Identification of a periodontal pathogen and bihormonal cells in pancreatic islets of humans and a mouse model of periodontitis

Results from epidemiological and prospective studies indicate a close association between periodontitis and diabetes. However the mechanisms by which periodontal pathogens influence the development of prediabetes/diabetes are not clear. We previously reported that oral administration of a periodontal pathogen, Porphyromonas gingivalis (Pg) to WT mice results in insulin resistance, hyperinsulinemia, and glucose intolerance and that Pg translocates to the pancreas. In the current study, we determined the specific localization of Pg in relation to mouse and human pancreatic α- and β-cells using 3-D confocal and immunofluorescence microscopy and orthogonal analyses. Pg/gingipain is intra- or peri-nuclearly localized primarily in β-cells in experimental mice and also in human post-mortem pancreatic samples. We also identified bihormonal cells in experimental mice as well as human pancreatic samples. A low percentage of bihormonal cells has intracellular Pg in both humans and experimental mice. Our data show that the number of Pg translocated to the pancreas correlates with the number of bihormonal cells in both mice and humans. Our findings suggest that Pg/gingipain translocates to pancreas, particularly β-cells in both humans and mice, and this is strongly associated with emergence of bihormonal cells.

cyctic fluid from Intraductal papillary mucinous neoplasm 8,9 . Although the presence of Pg in the pancreas has not been investigated, increased antibody to Pg has been detected in the plasma of subjects with pancreatic cancer 10 .
We have recently determined that mice orally administered Pg develop insulin resistance and hyperinsulinemia while maintaining normal glucose levels indicating a prediabetic condition 11 and that Pg translocates to the pancreas 12 . These results suggest that Pg may influence β-cell function. To gain understanding of how Pg interacts with islet cells, we set out to determine the specific localization of Pg in αand β-cells in mouse pancreatic islets and human pancreatic islet cells. In this process we quantitated the relative number of αand β-cells containing Pg and the emergence of bihormonal cells which express both insulin and glucagon in response to translocated Pg.
The emergence of bihormonal cells in animal models has been reported following near complete destruction of β-cells (99% ablation) by chemical agent 13 or by forced expression/deletion of βor α-cell specific transcription factors [14][15][16][17] using conditional knockout and/or lineage tracing mice. Re-differentiation of α-cells from de-differentiated β-cells 18 also represents another means of developing intermediate/bihormonal cells. Beta-to α-cell conversion has also been reported as a result of DNMT1 deletion 19 . Taken together, these studies show plasticity of pancreatic islet cells under defined conditions. Most recently, emergence of bihormonal cells was observed in a mouse model of experimental autoimmune diabetes 20 . In contrast to animal studies, quantitative data on human pancreatic bihormonal cells are scarce 21,22 . A recent study using human pancreatic samples obtained following pancreatoduodenectomy reported the higher percentage of bihormonal cells in an insulin resistant group compared with an insulin sensitive group, suggesting a possible adaptive response to insulin resistance 23 .
Here we show that orally applied Pg in mice translocates to and resides in intra-and peri-nuclear locations primarily in islet β-cells. The emergence of bihormonal cells was strongly associated with the presence of Pg/ gingipain in pancreatic islets of these animals as well as in human post-mortem pancreatic samples. These observations support the novel concept that oral bacteria causing periodontal infection can translocate to pancreatic islets where they may impact islet pathophysiology and the development of bihormonal cells.

Results
Pg/gingipain translocates to nuclear-and peri-nuclear regions of β-cells but not to α-cells in animals administered Pg. Following oral application of Pg 3 times per week for 22 weeks to simulate chronic periodontitis, the presence of Pg/gingipain was determined. Pg/gingipain was identified in pancreata of all mice that were administered Pg (N = 9) but none in control mice treated with vehicle alone (N = 10) by immunofluorescence (IF) microscopy and qPCR (N = 3/group) (Fig. 1A,B). 3-D confocal microscopy and orthogonal analyses revealed nuclear-or peri-nuclear localization of Pg/gingipain in β-cells (Fig. 1C,D, respectively).
Interestingly, Pg/gingipain was localized intracellularly in β-cells but not in α-cells that were identified based on insulin and glucagon staining, respectively (p < 0.001) ( Fig. 2A,B). Again, no Pg/gingipain was detected in pancreatic samples from control animals.
