terminal phosphorylation of RB by p38 bypasses its inactivation by CDKs and prevents proliferation in cancer cells
.
Mol Cell
2016
;
64
:
25
36
.
9.
Lin
SH
,
Cheng
CJ
,
Lee
YC
,
Ye
X
,
Tsai
WW
,
Kim
J
, et al
A 45-kDa ErbB3 secreted by prostate cancer cells promotes bone formation
.
Oncogene
2008
;
27
:
5195
203
.
10.
Lee
YC
,
Jin
JK
,
Cheng
CJ
,
Huang
CF
,
Song
JH
,
Huang
M
, et al
Targeting constitutively activated beta1 integrins inhibits prostate cancer metastasis
.
Mol Cancer Res
2013
;
11
:
405
17
.
11.
Chu
K
,
Cheng
CJ
,
Ye
X
,
Lee
YC
,
Zurita
AJ
,
Chen
DT
, et al
Cadherin-11 promotes the metastasis of prostate cancer cells to bone
.
Mol Cancer Res
2008
;
6
:
1259
67
.
12.
Somwar
R
,
Koterski
S
,
Sweeney
G
,
Sciotti
R
,
Djuric
S
,
Berg
C
, et al
A dominant-negative p38 MAPK mutant and novel selective inhibitors of p38 MAPK reduce insulin-stimulated glucose uptake in 3T3-L1 adipocytes without affecting GLUT4 translocation
.
J Biol Chem
2002
;
277
:
50386
95
.
13.
Nakagawa
T
,
Kollmeyer
TM
,
Morlan
BW
,
Anderson
SK
,
Bergstralh
EJ
,
Davis
BJ
, et al
A tissue biomarker panel predicting systemic progression after PSA recurrence post-definitive prostate cancer therapy
.
PLoS One
2008
;
3
:
e2318
.
14.
Taylor
BS
,
Schultz
N
,
Hieronymus
H
,
Gopalan
A
,
Xiao
Y
,
Carver
BS
, et al
Integrative genomic profiling of human prostate cancer
.
Cancer Cell
2010
;
18
:
11
22
.
15.
Lapointe
J
,
Li
C
,
Higgins
JP
,
van de Rijn
M
,
Bair
E
,
Montgomery
K
, et al
Gene expression profiling identifies clinically relevant subtypes of prostate cancer
.
Proc Natl Acad Sci U S A
2004
;
101
:
811
6
.
16.
Bragado
P
,
Estrada
Y
,
Parikh
F
,
Krause
S
,
Capobianco
C
,
Farina
HG
, et al
TGF-beta2 dictates disseminated tumour cell fate in target organs through TGF-beta-RIII and p38alpha/beta signalling
.
Nat Cell Biol
2013
;
15
:
1351
61
.
17.
Kobayashi
A
,
Okuda
H
,
Xing
F
,
Pandey
PR
,
Watabe
M
,
Hirota
S
, et al
Bone morphogenetic protein 7 in dormancy and metastasis of prostate cancer stem-like cells in bone
.
J Exp Med
2011
;
208
:
2641
55
.
18.
Ghajar
CM
,
Peinado
H
,
Mori
H
,
Matei
IR
,
Evason
KJ
,
Brazier
H
, et al
The perivascular niche regulates breast tumour dormancy
.
Nat Cell Biol
2013
;
15
:
807
17
.
19.
Barkan
D
,
Kleinman
H
,
Simmons
JL
,
Asmussen
H
,
Kamaraju
AK
,
Hoenorhoff
MJ
, et al
Inhibition of metastatic outgrowth from single dormant tumor cells by targeting the cytoskeleton
.
Cancer Res
2008
;
68
:
6241
50
.
20.
El Touny
LH
,
Vieira
A
,
Mendoza
A
,
Khanna
C
,
Hoenerhoff
MJ
,
Green
JE
. 
Combined SFK/MEK inhibition prevents metastatic outgrowth of dormant tumor cells
.
J Clin Invest
2014
;
124
:
156
68
.
21.
Poole
KE
,
van Bezooijen
RL
,
Loveridge
N
,
Hamersma
H
,
Papapoulos
SE
,
Lowik
CW
, et al
Sclerostin is a delayed secreted product of osteocytes that inhibits bone formation
.
FASEB J
2005
;
19
:
1842
4
.
22.
Salm
SN
,
Burger
PE
,
Coetzee
S
,
Goto
K
,
Moscatelli
D
,
Wilson
EL
. 
