CD9, a key actor in the dissemination of lymphoblastic leukemia, modulating CXCR4-mediated migration via RAC1 signaling

Acute lymphoblastic leukemia (ALL) is the most common childhood cancer, accounting for 25% of all pediatric cancers and 75% of all cases of childhood malignant blood disease. In most cases (85%), ALL affects the B lineage, giving rise to a heterogeneous group of diseases displaying different genetic abnormalities. Five-year event-free survival rates have improved,¹  but about 20% of children suffer from ALL relapse in the bone marrow (BM) or at extramedullary sites, such as the testes or ovaries, these sites being particularly frequently affected in cases of late relapse.^2-4  It has been suggested that late relapses originate from a preleukemic clone that persists after treatment.^5,6  This clone is present at the extramedullary site, where it may undergo new genetic alterations, generating a “new” leukemic clone.

We have shown that several lymphoblastic leukemias can be distinguished on the basis of their CD9 expression.^7,8  CD9 is a protein belonging to the tetraspanin superfamily. Tetraspanins are small proteins with a highly conserved tertiary structure consisting of 4 transmembrane domains.⁹  They are involved in many biological pathways, due to their ability to associate with a wide variety of other transmembrane proteins, most of which are tissue-specific.¹⁰  Together, these proteins form a structural platform, referred to as the tetraspanin web, and play an important role in signaling pathways.^10,11  CD9 has been implicated in various physiological processes, such as motility, adhesion, and fertilization.¹²  It also plays a critical role in several pathological processes, including tumor progression and solid tumor metastasis.^13,14  Nevertheless, its modes of action remain unclear and are highly dependent on cell type and molecular partner.^10-12,15,16  In hematopoietic cells, CD9 expression depends on the stage of differentiation and the location of the cell.¹⁶  In leukemia, CD9 enhances metastatic medullary invasion and decreases the survival of mice receiving an injection of pre-B leukemic cells.^17,18  CD9 also promotes the chemotactic migration induced by C-X-C motif chemokine ligand 12 (CXCL12) in vitro and the homing to the BM in vivo of human CD34⁺ hematopoietic stem and progenitor cells.¹⁹  CXCL12 or stromal cell-derived factor-1α is a chemokine from the CXC family. CXCL12 was initially found to affect the growth of pro-B cells,²⁰  but has since been shown to be a potent chemoattractant of lymphocytes.²¹  Its receptor, C-X-C motif chemokine receptor 4 (CXCR4), is highly expressed on B-cell progenitors and acute leukemia cells and plays a major role in their migration.^22,23  The CXCL12/CXCR4 axis is thus crucial in hematopoiesis, supporting the homing and engraftment of normal hematopoietic stem cells (HSCs) and leukemic stem cells (LSCs) in BM.^23-25  Disruption of the interaction between CXCR4 and CXCL12 with a selective chemical antagonist of CXCR4 (AMD3100 or plerixafor) promotes the massive and rapid mobilization of HSCs into the peripheral blood of healthy mice and human volunteers.²⁶ 

The purpose of this study was to investigate the effect of CD9 on the CXCL12-induced motility and engraftment of pre-B leukemic lymphocytes. We identified CD9 as a regulator of pre-B lymphoblast adhesion and a new actor in CXCR4-mediated migration and homing, promoting RAC1 activation in response to CXCL12. We confirmed the activation of the CD9-RAC1 signaling pathway in samples from patients. We demonstrated that CD9 enhanced the dissemination of pre-B leukemic cells to the testis, one of the preferential sites for late relapses of ALL. We therefore hypothesized that CD9, through its effects on migration and homing, may be a key player in late relapses of B acute lymphoblastic leukemia (B-ALL).

The REH and NALM6 pre-B acute lymphoblastic leukemic cell lines were cultured in RPMI 1640 (Life Technologies) supplemented with 10% decomplemented fetal calf serum (FCS) and 1% antibiotics.