Bihormonal cells express the β-cell transcription factor Nkx6.1. Nkx6.1 is a transcription factor that is expressed in islet β-cells to repress expression of the α-cell transcription factor, Arx, and is not expressed in mature α-cells. To confirm that bihormonal cells exhibit a characteristic of β-cells, we stained for expression of the β-cell specific transcription factor Nkx6.1. IF microscopy revealed that bihormonal cells that were positive for both insulin and glucagon co-expressed Nkx6.1 (Fig. 4A,D). 3-D and orthogonal analysis of bihormonal cells showed the intra-nuclear localization of Nkx6.1 (Fig. 4B,C respectively).
Pg/gingipain is identified in human pancreata of both diabetic and non-diabetic subjects. We obtained a total of 45 postmortem human pancreata samples of which 32 were FFPE (19 Non-DM, 13 DM) and 13 frozen (11 Non-DM, 2 DM). Thus, of these 45 samples, 30 were from non-diabetic (Non-DM) and 15 from diabetic (DM) subjects (Table 1). In Non-DM subjects, 23% had Pg in the pancreas (Pg positive) and 77% were Pg negative (Table 1). In DM subjects, 40% were Pg positive and 60% Pg negative ( Table 1). The Chi-square analysis reveals a Chi-square value of 5.662 with 1 degree of freedom. The two-tailed P value equals 0.01. Thus, the proportion of subjects with Pg was significantly higher in DM compared to Non-DM groups.
Overall Pg/gingipain was identified by immunofluorescence microscopy or by qPCR in 13 of 45 (29%) of human pancreata.
Pg/gingipains are more strongly associated with β-cells than α-cells in human pancreata. 2-D IF microscopy and 3-D confocal microscopy analysis of human pancreatic tissue sections show the presence of peri-and intra-nuclear Pg/gingipain in β-cells (Fig. 5A,C). As was observed in mouse islets, the presence of intraand peri-nuclear localization was confirmed by orthogonal analysis (Fig. 5D). Of 13 frozen samples for which we performed qPCR analysis for Pg positivity, 4 were positive for Pg and 9 were negative. The copy number of the Pg 16S rRNA gene varied from 2.6 × 10 4 to 8.8 × 10 6 (mean 2.8 × 10 6 ) (Fig. 5B).
We had determined that Pg/gingipain signal was exclusively localized to β-cells in samples from mice ( Fig. 2) and addressed this same question in samples from humans. In contrast to what we found in mice, Pg/gingipain was mainly localized to β-cells but was also present in α-cells (Fig. 6A,B) in human islets. The mean percentage of Pg positive βand α-cells was 30% and 7.5% respectively and this difference was statistically significant (p < 0.001) (Fig. 6B).    (Fig. 7B). The percentage of bihormonal cells was significantly higher in DM (17 ± 7.6%) compared with Non-DM subjects (9 ± 7.2%) (p < 0.001) (Fig. 7C). 2-D IF images (Fig. 7A,B) as well as 3-D and orthogonal analysis (Fig. 7D,E, respectively) show the presence of bihormonal cells. IF microscopy data indicate that Pg/gingipain is present in β-cells and in a low percentage of α-cells as well as bihormonal cells (Fig. 7F,G).

A strong correlation exists between Pg invasion and the emergence of bihormonal cells in both human and mice.
We next determined the relationship between the number of bihormonal cells and the presence or absence of Pg in the pancreas in both humans and mice (Fig. 9A,B). In both humans and mice, the number of bihormonal cells was significantly higher in Pg+ compared with Pg-samples (p < 0.0001).
