Transcriptional Regulation of X-Box-binding Protein One (XBP1) by Hepatocyte Nuclear Factor 4α (HNF4Α) Is Vital to Beta-cell Function*

The transcription factor, X-box-binding protein-1 (XBP1), controls the development and maintenance of the endoplasmic reticulum (ER) in multiple secretory cell lineages. We show here that Hepatocyte Nuclear Factor 4α (HNF4α) directly induces XBP1 expression. Mutations in HNF4α cause Mature-Onset Diabetes of the Young I (MODYI), a subset of diabetes characterized by diminished GSIS. In mouse models, cell lines, and ex vivo islets, using dominant negative and human- disease-allele point mutants or knock-out and knockdown models, we show that disruption of HNF4α caused decreased expression of XBP1 and reduced cellular ER networks. GSIS depends on ER Ca2+ signaling; we show that diminished XBP1 and/or HNF4α in β-cells led to impaired ER Ca2+ homeostasis. Restoring XBP1 expression was sufficient to completely rescue GSIS in HNF4α- deficient β-cells. Our findings uncover a transcriptional relationship between HNF4α and Xbp1 with potentially broader implications about MODYI and the importance of transcription factor signaling in the regulation of secretion.

The transcription factor, X-box-binding protein-1 (XBP1), controls the development and maintenance of the endoplasmic reticulum (ER) in multiple secretory cell lineages. We show here that Hepatocyte Nuclear Factor 4␣ (HNF4␣) directly induces XBP1 expression. Mutations in HNF4␣ cause Mature-Onset Diabetes of the Young I (MODYI), a subset of diabetes characterized by diminished GSIS. In mouse models, cell lines, and ex vivo islets, using dominant negative and human-disease-allele point mutants or knock-out and knockdown models, we show that disruption of HNF4␣ caused decreased expression of XBP1 and reduced cellular ER networks. GSIS depends on ER Ca 2؉ signaling; we show that diminished XBP1 and/or HNF4␣ in ␤-cells led to impaired ER Ca 2؉ homeostasis. Restoring XBP1 expression was sufficient to completely rescue GSIS in HNF4␣deficient ␤-cells. Our findings uncover a transcriptional relationship between HNF4␣ and Xbp1 with potentially broader implications about MODYI and the importance of transcription factor signaling in the regulation of secretion.
Cells use transcription factors to regulate expression of gene cohorts that coordinate response to stress, determine specific developmental fates, and scale intracellular architecture during development and disease (1,2). When the biosynthetic load of a cell is increased and misfolded proteins accumulate in the endoplasmic reticulum (ER), 2 the volume and composition of the ER is altered to facilitate the synthesis and processing of nascent polypeptides via the unfolded protein response (UPR) pathway. One of the principal components of the UPR is the transcription factor X-box-binding protein 1 (XBP1), which is canonically activated via IRE1 splicing of the XBP1 transcript during ER stress (3,4). However, XBP1 also establishes and maintains the subcellular machinery for synthesizing large quantities of protein during the normal development of professional secretory cells (2,5). While the majority of genes activated by XBP1 are involved in ER biogenesis, up to 40% of its targets are not directly linked to the ER-stress response (6), further supporting its functions outside the UPR. In addition, the increased XBP1 that coordinates the scaling up of the secretory apparatus during development and homeostasis of dedicated secretory cells does not seem to require activation of the UPR (7). How XBP1 is induced and maintained during differentiation of secretory cells even in the absence of substantial ER stress is unclear, but a potential alternative mechanism is that Xbp1 may also be transcriptionally regulated. Hepatocyte Nuclear Factor 4-alpha (HNF4␣) is a highly conserved transcription factor responsible for orchestrating the early development and maintenance of multiple adult organs. As a master developmental regulator, HNF4␣ likely acts upstream of the factors that establish the extensive cellular machinery required in professional secretory cell lineages within those organs. Despite overlapping expression and function, no direct relationship between HNF4␣ and Xbp1 has yet been described.
HNF4␣ is vital for ␤-cell function, and indeed, human mutations in HNF4␣ cause Mature-Onset Diabetes of the Young 1 (MODYI), a subset of diabetes characterized by diminished glucose-stimulated insulin secretion (GSIS) in pancreatic ␤-cells, and susceptibility to type II diabetes (8,9). While we know that ␤-cells require HNF4␣ to function, we understand little about the mechanistic/physiological role of HNF4␣ in these cells. Previous work showed that disrupting HNF4␣ expression in vivo in mouse islets resulted in diminished GSIS similar to that observed in MODY patients with HNF4␣ mutations (10,11). Loss of HNF4␣ also was observed to disrupt Ca 2ϩ signaling, though the mechanisms underlying those defects remain unclear. Decreased ER function is a plausible mechanism for the loss of function in MODYI ␤-cells, because insulin secretion in ␤-cells is diminished if ER homeostasis is disturbed (12,13).
