Exocrine specific expression of Connexin32 is dependent on the basic helix-loop-helix transcription factor Mist1

Gap junctions are intercellular channels that provide direct passage of small molecules between adjacent cells. In pancreatic acini, the connexin26 (Cx26) and connexin32 (Cx32) proteins form functional channels that coordinate the secretion of digestive enzymes. Although the function of Cx26/Cx32 gap junctions are well characterized, the regulatory circuits that control the spatial and temporal expression patterns of these connexin genes are not known. In an effort to identify the molecular pathways that regulate connexin gene expression, we examined Cx26 and Cx32 gene activities in mice lacking the basic helix-loop-helix transcription factor Mist1 (Mist1KO). Mist1, Cx26 and Cx32 are co-expressed in most exocrine cell types, and acinar cells from Mist1KO mice exhibit a highly disorganized cellular architecture and an altered pattern of expression for several genes involved in regulated exocytosis. Analysis of Mist1KO mice revealed a dramatic decrease in both connexin proteins, albeit through different molecular mechanisms. Cx32 gene transcription was greatly reduced in all Mist1KO exocrine cells, while Cx26 gene expression remained unaffected. However, in the absence of Cx32 protein, Cx26 did not participate in gap junction formation, leading to a complete lack of intercellular communication among Mist1KO acinar cells. Additional studies testing Mist1 gene constructs in pancreatic exocrine cells confirmed that Mist1 transcriptionally regulates expression of the Cx32 gene. We conclude that Mist1 functions as a positive regulator of Cx32 gene expression and, in its absence, acinar cell gap junctions and intercellular communication pathways become disrupted.


Introduction
Pancreatic acini consist of exocrine cells whose main function is to produce and deliver the enzymes necessary for digestion. The appropriate release of these enzymes is accomplished through a complex and highly regulated system involving protein packaging and storage, vesicle trafficking, sensing of feeding signals and coordination of cellular secretion (Castle, 1990;Wasle and Edwardson, 2002). Enzymes are stored in granules as inactive zymogens until secretagogues, such as cholecystokinin (CCK), stimulate their release in a process known as regulated exocytosis. Binding of CCK to G-protein coupled receptors initiates a signaling cascade that culminates in the release of intracellular calcium, which in turn promotes movement of the zymogen granules (ZG) toward the apical cell surface and exocytosis of enzymes into pancreatic ducts (Joseph, 1996;McNiven and Marlowe, 1999). Improper signaling leads to impaired exocytosis and premature enzyme activation (Saluja et al., 1989). Owing to this specialized function, acinar cells exhibit extensive extra-and intracellular organization, with the granules aligned to the apical border adjacent to a centrally located duct while the receptors for secretagogue binding are located on the basal aspect of the cell.
It is this highly organized polarity that leads to efficient enzyme release. The process of regulated exocytosis is controlled in part by the cellular junctions that exist between the individual acinar cells, with gap junctions playing a major role in modulating the exocytotic process (Meda, 1996).
Gap junctions are trans-membrane protein structures that allow passage of small solutes between adjacent cells (Kumar and Gilula, 1996). Each gap junction consists of a plaque of connexons, which are composed of hexameric complexes of connexin proteins. Nineteen different connexin genes, each with a specific tissue distribution, exist in the mouse . The exocrine pancreas expresses connexin32 (Cx32; also referred to as Gjb1) and connexin26 (Cx26; also referred to as Gjb2), which is typical of most serous exocrine cells, including cells of the salivary glands, seminal vesicles and lacrimal glands (Hsieh et al., 1991;Meda et al., 1993). The existence of many different connexins with unique tissue specificities highlights the ability of gap junctions to provide a selective permeability between cells. Specific connexin expression patterns, coupled with various combinations of hexameric junctions and states of protein phosphorylation, contribute to a highly regulated process of intercellular Gap junctions are intercellular channels that provide direct passage of small molecules between adjacent cells. In pancreatic acini, the connexin26 (Cx26) and connexin32 (Cx32) proteins form functional channels that coordinate the secretion of digestive enzymes. Although the function of Cx26/Cx32 gap junctions are well characterized, the regulatory circuits that control the spatial and temporal expression patterns of these connexin genes are not known. In an effort to identify the molecular pathways that regulate connexin gene expression, we examined Cx26 and Cx32 gene activities in mice lacking the basic helix-loophelix transcription factor Mist1 (Mist1 KO ). Mist1, Cx26 and Cx32 are co-expressed in most exocrine cell types, and acinar cells from Mist1 KO mice exhibit a highly disorganized cellular architecture and an altered pattern of expression for several genes involved in regulated exocytosis. Analysis of Mist1 KO mice revealed a dramatic communication (Goodenough et al., 1996;Lampe and Lau, 2000). Stimulated acinar cells utilize gap junctions to communicate with adjacent cells in a common acinus to carefully regulate and coordinate the secretion of their protein products. Loss of connexin expression leads to increased levels of basal and stimulated exocytosis in acinar cells (Chanson et al., 1998), confirming the importance of gap junctions in the exocytosis pathway.
