Tenascin-C Protects Cancer Stem–like Cells from Immune Surveillance by Arresting T-cell Activation

Metastatic disease is a dreadful complication of primary cancer, and the principal cause of cancer-related death. Two models explain the process of systemic cancer progression (1). The linear progression paradigm posits that tumor ontogeny occurs in the primary tumor, and only thereafter, disseminated tumor cells (DTC) deploy to sites of metastasis. Thus, the larger is the size of the primary tumor, the higher is the chance of metastatic spread. Conversely, the parallel progression model claims that tumor cells abandon the primary lesion before the acquisition of full malignant phenotype, and seed secondary growths where DTC undergo a stepwise progression of morphologic abnormalities. This model predicts greater genetic and epigenetic disparities between primary tumor cells and metastasis founders. Indeed, in several solid tumors, DTC exhibit significantly fewer genetic abnormalities than primary tumor cells, and heterogeneous chromosomal rearrangements can be found in primary tumors and DTC from different sites. The concept of early dissemination of tumor cells is also supported by the demonstration that in breast cancer models shortly after expression of the oncogenic transgene, neoplastic epithelial cells from atypical ductal hyperplasia disseminate to the lungs and bone marrow (2). Also, in women, bone marrow dissemination can occur at the stage of ductal carcinoma in situ (2).

Recent evidence strongly supports the role of cancer stem–like cells [CSC; (3)] in the metastatic process (4), and pancreatic epithelial cells with stem cell properties may invade the liver before frank malignancy is detected in the pancreas (5).

Once the DTC has reached the site of future metastasis, it has to survive and protect itself from immune surveillance (6). Indeed, in vitro CSC are killed by NK and T cells (7, 8), but also inhibit T-cell proliferation and effector function (9) through yet undefined mechanisms. Nothing is known about CSC–T cell interactions in vivo.

The transgenic adenocarcinoma of the mouse prostate (TRAMP) model relies on the SV40 early genes (small and large T antigens; Tag) expressed under the control of the rat probasin regulatory element, so that Tag appears at puberty selectively on the prostate epithelium (10). TRAMP mice progressively develop mouse prostate intraepithelial neoplasia (mPIN; week 6–12), adenocarcinoma (week 12–18), and lymph node (LN) and lung metastases (week 18–30) resembling the human pathology (11). CSCs obtained from prostates of 10- to 11-week-old TRAMP mice affected by mPIN [hereafter named TPIN-SC; (12)] express prostate cancer–associated antigens, MHC I and MHC II molecules, and ligands for NK cells. Indeed, CSC can be targeted by NK and cytotoxic T lymphocytes (CTL) both in vitro and in vivo (13). Nonetheless, they generated tumors when injected in immunocompetent mice (13), suggesting they possess mechanisms of immune evasion. Concomitantly, at that age, TRAMP males develop full CTL tolerance to Tag (14), which behaves in this model as a tissue-restricted tumor-associated antigen. We then asked whether TPIN-SC adopt strategies to overcome immune surveillance. Here, we show that TPIN-SC precociously disseminated to prostate draining LN (PDLN), and used Tenascin-C (TNC), an extracellular matrix (ECM) disulfide-linked hexameric glycoprotein (15) to inhibit T-cell proliferation and effector function.

C57BL/6, C57BL/6-Tg(TcraTcrb)425Cbn/Crl, C57BL/6-Tg(TcraTcrb)1100Mjb/Crl (Charles River), and B6.129S7-Rag1tm1Mom/J mice (16) were housed in a pathogen-free animal facility. The latter two were crossed to obtain RAG-1^−/−OT1 mice. Heterozygous TRAMP mice (10) were generated as described previously (14). Animals were treated in accordance with the European Community guidelines and with the approval of the Institutional Ethical Committee. Prostate CSC were cultured as described previously (12). Human peripheral blood mononuclear cells (PBMC) and B16-OVA melanoma cells (17) were cultured in IMDM (Lonza).