Defects in ER-related proteins contribute to multiple diabetic phenotypes in humans (14), and HNF4␣ is known to be important for maintaining ER stress response (15). In addition, knocking down XBP1 specifically in ␤-cells also leads to significantly reduced GSIS (16). Finally, disruption of calcium homeostasis in the ER also leads to impaired GSIS (17). Here we have characterized how XBP1 expression is governed at the transcriptional level and establish HNF4␣ as a direct transcriptional regulator of its expression. This implicates HNF4␣ in the maintenance and establishment of secretory cell ER networks. Accordingly, we report that both HNF4␣ and XBP1 are required to maintain ER calcium homeostasis and GSIS in ␤-cells. In addition, we show that restoration of XBP1 expression alone in islets deficient for HNF4␣ is sufficient to rescue impaired GSIS. Thus, the results may provide new insight toward discerning why dysfunction in HNF4␣ causes the pathophysiological findings in MODYI patients.
Constructs were verified to be correct by DNA sequencing. For siRNA we transfected MIN6 and INS-1 cells with 10 nM HNF4␣ siRNA (silencer select, Invitrogen) using Lipofectamine 2000 according to the manufacturer's protocol.
In Silico Identification of HNF4␣ Binding Sites in XBP1 Promoter-Areas of high conservation among multiple mammalian species (human, rhesus, mouse, rat, dog) 10 kb upstream and downstream of the XBP1 transcription start site were identified using ECR browser. These areas were then scanned with the Transfac transcription factor binding database for known transcription factor binding site sequences (18). Two consensus HNF4␣ sequences were identified; one 1.2-kb upstream (hg19 chr22: 29,198,941) of the XBP1 transcription start site and one 2.4-kb upstream (hg19 chr22:29,197,731).
qRT-PCR and Western Blot-RNA was isolated using RNeasy (Qiagen) per the manufacturer's protocol. RNA was treated with DNase I (Invitrogen) and then reverse transcribed using the SuperScript III (Invitrogen) standard protocol (most cDNA syntheses started with 1 g of total RNA). Measurements of cDNA levels were performed by qRT-PCR using a Stratagene MX3000P detection system. Absolute QPCR SYBR green mix (Thermo Scientific) fluorescence was used to quantify relative amplicon amounts of each gene.
Cells for Western blot analysis were lysed in RIPA buffer. Proteins were quantified by DC protein assay (Bio-Rad) and then separated on NuPAGE Bis-Tris gels (Invitrogen), transferred onto Amersham Biosciences Hybond ECL nitrocellulose (GE Healthcare) membranes, and detected by Immobilon chemiluminescence (Millipore). Primary antibodies used were rabbit anti-XBP1 (Santa Cruz Biotechnology), mouse antimyc-tag (Cell Signaling), and rabbit anti-␣and ␤-tubulin (Cell Signaling). Secondary antibodies were horseradish peroxidaseconjugated donkey anti-rabbit and anti-mouse Ig (Santa Cruz Biotechnology).
Chromatin Immunoprecipitation-Chromatin immunoprecipitation (ChIP) was performed as described previously (47). Approximately 100 mg of tissue from the pancreata of 5, 6 -8week-old WT mice were homogenized and used for this ChIP experiment. Ten microliters of anti-HNF4␣ (Santa Cruz Biotechnology) or whole rabbit serum (preimmune control) together with protein A/G plus-agarose (Santa Cruz Biotechnology) was added to the homogenized tissue for immunoprecipitation. Quantitative real-time PCR (qRT-PCR) was performed (primer sequences are available from corresponding author) to assess the quantity of genomic sequences immunoprecipitated by either preimmune control or HNF4␣ antiserum, as well as a 1:10 dilution of the cell extract prior to immunoprecipitation (input). The two predicted HNF4␣ binding sites described above were probed in addition to an intronic control region with no predicted HNF4␣ binding sites nearby. Data are graphed as a percentage of precipitated DNA:total input (genomic DNA).
Beta-cell Morphological Characterization using Immunofluorescence-Pancreata were prepared and stained as described previously (49). Briefly, they were fixed with freshly prepared formalin and suspended in fixative for 24 h at room temperature, followed by multiple rinses in 70% ethyl alcohol (EtOH), arrangement in 2% agar in a tissue cassette, and routine paraffin processing. Sections (5 m) were deparaffinized and rehydrated, and then antigen retrieval was performed by boiling in 50 mM Tris-HCl, pH 9.0. Slides were blocked in 1% bovine serum albumin (BSA) and 0.3% Triton X-100 in phosphatebuffered saline (PBS) and then incubated in goat anti-Calregulin (Santa Cruz Biotechnology) followed by AlexaFluor594 antigoat. Fluorescence microscopy and imaging were performed using a Zeiss Axiovert 200 microscope with Axiocam MRM camera and Apotome II instrument for grid-based optical sectioning.