BM leukemia cells were collected at diagnosis, after informed consent had been obtained, in accordance with the declaration of Helsinki. The protocol was approved by the ethics committee of Rennes Hospital (Rennes, France). Mononuclear cells were isolated from BM by successive centrifugations through lymphocytes separation medium (Eurobio).

REH and NALM6 cells were infected with lentiviral vectors bearing MISSION pLKO.1 shRNA-puro constructs targeting human CD9 (#TRCN000000296958, #TRCN00000057470) and RAC1 (#TRCN 0000004870; #TRCN 0000004873) or with a nontargeted short hairpin RNA (shRNA) from Sigma-Aldrich. Transduced REH and NALM6 cells were maintained in medium containing 0.5 and 0.25 μg/mL puromycin (Invivogen), respectively.

Cells were labeled with anti-CD9–fluorescein isothiocyanate (ALB6; Beckman Coulter), anti-CD9-allophycocyanin (M-L13; Becton Dickinson), anti-CXCR4-allophycocyanin (12G5; Becton Dickinson), anti-CD10-phycoerythrin cyanine 7 tandem (HI10a; Becton Dickinson), and anti-CD19–fluorescein isothiocyanate (J3-119; Beckman Coulter) antibodies. Cell immunofluorescence was measured using a Beckman Coulter FACS FC500 and analyzed with the CXP analysis software package.

We incubated 2 × 10⁶ cells with 1 μg/mL anti-CD9 antibody (MEM-61; Abcam), 1 μg/mL immunoglobulin G (IgG) isotype control (IgG1; Sigma-Aldrich), 20 μM plerixafor (Genzyme), a specific CXCR4 inhibitor, 100 nM CCX771, a specific CXCR7 inhibitor, or 100 nM of the control CCX704 (ChemoCentryx), in the absence of serum, for 30 minutes at 4°C, or with 10 or 25 μM NSC23766 (Santa Cruz Biotechnology), a specific RAC inhibitor, for 3 hours at 37°C. Cells were resuspended in RPMI 1640 supplemented with 1% bovine serum albumin and loaded into the upper chamber of a migration system equipped with 8-μm Transwell microporous polycarbonate membranes (Corning-Costar). We added 0.6 mL of CXCL12 (100 ng/mL; Abcam) or testis tissue-conditioned medium (TCM) to the lower chamber. After 5 hours of incubation at 37°C, the cells that had migrated were subjected to trypan blue exclusion and counted. The migration rate was determined as the percentage of cells from the input that had migrated.

We coated 96-well plates with 0.1 μg of superfibronectin diluted in phosphate-buffered saline (Sigma-Aldrich), by incubation overnight at 4°C. We then incubated 5 × 10⁵ cells with 1 μg/mL blocking antibody against CD9 or IgG isotype control, in the absence of serum, for 30 minutes at 4°C. The cells were resuspended in medium supplemented with 5% FCS and allowed to adhere to the fibronectin for 90 minutes at 37°C. Nonadherent cells were removed by washing and adherent cells were quantified by colorimetric analysis with the CellTiter 96 AQ One Solution Cell Proliferation Assay (Promega).

Cells were stimulated by incubation with CXCL12 (200 ng/mL) for 20 minutes at 37°C and washed in glycine-HCl buffer (50 mM glycine; 150 mM HCl). CXCR4 expression was analyzed by flow cytometry.

Cytospin preparations of cells were fixed in 4% paraformaldehyde for 20 minutes. Preparations were blocked for 1 hour with 5% FCS, supplemented with 0.2% saponin (Sigma-Aldrich) if permeabilization was required. Slides were incubated overnight at 4°C with a mouse anti-CD9 antibody (1:50; ALB6; Beckman Coulter) and a rabbit anti-CXCR4 antibody (1:50; polyclonal; Abcam) or a rabbit anti-CXCR7 antibody (1:500; polyclonal; Abcam). They were then incubated for 1 hour at room temperature with anti-rabbit DyLight594 antibody or anti-mouse DyLight488 secondary antibodies (1:500; Jackson ImmunoResearch Laboratories). Actin filaments were decorated by incubation for 1 hour at room temperature with rhodamine-phalloidin (1:100; Interchim).

