Transient PAX8 Expression in Islets During Pregnancy Correlates With β-Cell Survival, Revealing a Novel Candidate Gene in Gestational Diabetes Mellitus

During gestation, maternal metabolic adaptations are essential to ensure the health of the mother and the viability of the fetus. One of the organs that must undergo a profound adaptation during pregnancy is the endocrine pancreas. The islet β-cell mass expands as an adaptive response to the progressive insulin resistance that develops in the pregnant female to favor nutrient accessibility to the fetus (1). This mechanism of β-cell mass expansion includes β-cell hypertrophy, increased β-cell proliferation, and increased prosurvival/antiapoptotic signaling (2). Failure to adapt may result in the development of gestational diabetes mellitus (GDM), with concomitant health issues not only for the mother but also for the fetus. The underlying molecular mechanisms triggering GDM remain largely uncharted (3). Toward identifying these molecular pathways, two independent studies have analyzed the transcriptome landscape of islets isolated from gestating mice (2,4). Of note, one of the most upregulated genes was the transcription factor paired box 8 (PAX8) (2). Historically, PAX8 is known for its essential role in the development and function of the thyroid and urogenital system, maintaining its expression in adult kidney and thyroid gland (5). In agreement with this described pattern of expression, mutations/polymorphisms in PAX8 have been associated with hypothyroidism and urorenal abnormalities, whereas its overexpression correlates with tumor formation in these and other tissues (6,7).

We have previously reported that Pax8 is expressed neither during mouse pancreatic development nor in adult islets under normal physiological conditions (8,9). This fact raises the prospect that PAX8 expression likely will be induced transiently under specific physiological conditions, such as pregnancy, which may confer benefits to islets. Consistent with this beneficial effect of PAX8 under situations of increased insulin demand, a single genome-wide linkage and admixture mapping study has highlighted PAX8 as a putative type 2 diabetes mellitus (T2DM) candidate gene in African American descendants (10). In the current study, we sought to validate PAX8 expression and determine its physiological function during pregnancy. We observed that PAX8 is transiently expressed in mouse under gestating conditions as well as in human islets exposed to conditions mimicking gestation in vitro. Lentiviral-mediated PAX8 overexpression in islets modulates genetic pathways–implicated immunomodulation, protein processing, and cell survival. More importantly, we identified a yet undescribed missense mutation of PAX8 (p.T356M, c.1067C>T) in a family pedigree in which a female developed GDM while her father developed T2DM. Functional analysis demonstrated that the p.T356M mutation significantly decreased PAX8 transactivation potential. In addition, a retrospective clinical analysis uncovered that GDM developed in a pregnant female of a familial pedigree that harbors a previously described PAX8-P25R mutation associated with congenital hypothyroidism.

C57BL/6J mice were purchased from JANVIER LABS (Saint-Berthevin, France). Five-month-old C57BL/6J pregnant females were euthanized at post coitum day (pcd) 10.5, 14.5, 16.5, and 18.5 for islet isolation. Pax8 knockout mice were procured from the European Mouse Mutant Archive network. Mouse experiments were approved by the Centro Andaluz de Biología Molecular y Medicina Regenerativa (CABIMER) animal committee and performed in accordance with Spanish law on animal use (RD 53/2013).

Pregnant wild-type (Wt) and heterozygous Pax8 mice were subjected to an oral glucose tolerance test (OGTT) at pcd 14.5 as previously reported (11).

Mouse islets were isolated using the intraductal collagenase perfusion protocol (12,13). Human islets were procured from the Cell Isolation and Transplantation Center (Geneva, Switzerland). Characteristics of human islet preparations are presented in Supplementary Table 1. Islets were cultured as previously described (14). Where indicated, murine islets were treated for 16 h with cytokines at 48 h after lentiviral infection (interferon-γ [IFN-γ 1,000 units/mL, interleukin-1β [IL-1β 100 units/mL, and tumor necrosis factor-α [TNF-α 1,000 units/mL). Where indicated, human islets were treated with recombinant human prolactin (200 ng/mL daily) for up to 96 h (Sigma-Aldrich, Madrid, Spain). Before treatments, human islets were precultured in media supplemented with charcoal-treated serum for 2 days to remove hormonal contaminants.

