Lentivirus-mediated transduction of connexin cDNAs shows level- and isoform-specific alterations in insulin secretion of primary pancreaticβ-cells

Connexin (Cx) channels are the basic components of gap junctions, which, in vertebrates, are widely distributed among most types of cells. The primary function of these structures is to couple adjacent cells by allowing the exchange of low mass molecules including current-carrying ions, nucleotides and metabolites (Kumar and Gilula,1996; Willecke et al.,2002). Cxs form a multigene family of at least 20 proteins that share a highly conserved sequence and topography throughout the phylogenetic scale (Richard, 2000; Willecke et al., 2002). However, the molecular weight, unitary conductance of the resulting channel,voltage dependency and tissue distribution differ between individual Cxs(Harris and Bevans, 2001; Shibata et al., 2001; Willecke et al., 2002). Studies of spontaneous Cx mutations and of transgenic mice featuring altered expression of selected Cxs have indicated that these proteins, and/or the intercellular communication that they permit, contribute to the in vivo homeostasis of several tissues (Kelsell et al., 2001b; Willecke et al.,2002). They have further revealed that all Cxs can sustain certain common functions but that individual Cx isoforms also have specific functions that cannot be fulfilled by others, at least when the proteins are replaced after the very beginning of development(Plum et al., 2000; White, 2002).

It has not yet proved feasible to assess whether this selective versatility also applies to acute changes in the Cx pattern of adult cells, mostly because it is difficult to obtain rapid, stable transfection of a foreign cDNA in primary cells without losing the expression of the native differentiation genes, particularly if the target cells divide poorly(Anderson et al., 2002; Nielsen et al., 1989). To overcome these problems, we have taken advantage of novel lentiviral vectors which sustain expression of foreign genes in primary, fully differentiated and non-dividing cells (Gallichan et al.,1998; Ju et al.,1998; Naldini et al.,1996a; Naldini et al.,1996b; Salmon et al.,2000). Here, we show that these vectors can be designed to express efficiently a range of Cx-encoding cDNAs and to change both the number of gap junction plaques and the coupling extent of primary, fully differentiated and non-dividing cells.

In these experiments, we have used as a model the insulin-producingβ-cells of pancreatic islets. These endocrine micro-organs, which are mostly made up of β-cells, express Cx36(Serre-Beinier et al., 2000)but no Cx26 or Cx32, which are found in pancreatic acini(Meda et al., 1993). The distribution of Cx43 is less clear: it is expressed in some pancreatic vessels and fibroblasts (Theis et al.,2001) and possibly also by β-cells, at least under certain conditions (Collares-Buzato et al.,2001; Meda et al.,1993). This distribution suggests that specific Cx isoforms and the selective coupling that they mediate are required for the proper functioning of different pancreatic cell types. In particular, several lines of evidence suggest that Cx-mediated communication between β-cells is required for control of insulin secretion. In rats, sequential changes of pancreatic function were found to parallel changes in Cx36 expression(Calabrese et al., 2001). Expression of Cx32 in β-cells of transgenic mice reduced insulin secretion in response to physiological concentrations of glucose, indicating that the in vivo expression of a Cx, which is exogenous to pancreatic islets,perturbed b-cell functioning, in spite of the persistence of the native Cx36 and increased cell-to-cell coupling(Charollais et al., 2000).

