By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
IL-33 has recently been identified as a cytokine endowed with pro-Th2 functions, raising the question of its effect on invariant natural killer T cell (iNKT), which are potent IL-4 producers. Here, we report a two-fold increase of iNKT-cell counts in spleen and liver after a 7-day treatment of mice with IL-33, which results from a direct effect, given that purified iNKT cells express the T1/ST2 receptor constitutively and respond to IL-33 by in vitro expansion and functional activation. Conversely to the expected pro-Th2 effect, IL-33 induced a preferential increase in IFN-γ rather than IL-4 production upon TCR engagement that depended on endogenous IL-12. Moreover, in combination with the pro-inflammatory cytokine IL-12, IL-33 enhanced IFN-γ production by both iNKT and NK cells. Taken together these data support the conclusion that IL-33 can contribute as a co-stimulatory factor to innate cellular immune responses.
IL-33 (or IL-1F11) has recently been identified as a ligand of the orphan T1/ST2 receptor, a member of the IL-1 receptor (IL-1R) family 1 that was initially described as a nuclear factor, nuclear factor from high endothelial venules, abundantly expressed by endothelial cells in lymphoid tissues 2, 3.
IL-33 induces its biological effects through a heterodimeric complex comprising the T1/ST2 receptor 1 and the IL-1R accessory protein (IL-1RAcP), another member of IL-1R family 4, 5. T1/ST2 engagement triggers a signalling pathway that requires MyD88 and NF-κB 1, 4, 6. It has long been known that T1/ST2 is expressed primarily in mast and Th2 cells and is associated with important Th2 effector functions 7–9. Accordingly, IL-33 has been found to promote Th2 cytokine production by mast cells and polarized T cells in vitro, and to induce pulmonary and mucosal Th2 inflammation when administered in vivo1.
iNKT cells constitute a distinctive subpopulation of mature αβ-T cells bearing an invariant TCR α-chain together with NK-cell receptors 10, 11. They recognize glycosphingolipid Ags presented by CD1d, a non-classical class I-like Ag-presenting molecule, and respond rapidly to TCR stimulation with α-galactosylceramide (α-GC) by generating a number of cytokines, particularly IFN-γ and IL-4 10, 11. In most disease models in which iNKT cells have been implicated their beneficial or detrimental effects have been ascribed to either Th1 or Th2 cytokines 10, 11. It has also been established that the balance between these two profiles depends essentially on the microenvironment, which favours IL-4 or IFN-γ production 12–17.
Given its previously established pro-Th2 functions, IL-33 seemed a plausible candidate for the regulation of iNKT-cell activities, prompting us to investigate whether it could directly interact with this regulatory cell subset to drive IL-4 production. Starting from the observation that the incidence of iNKT cells was increased in spleen and liver of mice injected with IL-33, we examined how this treatment affected their functional status in terms of activation, as well as IFN-γ and IL-4 production. We found that IL-33 is a potent co-stimulator of iNKT cells that induces, in combination with IL-12, a preferential increase of IFN-γ production in vivo.
IL-33 targets and activates iNKT cells to increase their numbers both in vivo and in vitro
Since repeated injections of IL-33 promote striking pathological modifications associated with Th2 differentiation 1, we addressed the question whether iNKT cells, which are a potent source of IL-4, were affected by this treatment. To this end, after seven daily i.p. injections with IL-33, spleen and liver iNKT cells were counted as α-GC-loaded CD1d TT+ cells expressing an intermediate level of TCR-β. Mice treated with IL-33, developed splenomegaly, as previously described 1 as well as hepatomegaly (lymphocyte counts: IL-33: 1.92±0.15×105versus 0.99±0.14×105 after vehicle, p<0.005). IL-33 promoted a nearly two-fold increase in iNKT-cell counts in both spleen and liver (Fig. 1A, middle panels), concomitant with cellular activation, attested by up-regulation of the early activation marker CD69 (Fig. 1A, right panels). In the same experimental set-up, IL-33 did not modify mainstream T-cell (defined as α-GC-loaded CD1d TT− TCR-β+ cells) counts, even though it led to a modest increase in the expression of CD69. The incidence of iNKT cells in spleen and liver was not altered by the treatment with IL-33 (Fig. 1A, left panels; Fig. 1B), probably because of the concomitant increase of B cells (data not shown), and the significant decrease in the percentage of mainstream spleen T cells (Fig. 1A, left panels).
