We describe herein a new transgenic mouse tumor model in which fibroblast growth factor (FGF) receptor activity is selectively inhibited. Tyrp1-Tag mice that develop early vascularized tumors of the retinal pigment epithelium were crossed with tyrp1-FGFR1-DN mice that express dominant-negative FGF receptors in the retinal pigment epithelium to generate bigenic mice. Initial angiogenesis-independent tumor growth progressed equally in tyrp1-Tag and bigenic mice with no significant differences in the number of dividing and apoptotic cells within the tumor. By contrast, at a later stage when tyrp1-Tag tumors rapidly expanded to fill the entire eye posterior chamber and migrate along the optic nerve toward the chiasma, bigenic tumors remained small and were poorly vascularized. Secondary tumors of small size developed in only 20% of bigenic mice by 1 month. Immunohistochemical analysis of secondary tumors from bigenic mice showed a reduction of angiogenesis and an increase in apoptosis in tumor cells. Tumor cells from bigenic mice expressed high levels of truncated FGF receptors and did not induce endothelial tube formation in vitro. All in all, this indicates that the tyrp1-Tag mouse may be a useful model to study selective tumor inhibition and the effect of antitumor therapy that targets a specific growth factor pathway. FGF receptors are required at the onset of tumor invasion and angiogenesis in ocular tumors and are good therapeutic targets in this model. The bigenic mouse may also constitute a useful model to answer more fundamental questions of cancer biology such as the mechanism of tumor escape.
Fibroblast growth factors (FGF) stimulate growth, survival, and/or differentiation of a number of mesenchyme-derived cells and neurons. Most of these functions have been demonstrated in cultured cells and in transgenic mice (1). It has long been known that exogenous FGF is a good inducer of capillary formation in vitro and in vivo, but its function in physiology or pathology remained ill defined (1, 2). We recently reported that tyrp1-FGFR1-DN mice overexpressing a mutant FGFR1 receptor in the retinal pigment epithelium (RPE) suffer from defects in the eye vasculature (3, 4). These mice have a defect in embryonic choroidal angiogenesis that is characterized by a decrease in the number of capillaries and poorly branched vascular network. In addition, retinal angiogenesis is also strongly inhibited. These observations suggest that FGF is a key component of blood vessel development in the eye.
The function of growth factors, receptors, or inhibitors in tumor development is mostly studied by using tumorgrafts in immunodeficient or syngeneic mice. These models suffer several drawbacks, i.e., they do not recapitulate the exact environment in which the tumor grows, invades and metastasizes. Thus, there is a need for alternative models that more accurately reflect the correct relationship of the tumor to its host environment. The tyrp1-Tag mice develop highly vascularized ocular tumors that start to develop from the RPE before birth and progressively invade the optic nerve and the brain and metastasize in lymph nodes and the spleen (5). Mice die usually at 2 months of age. In the present study, we aimed to validate this model for selective tumor inhibition by targeting a specific growth factor signaling pathway. By using a genetic approach, we asked whether specific inhibition of FGFR activity could delay tumor growth and inhibit tumor invasion and angiogenesis in this model. Bigenic mice were obtained by crossing tyrp1-Tag mice with homozygous tyrp1-FGFR1-DN mice. We herein provide a detailed phenotypic analysis and demonstrate that the inhibition of FGF activity efficiently impairs tumor development, angiogenesis, and invasion in this mouse model.
MATERIALS AND METHODS
Tyrp1-Tag mice are all male hemizygotes due to the insertion of the transgene on the Y chromosome (5). These were crossed to tyrp1-FGFR1-DN homozygous females (3, 4) and to nontransgenic control females. In the first set of crosses, the progeny was composed of double hemizygous bigenic males and tyrp1-FGFR1-DN control females; whereas in the second, males were tyrp1-Tag controls, and females were nontransgenic controls.
Histology and Immunohistochemistry.
Eyes were fixed in 4% paraformaldehyde, dehydrated, and embedded in paraffin. Five-μm median sections were prepared and stained with Masson’s trichrome or H&E or prepared for immunohistology according to standard procedures. Tumor volumes were calculated from media tumor sections using the 4/3πR3 formula.
