SDF-1/CXCL12 induces directional cell migration and spontaneous metastasis via a CXCR4/Gαi/mTORC1 axis

Abstract

Multiple human malignancies rely on C-X-C motif chemokine receptor type 4 (CXCR4) and its ligand, SDF-1/CXCL12 (stroma cell–derived factor 1/C-X-C motif chemokine 12), to metastasize. CXCR4 inhibitors promote the mobilization of bone marrow stem cells, limiting their clinical application for metastasis prevention. We investigated the CXCR4-initiated signaling circuitry to identify new potential therapeutic targets. We used HeLa human cancer cells expressing high levels of CXCR4 endogenously. We found that CXCL12 promotes their migration in Boyden chamber assays and single cell tracking. CXCL12 activated mTOR (mechanistic target of rapamycin) potently in a pertussis-sensitive fashion. Inhibition of mTOR complex 1 (mTORC1) by rapamycin [drug concentration causing 50% inhibition (IC₅₀) = 5 nM] and mTORC1/mTORC2 by Torin2 (IC₅₀ = 6 nM), or by knocking down key mTORC1/2 components, Raptor and Rictor, respectively, decreased directional cell migration toward CXCL12. We developed a CXCR4-mediated spontaneous metastasis model by implanting HeLa cells in the tongue of SCID-NOD mice, in which 80% of the animals develop lymph node metastasis. It is surprising that mTORC1 disruption by Raptor knockdown was sufficient to reduce tumor growth by 60% and spontaneous metastasis by 72%, which were nearly abolished by rapamycin. In contrast, disrupting mTORC2 had no effect in tumor growth or metastasis compared with control short hairpin RNAs. These data suggest that mTORC1 may represent a suitable therapeutic target in human malignancies using CXCR4 for their metastatic spread.—Dillenburg-Pilla, P., Patel, V., Mikelis, C. M., Zárate-Bladés, C. R., Doçi, C. L., Amornphimoltham, P., Wang, Z., Martin, D., Leelahavanichkul, K., Dorsam, R. T., Masedunskas, A., Weigert, R., Molinolo, A. A, Gutkind, J. S. SDF-1/CXCL12 induces directional cell migration and spontaneous metastasis via a CXCR4/Gαi/mTORC1 axis.

Chemokines are master regulators of cell migration (1). Currently, there are 48 known molecules that belong to this cytokine superfamily, which are divided in 2 groups depending on their function: inflammatory and homeostatic. Inflammatory chemokines, such as IL-8, C-X-C motif chemokine (CXCL) 1, and CXCL2, are induced during inflammatory events, and multiple chemokines can bind and activate shared receptors (2). In contrast, homeostatic chemokines are constantly expressed and released as they play crucial physiologic roles, such as tissue regeneration and stem cell maintenance, which also results in a more restricted ligand usage by its receptors (3). The C-X-C motif chemokine receptor type 4 (CXCR4) 7-transmembrane G protein–coupled receptor binds to the homeostatic C-X-C motif chemokine 12 (CXCL12, also known as stromal cell–derived factor 1, or SDF-1). CXCL12 is secreted in normal, nonpathologic state by many organs such as liver, bone marrow, lung, and lymph nodes (2). CXCR4 expression correlates with poor prognosis in many tumor types, and most of the sites that secrete its ligand, CXCL12, are frequently colonized by metastatic cells (4). Given the key role that CXCR4 plays in cancer metastasis, therapeutic targeting of CXCR4 has been tested in clinical trials (5, 6). Although effective in decreasing metastatic disease, CXCR4 inhibitors promote the mobilization of bone marrow stem cells from their niche (5), a side effect that limits their clinical application. In fact, CXCR4 antagonists are currently approved by the U.S. Food and Drug Administration for hematopoietic stem and progenitor cell mobilization for autologous transplantation in patients with non-Hodgkin lymphoma and multiple myeloma (6–8). To circumvent this, we hypothesized that the study of the CXCR4 downstream signaling circuitry may help identify targets that could be affected by drugs and that could potentially be explored as new therapeutic options for many human malignancies that depend on CXCR4 for their metastatic spread without provoking hematopoietic stem and progenitor cell mobilization.

The PI3K/Akt/mTOR (phosphoinsoitide 3-kinase/AKT/mechanistic target of rapamycin) signaling pathway plays a major role in regulating cell growth and survival and is one of the most frequently deregulated biochemical routes in cancer (9). It is noteworthy that many oncogenes and tumor suppressor genes directly or indirectly affect this pathway, and mutations of and genetic and epigenetic alterations in components of the PI3K/mTOR pathway are among the most prevalent alterations in human malignancies (10–13).

mTOR is a major effector of PI3K/Akt pathway, and CXCL12 signaling through CXCR4 has been shown to induce Akt phosphorylation/activation (14, 15). However, mTOR exerts different functions depending on the complex in which it is engaged. Two accessory proteins, known as Raptor (regulatory-associated protein of mTOR) and Rictor (rapamycin-insensitive companion of mTOR), define mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2), respectively (16, 17). Both complexes have been associated with a variety of functions, including cell metabolism, growth, proliferation, autophagy and protein synthesis for mTORC1, and survival, metabolism, and cytoskeletal organization for mTORC2 (9, 18). The effectiveness of inhibitors of the mTOR pathway for cancer treatment is under current evaluation in multiple clinical settings, and some mTOR inhibitors have already been approved for clinical use (19–23). Nonetheless, the precise role of mTOR and its complexes in each tumor type is yet to be fully understood, as is the possibility that mTOR may contribute to cancer metastasis.