Prostate CSCs were cultured in serum-free DMEM/F12 (Lonza) containing EGF and FGF2. Murine splenocytes were cultured in T-cell medium (TCM), composed by RPMI (Lonza), with 8% fetal bovine serum (FBS; Invitrogen), 2 mmol/L l-glutamine, 150 U/mL streptomycin, and 200 U/mL penicillin (Cambrex), 10 mmol/L Hepes, 10 mmol/L sodium pyruvate and 5 μmol/L β-mercaptoetanol (Gibco-Invitrogen). Human PBMC and B16-OVA melanoma cells were cultured in IMDM (Lonza), with 10% FBS (Invitrogen). Unless specified, all chemical reagents were from Sigma-Aldrich. Peptides were kindly provided by R. Longhi (CNR, Milan, Italy).

Splenocytes were labeled with CFSE, and activated in vitro with anti-CD3 and anti-CD28 beads (Invitrogen) and IL2 (R&D Systems) according to the manufacturer's instructions. When needed, irradiated (50 Gy) prostate CSCs were added in coculture or in a Transwell system (0.4 μm) at the indicated CSC:splenocyte ratio. When indicated, soluble TNC (0.5μg/mL; R&D Systems) was added, or splenocytes were incubated with 0.5 μg/mL of isoDGR (18) and/or beads coated with anti-α₅ (CD49e) antibody (clone 5H10-27; 66μg/mL; BioLegend) 30 minutes before the addition of prostate CSC. CFSE-labeled splenocytes from transgenic OTII or Rag-1^−/− OTI mice were cocultured with irradiated CSC in the presence of OVA_323–339 (20 ng/mL) or OVA_257–264 (10 ng/mL) peptides, respectively, and 3.5 ng/mL of IL12 (R&D Systems). After 3 or 5 days, respectively, cells were analyzed by FACS. PDLN from TRAMP or wild-type (WT) mice were digested with collagenase IV (1,600 U/mL) for 1 hour at 37°C, labeled with CFSE, activated as described above, and analyzed after 3 days by FACS. When indicated, mice were treated i.p. (5 d/wk) with either AMD3100 (3.5 mg/kg) or vehicle from week 4 to 12 of age. Human PBMC isolated from buffycoats of healthy subjects, who gave their informed consent, were labeled with CFSE, activated with concanavalin A (5 μg/mL), and cultured and analyzed as described above.

Procedures are reported in Supplementary Materials and Methods.

OTI cells were activated as described above and cultured in SILAC (stable isotope labeling with amino acids in cell culture) medium containing light (¹²C and ¹⁴N) or heavy (¹³C ¹⁵N) labeled l-lysine and l-arginine, in the presence of irradiated prostate CSC. After 7 days of coculture, CD8⁺ T cells from either condition were sorted for activated (CD8⁺CD62L^low) or inhibited (CD8⁺CD62L^high) phenotype and lysed in RIPA buffer (Cell Signaling Technology). Heavy and light lysates were mixed at 1:1 for protein content; proteins were then resolved onto a 4% to 12% NuPAGE precast gel (Invitrogen) and stained by Coomassie colloidal blue. The gel lane was cut into 12 slices, each of which was reduced, alkylated, and digested with trypsin as reported previously (19). Mass spectrometry (MS) procedures, according to ref. 20, and analysis of differentially expressed proteins in T cells from SILAC cultures are described in Supplementary Material and Methods.

Procedures are reported in Supplementary Materials and Methods.

TPIN-SC were stably infected with TNC shRNA Lentiviral Particles or with control shRNA Lentiviral Particles (Santa Cruz Biotechnology, Inc.) as described in Supplementary Materials and Methods.

Procedures are described in Supplementary Material and Methods.

PDLN, or axillary LN as control, were isolated and digested with collagenase IV (1,600 U/mL) for 1 hour at 37°C. Cells were seeded in a serum-free medium containing EGF and FGF2, and cultures were splitted every 2 to 10 days, according to the stage of origin (12).

Real-time PCR was performed and analyzed using the ΔΔC_T method as described in Supplementary Experimental Procedures.

Statistical analyses were performed using the Student t or one-way ANOVA followed by Tukey tests. Values were considered statically significant for *, P < 0.05; **, P < 0.01; ***, P < 0.001.

To investigate potential interactions between CSC and the immune system in the early phases of neoplastic transformation, we took advantage of TPIN-SC, which were obtained from young TRAMP mice affected by mPIN, and TNE-SC isolated from neuroendocrine lesions developed in 30-week-old TRAMP mice (12). Both CSC types fulfilled the criteria of endless self-renewal ability, multi-lineage differentiation, and tumorigenic potential (12). However, gene expression analysis revealed that the two CSC types were remarkably different, resembling the stages of progression of human prostate cancer (12).