Total RNA was extracted using the RNeasy Micro Kit (QIAGEN). cDNA using 0.5–1 μg RNA was synthesized using SuperScript II Reverse Transcriptase (Thermo Fisher Scientific, Madrid, Spain). The RT-PCR was performed on individual cDNAs using SYBR Green (Roche Diagnostics, Mannheim, Germany). Primer sequences are presented in Supplementary Table 2. The mRNA expression was calculated by the 2^−ΔΔCT method and normalized to the expression of cyclophilin, RPS29, and/or β-actin (12).

The human and murine PAX8Wt cDNAs (Origene, Rockville, MD) were subcloned into the pHRSIN DUAL-GFP easy vector provided by J.A. Pintor-Toro (CABIMER). To generate control lentiviruses for mock transductions, the empty vector was used. This vector allows the cloning and expression of a gene of interest under the control of the spleen focus-forming virus promoter while the constitutive ubiquitin promoter regulates expression of the reporter green fluorescent protein (GFP). Lentivirus amplification, purification, and transduction were performed as previously described (13). Islet transduction efficiency was estimated by flow cytometry (FACSCalibur; BD Biosciences, Madrid, Spain) using ∼20 disaggregated islets.

Apoptosis and proliferation were assessed using cell death detection and BrdU cell proliferation ELISA, respectively (catalog #11647229001; Sigma-Aldrich, Madrid, Spain). BrdU (10 μmol/L) was added during the last 24 h of the assay at each time point. Glucose-stimulated insulin secretion (GSIS) was performed as previously described (14).

Islet embedding and immunohistochemistry analysis were performed as previously described (12). A list of primary and secondary antibodies is provided in Supplementary Table 3. Images were acquired using a Leica microscope (AF6000; Leica, Milton Keynes, U.K.).

cRNA preparations from human and murine islets were hybridized to GeneChip Human Gene 2.0 ST Array Chip and GeneChip Mouse Gene 1.0 ST Array Chip, respectively (Affymetrix, Santa Clara, CA), using the standard protocols of the Genomic Core Facility, CABIMER. Image analysis, quality control, and data quantification were performed using Affymetrix GeneChip Command Console version 4.0 software. Additional statistical data analysis was performed using the Transcriptome Analysis Console (TAC) developed by Affymetrix. Modulated pathways using TAC software were obtained by applying significantly modulated genes and pathways with at least three modulated genes. The Ingenuity Pathway Analysis platform was used to determine modulations on annotated canonical pathways and annotated diseases and functions using the significantly modulated genes.

Women in the second and third trimester of gestation were invited to participate in the genetic screening of PAX8 if they were diagnosed with GDM by OGTT per defined by international criteria or by HbA_1c >6.5% (48 mmol/mol) (criterion 1) (15). Indexed patients were then selected on the basis of hypothyroidism defined as free levothyroxine (FT4) <0.90 ng/dL or thyroid stimulating hormone (TSH) >2.5 mU/L during the first trimester of pregnancy or TSH >3 mU/L during the second to third trimesters of pregnancy of nonautoimmune nature (absence of circulating thyroglobulin [TG] and thyroperoxidase antibodies) (criterion 2). The rationale for this second criterion was to increase the probability of identifying PAX8 mutations, a factor for which mutations are associated with thyroid disorders. Exclusion criteria were 1) patients not taking iodine supplements and 2) patients with a BMI >30 kg/m².

In addition, accessible relatives of the index patient in familial pedigree 1 (T356M) also were invited to participate in the genetic screening. After obtaining informed consent at Hospital La Paz, Madrid, using protocols approved by the local ethics committee, blood was collected and genomic DNA extracted from peripheral leukocytes using the Chemagen DNA blood kit in the Chemagic automated system (PerkinElmer). The entire coding sequence of the human PAX8 gene (reference cDNA: NM_003466.3; Genome location chr2: 113992991) was amplified by PCR under standard conditions. DNA sequencing was performed in an ABI 370 sequencer and analyzed with Sequencher software.