We next performed in vitro experiments to assess whether IL-33 contributed directly to iNKT-cell expansion. To address this question, we evaluated its effect on IL-7-induced thymocyte proliferation, an experimental procedure leading to the expansion of both iNKT and mainstream mature T cells (18; Fig. 2, panel A versus panel H). Even though IL-33 did not support the survival of mature thymocytes on its own (data not shown), it clearly increased the number of iNKT cells after a 5-day exposure to IL-7 (IL-7+IL-33: 2.82±0.34×106versus IL-7: 1.07±0.14×106, p<0.05). Furthermore, iNKT-cell frequency was significantly higher in cultures supplemented with IL-33 (Fig. 2B), whereas that of mainstream T cells remained unchanged (Fig. 2C), indicating a preferential effect of IL-33 on the iNKT-cell fraction. Once again, as in vivo, the surface expression of the early activation marker CD69 was up-regulated preferentially on iNKT cells (Fig. 2D) compared with mainstream T cells (Fig. 2E). Lastly, in support of a direct interaction, iNKT cells sorted from thymocyte cultures with IL-7, expressed the T1/ST2 receptor (Fig. 2F) and proliferated in response to IL-33 (Fig. 2G). Note that freshly isolated thymic iNKT cells (Fig. 2H) expressed the T1/ST2 receptor constitutively (Fig. 2I) and the percentage of T1/ST2+ cells in this population was higher than in mainstream T cells (Fig. 2J). IL-33R gene expression was further confirmed by real-time RT-PCR since both ST2 and IL-1RAcP transcripts were consistently detected in purified thymic iNKT cells using 23.3 and 26.9 threshold cycles, respectively. A similar result was obtained with spleen and liver iNKT cells that did likewise express the T1/ST2 receptor constitutively, as assessed by flow cytometry (Fig. 3A) and real-time RT-PCR (data not shown).
IL-33 directly enhances IL-4 and IFN-γ production by iNKT cells upon TCR engagement
One of the original features of iNKT cells consists in their capacity to release large amounts of both Th1 and Th2 cytokines, mainly IFN-γ and IL-4, following TCR engagement. We examined the effect of IL-33 on this particular Th1/Th2 cytokine profile, by measuring IFN-γ and IL-4 production by total spleen cells stimulated with the specific iNKT-cell ligand α-GC. We observed a more than two-fold augmentation of IL-4 levels in these conditions (Fig. 3B), in agreement with the previously reported pro-Th2 effect of IL-33 on other cell populations, such as mast cells 5, 6, 19, 20 and Th2 lymphocytes 1, 4. Surprisingly, IL-33 induced much more IFN-γ than IL-4 since it induced on average a 30-fold increase in response to α-GC, which is 15 times more than the effect on IL-4. The enhancement was strictly iNKT-cell-dependent since IFN-γ was detected at very low levels and IL-4 was virtually absent in the supernatants of cells recovered from the iNKT-cell-deficient mice (Fig. 3B). The up-regulation of IFN-γ and IL-4 production by IL-33 alone or in combination with α-GC was also abrogated in the absence of the MyD88 adaptor molecule (Fig. 3B), a member of the IL-33 signalling receptor complex 1, 4, 6. The preferential effect of IL-33 on IFN-γ was maintained when splenocytes from BALB/c rather than C57Bl/6 mice were used (Fig. 3C), showing that the pro-Th1 effect of IL-33 persists in strains biased towards Th2-mediated immunity.
We further investigated the effect of IL-33 on cytokine production by iNKT cells in vivo. As shown in Fig. 4, both IFN-γ and IL-4 were rapidly and markedly increased in the serum of mice having received IL-33 together with α-GC, as compared with α-GC-injected controls. To prove that iNKT cells were responsible for this cytokine production, we analysed IFN-γ expression in single cells using intracellular staining. As shown in Fig. 4B, IL-33 augmented not only the percentage of IFN-γ+ cells among the gated iNKT subset of the spleen, but also upregulated their CD69 surface expression (Fig. 4C).