TGF-{beta} maintains dormancy of prostatic stem cells in the proximal region of ducts
.
J Cell Biol
2005
;
170
:
81
90
.
23.
Aguirre-Ghiso
JA
. 
Models, mechanisms and clinical evidence for cancer dormancy
.
Nat Rev Cancer
2007
;
7
:
834
46
.
24.
Shiozawa
Y
,
Pedersen
EA
,
Patel
LR
,
Ziegler
AM
,
Havens
AM
,
Jung
Y
, et al
GAS6/AXL axis regulates prostate cancer invasion, proliferation, and survival in the bone marrow niche
.
Neoplasia
2010
;
12
:
116
27
.
25.
Delworth
M
,
Nishioka
K
,
Pettaway
C
,
Gutman
M
,
Killion
J
,
Voneschenbach
A
, et al
Systemic administration of 4-amidinoindanon-1-(2′-amidino)-hydrazone, a new inhibitor of s-adenosylmethionine decarboxylase, produces cytostasis of human prostate-cancer in athymic nude-mice
.
Int J Oncol
1995
;
6
:
293
9
.
26.
Hayward
SW
,
Dahiya
R
,
Cunha
GR
,
Bartek
J
,
Deshpande
N
,
Narayan
P
. 
Establishment and characterization of an immortalized but non-transformed human prostate epithelial cell line: BPH-1
.
In Vitro Cell Dev Biol Anim
1995
;
31
:
14
24
.
27.
Lopez-Casillas
F
,
Wrana
JL
,
Massague
J
. 
Betaglycan presents ligand to the TGF beta signaling receptor
.
Cell
1993
;
73
:
1435
44
.
28.
Upadhyay
G
,
Yin
Y
,
Yuan
H
,
Li
X
,
Derynck
R
,
Glazer
RI
. 
Stem cell antigen-1 enhances tumorigenicity by disruption of growth differentiation factor-10 (GDF10)-dependent TGF-beta signaling
.
Proc Natl Acad Sci U S A
2011
;
108
:
7820
5
.
29.
Cheng
CW
,
Hsiao
JR
,
Fan
CC
,
Lo
YK
,
Tzen
CY
,
Wu
LW
, et al
Loss of GDF10/BMP3b as a prognostic marker collaborates with TGFBR3 to enhance chemotherapy resistance and epithelial-mesenchymal transition in oral squamous cell carcinoma
.
Mol Carcinogen
2016
;
55
:
499
513
.
30.
Li
S
,
Nie
EH
,
Yin
Y
,
Benowitz
LI
,
Tung
S
,
Vinters
HV
, et al
GDF10 is a signal for axonal sprouting and functional recovery after stroke
.
Nat Neurosc
2015
;
18
:
1737
45
.
31.
Andres
JL
,
Ronnstrand
L
,
Cheifetz
S
,
Massague
J
. 
Purification of the transforming growth factor-beta (TGF-beta) binding proteoglycan betaglycan
.
J Biol Chem
1991
;
266
:
23282
7
.
32.
Aguirre-Ghiso
JA
,
Liu
D
,
Mignatti
A
,
Kovalski
K
,
Ossowski
L
. 
Urokinase receptor and fibronectin regulate the ERK(MAPK) to p38(MAPK) activity ratios that determine carcinoma cell proliferation or dormancy in vivo
.
Mol Biol Cell
2001
;
12
:
863
79
.
33.
Aguirre-Ghiso
JA
,
Estrada
Y
,
Liu
D
,
Ossowski
L
. 
ERK(MAPK) activity as a determinant of tumor growth and dormancy; regulation by p38(SAPK)
.
Cancer Res
2003
;
63
:
1684
95
.
34.
Ranganathan
AC
,
Adam
AP
,
Aguirre-Ghiso
JA
. 
Opposing roles of mitogenic and stress signaling pathways in the induction of cancer dormancy
.
Cell Cycle
2006
;
5
:
1799
807
.
35.
Massague
J
. 
TGFβ in cancer
.
Cell
2008
;
134
:
215
30
.
36.
Moustakas
A
,
Heldin
CH
. 
The regulation of TGFbeta signal transduction
.
Development
2009
;
136
:
3699
714
.
37.
Yamashita
M
,
Fatyol
K
,
Jin
C
,
Wang
X
,
Liu
Z
,
Zhang
YE
. 
TRAF6 mediates Smad-independent activation of JNK and p38 by TGF-beta
.
Mol Cell
2008
;
31
:
918
24
.
38.