Site-directed mutagenesis (GenScript, Piscataway Township, NJ) was performed on the human PAX8 cDNA to generate mutant variants PAX8-P25R and PAX8-T356M (16). These variants were subcloned into the pCDNA3.1 vector, and the fidelity of the mutated constructs was confirmed by sequencing. To assess protein stability and subcellular localization, MCF-7 cells that do not express PAX8 (8) were transfected with the human PAX8Wt, T356M, and P25R using lipofectamine (Thermo Fisher Scientific) according to the manufacturer’s protocol. Twenty-four hours posttransfection, cells were either treated or not with cycloheximide (80 μg/mL) for 24 and 48 h. Protein extracts were then prepared and Western blotting performed against PAX8. Alternatively, transfected cells were processed for immunofluorescence analysis as previously described (12).

The human TG gene promoter, a PAX8 target, was subcloned into the firefly luciferase reported plasmid (17) and transfected in HEK293T cells along with expression vectors for thyroid transcription factor 1 (TTF1 [also known as NKX2.1]) in combination with PAX8Wt, T356M, or P25R using the FuGENE 6 reagent (Promega Biotech Ibérica, Madrid, Spain). To correct for transfection efficiency, Renilla-encoding pRL-cytomegalovirus vector was cotransfected in all cases. Cellular extracts were collected 48 h posttransfection, and Dual-Luciferase Reporter Assay (Promega) was performed according to the manufacturer’s instructions. The ratio between the luciferase and Renilla activities was expressed relative to the ratio obtained in cells transfected with reporter and empty expression vector (pCDNA3).

Data are shown as mean ± SD. Paired and unpaired Student t test, Wilcoxon rank sum test, Kruskal-Wallis one-way ANOVA, and one-way ANOVA were performed when appropiate using SigmaPlot 12.0 software (SigmaPlot, Barcelona, Spain).

We initially validated that PAX8 expression was transiently induced in mouse islets during pregnancy, reaching maximal fold expression at pcd 14.5 and thereafter returning to basal levels by pcd 18.5 (Fig. 1A). As previously reported, Tph1 transcript levels also were increased, whereas the Pdx1 transcript did not show significant alterations (4) (Fig. 1B and C). To extend these findings to human, isolated human islets were treated with recombinant human prolactin to mimic conditions of the third trimester of pregnancy (18). Prolactin produced a significant 2.5-fold increase in PAX8 expression at 72 h of treatment compared with untreated islets (Fig. 1D). TPH1 expression also was increased, starting at 24 h of treatment, whereas no significant changes were observed in PDX1 expression levels (Fig. 1E and F). Prolactin treatment caused a rapid and transient increase in the proliferation rate of human islet cells, peaking at 24 h of treatment (Fig. 1G). Of note, the PAX8 expression pattern inversely correlated with proliferation rates, suggesting that the transcription factor is not involved in human islet cell replication. Taken together, PAX8 expression can be induced in human islets under conditions mimicking pregnancy.

To establish the putative function of PAX8 in the endocrine pancreas, we conducted global gene profiling of isolated human and mouse islets transduced with lentiviral constructs harboring either the human or the mouse PAX8 cDNA within a bicistronic cassette with GFP. The rationale of using both species-derived factors stemmed from our previous studies that revealed interspecies differences between human and murine PAX4 (19). Flow cytometry analysis demonstrated that 48 h after infection, ∼75% of islet cells expressed GFP (Fig. 2A and B). Principal component analysis (PCA) clearly separated islets overexpressing PAX8 from control-infected islets, illustrating the robust effect of PAX8 on overall gene expression (Fig. 2C and D). Although several genes were similarly regulated by PAX8 in both mouse and human islets (Fig. 2E), the functional enriched pathways were species specific (Fig. 2F and G and Supplementary Fig. 1A and B). Of note, we determined that lentiviral-mediated murine Pax8 and human PAX8 overexpression resulted in a similar modulation of the most upregulated and downregulated transcripts in murine islets (Fig. 2H and Supplementary Tables 4 and 5). Analysis of annotated “Disease and Functions” (Ingenuity Pathway Analysis module) confirmed that the transcriptome profile of mouse islets overexpressing Pax8 was associated predominantly with immune-related diseases, whereas the one from PAX8 overexpressing human islets included immune-related diseases; endocrine/metabolic disorders; and cell cycle, cell death, and cancer (Supplementary Fig. 1C and D and Supplementary Tables 6 and 7). To analyze the functional effect of PAX8 in islets, we studied the effect of either murine or human PAX8 overexpression in mouse islets. Functional cell metabolic activity as well as GSIS were unaltered in murine islets transduced with either the human or the mouse PAX8 compared with mock-transduced islets (Fig. 2I and J). Of note, overexpression of mouse PAX8 in murine islets reduced apoptosis by ∼40% under basal conditions and blunted cytokine-induced apoptosis (Fig. 2K and L). A similar reduction also was confirmed by immunostaining the apoptotic marker cleaved caspase-3 in transduced β-cells (Supplementary Fig. 2). Human PAX8 overexpression in both mouse (Fig. 2K) and human islets (Fig. 2M) also significantly decreased by ∼20% apoptosis. Collectively, these data suggest that both human and mouse PAX8 promote islet cell survival in both murine and human pancreatic islets.