For morphological analysis, three 5-m sections taken 100-m apart were stained with hematoxylin and eosin to allow identification of islets. Whole slides were scanned with the Nanozoom microscope and the cross-sectional area of islet/ total pancreas tissue was measured across each slide using Nanozoom Digital Pathology and software (Hamamatsu). Samples were randomized, and the scorer was blinded to ensure unbiased quantification. Values are expressed as %␤-cell area.
Mouse Studies-All experiments involving animals were performed according to protocols approved by the Washington University School of Medicine Animal Studies Committee. Floxed Hnf4␣, CAGGCreERTM transgenic mice were generated by crossing Hnf4␣ floxed/floxed mice (a gift from Frank Gonzalez, NIH)(23), with CAGGCreER TM ;Hnf4␣ floxed/ϩ (48) mice to allow systemic, tamoxifen-inducible knock out of HNF4␣. 6 -8-week-old CAGGCreER TM ;Hnf4␣ floxed/floxed mice and CAGGCreER TM ;Hnf4␣ floxed/ϩ littermate controls were injected intraperitoneally with tamoxifen (5 mg/20g body weight, 5 consecutive days) to induce cre-mediated Hnf4␣ deletion. Mice were sacrificed 4 weeks after first tamoxifen injection. No mouse samples were excluded from analysis in this study. Immunofluorescent Quantification: For quantification of ER in islets, the pancreata of HNF4␣ KO and littermate control heterozygous mice were fixed, mounted, and stained as described above. 16-bit images captured in Zeiss Axiovision software were analyzed with ImageJ software as follows; Insulin positive regions (Santa Cruz Biotechnology rabbit anti-insulin) were measured as regions of interest, and mean fluorescence intensity of the global ER marker Calregulin in each region was determined and then subtracted from the median fluorescent intensity of acinar cell regions in the same image to normalize fluorescence intensity on each slide. The mean fluorescent intensity was measured in every islet in a 5-m section (three mice/condition). Analysis of Calregulin fluorescence was restricted to the ␤-cell cytoplasm by excluding Hoechst-positive (nuclear) regions. After capture, each image was assigned a random number so that subsequent fluorescent quantification was blind relative to condition. Nuclear areas were identified by Hoescht staining, and pixels with an intensity of Ͼ30 gray value as determined by the "plot profile" tool in ImageJ were excluded from measurement. Hoescht-negative, cytoplasmic pixels were measured and normalized by subtracting the mean fluorescence of the surrounding acinar tissue.
Endoplasmic Reticulum Calcium FRET Measurement-INS-1 832/13 cells stably expressing the D1ER calcium sensor (12,34), were cultured as described above. Cells were transiently transfected with 10 nM HNF4␣ siRNA or scrambled control siRNA to knockdown HNF4␣ expression. Cells were also treated with vehicle or 16 M 4-methyl umbelliferone 8-carbaldehyde (48C) to inhibit XBP1 splicing as previously described (50). Five days post-transfection/treatment, 1 ϫ 10 5 cells were seeded in transparent bottom 96-well plates to achieve 70% confluency 6 h pre-measurement. Cells were washed twice in PBS and incubated in Krebs-Ringer buffer supplemented with 3 mM glucose immediately before measurement. To establish maximum and minimum ER Ca 2ϩ levels, at this time, cells were respectively treated with 10 M membranepermeabalizing ionomycin and subsequently 10 mM CaCl 2 or 5 mM EGTA. As an additional control for low ER Ca 2ϩ , cells were treated with 1 M thapsigargin to specifically diminish ER Ca 2ϩ . FRET ratio of the D1ER cameleon was measured using the Tecan Infinite M1000Pro microplate reader. Fluorophores were excited at 434 nm, and emission was quantified at 530 nm (YFP) and 477 nm (CFP). The ratios were measured across 4 fields/well, and the values were averaged from 4 wells per experimental condition. After FRET microplate measurement, RNA was isolated as described above and quantified by qPCR to measure Hnf4␣ and Xbp1 knockdown efficiency.
Ratiometric Calcium Imaging and Data Analysis-Studies were performed 24 h after plating INS-1 823/13 cells at 50% confluency and carried out at 37°C with 5% CO2 in a perifusion chamber with a flowrate of 2 ml/min. Cells were loaded with Fura-2AM by incubation at 37°C in Krebs-Ringer buffer supplemented with 3 mM glucose, 1 M Fura-2AM, and 0.1% Pluronic F-127 for 30 min, washed in HBSS, and incubated for another 30 min to allow for ester hydrolysis. After loading, cells were imaged on an inverted microscope (Till Photonics; Munich, Germany) equipped with a cooled CCD camera (Cooke, Auburn Hill, MI) using a ϫ20/0.45 Plan Fluor objective (Nikon). The fluorescence excitation (340 and 380 nm) was provided by a Polychrome V Monochromator (Till Photonics). After the matching background was subtracted, the image intensities from each pair of images, measured at 520 nm, were divided by one another to yield ratio values for individual cells.