Although the expression and function of gap junctions are well characterized, the molecular regulatory circuits that control the spatial and temporal expression patterns of the connexin genes are not known. Elucidating these regulatory factors is of significant interest since much of the function associated with a specific gap junction complex is dependent on the types and amounts of connexins it contains (Hill et al., 2002;. Expression of the connexin gene family is probably regulated by key transcription factors that exhibit similar patterns of spatial expression. As previously mentioned, the Cx32 and Cx26 genes are co-expressed in cells known to exhibit regulated exocytosis (Meda et al., 1993). One transcription factor that shares a similar, but not identical, expression pattern is the basic helix-loop-helix (bHLH) protein Mist1 (Pin et al., 2000). Members of the bHLH family are instrumental in the development of numerous organ systems, and gene ablation studies in mice have revealed an essential role for bHLH proteins in the development of different cell types within the pancreas (Jenny et al., 2002;Kageyama et al., 2000;Krapp et al., 1998). Mice deficient in BETA2/NeuroD or neurogenin3 fail to develop insulinproducing β-cells, develop diabetes and die within a few days of birth (Jenny et al., 2002;Naya et al., 1997). Similarly, targeted disruption of the exocrine-specific factor PTF1-p48 results in a complete lack of exocrine cells within the pancreas, causing early postnatal death (Kawaguchi et al., 2002;Krapp et al., 1998).
Deletion of the Mist1 gene (Mist1 KO ), while not lethal, causes disorganization in acinus formation and numerous defects in pathways controlling regulated exocytosis (Pin et al., 2001). These include altered expression of genes involved in calcium mobilization (IP3 Receptor Type 3) and secretagogue signaling (CCK A receptor). In an effort to establish if there is a relationship between gap junction communication and the Mist1 KO phenotype, we examined the expression and function of connexin molecules in Mist1 KO acinar cells. In this report, we show that Mist1 KO acini are deficient in gap junctionmediated intercellular communication, mainly due to the loss of Cx32 expression. At all time points examined, Cx32 mRNA and protein levels are greatly reduced in the Mist1 KO exocrine pancreas, as well as in all exocrine organs that normally express the Mist1 gene. Although Mist1 KO acinar cells continue to express the Cx26 gene, Cx26 protein does not accumulate in gap junction plaques, suggesting that Cx32 is required for stable incorporation of Cx26 into the membrane. Additional studies using pancreatic acinar cell lines, transgenic mice and Mist1 expression plasmids have confirmed that Mist1 transcriptionally regulates expression of the Cx32 gene. We conclude that Mist1 functions as a positive regulator of Cx32 gene expression and, in the absence of Mist1, acinar cell gap junctions and intercellular communication pathways become disrupted.

Mist1 KO mice
The generation of Mist1 lacZ (Mist1 Het ) and Mist1 KO mice has been described previously (Pin et al., 2001). These mice contain a lacZ gene encoding a nuclear localized β-galactosidase (β-gal) protein driven from the Mist1 gene locus. Mist1 Het mice were maintained on a C57BL/6 background. The elastasep-Mist1 mut basic mice were generated through standard pronuclear injection protocols using a -500 -+8 fragment of the rat elastase 1 promoter (Heller et al., 2001;Kruse et al., 1995). Mist1 mut basic mice express the Mist1 mut basic protein exclusively in pancreatic acinar cells. The complete characterization of these mice will be reported elsewhere.