For immunodetection, the following antibodies were used: for CD31, rat antimouse MEC 13.3 (PharMingen, San Diego, CA) monoclonal antibody (1/500 dilution) and goat polyclonal antirat Alexafluor 488 secondary antibody (1/500 dilution; Molecular Probes, Eugene, OR); for proliferating cell nuclear antigen (PCNA), mouse monoclonal clone PC10 (Santa Cruz Biotechnology, Santa Cruz, CA; 1/500 dilution) and goat antimouse 546 Alexafluor secondary antibody (Molecular Probes); for SV40, mouse monoclonal large T SV40 antibody clone PAB101 (Santa Cruz Biotechnology, 1/200 dilution); amplification was performed with goat antimouse biotinylated antibodies and amplification by ABC kit (Vector Laboratories, Burlingame, CA). For in vitro angiogenesis, vessels were also stained with biotinylated Bandereira simplicifolia isolectin-1 (BS-1; Sigma, St. Louis, MO). Detection was performed with Alexafluor 488 Streptavidin (Molecular Probes).
The effect on angiogenesis was quantified by calculating the percentage of the total vascular surface of CD31-positive vessels to the total tumor surface of four tumors of each group and at least on 1-mm2 sections using Nikon software analysis. Proliferation was quantified by calculating the percentage of PCNA-positive cells from the total amount of cells. A total of 3364 and 4791 nuclei were counted on four tumors for SV40T control and bigenic mice, respectively.
Apoptosis was determined using terminal deoxynucleotidyltransferase-mediated nick end labeling (TUNEL) staining (Roche, Germany). The results were quantified by calculating the percentage of TUNEL-positive cells. A total of 10,361 nuclei of six tumors were counted for bigenic mice and 30,544 nuclei of five SV40T control tumors.
Tumor Cell Lines.
Eye tumors were isolated from 1-month-old mice, homogenized, and trypsinized. Cells were plated and cultured in DMEM containing 20% heat-inactivated fetal bovine serum. They were trypsinized and passaged at confluency every week and kept in culture for 3 months.
In Vitro Angiogenesis Coculture Assay.
The experiments were carried according to the method of Sakamoto et al. (6) with modifications. In brief, 500,000 tumor cells were mixed with a collagen suspension (2 mg/ml) and plated in 24-well plates (500 μl/well). Endothelial cells were plated on the top of the gel [500,000 bovine capillary endothelial cells (BCE)]. Tube formation was monitored after 48 h of coculture. Photomicrographs were taken using an inverted microscope equipped with a digital camera. Tubular structures were also visualized using the biotinylated BS-1. Detection was performed with Alexafluor 488 streptavidin (Molecular Probes).
For detection of dominant-negative FGFR1 (FGFR1-DN) by PCR amplification, the following primers were used: P3 primer, 5′-TCCTGAGAACTTCAGGCTCC-3′; P4 primer, 5′-TCGTCATCATCTTC CGAGG-3′. Reverse transcription and PCR were performed according to standard techniques.
Quantification of the PCR products was done in linear range of PCR amplification after staining with ethidium-bromide using NIH image 1.60 freeware.
Human recombinant FGF-2 (a gift of Dr. H. Prats Institut National de la Santé et de la Recherche Médicale, Toulouse, France) was labeled with Na125I using Iodogen (Pierce, Rockford, IL) as a coupling agent. 125I-FGF-2 binding experiments were performed as described by Moscatelli (7). Cross-linking experiments were done and analyzed as described by Auguste et al. (8), using 0.1 mm Bis(sulfosuccinyl) suberate (Pierce). Nonspecific binding was determined by competition with 200-fold excess of cold ligand or by competition with 100 μg/ml protamin sulfate. One hundred μg of protein from each sample were separated by SDS-PAGE. The gel was dried and analyzed by PhosphorImager (Molecular Dynamics, Sunnyvale, CA). Quantification of FGFR1-DN expression was done by comparing the intensity Mr ∼60,000 band (monomeric FGFR1-DN) with the higher Mr ∼120,000 band using NIH image 1.60 freeware.
Statistical analyses were performed by using the Student’s t test with Statistica 6.0 software.
Growth Pattern of Tumors in Tyrp1-Tag Mice.