CFSE-labeled naïve splenocytes were cultured with anti-CD3/CD28 beads alone or in the presence of irradiated TPIN-SC or TNE-SC. Flow cytometry analysis of CD4⁺ and CD8⁺ T cells showed that in the presence of TPIN-SC, most of the T cells underwent growth arrest soon after activation (Fig. 1A). This phenomenon was highest at the CSC:splenocyte ratio 1:10 (Fig. 1B), and required cell-to-cell contact (Fig. 1A). Interestingly, TNE-SC did not inhibit T-cell proliferation (Fig. 1A), thus demonstrating that dampening T-cell activation is a peculiar characteristic of TPIN-SC.

Both CD4⁺ and CD8⁺ T cells in the presence of TPIN-SC and CD3/CD28 beads received an activation signal strong enough to induce upregulation of several activation markers, including CD44 and LFA-1 (Fig. 1C and Supplementary Fig. S1A). However, in those cells, CD69 remained exceedingly upregulated, whereas CD25 did not reach the expression levels of fully activated T cells (Fig. 1D and Supplementary Fig. S1B). Aborted activation was confirmed by the lack of CD62L downregulation, (Fig. 1D and Supplementary Fig. S1B). Moreover, T cells cultured with TPIN-SC showed a higher expression of CCR6 and a lower expression of CXCR3, the latter being a characteristic of very recently activated T cells, while, as expected, it was upregulated in fully activated T cells (Fig. 1E and Supplementary Fig. S1C). In line with the observation that CXCR3 is required for optimal IFNγ production by T cells (21), the frequency of IFNγ⁺ T cells was dramatically reduced in TPIN-SC-splenocyte cocultures (Fig. 1E and Supplementary Fig. S1C), confirming that TPIN-SC interfered with full activation of T cells. TPIN-SC–mediated T-cell inhibition was not species specific, because also proliferation of human PBMCs was inhibited by TPIN-SC and not by TNE-SC (Fig. 1F and Supplementary Fig. S1D).

To verify that T-cell inhibition ability was a peculiarity of TPIN-SC and not an artifact of irradiation, which can induce release of inhibitory molecules, we repeated the above-described experiments with nonirradiated CSC. In line with the results reported in Fig. 1, nonirradiated TPIN-SC inhibited proliferation (Supplementary Fig. S2A) and IFNγ production (Supplementary Fig. S2B and S2C) of both CD4⁺ and CD8⁺ T cells in a cell-to-cell contact fashion (Supplementary Fig. S2A). As expected, nonirradiated TNE-SC had no effect on T-cell proliferation and cytokine production (Supplementary Fig. S2).

Also, antigen-specific proliferation and IFNγ production induced by ovalbumin (OVA) in both OTII and OTI transgenic T cells, which express TCR specific for OVA (22, 23), were inhibited by TPIN-SC (Fig. 2A) and not by TNE-SC (data not shown). Of relevance, coculture with TPIN-SC did not increase the number of Annexin V⁺ T cells, ruling out the possibility that CSC favor T-cell apoptosis (Fig. 2B).

To verify whether inhibition occurs also in vivo, a mixture (10:1 ratio) of naïve CFSE-labeled OTI cells and irradiated TPIN-SC was injected in the spleen of C57BL/6J mice bearing a well-established B16-OVA melanoma (17). Their spleens and tumors were collected 5 days later, to quantify infiltrating CFSE⁺ cells. The presence of TPIN-SC associated with a reduced expansion of OTI cells in the spleen, and a reduced tumor infiltration by OTI cells, thus confirming the immunosuppressive activity of TPIN-SC also in vivo (Fig. 2C).

TPIN-SCs also inhibited restimulation of previously activated T cells. When splenocytes were collected from mice that had been vaccinated with dendritic cells pulsed with Tag (14), and restimulated in vitro, coculture with TPIN-SC dampened proliferation and IFNγ production (Supplementary Fig. S3A) and cytolytic activity (Supplementary Fig. S3B), whereas TNE-SC were ineffective. As for naïve T cells, inhibition of antigen-experienced T cells required cell-to-cell contact (Supplementary Fig. S3A and S3B). Notably, when TPIN-SC were added to the culture of fully activated CD4⁺ and CD8⁺ T cells, no inhibitions in subsequent T-cell proliferation and cytokine production were measured (Supplementary Fig. S3C).