Cells were stimulated with CXCL12 (100 ng/mL) for 30 seconds and lysed in 100 μL of lysis buffer (5× Mg²⁺ lysis/wash buffer [Millipore], 1× protease inhibitor cocktail [Roche], 1 mM sodium orthovanadate, 40 mM β-glycerophosphate, 30 mM sodium fluoride, and 10% glycerol). RAC1-GTP was immunoprecipitated with RAC/CDC42-PAK1 beads (Millipore), used according to the manufacturer’s instructions. Immunoprecipitates were analyzed by western blotting.

Cell lysates were separated by electrophoresis in 15% polyacrylamide gels under standard conditions. The separated proteins were blotted onto Hybond-C Extra nitrocellulose membranes (GE Healthcare). Immunoblotting was performed with a 1:1000 dilution of mouse anti-RAC1 or rabbit anti-CXCR4 antibody.

Permeant peptides labeled at the N terminus with tetramethylrhodamine (TAMRA) and bearing the C-terminal sequence of CD9 or a scrambled sequence were purchased from Lifetein: YGRKKRRQRRR-CCAIRRNREMV (CD9); YGRKKRRQRRR-NRACIRERVCM (scrambled).

NOD/scid IL2 Rg^null mice (Charles River Laboratories) were maintained in the ARCHE Animal Center (UMS CNRS 3480 Biosit, Rennes, France) at Rennes 1 University. Four-week-old mice received 2 intraperitoneal injections of 20 μg/g busulfan (Busilvex; Pierre Fabre) on 2 days. They were then allowed to rest for 2 days before the IV injection of REH cells (1 × 10⁵, 10 × 10⁶ cells).

Survival analysis was performed on the mice into which 1 × 10⁵ cells were injected. The mice into which 10 × 10⁶ cells were injected were killed 4 weeks after the injection, and single-cell suspensions were prepared from their BM and testes. Pieces of testis tissue were dissociated in 0.25% trypsin (Life Technologies) supplemented with collagenase type 1 (Sigma-Aldrich) for 25 minutes at 37°C. Red blood cells of BM aspirates from mouse femurs were lysed in ammonium-chloride-potassium lysis buffer (155 mM NH₄Cl, 10 mM KHCO₃, 500 mM EDTA). Human cells were detected by flow cytometry. Dead cells were excluded from the analysis on the basis of their size and granularity characteristics.

These experiments were approved by the ethics committee for animal experimentation of the French Ministry for Higher Education and Scientific Research.

Statistical significance was analyzed using nonparametric tests (Wilcoxon or Mann-Whitney), with values of P < .05 considered significant. For survival studies, Kaplan-Meier curves were generated with Prism 6 software and analyzed with Mantel-Cox tests.

CD9 has been reported to regulate the adhesion of several pre-B cell lines to fibronectin, by modulating β1 integrin avidity.^27,28  For confirmation of these observations in the context of B-ALL, we performed adhesion experiments with the REH cell line, a CD9-positive pre-B-ALL cell line established from an ETV6/RUNX1-positive late relapse. We generated 2 stable cell lines depleted of CD9 protein with 2 different shRNAs targeting CD9 (Figure 1A). We found that higher levels of CD9 protein resulted in the adhesion of larger numbers of cells to the fibronectin coat (Figure 1B). We assessed the role of CD9 in the migration of B-ALL cells, using the CXCL12 chemokine as a chemoattractant. REH cells express high levels of CXCR4 and this expression was not modified by transduction with lentiviruses (Figure 1A). Passive migration was initially excluded through the use of CXCL12-depleted medium (average migration rate: 3% ± 1.1%). CD9 enhanced chemotactic migration in vitro (Figure 1C).

