cortocontributes to intervein tissue differentiation

We first investigated the ectopic vein phenotype of corto mutants using three different recessive lethal alleles: corto ⁴²⁰ , corto ^07128b and corto ^L1 . corto ⁴²⁰ is a deletion of the corto locus [25], corto ^07128b a P-element insertion located 0.5 kb upstream of corto 5'-UTR [37], and corto ^L1 an EMS-induced mutation [38]. As already described for corto ⁴²⁰ [24, 25], heteroallelic combinations using corto ^07128b , corto ^L1 and a deficiency encompassing corto [Df(3R)6-7] are poorly viable, since we observed 0% to 10% escapers depending on combinations. Therefore, these three alleles are true loss-of-function alleles. This was confirmed by quantitative RT-PCR analysis on wing discs from third instar larvae, that showed absence of corto transcripts in corto ⁴²⁰ /Df(3R)6-7 and corto ^07128b /Df(3R)6-7 larvae (Figure 1). In contrast, corto ^L1 /Df(3R)6-7 larvae exhibited the same level of corto transcripts as wild-type flies, which suggests that the mutation in corto ^L1 rather affects the level or activity of Corto protein.

As already described [25], corto ⁴²⁰ /+ heterozygous flies exhibited very few ectopic veins (2.2% to 8.6%). This phenotype was more penetrant in corto ^L1 /+ (49% to 52.2%) and corto ^07128b /+ (97.4% to 97.8%) heterozygous flies (Table 1 and Figure 2A). Since both corto ⁴²⁰ and corto ^07128b are devoid of corto transcripts, the discrepancy between these alleles may be a consequence of an interaction with the genetic background. For all combinations, the few heteroallelic corto escapers displayed a stronger ectopic vein phenotype than corto heterozygous flies (Figure 2B-D). Ectopic veins mainly arose close to longitudinal veins 2, 3, 5 and to the posterior cross-vein, which seems to be the case for most mutations that induce ectopic vein phenotypes [32]. Interestingly, over-expressing corto using a UAS::corto construct and the wing specific Beadex::Gal4 (Bx::Gal4, data not shown) or scalloped::Gal4 (sd::Gal4) driver also induced extra pieces of vein tissue in all flies (Figure 2E). Since both corto over-expression and loss-of-function induced the same phenotype, one possibility is that Corto may be required in stoechiometric amount to allow correct wing tissue differentiation. This feature characterizes proteins that act through formation of complexes. Indeed, complexes are very sensitive to the relative amounts of their components, and can be disrupted either by an excess or a shortage of one of these [39].

Genotype	Total females observed	% females with ectopic veins only	% females with blistered wings
corto 420 /+	70	8.6	0
+/corto 420	45	2.2	0
corto L1 /+	51	49	0
+/corto L1	90	52.2	0
corto 07128b /+	78	97.4	0
+/corto 07128	46	97.8	0
+/bs EY23316	136	100	0
+/bs EY23316 ; corto 07128b/+	80	67.5	32.5a
+/rho EP3704	68	0	0
sd::Gal4/+; +/rho EP3704	105	90.5	9.5
sd::Gal4/+; corto 420 /rho EP3704	34	58.8	41.2 a
sd::Gal4/+; corto 07128b /rho EP3704	90	51.1	48.9 a

For all genotypes, the first allele was brought by the mother. Only female phenotypes are reported here, but similar results were obtained for males. Numbers of females with blistered wings in flies containing a corto mutation and the bs ^EY23316 or rho ^EP3704 allele were compared to those of flies containing the bs ^EY23316 or rho ^EP3704 allele only (Z-test, ^a p < 0.001).

In conclusion, corto misregulation (either loss-of-function or over-expression) induced ectopic veins that formed within intervein tissue and never truncated veins. This observation suggested that Corto contributes to intervein tissue differentiation, whereas it does not seem to be involved in vein formation. We have previously shown that corto interacts with some TrxG genes during wing tissue formation. Indeed, moira, kismet and ash1 mutants enhance the ectopic vein phenotype of corto ⁴²⁰ [25]. Furthermore, several corto alleles enhance the ectopic vein phenotype of mutations in snr1 that encodes a component of the SWI/SNF complex [38, 40], a chromatin-remodeling complex also involved in wing tissue differentiation [41–43]. One hypothesis is that Corto, as an ETP, could participate in the recruitment of TrxG complexes to regulate expression of genes involved in wing tissue differentiation.

