mRNA profiling of rat islet tumors reveals nkx 6.1 as a beta-cell-specific homeodomain transcription factor.

Development of a high capacity multiplex reverse transcriptase-polymerase chain reaction protocol has allowed us to screen lineage related rat islet tumors classified as α-, β-, and δ-like as judged by their hormone profile for differential expression of more than 50 selected genes. We find that in addition to insulin the insulinoma express the normal β-cell markers Pdx-1, IAPP, and Glut-2, and that these markers are absent from the glucagonoma: a reflection of the normal α-cell. Furthermore, this study suggests that the GLP-1, glucagon, GIP, IGF-1, and insulin receptors as well as E-cadherin, R-cadherin, Id-1, and Id-2 are differentially expressed within the islet of Langerhans. Importantly, insulinoma-specific expression of the recently cloned homeodomain protein Nkx 6.1 predicted β-cell-specific expression in the normal islet. Immunohistochemistry using antibodies raised against recombinant Nkx 6.1 did indeed localize Nkx 6.1 expression exclusively to the nuclei of normal islet β-cells. Apart from pancreatic islets only the antral part of the stomach contained Nkx 6.1 mRNA. We conclude that multiplex reverse transcriptase-polymerase chain reaction-based mRNA profiling is a powerful tool to identify differentially expressed genes within phenotypically related cells and propose that Nkx 6.1 is involved in specifying the unique characteristics of the β-cell.

The developmental maturation of precursor cells into more specialized phenotypes during ontogeny is characterized by highly specific changes in gene expression. Thus, a detailed picture of mRNA levels in developmentally related cells may provide insight into both mechanisms and pathways of differentiation. The four distinct cell types (␣-␤-␦-, and PP-cells) present in the pancreatic islet of Langerhans are thought to derive from common precursor cells through stages of transient phenotypes expressing multiple islet as well as gut and brain peptide hormones (1,2). Such multi-hormonal and presumably immature phenotypes are also hallmarks of pancreatic endocrine tumors (3,4). From the clonal heterogeneous rat islet tumor culture MSL-G2 (5) of identical clonal origin as the widely used RIN-5AH insulinoma cell line (6,7), several distinct and more homogeneous phenotypes comprising transplantable insulinomas (referred to later as "IN insulinoma") (8), glucagonomas ("AN glucagonoma") (9), as well as a somatostatin-producing in vitro culture ("Tu-6") (10) have been derived.
In this study we have: 1) determined the degree by which these transformed phenotypes reflect the gene expression of their normal islet counterparts, 2) identified islet tumor phenotype-specific expression of selected genes and thereby predicted those involved in normal islet cell specialization.
The genes selected comprise those coding for products known or thought to be involved in specialized functions of the individual islet cells as well as transcription factors, which might be involved in specifying the process of islet cell differentiation. Since transcription factors regulating key differentiation steps are often involved in activating the genes that confer the specialized functions of terminally differentiated cells (11)(12)(13)(14), we have included all cloned factors, which have been implicated in insulin gene regulation.
Analysis of the mRNA levels of 54 selected genes in the different tumor phenotypes has led to the identification of several candidate genes with a predicted cell-specific expression in the normal islet. Importantly, we show by immunohistochemistry on sections of adult rat pancreas that the newly identified homeodomain protein Nkx 6.1 is expressed specifically in ␤-cells as predicted from the tumor analysis. The tissue distribution of Nkx 6.1 mRNA showed a remarkable tissue restriction (comparable to that of insulin and IAPP), suggesting that this factor plays a crucial role in specifying the unique nature of the pancreatic islet ␤-cell.
RNA Isolation-RNA isolation was carried out by lysing cells or tissue in guanidinium thiocyanate/phenol buffer (RNAzol, Cinna Biotecx) according to manufacturer's instructions. The RNA was resuspended in diethyl pyrocarbonate-treated water, reprecipitated with 2.5 volumes of EtOH, 0.1 volume of 4 M NaCl, recovered by centrifugation, and washed with 70% ice-cold EtOH.
cDNA Synthesis-Total RNA was diluted in diethyl pyrocarbonatetreated water to 0.2 g/l, denatured at 85°C for 3 min, and quickly chilled on ice. 5 l of the total RNA was mixed with 20 l of RT-mix (50 mM Tris-HCl, pH 8.3, 75 mM KCl, 3 mM MgCl 2 , 10 mM dithiothreitol, 200 units of Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.), 40 units of RNAsin (Promega), 3 g of random hexamers (Life Technologies, Inc.), and 0.9 mM dNTPs (Pharmacia Biotech Inc.) all in final concentrations) and left 10 min at room temperature and subsequently incubated at 37°C for 1 h. After cDNA synthesis the reaction was diluted with 50 l of H 2 O.