Our data demonstrating that PAX8 is transiently induced in islets specifically during pregnancy combined with its prosurvival properties provide strong arguments for PAX8 as a candidate gene for the study of GDM. Toward addressing this possibility, we assessed the impact of Pax8 ablation in mouse on glucose homeostasis during gestation. Because Pax8^−/− pups die early postnatally (5) and are infertile despite thyroid hormone supplementation (20,21), we worked with Pax8 heterozygous mice (Pax8^+/−). Pax8^+/− females remained normoglycemic and did not suffer glucose intolerance throughout pregnancy (Supplementary Fig. 3). However, the fact that thyroid hormone levels, for which biosynthesis is directly regulated by PAX8, were similar in both Pax8^+/− and Pax8^+/+ (Wt) mice suggests a potential genetic compensation for Pax8 haploinsufficiency (data not shown). Consequently, we opted to perform a genetic screening for mutations in the PAX8 gene in patients with GDM and gestational thyroid dysfunction (GTD) (for family pedigrees, clinical characteristics, and glucose tolerance tests during pregnancy, see Supplementary Table 8 and Supplementary Figs. 4 and 5). The rationale of using patients with GTD, a disease associated with PAX8 mutations, was to optimize probabilities of identifying novel PAX8 gene mutations in a small cohort. Consistent with this premise, exon sequencing of PAX8 in members of seven pedigrees (pedigrees 1, 3, 4, 5, 6, 7, and 8) identified a woman who belongs to pedigree 1 and bears a novel heterozygous variant (c.1067C>T) (Fig. 3A and B). This missense mutation causes a threonine-to-methionine substitution at codon 356 (p.T356M) located at the carboxy-terminal region of the protein, distant from the majority of identified PAX8 mutations that are located within the DNA-binding paired domain (Fig. 3C). The index patient, a 30-year-old south occidental European Caucasian woman, was diagnosed with GTD in all four pregnancies and was prescribed FT4 (25–50 μg/day) (Table 1). Although no data are available for the first pregnancy, the patient exhibited impaired glucose tolerance after OGTTs in the other three pregnancies and was diagnosed with GDM according to National Diabetes Data Group (NDDG) and Carpenter-Coustan (CC) guidelines (22,23) (Fig. 3D–F). She could control hyperglycemia through a carbohydrate-restricted diet and physical exercise. To date, the patient has not been diagnosed with DM under nonpregnant conditions, and her HbA_1c levels are normal (Table 1). Her father, who also carries the mutation, was diagnosed with T2DM (HbA_1c 6.7% [50 mmol/mol]) at age 55 years (Fig. 3A). A 20-year-old sister of the index patient also harbors the mutation but has never been pregnant. Of note, numerous family members of the pedigree who are deceased and therefore not accessible for genotyping also were diagnosed with T2DM and/or nonautoimmune hypothyroidism (Fig. 3A).