[Ca 2ϩ ] i in individual cells was estimated based on the formula: ; R is ratio of fluorescence intensity at two different wavelengths (340/380 nm); R max and R min are the ratios of Ca 2ϩ -free and Ca 2ϩ -bound Fura-2, respectively; and B is the ratio of the fluorescence intensity of the second excitation wavelength at zero and saturating Ca 2ϩ concentrations. The calibration constants were determined as previously described (51), and the ratio values were plotted against time.
Islet Isolation and Culture-Pancreatic islets from CAG-GCreERTM;Hnf4␣ floxed/floxed mice and CAGGCreERTM; Hnf4␣ floxed/ϩ littermate controls were isolated as previously described (52), by pancreatic duct injection of 1000 U/ml of collagenase solution (Sigma) followed by digestion at 37°C for 15 min with mild shaking. Islets were washed several times with Hanks' balanced salt solution, separated from acinar cells by straining through a 100-m filter, viewed under a dissecting microscope, and handpicked for culture (yield ϭ 200 -300 islets/mouse). Isolated Islets were maintained in RPMI1640 supplemented with 10% FBS and penicillin and streptomycin at 37°with 5% CO 2 . All islets were allowed to recover from isolation for 24 h before analysis. Islets were isolated from mice in random order relative to condition.
Adenoviral Transduction-Unspliced XBP1 adenovirus (Applied Biological Materials), LacZ Adenovirus (Applied Biological Materials), and spliced XBP1 (a gift from Laurie Glimcher) (53), were amplified in HEK293 cells, cultured as described above. Infected cells were lysed by three cycles of freezing and thawing and then centrifuged. Viral titer was determined by infecting HEK-293 cells with serially diluted viral stock and overlaying with agar and subsequently counting the resulting plaques. INS-1 cells containing dnHNF4␣ were treated with doxycycline or vehicle to induce expression of dnHNF4␣ as described above. Five days post-treatment, cells were infected with either XBP1u or LacZ adenovirus at a mul-tiplicity of infection (MOI) of 100. Viral stock was replaced with complete medium after 2 h of infection. Isolated murine islets were infected as described previously (54). Briefly, 70 islets/condition were washed in cold PBS, pretreated with HBSS containing 2 mM EGTA at 37°with 5% CO 2 for 15 min, then infected with adenovirus in serum-free RPMI 1640. Following a 15-min incubation, complete medium was added to islet culture. Islets were infected for 24 h before GSIS assay and harvesting RNA. Adenovirus was used in isolated islets at the following MOIs: LacZ MOI ϭ 50, XBP1u MOI ϭ 50, XBP1s MOI ϭ 10.
Glucose-stimulated Insulin Secretion Measurement-For INS-1 GSIS assay, 2 days post-adenoviral infection cells were washed with PBS, then incubated for 1 h in Krebs-Ringer Buffer containing 3 mM glucose. After 1 h, cells were washed with PBS, and basal insulin secretion was measured by incubating cells for 1 h in Krebs-Ringer Buffer containing 3 mM glucose. Medium was sampled, then replaced with medium containing 16.7 mM glucose, or 200 M Tolbutamide for 1 h. Medium was collected and analyzed for insulin content by ELISA using the Singulex Erenna platform by the Washington University Diabetes Research Center Immunoassay Core. Static GSIS was similarly measured in isolated islets as previously described (55). 24 h post-infection, fifty islets were placed in Krebs-Ringer Buffer containing 3 mM glucose to measure basal insulin secretion, then stimulated with 16 mM glucose. Insulin secretion was measured in each condition as described above.
Graphing and Statistics-All graph values represent the mean of the sample, and error bars represent S.E. where indicated. Significance was determined using Student's t test or ANOVA with Dunnet's posthoc comparison as indicated. Wherever possible, samples were randomized and measurements were blinded to prevent the introduction of experimental bias. Sample sizes were determined based on statistical significance and practicality.