Although the ability of CXCL12 to stimulate mTOR is well established, the functional contribution of mTOR signaling to CXCR4-mediated migration and metastasis is poorly understood. The latter may be of direct cancer relevance, as mTOR blockade is not known to cause bone marrow stem cell mobilization (24, 25), whereas it has been reported in gastric carcinomas that mTOR is required for CXCL12-mediated migration in vitro (26). By use of cells that express CXCR4 endogenously, we show that CXCR4/Gαi activates mTORC1 and mTORC2, and that this activation is required for chemotaxis. Moreover, we took advantage of a novel in vivo system to monitor CXCR4-mediated spontaneous metastasis to the lymph nodes to investigate whether mTOR represents a suitable antimetastatic target. It is surprising that we found that, although the 2 mTOR complexes play a role in CXCR4-mediated migration in vitro, only mTORC1 disruption decreases tumor growth and the ability of tumor cells to spontaneously metastasize to lymph nodes. This suggests that rapamycin and its analogs, which inhibit primarily mTORC1, may represent promising targeted agents preventing metastasis of many highly aggressive cancers that use CXCR4 for the guided migration of cancer cells from their primary tumors to their secondary colonization sites.

MATERIALS AND METHODS

Reagents

All chemical and reagents were purchased from Sigma-Aldrich (Woodlands, TX, USA) and all antibodies were purchased from Cell Signaling Technology (Beverly, MA, USA) unless otherwise stated. mTOR inhibitors rapamycin and Torin2 were purchased from LC Laboratories (Woburn, MA, USA) and Tocris Bioscience (Ellisville, MO, USA), respectively. CXCL12, epidermal growth factor (EGF), and lysophosphatidic acid (LPA) were purchased from R&D Systems (Minneapolis, MN, USA).

Cell culture, transfection, and lentivirus infection

HeLa cells were cultured in DMEM supplemented with 10% fetal bovine serum at 37°C in 95% air/5% CO₂ (Invitrogen, Carslbad, CA, USA). Small interfering RNA (siRNA) transfection was performed using Lipofectamine RNAiMAX reagent and 50 nM of SMARTpool siRNA for Raptor or Rictor (Thermo Fisher Scientific, Woburn, MA, USA). All analyses were performed between 48 and 72 h after transfection. Stable knockdown of Raptor, Rictor, and CXCR4, and H2B-GFP stable cell lines were achieved by infecting HeLa cells with lentivirus expressing the respective short hairpin RNA (shRNA) (Open Biosystems, Huntsville, AL, USA) or H2B-GFP (Addgene, Cambridge, MA, USA). Selection was started 7 d after infection using puromycin (1 μg/ml). Experiments using knockdown cells were performed 5 to 7 passages after selection was done, always in the presence of puromycin.

Chemokine receptor expression profile

Gene expression analysis was performed by RNA sequencing. RNA was isolated from cell lines during exponential growth and then submitted to RNA sequencing. Indexed RNA sequencing libraries were prepared using Truseq RNA sample Prep Kit, version 2 (Illumina, San Diego, CA, USA) and sequenced in paired-end mode on a Illumina Hiseq2000 sequencer. Raw data were mapped to the human genome (hg19 version) using GSNAP software in SNP-aware mode. Aligned reads were imported into the AVADIS NGS v1.6 software for read filtering and transcript quantification using the DESeq algorithm. Clustering and other statistical analysis was performed by Avadis NGS.

Immunoblot analysis

Proteins from subconfluent HeLa cells were extracted at 4°C using lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1% Nonidet P-40) supplemented with proteases and phosphatase inhibitors (protease inhibitor cocktail, Sigma-Aldrich; 1 mM Na₃VO₄ and 1 mM NaF). Protein quantification was assessed using DC protein assay (Bio-Rad, Hercules, CA, USA) following the manufacturer’s protocol. A total of 30 μg of protein was loaded in an SDS-PAGE, and proteins were transferred to nitrocellulose membranes. All membranes were blocked in milk and incubated with primary antibody (1:1000) for 1 h at room temperature. The following primary antibodies were used: phospho-S6, S6, phospho-Akt 473 and 308, glyceraldehyde phosphate dehydrogenase, and α-tubulin. To detect signal, we used secondary antibodies conjugated to fluorochrome: monkey anti-rabbit coupled to IRDye800CW and goat anti-mouse coupled to IRDye 700CW (Li-Cor Bioscience, Lincoln, NE, USA). Images were acquired using Odyssey (Li-Cor Bioscience) and processed using Odyssey Application Software, v3 (Li-Cor Bioscience).

Boyden chamber chemotaxis assay

Chemotaxis assay was performed using a 48-well Boyden chamber with an 8 μm pore size polyvinyl pyrrolidone–free polycarbonate membrane (NeuroProbe, Gaithersburg, MD, USA). Cells were serum starved, and 8 μm membranes were coated with collagen type I (10 μg/ml) at 4°C for 6 to 8 h (BD Bioscience, San Jose, CA, USA). Next, cells were added to the upper chamber and chemoattractant was added to the lower chamber in serum-free medium, and cells were allowed to migrate for 16 h in humid chambers at 37°C. Membrane was then fixed, stained with hematoxylin, and cells from the upper chamber removed with a cotton swab. Migrated cells (in the lower chamber) were counted under a high-magnification field.

Chemotaxis assay using μ-Slides

Chemotaxis assay was performed following manufacture's instruction. In brief, the observation area of the chemotaxis chamber (IBIDI LLC, Verona, WI, USA) was coated with fibronectin (100 μg/ml) or collagen type I (10 μg/ml) for 1 h followed by 1 wash with water, then let dry for 1 h. HeLa cells transfected with siControl, siRaptor, or siRictor 48 h before the assay were seeded in serum-free condition and left to adhere for 3 to 4 h at 37°C in a humid chamber. CXCL12 (90 ng/ml) was used as a chemoattractant and was added to the upper reservoir. Images were collected every hour for 24 h with a 10× objective using a Zeiss LSM 700 confocal microscope equipped with a CO₂- and temperature-controlled chamber. Data were analyzed by the ImageJ (http://rsb.info.nih.gov/ij/) plug-in Manual Tracking (http://rsbweb.nih.gov/ij/plugins/track/track.html), followed by the Chemotaxis and Migration tool from IBIDI (http://www.ibidi.de/applications/ap_chemo.html).