Collectively, these data suggested that TPIN-SC acted during the early steps of T-cell activation. Phosphorylation of ZAP70, ERK2, and STAT5, all proximal signaling events linked to TCR and IL2 receptor triggering, respectively, was significantly reduced in T cells activated in the presence of TPIN-SC (Supplementary Fig. S4A). Thus, TPIN-SC–conditioned T cells might have acquired phenotypic and functional characteristics of hypo-responsive lymphocytes (24). To assess this possibility, SILAC and MS procedures were implemented to quantify protein variations in cocultures of T cells with CSCs. Proteins extracted from T cells inhibited by the presence of TPIN-SCs (sorted for being CD62L^hi) and from fully activated T cells in the presence of TNE-SCs (sorted for being CD62L^low; Supplementary Fig. S4B) were analyzed by high-resolution MS, allowing the reliable identification and quantification of about 1,000 proteins (Supplementary Fig. S4B). Statistical analysis by MaxQuant highlighted 100 of them to be differentially expressed (P < 0.05; Supplementary Table S1) and query of Gene Ontology and KEGG databases showed that proteins involved in cytoskeleton organization and glucose metabolism were significantly modified in T cells after coculture with TPIN-SCs (Supplementary Table S2). Interestingly, TPIN-SC–conditioned T cells maintained upregulated proteins involved in pyruvate metabolism (Supplementary Table S2), which is a peculiarity of resting T cells (25), and failed to upregulate the amino acid transporter CD98 and the transferrin receptor CD71 (Supplementary Fig. S4B), suggesting an impaired change of the metabolic machinery, a characteristic of anergic T cells (25). When CD8⁺ T cells were sorted from cultures of either T cells alone or with TPIN-SCs, and subjected to a second round of stimulation in the absence of TPIN-SCs, they proliferated and downregulated CD62L in comparable way (Supplementary Fig. S4C), demonstrating that the TPIN-SC–mediated T-cell hyporesponsiveness is reversible.

We compared the transcriptome of TPIN-SC with that of TNE-SC (12), founding 37 differentially expressed genes (fold change > 2; P < 0.001), among which we identified the ECM protein TNC, upregulated in TPIN-SC (Fig. 3A) as a potential immunosuppressive molecule. TNC provides physical and signaling support for CSC and metastasis initiating cells (26), and can be immunosuppressive in vitro (27), yet each of these functions had been independently reported.

Expression of TNC in TPIN-SC and not in TNE-SC was confirmed by both Western blot analysis (Fig. 3B), and immunofluorescence (Fig. 3C). Moreover, when soluble TNC, whose sequence is conserved between humans and mice, was added to the cultures of human (Fig. 3D) or mouse T cells (Fig. 3E), it mirrored the effects of TPIN-SC on proliferation, expression of CD62L, and cytokine production. Thus, TNC was a strong candidate for mechanistic TPIN-SC–mediated immunosuppression.

Whereas in the prostate of WT mice at any age (Fig. 4A), expression of TNC was absent (Fig. 4B); in mPIN lesions from TRAMP mice (Fig. 4A), TNC showed a patchy distribution mostly within the stroma surrounding affected acini, and in a few transformed epithelial cells (Fig. 4B). A faint cytoplasmic expression of TNC was present in TRAMP adenocarcinoma cells (Fig. 4B), and absent in neuroendocrine tumors (Fig. 4B). In good correlation with the TRAMP model, TNC was expressed in human PIN (Fig. 4B), and not in human adenocarcinoma (Fig. 4B). TNC was also expressed in metastatic LN, confirmed by cytokeratin staining (Fig. 4C), of both TRAMP (Fig. 4C) and prostate cancer patients (Fig. 4C), but not in metastatic LN from TRAMP mice affected by neuroendocrine tumors (Fig. 4C, middle).