We confirmed that IL-33 targeted iNKT cells specifically by sorting them electronically from both spleen and liver, using the α-GC-loaded CD1d TT and the receptor CD5 for positive selection (Fig. 5A). DC loaded with α-GC (DC α-GC) promoted a substantial IL-4 production by sorted α-GC-loaded CD1d TT+ CD5+ spleen cells that was about two-fold higher in the presence of IL-33. Once again, IFN-γ rather than IL-4 production was enhanced, as attested by its ten-fold increase relative to control cultures set up with DC α-GC alone.
As shown in Fig. 5B and C, a significant but less marked increase in cytokine production in response to IL-33 was obtained when freshly isolated α-GC-loaded CD1d TT+ CD5+ cells from spleen and liver were stimulated with coated anti-CD3 mAb instead of α-GC-loaded DCs, suggesting a possible contribution of DC. IL-33 enhanced the effect of anti-CD3 mAb at both suboptimal (1–2 μg/mL) and optimal (5–10 μg/mL) concentrations (data not shown) and promoted preferential IFN-γ production, which increased by a factor 10 instead of 2 for IL-4. This ratio remained the same at different doses of anti-CD3 mAb (data not shown). Lastly, IL-33 amplified the proliferation of sorted liver or spleen iNKT cells in response to TCR engagement (Fig. 5, right panels) in accordance with a direct interaction.
Endogenous IL-12 contributes to IL-33-dependent enhancement of IFN-γ production by iNKT cells
It has been reported that α-GC-loaded DC produce IL-12 upon interaction with iNKT cells, which in turn enhances their IFN-γ production 12. To examine whether this mechanism was involved in our experimental protocol, we performed DC/iNKT co-culture experiments in the presence of neutralizing anti-IL-12 mAbs. As shown in Fig. 6A, the blockade of endogenous IL-12 largely reduced IFN-γ levels generated during incubation of iNKT cells stimulated with α-GC-loaded DC and IL-33, whereas IL-4 production remained unchanged. As a definite proof for the implication of IL-12, we repeated the experiments in C57Bl/6 mice in which the IL-12 p40 chain had been deleted (Fig. 6B). Once again, the loss of IL-12 reduced IFN-γ but not IL-4 production upon costimulation with α-GC and IL-33.
IL-33 enhances IL-12-induced IFN-γ production by iNKT and NK cells
iNKT cells participate not only in acquired but also in innate immune responses, since, like classical NK cells, they can be fully activated and produce large amounts of IFN-γ without TCR engagement, after exposure to the pro-inflammatory cytokine IL-12 12, 13, 21. When assessed in this experimental set-up, IL-33 markedly enhanced the synthesis of IFN-γ by sorted α-GC-loaded CD1d TT+ CD5+ cells from spleen and liver, stimulated with IL-12 (Fig. 7A). However, it is important to note that the sorting of iNKT cells using α-GC-loaded CD1d TTs implies a TCR engagement, which may have conferred responsiveness to IL-33, as previously reported 22. To avoid this stimulation, we purified NK1.1+CD5+ iNKT cells and found that unlike their α-GC-loaded CD1d TT+ counterpart (see Fig. 5B and C), they produced neither IFN-γ nor IL-4 spontaneously or in response to IL-33 alone but conserved their reactivity to IL-12 and its enhancement by IL-33 (Fig. 7B). These data are in accordance with a possible involvement in early immune responses occurring independently from TCR engagement. Consistent with this view, we found that other typical IL-12 responder cells, namely the NK- (NK1.1+CD3−) cell population, expressed the T1/ST2 receptor constitutively, in both spleen and liver (Fig. 8A). Furthermore, its functionality was assessed by the remarkable 250-fold increase of IFN-γ production that occurred when IL-33 was added to sorted NK1.1+CD5 cells from spleen and liver stimulated with IL-12 (Fig. 8B).
Our study provides the first evidence that IL-33 directly targets immunoregulatory iNKT cells to increase their state of activation and their incidence in spleen and liver after a 7-day in vivo treatment. This effect resulted from a direct interaction, as assessed by in vitro experiments with purified iNKT cells which express the T1/ST2 receptor constitutively and are expanded and functional in response to IL-33.