Acini isolation
Pancreatic acini were obtained following standard collagenase protocols (Ohnishi et al., 1997). Briefly, animals were euthanized and the pancreas was removed and placed into KRB (0.1 M NaCl, 1 mM MgCl2, 5 mM KCl, 0.5 mM Na2HPO4, 33 mM NaHCO3, 0.5 mM CaCl2, 4 mg/ml glucose, 15 mg/ml glutamine and MEM amino acids) buffer containing 0.1 mg/ml trypsin inhibitor. Pancreatic tissue was injected with a collagenase solution (100 U/ml in KRB plus 2.5 mg/ml BSA), gassed with O2 and incubated at 37°C at 110 rpm for 10 minutes. Tissue fragments were transferred to a fresh collagenase solution, gassed and incubated for an additional 40 minutes at 37°C with shaking. Acini were separated by manual pipetting through decreasing pipette orifices, and passed through a 150 µm nylon net filter. The filtrate was centrifuged at 500 rpm for 3 minutes and incubated at 37°C until ready to use. Cells were plated on glass coverslips in Waymouth's medium containing 1% penicillin and streptomycin (Pen/Strep) and 0.5% fetal bovine serum (FBS).

Dye coupling and electrophysiology
Glass microelectrodes were prepared from filamented microcapillary tubing (World Precision Instruments, Kwik-Fill 1B100F-3) using a Narishige vertical microelectrode puller. The glass microelectrodes were backfilled with 10 mM 6-carboxyfluorescein and individual acinar cells were injected using hyperpolarizing pulses of current (50 nA; 100 ms pulse per second) for 1 minute. Transfer of dye was monitored for at least three minutes and images were recorded using a digital camera.
Experiments to detect electrical coupling between acinar cells was performed as previously described (Mao et al., 2000). Briefly, glass microelectrodes were backfilled with 1 M KCl, single cells were injected using hyperpolarizing pulses of current (50 nA; 100 ms) and the transfer of current to adjacent cells was recorded. The recording electrode was inserted into either an adjacent cell or into a cell that was positioned ~200 µm away from the current-injecting electrode. Cell membrane potentials were recorded and plotted as current (mV) vs. time.

Gene constructs
The mouse Cx32 P1 promoter spanning -680 to +20 (Hennemann et al., 1992;Neuhaus et al., 1995) was cloned from genomic DNA using PFU DNA polymerase and promoter-specific primers containing flanking restriction sites. The 700 bp PCR product was digested with BamHI and BglII and cloned into the pGL2-Basic luciferase reporter plasmid from Clontech. The entire promoter region was verified by DNA sequencing. All other DNA constructs used in this report have been described previously (Krapp et al., 1998;Lemercier et al., 1998).

RNA isolation and hybridization blots
Total organ RNA was isolated using the QIAGEN RNeasy isolation system following the manufacturer's recommended protocol. Mice were euthanized and perfused with 12 ml of cold PBS and then tissues were removed and disrupted using a Tissue-Tearor mechanical homogenizer. Samples were resuspended in RNase-free water and stored at -80°C. For RNA blot analysis, 20 Μg of total RNA was precipitated in ethanol, resuspended in 20 µl of RNA loading buffer and heated at 65°C for 15 minutes until completely dissolved. RNA was loaded on a 1.2% formaldehyde agarose gel and electrophoresed in 1× MOPS buffer (0.5 M Mops, 0.01 M EDTA). After separation, RNA was transferred by capillary action to Hybond + membranes in 10× SSC. Hybridizations were performed using Clontech's ExpressHyb buffer according to the manufacturer's recommendations. Isotopically labeled probes were synthesized by standard procedures using the Ambion Decaprime kit.
Protein immunoblot assays Tissue protein extraction, protein electrophoresis and immunoblotting were performed as described previously (Pin et al., 2000). For immunoblot analysis, 50 µg of whole cell protein extracts were electrophoresed on acrylamide gels, transferred to PVDF membranes and incubated with primary antibodies against Mist1, Cx32, β-gal and β-actin. Following secondary antibody incubation the immunoblots were developed using an ECL kit (Pierce) as per manufacturer's instructions.