TPIN-SC were infected with lentiviral vectors encoding either a TNC-specific or a scrambled short hairpin RNA (shRNA-TNC and shRNA-ctr, respectively). Both Western blot analysis (Fig. 3B) and immunofluorescence (Fig. 3C) confirmed the specific silencing of TNC. Reduction of TNC expression in TPIN-SC did not influence their in vitro growth, because TPIN-SC infected with either shRNA-TNC or shRNA-ctr showed similar growth rate (data not shown). However, specific silencing of TNC substantially dampened the effects of TPIN-SC on T-cell proliferation and modulation of activation markers (Fig. 5A), and fully rescued the capacity of T cells to produce IFNγ upon activation (Fig. 5A). This was confirmed by the increase of ZAP70 and ERK2 phosphorylation levels (Fig. 5B and Supplementary Fig. S5A and S5B). Thus, TNC is directly involved in TPIN-SC–mediated immunosuppression. Interestingly, silencing of TNC did not abrogate the inhibition of TPIN-SC of STAT5 phosphorylation (Fig. 5B and Supplementary Fig. S5C), suggesting that TNC does not affect IL2 signaling pathway, and that additional mechanisms cooperate in the immunosuppressive activity of TPIN-SC on T cells.

TNC silencing abrogated the inhibitory activity of TPIN-SC on T cells also in vivo. Indeed, the frequency of CFSE-labeled OTI cells was similar in the spleens and tumors of mice bearing B16-OVA melanoma, irrespective of the fact that they were infused 5 days earlier with naïve CFSE-labeled OTI cells either admixed or not to TPIN-SC-shRNA-TNC cells (Fig. 5C).

TPIN-SC infected with either shRNA-TNC or shRNA-ctr were injected subcutaneously within Matrigel plugs into fully immunocompetent C57BL/6 mice or RAG-1^−/−OT1 mice, which lack B and T cells but OVA-specific CD8⁺ T cells. In agreement with our previous findings (13), TPIN-SC-shRNA-ctr cells generated tumors in 100% of RAG-1^−/−OT1 mice, and 60% of C57BL/6 mice (Fig. 5D), supporting a role for adaptive immunity in surveillance against TPIN-SC–derived tumors. Because no tumor grew in C57BL/6 mice challenged with TPIN-SC-shRNA-TNC cells (Fig. 5D), we conclude that TNC produced by TPIN-SCs negatively affected adaptive immune surveillance. TNC also has a function on tumor biology, because only 60% of RAG-1^−/−OT1 mice developed tumor upon challenge with TPIN-SC-shRNA-TNC cells (Fig. 5D).

TPIN-SC required a cell-to-cell contact to arrest T-cell activation (Fig. 1A). To further investigate the role of TNC in this cell-to-cell interaction, contacts between CFSE-labeled prostate CSC and CMTMR-labeled CD8 T cells were recorded by multichannel time-lapse fluorescent live cell imaging during the first 4 hours of coculture (Fig. 6A). Stimulated T cells were engaged in more prolonged contacts with TPIN-SC than with TNE-SC, and, more importantly, silencing of TNC in TPIN-SC reduced the duration of contacts to the level of TNE-SC (Fig. 6B and C). Thus, TNC appeared to be directly involved in CSC–T cell contacts, eventually leading of T-cell inhibition.

Intracellular signaling triggered by TCR stimulation is intimately related to the reorganization of the actin/myosin and microtubule cytoskeleton through small GTPases, including Rho, Rac, and Cdc42 (28). Inhibition of Rho in T cells suppresses T-cell proliferation (29), and TNC inhibits Rho activation (30). Thus, we hypothesized that TNC arrests T-cell proliferation by inhibiting cytoskeleton reorganization. In support of this hypothesis, SILAC analysis on T cells inhibited by TPIN-SC showed downregulation of actin gamma 1 (Actg1) and myosin heavy chain 10 (Myh10), two molecules involved in actin stress fibers assembly and focal adhesion formation (Supplementary Table S1). SILAC analysis revealed that other molecules involved in Rho and Rac pathways were differentially expressed in T cells inhibited by TPIN-SCs. Indeed, in TPIN-SCs conditioned T cells, we found upregulation of the Rho GDP dissociation inhibitors 1 and 2 (Arhgdia and Arhgdib), which prevent Rho activation by blocking the release of GDP, an essential step for GTP loading and activation (Supplementary Table S1; ref. 31). Cofilin, a downstream effector of both RhoA and Rac, which inhibits actin polymerization (32) was also upregulated in TPIN-SC–conditioned T cells (Supplementary Table S1). Conversely, the Rho GTPase-activating protein 1 (Arhgap1) was downregulated (Supplementary Table S1). Also Crk, which can activate small GTPAses including Rac (33), and the Rac downstream effectors Vinculin and Gelsolin were downregulated in TPIN-SCs conditioned T cells (Supplementary Table S1).