Based on the current view that IL-33 exerts its biological activity by promoting a Th2 immune response 1, we expected that it would also drive a preferential IL-4 production by iNKT cells. However, it turned out that exposure to IL-33 privileged the production of IFN-γ production in response to TCR engagement, revealing a marked pro-Th1 effect that was enhanced on average 30-fold, that is 10–15 times the increase of IL-4, as well as that of other pro-Th2 cytokines, namely IL-5 and IL-13 (data not shown). Finally, the capacity to promote a pro-Th1 cytokine profile constitutes a general feature of IL-33 applying likewise to C57Bl/6 and BALB/c mice, which are biased towards a pro-Th1 and a pro-Th2 response, respectively.
It is not yet clear whether this activity is restricted to iNKT cells or applies to conventional T-cell populations. Our preliminary data indicate that IL-33 can also amplify IFN-γ production by conventional CD4+ memory T cells, sorted as CD44+CD62L−CD4+ α-GC-loaded CD1d tetramer− (TT) cells, in response to primary TCR stimulation (data not shown). The discrepancy between these findings and those reported by Schmitz et al.1 might be explained by the use of polarized Th1 and Th2 cells rather than freshly isolated, naturally activated CD4+ T cells, as in our experimental setup.
The preferential IFN-γ production induced by IL-33 in iNKT cells indicates that the immunomodulatory functions of this newly discovered cytokine might actually be more complex than initially assumed. Depending on the local cytokine environment it might indeed behave like an enhancer of Th1-mediated inflammation, as is the case in our study where endogenous Il-12 mediates the enhancement of IFN-γ production by iNKT cells in response to IL-33 both in vitro and in vivo. Further investigations are now needed to determine to what extend the contrasting activities of iNKT cells regarding Th1 21versus Th2 23 adaptive immune responses are regulated by endogenous IL-33 in vivo. Our results also support the notion that IL-33 may contribute not only to acquired but also to innate immune responses by targeting iNKT cells. Indeed, we demonstrate that IFN-γ production can be induced without TCR cross-linking, provided that the proinflammatory cytokine IL-12 is present during stimulation. Rapid regulation of ST2 expression during human NK-cell activation by IL-12 is one mechanism that has been proposed to explain the ability of IL-33 and IL-12 to synergistically induce IFN-γ 24. However, we detected no up-regulation of ST2 surface expression on mouse iNKT cells in response to IL-12 both in vitro and in vivo (data not shown), suggesting that other mechanisms must be involved.
The capacity of IL-12-induced IFN-γ production by iNKT cells exposed to IL-33 is of potential interest in a patho-physiological setting, considering the well-established regulatory functions of this cell population in anti-infectious and anti-tumor immune responses 10, 11, 13, 25. The potential involvement of the IL-33/ST2 pathway in innate cellular immune responses is also supported by our observation that IL-33 targets not only iNKT but also NK cells, both in vitro and in vivo, and by previous reports on the release of pro-inflammatory mediators by mast cells in response to IL-33 5, 20, 26. It is not yet clear how IL-33 increases cytokine production by iNKT or NK cells. According to our in vitro data, the interaction with its receptor alone is not sufficient to induce IFN-γ secretion, which failed to occur in the absence of TCR or IL-12 signalling in cultures set up with naive iNKT (CD5+ NK1.1+) (Fig. 7B) or NK cells (Fig. 8). The same conclusion applies to IL-4 production (data not shown), suggesting that IL-33 acts as a cofactor rather than an inducer per se. Yet, somewhat paradoxically, IL-33 administration alone activated both iNKT (Fig. 1) and NK cells (data not shown) in vivo. It is conceivable that this occurs indirectly through endogenous factors generated by other target cells of IL-33. This hypothesis is consistent with the fact that IFN-γ could only be detected in response to IL-33 alone in cultures set up with whole spleen cells (Fig. 3B and C) but not with sorted iNKT cells. It remains to be determined whether endogenous IL-33 provides a general means of amplifying inherent iNKT-cell activities, by enhancing both IL-4 and IFN-γ secretion, as previously proposed for IL-12 16, 17. This hypothesis is consistent with the constitutive surface expression of T1/ST2 receptors by iNKT cells.
As to the molecular events leading to the pro-Th1 effect, our data support the conclusion that IL-33 amplifies the signalling pathway initiated by TCR cross-linking or IL-12 since it has no effect on its own. It remains to be elucidated whether this occurs through the recruitment of MyD88 adaptor protein, which we found to be required for the activity of IL-33.
