Co-expression of insulin and somatostatin genes in pancreatic endocrine cells selected for their high level of insulin gene transcription

There are several insulin-producing cell lines that are used as models for the study of the regulation of insulin synthesis. The HIT cell line is derived from a primary hamster islet cell culture transformed by SV40 (Santerre et al. 1981). More recently another cell line has become available, this being derived from a tumor induced in transgenic mice by expressing the SV40 T antigen under the control of the insulin gene regulatory sequences (Efrat et al. 1988). In both lines T antigen is synthesized, and this may alter the normal transcription of the insulin genes. This problem makes the RIN cell line more attractive, since these cells come from a rat tumor induced by X-rays (Chick et al. 1977). However, clonal selection has shown that RIN cells spontaneously differentiate into any one of the pancreatic endocrine cell types (Gadzar et al. 1980; Nielsen et al. 1985; Philippe et al. 1987). This feature is therefore a problem in the study of gene regulation that is specific to βcells.

To select cells expressing the insulin gene and counterselect those in which it is transcriptionally inactive, a RIN cell line was transfected with a gene coding for a dominant selective marker, the gpt gene, which codes for the bacterial xanthine-guanine phosphoribosyltransferase (XGPRT), placed under the control of the insulin gene regulatory sequences; the analysis of several selected recombinant cell lines had previously led to the conclusion that a pure βcell line could be isolated in this way (Besnard et al. 1991).

We have analysed the pattern of cell types present in such a recombinant line by immunocytochemistry. We found that, as for the RIN wild-type cells, the cells of the selected line form a heterogeneous population. In addition they have a large proportion of somatostatinproducing cells, some of which also synthesize insulin.

Rabbit anti-glucagon, rabbit anti-pancreatic polypeptide and rabbit anti-somatostatin (SRIF 16) sera were gifts from Dr Garaud (INSERM, Strasbourg), Dr Chance (Lilly Research Laboratories, USA) and Dr Dubrasquet (INSERM, Paris), respectively. The following reagents were purchased: guinea pig anti-insulin serum from Chemical Credential (no. 65–104); rabbit anti-guinea pig peroxidase-labelled immunoglobulins (P141) from DAKO; goat anti-rabbit peroxidase-labelled immunoglobulin (04–15–06); goat anti-guinea pig fluorescein-labelled serum (02–17–06) and goat anti-rabbit rhodamin-labelled serum (03-15-06) from KPL; auroprobes EM antiguinea pig (Ga Gp G15) and anti-rabbit (GAR G10) from Janssen.

The RIN2A cell line is a clonal descendant of the predominantly insulin-producing RIN-5F line (Praz et al. 1983) that was used to obtain an HGPRT^- (RTN2A) cell line, (a gift from Dr J. Jami). This line was used to select cells stably transformed by a chimeric insulin gene, constructed by replacing the SV40 Pvidl-Hmdlll enhancer-containing fragment of pSV2gpt that codes for XGPRT (Mulligan and Berg, 1980) with the 5’ region (−879 to +241) of the human insulin gene (Bell et al. 1980). This region contains all the regulatory sequences necessary for /J-cell-specific expression in transgenic mice (Fromont-Racine et al. 1990). The details of the construction, the conditions of transfection and selection as well as the characteristics of the cell have been published (Besnard et al. 1991). The RW line (a gift from Dr D. Daegelen) is the recombinant line that synthesized the highest level of insulin mRNA. The cells were cultivated in Dulbecco’s modified Eagle’s medium complemented with 8% foetal bovine serum, antibiotics and either with (RW) or without (RIN2A) 80 μM mycophenolic acid.

Single indirect immunoperoxidase labels for the four types of pancreatic hormones were applied to both RIN2A and RW lines. Cells were grown in culture dishes and fixed in situ at room temperature in 3% paraformaldehyde/phosphate buffer for 3 min. Immunolabelling was performed on Triton X-100 permeabilized cells by successive incubations with normal goat serum 1:20 (30 min), one of the specific antisera (60 min) and rinsing with PBS before coupling (60 min) with peroxidase-labelled anti-guinea pig (for insulin) or anti-rabbit serum (for glucagon, pancreatic polypeptide and somatostatin). Peroxidase activity was revealed with α-chloronaphthol (Van Noorden and Polak, 1983). Counting of the four cell types was performed independantly on three fields of 1000 cells each.