Animal studies

All animal studies were carried out according to U.S. National Institutes of Health (NIH)–approved protocols (ASP 07-442), in compliance with the NIH Guide for the Care and Use of Laboratory Animals. Female SCID-NOD (NCI, Frederick, MD, USA), 4 to 6 wk of age and weighing 18 to 20 g, were used in the study and were housed in appropriate sterile filter-capped cages and permitted food and water ad libitum. All handling and tumor implantation procedures were conducted in a laminar-flow biosafety hood.

Establishment of spontaneous metastasis model in SCID-NOD mice

HeLa cells were cultured to 70–80% confluence, trypsinized, washed in PBS, and suspended in serum-free DMEM at 50,000 cells/50 μl. For implanting, animals were anesthetized with 2% to 5% isoflurane (Baxter Healthcare Corporation, Deerfield, IL, USA), the tongue was exposed using tweezers, and 50 μl of media containing the cells was injected submucosally into the posterior area of the tongue. The animal was then monitored until recovery (1 to 2 min). All animals were then assessed weekly for tumor formation and progression. All animals developed visible tumors in the tongue after 3 wk. For the initial analysis of tumor metastasis, animals were euthanized at wk 5 after tumor implantation, and inguinal, abdominal, and cervical lymph nodes were retrieved and analyzed histopathologically. For all further studies, only cervical lymph nodes were analyzed.

Flank injections were performed initially with different cell concentration injected subcutaneously in SCID-NOD mice. A total of 2,000,000 cells/100 μl was the condition where all animals developed visible tumors after 3 wk, thus resembling the tongue model. Animals were monitored weekly and euthanized at wk 5 after tumor implantation. Inguinal, abdominal, and cervical lymph nodes were retrieved and analyzed histopathologically.

Immunofluorescence, immunohistochemistry, and histopathologic analysis

Tongues were cut into 4 sections of approximately the same thickness, following its major axis. These sections were fixed and embedded in a single paraffin block for histopathologic analysis and immunohistochemistry or in optimal cutting temperature compound for immunofluorescence. Multiple 8 μm sections were cut and stained as previously described (27, 28). For immunohistochemistry, we used rabbit polyclonal anti-LYVE1 1:200 (Abcam, Cambridge, MA, USA) and rat monoclonal anti-CD31 1:50 (BD Bioscience), and for immunofluorescence, we used rabbit polyclonal anti-LYVE1 1:250 (Abcam).

Two-photon microscopy

Time-lapse and 3-dimensional acquisitions were performed using an Olympus IX81 microscope equipped with FluoView 1000 scanning head (Olympus America Inc., Center Valley, PA, USA) that was customized for 2-photon microscopy as previously described (29).

Flow cytometry (FACS) analysis

Cells were collected and stained using anti-CXCR4 (clone 44717) or isotype control (clone 133303) antibody conjugated to phycoerythrin for 1 h at 4°C (R&D Systems) in PBS 2% fetal bovine serum. After staining, cells were washed, and 100,000 events were acquired in FACSCalibur using CellQuest software (BD Bioscience) and analyzed by FlowJo software version 9.4.11 (TreeStar, Ashland, OR, USA).

Population doubling assay

To calculate the population doubling, 30,000 cells were seeded and kept in culture in DMEM 10% fetal bovine serum for 24, 48, or 72 h. The number of cells per well was counted at each of the end points and applied to the formula x = [log₁₀(NH/N1)]/log₁₀(2)] (30), where N1 is the inoculum cell number (30,000) and NH the number of collected cells. To yield the cumulated doublings, the population doubling for each passage was calculated and then added to the population doubling of the previous day (28).

Statistical analysis

Data analysis was done using GraphPad Prism version 5.00 for Windows (GraphPad Software, La Jolla, CA, USA). One-way ANOVA followed by Newman-Keuls multiple comparison tests was used, and P values of <0.05 were considered statistically significant.

RESULTS

CXCL12 induces HeLa cell migration and spontaneous metastasis through CXCR4

To investigate the underlying mechanisms by which CXCL12 induces tumor cell migration and metastasis, we took advantage of HeLa cells that express CXCR4 endogenously. We first performed FACS analysis to confirm CXCR4 expression, and as seen in Fig. 1A, HeLa cells express high levels of CXCR4. Moreover, when challenged in a chemotaxis assay, HeLa cells migration toward CXCL12 was remarkable, while control cells without chemoattractant barely moved. Of note, treatment of HeLa cells with pertussis toxin (PTX) completely prevented CXCL12-mediated migration without affecting EGF-induced migration, demonstrating that HeLa cell migration to CXCL12 is mediated by coupling of the receptor to Gαi (Fig. 1B). To further explore the role of CXCR4 in HeLa cell migration and metastasis, we generated stable CXCR4 knockdown cells using lentivirus infection. Upon puromycin selection, a pool of shCXCR4⁺ cells was established and validated by FACS for decreased expression of the chemokine receptor (Fig. 1C). As expected, cells with decreased CXCR4 expression had defects on CXCL12-mediated chemotaxis (Fig. 1D). Although it has been reported that CXCL12/CXCR4 are critical for survival and cell proliferation in connective tissue sarcoma, gliomas, and kidney and prostate tumors (31–34), HeLa cells do not require CXCR4 signaling for cell proliferation in vitro (Fig. 1E). Given the high levels of CXCR4 expression in HeLa cells and their propensity to spontaneous metastasis in an oral xenograft model (29), we hypothesized that their dissemination to locoregional lymph nodes could be a CXCR4-mediated process. To test this, we first confirmed the ability of HeLa cells to spontaneous metastasize by injecting mice in the tongue and screening them for metastasis by performing full necropsies. The primary tumors were large and aggressive, and the majority of the animals had at least 1 invaded lymph node. The secondary site in all cases was cervical lymph nodes, and the incidence of metastasis was between 70% and 100% (Fig. 1F). Histologic analysis showed poor differentiated primary tumors and the presence of tumor cells in cervical lymph nodes (Fig. 1G). Because of the easy accessibility permitted by this tongue model system, live images from primary tumor and lymph node metastasis were acquired using a 2-photon microscope. In its upper panels, Fig. 1H shows snapshots from Supplemental Movies 1 and 2, in which H2BGFP (histone and GFP fusion protein) tumor cells can be visualized in the tongue and within a cervical lymph node, respectively. Immunohistochemical analysis revealed that HeLa xenografts are highly positive for the lymphatic marker LYVE1, suggesting that those tumors have a complex lymphatic network. Moreover, immunofluorescence using H2BGFP tumors captured the presence of tumor cells inside LYVE1⁺ vessels within the primary tumor. Using shCXCR4⁺ and shControl⁺ cells, we induced tongue xenografts to assess the contribution of CXCR4 in this experimental spontaneous metastasis model. Corroborating the in vitro data, shCXCR4⁺ tumor cells showed no proliferation defect in vivo, and primary tumors from shControl and shCXCR4 cells were similar regarding their growth and tumor size at the experimental end point (Fig. 1I, J). Although effects in the primary tumor were absent, remarkably, CXCR4-deficient HeLa cells were not capable of establishing lymph node metastasis (Fig. 1K). These data suggest that spontaneous lymph node metastasis of HeLa tongue xenografts rely on CXCR4 to metastasize to cervical lymph nodes; therefore, it may represent a novel, short-term, and suitable CXCR4-dependent spontaneous metastasis model.