To assess whether the type and level of Cxs affect insulin secretion differently, we tested the expression of distinct Cx cDNAs in primaryβ-cells. Because these experiments are not feasible using standard transfection methods owing to the very low proliferation rate of fully differentiated islet cells (Nielsen et al., 1989), we transduced the cells with Cx-encoding lentiviral vectors. For these experiments, we did not choose a standard preparation of isolated pancreatic islets, because the central β-cell-rich core of these micro-organs undergoes rapid necrotic changes during culture times shorter than those required for transduction and analysis of gene expression(Ono et al., 1979). Furthermore, infection of isolated islets has so far resulted in a variable cDNA targeting of those β-cells, which are located in the centre of the islets (Curran et al., 2002; Gallichan et al., 1998; Giannoukakis et al., 1999; Ju et al., 1998; Salmon et al., 2000) and are responsible for most of the acutely induced secretion(Stefan et al., 1987). To bypass these limitations, still preserving the three-dimensional (3D) context that is required for proper function of contacting islet cells(Halban et al., 1982; Hauge-Evans et al., 1999; Hopcroft et al., 1985), we studied pseudo-islets formed by the spontaneous reaggregation of cells that had been previously transduced in suspension. We found that the glucose-induced insulin secretion of pseudo-islets was altered depending on the type of transduced Cx and its actual level of expression.

Islets of Langerhans were isolated by collagenase digestion of pancreases from male Sprague-Dawley rats (weighing 250-350 g) as previously described(Vozzi et al., 1995). For cell preparation, the isolated islets were rinsed three times in a Mg²⁺-and Ca²⁺-free phosphate buffered saline (PBS) and resuspended in 1.5 ml of this buffer containing 0.16% trypsin (w/v) (Life technologies,Scotland) and 0.0066% EDTA (w/v). Digestion (under repeated pipetting) was carried out for 7 minutes at 37°C and was stopped by addition of 10 ml ice-cold culture medium, described below. Cells were washed three times in 15 ml sterile culture medium, before plating for viral infection. For pseudo-islet formation, cells were seeded into 60-mm dishes to which cells do not adhere at an initial density of 10⁵ cells ml^–1. Pseudo-islets were harvested after 5 days of static culture at 37°C (Halban et al.,1987).

RIN2A cells (Gazdar et al.,1980) were cultured in RPMI 1640 medium containing 10% foetal calf serum (FCS), 110 U ml^–1 penicillin, 110 μg ml^–1 streptomycin and 2 mM L-glutamine.

Primary β-cells, intact islets and pseudo-islets were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 11.2 mM glucose, 10%heat-inactivated FCS, 2 mM L-glutamine, 110 U ml^–1 penicillin and 110 μg ml^–1 streptomycin.

293T cells were also cultured in the DMEM medium described above, except that the glucose concentration was 25 mM(Naldini et al., 1996a). All cells were kept at 37°C in a humidified incubator, gassed with air and CO₂ to maintain medium pH at 7.4.

Three self-inactivating HIV-derived gene-transfer plasmids(pHR′-CMV-Cx-W-sin18) were constructed by inserting the cDNA for rat Cx32, Cx36 or Cx43 downstream of the cytomegalovirus (CMV) promoter elements and upstream of the post-transcriptional regulatory element of woodchuck hepatitis virus (W) (Fig. 1A). To express Cx43, plasmid (pHR′-CMV-Cx43-W-sin18) was made by digesting pHR′-CMV-GFP-W-sin18 (Zufferey et al., 1999) with BamHI and XhoI, and inserting rat Cx43 cDNA (Beyer et al.,1987) in place of the green fluorescent protein (GFP) cDNA. To express Cx36, plasmid (pHR′-CMV-Cx36-W-sin18) was constructed by digesting pHR′-CMV-Cx43-W-sin18 with SnaBI and XhoI,and inserting the rat Cx36 cDNA(Condorelli et al., 1998) from pcDNA 3.1 (Invitrogen, The Netherlands) in place of the Cx43 cDNA. To express Cx32, plasmid (pHR′-CMV-Cx32-W-sin18) was constructed by partially digesting pHR′-CMV-Cx43-W-sin18 with EcoRI and inserting the rat Cx32 cDNA (Paul, 1986) in place of the Cx43 cDNA.