Double-staining experiments were used to reveal the concomitant presence of insulin and somatostatin in the recombinant RW cells. Cells were grown on sterilized coverslips placed in the culture dishes and treated as previously up to the specific sera (a mixture of guinea pig antiinsulin and rabbit anti-somatostatin antisera) addition step. Labelling was performed in a mixture of fluorescein isothiocyanate (FTTC)-coupled anti-guinea pig antiserum and tetra-methyl-rhodamine isothiocyanate (TRITC)-coupled anti-rabbit antiserum. The specificity of the labelling was first established by using exactly the same protocol on sections of rat pancreas. Controls were performed with absorbed specific sera or by omitting the specific sera. After washing, the coverslips were mounted in Mowiol (Calbiochem no. 475904) on microscope slides and observed with an epifluorescence microscope using a 100 W mercury burner for excitation and selective interference filters Leitz 13 for FITC and N2 for TR1TC. Photographs were taken with a Kodak TMAX film 400 ASA.

The cells were fixed in situ in 1.5% glutaraldehyde 0.1 M phosphate buffer, pH 7.3, for 20 min at room temperature postfixed in 1% osmium tetroxide and embedded in Epon. Areas of the culture were cut and sections collected on 300 mesh nickel grids. They were either directly counterstained with uranyl acetate and lead citrate’ or first processed for immunocytochemistry by etching with water-saturated sodium metaperiodate (Bendayan and Zollinger, 1983). All the samples were then preincubated in normal goat serum before successive incubation in anti-insulin guinea pig serum followed by anti-guinea pig (15 nm) auroprobe and then antisomatostatin rabbit serum followed by anti-rabbit (10 nm) auroprobe. Controls were performed as described for light microscopy. Observations were made with a Philips EM410.

The recombinant RW cell line was analysed for cell types and compared with the parental (RIN2A) cell line. Results after single immunoperoxidase labelling are presented in Table 1. The RIN2A cell population is heterogeneous with four reactive cell types: insulin (56.5%), somatostatin (4.1%), glucagon (0.2%) and pancreatic polypeptide (0.3%). The remaining cells are unreactive (38.9%). The distribution of the immunoreactive material is different according to the hormone: insulin is mostly present in a small crescent-shaped reactive area of the cytoplasm while somatostatin is more dispersed throughout the cytoplasm (Fig. 1A and B). The recombinant RW cell population remains heterogeneous. There is no change in the proportion of glucagon and pancreatic polypeptide cells, and a small enhancement (about 10%) of the proportion of insulin cells (62%). However, in contrast to RIN2A there is a large proportion (28%) of cells containing somatostatin associated with an apparently strong reduction in the number of non-reactive cells (9.5%). Because the counts of cells containing the four hormones were performed on different samples, it is notz possible to distinguish between two possibilities: (a) that the somatostatin and insulin cells are distinct entities and thus there are about 10% of nonreactive cells; or (b) that some or all somatostatin cells also synthesize insulin, in which case the number of nonreactive cells would be higher in relation to the number of cells containing both hormones, up to the level found in the parental line.

To determine if the somatostatin-producing cells are part of the insulin-producing cell population, the cells were grown at low density to distinguish individual cells better, and processed for immunofluorescence doublelabelling using anti-insulin and anti-somatostatin antisera. The results show that there are cells that contain only one hormone, either insulin or somatostatin, and cells that contain both insulin and somatostatin at various relative concentrations (Fig. 1 C to F). However, determination of the exact number of cells containing both hormones is technically difficult. First, excitation by UV light decreases the signal rapidly; thus the intensity of the fluorescence that is recorded last is reduced, leading to an underestimation of the content of the corresponding hormone. Second, the separation power of the filters used to record the individual fluorescence, of rhodamine and fluorescein, does not allow perfect discrimination between the two signals: a high cellular content of one hormone could produce a low background level sufficient to create false positives in the screening for the other. Owing to these restrictions, the number of somatostatin-containing cells that also contain insulin cannot be precisely determined. However, the results clearly indicate that these cells are numerous, representing between 25% and 50% of the somatostatin cells and thus at least 10% of the total cell population.