Figure 1.

CXCL12 induces HeLa cell migration and spontaneous metastasis through CXCR4. A) FACS data showing membrane expression of CXCR4 in HeLa cells. B) Chemotaxis assay using Boyden chamber after 16 h exposure to vehicle, CXCL12 (50 ng/ml), or EGF (50 ng/ml) showing that CXCL12-mediated migration is PTX sensitive. C) Stable knockdown of CXCR4 was achieved using shRNA, and FACS analysis was performed to confirm decreased expression of CXCR4 in HeLa cells stably expressing shCXCR4 compared to shControl. D) Chemotaxis assay measured after 16 h exposure to CXCL12 (50 ng/ml) in Boyden chamber, showing decreased cell migration to CXCL12 in shCXCR4 cells. E) Cell proliferation was assessed by population doubling, and it was not changed by CXCR4 knockdown. F–H) HeLa cells were injected in the tongue of SCID-NOD mice, and full necropsy was performed at the experimental end point. F) Graph represents the percentage of invaded lymph nodes per mouse. G) Histology from primary tumor and lymph nodes showing in the upper panel a poorly differentiated primary tumor and in the lower panels a normal and a metastatic lymph node. H) Live animal imaging of H2B-GFP expressing primary tumor (upper left) and lymph node metastasis (upper right) using 2-photon microscopy. LYVE1 immunohistochemistry showing high density of lymphatic vessel in the primary tumor (lower left). LYVE1 immunofluorescence analysis revealed H2B-GFP⁺ tumor cells inside the LYVE1⁺ lymphatic vessel. I) CXCR4 knockdown did not affected primary tumor growth in tumor xenografts. J) Histology from 4 representative primary tumors (dashed lines represent tumor limits). K) Percentage of invaded cervical lymph nodes per mouse, showing that CXCR4 is required for lymph node metastasis. In each case, ANOVA was performed on 2 or 3 independent experiments in vitro and n = 10 from 2 independent experiments in vivo. Data represent mean ± sem. **P < 0.01, ***P < 0.001.

CXCR4/Gαi induces mTORC1 and mTORC2 activation and the use of mTOR inhibitors abrogate CXCL12 mediated chemotaxis.

Steady-state HeLa cells in complete media were treated with a mTOR kinase inhibitor that blocks mTORC1 and mTORC2 activity, Torin2 (35); or with mTORC1 inhibitors rapamycin and Torin2. Long-term blockade of mTOR downstream targets were achieved with both pharmacologic approaches as judged by up to 72 h decrease in Akt^S473 phosphorylation upon Torin2 treatment and S6 phosphorylation upon Torin2 and rapamycin treatment (Fig. 2A). HeLa cells were starved of serum overnight and treated with CXCL12. CXCR4 activation induced S6 and AKT^S473 phosphorylation, which was completely dependent of Gαi, as it was abolished by pretreatment of cells with PTX (Fig. 2B). As expected, CXCL12-induced activation of mTORC1 was blocked by rapamycin and Torin2; and activation of mTORC2 was blocked by Torin2, as judged by decreased pAkt^S473 (35) (Fig. 2C). In order to test whether CXCR4-induced mTOR activation was required for CXCL12 induced chemotaxis, we challenged HeLa cells in a Boyden chamber chemotaxis assay. Interestingly, blockade of mTORC1 and -2 by Torin2 decreased CXCR4-mediated migration but had no effect on EGF receptor– and LPA receptor–mediated migration (Fig. 2D). Moreover, treatment with rapamycin also decreased CXCR4-mediated migration in HeLa cells (Fig. 2), suggesting that blockade of mTORC1 is sufficient to abrogate CXCR4/Gαi-mediated migration. Similar findings were observed when analyzing the role of mTOR in CXCL12-initated migration in MCF-7 (Supplemental Fig. S1A) and T47D cells (Supplemental Fig. S1I), 2 representative breast cancer cells in which CXCR4 uses Gαi for chemotaxis (36). However, interestingly, rapamycin had no effect on CXCL12-mediated migration in SUM-159 and MDA-MB-231 (Supplemental Fig. S1J, K), 2 triple-negative breast cancer cells that have been previously reported to engage Gα13 instead of Gαi downstream of CXCR4 for cell migration (37), supporting a more specific role for mTOR downstream of Gi signaling. None of the mTOR inhibitors induced changes in the CXCR4 expression levels (Fig. 2F). These data indicate that CXCR4/Gαi activates mTOR, which is required for CXCL12-mediated migration.

Figure 2.