A correlation analysis was performed to determine if there is a relationship between the number of Pg and the number of bihormonal cells. This analysis showed a strong correlation between these two variables in both mice and human samples (correlation coefficient r = 0.821, p < 0.0001 for humans; r = 0.853, p < 0.0001 for mice) (

Discussion
Nearly 50% of human adults in the U.S. have periodontitis 24 with 20% having moderate, and 10% severe periodontitis. Periodontitis is a disease characterized by destruction of gingiva and tooth-supporting bone by a host immunological response to periodontal pathogens. It is reported that periodontal bacteria (gram-negative bacteria) are translocated to the liver 1 , aorta 2 , and arteries 3,4 via the systemic circulation [25][26][27] or possibly via the gut axis 1 , but the effects of periodontal pathogens in these tissues are poorly understood. We have recently reported the presence of Pg/gingipain in the brain (hippocampus) of young wild type mice that were administered Pg orally 3 times per week for 22 weeks and these mice developed neuropathology consistent with that of Alzheimer's disease 5 . Our current study is the first to show the intra-cellular, peri-nuclear and intra-nuclear localization of a periodontal pathogen/product in pancreatic β-cells in mice as a result of oral application of Pg as well as in human pancreatic β-cells in both non-diabetic and diabetic subjects. Interestingly, the presence of Pg/gingipain in 29% of the randomly selected human pancreata used in this study corresponds to the percentage of humans with moderate to severe periodontitis. Unfortunately, in our current study, data regarding the periodontal condition were not available from these subjects. Nonetheless, given that the primary niche of Pg is the oral cavity, Pg/gingipain identified in the human pancreas most likely originated from this site.
In humans, the presence of Pg in the oral cavity or increased level of serum antibody to Pg is associated with pancreatic cancer 10 but the actual presence of Pg in pancreatic cancer tissues has not been reported. However, a periodontal pathogen Fusobacterium sp. has been detected in pancreatic ductal adenocarcinomas and in pancreatic cystic neoplasm cyst fluid 8,9 . The result of our analysis of human post-mortem pancreatic samples is the first to show the presence of a periodontal pathogen in non-cancerous subjects, and specifically in human islets.
We used both a mouse monoclonal antibody (61BG1.3) which is specific to an epitope in the beta-adhesin domain of gingipains (G907-T931) and a rabbit polyclonal antibody raised against the active site of His sequence of gingipain 28 to confirm the presence of Pg/gingpains in islets. Although positive signals using these antibodies reflect the presence of gingipains, this does not preclude the presence of whole bacteria residing peri-nuclearly or intra-nuclearly in β-cells. Gingipains are part of the outer cell membrane or reside inside membrane vesicles and thus it is possible that our staining detects whole Pg, a fragment of Pg cell membrane and/or secreted gingipains. Pg is a non-motile, coccobacillus gram-negative obligate anaerobe with an approximate size of 1.2 μm × 0.5 μm with an outer membrane vesicle of approximately 40 nm 29 . Since the size of a human β-cell is approximately 10-12 μm with a nuclear diameter of 6.0-8.0 μm 30-32 , our antibodies may detected whole Pg. In support of this, we detected Pg specific 16S rRNA genes in both mouse and human islets. However, we are not certain if this reflects the presence of either actively dividing or dormant bacteria.
Identification of a periodontal pathogen or its product in the nucleus of pancreatic islet cells is surprising. However, Pg is known to invade host cells such as epithelial 7 and endothelial cells in vitro 33 . Further, a heterodimer derived from gingipain RgpA designated as HRgpA is known to enter the nucleus of epithelial cells and localize in and around the nucleus of these cells in vitro 7 . We recently reported the peri-and intra-nuclear localization an experimental mouse. Yellow: glucagon, Green: insulin, Red: Pg, Blue: nucleus. Scale bar represents 10 μm. (F) Percentage of Pg positive α-, βand bihormonal cells. Y-axis: % cells positive for Pg, X-axis: cell types. 449 α-cells, 968 β-cells and 72 bihormonal cells were counted. N = 5 experimental mice. Data are presented as mean + S.D. ***p < 0.001. (2020) 10:9976 | https://doi.org/10.1038/s41598-020-65828-x www.nature.com/scientificreports www.nature.com/scientificreports/ of Pg/gingipain in neurons, microglial cells and astrocytes in the brain (hippocampus) of mice orally administered Pg 5 .
A number of bacteria and their products are known to enter and reside in host cell cytoplasm or the nucleus 34,35 . These bacteria include E. coli 36,37 , Legionella penumophila 38 and its product OrfX which specifically targets the nucleus 39 , and Rickttsia richettsii which can invade the nucleus and form endonuclear colonies in human cells 40,41 . The mechanism that facilitates entry of bacteria or its product, referred to as "nuclear modulin (nucleomodulin), " into the nucleus is not clear as nucleomodulins usually lack a classical nuclear localization   42 . Pg/gingipains also do not have canonical NLS 7 . Study of the mechanism by which Pg and/or gingipain gains access to the nucleus primarily in β-cells may give us insight into the development of prediabetes and diabetes associated with oral bacteria. It is not clear why Pg/gingipain specifically invades β-cells in mice and significantly more in βthan α-cells in humans.