To clarify the role of corto in the formation of intervein tissue, we performed genetic interaction assays between corto and the intervein-promoting gene blistered (bs), or the vein-promoting gene rhomboïd (rho). As expected for a bs loss-of-function allele [35, 36], wings of flies heterozygous for bs ^EY23316 exhibited a moderate ectopic vein phenotype, but none showed blisters in the wings (Table 1 and Figure 2F). corto ^07128b enhanced the ectopic vein phenotype induced by bs ^EY23316 (compare Figure 2G to Figure 2F). In addition, 32.5% of these trans-heterozygous flies had blisters in the wings (Table 1 and Figure 2H). These blisters, which result from impaired adhesion between the ventral and dorsal wing surfaces, could be caused by formation of many vein cells within intervein tissue. They are frequently observed in bs mutants [35] or when rho is over-expressed [36]. This result therefore showed that bs and corto act synergistically to promote intervein cell fate. Ectopic over-expression of rho using the rho ^EP3704 allele and the sd::Gal4 driver induced ectopic veins for most of the flies and in a few cases (9.5%) formation of blisters (Figure 2I and Table 1). This phenotype was similar to that induced by over-expressing rho under control of a heat-inducible promoter [32]. Both corto ⁴²⁰ and corto ^07128b alleles enhanced this phenotype since the number of flies with blisters in the wings significantly increased (Table 1 and Figure 2J). This observation showed that corto antagonizes rho in vein formation.

Taken together, these results suggest that corto might antagonize rl vein-promoting function in future intervein cells. corto misregulation could therefore lead to deregulation of certain vein and intervein-promoting genes. Indeed, we observed deregulation of bs and rho in some intervein cells of pupal wings from corto ^L1 /Df(3R)6-7 escapers: in these cells, bs is down-regulated (Figure 2K) whereas rho is ectopically expressed (Figure 2L).These cells could thus acquire a vein fate.

Corto and dMP1 act together and participate in the control of wing tissue differentiation

Since the wing phenotype of corto mutants resembles the one induced by hyperactivation of ERK signaling pathway, we wondered whether corto was involved in the regulation of this pathway during wing development. We first tested genetic interactions between corto and rolled using the UAS::rolled strain which allows targeted ERK over-expression when crossed with a Gal4 driver. All flies over-expressing rolled with the sd::Gal4 driver at 25°C exhibited a mild ectopic vein phenotype ([3] and Figure 3A). Expressivity of this phenotype was enhanced by the corto ^07128b allele (Figure 3B). This result suggests that the roles of corto and rolled in vein-promoting function are antagonistic.

We have recently shown that the Drosophila ortholog of MP1, dMP1, antagonizes rl vein-promoting function in the future intervein cells of the wing [3]. Furthermore, we isolated dMP1 in a two-hybrid screen using Corto as bait (see below and Figure 4B). We thus tested the genetic interactions between corto and dMP1. As already described [3], down-regulation of dMP1 by RNA interference using the sd::Gal4 driver induced ectopic veins in 78.5% of flies (Table 3 and Figure 3E). This percentage increased to 92.2% and 100% in combination with corto ⁴²⁰ or corto ^07128b , respectively (Table 3). With corto ^07128b , the expressivity of the ectopic vein phenotype was also enhanced (compare Figure 3F to Figure2A and 3E). Therefore, these results showed that corto and dMP1 act synergistically and participate in intervein tissue differentiation in response to ERK signaling.

Corto interacts in vitrodirectly with ERK and indirectly with dMP1

We have previously shown that dMP1 forms a complex with ERK, which is required for the proper development of intervein cells [3]. To understand the molecular bases of the relationship between Corto, dMP1 and ERK, we first questioned the physical interaction between Corto and ERK. We carried out GST pull-down assays using in vitro translated ERK and GST-Corto fusion proteins. Structural analysis of Corto has shown that this 550 amino-acid protein contains three globular domains that might correspond to functional domains ([26] and Figure 4A). The first one is located at position 127-203 and exhibits strong structural similarities with chromodomains, that are chromatin targeting modules found in some regulators of chromatin structure (for a review, see [45]). The two others, located at positions 418-455 and 480-550, present no obvious similarities with known protein domains. In vitro translated ERK protein was retained on GST-C1/324 and GST-C325/550 beads containing the NH₂-terminal half and the COOH-terminal half of Corto, respectively (Figure 4A). In contrast, ERK was not retained on GST-C127/207 beads containing the Corto chromodomain, or on GST-C418/503 beads containing part of the two COOH-terminal globular domains. The lack of interaction with GST-C127/207 and GST-C418/503 suggested that none of these domains was sufficient to mediate Corto-ERK interaction, either because of inappropriate folding of these short domains in the GST fusion proteins, or because none of these two fragments contains the sequences that mediate ERK binding. Taken together, these results showed that Corto interacts directly with ERK in vitro. Further experiments are needed to determine the precise domains or residues that mediate the interaction between Corto and ERK.