CXCL12 activates mTOR pathway downstream of CXCR4/Gαi, and pharmacologic inhibition of mTOR abrogates CXCR4-mediated migration. A) Rapamycin (100 nM) and Torin2 (100 nM) decreased the phosphorylation of mTOR downstream targets S6 (mTORC1) and Akt^S473 (mTORC2) in HeLa cells grown in complete media. B) CXCL12-induced mTORC1 and mTORC2 activation in serum-free conditions. S6 and Akt^S473 phosphorylation induced by CXCL12 is PTX sensitive, showing that CXCL12 activates the mTOR pathway through a CXCR4/Gαi axis. C) CXCL12-induced mTOR activation was blocked by mTOR inhibitors rapamycin (mTORC1) and Torin2 (mTORC1 and mTORC2). D) Blockade of mTORC1 and mTORC2 by Torin2 decreased CXCR4-mediated chemotaxis of HeLa cells. E) Inhibition of mTORC1 by rapamycin was sufficient to abrogate CXCR4-mediated migration in HeLa cells. F) FACS data showing membrane expression of CXCR4 in control, rapamycin-, and Torin2-treated HeLa cells. In each case, ANOVA was performed on 3 independent experiments. Data represent mean ± sem. ***P < 0.001.

mTORC1 and mTORC2 are required for migration directionality during CXCR4-mediated chemotaxis in vitro

Although rapamycin specifically blocks mTORC1, it has been reported that long-term treatment with rapamycin has also an inhibitory effect on mTORC2 (38). This indirect effect of rapamycin on mTORC2 makes it difficult to define which of the mTOR complexes are involved in the downstream pathway of CXCR4. To address this question, we used a loss-of-function approach using pools of siRNA sequences to knock down key components of each of the complexes: Raptor to affect mTORC1, and Rictor to affect mTORC2 (16, 17). Therefore, we generated cells that had nonfunctional mTORC1 showing a decrease in S6 phosphorylation, or that had nonfunctional mTORC2 with decreased Akt phosphorylation at serine 473 (Fig. 3A and Supplemental Fig. S1B, C). Similar to the pharmacologic approach, disruption of either mTOR complexes had no effect on CXCR4 expression levels (Fig. 3B) and decreased CXCR4-mediated migration without affecting EGF- or LPA-mediated migration (Fig. 3C). To better understand the involvement of mTOR complexes in CXCR4-mediated migration, we used a single-cell tracking-based strategy, which allowed us to study various aspects of cell migration during chemotaxis. Figure 3D and Supplemental Fig. S1D show dot-plot graphs representing the final position of individual cells after 24 h chemotaxis in μ-Slides. Black dots represent cells that migrated forward, and red dots represent cells that migrated backward to the chemoattractant gradient; the inner panel in each plot is the vector analysis, showed as a rose diagram. We observed that in the absence of chemoattractant, these epithelial-derived tumor cells had limited movement; however, when a gradient of chemoattractant (CXCL12) is applied, they migrate readily in the direction of the gradient (Fig. 3D and Supplemental Fig. S1D). In contrast, tumor cells with a nonfunctional mTORC1 or mTORC2 were still able to move, but they lacked directionality. Indeed, analysis of statistical parameters of μ-Slides chemotaxis revealed a minor, albeit significant, decrease in velocity (Fig. 3E and Supplemental Fig.S1E), as well as accumulated (Fig. 3F and Supplemental Fig. S1F) and Euclidean (Fig. 3G and Supplemental Fig. S1G) distances; however, the major defect observed in Raptor and Rictor knockdown cells was in migration directionality, as judged by a forward migration index similar to siControl cells in the absence of CXCL12 (Fig. 3H and Supplemental Fig. S1H). Taken together, these data show that mTOR activation downstream CXCR4/Gαi is required for chemotaxis (directional migration) but not for chemokinesis (random cell movement). Moreover, Raptor knockdown was sufficient to decrease CXCR4-mediated directional cell migration, suggesting that pharmacologic inhibition of mTORC1 could potentially be sufficient to decrease CXCR4-mediated migration/metastasis.

Figure 3.

Loss-of-function and single cell tracking approaches reveled that mTORC1 and mTORC2 are required for migration directionality during CXCR4-mediated chemotaxis. A) Western blot analysis showing Raptor and Rictor knockdown and their effect on the mTOR downstream targets S6 and Akt^S473 after transfection with control or with Raptor or Rictor siRNA (50 nM). HeLa cells with nonfunctional mTORC1 or mTORC2 upon Raptor and Rictor knockdown, respectively, were tested for CXCR4 expression and challenged in chemotaxis assays. B) FACS data showing unchanged membrane expression of CXCR4 upon Raptor or Rictor knockdown. C) Chemotaxis was measured after 16 h exposure to CXCL12 (50 ng/ml), EGF (50 ng/ml), or LPA (1 μM) in Boyden chamber. Raptor and Rictor knockdown decreased CXCL12-mediated chemotaxis. D) Dot-plot graphs representing the final position of individual cells after 24 h chemotaxis in μ-Slides as described in Materials and Methods. Black dots represent cells that migrated forward, and red dots represent cells that migrated backward to the chemoattractant gradient. Rose diagrams of each group are shown in the inner panel of each dot-plot. E–H) Statistical analysis of μ-Slides chemotaxis showing a small decrease in velocity (E), accumulated distance (F), and Euclidean distance (G) and a major defect in directionality in Raptor and Rictor knockdown cells, as judged by a forward migration index that was similar to siControl cells in the absence of CXCL12. In each case, ANOVA was performed on 3 independent experiments. Data represent mean ± sem. *P < 0.05, **P < 0.01, ***P < 0.001.

Pharmacologic blockade of mTORC1 by rapamycin decreases primary tumor growth and CXCR4-mediated lymph node metastasis, and increases animal survival

We next used our CXCR4-dependent metastasis model to test whether our in vitro observation that blocking mTORC1 is sufficient to decrease CXCR4-mediated migration were relevant in the context of metastasis. Mice bearing tumor xenografts were treated with vehicle or rapamycin (5 mg/kg) daily. As shown in Fig. 4A–D, all animals developed primary tumors; however, tumor size in the rapamycin-treated group was significantly smaller than in the vehicle-treated group. Moreover, histologic analysis of cervical lymph nodes revealed that the incidence of invaded lymph nodes among the rapamycin-treated group was dramatically decreased (Fig. 4E), suggesting that pharmacologic blockade of mTORC1 is sufficient and effective to decrease CXCR4-mediated metastasis. As a consequence of decreasing primary tumor size and metastasis, animals from the rapamycin group had a major improvement in survival, with 100% of animals alive 100 d after the tumor xenograft was induced, while all mice from the vehicle-treated group had to undergo euthanasia due to tumor size to minimize animal suffering before d 90 (Fig. 4F). These data suggest that mTORC1 blockade by rapamycin decreases primary tumor growth and CXCR4-mediated metastasis.