Vectors were prepared by co-transfecting 293T cells with the following three plasmids at a time (Naldini et al.,1996b): the three pHR′-CMV-Cx-W-sin18 (or the original pHR′-CMV-GFP-W-sin18); a second-generation packaging plasmid p8.91(Zufferey et al., 1997); a VSV-G envelope-protein-expression plasmid pMDG(Naldini et al., 1996b). After an overnight incubation in the presence of the transfection precipitate, the culture medium was changed. On the following day, the medium of the transfected cells was harvested, filtered through a polyethersulfone membrane(0.45 μm pores) and stored in 1-10 ml aliquots at –80°C. The concentration of the vector stock [transducing infectious units (TIU)] was determined by adding aliquots of vector on monolayers of RIN2A cells and by assessing: (1) the percentage of GFP-positive cells by fluorescence-activated cell sorting (FACS) using a Beckton Dickinson FACScan(Fig. 1B) as per the guidelines discussed by Klages et al. (Klages et al.,2000); (2) the immunofluorescence labelling (between neighboring cell pairs) of different Cxs 48 hours after the infection.

Infection was carried out by adding the vectors to either established monolayer cultures (for dye-coupling experiments) or single-cell suspensions(for experiments involving pseudo-islets). In both cases, the infection was performed the day following cell isolation and lasted 24 hours. Based on the vector concentration determined on RIN2A cells (see above),4×10⁵ TIUs were used to infect 10⁵ primary islet cells with vectors encoding GFP, Cx32, Cx36 or Cx43. After several rinsing periods, infected cells were allowed to reaggregate for 5 days. As control,uninfected cell batches were cultured for pseudo-islet formation, in parallel with the transduced cells. Thus, each experiment compared pseudo-islets made of a single original batch of cells from which some aliquots were not infected whereas others were transduced for GFP, Cx32, Cx36 or Cx43.

For indirect immunofluorescence, pancreatic β-cells were attached to coverslips coated with matrix 804G (Bosco et al., 2000) and exposed for 3 minutes to acetone at–20°C. All cells were first rinsed in cold (4°C) PBS, blocked for 30 minutes with PBS supplemented with 2% bovine serum albumin (BSA) and incubated for 2 hours at room temperature with one of the following antibodies: (1) polyclonal rabbit antibodies against Cx43, diluted 1:500(Zymed Laboratories, South San Francisco, CA); (2) polyclonal rabbit antiserum against rat Cx32, diluted 1:400(Dermietzel et al., 1984); and(3) polyclonal rabbit antiserum against rat Cx36, diluted 1:200(Serre-Beinier et al., 2000). Pseudo-islets were processed in the same way except that the fixation was performed for 1 hour with 100% ethanol at –20°C, and that the incubation with the anti-Cx antibodies was run overnight at room temperature.

Cells and pseudo-islets were then rinsed in PBS and incubated for 1 hour at room temperature with fluorescein-conjugated antibodies against rabbit Ig_s, diluted 1:500. After further rinsing, sections were stained with a 0.03% Evans' blue solution (which gives a red background staining when viewed by fluorescence using filters for fluorescein detection), covered with 0.02% paraphenylenediamine in PBS-glycerol (1:2 vol:vol) and photographed using filters for fluorescein detection with either an Axioplan microscope(Zeiss, Germany) or a confocal scanning laser microscope (LSM 510, Zeiss,Germany) equipped with a 30 mW ArKr and a 1 mW HeNe laser, and a 40×inverted objective. Optical scans were continuously collected at a scan speed of 7.2 seconds per image. Section planes were collected in 2 μm steps over a total thickness of about 40-50 μm, at both 488 nm and 543 nm excitation wavelengths. The focus, contrast and brightness settings were kept constant during image acquisition. For 3D reconstruction and analysis, images were processed for surface and alpha rendering and arranged in a movie sequence,using the LSM 510 software (Zeiss, Germany).