Since we isolated dMP1 in a two-hybrid screen using the NH₂-terminal part of Corto as a bait (Figure 4B), we next questioned the physical interaction between Corto and dMP1. We performed GST pull-down assays using GST-dMP1 fusion protein and in vitro translated Corto to see whether their interaction was direct or indirect. Indeed, indirect interactions via yeast proteins have already been observed in two-hybrid experiments [46]. As shown in the middle panel of Figure 4C, the same result was obtained using GST or GST-dMP1 beads indicating that there was no specific direct interaction between Corto and dMP1. However, by incubating GST-dMP1 beads with total embryonic protein extract, we observed after blotting with anti-Corto antibodies that Corto was specifically retained on GST-dMP1 beads (bottom panel of Figure 4C). Therefore, we concluded that Corto and dMP1 interact via additional factors. One potential candidate could be ERK, since it directly interacts with Corto (as shown above) and with dMP1 [3].

Corto is located both in the cytoplasm and the nucleus and is phosphorylated

Upon activation of the MAPK cascade, ERK is phosphorylated in the cytoplasm. Di-phosphorylated ERK (dP-ERK) phosphorylates in turn a large number of targets with diverse functions and different subcellular localizations. In particular, part of dP-ERK is translocated into the nucleus where it phosphorylates some transcription factors (for reviews, see [1, 4]). Mammalian MP1 is present in the cytoplasm in association with ERK but its possible nuclear localization has not been reported [47, 48]. Nevertheless, we have recently shown that dMP1 is present both in the cytoplasm and the nucleus [3]. We thus asked whether Corto was present in the same compartments as ERK and dMP1. Corto was detected in nuclear and cytoplasmic extracts from embryos (Figure 5A), third instar larvae (data not shown) and Schneider S2 cells (Figure 5B). A similar nuclear and cytoplasmic distribution was observed for a Corto-FLAG fusion protein expressed in S2 cells (Figure 5C). Both Corto and Corto-FLAG exhibited several isoforms very close to each other in size, with the lowest isoforms being more abundant in the cytoplasm than in the nucleus (Figure 5A-C). Corto is very rich in serine (16%), threonine (4.5%), tyrosine (2.9%), and presents many predicted phosphorylation sites for several kinases distributed all along the sequence (according to NetPhos predictions: 28 phosphorylable serines, 6 threonines, and 3 tyrosines). Hence we checked whether the isoforms we observed could indeed correspond to differentially phosphorylated molecules. As shown in Figure 5D, the upper bands of Corto-FLAG disappeared after lambda phosphatase treatment. Altogether, these results showed that Corto is a phosphorylated protein present both in the cytoplasm and in the nucleus. In addition, the Corto phosphorylation pattern seemed to be different between the two cellular compartments, with enrichment in highly phosphorylated forms in the nucleus. Phosphorylation of M33, a chromatin regulator, has been shown to regulate its nuclear translocation [49]. Further experiments are required to determine if Corto localization is regulated through a similar mechanism.

The phosphorylation pattern of Corto is controlled at least partially by ERK pathway

Since Corto is phosphorylated and interacts in vitro with ERK, it is tempting to speculate that Corto could be phosphorylated in response to activation of the corresponding MAP kinase cascade. To answer this question, we transfected S2 cells in absence of serum with a plasmid allowing Corto-FLAG expression. After two days, we transiently activated the MAP kinase cascade by a short (15 minutes) treatment either with serum, with serum plus insulin or with serum plus PMA. In all three conditions, faint upper Corto isoforms appeared (brace on Figure 6A, upper panel), showing that Corto phosphorylation was induced very rapidly upon ERK activation. To confirm the existence of these faint upper isoforms, we immunoprecipitated Corto from these extracts using anti-FLAG antibodies and blotted the immunoprecipitates with antibodies directed against phosphoproteins (Figure 6A, lower panel). This experiment revealed first that Corto is constitutively phosphorylated even without MAP kinase activation. Second, ERK activation induced an hyperphosphorylation of Corto since an upper smear containing several bands very close in size appeared in all three conditions of activation tested.