Results
Regulation of Xbp1 by HNF4␣-To elucidate the potential transcriptional regulation of XBP1, we identified evolutionarily conserved binding sites in the human XBP1 promoter by aligning regions of synteny, then screening them with the Transfac transcription factor binding site database (18). Two regions with high conservation containing putative HNF4␣ binding sites, 1.4 and 2.6 kilobases upstream of the Xbp1 transcription start site, were identified using first the Transfac transcription factor binding library and then affirmed using a previously published algorithm developed to search for sites of high HNF4␣ binding affinity (19). These putative binding sites were constitutively occupied by HNF4␣ in mouse pancreas, measured via chromatin immunoprecipitation (Fig. 1A). Overexpression and knockdown experiments in vitro showed HNF4␣ was both sufficient and necessary for normal Xbp1 expression in pancreatic ␤-cell derived-cell lines. Disrupting HNF4␣ either by siRNA knockdown (Fig. 1, B and C) or by overexpressing a doxycycline-inducible dominant-negative version of HNF4␣ that lacks a DNA-binding domain and has been shown to bind to endogenous HNF4␣ (Fig. 1, D and E) (36), resulted in a 65-75% decrease in Xbp1 expression in INS-1 and MIN-6 cells. Conversely, overexpression of HNF4␣ via transient transfection caused a 5-fold increase in Xbp1 expression (Fig. 1F). To further substantiate this transcriptional relationship, we analyzed the effects of Hnf4␣ deletion on Xbp1 expression in other tissues by mining published microarray studies and by direct qRT-PCR analysis of adult (Fig. 1G) and embryonic liver (20) as well as adult small intestine (data not shown but available on Gene Expression Omnibus under accession numbers GSE34581, GSE3124, GSE3126) (21)(22)(23). Again, the results showed a consistent trend toward correlation of Xbp1 expression decrease with loss of Hnf4␣ in multiple secretory tissues.
Various single point mutations in the HNF4␣ locus have been identified in human patients afflicted with impaired ␤-cell function. To better understand the impact of these mutations, we designed two HNF4␣ expression vectors, each containing one of the most prevalent mutations (24,25). Overexpression of each of these individual mutants in INS-1 cells resulted in a ϳ4-fold decrease in the expression of XBP1 mRNA (Fig. 1H). Interestingly, although we could detect the transfection-mediated increase in mRNA for both mutant and wild-type HNF4␣ qPCR, we could detect only wild-type HNF4␣ protein by Western blot ( Fig. 4D and data not shown). We speculate the mutant genes may encode protein products that are abnormally folded and/or truncated and degraded. In any case, all the data support the conclusion that common HNF4␣ mutations found in human patients decrease abundance of Xbp1.
HNF4␣ Is Required for ER Maintenance in Vivo-The mechanisms whereby disruptions of HNF4␣ in MODYI cause ␤-cell dysfunction remain an area of open debate. As XBP1 is critical in scaling up and maintaining the ER of professional secretory cells, we hypothesized mutations in HNF4␣ may impair ␤-cell function via dysregulation of Xbp1 and consequent ER dysfunction. Previous studies of effects of loss of Hnf4␣ in islets focused on deletion of the gene during development. As we were concerned here with the effects of adult onset of aberrant Hnf4␣ on islet function, we induced deletion of Hnf4␣ flox/flow mice using a global, tamoxifen-inducible Cre in 6 -8-week-old mice. As predicted, deleting Hnf4␣ in existing islets resulted in significant loss (ϳ60%) of both Xbp1 expression in islets and, in turn, of XBP1 transcriptional targets like Edem1 ( Fig. 2A) when compared with littermate Hnf4␣ flox/ϩ controls (referred to hereafter as "WT") (26). Supporting previous findings, insulin and Hnf1␣ mRNA levels were unaffected by loss of Hnf4␣ (10,11). While XBP1 establishes secretory cell machinery and maintains cell architecture (2,27), likely in large part independent of activation of the UPR (28), in chronic/long-term ER stress conditions, XBP1 is a fundamental component of the UPR (29). We sought to characterize the ER-stress state of ⌬HNF4␣ ␤-cells to determine whether other branches of the UPR were activated to compensate for loss of Xbp1. One of the most reliable methods of measuring UPR activation in cells is by quantifying the expression of genes whose transcription is enhanced by activated master regulators of the UPR (30). Decreased Xbp1 in Hnf4␣ ⌬/⌬ mice did not cause increase in such UPR transcripts like Chop, Bip, or Atf4, nor a decrease in mRNA levels of the ER-marking Calregulin (CRP55) (Fig. 2B). However, loss of XBP1 following deletion of Hnf4␣ did correlate with a nearly 7-fold reduction in ER area in ␤cells (Fig. 2, C and D). Despite decreased ER in each cell, overall islet mass, as measured by HNF4␣ maintains ER homeostasis in ␤-cells through XBP1 MARCH 18, 2016 • VOLUME 291 • NUMBER 12 JOURNAL OF BIOLOGICAL CHEMISTRY 6149 microscopic area through cross-sectioned pancreata, was not changed in Hnf4␣ ⌬/⌬ pancreata (Fig. 2E), indicating that Hnf4␣ is not required to maintain islet number or size. Thus, loss of Hnf4␣ caused diminished ER network, a phenotype similar to that caused by deleting Xbp1 from existing adult secretory cells (2,5).