We also used scanning electron microscopy to explore whether CD9 expression induced morphologic changes. CD9 downregulation resulted in a significantly smaller number of villi, which were also significantly thinner on REH CD9^low cells than on control cells (Figure 1D).

Finally, we investigated whether the phenotype and mobility properties induced by CD9 downregulation affected engraftment capacity in vivo. We performed xenotransplantation in NSG mice. We first checked that proliferation was not affected by CD9 expression (supplemental Figure 1A, see supplemental Data available at the Blood Web site). We then assessed BM invasion after transplantation, at death. BM invasion rates of 97% were recorded at the time of death, consistent with death from leukemia (supplemental Figure 2). We carried out an analysis of survival curves and demonstrated that the injection of REH cells in which CD9 was downregulated was associated with significantly longer survival (Figure 1E).

We investigated the contribution of CD9 and CXCR4 to cell migration in response to CXCL12. The inhibition of CXCR4 signaling with plerixafor abolished cell migration. The anti-CD9 antibody decreased migration by almost 47% (average migration rate: 23.51% ± 10.57%), suggesting a possible role of CD9 in this major migration pathway (Figure 2A). We also explored the contribution of a second CXCL12 receptor, CXCR7, the function of which in hematopoietic cell migration remains a matter of debate. We added a specific synthetic inhibitor of CXCR7 (CCX771) to the experimental system described in “Materials and methods.” No effect on REH migration was detected (Figure 2B). We used confocal imaging to localize CD9, CXCR4, and CXCR7 on the cell membrane. Cells were stabilized by cytospin centrifugation at low speed, to conserve their morphology and membrane structures. A merged image of the CD9 and CXCR4 signals showed tight colocalization, confirming that the proteins were located close together on the membrane unlike CD9 and CXCR7 (Figure 2C). The same results have been obtained with NALM6 cells, another pre-B cell line (supplemental Figure 3). We investigated the possible effect on CXCR4 trafficking of the presence of CD9 on the cell membrane, by assessing CXCR4 internalization after stimulation with CXCL12 (Figure 2D). The levels of CXCR4 expression after stimulation were similar in REH CD9^high, CD9^medium, and CD9^low cells, suggesting that CD9 does not affect CXCR4 signaling through enhanced internalization.

We used confocal microscopy to explore the distributions of CD9 and actin. Following stimulation with CXCL12, REH CD9^high cells were found to have long cytoplasmic extensions displaying colocalized CD9 and actin signals (Figure 3A). These structures were numerous on REH CD9^high cells, but were clearly less abundant on REH CD9^low cells (Figure 3B). The extensions of REH CD9^high cells were also significantly longer, increasing the width of the cell by 77% ± 36% rather than by the 21% ± 36% observed for REH CD9^low cells (Figure 3C). REH CD9^high cell extensions were also essentially unidirectional, with 88% orientated in the same direction (Figure 3D). By contrast, the extensions of REH CD9^low cells were more uniformly organized around the cell. In parallel, we measured a significantly larger increase in F-actin content following CXCL12 stimulation in CD9^high than in CD9^low cells (supplemental Figure 4). These data suggest that CD9 promotes the formation of cytoplasmic extensions and the rearrangement of actin in response to CXCL12.

RAC1 is a Rho-GTPase protein known to cycle between an inactive state (guanosine diphosphate-bound) and an active state (guanosine triphosphate [GTP]-bound). RAC1 is activated by CXCL12 in pre-B cells. This activation enhances the formation of actin-rich lamellipodia in diverse cellular functions, including migration.²⁹ 

We investigated the effect of CD9 depletion on RAC1 activation. CXCL12 stimulation increased RAC1-GTP protein levels by a factor of >2.5 ± 0.2 in REH CD9^high cells (Figure 4A). No such response was observed in REH CD9^medium and CD9^low cells, suggesting that the modulation of CD9 expression affects RAC1 activation.