As another way to activate the ERK pathway, we co-expressed Corto-FLAG and tagged forms of either ERK, ERK^Sem or Ras^V12 in S2 cells. Similar to ERK and ERK^Sem over-expression, over-expression in flies of a constitutively active form of Ras, Ras^V12, induces ectopic vein cells [50]. Surprisingly, smaller Corto isoforms appeared when constitutively activating the ERK pathway with either ERK, ERK^Sem or Ras^V12 (Figure 6B). One possibility is that these smaller isoforms could correspond to partially dephosphorylated Corto molecules. Taken together, these experiments demonstrate that Corto presents a complex phosphorylation pattern that depends at least on ERK signaling. It is tempting to speculate that some phosphorylations are performed directly by ERK. Indeed, Corto contains 3 SP sites at positions 139, 190 and 428 that correlate with theoretical ERK1/ERK2 phosphorylation sites [51]. Identification of Corto phosphorylation sites as well as phospho-mutant analysis and determination of Corto phosphorylation status when bound to chromatin would help to better understand the role of these phosphorylation events.

Interaction between ERK and dMP1 or Corto takes place in the nucleus only

We next performed co-immunoprecipitation experiments to see whether Corto interacts with ERK and dMP1 in vivo. We have previously shown that ERK and dMP1 co-immunoprecipitate in a total protein extract [3]. In order to determine the subcellular localization of this dMP1/ERK complex, we carried out co-immunoprecipitation experiments using cytoplasmic or nuclear extracts of S2 cells expressing dMP1 and ERK tagged proteins. Surprisingly, dMP1-Myc co-immunoprecipitated with ERK-FLAG in the nuclear extract only (Figure 7A). Corto-Myc and ERK-FLAG also co-immunoprecipitated in the nuclear extract only (Figure 7B). In this last experiment, all Corto isoforms were co-immunoprecipitated with ERK. In both cases, the lack of co-immunoprecipitation in cytoplasmic extracts was confirmed by using two different protocols to prepare nuclear and cytoplasmic extracts (see the Methods section). The co-immunoprecipitation observed between Corto and ERK fusion proteins was confirmed with endogenous proteins from a total embryonic extract using anti-dP-ERK antibody (Figure 7C). Furthermore, this experiment showed that Corto was able to interact with dP-ERK. Lastly, we observed no co-immunoprecipitation between Corto-FLAG and dMP1-Myc whether in total, cytoplasmic or nuclear extracts (data not shown). Labile protein interactions can be stabilized by cross-linking before performing cell extracts, although such treatment does not allow separating cytoplasmic and nuclear extracts. When using a total extract from cross-linked cells, we could detect co-immunoprecipitation between Corto-FLAG and dMP1-Myc (Figure 7D). This observation was consistent with our GST pull-down and peptide pull-down assays showing that the interaction between dMP1 and Corto was not direct but was probably mediated by other proteins. Altogether, our co-immunoprecipitation results suggest that a complex containing Corto, ERK and dMP1 might exist in the nucleus only. The core protein of this complex should be ERK, since it interacts directly with both Corto and dMP1.

ERK and dMP1 bind polytene chromosomes where they partially co-localize with Corto

To further analyze the relationship between Corto, dMP1 and ERK, and since Corto has been shown to bind polytene chromosomes [26], we analyzed the binding of ERK and dMP1 onto polytene chromosomes. We observed that ERK and dMP1 bound polytene chromosomes on many sites where they completely co-localized (Figure 8A). Furthermore, Corto and dMP1 co-localized on several sites (Figure 8B-C). These results suggest that a dMP1/ERK/Corto complex might regulate targets directly on chromatin. Since the co-localization of Corto with dMP1 is not complete, Corto appears to be an optional partner of dMP1/ERK on chromatin. Corto association may require additional factors or may be controlled by signaling events.