Another important issue concerns the relevance of our results in humans. According to preliminary experiments, IL-33 amplifies the expansion of iNKT cells in PBMC in response to the α-GC ligand and enhances the production of IFN-γ by IL-12-activated NK cells (J. M. G. and A. B.; unpublished data). These data, along with those recently reported by Smithgall et al. 24, support a similar mechanism of action in mice and humans.
In conclusion, we demonstrate a new pro-Th1 activity of IL-33, which can act as a co-stimulatory factor in innate cellular immune responses. Our findings challenge the prevailing opinion that IL-33 is strictly a pro-Th2 cytokine and provide further evidence for the importance of the microenvironment and the cellular context for determining both Th1 and Th2-oriented immune responses.
Materials and methods
Murine rIL-12, rIL-33 and anti-IL-12 mAb (clone C17.8) were purchased from R&D Systems (Abingdon, UK). α-GC (KRN 7000) was provided by the Pharmaceutical Research Laboratory of Kirin Brewery (Gunma, Japan). Fluorochrome-conjugated anti-βTCR (clone H57-597), anti-CD3 (clone 500-A2), anti-CD4 (clone RM4-5), anti-CD5 (clone 55-7.3), anti-CD69 (clone H1-2F3), anti-CD44 (IM7), anti-IFN-γ (clone XMG1.2) and anti-NK1.1 (clone PK136), and corresponding isotype controls were from BD Pharmingen (San Diego, CA). The FITC-conjugated anti-T1/ST2 and its FITC-conjugated isotype control Ab were obtained from MD Bioscience and BD Pharmingen, respectively. APC-conjugated α-GC-loaded CD1d TT was obtained through the NIH tetramer facility.
Production of human rIL-33
The human cDNA encoding IL-33 aa 112–270 3 was subcloned into expression vector pET-15b (Novagen), and human rIL-33 was produced in E. coli BL21pLysS (Novagen) and purified on Ni-NTA agarose (Qiagen), according to the manufacturer's instructions.
Mice and in vivo treatments
Seven- to eight-wk-old mice were used in this study. Wild-type, mutant Jα−/−18, MyD88−/− female mice on a C57Bl/6 genetic background and wild-type female mice on a BALB/c genetic background were bred and maintained in our animal facility under specific pathogen-free conditions. Mutant C57Bl/6 IL-12 p40−/− mice and their wild-type male controls from the Jackson Laboratory were provided by Jean-François Arnal (Institut National de la Santé et de la Recherche Médicale, U589, Toulouse, France). The in vivo efficiency of IL-33 was controlled by measuring the increase of the hematopoietic cell counts in the spleen of treated mice. Based on preliminary experiments using a seven-day treatment with various doses of IL-33 (1, 2, 4 and 8 μg daily) followed by sacrifice 12 h after the last injection, we currently injected a dose of 4 μg per day and per mouse, which was sufficient to induce reproducibly a maximal splenomegaly, as previously described by Schmitz et al. 1. In another series of experiments, mice received a single i.p. injection of 2 μg α-GC in combination or not with 4 μg of IL-33, and were sacrificed 2 h later. Animal experiments were performed according to the French institutional committee.
All cell preparations were carried out in complete RPMI ((RPMI 1640; Invitrogen Life Technology, Grand Island, NY) supplemented with 10% FCS, antibiotics and 2-ME). After perfusion with PBS, livers were homogenized through a 100-μm cell strainer and washed. Parenchymal cells were removed by centrifugation at 50g for 5 min. The suspension was centrifuged over a 35% Percoll (Amersham Biosciences Europe, Orsay, France) gradient and mononuclear cells (MNC) were isolated by harvesting the interphase. Spleen cells and thymocytes recovered after homogenization and lysis of RBC in ammonium chloride buffer were suspended in complete RPMI 1640.