Figure 4.

Rapamycin decreased primary tumor growth, decreased CXCR4-mediated lymph node metastasis, and increased animal survival. A–D) HeLa xenografts were induced, and vehicle and rapamycin (5 mg/kg per d) treatment started after the primary tumor was established (2 mm³). A) Primary tumor growth curve. B) Representative images at the experimental end point. C, D) Histologic analysis of the tumor area showing a decrease in primary tumor size upon rapamycin treatment. E) Number of invaded lymph nodes per mouse at the experimental end point. F) Rapamycin treatment increased animal survival. Statistical analysis was performed by ANOVA. n = 10 from 2 independent experiment. Data represent mean ± sem. ***P < 0.001.

Stable disruption of mTORC1 by Raptor shRNA is sufficient to decrease cell proliferation and CXCR4-mediated migration in vitro

In order to study the long-term inhibition of mTORC1 and mTORC2, we generated stable Raptor and Rictor knockdown cells by infecting HeLa cells with lentivirus expressing Raptor and Rictor shRNA-targeting sequences. After selection of cell pools resistant to puromycin, we analyzed Raptor and Rictor expression by Western blot analysis to validate knockdowns, and mTORC2 and mTORC1 downstream signaling by analyzing S6 and Akt^S473 phosphorylation status. As expected, Rictor shRNA abrogated Akt^S473 phosphorylation (Fig. 5A) and Raptor shRNA decreased S6 phosphorylation (Fig. 5E), indicating that these cells had stable nonfunctional mTORC2 and mTORC1, respectively. Consistently, stable Rictor knockdown cells showed defects in CXCR4-mediated chemotaxis without affecting EGF-mediated migration (Fig. 5B) or CXCR4 expression levels (Fig. 5C). Interestingly, cells with nonfunctional mTORC2 exhibited no changes in proliferation rate in vitro (Fig. 5D), as they had similar population doubling rate compared to the pool of shControl cells. Regarding cells with stable nonfunctional mTORC1, defects on CXCR4-mediated chemotaxis were observed, corroborating the pharmacologic and transient knockdown approaches, with no effect on EGF-mediated migration (Fig. 5F) or CXCR4 membrane expression (Fig. 5G). Nonetheless, Raptor ablation significantly decreased the proliferation rate of the cells. As shown in Fig. 5H, shRaptor⁺ cells required almost twice the time to double their population compared with shControl cells. These data suggest that in addition to be sufficient to decrease CXCR4-mediated chemotaxis, targeting mTORC1 to impair tumor progression and dissemination also confers growth disadvantages to tumor cells, thus limiting the ability to establishing primary and metastatic colonies.

Figure 5.

Stable disruption of mTORC1 by Raptor shRNA was sufficient to decrease cell proliferation and CXCR4-mediated migration in vitro. A) Western blot analysis showing decreased expression of Rictor and down-regulation of mTORC2 downstream pathway upon stable knockdown of Rictor using shRNA. B) Boyden chamber chemotaxis assay showing decreased migration of shRictor cells to CXCL12 (50 ng/ml) and unchanged migration to EGF (50 ng/ml). C) FACS data showing unaltered CXCR4 membrane expression upon Rictor stable knockdown. D) Down-regulation of Rictor and mTORC2 downstream pathway had no effect on cell proliferation in vitro. E) Decreased expression of Raptor and down-regulation of mTORC1 downstream pathway demonstrated by Western blot analysis in cells expressing Raptor shRNAs. F) Stable Raptor knockdown decreased CXCL12-mediated migration without affecting the ability of cells to migrate to EGF. G) Membrane expression of CXCR4 assessed by FACS was not changed upon stable Raptor knockdown. H) Raptor stable knockdown was sufficient to decrease cell proliferation in vitro. In each case, ANOVA was performed on 2 to 3 independent experiments. Data represent mean ± sem. **P < 0.01.

Stable disruption of mTORC1 by Raptor shRNA is sufficient to decrease tumor burden, angiogenesis, lymphangiogenesis, and lymph node metastasis

To rule out the possibility that an indirect effect of rapamycin long-term administration in mTORC2 could contribute to the decrease in tumor burden and metastasis, we took advantage of our stable Raptor and Rictor knockdown cells. We induced tumor xenografts using the CXCR4-mediated spontaneous metastasis model by injecting shControl, shRictor, or shRaptor cells and monitored animals weekly. At the experimental end point, primary tumor and cervical lymph nodes were collected for histologic analysis. Surprisingly, the disruption of mTORC2 by Rictor knockdown did not affect tumor progression in vivo. As seen in Fig. 6A, B, shControl and shRictor cells generated primary tumors with similar size. Moreover, despite affecting CXCR4-mediated migration in vitro (Fig. 6B), Rictor knockdown did not prevent CXCR4-mediated spontaneous metastasis in vivo (Fig. 6C). Interestingly, cells with nonfunctional mTORC2 were still able to induce changes in the tumor microenviroment and induce angiogenesis and lymphangiogenesis, as shControl and shRictor tumors showed similar levels of the endothelial marker CD31 and the lymphatic marker LYVE1 (Fig. 6D–F). A completely different scenario was observed upon mTORC1 disruption in the tumor cells. Primary tumors from cells with nonfunctional mTORC1 due to Raptor knockdown were significantly smaller compared to the shControl group (Fig. 6G, H). Interestingly, disruption of mTORC1 function had a major effect on CXCR4-mediated metastasis, as Raptor knockdown tumors showed a significant decrease in lymph node metastasis, with most of the animals (16 of 20) free of metastases at the end of the observation period (Fig. 6I). In addition, a dramatic impact of a dysfunctional mTORC1 within the tumor cells was observed in the tumor microenvironment. Primary tumors from the shRaptor group had a significant decrease in the number of CD31⁺ endothelial cells (Fig. 6J, K) as well as LYVE1⁺ lymphatic vessels (Fig. 6J, L). Taken together, our data show that mTOR is activated downstream CXCR4/Gαi axis and that it is essential to CXCR4-mediated directional migration and metastasis. Our data also suggest that although mTORC1 and mTORC2 are activated upon CXCR4 stimulation, mTORC1 seems to play a more important role in tumor progression in vivo (Fig. 6M). While mTORC2 blockade affected chemotaxis, disruption of mTORC1 significantly decreased CXCR4-mediated migration and metastasis. In addition, mTORC1 disruption using pharmacologic or loss of function approaches also impacted cell proliferation, which is required for primary tumor growth and the establishment of secondary colonies. The ability of tumor cells to modify their microenviroment was also compromised by Raptor ablation, thus suppressing angiogenesis and lymphangiogenesis, which limits the access of tumor cells to circulation.