For extraction of total proteins, cell cultures and control tissues were homogenized by sonication in 0.1 M Tris-HCl, pH 7.4, supplemented with 20 mM EDTA, 1 μg ml^–1 pepstatin A, 1 μg ml^–1antipain, 1 mM benzamidine, 4.5 Tiu ml^–1 (Tiu=trypsin inhibitor unit) aprotinin, 2 mM PMSF and 1 mM DFP, and stored at–20°C. For extraction of membrane proteins, the sonicate was centrifuged for 10 minutes at 3000 g and 4°C, the supernatant was collected and centrifuged for 60 minutes at 100,000 g and 4°C. Pelleted material was resuspended in PBS and solubilized in 0.1 M Tris-HCl containing 10 mM EDTA and 20% SDS, and stored at–20°C. Protein content was measured by the DC protein assay kit(Bio-Rad Laboratories). Aliquots of membrane or total proteins were fractionated by electrophoresis in a 12% polyacrylamide gel and either stained with Coomassie blue or immunoblotted, as previously described.

To this end, electrophoresed samples were transferred onto PVDF membranes(Immobilon™-P, Millipore) for 2 hour in the presence of 0.01% SDS and 20% methanol, using a constant current of 400 mA. Thereafter, the membranes were saturated for 30 minutes at room temperature in a buffer containing 10 mM Tris (pH 7.4), 150 mM NaCl, 0.1% Tween 20 (TBS-Tween) and 5% dry milk, and incubated overnight at 4°C with one of the following antibodies: (i) mouse monoclonal antibodies against Cx43 (Zymed Laboratories, South San Francisco,CA), diluted 1:500; (ii) polyclonal rabbit antiserum against rat Cx32, diluted 1:1000 (Dermietzel et al.,1984); (iii) polyclonal rabbit antiserum against rat Cx36(Serre-Beinier et al., 2000),diluted 1:400. After repeated rinsing in TBS-Tween, the immunoblots were incubated for 60 minutes at room temperature with a goat serum against either mouse or rabbit Ig_s and conjugated to horseradish peroxidase(Bio-Rad Laboratories), diluted 1:6000. Membranes were then washed and developed by enhanced chemiluminescence using kit ECL™ (Amersham Pharmacia Biotech) according to manufacturer's instructions.

For assessment of junctional coupling, individual cells were microinjected by iontophoresis (Meda, 2001)within monolayer cultures, using microelectrodes containing either 4% Lucifer Yellow CH (Sigma Chemical, St Louis, MO) or a mixture of 5% Neurobiotin(Vector Laboratories) and 0.4% rhodamine 3-isothiocyanate dextran 10S (Sigma)in 150 mM LiCl. After injection of neurobiotin, cells were fixed in 4%paraformaldehyde in phosphate buffer (0.1 M, pH 7.4) for 20 minutes, washed,incubated in 0.25% Triton X-100 in PBS for 30 minutes, washed again and incubated with fluorescein-conjugated streptavidin (Jackson ImmunoResearch Laboratories), diluted 1:400 for 60 minutes.

The percentage of microinjected cells that exhibited cell-to-cell transfer of Lucifer Yellow or Neurobiotin, as well as the order of dye transfer (first order means cells contacting the microinjected cell; second and third orders means cells distant from the microinjected cell by one or two cell diameters,respectively) were determined on photographs taken either immediately after each microinjection (Lucifer Yellow) or after the streptavidin incubation(Neurobiotin).

For assessment of gap junction plaques, cells were fixed for 60 minutes in a 2.5% glutaraldehyde solution in 0.1 M phosphate buffer, infiltrated with 30%glycerol, frozen in Freon22 cooled in liquid nitrogen and processed for freeze-fracture using a Balzers BAF (Balzers High Vacuum, Balzers,Liechtenstein) (Vozzi et al.,1995). Replicas were examined in a Philips EM 300 microscope.