Tyrp-Tag mice developed two tumor foci in two distinct regions of the eye. One tumor grew at the ciliary margin and the medial part of the eye (anterior tumor). There was no sign of neovascularization in the anterior tumor, which did not grow any further than what was observed at 4 days after birth (dpn; Fig. 1, A, C, and E). The second tumor (posterior tumor) developed around the optic nerve at the base of the retina (Fig. 1,G). At around 2 weeks after birth, this tumor reached a critical size beyond which it cannot expand unless it turns on the angiogenic switch. It then rapidly invaded the optical cavity (Fig. 1,G; Fig. 2 E), and tumor cells migrated along the optic nerve toward the brain where they formed secondary lethal tumors (data not shown).
Metastasis were seen in the spleen and inguinal lymph nodes (data not shown). All mice were moribund by the age of 2 months (5).
Expression of Dominant-Negative FGFR in Tumor Cells from Bigenic Mice.
Bigenic mice were established by crossing tyrp1-Tag mice with tyrp1-FGFR1-DN mice as indicated in “Materials and Methods.” Expression of FGFR1-DN in bigenic mice was first evidenced by reverse transcription-PCR using P3 and P4 primers. A band of 250 kb was only detected in monogenic tyrp1-FGFR1-DN and bigenic mice but not in tyrp1-SV40T mice (data not shown). To ascertain that FGFR1-DN is sufficiently expressed in bigenic tumors, tumor cells were isolated from 1-month-old bigenic and tyrp1-Tag mice and cultured in vitro. Cultured tumor cells were also characterized by labeling with anti-Tag antibodies (Fig. 3, right panel, A–D). Tumor cells from tyrp1-Tag and bigenic mice retained their Tag expression in culture. FGFR1-DN expression was checked by binding of 125I-FGF-2 and cross-linking. Fig. 3, left panel, shows that the mutant receptor is clearly identified as Mr ∼60,000 band and is overexpressed in the cultured cells at passage 8. This band was exactly the same size as that detected in C6 glioma cells overexpressing FGFR1-DN (8). Furthermore, a second band of lesser intensity migrating above Mr 116,000 was detected. This corresponds to the endogenous receptor and to dimers between endogenous receptors and/or FGFR1-DN. The intensity of signal of the Mr 60,000 band was 3.8 times higher than that of the latter band. This indicates that the expression of FGFR1-DN, in tumor cells from bigenic mice, is at least 3.8-fold higher than that of the endogenous FGF receptors.
Tumor Growth in Bigenic Mice.
We next compared tumor growth in tyrp1-Tag and in tyrp1-Tag/tyrp1-FGFR1-DN bigenic mice. Initial tumor growth was unchanged in bigenic mice up to 1 week after birth. Fig. 1 shows that tumor size within the RPE layer and at the ciliary margin was similar in tyrp1-Tag and bigenic mice aged 4 or 15 dpn (Fig. 1, A–D). We also investigated proliferation and/or apoptosis in anterior tumors. Proliferation status was evaluated by quantifying PCNA-positive cells in both groups of tumors at 15 dpn. No differences in the percentage of PCNA-positive cells were observed in both groups (PCNA-positive nuclei/total nuclei, 48 ± 6% in tyrp-Tag and 44 ± 10% in bigenics; data not shown). Furthermore, no TUNEL-positive cells were found in tumors in either group (data not shown). We also investigated angiogenesis in anterior tumors using CD31 labeling. Again, no vessels were detected (data not shown). These results indicate that FGFR1-DN does not affect proliferation, apoptosis, and angiogenesis in early tumors.
Posterior tumors were studied in mice aged 15 dpn. As shown in Fig. 1, there is a significant difference in tumor size from 15 days on (Fig. 1, G and H). Differences in tumor volume were estimated from the average diameter value. Posterior tumors from bigenic mice display an 18-fold reduction in tumor volume compared with tumors from tyrp1-Tag mice at 15 dpn (0.05 ± 0.01 mm3 for bigenic versus 0.89 ± 0.15 mm3 for tyrp1-Tag; Fig. 1; Fig. 4,A). At 1 month, this difference was even more pronounced, because tyrp1-Tag tumor grew rapidly and filled the eyeball, whereas bigenic posterior tumors were similar to stage 15 dpn (Fig. 2, E and F).
Tumor vascularization was evaluated in both groups at 15 dpn. Fig. 4 B shows an approximately 2-fold reduction in vascularized tumor surface in bigenic mice compared with tyrp-Tag controls (vascular surface, 6.2 ± 0.6% in tyrp1-Tag tumors; 3.7 ± 1% in bigenic tumors).