Bihormonal cells are considered an intermediate stage in αto β-cell transdifferentiation 16 and may be important as a β-cell compensatory response to IR 22,43 . Thus, the emergence of bihormonal cells is potentially a physiologically important mechanism to cope with IR and/or loss of functional β-cells. IF and confocal microscopy with orthogonal analysis showed nuclear localization of Nkx6.1 in a low percentage of bihormonal cells in both human and experimental mice. Nkx6.1 is a transcription factor specific to β-cells within the islets, which binds and inhibits transcription of the α-cell specific Arx gene thereby repressing α-cell identity 44 . However, secretion of insulin from bihormonal cells isolated from mice subjected to oral application of Pg has yet to be confirmed by functional analysis.
The literature reporting quantitative data on human bihormonal cells is limited 21,22 , and previous studies have not considered the possibility that bacterial pathogens may contribute to their emergence. Two possible means to induce conversion of α-cell to β-cell or β-cell to α-cell have been identified: (1) near complete ablation of β-cells using chemical agents 13,45,46 and (2) forced expression/deletion of βor α-cell specific transcription factors and/ or deletion of Dnmt1 or FOXO1 transcription factor in β-cells in mice [14][15][16]18,19 . We found that Pg/gingipain is not detected in mouse α-cells and at a low % in human α-cells, making it unlikely that Pg directly induces αto www.nature.com/scientificreports www.nature.com/scientificreports/ β-cell transdifferentiation. However, whether the emergence of bihormonal cells is based on αto β-cell transdifferentiation or β-cell de-differentiation/conversion in the context of an oral bacteria reaching the pancreas will require additional studies as some of our bihormonal cells in mice and human did contain Pg/gingipain. Thus, the study using lineage tracing methods would be necessary to identify the direction or bi-direction of islet cell transdifferentiation and/or dedifferentiation. In our mouse model of experimental periodontitis the percentage of β-cell destruction/apoptosis was a mean of 8% 12 and yet bihormonal cells (mean 17.7% of total count of gluca-gon+ cells) emerged, suggesting that near complete destruction of β-cells may not be necessary for the emergence of bihormonal cells in our model system. Interestingly, Pg-LPS has been shown to impact gene expression by significantly reducing DNA methyltransferase (DNMT1) expression 47 . Inactivation of both Dnmt1 and Arx in adult mice has been shown to drive transdifferentiation/conversion of αto β-cells 16 . It has also been shown that β-cells deficient in DNMT1 are converted to α-cells 19 . Thus, Pg with cell membrane associated LPS present in the nucleus of β-cells may be altering the methylation signature and thus βand α-cell identity. Further, intra-cellular localization of Pg/gingipain suggests other epigenetic functions of Pg/product. For example, it is reported that Pg decreases trimethylation of histone H3K4 48 . Furthermore, treatment of cultured pancreatic islets with a histone methyltransferase inhibitor results in colocalization of both glucagon and insulin in mouse islets, i.e. bihormonal cells 49 . Based on these data, it is possible that Pg translocation from the oral cavity to the nuclei of islet cells may lead to the decreased trimethylation of H3K4 and thereby contribute to the emergence or maintenance of bihormonal cells. In summary, we present data obtained by IF, confocal microscopy and qPCR that Pg is present in human pancreata and is at a higher incidence in the pancreata of DM than Non-DM humans. This is the first study to investigate the presence and intra-cellular localization of a periodontal pathogen and/ or its product in human pancreata and also in the pancreas of an animal model of experimental periodontitis. Our data indicate that Pg/gingipain is translocated from the oral cavity to pancreatic islets and localized primarily in β-cells in both humans and mice. In this process, emergence of bihormonal cells occurs. The results from a recent study show that Pg expresses dipeptidyl peptidase 4 (DPP4) which degrades incretins and thus reduces insulin levels which is accompanied by a substantially elevated glucose levels at a 15 min time point following oral glucose administration 50 . PgDPP4 does not appear to account for our results since our experimental mice show high insulin levels and normal glycemic levels, thus suggesting that non-DPP4 factors/effects may be involved. However, we can't exclude the possibility that PgDPP4 may be present but not contributing to net effects on insulin levels. Finally, based on the published literature on nuclear invasion of bacteria/products in various cell types, it is interesting to speculate that Pg/gingipains may epigenetically influence development of bihormonal cells.