Expression of Cx32 is significantly reduced in Mist1 KO pancreatic acini
The unusual morphology of the Mist1 KO pancreatic acini, coupled with the altered expression of proteins involved in regulated exocytosis, suggests that cell junction complexes may be disrupted (Pin et al., 2001). As a first step to examining cell junctions, we assessed the expression of gap junction proteins in pancreatic acini derived from wild-type (WT) or Mist1 KO animals. Connexin32 (Cx32) is one of two connexin proteins expressed in serous exocrine cells and increases in gap junctions are readily detected during acinar cell development at the stage coinciding with cell organization (Bock et al., 1997;Yamamoto and Kataoka, 1985). Immunofluorescence of Cx32 in adult WT pancreatic sections revealed extensive expression, with gap junctions aggregated along the acinar cell membrane (Fig. 1A). In contrast, no Cx32-containing gap junctions were observed within Mist1 KO exocrine tissue (Fig.  1B). The loss of Cx32 protein was confirmed by immunoblot assays in which Cx32 protein was not detected in Mist1 KO extracts whereas it was readily observed in samples from WT pancreas (Fig. 1C). Previous work from our group described the gradual loss of β-catenin expression in older Mist1 KO animals (Pin et al., 2001) and it has been suggested that the loss of one type of junctional complex (e.g. adherens) can lead to the loss of other complexes (Fujimoto et al., 1997;Kojima et al., 1999). In order to determine if the loss of Cx32 expression was a result of disruption in other cell junctions, Mist1 KO and WT pancreas samples were co-stained for Cx32 and markers for zymogen granules (amylase), adherens junctions (β-catenin) and tight junctions (occludin). As previously reported (Pin et al., 2001), amylase staining is not restricted to the apical border of Mist1 KO acinar cells, indicating that ZG targeting is disrupted (Fig. 1D,E). In contrast, the accumulation of β-catenin and occludin was normal in Mist1 KO acinar cells, despite the absence of Cx32 protein ( Fig. 1F-I). These findings indicate that the loss of Cx32-containing gap junctions is an independent event in response to the absence of Mist1, and is not linked to the observed loss of other junctional proteins (specifically βcatenin) that occurs in older Mist1 KO mice (Pin et al., 2001).
The progressive deterioration of tissue architecture that is observed in Mist1 KO mice resembles chronic pancreatic injury and is associated with changes in a number of gene products (Pin et al., 2001). In many cases, the timing of the observed molecular changes suggests that they are secondary events occurring as a consequence of cell morphology, and not as a consequence of direct regulation by Mist1. In order to investigate whether the loss of Cx32 protein represents an early or late event associated with deletion of the Mist1 gene, we performed a time course analysis of Cx32 expression in WT and Mist1 KO pancreatic tissue (Fig. 2). Tissue samples taken between postnatal day 1 (PN1) and adulthood revealed strong Cx32 immunofluorescence in WT pancreas, whereas virtually no Cx32-containing gap junctions were detected in Mist1 KO sections, regardless of age ( Fig. 2A-F).
Several mechanisms regulate protein accumulation, including changes in protein stability or mRNA transcription. To establish if the loss of Cx32 protein in Mist1 KO acini was due to a corresponding loss of Cx32 transcripts, RNA blot analyses were performed on total pancreas RNA isolated from WT and Mist1 KO mice. As shown in Fig. 2G, Cx32 mRNA levels in all Mist1 KO samples were reduced by 75-85% compared to WT controls. This was particularly evident in the samples taken from 3-week-old mice, at which time the expression of Cx32 peaks in WT pancreas possibly because of the maturation of the regulated exocytosis process (Jamieson et al., 1988). The difference in Cx32 mRNA levels was consistent across all time points examined, with the first indication of a significant decrease detected at PN1 (Fig.  2G). These results indicate that in Mist1 KO acini, Cx32 gene expression is never activated to levels in WT acinar cells, and that the decrease in Cx32 protein most probably reflects changes in mRNA transcription or mRNA stability.

Mist1 KO pancreatic acinar cells are defective in intercellular communication
Given the significant reduction in Cx32 gene expression, we set out to examine whether Mist1 KO cells remain functionally coupled. For these initial experiments, individual acini were isolated from WT and Mist1 KO mice and stained for Cx32 protein. As observed previously in the pancreatic sections, Cx32 was detected in the membranes of WT acini while Journal of Cell Science 116 (16) Mist1 KO acini were devoid of Cx32 staining (Fig. 3A,B). WT and Mist1 KO acini then were tested for the ability to transfer 6carboxyfluorescein, a small molecular mass dye, into adjacent cells. As expected, dye transfer rapidly occurred in acinar cells isolated from control mice (Fig. 3C,D). In all instances (n=15 acini from 4 WT mice) three or more adjacent cells exhibited dye transfer. In contrast, acini isolated from Mist1 KO mice were highly defective in dye transfer. No transfer of 6carboxyfluorescein was detected in any of the neighboring acinar cells of Mist1 KO mice (n=12 acini from 4 mice), even after extended incubation times (Fig. 3E-H).