This result was confirmed in blasts extracted from fresh BM collected from B-ALL patients on diagnosis (Figure 4B). Linear regression analysis revealed a strong correlation between CD9 protein levels (mean fluorescence intensity [MFI] for CD9) and RAC1-GTP:RAC1 total ratio (r² = 0.87, Spearman r = 0.76 and P = .0377).

We further assessed the effect of RAC1 activation on the chemotactic migration of REH cells. REH CD9^high cells were incubated with various concentrations of NSC23766 before the migration assay. NSC23766 prevents RAC1 activation and chemotactic migration in vitro, in a dose-dependent manner (Figure 4C). Moreover, the chemotactic migration of REH CD9^high cells was more strongly altered by NSC23766 treatment than that of REH CD9^medium and CD9^low cells, resulting in higher rates of migration for these cells with low levels of CD9 (Figure 4D). This finding supports the hypothesis that CD9 is involved in RAC1 activation and affects CXCL12-induced migration.

Finally, we explored the impact of RAC1 downregulation on mouse survival. We used 2 different shRNAs to generate 2 cell lines displaying a stable depletion of RAC1 protein (Figure 4E). CXCR4 expression was not affected by transduction or cell proliferation in vitro following RAC1 depletion (supplemental Figure 1B). Finally, xenograft studies in NSG mice showed that survival was significantly longer for mice receiving injections of cells depleted of RAC1 than for mice receiving control cells (Figure 4F). Thus, RAC1 depletion results in phenotypes identical to those observed following CD9 disruption.

We investigated the mechanism by which CD9 protein affects RAC1 activation, using a permeant TAMRA-coupled peptide containing either the C-terminal amino-acid sequence of CD9 or a scrambled sequence. Cells were incubated with each of these peptides separately, at a concentration of 1 µM. After 2 hours of incubation, the peptides were detectable in the cytoplasm of REH cells and were not trapped in the cytoplasmic membrane (Figure 5A).

We investigated RAC1 activation in the presence of the 2 peptides (Figure 5B). The CD9 peptide significantly decreased basal levels of RAC1-GTP (P = .04) and lower levels of RAC1 activation in response to CXCL12 were reported for this peptide than for the scrambled peptide. Cells expressing the scrambled peptide had long cytoplasmic extensions (Figure 5C), most of which (41.6%) were oriented in the same direction (Figure 5D). By contrast, cells expressing the CD9 peptide had a smaller number of extensions and their extensions were shorter (Figure 5C) and more uniformly distributed around the cell (Figure 5D). These findings suggest that CD9 enhances RAC1 activation and actin remodeling via its C-terminal sequence, in response to CXCL12 (Figure 5E).

As the testis is a major site of late relapses and CXCL12 is expressed in mouse testis tissue (supplemental Figure 5), we assessed the biological relevance of this molecule in the migration induced by a TCM. The cells that had migrated were detected by flow cytometry. We determined the migration rate as the percentage of REH cells present in the lower chamber after incubation (Figure 6A).

On 3 independent experiments, the migration induced by the conditioned medium was impaired to similar extents by plerixafor, NSC23766, and the antibody targeting CD9 (Figure 6B). Complete CD9 depletion also significantly decreased migration (Figure 6C). Similar results were obtained for the chemotactic migration induced by CXCL12 (supplemental Figure 6A-B) and the conditioned medium (Figure 6D) with NALM6 cell line, which expresses both CD9 and CXCR4.