Methods
Animals. This study was carried out in strict accordance with the recommendations outlined in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee at the University of Illinois at Chicago (Protocol approval #15-142). Twenty 8-week old male C57BL/6 mice were used for this study. Oral application of Pg was performed 3 times per week for 22 weeks. At baseline and at week 22, a glucose tolerance test was performed by intraperitoneal injection of 50% dextrose solution (2 g/kg body weight) and glucose levels in tail blood were determined at 15, 30, 60, 90 and 120 min post-dextrose injection using a glucometer. Fasting insulin levels were determined by ELISA (10-1247-01, Mercodia AB, Uppsala, Sweden) and insulin resistance was calculated using the homeostasis model assessment (HOMA-IR) 51 .  www.nature.com/scientificreports www.nature.com/scientificreports/ Appropriate informed consent was obtained from next of kin for autopsies and all methods were carried out in accordance with University of Illinois Hospital Institutional guidelines and regulations involving dead bodies.
Bacterial culture and oral application of Pg. Briefly, Pg (strain W83) was grown anaerobically as described in our previous study 11 . Cell density was determined by spectrophotometry at an optical density of 550 nm based on a standard curve established using colony forming units (CFU). 1 × 10 9 Pg were transferred to microfuge tubes, vortexed briefly and centrifuged at 10,000 g for 2 min at 4 °C. Supernatants were discarded and the pellets were re-suspended in 4 o C PBS, and then re-pelleted by centrifugation. Supernatants were removed, and bacteria re-suspended in 100 μl of 2% carboxymethyl cellulose (CMC) in PBS and tubes immediately placed on ice until administered to mice. 100 μl of Pg in CMC containing 10 9 Pg (experimental group) or CMC alone (vehicle alone control group) was placed (2 applications of 50 μl) in the oral cavity every other day (Monday, Wednesday, and Friday) for 22 weeks. Animals were sacrificed using isoflurane anesthesia followed by cervical dislocation following 22 weeks of Pg application and pancreata were removed and fixed in paraformaldehyde to be embedded in paraffin.
Mouse pancreata. Mouse pancreata samples were obtained from C57BL/6 WT mice that received Pg orally (experimental animals, N = 10) or vehicle alone (control animals N = 10) 11 . Pancreata from these animals were harvested, fixed in formalin and embedded in paraffin (FFPE).
Immunofluorescence, confocal microscopy, and orthogonal analysis of human and mouse pancreas. Immunofluorescence microscopy (IF) was performed to detect Pg/gingipain and αand β-cells.
Briefly, 5μm thick sections were first de-paraffinized with xylene and rehydrated through a series of decreasing percentages of ethanol. Antigen retrieval was performed by microwaving sections in 1 mM ethylenediaminetetraacetic acid (EDTA), pH 8.0 or in 10 mM citrate buffer pH 6.0 for 5 min repeated 4 times. Cell and nuclear membrane permeabilization was performed by incubating sections in 0.25% Tween 20 in PBS for 30 min. To detect Pg/gingipain, tissue sections were incubated overnight at 4 o C either with mouse anti-Pg monoclonal antibody (DSHB, Iowa City, Iowa) or rabbit anti-Pg polyclonal antibody as primary antibody 28 at a dilution of 1:100, followed by donkey anti-mouse conjugated to Alexa Fluor 647 (A31571, Invitrogen, Grand Island, NY) or donkey anti-rabbit IgG conjugated to Alexa Fluor 594 (A21206, Invitrogen, Grand Island, NY) respectively as secondary antibody at room temperature for 1 hr. To detect β-cells, tissue sections were incubated overnight at 4 °C with goat anti-insulin antibody (sc-7838, Santa Cruz Biotechnology, Inc., Dallas, TX) at a concentration of 1:100, followed by donkey anti-goat IgG conjugated to Alexa Fluor 488 (A11055, Invitrogen, Grand Island, NY), as secondary antibody at room temperature for 1 hr. To detect α-cells, tissue sections were incubated overnight at 4 o C with anti-glucagon antibody at a concentration of 1:100 (ABCAM, Cambridge, MA), followed by donkey anti-mouse IgG conjugated to Alexa Fluor647 (A31571, Invitrogen, Grand Island, NY) as secondary antibody. To detect the β-cell specific transcription factor Nkx6.1, goat anti-Nkx6.1 antibody (AF5857, R&D Systems, Inc, Minneapolis, MN USA) was used as primary antibody at a concentration of 1:100. Sections were incubated for 1.5 hr in a humidified chamber at room temperature, followed by anti-goat IgG conjugated with Alexa Fluor 488 at a 1:800 dilution as secondary antibody. Isotypic controls were used to determine non-specific binding of antibodies for all experiments. Nuclei were identified using DAPI as described 12 . For 3-D image reconstruction, an isovoxel data-set was acquired. Orthogonal analysis were performed based on compiled z-sections (0.08μm step size) obtained by confocal microscopy using the Imaris software (Bitplane, Inc) version 7.7.2.