We next characterized WT and Mist1 KO cells with respect to electrical coupling. Hyperpolarizing pulses of current (50 nA; 100 msec) were injected into one cell within an isolated acinus and the spread of current to neighboring cells was monitored. As shown in Fig. 3I, electrical coupling between adjacent cells from WT acini was readily detected. Electrical coupling was also evident at interelectrode distances of ~200 µm, suggesting that widespread coupling of WT acinar cells occurred within a single acinus (data not shown). In contrast, Mist1 KO acini showed a complete absence of electrical coupling. In all cases, no transfer of the hyperpolarizing current was ever detected, even when multiple adjacent cells were tested (Fig. 3J). The complete loss of cell-cell communication in Mist1 KO acinar cells is surprising since acini from Cx32 KO mice maintain some coupling properties (Chanson et al., 1998). These findings suggest that other factors that are instrumental in establishing communication networks may also be affected in Mist1 KO mice.
Connexin26 mRNA levels are retained, but protein abundance is altered, in Mist1 KO mice The complete loss of dye transfer and electrical coupling in Mist1 KO acinar cells suggested that all gap junctional proteins were affected in these cells. Formation of normal pancreas gap junctions requires the co-expression of Cx32 and a second connexin protein, Cx26 (Meda, 1996;Zhang and Nicholson, 1994). Cx26 gene expression occurs in all acinar cells of the pancreas and the protein often cooperates with Cx32 to form heteromeric channels (Meda et al., 1993;Stauffer, 1995). In fact, Cx26 and Cx32 are often co-localized in the same cells (Zhang and Nicholson, 1994) and our immunohistochemical analysis confirmed that both proteins were localized to the same gap junction plaques in WT pancreatic acini (Fig. 4A-C). To determine if Mist1 regulates the expression of both genes in a similar fashion, we characterized the expression and localization of Cx26 in Mist1 KO acinar tissue. As shown in Fig. 4D, RNA blot analysis revealed comparable Cx26 transcript levels in both WT and Mist1 KO acinar cells. Surprisingly, no difference in Cx26 mRNA levels could be detected between normal and Mist1 KO mice at any time point examined (PN1 to adult) (data not shown), suggesting that transcriptional regulation of the Cx32 and Cx26 genes remains under separate control. To confirm if this expression pattern was also observed at the protein level, Cx26 antibodies were used to detect Cx26 in Mist1 KO pancreatic acini. Immunoblot analysis of 3-week old pancreatic tissue revealed equivalent levels of Cx26 protein in WT and Mist1 KO samples (Fig. 4E). Surprisingly, Cx26 protein was not detected in pancreatic acinar cells obtained from older (10-weeks old and adult) Mist1 KO mice, despite the continued presence of Cx26 transcripts (Fig. 4D). Examination of additional young animals

Fig. 2. Cx32 protein is absent in Mist1 KO pancreas acinar cells at birth. (A-F) Mist1
heterozygous (Mist1 Het ) and Mist1 KO pancreas samples were subjected to Cx32 immunofluorescence. Cx32-containing gap junctions (green, arrows) are absent in all Mist1 KO samples, regardless of age. Nuclei are stained with the DNA fluorochrome DAPI (blue). (G) Total RNA was isolated from wild type (WT) and Mist1 KO pancreas samples from postnatal day 1 (PN1), 3-week old, 4-month and 8month-old animals and then subjected to RNA blot analysis using a mouse Cx32 probe. In all cases, Cx32 mRNA levels are significantly lower in the Mist1 KO samples than in the WT control mice. 12S rRNA hybridization was used as an internal control.
(PN1 to 6-weeks old) by immunofluorescence confirmed the presence of Cx26 protein within pancreatic acinar cells, but instead of the expected gap junction staining pattern observed in WT and Mist1 heterozygous (Mist1 Het ) samples (Fig. 4F,H), Mist1 KO acinar cells exhibited a diffuse cytoplasmic staining (Fig. 4G,I). As predicted from the immunoblot analysis (Fig.  4E), this intracellular staining pattern persisted until 6 weeks of age, after which time the Cx26 protein was no longer detected in Mist1 KO cells (Fig. 4K). The significant mislocalization and reduction in Cx26 protein most likely contributes to the defects in cell communication observed in Mist1 KO acini. These results suggest that the loss of Cx32, coupled with the overall disorganization of the Mist1 KO acinar cells, affects the assembly, trafficking, and/or stability of Cx26-containing gap junctions to the plasma membrane, resulting in the eventual loss of unassembled Cx26 protein.