Previously, other scaffold proteins have been reported to bind chromatin. This is the case of the scaffold protein Ste5p in the pheromone pathway of yeast which interacts with the MAPKs Fus3p and Kss1p and occupies the same mating-type genes. Ste5p has then been suggested to function as an adaptor for protein-protein interactions both at the plasma membrane and in the nucleus [10]. In mammals, the scaffold protein β-arrestin, localized both in the cytoplasm and in the nucleus, is also recruited to target promoters under opioid receptor stimulation thus enhancing gene transcription [52]. The scaffold protein dMP1 could serve as an adaptor to connect ERK with other partners directly on chromatin. It could also allow ERK to form dimers, as mammalian scaffold proteins have been shown to be essential to connect ERK dimers to cytoplasmic targets [53]. It would therefore be interesting to know if ERK is monomeric or dimeric when bound to chromatin.

Drosophilastrains and genetic crosses

Flies were raised on standard yeast-cornmeal medium at 25°C. w ¹¹¹⁸ was used as control strain. The corto ^L1 (EMS-induced allele), corto ^07128b , bs ^EY23316 , rho ^EP3704 (P-insertion alleles) lines were from the Bloomington Stock Center. The corto ⁴²⁰ line results from imprecise P-element excision [24, 25]. The transgenic lines UAS::corto [24] (transgene on the third chromosome), UAS::rolled (transgene on the X chromosome) and UAS::rl ^Sem [54] (transgene on the third chromosome) were gifts from Dr. R. Rosset (UAS::corto) and Dr. K. Moses (UAS::rl and UAS::rl ^Sem ). The UAS::dsMP1 line allowing dMP1 down-regulation by RNA interference (transgene on the second chromosome) was described previously [3]. Lines containing a transgene with UAS sequences were crossed with the hs::Gal4, scalloped::Gal4 (sd::Gal4, [55]) and Beadex::Gal4 (Bx::Gal4, [56]) drivers. All crosses were performed at 25°C, except those of UAS::rl ^Sem with sd::Gal4 that were performed at 18°C to decrease Gal4 activity and therefore lower transgene expression. To perform staged corto expression with the hs::Gal4 driver, Gal4 was induced by 20 minute heat-shocks applied at various moments during larval and pupal development.

In situhybridization experiments on pupal wings

White pupae were collected and maintained for 30 h at 25°C. After puparium dissection, pupae were fixed in 8% formaldehyde for 12 h at 4°C. In situ hybridization with blistered (EST SD23611) and rhomboïd (EST RE59529) DIG-UTP labeled RNA probes was performed according to standard protocols [57].

Q-RT PCR experiments

Total RNA were extracted from 20 third instar larval discs of each genotype using the PureLink RNA Microkit (Invitrogen) according to the manufacturer's instructions. 1 μg of RNA was reverse-transcribed with the SuperScript^® VILO™ cDNA Synthesis Kit (Invitrogen). Q-RT PCR experiments were carried out on a Light Cycler 480 (Roche Diagnostics) with the Maxima SybrGreen mix (Fermentas). The primers used were: cortoF (5'-TGGCCACAGTTCCTAGCATT-3') and cortoR (5'-GCATGGGATTGGTGTCAGG-3'); rp49F (5'-CCGCTTCAAGGGACAGTATC-3') and rp49R (5'-GACAATCTCCTTGCGCTTC-3'); spt6F (5-'CGGAGGAGCTCTTCGATATG-3') and spt6R (5'-GACAGCTCTGGGAAGTCGTC-3'). A standard curve of amplification efficiency for each set of primers was generated with a serial dilution of cDNA. rp49 or spt6 levels were used for normalization according to the standard curve method. Three independent experiments were performed.

Plasmids and S2 cell transfection

The corto, dMP1, rl and rl ^Sem cDNAs were cloned into Gateway^® Drosophila vectors allowing expression of the fusion proteins under control of the actin5C promoter, as previously described [3]. The rl ^Sem sequence was obtained by in vitro mutagenesis using the QuickChange^® Site Directed Mutagenesis kit (Stratagene) according to the manufacturer's instructions. This gain-of-function mutation is a G to A transition resulting in a D to N substitution at position 334 [28]. pMT-Ras^V12 (a gift from Dr. A. Nagel) allowed transient expression of the constitutively active form of Ras, Ras^V12, under the control of the heavy metal inducible promoter metallothionein [58]. Ras^V12 was induced by treating transfected cells for 24 h with CuSO₄ 0.5 mM. For transfection, S2 cells were cultivated at 25°C in Schneider medium with or without 10% fetal calf serum as indicated. 5.10⁶ cells were transfected with 2 μg of DNA using Effecten^® transfection reagent (Qiagen) according to the manufacturer's instructions (1/10 DNA-Effecten^® ratio). Cells were collected 48 h (ERK pathway activation) or 72 h (immunoprecipitation) after transfection.