Figure 6.

Stable disruption of mTORC1 by Raptor shRNA was sufficient to decrease tumor burden, angiogenesis, lymphangiogenesis, and lymph node metastasis. A) Histology of 4 representative primary tumors from shControls and shRictor (dashed lines represents the tumor limits). B) Tumor area assessed at the experimental end point. C) Percentage of invaded lymph nodes per mouse showing that disruption of mTORC2 by Rictor knockdown did not affect CXCR4-dependent lymph node metastasis. Immunohistochemistry data revealed similar expression of CD31⁺ blood vessels (D, E) and LYVE1⁺ lymphatic vessels (D, F) within the primary tumor. G) Histology of 4 representative primary tumors from shControls and shRaptor (dashed lines represent tumor limits). H) Tumor area assessed at the experimental end point. I) Percentage of invaded lymph nodes per mouse showing that disruption of mTORC1 by Raptor knockdown decreased CXCR4-dependent lymph node metastasis. Immunohistochemical data showing decreased expression of CD31⁺ blood vessels (J, K) and LYVE1⁺ lymphatic vessels (J, L) within the primary tumor. M) Schematic representation of CXCR4/Gαi/mTOR pathway in CXCL12-mediated migration and metastasis. In each case, ANOVA was performed. n = 10 from 2 independent experiments. Data represent mean ± sem. *P < 0.05, **P < 0.01, ***P < 0.001.

DISCUSSION

In the present study, we used HeLa cells that express high levels of CXCR4 endogenously as an experimental model system to investigate the downstream signaling mechanism involved in CXCL12-mediated migration and metastasis. Our data show that binding of CXCL12 to CXCR4 induces mTOR pathway activation in these epithelial-derived tumor cells. Moreover, mTOR activation downstream of CXCR4 is strictly required for CXCL12-mediated migration and is completely dependent of G proteins from the Gαi family. Similar findings were observed in human breast cancer cells exhibiting Gαi-dependent migration and signaling downstream from CXCR4. Because mTOR exists in 2 different complexes, mTORC1 and mTORC2, and because CXCL12 induces phosphorylation of downstream targets of both complexes, we used pharmacologic and loss-of-function approaches to address the contribution of mTORC1 and mTORC2 in CXCR4-mediated directional migration. We demonstrated that disruption of either mTORC1 or mTORC2 has a profound effect on directionality during chemotaxis. To study the CXCR4/mTOR axis in vivo, we developed a model of CXCR4-dependent spontaneous metastasis in which 70–100% of the animals have lymph node metastasis at 1 month after cell inoculation. Surprisingly, mTORC1 blockade was sufficient to decrease CXCR4-mediated metastasis as well as tumor cell proliferation, angiogenesis, and lymphangiogenesis, whereas mTORC2 impairment had no demonstrable effect in tumor growth or dissemination in vivo.

The PI3K/Akt/mTOR pathway is a major player in many tumor types. It regulates cell metabolism, growth, survival, and proliferation as well as cytoskeleton organization (9). Although many functions that are regulated by mTOR are easily associated with tumor progression, the precise mechanisms by which the mTOR pathway mediates tumor development and dissemination are not fully elucidated. Most of the effects of mTOR in cell migration have been attributed to mTORC2 (17, 18, 39–42) due to its ability to activate small GTPases, such as Rac, and to control actin cytoskeleton organization (18). Recently, Liu et al. (40) reported that in neutrophils mTORC2 is also critical for cell migration through a mechanism that is dependent on cAMP and RhoA activation. Concerning mTORC1, little is known regarding its role in cell migration and metastasis. In particular, as a result of its important function in protein synthesis (43, 44), mTORC1 has been described to influence cell migration mainly by regulating expression of key components of the migratory machinery, such as small GTPases and chemokine receptors (26, 45). By use of loss-of-function strategies, our study shows that both mTOR complexes are important for cell migration downstream of CXCR4. Of note, no changes in CXCR4 or in small GTPases RhoA, Rac, or CDC42 expression were observed after rapamycin and Torin2 treatment or Raptor/Rictor knockdown (data not shown). Moreover, by use of a single-cell tracking strategy, we showed that blockade of mTORC1 or mTORC2 decreases the ability of cells to migrate directionally toward CXCL12, but it does not abrogate the ability of the cells to move, indicating that mTORC1 and mTORC2 are essential for chemotaxis but not for chemokinesis. The underlying mechanism of such an effect is not yet known, but it is tempting to speculate that it might be related to the ability of the cells to polarize and sense the chemoattractant gradient, as cells lose directionality and not motility upon mTOR inhibition. Supporting this hypothesis, the connection between mTOR pathway and cell polarity has already been reported, as an mTOR-regulating kinase, LKB1, is the human ortholog of Drosophila melanogaster’s PAR-4 (46, 47), and it has been shown that once activated, LKB1 can fully polarize a single cell in the absence of cell–cell contact (48). However, similar to 20% of all cervical carcinomas (49), HeLa cells do not express LKB1, suggesting the existence of a yet-to-be identified mechanism linking mTOR and cell polarity, which warrants further investigation.