Primers-Primer sequences are given in Table I. In order to obtain as identical amplification kinetics for individual amplicons as possible, we set rules for the design of primers as follows. The length of the oligonucleotides should be within 19 -23 bases, with a GC content around 50%. Furthermore, the length of the products was restricted to be between 150 and 320 base pairs. When possible primers were synthe-sized based on published rat sequences. In cases where no rat sequences were available, primers were synthesized based on published mouse or hamster sequences. All of these amplify the homologous rat genes with the expected size in exact base pairs. However, due to possible mismatched bases, it remains a possibility that the ratio to the internal standard is reduced, as a result of less efficient amplification during the first cycles. This does not, however, affect the relative difference in expression between the samples tested, and will still allow a comparative expression analysis.
Multiplex RT-PCR 1   Primer sequences for each gene product is listed except for those not expressed in the samples tested. These genes were: GH, GHRH, VIP, oxytocin, PACAP, and Glut-4. The number of cycles listed were used during RT-PCR. These vary according to the difference in abundance of the mRNAs analyzed. R, M, and H refer to rat, mouse, and hamster, respectively. step at 94°C/1 min was followed by the chosen number of cycles as given: 94°C/30 s; 55°C/1 min; 72°C/1 min, 30 s, with the exception of reactions amplifying GC-rich amplicons (Ͼ60% GC content) where a thermal cycle profile of 95°C/1 min; 55°C/1 min; 72°C/1 min, 30 s was required for reproducible amplification.
Reaction products were separated on 0.4-mm 7 M urea, 1 ϫ TBE (0.13 M Tris base, 80 mM boric acid, 0.25 mM EDTA), 6% polyacrylamide gels. These were dried and exposed overnight to a phosphorimage storage screen. Screens were scanned by a Molecular Dynamics PhosphorImager series 400, and band intensities were calculated using the Imagequant software by the use of rectangle mode/local background/volume integration. All amplicons were measured against internal standards. ␣-Tubulin was used at 18 cycles, glucose-6-phosphate dehydrogenase at 22 and 25 cycles, and TATA-binding protein at 26 cycles. Expression data are given as ratios to the co-amplified internal standard, and normalized for dC content. Negative controls were always included. Second, by the use of primers specific for genomic sequences, we do not observe genomic DNA amplified even at high (Ͼ 30) cycle numbers (not shown). Positive controls of all primer sets were performed on cDNA prepared from tissue known to express the gene.
The present multiplex RT-PCR method allows the co-amplification of several cDNA products from total RNA preparations in a single tube. To avoid biased results due to potential interference between individual amplicons, we analyzed the amplification kinetics of individual amplicons in reactions where several products are co-amplified. Representative experiments of simultaneous amplification of either high or low abundance cDNAs (Fig. 1, A and B, respectively) directly show the non-competitive amplification of individual products, and also that these share an almost identical rate of amplification, highlighted by the almost identical slopes observed from a linear regression analysis within the exponential phase (Fig. 1C). However, as data obtained during the exponential phase may fail to detect an interference occurring during the first cycles, we combined multiplex sets with increasing numbers of products (from 1 to 5) co-amplified. Two representative sets are shown in Fig. 1D. No initial interference is detected, in any combinations, as no decrease in absolute product yield is observed for any product. Thus, all of these amplicons are independent of the presence of other products. However, as interference detection based on these assays is time-consuming, we devised a cross-combination assay (not shown) as a mean to detect amplicon interference. Principally, 25 amplicons were randomly grouped in 5 combinations (A-E) of 5 different amplicons, avoiding any co-migration, and co-amplified with an internal standard. These combinations are termed vertical (as A1-A5, B1-B5; etc.). All the amplicons were additionally combined in a horizontal manner, (i.e. A1-E1, A2-E2, etc.). If a significant difference in product yield of a given amplicon were observed between the vertical and the horizontal combinations, we accredited this to interference, which allowed us to avoid unsuitable combinations of particular primer sets. Furthermore, the cross-combination assay allows identification of products being amplified with combinations of primers in between sets. If any such significant products were detected, co-amplification of these primer sets was not allowed.