Mist1 functions as a positive regulator of Cx32 transcriptional activity
Our studies have shown that Cx32 gene expression, but not Cx26, is altered in the Mist1 KO pancreas. The specific loss of RNA expression for only one of the connexins strongly suggests that Mist1 exerts its control directly on Cx32 gene transcription. Indeed, analysis of Mist1 and Cx32 gene expression profiles revealed a striking overlap in tissue distribution, with co-expression observed in all secretory exocrine cells, including the pancreas, submandibular gland, parotid gland and seminal vesicles (Fig. 5A). The exception to this pattern of common expression was in liver hepatocytes (Fig. 5A) and neuronal supporting cells (data not shown), both of which do not undergo regulated exocytosis and only express Cx32 but not Mist1 (Fig. 5A) (Evert et al., 2002;Meda et al., 1993;Nicholson et al., 2001;Pin et al., 2000). Analysis of various tissues in Mist1 KO animals also revealed a striking correlation between Mist1 and Cx32 gene expression. Tissues that normally express Mist1 mRNA and protein, including the lacrimal and submandibular glands, exhibited greatly reduced levels of Cx32 transcripts (Fig. 5B) and an almost complete absence of Cx32-containing gap junction complexes (Fig. 5C-H). In contrast, expression of Cx32 transcripts and formation of Cx32-containing gap junction plaques remained unchanged in tissues that do not normally express Mist1, such as liver hepatocytes (Fig. 5B,I,J). These results support the hypothesis Journal of Cell Science 116 (16) that Mist1 functions as a tissue-specific regulator of Cx32 gene expression in secretory exocrine cells.
In order to examine whether Mist1 directly controls expression of the Cx32 gene, we cloned a portion of the mouse Cx32 P1 promoter (Neuhaus et al., 1995) spanning -680 to +20 into a luciferase reporter plasmid (Cx32p-Luc) and tested it for expression in the pancreatic exocrine cell line AR42J. Introduction of Cx32p-Luc into AR42J cells resulted in very low basal luciferase activity (Fig. 6A). Co-transfection of cells with the Cx32p-Luc reporter and an expression plasmid encoding Mist1 generated a 15-to 20-fold increase in Cx32p-Luc expression (Fig. 6A). Similar results also were obtained with the exocrine cell line ARIP (data not shown). Interestingly, no other bHLH transcription factor tested in this assay was able to activate expression of the Cx32p-Luc gene. These included both ubiquitously expressed Class A bHLH factors (E12, E47, HEB) and tissue restricted Class B bHLH factors (PTF1-p48, NeuroD, MyoD, Mash1) (Fig. 6B). The ability of Mist1 to activate expression of the Cx32p-Luc gene was dependent on its DNA binding and dimerization properties. Mutant Mist1 proteins that were defective in DNA binding (Mist1 mut basic ) or protein dimerization (Mist1 mut helix 1 ) (Lemercier et al., 1998), did not activate Cx32p-Luc expression (Fig. 6C). From these studies we conclude that an active Mist1 protein is required to generate full Cx32 gene expression in secretory acinar cells.
To more rigorously test the importance of active Mist1 protein for Cx32 gene transcription, we took advantage of the fact that the Mist1 mut basic protein functions as a dominantnegative factor to repress endogenous Mist1 activity (Lemercier et al., 1998). We reasoned that expression of Mist1 mut basic in pancreatic acinar cells would inhibit the endogenous Mist1 protein and down-regulate the expression of Mist1 target genes. Transgenic mice expressing a myc-tagged Mist1 mut basic protein under the transcriptional control of the pancreas-specific elastase 1 promoter (elastasep-Mist1 mut basicmyc) (Heller et al., 2001;Kruse et al., 1995), were generated and examined for Mist1 mut basic expression as well as for the presence of Cx32-containing gap junction plaques. As shown in Fig. 7, pancreatic acinar cells expressing the Mist1 mut basic protein showed a significant decrease in Cx32 expression, whereas adjacent acinar cells that did not express the Mist1 mut basic transgene exhibited normal expression and localization of Cx32. In addition, the Mist1 mut basic -expressing cells no longer produced Cx26-containing gap junction complexes and instead showed the diffuse cytoplasmic Cx26 staining that was evident in Mist1 KO samples (data not shown). These results demonstrate that loss of Mist1 activity, through either deletion of the Mist1 gene or through inhibition of endogenous Mist1 protein activity, leads to the loss of Cx32 expression. We conclude that Mist1 functions as a positive regulator of Cx32 gene expression in exocrine acinar cells and, in the absence of Mist1, gap junctions and intercellular communication pathways become disrupted.