HNF4␣ and XBP1 Are Necessary to Maintain ER Calcium Homeostasis-In previous reports, constitutive deletion of Hnf4␣ from islets early in development, as opposed to in the adult, caused impaired GSIS (10,11). The mechanism of decreased GSIS was hypothesized to be dysregulated cytoplasmic Ca 2ϩ signaling in response to glucose, but the molecular mechanism driving this impairment has remained unclear. Ca 2ϩ signaling depends on the ATP-dependent closure of KATP channels, triggering membrane depolarization and opening voltage-gated Ca 2ϩ channels (31). Thus, one mecha-nism that could mediate how loss of Hnf4␣ could cause GSIS could be via disruption of those channels. However the sulfonylurea Tolbutamide, which closes KATP channels leading to membrane depolarization, stimulated less insulin secretion in INS-1 cells expressing dominant negative-HNF4␣ (dnHNF4␣, described above) than in normal INS-1 cells (Fig. 3). The inhibition of Tolbutamide-stimulated insulin secretion caused by disrupted HNF4␣ was rescued by transduction of Xbp1, though transducing Xbp1 alone did not increase tolbutamide-stimulated insulin secretion (Fig. 3). Additionally, qPCR analysis showed abundance of transcripts for the KATP channel subunits Sur1 (Abbc8) and Kir6.2 (Kcnj11) in Hnf4␣⌬/⌬ islets were unchanged relative to littermate controls (Fig. 4A). These data, along with previous work showing impaired insulin secretion even upon depolarization with KCl (10), suggest the defect in the glucose response pathway due to disrupted HNF4␣ is distal to KATP channel dependent-membrane depolarization and that it depends, in part, on loss of XBP1. Though KATP channel expression was unchanged in Hnf4␣ islets, expression of the putative XBP1 transcriptional target, Serca2b(Atp2a2), was significantly reduced (Fig. 4A) (32). SERCA2b is an ER Ca 2ϩ pump, responsible for establishing and maintaining the large calcium gradient between the ER and cytoplasm (33). Intracellular stores of Ca 2ϩ are critical for GSIS, because Ca 2ϩ release from these stores triggers secretion of insulin granules. Accordingly, decreasing ER Ca 2ϩ stores and/or flux has been shown to disrupt GSIS (17). Thus, in Hnf4␣ ⌬/⌬ mice, the decreased expression of a key molecular driver of the ER Ca 2ϩ gradient suggested that disruption of ER Ca 2ϩ stores may play a critical role in the GSIS abnormalities seen in the absence of normal HNF4␣. We used an ER-specific FRET sensor to measure ER [Ca 2ϩ ] in ⌬Hnf4␣ ␤-cells (12,34

JOURNAL OF BIOLOGICAL CHEMISTRY 6151
other MODYI ␤-cell models (Fig. 4E). To identify the cause of this deficit in Ca 2ϩ signaling, and further explore our previous results indicating HNF4␣ is required for ER Ca 2ϩ homeostasis, we exposed these cells to 20 mM Caffeine. Caffeine is an agonist of the ryanodine receptor that stimulates release of Ca 2ϩ from stores in the ER and thus an increase in cytoplasmic [Ca 2ϩ ] (35). Caffeine induced diminished cytoplasmic [Ca 2ϩ ] increase in ⌬Hnf4␣ ␤-cells (Fig. 4F), indicating that ER Ca 2ϩ homeostasis is disrupted. In short, loss of ER Ca 2ϩ in ⌬Hnf4␣ ␤-cells may underlie their impaired GSIS. XBP1 Is Sufficient to Rescue Insulin Secretion in ⌬Hnf4␣ ␤-cells-We next sought to confirm the physiological relevance of the HNF4␣ 3 XBP1 relationship by determining if we could rescue aberrant GSIS in the absence of HNF4␣ simply by restoring XBP1. If XBP1 were decreased in the absence of HNF4␣ due to direct loss of transcriptional up-regulation by HNF4␣, as we hypothesized, then restoration specifically of the unspliced XBP1 (XBP1u) form should rescue GSIS, because XBP1u is the unmodified mRNA directly generated from transcription of the XBP1 gene (i.e. prior to activation by IRE1mediated splicing). Restoring the deficit of XBP1u transcript caused by loss of HNF4␣ would allow cells the freedom to use the normal IRE1 splicing mechanism to produce only as much active XBP1s transcript as they require for ER homeostasis. We returned to our in vitro model employing transgenic INS-1 cells containing a doxycycline inducible dnHNF4␣ described above (Figs. 1 and 3) (36). As shown above, the functional loss of HNF4␣ activity in insulin-secreting cells when doxycycline is added to induce dnHNF4␣, caused both decreased XBP1 (Fig.  1, D and E) and loss of GSIS (Fig. 5A). This decreased GSIS was completely rescued by adenoviral transduction of Xbp1u. As discussed above, forced expression of Xbp1u was also sufficient to rescue impaired insulin secretion in response to sulfonylurea treatment in these dnHNF4␣ ␤-cells (Fig. 3). We repeated this study ex vivo, using Hnf4␣ ⌬/⌬ islets isolated 3 weeks following tamoxifen-induced Hnf4␣ ⌬/⌬ deletion. As expected, cultured islets had impaired GSIS due to HNF4␣ deficiency. Xbp1u restoration increased the direct XBP1 transcriptional targets Edem1 and Serca2b (Fig. 5B) confirming that transduction of Xbp1u restored functional XBP1-mediated transcriptional activity to scale up expression of its normal transcriptional targets. Remarkably, as predicted, restoring unspliced Xbp1 expression in these ex vivo cultured, HNF4␣-deficient ␤-cells was also sufficient to completely rescue their GSIS, indicating that the impaired GSIS in the absence of HNF4␣ depends on transcriptional maintenance of XBP1 expression by HNF4␣ (Fig. 5C). Because of the direct transcriptional regulation of XBP1 by HNF4␣ and the lack of GSIS enhancement in WT ␤-cells transduced with Xbp1u, the ability of Xbp1u to rescue the phenotype caused by loss of HNF4␣ is likely because it corrects the diminished basal Xbp1 expression in ⌬HNF4␣ ␤-cells.