Antibodies, Immunohistochemistry, and Western Blotting-GST-Nkx 6.1 fusion protein was generated by insertion of a PCR fragment amplified from rat islet cDNA encoding 66 C-terminal Nkx 6.1 amino acids in-frame with GST in the pGEX-4T-1-plasmid (Pharmacia, Uppsala, Sweden). The sequence of the primers used was: upstream 5Ј-tccgaat-tcATGGCCACCGCCAAGAAGAAGCAG-3Ј and downstream 5Јtgc-caagtgcggccgcTCAGGAGCCCTCGGCCTCGG-3Ј. Lowercase letters indicate nucleotides included for cloning purposes. Fusion protein was purified using the RediPack GST purification module (Pharmacia) according to manufacturer's instructions. Immunization of rabbits was performed as described (19). Immunohistochemistry and Western blotting was performed as detailed elsewhere (5,20). Monoclonal antibodies to insulin (HUI-18), glucagon (GLU-001), and somatostatin (SOM-018) were from Novo-Nordisk Biolabs, Bagsvaerd, Denmark and used as described previously (21). Rabbit antiserum 1856 raised against rat Pdx-1 (as a recombinant GST fusion protein; Ref. 22), a nuclear marker for all adult islet ␤-cells as well as of few ␦-cells (23) was used in comparative analyses to characterize the expression profile of immunoreactive Pdx-1 versus Nkx 6.1 in pancreatic sections and islet cell cultures.

Phenotype-specific Islet Hormone mRNA Expression in MSL
Tumors-The RT-PCR assay showed expression of the islet hormones insulin and IAPP in the IN insulinoma, glucagon in the AN glucagonoma, and somatostatin in Tu-6, selectively ( Fig. 2A and Table II), concordant with previously described Northern and immunocytochemistry analyses (8,10,21,24). Furthermore, all three major islet hormones were simultaneously detected in islets. The fourth islet hormone, pancreatic polypeptide (PP), was detected in islets and unexpectedly also in the AN glucagonoma.
We then determined the expression profile of several neuroendocrine polypeptide hormones reported to be expressed during islet ontogeny. In agreement with previous reports we find that neuropeptide Y, peptide YY, and thyrotropin-releasing hormone are all highly expressed in newborn rat islets (25)(26)(27), and that CGRP-␤ but not CGRP-␣ is expressed in islets (25,28). Neither peptide YY nor CGRP-␤ were detected in any of the tumor clones, whereas thyrotropin-releasing hormone was detected at very low levels. In contrast, secretin was absent from newborn islets but found at low levels in all tumor phenotypes including RIN cells. Neuropeptide Y, however, is selectively expressed in the AN glucagonoma, which in addition also selectively expressed rather high levels of calcitonin. GH, GH-releasing hormone, oxytocin, pituitary adenylate cyclase activating peptide, and vasoactive intestinal peptide mRNA were detected neither in islets nor in the islet tumors. Finally, we confirmed previously reported expression levels of gastrin and cholecystokinin in the different MSL-derived tumor phenotypes and islets (5,9).
Cell-specific Expression of Hexokinases and Glucose Transporters-Glucose-sensing, which is an essential character of the mature ␤-cell, has been shown to rely on expression of a specific form of hexokinase (29) and possibly also a specific glucose transporter (30). We analyzed for potential tumor phenotype-specific expression levels of Glut-1, -2, -3, -4, glucokinase (hexokinase IV), and the low K m hexokinase II. In summary, we observed Glut-1 to be expressed at low levels in all tumor phenotypes as well as in islets. In contrast, Glut-2 was expressed at high levels in MSL-G2, IN insulinoma, and islets, at lower levels in Tu-6 and completely absent in AN glucagonoma and RIN-5AH cells (Fig. 2D). In these samples, we only detected very low levels of Glut-3 and no Glut-4 (data not shown). Glucokinase was expressed at lower levels in islets as well as in the various tumor phenotypes, although with a slightly reduced expression in RIN-5AH and Tu-6 cells. Furthermore, the exon-1a-containing glucokinase transcript initiated from the upstream promoter, which has been shown previously to be expressed in pancreatic ␤-cells, was the predominant form, as we observed similar quantitative data when using primers common to both the liver-and the ␤-cellspecific forms (Fig. 2D, product at 161 base pairs). Surprisingly, the AN glucagonoma showed the highest level of glucokinase expression. The hexokinase II gene expression level was vastly elevated in all tumor cells compared to normal islets (Fig. 2G).