We previously showed that CD9 enhanced the chemotactic migration induced by the CXCL12 present in the extracellular matrix of the testis in vitro. We carried out xenografting studies to assess, in vivo, the effect of CD9 depletion on REH cell engraftment in the testis. Four weeks after injection, testis tissues were collected, dissociated, labeled with human pre-B markers and analyzed by cytometry (Figure 7A). We also assessed CD9 expression. We showed that most of the REH cells able to migrate to testis were positive for CD9, even for the REH CD9^low cell line (Figure 7B). This suggests that a CD9-positive subclone within a predominantly CD9-negative population would engraft to the testis tissue more efficiently than the other cells. Similar results have been reported for NALM6 cells displaying stable downregulation of CD9 (CD9^medium cells with an MFI for CD9 < 60) (Figure 7C). Three weeks after injection, we observed a small but significant difference in BM homing between NALM6 CD9^medium cells and the control cell line (NALM6 CD9^high with MFI>100), with the control cells displaying slightly higher rates of BM homing. This effect was particularly marked in the testis, which contained more NALM6 CD9^high cells than NALM6 CD9^medium cells after xenografting (number of NALM6 cells detected in the testis per 10⁵ testis cells: median value of 24.8 for NALM6 CD9^high cells and of 3.43 for NALM6 CD9^medium cells).

CD9 is one of the best-characterized tetraspanins, but the functional involvement of CD9 in the mobility of leukemic blasts has never before been reported. We found that CD9 upregulated the adhesion of B leukemic cells to fibronectin and the migration of these cells in response to CXCL12, through a pathway dependent on RAC1.

Pre-B cells express a limited range of chemokine receptors, and CXCR4 is one of the most strongly expressed. The CXCL12/CXCR4 axis is of major importance in several biological processes. Its primary role is maintenance of the hematopoietic niche, through the surveillance of signaling and self-renewal and by allowing HSC homing to occur.²⁵  This axis is also essential for the retention of ALL cells in the BM³⁰  and it plays a role in the dissemination of leukemic cells.^31-33  We showed, in xenografting experiments on NSG mice, that reduced CD9 expression increased mouse survival, consistent with lower levels of dissemination and engraftment abilities. Our results are consistent with those of a recent study showing that CD9-positive cells are more tumorigenic than their CD9-negative counterparts.¹⁷ 

CD9 and CXCR4 were colocalized on the REH cell membrane. We therefore investigated the effect of CD9 on CXCR4 internalization following CXCL12 binding. CD9 has already been shown to attenuate epidermal growth factor receptor signaling by increasing receptor internalization following epidermal growth factor stimulation.³⁴  However, we found that CD9 depletion had no effect on CXCR4 internalization. By contrast, following CXCL12 stimulation, most of the CD9-positive cells had long cytoplasmic extensions, mostly oriented in the same direction and containing actin filaments not detected in the cells depleted of CD9. Such structures have been observed in B cells stimulated with interleukin-4 and lipopolysaccharide.³⁵  Actin extensions are finger-like protrusions containing parallel bundles of actin filaments that protrude from the leading edge of many motile cells.³⁶  We suggest that CD9 may be a partner in the CXCL12/CXCR4 axis, optimizing the formation of efficient motility structures. Severinson and Westerberg reported that actin filament assembly and growth in B cells was driven by cytokine receptor activation, leading to the production of PIP2 and active Rho-family GTPases.³⁵  Rho--family proteins are molecular switches that alternate between inactivated and activated states. Activation leads to the formation of new actin filaments and their stabilization.³⁶  RAC1 is one of the best studied member of the Rho family; it plays a critical role in actin remodeling during adhesion and migration events.²⁹  Some studies have strongly suggested that RAC1 is a key player in HSC homing and retention in the BM.^29,37,38  In leukemia, RAC1 is overexpressed in samples from ALL patients³⁹  and enhances CXCL12-induced migration and homing to BM niches.⁴⁰  Furthermore, Mangolini et al showed that ETV6/RUNX1 translocation enhances RAC1 signaling which promotes leukemogenesis.⁴¹  We show here that CD9 enhances the CXCL12-induced activation of RAC1 via its C-terminal tail. Moreover, as for CD9 depletion, changes in RAC1 activation impaired actin rearrangement, chemotactic migration in vitro, and BM homing in vivo. These results are consistent with the findings of Wang et al, who showed that the CD9 C-terminal tail was involved in the formation of microvilli and filopodia, in which CD9 was expressed and colocalized with F-actin.⁴²  The mechanism by which CD9 enhances RAC1 activation in response to CXCL12 signaling remains unknown, but our findings highlight the mechanisms underlying the motility of leukemic blasts, potentially accounting for chemotactic dissemination of B-ALL in vivo. Indeed, the CXCL12/CXCR4 pathway plays a critical role in cancer, by enhancing the tropism of cancer cells for organs expressing CXCL12.⁴³  Furthermore, in childhood ALL, CXCR4 expression on the cell surface is predictive for extramedullary infiltration.⁴⁴  CXCL12 is expressed in the testis,⁴⁵  one of the principal extramedullary sites of late ALL relapses. We demonstrate here that CXCR4 signaling plays a major role in the migration of pre-B lymphoblasts to the testis, as the chemotactic migration induced by a TCM was abolished by plerixafor. CD9 and RAC1 are also key players in this migration. In vivo, the engraftment of pre-B lymphoblasts in the testis is highly dependent on the level of CD9 expression. These results are consistent with those of recent studies on spermatogonial stem cells (SSCs) showing that (1) CXCL12 is a chemotactic factor inducing SSCs to migrate to testis niches,⁴⁵  (2) CD9 is a marker of the male germline, with testis-repopulating potential,⁴⁶  and (3) RAC1 enhances transendothelial migration across the blood-testis barrier.⁴⁷ 