We also transduced spliced XBP1 (Xbp1s) in isolated ⌬HNF4␣ mouse islets, bypassing the normal regulation of transcriptionally regulated Xbp1u by IRE1␣ splicing. Transduction of Xbp1s rescued the XBP1 targets, Edem1 and Serca2b (Fig.  6A) but resulted in GSIS roughly 50% lower than that in control WT islets (Fig. 6B). That result is consistent with previous reports that ␤-cell homeostasis is compromised by forced expression of spliced Xbp1 because cells must be able to dynamically regulate XBP1 levels via the endogenous IRE1 splicing mechanism, with direct forced expression of high levels of already activated Xbp1 potentially being toxic (37). Accordingly, Hnf4␣ ⌬/⌬ islets infected with Xbp1s exhibited GSIS rescue to the levels observed in WT islets infected with Xbp1s, suggesting that, while forced expression of XBP1s is detrimental to ␤-cell health, it is still able to compensate for GSIS defects in ␤-cells caused by the absence of HNF4␣.

Discussion
We report that Xbp1 is a direct transcriptional target of HNF4␣ in multiple secretory tissues. Given the importance of HNF4␣ mutations in diabetes, we have focused on the relationship between HNF4␣ and XBP1 specifically in insulin-secreting ␤-cells. Deletion of HNF4␣ in ␤-cells causes them to lose XBP1, which in turn causes dismantling of ER. HNF4␣ point mutants designed to match mutations that cause human diabetes also resulted in loss of XBP1 in vitro. Loss of either HNF4␣ or XBP1 leads to diminished ER Ca 2ϩ , which in turn impairs GSIS, a defect that can be completely rescued by reestablishing normal Xbp1 transcript levels (Fig. 5D). Together, our results identify a new transcriptional relationship between evolutionarily-conserved genes, Xbp1 and Hnf4␣ ⌬/⌬ , involved in fundamental development and disease in multiple tissues.
We also demonstrate specific cellular contexts during which Xbp1 expression is functionally regulated at the transcriptional level. XBP1 is induced in response to accumulation of unfolded protein in the ER by splicing of its message via the endonuclease IRE1␣, and the canonical view of how XBP1 abundance is modulated concern that mechanism. However, the transcriptional regulation of Xbp1 may be as important to its role in secretory cell development as its activation by IRE1. Xbp1 expression is significantly induced and maintained at much higher levels in  secretory cells, and large pools of unspliced Xbp1 mRNA are required to restore homeostasis in chronic ER stress conditions (38). Additionally, IRE1 is basally activated in dedicated secretory cells at levels comparable to those observed during acute ER stress in non-secretory cells (Ͼ40%) (39). When the presumably UPR-activating biosynthetic load-stimulus is removed in 〉-lymphocytes differentiating to plasma cells by disrupting expression of IgM, XBP1 transcript levels still increase, indicating that XBP1 activation is differentiation-dependent rather than UPR-dependent in certain secretory cells (40). Together, these studies suggest that transcription of Xbp1 is the rate-limiting step in its activation during secretory cell development and maintenance.