To assess the secretion of insulin, batches of 50 pseudo-islets were transferred to the chambers of a Brandel Suprafusion System (Brandel, USA) and perfused at 37°C and at a flow rate of 500 μl min^–1with a HEPES (10 mM)-buffered Krebs-Ringer buffer, pH 7.4, containing 0.1% BSA(KRB). Each chamber was initially perfused with KRB containing 2.8 mM glucose for 30 minutes, during which period the outflow was discarded. Pseudo-islets were then perfused for 5 consecutive periods of 20 minutes each with KRB containing increasing concentrations of glucose (2.8-16.8 mM) and eventually 1 mM isobutylmethylxanthine (IBMX) plus 5 μM forskolin (FSK). Insulin secreted from the pseudo-islets into the perfusate was collected in 1 ml aliquots. Insulin content of each aliquot was measured by radioimmunoassay(RIA) with a charcoal separation step, using rat insulin as standard and a guinea pig anti-rat-insulin (Linco Research) as antibody(Meda et al., 1990).

At the end of the perfusion, the pseudo-islets were collected from each chamber and extracted in acid ethanol for determination of insulin content. Each experiments tested in parallel and simultaneously the secretory response of native pancreatic islets (not shown) and of five groups of pseudo-islets made of uninfected and infected cells (GFP-, Cx32-, Cx36- and Cx43-transduced), respectively.

Data were expressed as mean ± s.e.m. Differences between means were assessed by Student's t-test and considered significant when P<0.05.

The insulin-producing line of RIN2A cells was used to assess the efficiency of the lentivirus-mediated gene expression. Using a vector coding for GFP, we found that ∼80% of the FACS-screened cells were strongly fluorescent two days after infection (Fig. 1B). After transduction with the same vector containing a Cx cDNA, immunostaining revealed a robust expression of either Cx32, Cx36 or Cx43 in most cells, and the expected localization of these proteins at membrane interfaces between adjacent cells (Fig. 1C). Consistent with the lack of Cx on immunoblots of membranes from wild-type RIN2A cells (Fig. 1D),freeze-fracture electron microscopy failed to detect gap junction plaques between these cells (not shown). By contrast, normal gap-junction plaques were seen between Cx-transduced cells (Fig. 1E), in agreement with the expression of high levels of transduced proteins, which had molecular weights similar to those of the native connexins(Fig. 1D).

Monolayer cultures of primary islet cells transduced using the same protocol applied to RIN2A cells revealed that ∼60% of the cells featured a green fluorescence two days after infection with a GFP-coding vector. When the vectors contained a cDNA coding for Cx32, Cx36 or Cx43, the cognate protein was easily detected by immunostaining in both the membrane (where immunolabelling was consistently intense) and the cytoplasm (where immunolabelling was of variable intensity) of most islet cells(Fig. 2). After microinjection of gap junction tracers, cells infected with a lentivirus encoding Cx showed dye coupling more frequently than uninfected wild-type controls(Fig. 3A,B). The extent of this coupling varied depending on the tracer used and the Cx type transduced. Thus,the intercellular exchange of Lucifer Yellow was markedly increased by over-expression of Cx32, less by the expression of Cx43 and not by that of Cx36 (Fig. 3A). However, the latter transduced Cx isoform led to a sizable increase in the exchange of neurobiotin between islet cells (Fig. 3B).

Suspensions of single wild-type islet cells spontaneously aggregate in vitro into 3D pseudo-islets which resemble native pancreatic islets(Fig. 4A). This spontaneous behaviour was not altered by transduction with lentiviral vectors coding for GFP and the pseudo-islets observed 5 days after infection were made mostly by cells that still expressed the transduced cDNA(Fig. 4A,C). Islet cells over-expressing transduced Cx32, Cx36 and Cx43 also formed pseudo-islets of standard size (Fig. 4B,C). Minute gap junctional plaques were detected by freeze-fracture electron microscopy at membrane interfaces between the uninfected β-cells of control pseudo-islets (Fig. 5A). When the micro-organs were formed by islet cells transduced with a lentiviral vector coding for Cx32, Cx36 or Cx43, unusual large plaques were revealed (Fig. 5A), in agreement with the expression of increased levels of transduced proteins observed by immunostaining and immunoblotting(Fig. 4B, Fig. 5B).