To rule out the possibility that the c.1067C>T variant is a common PAX8 polymorphism in the general population, we genotyped 100 control chromosomes of south-occidental European Caucasians. None of these individuals possessed this variant (minor allele frequency <0.00034). Moreover, this mutation is not documented in the Exome Variant Server (http://evs.gs.washington.edu/EVS), Database of Single Nucleotide Polymorphisms (www.ncbi.nlm.nih.gov/snp), Exome Aggregation Consortium (http://exac.broadinstitute.org), and Genome Aggregation Database (http://gnomad.broadinstitute.org). In silico analysis using consensus classifier prediction tools for disease-related mutations supported a pathogenic role of p.T356M (24) (Supplementary Table 9). Consequently, we tested whether the PAX8-T356M mutation affected protein stability and/or subcellular localization. For these studies, we included the PAX8-P25R mutant variant (P25R, c.74C>G) associated with congenital hypothyroidism that lacks DNA binding activity but retains nuclear localization (16). Of note, a retrospective clinical analysis of clinical data and familial pedigree (pedigree 2) of the family harboring the P25R mutation revealed that the only gestating female bearing this mutation also developed GDM (Fig. 3G and H and Supplementary Table 8). A recent second pregnancy resulted in a miscarriage at week 7. The stability of both mutant proteins was similar to that of PAX8Wt in transfected MCF-7 cells treated with the protein synthesis inhibitor cycloheximide (Fig. 3I and J). Nuclear localization of the mutant variants also was unaltered (Fig. 3K).

We then assessed the functional impact of PAX8-P25R and PAX8-T356M mutations. To this end, we analyzed by luciferase reporter assay the transactivation potential of PAX8Wt, PAX8-P25R, and PAX8-T356M on the TG gene promoter. The PAX8 cofactor TTF1 (NKX2.1) also was included to determine whether these mutations have an impact on the synergistic transcriptional activation observed between the PAX8Wt and TTF1 (25). Accordingly, cotransfection of both PAX8Wt and TTF1 resulted in a marked increase in luciferase activity (Fig. 3L). In contrast, addition of PAX8-T356M or PAX8-P25R along with TTF1 resulted in ∼60% and ∼35% of PAX8Wt and TTF1 combined activity, respectively, establishing the compromised functionality of PAX8-T356M and PAX8-P25R compared with PAX8Wt (Fig. 3L).

We previously refuted that PAX8 is expressed either in murine adult islets or in human pancreatic neuroendocrine tumors, which raises serious concerns about a potential role of this PAX family member in islet physiology in health and disease (6). Notwithstanding, we report that PAX8 is induced in human islets treated with the lactogenic hormone prolactin and that forced expression improves islet survival. Moreover, we identified a novel PAX8 missense mutation, p.T356M (pedigree 1), in a heterozygous female diagnosed with GDM and GTD, and a retrospective analysis of clinical data indicates that a female harboring the mutation PAX8-P25R (pedigree 2) developed GDM. Thus, our results highlight an unprecedented role for PAX8 in islet survival in response to a specific metabolic stress (e.g., pregnancy).

Although we found distinct transcriptome profiles between murine and human islets overexpressing the respective species-specific PAX8, both converged to reduce islet apoptosis. In particular, during pregnancy, the placenta, among other tissues, is known to increase the secretion of proinflammatory cytokines to induce peripheral insulin resistance and favor the entrance of nutrients to the fetus during gestation (4). Therefore, it is tempting to speculate that PAX8 could be involved in the protection of pancreatic islets from the proinflammatory milieu that occurs during pregnancy as well as in adaptive mechanisms that cope with the increased demand for insulin biosynthesis and secretion, contributing to prevent stress-induced cell death under these metabolic circumstances. In support of this premise, we found that lentiviral-mediated PAX8 overexpression reduces the apoptosis rate in both murine and human pancreatic islets while maintaining normal metabolic activity and functionality. Of note, because Pax8 is expressed only transiently during pregnancy, these beneficial effects are likely acute, which is in line with the concept that postpartum, β-cells undergo increased apoptosis to restore functional mass to prepregnancy conditions (26). The antiapoptotic attributes of PAX8 have been described in other cell types (27) and the lack of Pax8 expression has been associated with increased apoptotic death in the cardiomyocytes of Pax8 homozygous knockout mice (28); therefore, our findings suggest that specific interventions targeting PAX8 in pancreatic endocrine cells could be explored to enhance islet cell survival.