Discussion
The intricate cell-cell communication network that is associated with most epithelial cells involves the utilization of trans-membrane gap junctions that are made up of different connexin hexamers. Gap junctions are unique among membrane channels in that they span two discrete cells and form direct intercellular conduits. Their main function is to permit rapid cellular responses to external stimuli to generate a variety of coordinated events. This is accomplished by allowing the passage of small molecules, such as Ca 2+ , Na + , K + , cAMP and inositol triphosphates, from one cell to another (Wasle and Edwardson, 2002). Several mechanisms govern the functionality of gap junctions, including connexin stability and post-translational modifications that maintain gap junctions in an open or closed state. However, the tissue-specific expression of the various connexin proteins underscores the true role of the junctions in cell development, differentiation and function. Loss of individual connexins characterize the pathology of several human diseases, including Charcot-Marie-Tooth disease, oculodentodigital dysplasia, cardiac ischemia and Journal of Cell Science 116 (16) cardiac hypertrophy (Bergoffen et al., 1993;Kelsell et al., 2001;Paznekas et al., 2003;Severs, 1994). In addition, recent mouse genetic studies have revealed unique roles for specific connexins in embryonic development, cardiac function, and growth regulation (Plum et al., 2000;Willecke et al., 1999).
Each connexin protein is distinct and the expression pattern for each displays a tissue and cell-type specificity (Plum et al., 2000). The regulated expression of the connexin genes results in the selective passage of molecules because of differing permeability characteristics of the channels. For instance, Ins(1,4,5,)P3 travels through Cx32 homomeric channels more readily than channels composed of Cx43 or Cx26 . In many cases, connexin gene expression is also altered as cells respond to changes in local environmental conditions (Petrich et al., 2002). These observations suggest that connexin genes are regulated by tissue-specific transcription factors that are responsive to intracellular signaling pathways. Unfortunately, molecular details of the transcriptional control of individual connexin genes remain largely unknown.
In an effort to identify the importance of gap junctions in acinar cell morphology and function, we examined connexin expression in mice lacking the bHLH transcription factor Mist1. Mist1 KO pancreatic acinar cells are highly disorganized (Pin et al., 2001), suggesting that intercellular junctions could be compromised. Indeed, this study has revealed that Mist1 KO acinar cells exhibit a specific loss of connexin expression and gap junctional communication. This defect is primarily caused by reduced expression of the Cx32 gene, which leads to severe depletion of gap junction formation and gap junction-mediated cellular communication.
While it could be argued that the loss of Cx32 is a consequence of cellular disorganization and not directly of a loss of Mist1 activity, several pieces of evidence support a model of direct transcriptional regulation. First, one would predict that a simple disruption of the gap junctions by cellular disorganization would lead to decreases in both connexin genes expressed in pancreatic acini. Cx32 and Cx26 proteins are coexpressed in many of the same cell types and readily form heteromeric gap junctions (Meda et al., 1993;Stauffer, 1995). However, in Mist1 KO acinar cells, Cx26 gene transcription is maintained while Cx32 gene transcription is greatly reduced, implying that the transcriptional control of these two genes is distinct. Secondly, analysis of the Cx32 gene reveals two separate promoters, P1 and P2 (Neuhaus et al., 1995). The P1 promoter directs Cx32 gene expression in liver, salivary glands and pancreatic epithelial cells, while the downstream P2 promoter is responsible for Cx32 expression in cells of the spinal cord and brain (Neuhaus et al., 1995). Studies from the Ruch laboratory (Koffler et al., 2002;Piechocki et al., 2000) have shown that liver-specific expression of the Cx32 gene relies on the winged-helix transcription factor HNF-1, which is also present in the pancreas. However, adult pancreatic HNF-1 expression is primarily confined to cells of the endocrine compartment where Cx32 is not expressed (Edlund, 2002;   Nammo et al., 2002). Other HNF family members (HNF-6, HNF-3β) are present in the exocrine pancreas (Rausa et al., 1997), but they do not activate Cx32 gene expression in transfection assays (unpublished data). Instead, our data suggest that the P1 promoter is activated in acinar cells by Mist1. Sequence