Although tumor cell migration is a key and rate-limiting event in metastasis, it does not fully recapitulate the complex and multistep biologic process of metastasis. Therefore, we aimed to study the role of CXCR4/Gαi/mTOR axis in vivo. Unfortunately, the limited number of spontaneous metastasis models, together with the long-term duration and low incidence of advanced disease in most of the current protocols, represent substantial challenges in the study of metastasis (50). To overcome these difficulties, experimental approaches such as injection of tumor cells in the heart or in the tail vein of experimental animals have been established. Even though these models have been largely used in cancer research, injecting cells directly into the bloodstream does not fully recapitulate the complete process of tumor spread. In fact, how tumor cells gain access to blood and lymphatic vessels represents a key regulatory step for tumor metastasis (50).

Considering that HeLa cells express high levels of CXCR4, and considering our observation that they spontaneously metastasize to cervical lymph nodes after injection in the tongue (29), we hypothesized that CXCR4 could be the driver of HeLa cells spontaneous metastasis. Of interest, the rationale behind the tongue xenograft model relies in 3 important elements: 1) the tongue is a highly blood and lymphatic vascularized organ, 2) the neck area is fairly close to the primary tumor site and holds about one third of the body’s lymph nodes (51), and 3) carcinoma from uterus cervix metastasize to lymph nodes similarly to most carcinomas, including those from the head and neck, suggesting that the mechanism underlying lymph node metastasis could be shared by different carcinomas (52). Combining this model with a loss-of-function approach, we observed that HeLa require CXCR4 to metastasize. Indeed, shCXCR4⁺ cells showed severe defects to migrate to CXCL12 in vitro and were not capable of invading lymph nodes in vivo. Of note, the requirement of CXCR4 for metastasis, together with the complete absence of distant metastasis, especially considering lung and liver, which are common target organs when tumor cells are injected directly into the bloodstream, strongly suggest that this is an active and chemokine-regulated process, which requires tumor cells to gain access and follow a gradient of chemokine to reach secondary organs. Furthermore, as a result of the easy access, live imaging from primary tongue tumor and cervical lymph node metastasis are possible using 2-photon microscope, which allows multiple and sequential real-time images from primary and metastatic sites. Taken together, we can expect that this CXCR4-dependent spontaneous metastasis model, which results in a large fraction of mice exhibiting lymph node invasion, may provide a suitable experimental model for the future evaluation of antimetastatic agents. Certainly, careful standardization will be required for the future development of similar spontaneous metastasis models for each human cancer cell line of interest. In this context, we took advantage of the CXCR4-dependent spontaneous metastasis model to investigate whether mTOR activation was required to tumor spontaneous metastasis in vivo, and more importantly, what the contribution of mTORC1 and mTORC2 are to the process. Indeed, our data show that although both complexes are essential for CXCR4-mediated directional migration in vitro, mTORC1 may play a more important role in vivo. In this regard, disruption of mTORC1 function by either rapamycin or by Raptor knockdown decreased several features of aggressiveness, such as cell proliferation and chemotaxis in vitro, and tumor burden, vascularization, and ultimately lymph node metastasis in vivo. Although it is not possible to dissociate at this stage the effects of mTORC1 disruption on the primary tumor growth from tumor dissemination, it is important to consider that this may resemble the clinical scenario. In fact, standard cancer therapies successfully decrease primary tumor size, but often they have a more limited impact on treating or preventing metastatic disease, and consequently, metastasis remains the cause of more than 90% of all cancer-related deaths (50).

More studies are needed to understand the mechanisms by which mTORC1 contributes to CXCR4-mediated migration and metastasis; however, one possibility is that in addition to the loss of directionality in cell migration, tumors with nonfunctional mTORC1 are also inefficient in inducing changes in their tumor microenvironment and in eliciting autocrine/paracrine cytokine-initiated feedback loops that stimulate tumor progression and dissemination (33, 53, 54). Indeed, the remarkable effect of mTORC1 blockade in the primary tumor, and possibly cytokine-induced feedback loops, could help to explain the fact that only mTORC1 and not mTORC2 disruption have major consequences in CXCR4-mediated metastasis. Consistently, we have observed a significant decrease in angiogenesis and lymphangiogenesis in tumors with nonfunctional mTORC1 but no impact in endothelial or lymphatic vascular network in response to Rictor knockdown in cancer cells.

Although CXCR4 itself does not currently represent a drug-directed target for cancer therapy, identifying downstream targets involved in CXCR4-mediated migration and metastasis may afford a new window of opportunity for pharmacologic intervention for aggressive tumors. Taken together, our data suggest that mTORC1 blockade is sufficient to decrease primary tumor growth and CXCR4/Gαi/mTORC1-dependent migration and metastasis in vivo. Therefore, drugs that block mTORC1, such as the U.S. Food and Drug Administration–approved rapamycin, RAD001, and Torisel, then could be explored as an option for highly aggressive and metastatic tumors that rely on CXCR4 to spread. Of note, rapamycin has not been reported to promote expulsion of bone marrow stem cells from their niche (24, 25), a side effect observed when targeting CXCR4 itself (5–8). Indeed, if these findings can be translated into the clinic, such an approach would provide an opportunity to target both primary and metastatic tumors, thereby potentially improving patient outcome in the case of multiple highly aggressive malignancies that use CXCR4 for their metastatic spread.

Glossary

CXCL12

C-X-C motif chemokine 12

CXCR4

C-X-C motif chemokine receptor type 4

EGF

epidermal growth factor

FACS

flow cytometry

H2BGFP

histone and GFP fusion protein

LPA

lysophosphatidic acid

mTOR

mechanistic target of rapamycin

mTORC

mTOR complex

PTX

pertussis toxin

SDF-1

stroma cell–derived factor 1

shRNA

short hairpin RNA

siRNA

small interfering RNA