Protein extracts and phosphatase treatment

Embryos and S2 cell total extracts were prepared by sonication in RIPA buffer [50 mM Tris-HCl pH7.5, 150 mM NaCl, 25 mM NaVO₄, 25 mM NaF, 0.1% SDS, 0.5% NP40, complete protease inhibitors (Roche)]. Cytoplasmic extracts were prepared either with NE-PER Nuclear and Cytoplasmic Extraction Reagents (Pierce) according to the manufacturer's instructions or by homogenization using a Dounce potter in low salt buffer as described in [59]. When analyzed by immunoprecipitation, these two kinds of cytoplasmic extracts gave the same results. Nuclear extracts were obtained by sonication of the nuclear pellet in RIPA buffer. Phosphatase treatments of S2 cell extracts prepared without NaVO₄ and NaF phosphatase inhibitors were performed using Lambda Protein Phosphatase (Upstate) for a 10 minute incubation time at 37°C.

Western blot analysis and antibodies

Cell lysates or immunoprecipitated proteins were resolved on 8% SDS-PAGE Anderson gels [60] when separating different isoforms of Corto, or on 12% or 15% classical SDS-PAGE depending on the molecular weight of proteins. Western blot experiments were performed according to standard protocols. Antibodies used were monoclonal anti-FLAG (F3165, Sigma) or anti-Myc (sc-40, Santa Cruz Biotechnology) antibodies for fusion proteins, anti-β- tubulin (E7) and anti-lamin (ADL67.10) antibodies (Developmental Studies Hybridoma Bank) for control of cytoplasmic and nuclear fractions, rat anti-Corto antibodies [26], monoclonal phosphoserine/threonine/tyrosine antibody (MA1-38450, Pierce), rabbit anti-ERK antibodies (C-16, Santa-Cruz Biotechnology) or monoclonal anti-dP-ERK E10 antibody (9106, Cell Signaling). Anti-HA antibody (H3663, Sigma) was used as a negative control in immunoprecipitation experiments.

Protein-protein interactions

In vitro transcription-translation and GST pull-down assays were performed as previously described [26] using GST-Corto fusion proteins [61] and GST-dMP1 fusion protein [3].

For co-immunoprecipitation experiments, 500 μg of protein extracts (total, cytoplasmic or nuclear) were immunoprecipitated either with monoclonal anti-FLAG antibody, anti-Myc 9E10 antibody or anti-HA antibody using magnetic protein G-agarose beads (Ademtech). To co-immunoprecipitate Corto and dMP1, proteins were cross-linked before extraction by treating cells with 1% formaldehyde for 10 minutes followed by neutralization with 0.13 M glycine. To co-immunoprecipitate Corto and dP-ERK in embryonic extracts, monoclonal anti-dP-ERK E10 antibody was covalently bound onto protein G-agarose beads using standard protocols.

Two-hybrid experiments were performed as previously described [26], using leucine and X-Gal tests. Both tests gave the same result, and only the leucine test is shown. The full length dMP1 protein was fused with the B42 activation domain (B42AD). The NH₂-terminal half of Corto was fused with the LexA DNA binding domain. B42AD and LexA were used as negative controls. B42/SV40 large T-antigen (B42ADT) and LexA/p53 (LexA53) fusion proteins were used as positive controls.

Immunolocalization on polytene chromosomes

Co-immunostaining of w ¹¹¹⁸ polytene chromosomes was performed as previously described [26] using rabbit anti-ERK (1:20) (C-16; Santa-Cruz Biotechnology), guinea-pig anti-dMP1 (1:20) [3] and rabbit anti-Corto (1:20) [26] as primary antibodies. Secondary antibodies (Alexa Fluor^® 594 goat anti-rabbit IgG and Alexa Fluor^® 488 goat anti-guinea-pig IgG, Molecular Probes) were used at a 1:1000 dilution.