Staining for flow cytometry, FACS sorting and cell purification
Cells were re-suspended in staining buffer (PBS with 10% FCS) and anti-CD16/32 (BD Pharmingen) to block non-specific binding. Membrane labelling, as well as intracellular cytokine staining, was performed as described before 16. At least 1500 events gated among the population of interest were analysed on a FACS Canto cytometer using BD FACSDiva software (BD Bioscience). The proportion of α-GC-unloaded CD1d TT+ cells among gated β-TCR+ T cells was always below 0.05, 0.1, and 0.5% in thymus, spleen and liver, respectively. iNKT cells were sorted from thymus, liver or spleen as α-GC-loaded CD1d TT+CD5+ cells, while NKT and NK subsets were distinguished and sorted from the liver or spleen as CD5+NK1.1+ and CD5−NK1.1+ cells. In some experiments, thymic iNKT cells were sorted after a 5-day expansion of whole thymocytes with IL-7 (see section Cell culture and cytokine assays). Prior to sorting, freshly isolated thymocytes or splenocytes were enriched for iNKT and/or NK cells by magnetic depletion of CD8, CD11b, CD62L and CD19 cells (Invitrogen Life Technology), according to the manufacturer's instructions. Sorted cells were routinely >97% pure.
Cell culture and cytokine assays
Activation and expansion of thymic iNKT cells were assessed by flow cytometry after a 5-day culture of whole thymocytes (20×106/well) with or without IL-33 (10 ng/mL) in the presence of IL-7 (40 ng/mL), as previously described 18. Bone marrow-derived DC were prepared as reported 19. On day 6, DC were loaded or not with α-GC (100 ng/mL) for 18 h. A total of 2.5×104 sorted iNKT, NK and NKT cells were cultured for 48 h in 200 μL complete RPMI with or without coated anti-CD3 mAb (10 μg/mL, BD Pharmingen), α-GC-loaded or unloaded DC (5.0×103), in the presence or absence of IL-33 (10 ng/mL) or IL-12 (20 ng/mL) or both, and with or without anti-IL-12 mAb (5 μg/mL) in round-bottomed 96-well plates at 37°C and 5% CO2. In another set of experiments, total splenocytes (0.5×105 cells/well) were incubated for 48 h with or without α-GC (100 ng/mL) in the presence or absence of IL-33 (10 ng/mL). IL-4 and IFN-γ in supernatants were quantified using standard sandwich ELISA, as previously described 18. Cells were pulsed [3H]thymidine, then harvested and thymidine uptake was assessed as described 14.
RNA isolation and real-time RT-PCR
Total RNA was extracted from thymic iNKT and T lymphocytes or from liver, spleen and thymic iNKT cells using the SV Total RNA isolation Kit® (Promega, Charbonnieres-lesbains, France) and then subjected to a reverse transcription reaction using high-capacity cDNA archive kit® (Applied Biosystem, Foster City, CA). A total of 30 ng total complementary DNA was used as a template for amplification with primers specific for T1/ST2 (for.: 5′TACCAGGGTGGAGCCTACT3′, rev.: 5′GCCCAACCTTCTACCTCCTC3′); IL-1RacP (for.: 5′TGAGCTTTTTCATCCCCTTG3′, rev.: 5′ATAGATCTGGGGTGGCAATG3′) and 18S (for.: 5′CGCCGCTAGAGGTGAAATTC3′, rev.: 5′TTGGCAAATGCTTTCGCTC3′) used at a 250 nM final concentration. The real-time quantitative PCR were performed as previously described 16.
Data are expressed as means±SEM. Mean differences between experimental groups were evaluated by Student's unpaired t-test. Paired statistical analysis of the in vitro effect of IL-33 was performed according to Wilcoxon or Student's t-test, as appropriate.The p-values under 0.05 were considered statistically significant.
We gratefully acknowledge Drs. T. Nakayama and M. Taniguchi (Chiba University, Japan) for the gift of Jα18−/− C57Bl/6 mice. We also thank Dr. S. Akira (Osaka University, Japan) for the generous gift of MyD88−/− B6.129 mice and Dr. N. Thieblemont for providing us this mutant strain on the C57Bl/6 background. We thank the NIH tetramer facility, Kirin Brewery for α-GC (KRN7000), Pauline Général from B&Beud for helpful discussions. We thank A. F. Bertron, R. Bricard and M. Levasseur for technical assistance. We are grateful to Jérôme Megret for performing all cell sortings. This study was supported by CNRS, Université de Paris Descartes, Ligue Nationale contre le Cancer (to J. P. G., Equipe labellisée) and by institute funds from la Chancellerie des Universités de Paris (Legs Poix). E. B. and L. P. V. were recipients of a doctoral fellowship from the Ministère de l'Education Nationale de la Recherche et Technologie and from the Vietnam government, respectively.
Conflict of interest: The authors declare no financial or commercial conflict of interest.