Total insulin content of GFP-transduced pseudo-islets (1832±403 ng per chamber, n=4) was similar to that of uninfected pseudo-islets(1683±371 ng per chamber, n=5)(Fig. 6B). Uninfected pseudo-islets increased (P<0.05) insulin secretion twofold over the basal level observed in the presence of 5.6 mM glucose, when the concentration of the sugar was raised to 16.8 mM glucose(Fig. 6A,C). Addition to the sugar of 1 mM IBMX and 5 μM FSK further potentiated the insulin output,which was then increased eightfold (P<0.05) over that observed under basal conditions (Fig. 6A,C). Similar results were obtained with pseudo-islets made of GFP-transduced cells (Fig. 6A,C). Thus, when compared with pseudo-islets made of uninfected cells, pseudo-islets made of GFP-transduced cells did not show any alteration in insulin secretion, whether the hormone release was assessed after stimulation by glucose or by drugs that potentiate the sugar effect by raising the cytoplasmic concentration of cAMP (Fig. 6A).

Total insulin content of Cx-transduced pseudoislets (1480±263 ng per chamber, n=12; data from all Cx-transduced cells pooled) was similar to that of control pseudo-islets (1749±411 ng per chamber, n=9; data from GFP-transduced and uninfected cells pooled)(Fig. 7A). Pseudo-islets made of Cx43-transduced cells also featured a control secretory response (i.e. increasing secretion in the presence of 16.8 mM glucose)(Fig. 7B). By contrast,pseudo-islets made of cells transduced with either Cx32 or Cx36 failed to increase significantly their insulin release over the basal level when stimulated by 16.8 mM glucose (Fig. 7B). Under this condition, their insulin output was 48.5±5%(Cx32, n=3, P<0.005) and 51±5% (Cx36, n=3, P<0.005) the control values (data of GFP-transduced and uninfected pseudo-islets pooled), respectively(Fig. 7C). All Cx-transduced pseudo-islets significantly increased their insulin release after stimulation by glucose plus IBMX and FSK (Fig. 7B). However, this increase was smaller than that of controls,representing 58±3%, 67±17% and 71±16% of the normalised control value, in pseudo-islets transduced for Cx32 (n=3; P<0.05), Cx36 (n=3) and Cx43 (n=3)(Fig. 7C).

Previous studies have implicated Cx-mediated cell-to-cell coupling inβ-cell function (Bosco et al.,1995; Calabrese et al.,2001; Cao et al.,1997; Vozzi et al.,1995) and have suggested that adequate levels of specific Cx isoforms is required for the control of insulin secretion of permanent cell lines (Calabrese et al., 2001; Cao et al., 1997; Vozzi et al., 1995). As yet,however, it has not proved feasible to test this hypothesis directly in normal pancreatic β-cells, mostly because these highly differentiated cells proliferate poorly (Nielsen et al.,1989) and hence are not amenable to the stable expression of exogenous cDNAs by standard transfection procedures. Alternative approaches,which take advantage of viral vectors that can infect pancreatic islets, have been shown to be an effective solution to this problem(Curran et al., 2002; Gallichan et al., 1998; Leibowitz et al., 1999; Salmon et al., 2000). However,these approaches are still limited by the poor viability of β-cells within cultured islets (Ferguson et al.,1976; Ono et al.,1979) and by the variable degree of gene expression that the vectors induce in individual cells (Curran et al., 2002; Gallichan et al., 1998; Leibowitz et al.,1999; Salmon et al.,2000).