The infertility of Pax8^−/− mice treated with thyroid hormone has hampered the validation of Pax8 requirement to maintain normoglycemia during pregnancy. However, our assumption that PAX8 might act as a prosurvival gene in islets specifically during gestation is supported by the identification and functional characterization of an unprecedented PAX8 mutation (p.T356M) found in a pregnant female who developed GDM. This mutation also was detected in the index patient’s father who was diagnosed with T2DM, suggesting a potential association of PAX8 with T2DM as previously proposed in a genome-wide linkage and admixture mapping study (10). Of note, two sisters of the index patient also were genotyped for the PAX8 mutation. Although one young sister carries the mutation T356M but has never been pregnant, the second sister lacks the mutation and was diagnosed with GTD but not GDM during pregnancy. These data suggest that on this specific genetic background, which is prone to develop thyroid complications as seen in the paternal and maternal sides of familial pedigree 1, the p.T356M variant may trigger GDM. It will be imperative to monitor the youngest sister for signs of GDM in her first pregnancy. In addition to the p.T356M mutation, the only pregnant female of a familial pedigree harboring the mutation p.P25R developed GDM and elevated HbA_1c levels during pregnancy. Overall, although our results cannot be considered fully conclusive with regard to the pathogeny of PAX8 mutations in the development of GDM and the study of larger cohorts of patients harboring mutations on PAX8 is warranted, our data suggest that PAX8 could be considered as a candidate gene for the study of glucose metabolism disorders during pregnancy.

To date, most PAX8 mutations associated with thyroid dysgenesis are located in the highly conserved DNA-binding paired domain (29,30) (Fig. 3B). In contrast, p.T356M is located within the carboxy-terminal of the protein that was shown to be essential for synergistic transactivation of the TG gene promoter along with the homeodomain-containing transcription factor TTF1 (25). Accordingly, although the capacity of DNA binding might not be compromised in the PAX8-T356M variant, its synergistic transcriptional activity along with TTF1 was reduced compared with PAX8Wt, suggesting that the mutation alters protein-protein interactions. As such, the PAX8-T356M variant may exert a dominant-negative effect, which may render a more severe phenotype compared with other mutations that could either impede DNA binding or generate a truncated nonfunctional variant of the protein, as observed in Pax8 heterozygous mice.

In conclusion, our data indicate that PAX8 expression in pancreatic islets during pregnancy might favor islet survival under this challenging metabolic condition. Furthermore, PAX8 could be considered as a candidate gene for the study of GDM.

Acknowledgments. The authors acknowledge the full collaboration of patients and their family members in the study. The authors thank Isabel Moreno-Navarro and Mercedes Tanarro (Thyroid Molecular Laboratory, Institute for Medical and Molecular Genetics, La Paz University Hospital, Autonomous University of Madrid, Madrid, Spain) for excellent technical work. The authors also acknowledge the computer resources provided by the Andalusia Bioinformatics Platform.

Funding. This work was funded by grants from the Ministerio de Economía y Competitividad, Instituto de Salud Carlos III, co-funded by fondos FEDER (CP14/00105 and PI15/00134 to A.M.-M., PI16/00830 to J.C.M., and PI13/00593 and BFU2017-83588-P to B.R.G.). Human islets were procured through the European Consortium for Islet Transplantation funded by JDRF (3-RSC-2016-162-I-X).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.M.-M., L.L.-N., C.J.-M., A.H., P.I.L., N.C.-V., A.T., C.G.-G., J.S.W.R.H., F.L., D.B., M.G.T., L.H., J.A., J.C.M., and B.R.G. discussed the results and commented on the manuscript. A.M.-M., L.L.-N., C.J.-M., P.I.L., N.C.-V., and M.G.T. performed the experiments using pancreatic islets, MCF-7 cells, and site-directed mutagenesis. A.M.-M., J.C.M., and B.R.G. designed and supervised the study, secured funding, analyzed the data, and wrote the manuscript. A.H., A.T., C.G.-G., J.S.W.R.H., L.H., and J.C.M. designed and performed human studies and transactivation assays. F.L. and D.B. provided human pancreatic islets. J.A. performed the human studies related to the P25R variant. B.R.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Data Availability. The Microarray raw data on pancreatic islets overexpressing the murine or human PAX8 have been deposited in the Gene Expression Omnibus repository with the accession number GSE94846.