analysis of the P1 promoter reveals 5 E-box regulatory sites that potentially bind Mist1, and a mutant form of Mist1, which does not bind DNA, fails to activate Cx32 expression. The observed repression of the Cx32 gene in pancreatic acinar cells expressing a dominant-negative Mist1 protein confirms that endogenous Mist1 activity is essential for acinar cell Cx32 transcription. As predicted, the reduction in Cx32 gene expression in Mist1 KO mice is observed only in cell types that normally express Mist1, such as the exocrine cells from the pancreas, salivary glands, lacrimal gland and seminal vesicle. From these studies we conclude that Mist1 functions as a direct positive regulator of Cx32 gene expression and, in its absence, acinar cell gap junctions and intercellular communication pathways become disrupted. Although the DNA binding and dimerization domains of Mist1 are required to achieve Cx32 gene transcription, the precise mechanism(s) by which Mist1 activates Cx32 gene expression is not known. In this regard, it will be essential to determine if Mist1 functions as a homodimer or as a heterodimer with another bHLH protein expressed in acinar cells. Preliminary data suggest that Mist1 homodimers are the preferred protein complex in the pancreas (T. Tran, unpublished data). However, Mist1 homodimers alone are not sufficient to activate the endogenous Cx32 gene in non-acinar cell types (unpublished data), suggesting that acinar cells contain cofactors that modify the Mist1 protein, or that serve as accessory transcription factors, to allow full Cx32 gene expression. Indeed, regulation of the Cx32 promoter is probably complex, since in the absence of Mist1 the Cx32p-Luc reporter gene remains active at a low basal level. This expression is not surprising given that the 680 bp 5′ flanking sequence contains protein binding sites for many general transcription factors including Sp1, NF1, Stat5 and SRF (data not shown). Future studies will focus on determining if Mist1 interacts with these other transcription factors that are bound to the Cx32 promoter to influence acinar-specific gene expression. Identification of the specific Cx32 promoter elements and additional acinar cell transcription factors will be essential to establish their role in activating Cx32 gene expression.
These studies have also provided new insight into the differential regulation of acinar-specific connexin proteins. While Cx32 appears to be regulated at the transcriptional level, this is not the case for Cx26. Despite normal levels of Cx26 mRNA, very few Cx26-containing gap junctions are detected in Mist1 KO acinar cells. Instead, Cx26 protein accumulates within the cytoplasm and does not associate with the cellular membrane. This observation supports the hypothesis that Cx26 prefers to form gap junctions with Cx32 (Zhang and Nicholson, 1994). Similar events have also been reported for Cx32 KO mice, where significant reductions in Cx26 gap junctions are observed, resulting in decreased dye transfer (Chanson et al., 1998). Interestingly, lacrimal glands from Cx32 KO animals exhibit transient diffuse patterns of Cx26 protein accumulation that is similar to our observations with Mist1 KO mice (Walcott et al., 2002). However, the complete phenotypes associated with Mist1 KO and Cx32 KO mice are quite distinct. Cx32 KO acinar cells continue to display some electrical coupling (Chanson et al., 1998) and the disorganization observed in Mist1 KO mice is not evident in Cx32 KO pancreatic tissue (data not shown). The alterations in cell morphology in Mist1 KO acini probably contribute to the loss of cell communication and indicate that the loss of Cx32 protein is not solely responsible for the phenotypic abnormalities observed in Mist1 KO animals. It remains possible that several key defects in regulated exocytosis, including altered CCK AR and Ins(1,4,5)P3 receptor expression (Pin et al., 2001), may contribute to the inefficient dye transfer and loss of electrophysiological coupling found in this current study. The disorganization of Mist1 KO acinar cells is evident early in embryogenesis, prior to significant gap junction organization (C. Johnson, personal communnication). Indeed, results from our laboratory support the idea that Mist1 KO pancreatic acinar cells do not reach a fully differentiated state and are developmentally blocked as immature exocrine cells. Thus, although Mist1 clearly has a role in controlling Cx32 gene expression, additional Mist1 target genes are probably important in establishing and maintaining normal acinar cell polarity and function. Identification of these genes will be crucial to fully understanding the role of Mist1 in these complex cellular processes.