To bypass these difficulties, we have designed a novel generation of lentiviral vectors to transduce distinct Cx cDNAs in cells of the insulin-producing RIN2A line (Gazdar et al., 1980) which, in the wild-type configuration, do not transcribe Cx genes (Vozzi et al.,1995). Here, we document that these vectors efficiently upregulated the expression of Cx proteins in RIN2A cells and that the transduced connexins had a normal molecular weight and epitope structure. We further show that the transduced proteins were appropriately targeted to the membrane and became normally concentrated at sites of cell-to-cell contact,ultimately resulting in the development of bona fide gap junction plaques.

In a second step, we have shown that the same lentiviral vectors also efficiently transduced Cx cDNAs in primary islet cells, allowing over-expression of the cognate proteins. These proteins formed channels that significantly increased the exchange of different gap-junction-permeant tracers between β-cells. The degree of coupling, however, varied according to the gap junction tracer used and depended on both the type and level of the Cx expressed. Thus, exchange of Lucifer Yellow [a negatively charged molecule whose hydrated diameter approximates that of the Cx pore(Harris and Bevans, 2001)] was observed to spread up to three orders of islet cells after transduction of Cx32 and Cx43, whereas it was consistently restricted to the first order of cells in control cultures and after over-expression of Cx36. Moreover, the exchange of neurobiotin, a positively charged molecule that is smaller than Lucifer Yellow (Harris and Bevans,2001) and spreads as little as this tracer between control islet cells, increased up to three orders of cells after transduction of Cx36. Thus,over-expression of distinct Cxs provided increased islet cell coupling whichever connexin isoform was over-expressed, and led to a change in the specificity of the molecular transfer when Cx32 or Cx43 were involved. Hence,the lentiviral vectors we have generated open new perspectives for investigating Cx-dependent effects in primary cells that are not easily amenable to the experimental expression of foreign genes. Indeed, these vectors govern the stable transduction of normal cells, irrespective of their type and cycling status (Naldini et al.,1996b) and might be designed for both over-expression and downregulation of Cx proteins.

We have used these vectors to study the effect of various Cx changes on the main physiological function of β-cells, which is to synthesize, store and release insulin in a regulated way. Because these processes are multicellular functions that take place in a 3D context, we have first tested the transduced cells for their ability to reaggregate spontaneously into pseudo-islets, which feature morphological and functional characteristics reminiscent of those of native pancreatic islets (Halban et al.,1987; Hopcroft et al.,1985). We have found that transduction of a Cx-unrelated protein(GFP) did not impair the spontaneous ability of islet cells to aggregate into glucose-sensitive pseudo-islets. We have further observed that the functioning of these micro-organs which, under control conditions, expressed low levels of the islet native Cx36, was not affected by over-expression of Cx43, a protein whose presence in pancreatic islets is disputed, at least in control rodents(Charollais et al., 1999; Collares-Buzato et al., 2001; Meda et al., 1993; Theis et al., 2001). By contrast, we have observed that over-expression of either Cx32 or Cx36 reduced the response of pseudo-islets to glucose. Thus, in spite of a control content of insulin, the latter pseudo-islets were unable to enhance the release of the hormone in response to an increase in the glucose concentration of the perfusate. All types of Cx-over-expressing pseudo-islets significantly increased their insulin secretion over basal values in response to stimuli raising the intracellular concentration of cAMP(Gillis and Misler, 1993),even though somewhat less than controls. These findings show that the loss of glucose responsiveness observed in the same pseudo-islets cannot be accounted for by a nonspecific inhibition of the secretory machinery ofβ-cells.

Because these secretory alterations were paralleled by changes in the intercellular transfer of molecules that had a size and charge comparable to those of many endogenous metabolites(Bevans et al., 1998; Harris and Bevans, 2001; Veenstra et al., 1995), it is plausible that they resulted from the abnormal β-cell-to-β-cell exchange of some gap-junction-permeant signal. Ca²⁺ is a probable candidate, because this cation plays an important role in insulin secretion,diffuses through junctional channels and features abnormal stimuli-induced oscillations after Cx alterations(Charollais et al.,2000; Calabrese et al., 2003). Accordingly, the secretory defect we observed in this study is reminiscent of that observed in transgenic mice whose β-cells over-expressed Cx32. Under these conditions, the induced Cx32 channels improved the electrical synchronisation of β-cells and the glucose-induced increase in the intracellular concentration of Ca²⁺(Charollais et al., 2000). In spite of these changes, that would be expected to promote insulin release(Bergsten, 2000; Kanno et al., 2002; Satin, 2000), the secretion of the hormone was not stimulated by glucose, indicating that the signal(s) whose exchange had been increased between β-cells negatively affected secretion(Charollais et al., 2000).