Cell-specific Expression of Plasma Membrane Receptors-Glucagon, GLP-1, and GIP have all been suggested to act as incretins, potentiating glucose-induced insulin release. We therefore analyzed the mRNA expression profile of their plasma membrane receptors belonging to the seven-transmembrane G-protein-coupled receptor family in order to search for a heterogeneous distribution among the MSL clonal variants. The GLP-1-R, GIP-R, and the glucagon-R were detected in islets. Interestingly, expression of these receptors segregated predominantly with a particular tumor phenotype: GLP-1-R in the IN insulinoma and RIN-5AH cells (Fig. 2E), glucagon-R in Tu-6, and the expression of the GIP-R was 5-fold elevated in the AN glucagonoma (Fig. 2G)  coupled receptor and tyrosine kinase-R families were tested (see Table II). Particularly the IGF-1-R and insulin-R showed differential expression among the tumor phenotypes. Although insulin-R was expressed in all samples, a variation in splicing was observed. Using primers designed to distinguish between insulin-R splice variants (with or without exon 11), all tissues showed a balanced splicing yielding both forms, except the AN glucagonoma, which selectively expressed insulin-R mRNA without exon 11, similar to the splice pattern found in the brain. IGF-1-R expression was selectively absent from AN glucagonoma (Fig. 2D) while detected in other MSL-derived clones as well as in islets. In contrast, the GH-R was homogeneously expressed in islets as well as all the tumor phenotypes.
Finally, we analyzed the expression of the cell-adhesion molecules, NCAM and R-, N-, and E-cadherin (uvomorulin/L-CAM). These are all previously reported to be present in islets and thought to be involved in the specification of the islet architecture. In agreement with our earlier observations (20), we did not detect E-cadherin in the AN glucagonoma (Fig. 2I), whereas it was present in all other tumors and cell lines. R-cadherin was expressed in all lines but only at very low levels in Tu-6 and in RIN-5AH. In contrast, NCAM and N-cadherin (Fig. 2G) was homogeneously expressed in all tumor cells.
Tumor Phenotype-specific Expression Profiles of Hormone Gene Transcription Factors-To analyze for correlation between phenotype-specific markers and presence of specific transcription factors, we next determined the expression profile of several gene regulatory proteins implicated in islet hormone gene expression. Transcription factors from the helixloop-helix (HLH) family have been shown to be involved in islet hormone gene activation. We analyzed expression of the HLH factors Pan-1/Pan-2 (rat homologues of human E47/E12, respectively), BETA-2 (hamster)/NeuroD (mouse), and also the inhibitory factors Id-1, -2, -3, and -4. Both the Pan-1 and -2 splice forms were detected at low levels in all samples (Fig. 2I), although with a predominant (2-5-fold) expression of Pan-1. In comparison, NeuroD was similarly homogeneously expressed between samples (Fig. 2I), but found at much higher levels than the pan-mRNAs. In contrast, distribution of the Id factors was more heterogeneous, with Id-1 expressed at high levels in Tu-6 and RIN-5AH cells (Fig. 2H). Id-2 was found at high levels in Tu-6 and at lower levels in all other samples (Fig. 2E). Id-3 was only detectable in Tu-6, while Id-4 was selectively expressed in the IN insulinoma and Tu-6 (Fig. 2H). Within the islets, only Id-1 and Id-2 were expressed at significant levels. We also analyzed expression of factors belonging to the basic-leucine zipper family. None of the basic-leucine zipper family factors including CREB, ⌬CREB, and c-Jun, as well as CREBbinding protein (Fig. 2, E-G), showed marked differential expression within the tumor phenotypes, and all were detected at comparable levels in newborn islets.