In summary (Figure 5E), we show here, for the first time, that the tetraspanin CD9 can interfere with CXCL12/CXCR4 signaling by enhancing RAC1 activation in a pre-B ALL cell line and, more importantly, in samples from patients. RAC1 activation enhances actin rearrangement and the formation of cytoplasmic extensions, promoting chemotactic migration toward CXCL12 in the testis, a recognized site of late ALL relapses. These data indicate that CD9 increases the ability of pre-B leukemic cells to disseminate, through its effects on migration and homing, suggesting a possible key role in late relapses of B-ALL, which are currently poorly misunderstood.

The authors thank the Gene Expression and Oncogenesis team for helpful discussions and for providing scientific expertise, Prof Vincent Praloran and Dr Arnaud Villacrecès for the NALM6 cell line and the xenograft protocol for NOD/SCID mice, and Dr Elisabetta Dondi for providing technical expertise on flow cytometry. We thank the Animalerie Rennaise Centre d'Hébergement et d'Expérimentation (Stéphanie Tremulot, Gaëlle Bourgine), flow cytometry and cell sorting (Laurent Deleurme, Gersende Lacombe), and Microscopy Rennes imaging center (Agnès Burel, Marie-Thérèse Lavault) technological platforms from UMS CNRS 3480/US INSERM 018 and Centre de Microscopie Electronique à Balayage et de Microanalyse (Joseph Le Lannic) for technical support and scientific expertise.

This work was supported by la Ligue Régionale contre le Cancer (committees 22, 35, and 56; M.-P.A., V.G., M.-B.T.), European Union FP7 (REA grant no. 291851; M.-B.T.), Association Laurette Fugain (V.G., M.-P.A.), SFR Biosit UMS 3480 (M.-B.T., M.-P.A.), la Région Bretagne (M.-P.A.), l’Institut de la mère et l’enfant de Rennes (M.-P.A.), the CNRS, and Université de Rennes 1.

Contribution: M.-P.A. and M.-B.T. designed the study, performed experiments, analyzed the data, and wrote the paper; A.V. designed the study, performed experiments, and analyzed the data; G.R., J.B., and A.-G.R. performed experiments and analyzed the data; N.V.-B. participated in drafting and revising the article; C.L. contributed to the analysis and interpretation of data; M.-D.G. designed the study and wrote the paper; V.G. designed the study, analyzed the data, and wrote the paper; and all authors reviewed the paper and agree with its content.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Virginie Gandemer, IGDR-CNRS UMR 6290, Faculté de Médecine, 2 avenue du Professeur Léon Bernard, CS 34317, 35043 Rennes Cedex, France; e-mail: virginie.gandemer@chu-rennes.fr.