Apoptotic cells were quantified after TUNEL labeling. Tumors from bigenic mice displayed a 3-fold increase in the number of TUNEL-positive cells compared with tumors from tyrp1-Tag mice (apoptosis indices: tyrp1-Tag, 2.2 ± 0.5%; bigenic, 6 ± 0.8%; Fig. 4 C).
We next investigated the occurrence of secondary tumors in the brain in tyrp1-Tag and bigenic mice. All of the control tyrp1-Tag mice developed highly vascularized secondary tumors that migrate along the optic nerve and invade the brain. In bigenic mice, only 20% of the animals had secondary tumors. Limited migration on the optic nerve was observed, and only very small tumor foci were seen. Mean tumor volumes of secondary tumors from 2-month-old tyrp1-Tag mice were 69.48 ± 14.6 mm3. In bigenic mice, tumor volumes of only 0.37 ± 0.5 mm3 were observed (Fig. 5, left panel). This indicates that tumor expansion in the brain is significantly inhibited in bigenic mice. However, two tumors from bigenic mice out of 19 were 30.05 and 33.51 mm2 and two others were 19.56 and 14.4 mm2, respectively. This indicates the possibility for tumor escape. We also investigated tumor margins and centers of tyrp1-Tag and bigenic tumors for vessel number and shape. Numerous blood vessels were found in the tumor margin and tumor center of heterogenous size and shape, indicating the presence of a strong angiogenic response (Fig. 5, right panel, A and C). In secondary tumors from bigenic mice, the number of vessels was significantly reduced especially in the tumor center, and blood vessels were much more uniform in shape (Fig. 5, right panel, B and D).
Bigenic RPE Cells Are Not Able to Stimulate Angiogenesis in Vitro.
RPE tumor cell isolated from tyrp1-Tag and bigenic mice were investigated for their capacity to induce angiogenesis in vitro. Both lines grew in standard DMEM medium with 20% FCS and had similar morphology in culture.
RPE tumor cells derived from tyrp1-Tag or bigenic mice were cocultured in collagen gels with BCE, and tube formation was monitored according to the method of Sakamoto et al. (6). Rapid tube formation was observed in tyrp1-Tag/BCE cocultures (Fig. 6). In bigenic/BCE cocultures, tube formation was inhibited (Fig. 6). To visualize better vascular endothelial sprouts, cells were also stained with BS-1, a lectin that binds vascular endothelial cells. Endothelial cells were BS1 positive but not tumor cells. Sprouts observed in cocultures were clearly BS-1 positive. This indicates that angiogenic signaling is perturbed in tumor cells from bigenic mice.
We describe herein a novel bigenic mouse tumor model in which FGF receptor activity is inhibited in ocular tumors derived from the RPE. These mice display inhibition of tumor growth, invasion, and angiogenesis when compared with the monogenic tyrp1-Tag mice.
The growth patterns of tumors in bigenic mice and control tumors in the early stage are similar. Tumors have similar size and proliferation or apoptosis indexes. At a later stage, however, tumors from bigenic mice are considerably smaller and display an increase in tumor cell apoptosis. Furthermore, angiogenesis is inhibited, which is documented by two observations. First, the total vascular surface is significantly decreased in tumors from bigenic mice. Second, tumor cells derived from bigenic animals fail to induce capillary tube formation when cocultured in vitro with vascular endothelial cells. This indicates that in the tyrp1-Tag model, FGFR activity is required at a stage when tumors expand through invasion and angiogenesis-dependent growth. Thus, blockade of FGFR activity only impairs these two latter processes.
Cooption plays a significant role in the development of tumors of the central nervous system (9, 10). During cooption, tumor cells grow first around normal brain vessels and synthesize angiopoietin-2, which leads to vessel apoptosis and hypoxia. Hypoxia increases vascular endothelial growth factor synthesis and angiogenesis (9, 10). Tumor cells from tyrp1-Tag mice migrate along the optic nerve and coopt blood vessels to form secondary distant tumors. In secondary tumors from tyrp1-Tag mice, a significant high vascular density is found in the tumor center and margin with tumor cells surrounding blood vessels. Only a small fraction of bigenic mice develop secondary tumors in the brain, which are of small size. Examining the center and margin of the secondary tumors from bigenic mice, angiogenesis is much reduced. This suggests that inhibition of FGF activity leads also to a reduction of vessel cooption in this model that is possibly due to an inhibition of tumor cell invasion. Thus, impairing FGF function in tyrp1-Tag tumors primarily affects angiogenesis and tumor invasion but not cell division. This is in agreement with the fact that growth of tumor cells from bigenic or tyrp1-Tag mice is similar in early tumors.