The homeodomain protein Pdx-1, an insulin gene transcription factor that also is required for pancreas formation (14), was found to be expressed at comparable levels in the MSL-G2 stem-cell culture, the IN insulinoma, and the Tu-6 culture (from which it was originally cloned (31)) ( Fig. 2F) Importantly, this factor was absent from the AN glucagonoma as in the normal islet ␣-cell. Several other homeodomain proteins (Isl-1, cdx-3, lmx-1) cloned from islet cell lines and able to interact with the insulin promoter (32,33) were analyzed. Isl-1 was found at low levels in RIN-5AH cells but homogeneously expressed in all MSL-G2-derived tumor phenotypes (Fig. 2H) as well as in islets, while cdx-3 and lmx-1 were only found in RIN-5AH cells but not in islets (Fig. 2F, note that the narrow bands in lanes 1-4 migrating slightly faster than the doublet lmx-1 band in lane 5 are nonspecific). The LIM domain protooncogene, rhombotin (Lmo-1), was homogeneously expressed in all tumor phenotypes.
The Homeodomain Protein Nkx 6.1 as a Candidate Linked to ␤-Cell Differentiation-When analyzing the expression profile of the homeodomain factor Nkx 6.1, previously cloned by degenerate PCR from HIT cells (34), we found that expression of Nkx 6.1 was induced in the transition from MSL-G2 to the IN insulinoma (Fig. 2H). Additionally, Nkx 6.1 expression was detected in Tu-6. This expression profile marks the Nkx 6.1 gene as one of the few genes whose activity segregates differentially between the MSL-G2 culture and the insulinoma: a profile shared only by insulin, Id-4, and GLP-1-R genes. As Nkx 6.1 mRNA was readily detected in islets, we wanted to address the question whether the tumor data would predict a ␤-cellspecific expression pattern in the normal islet. We therefore raised antibodies to a GST-Nkx 6.1 fusion protein that included the 66 C-terminal amino acids of the rat Nkx 6.1 protein. Two distinct rabbit antisera, ␣-Nkx 6.1-173 and ␣-Nkx 6.1-174, reacted identically in Western blotting to a doublet band migrating at 44 and 46 kDa. We could thus confirm the expression profile of Nkx 6.1 at the protein level between the MSL-G2derived variants (Fig. 3). The staining of both bands were completely abolished when antisera were preincubated with the recombinant Nkx 6.1 fusion protein (Fig. 3). Moreover, when analyzing sections of mouse (data not shown) and rat pancreas, Nkx 6.1-like immunoreactivity was strictly confined to ␤-cell nuclei (Fig. 4, A and B). No glucagon-positive or somatostatin-positive cells co-expressed Nkx 6.1 as evidenced by FIG. 3. Western blot analysis of Nkx 6.1. ␣-Nkx-6.1-173 (1:1000) was used as primary antisera. 15 g of nuclear extract was applied per lane. NHI-6F is a MSL-G2 subclone stably transfected with the human insulin gene (8). NHI-6F-28 is an insulinoma culture derived from the NHI-6F line after ␤-cell maturation during in vivo passage. The NHI-6F/NHI-6F-28 cells are identical to the MSL-G2/IN system (8). The NHI-6F-28 nuclear extract is slightly degraded. GH 3 is a rat pituitary cell line. ␣-TC1 and ␤-TC3 are large T-antigen transformed cell lines derived from transgenic mice with heritable glucagonomas (17) and insulinomas (16), respectively. Absorption control is shown to the right. The anti-Nkx 6.1 staining pattern (two left lanes) was not affected by preabsorption to GST-Pdx-1 (data not shown) but completely blocked by preabsorption to GST-Nkx 6.1 (100 g/ml). co-immunostaining performed on dispersed rat islet cells in monolayer (Fig. 4, D and E). Additionally, we observed that all insulin-positive cells were Nkx 6.1-positive as well (Fig. 4C). The specificity control experiment shown in Fig. 5 demonstrates that the observed nuclear staining is due to the presence of Nkx 6.1-like immunoreactivity. Furthermore, when testing the expression profile in a large number of tissues by RT-PCR, Nkx 6.1 mRNA was selectively detected at high levels in pancreatic islets and in very low amounts in antrum (Fig. 6). Lack of Nkx 6.1 mRNA in MSL-G-AN is in accordance with absence of immunoreactive Nkx 6.1 in normal ␣-cells (Fig. 4A) and in mouse ␣-TC1 (Fig. 3); nonetheless, in accordance with Rudnick et al. (34), we do detect mRNA for Nkx 6.1 in the ␣-TC1.9 subclone derived from ␣-TC1 cells.