Thus, to corroborate the hypothesis that TNC inhibits cytoskeleton reorganization, polymers of F-actin were quantitated in T cells by staining with FITC-conjugated phalloidin. As expected, T cells increased actin polymerization after PMA stimulation and this process was inhibited by pretreatment with cytocalasin D (Fig. 6D and data not shown), which selectively blocks the ATP-dependent actin polymerization process. Actin polymerization was inhibited also when T cells were activated in the presence of TPIN-SC or TNC and, most importantly, silencing of TNC in TPIN-SC rescued actin polymerization in both CD4⁺ and CD8⁺ T cells (Fig. 6D).

Several molecules can bind to TNC, including integrins (e.g., α2β1, α5β1), annexin II, and EGFR (15). To assess the role of integrins in the TNC-mediated inhibition of T-cell functions, the peptide isoDGR, which selectively inhibits interaction between α5β1, αvβ6, and αvβ8 integrins and their ligands (18), or a blocking anti-α5 monoclonal antibody (34) were added to TPIN-SC-T cells cultures. Because both reagents inhibited the suppressive activity of TPIN-SC, and the combination of the two reagents did result in neither additive nor synergy effects (Fig. 6E), we conclude that TNC mostly interacts with α5β1 to deliver its immunosuppressive effect. Collectively, these findings demonstrate that TNC produced by CSC arrest T-cell activation by interacting with α5β1 and blocking reorganization of actin-based cytoskeleton.

Priming of naïve T cells and restimulation of memory T cells occur mostly in secondary lymphoid organs. Because TPIN-SC inhibited T-cell priming and restimulation, and in TRAMP mice, peripheral tolerance to Tag is reached at the stage of mPIN (14), we hypothesized that in TRAMP mice, CSC deploy early to PDLN, where they suppress T-cell activation. Indeed, TPIN-SC expressed CXCR4 (Supplementary Fig. S6A), which is essential for CSC-mediated tumor metastasis (4), and migrated in vitro through CXCL12. Moreover, the CXCR4 inhibitors AMD3100 and Peptide R (35) restrained their migration (Supplementary Fig. S6B and S6C). We found clear evidence that expression of CXCL12 was higher in PDLN from TRAMP mice, starting from week 8, than in NDLN of TRAMP mice or in PDLN of WT mice (Supplementary Fig. S6D). Interestingly, TPIN-SC did not express CCR7, and expressed low levels of CCR6 and CXCR3 (Supplementary Fig. S6E). Thus, CSC from mPIN lesions might primarily use the CXCR4–CXCL12 axis to migrate and deploy to PDLN.

Hence, we looked ex vivo for the presence of putative prostate CSC [enriched within the population of CD45⁻CD31⁻CD44⁺CD166⁺Sca1⁺ cells; (13, 36)] in PDLN of TRAMP mice either young enough not to show any prostate lesion (6-week-old), or affected by full-blown tumor (16-week-old). PDLN from both 6-week-old TRAMP and WT mice did not contain any of these cells (Fig. 7A); conversely, they were enriched in PDLN of 16-week-old TRAMP mice. To investigate whether prostate CSC can disseminate from early primary lesions, PDLN and LN not draining the prostate (NDLN) from 12-week-old TRAMP affected by mPIN and WT mice were processed to single-cell suspension and cultured to generate CSC lines (12). Although PDLN from TRAMP mice gave rise to CSC lines (TLN-SC) with efficiency close to 80%, neither TRAMP NDLN nor PDLN and NDLN from WT mice ever gave rise to prostaspheres (Fig. 7B). TLN-SC showed a phenotype comparable with TPIN-SC, and substantially different from ex vivo whole prostate epithelial cells collected from age-matched TRAMP mice (Fig 7C). As confirmation of their origin, TLN-SC expressed the prostate antigens STEAP (37) and PSCA (Fig. 7D; ref. 38). We found higher TNC in TRAMP PDLN than in NDLN or PDLN of age-matched WT mice (Fig. 7E). More importantly, TLN-SC expressed TNC (Fig. 7F) and recapitulated the immunosuppressive effects of TPIN-SC on mouse T cells (Fig. 7G).