A number of observations are in favor of a role of FGF in the development of solid tumors. Many tumor cell lines, such as glioma, lung carcinoma, or pancreatic carcinoma, synthesize FGF prototypes, particularly FGF-2 and FGF-1 (1, 2). Fibrosarcoma cells release FGF-2, which is correlated with tumor progression (11). In addition, FGF-2 urine or serum levels are correlated with the degree of malignancy and tumor progression (12, 13). Furthermore, overexpression of selective FGF forms such as Mr 18,000 FGF-2 in tumor cell lines that do not produce FGFs by themselves enhanced in vivo tumor growth and angiogenesis (14, 15, 16).
Inhibition of the biological activity by antisense molecules, neutralizing antibodies, soluble receptors, or dominant-negative FGFRs has provided a functional proof for a role of FGF prototypes in tumor growth and angiogenesis. Blockade of the FGF pathway in mice by FGF-2 or FGFR1 antisense molecules or by FGF-2 neutralizing antibody inhibits tumor angiogenesis and growth (17, 18). Adenoviral expression of soluble FGFRs does not inhibit tumor growth in vitro but decreases the number of capillaries in tumors in vivo, which suggests that in vivo tumor growth is impaired through an angiogenesis-dependent mechanism and is in agreement with the present study (19, 20). However, work from our laboratory has also demonstrated that in C6 tumors, FGFs may have a dual mode of action (8). C6 tumor cells transfected with FGFR1-DN or FGFR2-DN are growth inhibited in vitro, which indicates a direct autocrine action on tumor cells. In tumors grown in immunodeficient mice, angiogenesis and tumor development are strongly decreased, and vascular endothelial growth factor expression is down-regulated. Thus, the effect of FGF is context dependent and may vary according to the specific tumor type and the host environment.
Which FGF prototype or receptors are involved in RPE tumor development in the tyrp1-Tag model? Many FGF prototypes are able to control tumor development such as FGF-1, FGF-2, FGF-3, or FGF-4 (1, 2). For example, in the TRAMP mice in which SV40T is specifically expressed in the prostate, FGF-2 participate in interductal and intraductual tumor invasion and angiogenesis (21). Also, inducible expression of FGF-3 in the mammary gland in transgenic mice leads to hyperplasia of the mammary gland with cyst formation and multilayered ductal epithelium (22). In the RPE, various FGF prototypes are expressed, such as FGF-1, FGF-2, FGF-4, or FGF-9, that all are potential candidates for a role in tumor development in our model (23, 24, 25). These FGFs may activate either FGFR2IIIc or FGFR3IIIc on the RPE or FGFR1IIIc on the vascular stroma that may all be inhibited by FGFR1-DN in bigenic mice (25). Recent evidence indicates that FGF-2 alone is not involved in tumor development in the tyrp1-Tag mice. Indeed, bigenic mice from tyrp1-Tag mice and FGF-2 −/− mice in which FGF-2 expression is totally abolished have similar tumor development in vivo than monogenic tyrp1-Tag mice (26). Thus, the FGF prototypes that participate in the tumor development in this model are not known at the present time.
A number of transgenic mouse tumor models for the neural or ocular tissue have been described. These include mouse models for glioma, such as the GFAP-V ras mouse (27), the GFAP-V/EGFRvIII bigenic mouse, (28), the S100-v-erbB mouse (Ref. 29; the latter two are both models of oligodendroglioma), or the GFAP-SV40 T mouse, which develops high grade astrocytoma (30). Tumor formation in most of these models is dependent on a combination of genetic lesions or the genetic background such as INK4a-Arf deletion or heterozygosity for PTEN, INK-4a-Arf, or p53 (29, 30). These models also suffer from variable tumor incidence making them difficult to use for investigating the effect of selective therapeutics.