DISCUSSION
By multiplex RT-PCR, we have screened a large number of genes for differential expression in phenotypically distinct islet tumor phenotypes using neonatal rat islets of Langerhans as control tissue. Importantly, we find that the expression of the previously identified islet ␤-cell restricted genes IAPP, GLP-1-R, and Glut-2 and the transcription factor Pdx-1 segregate with insulin gene activity in the insulinoma phenotype. Derivation of the glucagonoma from pluripotent MSL-cells is associated with a shutdown of insulin, Glut-2, and Pdx-1 gene activity. These observations reflects the nature of normal islet ␤and ␣-cells. We therefore conclude that these phenotypically distinct transformed islet cells are useful representatives of the corresponding cell types in the normal islet.
Of the genes tested, a substantial fraction showed differential expression among the different tumor phenotypes, of which the most important are summarized in Fig. 7. It is striking that the MSL-G2 line is almost identical to the ␤-cell-like IN insulinoma in expression of most of the selected genes except for the inverse correlation between the two major pancreatic hormones insulin and glucagon and the presence of Nkx 6.1, Id-4, and GLP-1-R in the insulinoma. Several otherwise ␤-cell restricted factors, such as Glut-2, IAPP, and, most importantly, the Pdx-1 gene, are expressed prior to insulin gene activation. This supports our previous proposal that the MSL-G2 culture is representing a pre-␤-cell phenotype (21).
In contrast, the in vivo derivation of glucagonomas is associated with profound changes in the gene expression profile (Fig. 7). Several hormone genes (neuropeptide Y, PP, and calcitonin) as well as the GIP receptor gene are activated, while the expression of the calcitonin-R, IGF-1-R, R-, and E-cadherin, as well as Glut-2 and Pdx-1, are completely turned off. From these results we predict that these genes are differentially expressed in normal islet cells.
Glucokinase expression in the insulinoma as well as in the glucagonoma is interesting and suggests a role of the enzyme in sensing low glucose levels, which might be coupled to the regulated secretion of glucagon in the normal ␣-cell. Agreeing with the glucagonoma expression, the glucokinase promoter has also been found active in ␣-cells of transgenic mice (35).
During the ␦-cell-like maturation seen in Tu-6 cells, which proceeded through the establishment of an in vitro culture from the insulinoma tumor cells (10), we observe down-regulation of several hormonal genes (gastrin, cholecystokinin, IAPP, and insulin), whereas somatostatin is activated in most cells of the Tu-6 line. This is additionally accompanied by the reduction of Glut-2, GLP-1-R, and R-cadherin expression, and a concomitant activation of the glucagon receptor and the Id-1 and -2 genes. However, both Pdx-1 and Nkx 6.1 are still expressed in the Tu-6 line, indicating that some features of the ␤-cell phenotype are retained in this somatostatinoma culture. The fact that the Id factors are expressed in this material is interesting, in light of the HLH-protein-dependent hormone gene activation, e.g of the insulin gene. A possible role of these factors could be to evoke tissue-specific functions of the (in the islet) homogeneously expressed complex IEF-1 (36, 37), consisting of FIG. 5. Specificity control of the anti-Nkx 6.1 antisera for immunocytochemistry. NHI-6F-28 insulinoma cells (positive for Nkx 6.1 and Pdx-1 mRNA) were stained using antisera against Nkx 6.1 and Pdx-1. Both antisera were preabsorbed with either antigen (GST-Nkx 6.1, GST-Pdx-1) as indicated. As expected, only the corresponding antigen abolished staining. Identical results were obtained on pancreatic sections (not shown). Photomicrographs (A and D) were overexposed to visualize unstained nuclei.
the Pan1/2 and NeuroD factors. That is, these factors may act to negatively regulate the IEF-1 complex through binding to the individual components of IEF-1, inactivating these in non-DNA-binding heterodimeric complexes. In this respect, the low level insulin gene activation in Tu-6, despite presence of the known insulin gene transactivating proteins (Pdx-1, NeuroD, and the pan-factors) could be explained by the high level expression of the Id-1 and Id-2 genes (38).