Even though the nature of this signal remains to be elucidated, our data provide evidence that it does not transfer similarly from cell-to-cell whatever the Cx involved. Specifically, the data show that, in this respect,Cx32 channels cannot substitute the native Cx36 channel. In this perspective,it remains to be understood why over-expression of Cx36 also resulted in impaired glucose-induced insulin secretion. Because this over-expression increased the cell-to-cell exchange of some (e.g. neurobiotin) but not all gap-junction-permeant molecules (e.g. Lucifer Yellow), which is consistently restricted to small territories of β-cells within native islets(Charollais et al., 2000; Michaels and Sheridan, 1981),it is conceivable that extended coupling caused an excessive dilution of the Cx-dependent signals that positively control secretion. However, our data show that, even in this case, the specific characteristics of individual Cxs matter, in as much as no deleterious secretory effect was observed after transduction of Cx43, in spite of the fact that this protein increased coupling up to the levels elicited by the Cx32 and Cx36 transduction. The findings are consistent with the different molecular permeability of channels made by distinct Cx isoforms (Bevans et al., 1998; Goldberg et al.,2002). They imply that the signals diffusing through Cx43 channels are required for the cAMP-dependent potentiation of insulin release but not for the glucose stimulation of this event. Conversely, signals diffusing through either Cx32 or Cx36 channels contribute to modulate the two forms of insulin secretion. Additional experiments, in which different levels of specific Cx isoforms are compared in the absence of other connexins, will now be required unambiguously to establish the relative contributions to this fine tuning of the amount of each Cx isoform, on the one hand, and of its specific conductance and permeability characteristics, on the other. Addressing this important question might help us to identify the still-elusive gap-junction-permeant signals that influence insulin secretion(Meda and Spray, 2000;Serre-Beinier et al., 2003). As a first approach to this central issue, we have recently shown an essential role of Cx36 channels in synchronizing stimulus-induced Ca²⁺ oscillations between the insulin-producing cells of the MIN6 line (Calabrese et al.,2003).

In summary, our study first shows that lentiviral vectors are efficient tools to modulate the levels of Cxs and coupling in primary, non-dividing cells. With the availability of these tools, several questions about the(patho)physiological function(s) of gap junction proteins(Meda and Spray, 2000) can now be taken to direct experimental testing in primary tissues. In this perspective, we document that appropriate levels of specific Cx isoforms selectively influence distinct aspects of insulin secretion of primary pancreatic β-cells. Lentiviral vectors have also been shown to have some potential for gene therapy (Gallichan et al., 1998; Kordower et al.,2000). Therefore, their usefulness in a future correction of hereditary, Cx-linked diseases (Kelsell et al., 2001a; Kelsell et al.,2001b) should be envisaged.

We thank J. Bauquis, J. Cancela and P. Severi-De Marco for excellent technical assistance, P. Salmon for help with the lentivirus production and O. Brun for help with confocal data acquisition. Work by the Meda team is supported by grants from the Swiss National Foundation (3100-067788.02), the Fondation Romande pour la Recherche sur le Diabete, the Juvenile Diabetes Research Foundation International (1-2001-622), the European Union(QLK3-CT-2002-01777) and the National Institute of Health (DK 63443-01). D. Trono and R. Zufferey were supported by the Juvenile Diabetes Research Foundation International and the Swiss National Science Foundation.