It is somewhat surprising that HNF4␣, which is largely studied in developmental contexts as a master regulator of differentiation in endodermal organs, is required for continued maintenance of XBP1 in differentiated, adult cells. However, unbiased, comprehensive screens for genes whose expression depends on HNF4␣ have previously identified XBP1 as a potential target (14)(15)(16)(17), and chromatin immunoprecipitation fol-   lowed by sequencing (ChIP-Seq) has also shown peaks indicating potential binding of HNF4␣ to the putative XBP1 promoter (41). Thus, though the results of the previous screens have not been validated and the direct relationship between HNF4␣ and XBP1 has apparently never been specifically studied prior to the current work, our results are not entirely unprecedented. Indeed, HNF4␣ has also been shown to regulate expression of ankyrin repeat and sterile ␣ motif domain containing 4b (Anks4b), a protein that binds ER chaperones and augments the ER stress response, supporting our hypothesis that HNF4␣ is required for the establishment and maintenance of the ER in ␤-cells (42). In ongoing work in our lab, we also observe that Xbp1 expression depends on HNF4␣ in the stomach, and, exactly as in ␤-cells, loss of HNF4␣ in mature gastric chief cells causes dismantling of the ER. 3 The results are consistent with work by us and others showing that continued XBP1 is required for ER in gastric chief cells (2) and in ␤ cells (16). On the other hand, though we observe Xbp1 expression clearly depends on HNF4␣ in the liver (see Fig. 1G plus previous studies available in GEO that we have analyzed (21-23)), we have not been able to detect substantial effects on ER in that organ when HNF4␣ is deleted from mature hepatocytes (data not shown). Actually, this is in agreement with previous studies showing that direct depletion of XBP1 itself from mature hepatocytes is not sufficient to cause ER dismantling in existing cells (53). Together, the results suggest that there are tissue-specific differences in the role of XBP1 in regulating the ER: perhaps XBP1 is generally required for establishment of an elaborate ER network in dedicated secretory cells but is only required in certain cell types for its continued, homeostatic maintenance.
While previous studies suggest alterations in the KATP channel may play a role in MODYI pathology (10,11) they also show that MODYI ␤-cells have impaired GSIS even upon membrane depolarization with KCL, suggesting the cause of the diminished GSIS is distal to the KATP channel/membrane depolarization event. ER Ca 2ϩ homeostasis is important for myriad cellular processes and plays a pivotal role in intracellular Ca 2ϩ signaling. Our data indicate that XBP1 and HNF4␣ are required for maintaining this homeostasis. Though inflamma-   tion models have shown that activation of XBP1 leads to expansion of ER calcium stores in bronchial secretory cells (43), to our knowledge we present the first evidence that XBP1 is required to maintain ER Ca 2ϩ stores. The targets whose expression are dictated by XBP1/HNF4␣ and help maintain ER homeostasis are not clear, but one such candidate target may be SERCA2b. Depleting ER calcium stores by inhibiting SERCA2b with thapsigargin reduces GSIS, similar to our MODYI model (33 4, A, C, F). These data outline a potential mechanism wherein altered ER Ca 2ϩ homeostasis from loss of HNF4␣ function disrupts Ca 2ϩ signaling and insulin release in ␤-cells of patients with MODYI. There are no faithful animal models of MODYI. Previously mouse models have completely disrupted HNF4␣ expression in ␤-cells constitutively, characterizing its role in development (10,11), rather than in the adult mouse, highlighting its role in ␤-cell maintenance as we show above. Some mutations in HNF4␣ in humans result in alleles thought to cause diabetes via a dominant negative mechanism (24,25), whereas others would be expected to act via haploinsufficiency (44,45). Thus, it may be premature to reach any firm conclusions about MODYI mechanisms based on our results using known MODYI-inducing human polymorphisms and adult-onset knock-out of Hnf4␣ in mice. However, if we are to speculate on the implications of our findings, we might suggest that they indicate consideration of a new angle on MODYI therapy. Currently, like type II diabetes, MODYI is responsive to treatment with sulfonylureas, though treatment often eventually involves insulin therapy to manage hyperglycemia, presumably because ␤-cells eventually become dysfunctional or die (46). Our results suggest that various therapies targeting ER homeostasis, a rapidly developing therapeutic avenue with several drugs at various stages of development, may augment or improve existing approaches to managing MODYI.  and heterozygote control islets were isolated and treated as in (Fig. 5B) but infected with adenovirus harboring plasmid a transcript for spliced XBP1 (XBP1s). XBP1s expression restores mRNA abundance of downstream targets, Edem and Serca2b, to normal levels in ⌬HNF4␣ islets ( means Ϯ S.E. of n ϭ 3 experiments depicted). B, restoring expression of XBP1s sufficient to partially rescue GSIS in isolated ⌬HNF4␣ islets, but GSIS (i.e. insulin release following switching of islets to 16 mM glucose) is diminished in both WT and ⌬Hnf4␣ islets upon transducing expression of XBP1s when compared with WT GSIS. LacZ vector-containing adenovirus was used as a control for transduction (means Ϯ S.E. of n ϭ 3 experiments depicted, statistical significance was determined by one-tailed Student's t test).