Electron microscopy indicates that for both cell types most secretory granules are atypical. They are elongated and uniformely dense; a few present a peripheral halo that has no resemblance with the halo that normally surrounds insulin β granules. In contrast to what is observed in the RIN2A population, many RW cells contain numerous granules (Fig. 1 G). The results of double labelling for insulin and somatostatin are in agreement with the light-microscopic observations: they show that in many cells both hormones are present in the same granule (Fig. 1 H), indicating at least partial cosegregation during packaging.

The RW cell line was selected for this analysis, because among the insulin gpt gene recombinant cell lines it contains the highest level of insulin mRNA. This several-fold increase was attributed to the selection by the hybrid gene of a β -cell line that actively transcribes the insulin gene, since both the insulin and chimeric genes were co-regulated. This was demonstrated by the fact that stimulation of insulin gene transcription by sodium butyrate also increased expression of the gpt gene (Besnard et al. 1991).

Clonal analysis has shown that the RIN cells can differentiate into any of the four pancreatic endocrine cell types. As individual clones were obtained simply by limiting dilution from the parent line (Philippe et al.1987), the clones can originate from undifferentiated cells. Alternatively, as proposed by the authors, the cells are capable of dedifferentiating and switching phenotype to express another hormone gene. With the RW cell line the situation should be different, since in the presence of mycophenolic acid, insulin and XGPRT are supposedly co-regulated. Thus it was of interest to look at individual cells; first, to determine if it really is possible to select and maintain a pure insulin-producing cell line; and second, to determine if one cell can synthesize more than one hormone or if the expression of the two genes is mutually exclusive.

The results of this analysis show that the method does not exclusively select β cells, since there are cells with undetectable levels of insulin that are resistant to mycophenolic acid. This observation suggests different possibilities. First, trans- or dedifferentiation could occur through one cyle and the observed cells could represent a population of cells condemned but not yet killed. This possibility is unlikely, or not the only one, since pairs or clusters of cells are frequently observed (Fig. 1E and F), indicative of the clonal origin of these insufin-negative cells. Second, the transcription of gpt could be controlled not only by regulatory sequences from the insulin gene but also in part from a gene adjacent to the insertion site. This environment would permit transcriptional activity sufficient to confer resistance to the antibiotic, allowing cells to multiply independently of insulin gene expression. However, this explanation is not tenable in view of the observed concomitant extinction of insulin and gpt expression in fibroblast × RW cell hybrids (and in hybrids with the other recombinants lines as well) leading to cell death in selective medium (Besnard et al. 1991). Furthermore, this hypothesis does not explain the increase in the number of somatostatin-synthesizing cells.

The failure to obtain these cells as a pure population precludes an in-depth analysis of the mechanism responsible for this differential expression. We have previously observed another example of discrepancy between the regulation of an exogenous (transfected) and an endogenous gene controlled by the same sequences, in this case for the hepatocyte-specific tyrosine aminotransferase (TAT) gene (Grange et al.1989). Fibroblasts that had stably integrated a gpt hybrid gene placed under the control of TAT regulatory sequences were cultured in the presence of mycophenolic acid and resistant clones synthesizing XGPRT were isolated. However, in these cells TAT gene expression was not induced and expression of gpt was associated with gene amplification and the use of cryptic promoters. Taken together, these experiments illustrate the difficulty there is in maintaining normal gene regulation when the regulatory sequences are under conditions of selection. They also show that strategies based upon the use of hybrid genes, designed to use selective pressure to detect inducers or repressors of a specific gene, may be very deceptive.

The presence of cells containing both somatostatin and insulin is surprising. Although their number can only be estimated due to technical difficulties (see Results), it is large enough to demonstrate clearly that co-expression of more than one cell-specific gene is possible. This suggests that the differentiation pathways of both β - and δ -cell types, whether derived from endoderm (Pictet et al. 1976) or neural crest (Alpert et al. 1988), are closely related. It also suggests that the regulation of transcription of the two insulin and somatostatin genes share common determinants. Of course, this similarity between the two genes may also be common to other cell-specific genes.

We thank Mrs E. Pierre for invaluable technical help. We also thank Drs D. Daegelen and J. Jami for the cell lines, Drs Garaud, Chance and Dubrasquet for the antibodies and Dr R. Buckle for carefull reading of the manuscript. This work was in part supported by a grant from the ARC.