Two other models of RPE tumorogenesis have been described, the tyr-SV40T and the tyrp1-RET mouse (31, 32). In tyr-SV40T mice, bilateral ocular tumors are observed similar as in tyrp1-Tag mice. However, tumor formation starts later from dpn 7, and mice die at 4 months of age. Although skin, lung, and brain metastasis are seen, mice rarely develop secondary tumors that invade the brain from the primary ocular tumor, which is a constant feature in tyrp1-Tag mice. In the tyrp1-RET mouse, the tyrp1 promoter is also used to target the expression of the RET oncogene in the RPE. However, these mice give rise to microphthalmia and ocular tumors that do not metastasize or invade the surrounding tissue. Thus, the tyrp1-Tag mouse seems to be a better model to study tumor development, invasion, and angiogenesis than these two latter models.
FGF as a target for antitumor or antiangiogenesis molecules has been validated in one other transgenic mouse model, the Rip-Tag transgenic mouse, which gives rise to insulinomas that undergo multistage carcinogenesis (19). Using adenoviral expression of soluble FGF receptors in vivo, Compagni et al. (19) have demonstrated that the growth and angiogenesis of this tumor type are also inhibited through blockage of FGF activity.
The interest of our study is 3-fold. First, the tyrp1-Tag seems to be a good model to explore more selective therapeutics that especially target the FGF receptor pathway. A step toward this direction has been made by Bouquet et al. (33), who demonstrated that inhibiting angiogenesis with HSA-kringel 1–3 angiostatin in this model also impairs the development of secondary tumors in the brain. However, HSA-kringel 1–3 angiostatin was much less potent than selective FGF receptor inhibition as done in our study. Furthermore, angiostatin has an ill-defined mechanism of action and is able to interact with a number of target molecules such as cell surface ATPase, integrins, or the met receptor (34). Second, the tyrp1-Tag mouse may also serve as a model for tumors of the basal-region of the brain, which are among the most difficult cerebral tumors to treat (35), and could allow to investigate the effect of novel therapeutics for this tumor type. Third, the bigenic mouse may constitute a useful model to answer more fundamental questions of cancer biology. FGF is a multifunctional molecule and interacts with tumor cells and the tumor stroma. This may provide stringent inhibition of tumor development. Among the outstanding questions that may be addressed with this model are the following: How stringent is inhibition of tumor growth in vivo provided by growth factor loops, and what are the mechanisms that allow the tumor to escape? We found that four out of 19 of the secondary tumors from bigenic mice had increased tumor volumes in comparison with the other bigenic tumors. This may indicate the possibility of tumor escape under certain circumstances. Modern genomic or proteomic approaches may help to identify molecules that are implicated and unravel the mechanisms underlying this process. Furthermore, bigenic mice may be crossed with transgenic mice that express activated signaling molecules downstream of the FGF receptor and investigated for tumor escape. This may also help to identify candidate molecules that are involved in this process.
Grant support: Association de la Recherche sur le Cancer, Retina France, Institute National de la Santé et de la Recherche Medicale, Ministère de la Recherche et de la Technologie, Conseil Régional d’Aquitaine, The Swiss Cancer League (F. Beermann), The Swiss National Science Foundation (F. Beermann), and The National Center of Competence in Research Molecular Oncology, a research instrument of The Swiss National Science Foundation (F. Beermann). B. Rousseau was a recipient of a fellowship from Association de la Recherche sur le Cancer, and F. Larrieu-Lahargue was a recipient of a fellowship from Ministère de la Recherche et de la Technologie.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Andreas Bikfalvi, Laboratoire des Mécanismes Moléculaires de l’Angiogenèse, avenue des Facultés, Université de Bordeaux 1, 33405 Talence, France. Phone: 33-5-40-00-87-03; Fax: 33-5-40-00-87-05; E-mail: email@example.com
We thank Pierre Costet (Transgenic mice facility, University Bordeaux II) for invaluable help with breeding follow-up, Hervé Prats (Institut National de la Santé et de la Recherche Médicale, Toulouse) for providing recombinant FGF-2, Jean-Marie Daniel-Lamazière (Institut National de la Santé et de la Recherche Médicale, Bordeaux) for assistance with confocal microscopy, and Xavier Canron for technical advice.