Parvani
JG
,
Taylor
MA
,
Schiemann
WP
. 
Noncanonical TGF-beta signaling during mammary tumorigenesis
.
J Mammary Gland Biol Neoplasia
2011
;
16
:
127
46
.
39.
Satcher
RL
,
Pan
T
,
Bilen
MA
,
Li
X
,
Lee
YC
,
Ortiz
A
, et al
Cadherin-11 endocytosis through binding to clathrin promotes cadherin-11-mediated migration in prostate cancer cells
.
J Cell Sci
2015
;
128
:
4629
41
.
40.
Soderberg
O
,
Gullberg
M
,
Jarvius
M
,
Ridderstrale
K
,
Leuchowius
KJ
,
Jarvius
J
, et al
Direct observation of individual endogenous protein complexes in situ by proximity ligation
.
Nat Methods
2006
;
3
:
995
1000
.
41.
Fidler
IJ
. 
The pathogenesis of cancer metastasis: the ‘seed and soil’ hypothesis revisited
.
Nat Rev Cancer
2003
;
3
:
453
8
.
42.
Takao
M
,
Hino
J
,
Takeshita
N
,
Konno
Y
,
Nishizawa
T
,
Matsuo
H
, et al
Identification of rat bone morphogenetic protein-3b (BMP-3b), a new member of BMP-3
.
Biochem Biophys Res Commun
1996
;
219
:
656
62
.
43.
Hino
J
,
Matsuo
H
,
Kangawa
K
. 
Bone morphogenetic protein-3b (BMP-3b) gene expression is correlated with differentiation in rat calvarial osteoblasts
.
Biochem Biophys Res Commun
1999
;
256
:
419
24
.
44.
Kaihara
S
,
Bessho
K
,
Okubo
Y
,
Sonobe
J
,
Komatsu
Y
,
Miura
M
, et al
Over expression of bone morphogenetic protein-3b (BMP-3b) using an adenoviral vector promote the osteoblastic differentiation in C2C12 cells and augment the bone formation induced by bone morphogenetic protein-2 (BMP-2) in rats
.
Life Sci
2003
;
72
:
1683
93
.
45.
Lawson
MA
,
McDonald
MM
,
Kovacic
N
,
Hua Khoo
W
,
Terry
RL
,
Down
J
, et al
Osteoclasts control reactivation of dormant myeloma cells by remodelling the endosteal niche
.
Nat Commun
2015
;
6
:
8983
.
46.
Taichman
RS
,
Patel
LR
,
Bedenis
R
,
Wang
J
,
Weidner
S
,
Schumann
T
, et al
GAS6 receptor status is associated with dormancy and bone metastatic tumor formation
.
PLoS One
2013
;
8
:
e61873
.
47.
Yumoto
K
,
Eber
MR
,
Wang
J
,
Cackowski
FC
,
Decker
AM
,
Lee
E
, et al
Axl is required for TGF-beta2-induced dormancy of prostate cancer cells in the bone marrow
.
Sci Rep
2016
;
6
:
36520
.
48.
Adam
AP
,
George
A
,
Schewe
D
,
Bragado
P
,
Iglesias
BV
,
Ranganathan
AC
, et al
Computational identification of a p38SAPK-regulated transcription factor network required for tumor cell quiescence
.
Cancer Res
2009
;
69
:
5664
72
.
49.
Chery
L
,
Lam
HM
,
Coleman
I
,
Lakely
B
,
Coleman
R
,
Larson
S
, et al
Characterization of single disseminated prostate cancer cells reveals tumor cell heterogeneity and identifies dormancy associated pathways
.
Oncotarget
2014
;
5
:
9939
51
.
50.
Dick
FA
,
Rubin
SM
. 
Molecular mechanisms underlying RB protein function
.
Nat Rev Mol Cell Biol
2013
;
14
:
297
306
.
51.
Lu
X
,
Mu
E
,
Wei
Y
,
Riethdorf
S
,
Yang
Q
,
Yuan
M
, et al
VCAM-1 promotes osteolytic expansion of indolent bone micrometastasis of breast cancer by engaging alpha4beta1-positive osteoclast progenitors
.
Cancer Cell
2011
;
20
:
701
14
.
52.
Gao
H
,
Chakraborty
G
,
Lee-Lim
AP
,
Mo
Q
,
Decker
M
,
Vonica
A
, et al
The BMP inhibitor Coco reactivates breast cancer cells at lung metastatic sites
.
Cell
2012
;
150
:
764
79
.