For samples from mice, the percentage of cells that exhibited perinuclear/intranuclear Pg/gingipain in αand β-cells was calculated under 40X magnification by counting the number of insulin+/Pg+ cells divided by the total number of insulin+ cells per islet or glucagon+/Pg+ cells divided by the total number of glucagon+ cells. A total of 5 islets per mouse were examined in 5 mice/group.
In human post-mortem pancreatic samples, the percentage of αand β-cells exhibiting Pg+ was calculated as for mouse samples except that cells were counted in four rectangular areas of 400 × 300 μm per sample.
The percentage of bihormonal cells in mice were counted under 40X magnification by identifying glucagon+/ insulin+ cells relative to the total number of glucagon+ cells per islet. 4-5 islets/animal were examined in 7 animals/group.
For human samples, the percentage of bihormonal cells positive for Pg/gingipain was determined by counting the number of cells identified by immunofluoresnce microscopy that were positive for both insulin and glucagon staining relative to the number of cells that stained positive for glucagon in a 400 × 300 μm rectangle under 40 X magnification.  52 . The sequences of the Pg16S rRNA gene forward primers were GCGCTCAACGTTCAGCC, reverse primers were CACGAATTCCGCCTGC, and probe was CACTGAACTCAAGCCCGGCAGTTTCAA. PCR primers and TaqMan probe for detection of Pg were custom made by Integrated DNA Technologies (IDT, Coralville, IA) with a 6-FAM fluorescent label and both Zen internal and 3' Iowa Black fluorescence quenchers. A synthetic double-stranded DNA standard for 16S rRNA gene was synthesized as a gBLOCK fragment (IDT, San Jose, CA) 5 . The standard contained 243 bp of the16S rRNA gene from the Pg strain W83, and greater than 60 bp of adjacent DNA on either side of the target sites. The double-stranded gBLOCK oligonucleotide was diluted across 8 orders of magnitude, and used as a standard for quantitation of Pg 16S rRNA genes in genomic DNA extracts. The concentration of the standards was assessed using Qubit fluorimetry, and the copies per microliter assessed by estimation of the DNA weight/molecule based on sequence and sequence length and the DNA concentration. Together with a dilution series, absolute quantification (copies per microliter of nucleic acid extract) of samples was assessed through linear regression analysis from the standard curve. All assays were performed in technical triplicates. PCR amplification was performed in a total reaction mixture volume of 25 μl. The reaction mixtures contained 12.5 μl of 2x TaqMan universal PCR master mix (4304437, Thermo Fisher Scientific), 300 nM each Pg-specific primer, 100 nM Pg-specific probe, and purified DNA. The samples were subjected to an initial amplification cycle of 50 °C for 2 min and 95 °C for 10 min, followed by 40 cycles at 95 °C for 15 s and 60 °C for 1 min 5 .
Statistical analysis. Nonparametric data, Pg 16S copy number in mice and human samples as well as ipGTT, glucose, insulin and HOMA-IR, were evaluated by the Mann-Whitney U-test for two-group comparisons. The proportion of subjects with Pg+ in pancreas was analyzed between DM and Non-DM groups using a Chi-square test. Shapiro-Wilk test was used to determine the distribution of the data. Based on the results of Shapiro-Wilk test, the Pearson's correlation analysis was used to analyze the correlation between the number of bihormonal cells and the number of Pg in humans, and Spearman's correlation analysis was used for mice samples. All other statistical analyses were done using a Student's t-test with a cutoff value of p < 0.05 considered significant.

Data availability
The data sets analyzed for this study are available from the corresponding author upon reasonable request.