Together, our results show that the maturation pathways of the MSL-G2 tumor cells closely resembles the hypothetical model of islet ontogeny (1,39,40) proposing that cells positive for Pdx-1 (40) are precursors for all four islet cell types, and that ␣-cell maturation involves inactivation of the Pdx-1 gene. The Pdx-1/glucagon double-positive cells constituting the majority of the cells of the MSL-G2 culture may reflect an endocrine precursor phenotype as previously suggested (2,40). Furthermore, Glut-2-positive and hormone-negative cells in the developing pancreas have been proposed as pre-␤-cells (39). The fact that ␦-cell-like maturation of the Tu-6 somatostatinoma culture occurred through the insulinoma phenotype is in accordance with the cell lineage model of islet ontogeny as ␦-cell maturation was proposed to proceed through an insulin/ somatostatin double-positive phenotype (1). PP-cell maturation has earlier been suggested to proceed through insulin and somatostatin-expressing precursor cells with terminal differentiation occurring late in islet development (1,40). This is contrasted by our finding that, within the tumor system, only the AN glucagonoma expresses PP, suggesting that PP-and ␣-cell types could be linked more closely. Supporting this hypothesis are findings of a cell type co-expressing glucagon and PP in man (41), and that a high fraction of the earliest PP-positive cells during rat pancreatic development are co-expressing glucagon. 2 An important observation of this study is the selective coactivation of the homeodomain protein Nkx 6.1 with insulin in the IN insulinoma. This suggests a role for Nkx 6.1 in the specification of the normal ␤-cell, substantiated by our finding that Nkx 6.1 immunoreactivity within the islet is confined to the nuclei of pancreatic ␤-cells exclusively. Several otherwise ␤-cell restricted genes, such as Glut-2, IAPP, and Pdx-1 were found to be expressed in the MSL-G2 culture prior to insulin and GLP-1-R gene activation. It thus appears that differentiation of the ␤-cell can be divided into discrete steps, each requiring the action of appropriate key regulatory factors. Both Pdx-1 and Nkx 6.1 are likely mediators of these processes. However, although this study may suggest that Nkx 6.1 is playing a role late in ␤-cell development, its action is probably not limited to the specification of the ␤-cell phenotype, as developmental studies have shown that Nkx 6.1 protein is expressed in the majority of the endodermal pancreatic precursor cells (Pdx-1and Glut-2-positive-cells), during early stages of pancreatic organogenesis, and at subsequent stages segregates to ␤-cells exclusively. 2 In addition to the MSL-G2-derived cells, we analyzed the commonly used rat insulinoma cell line, RIN-5AH, which is syngeneic to MSL cells. In many respects, the RIN-cells appear to represent much less differentiated ␤-cells when compared to the MSL-G2-derived IN tumor. This is most notably exemplified by the low insulin production, and lack of IAPP, Glut-2, Nkx 6.1, R-cadherin, and IGF-1 receptors. Additionally, RIN-5AH cells express the homeodomain proteins cdx-3 and lmx-1, which we do not find in islets. We believe that the prolonged in vitro culture of RIN-5AH cells combined with their fast growth rate are major determinants of these differences. From this comparison, we suggest that the MSL-G2-IN phenotype in many respects is superior to RIN-5AH cells when addressing aspects of ␤-cell biology by the use of transformed cells.
In summary, we have, by the use of mRNA profiling, successfully predicted the ␤-cell-specific expression of Nkx 6.1. We also predict a number of other genes to be expressed in particular islet cell types. Whether these predictions are accurate is currently being investigated using in situ hybridization and immunohistochemistry.  Table II). One arrow represents a significant change in level of expression (up or down), two arrows up denotes a marked induction from undetectable expression, and two arrows down denote a marked reduction to undetectable levels.
FIG. 6. RT-PCR (25 cycles) analysis of Nkx 6.1 expression in extrapancreatic tissues. Using glucose-6-phosphate dehydrogenase as internal standard, 22 different rat tissues were analyzed for Nkx 6.1 expression. Various tumor lines (first seven lanes) served as control. High level expression was confined to pancreatic islets, with antrum as the only other tissue showing detectable expression.