The existence of an immunosuppressive environment in PDLN of 12-week-old TRAMP mice was confirmed by the finding that CD8⁺ T cells from TRAMP PDLN proliferated less than T cells from PDLN of age-matched WT mice (Fig. 7H). Importantly, when TRAMP mice were treated with AMD3100 starting at 4 weeks of age, CD8⁺ T-cell proliferation in their PDLN at 12 weeks returned comparable with WT mice either treated or not with the drug (Fig. 7H).

Thus, CSCs early disseminate to PDLN, where they use TNC to avoid immune surveillance, and a precocious treatment with a CXCR4 inhibitor can positively affect local CSC-mediated immunosuppression.

We have found that CSC, already at the stage of mPIN, deploy to PDLN that at pathologic examination appear as nonmetastatic. Thus, prostate cancer appears to adhere to the model of parallel metastatic spread (1). In support of this model, 13.3% of patients with pT3 prostate carcinoma treated by radical prostatectomy and bilateral LN dissection, and classified as node negative, harbored instead occult LN metastases, which were an independent predictor of recurrence and death in a multivariable analysis (39).

We also found that CSC either from mPIN or PDLN of mice affected by mPIN likely overcome immune surveillance by inhibiting T-cell proliferation and effector functions. Several experimental evidences let us conclude that TNC is one major mechanism by which CSCs inhibit T-cell response in TRAMP mice. Indeed, prostate CSC overexpressed TNC and TNC silencing abrogated CSC-mediated immunosuppression. Furthermore, TNC in vitro recapitulated the immunosuppressive activity of CSC.

Whereas TNC does not correlate with prostate cancer Gleason score (40), data reported herein in ref. 40 indicate that in normal prostate and low-grade PIN, TNC is weakly expressed in the ECM and acinar basement membrane, its expression increases in high-grade PIN, returns to basal levels in prostate adenocarcinoma, and is again overexpressed in metastases. TNC also is overexpressed in melanoma spheres enriched in CSC, supporting their growth, metastatic potential, and resistance to chemotherapy (41). Accordingly, while TNC was dispensable for TPIN-SC growth in vitro, it had a role in supporting their tumor formation in vivo. Thus, TNC has a dual role for prostate CSC in vivo; on the one hand, it sustains tumor growth, and on the other hand, it inhibits T cell–mediated immune surveillance. TNC is therefore a very interesting target of cancer therapy, and anti-TNC antibodies are already in the clinic (42). It is also tempting to speculate that overexpression of TNC in high-grade PIN and low Gleason score patients, because of the role of TNC in tissue remodeling and metastasis (26), might identify a subgroup of patients with increased chances of early LN invasion. This hypothesis should be investigated in a large cohort of patients.

CSCs use TNC to block T-cell activation during priming/restimulation, but are ineffective on fully activated T cells, as it has been reported for TNC (27). Hence, CSC do not appear to directly interfere with TCR–MHC interactions, nor TNC interferes with LFA1–ICAM1 interaction (27). It is well known that TNC can bind several molecules among integrins (15). In particular, the fragment FnIII 1–5 of TNC inhibits α5β1-dependent T-cell adhesion to fibronectin (43), while the fragment FnIII A1A2 inhibits T-cell proliferation through a yet unknown mechanism (44). Although we did not investigate which TNC fragment is involved in the CSC-mediated immunosuppressive activity, our in vitro findings strongly support a mechanism by which TNC interacts with α5β1 integrins on T-cell surface, thus likely inhibiting Rho activation as previously reported in fibroblasts (30), and blocking actin stress fibers assembly and focal adhesion formation. The signaling cascade triggered by TCR engagement activates Rho, Rac, and Cdc42 GTPases (31). In particular, phosphorylated Zap70 can activate a pathway leading to the formation of a complex of proteins, including VAV1, which facilitates the exchange from GDP to GTP and thus the activation of Rho, Rac, and Cdc42 (32). As TNC inhibits TCR signaling and Zap70 phosphorylation in particular, this could be a complementary mechanism by which TNC inhibit Rho, Rac, and Cdc42 GTPases.

Both CSCs and TNC significantly reduced phosphorylation of ZAP70 and ERK2 and induced substantial metabolic alterations in T cells, as those described in T-cell anergy (24). The state of T-cell hyporesponsiveness induced by TPIN-SCs does not corresponds to the physiologic state of activation-induced nonresponsiveness (45), because T cells lost not only proliferation competence, but also IFNγ production and cytotoxic activity; neither can it be considered as adaptive tolerance, because removal of CSC/TNC allowed full T-cell recovery. Thus, the effect mediated by CSC on T cells is characteristic more of cell-mediated immunosuppression (24).

Silencing of TNC did not completely abrogate the immunosuppressive activity of CSC and did not rescue STAT5 phosphorylation, suggesting that other molecules are involved in CSC-mediated immunosuppression. Indeed, microarray-based whole transcript showed that both TPIN-SC and TLN-SC produce several other immunomodulatory factors, currently investigated by us.

Peripheral tolerance to tumor-associated antigens in prostate cancer may be induced and maintained by several additional mechanisms. Indeed, CSC and/or other more differentiated tumor cells may generate immunosuppressive networks within the primary lesion, and anergic Tag-specific CD8⁺ T cells are found in the prostate of TRAMP mice (14). Both regulatory T cells and myeloid-derived suppressor cells accumulate in the prostate of cancer patients and TRAMP mice, although their targeting revealed insufficient for breaking T-cell tolerance and sustaining antitumor immunity (46, 47). Moreover, in TRAMP and in human prostate cancer lesions, a population of tolerogenic DCs induces CD8⁺ T cells to acquire suppressive functions through indoleamine-2,3-dioxygenase, arginase, TGFβ, and PDL-1 (48).

Our results merge the evidence of early tumor cell dissemination with tumor-induced immune escape, with relevant clinical implications. In our cohort, TRAMP mice never developed LN metastasis from prostate adenocarcinoma before week 17 (data not shown), suggesting that CSC that have colonized PDLN remain quiescent for several weeks. Immune surveillance appears to be involved in limiting the metastatic outgrowth, as cytostatic CD8⁺ T cells have been shown to maintain DTC dormancy in visceral organs (49). Thus, strategies that either block the interaction between TNC and α5β1, such as antibodies (42) or isoGDR peptides, or the CXCR4–CXL12 axis (50), as the one described herein, or both TNC and CXCR4 should unsheathe early-disseminated CSC and avoid full-blown metastasis development, and might be proposed either alone or in combination to subjects with high risk to develop prostate cancer or to patients at high risk for disease recurrence after radical prostatectomy. Although in our cohort of TRAMP mice, the earliest metastasis from adenocarcinoma was found in the kidney at 16 weeks of age, metastasis to soft tissues in this model is rather infrequent and stochastic. Hence, efficacy of the proposed treatments in preventing metastasis occurrence to soft tissues might better be investigated in other models.

No potential conflicts of interest were disclosed.

Conception and design: E. Jachetti, S. Caputo, A. Calcinotto, M. Bellone

Development of methodology: E. Jachetti, S. Caputo, S.M. Parigi, A. Calcinotto, A. Bachi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Jachetti, S. Caputo, C.S. Brambillasca, S.M. Parigi, M. Grioni, U. Restuccia, A. Calcinotto, M. Freschi, A. Bachi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Jachetti, S. Caputo, C.S. Brambillasca, S.M. Parigi, I.S. Piras, U. Restuccia, M. Freschi, A. Bachi, M. Bellone

Writing, review, and/or revision of the manuscript: E. Jachetti, S. Caputo, U. Restuccia, M. Freschi, A. Bachi, M. Bellone

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Jachetti, M. Grioni

Study supervision: E. Jachetti, R. Galli, M. Bellone

Other (prostate cancer stem cell lines establishment and characterization): S. Mazzoleni

The authors thank Drs. A. Mondino, M.P. Protti, R. Pardi, I. De Curtis, A. Corti, and F. Curnis (San Raffaele Scientific Institute, Milan, Italy) and R. Longhi (Consiglio Nazionale Ricerche, Milan, Italy) for critical comments, helpful suggestions, and reagents, and Dr. S. Scala (Fondazione Pascale, Napoli, Italy) for Peptide R.

The work was supported by Associazione Italiana per la Ricerca sul Cancro (AIRC; # IG11692) and Ministero della Salute. E. Jachetti and A. Calcinotto were awarded an AIRC/FIRC fellowship. S. Caputo and A. Calcinotto conducted this study in partial fulfillment of their Ph.D. at San Raffaele University.

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.