In the setting of β-cell stress, the pancreatic duct gland transcriptome shows characteristics of an activated regenerative response

Abstract

The pancreatic duct gland (PDG) compartment has been proposed as a potential stem cell niche based on its coiled tubular structure embedded in mesenchyme, its proliferation and expansion in response to pancreatic injury, and the fact that it contains endocrine and exocrine epithelial cells. Little is known of the molecular signature of the PDG compartment in either a quiescent state or the potentially activated state during β-cell stress characteristic of diabetes. To address this, we performed RNA sequencing on RNA obtained from PDGs of wild-type vs. prediabetic HIP rats, a model of type 2 diabetes. The transcriptome of the PDG compartment, compared with a library of 84 tissue types, placed PDGs midpoint between the exocrine and endocrine pancreas and closely related to seminiferous tubules, consistent with a role as a stem cell niche for the exocrine and endocrine pancreas. Standard differential expression analysis (permissive threshold P < 0.005) identified 245 genes differentially expressed in PDGs from HIP rats vs. WT rats, with overrepresentation of transcripts involved in acute inflammatory responses, regulation of cell proliferation, and tissue development, while pathway analysis pointed to enrichment of cell movement-related pathways. In conclusion, the transcriptome of the PDG compartment is consistent with a pancreatic stem cell niche that is activated by ongoing β-cell stress signals. The documented PDG transcriptome provides potential candidates to be exploited for lineage tracing studies of this as yet little investigated compartment.

NEW & NOTEWORTHY The pancreatic duct gland (PDG) compartment has been proposed as a potential stem cell niche. Transcriptome analysis of the PDG gland placed it midpoint between exocrine and endocrine tissues with adaptation toward response to inflammation and increased cell movement in a model of type 2 diabetes with ongoing β-cell apoptosis. These findings support the proposal that PDGs may act as a pancreatic stem cell niche.

INTRODUCTION

It has been known for many years that there are glandular-like structures arising from larger pancreatic ducts that undergo proliferation following pancreatic injury with an increase in cells expressing the transcription factor Pdx-1, and containing occasional endocrine cells (22, 23). More recently, these glandular structures were proposed as being a distinct anatomical compartment with molecular features consistent with a stem cell niche, and named pancreatic duct glands (PDGs) (20). PDGs are crypt-like invaginations off the pancreatic ductal tree, embedded in mesenchyme and as such are anatomically reminiscent of the gastric, ileal, and colonic crypt stem cell niches (Fig. 1) (3). Similar glandular structures (peribiliary glands) are present as crypt-like outpouches off the biliary tree and have been reported to bear multiprogenitor cells with a potential ductal epithelium or endocrine fate (8). PDGs, like ileal crypts, have zones of increased replication, consistent with a transit amplifying zone, and a predominance of exocrine epithelial cells with a minority of endocrine cells. The PDG compartment is expanded with increased proliferation in humans with both type 1 diabetes (17) and type 2 diabetes (19).

Fig. 1.

Comparison of pancreatic duct gland (PDG) and ileal crypt. A and B: human ileal crypts and PDGs share many properties. Both are crypt-like structures embedded in mesenchyme. Ileal crypts and PDGs are composed primarily of gut or pancreas epithelium, respectively, but with occasional endocrine cells (A, inset stained brown, chromogranin) and (B, stained pink, insulin). C and D: both ileal crypts and PDGs have zones of increased replication (Ki-67) compared with surrounding structures, consistent with a role as a transit amplifying zone. The variability in frequency of proliferating cells in PDG compartments was detected.

Relatively little is known about the molecular signature of PDGs. Thus far, molecular characterization of PDGs has relied on candidate immunohistochemistry or RT-PCR in rats or mice (20, 22, 23) and immunohistochemistry in humans (17). To date this approach has revealed Pdx-1, nestin, mucin-6, Hes1, and Ngn3 expression in rodent PDGs and Sox9, GATA4, NKx6.1, NKx2.2, and chromogranin A in humans PDGs.

Powerful new tools are available to characterize the molecular signatures of tissues of interest based on RNA and/or protein profiling. Laser capture microdissection (LCM) is a technique that enables isolation of RNA and/or protein from a compartment of interest within an organ. After identification of the compartment by microscopy, laser dissection is used to procure a sample of the cells of interest to permit subsequent gene expression and/or proteome profiling (11, 21). In the present study, we employed that approach to procure high-quality RNA by a protocol validated for PDGs (6). We then characterized the transcriptional profile of these samples using RNA sequencing (RNA-seq) to establish the molecular identity of the PDG compartment in an unbiased manner. We used these data 1) to compare the molecular signature of PDGs with that of multiple tissues; 2) to identify possible markers overrepresented in the PDG compartment; and 3) to compare PDG samples from a rat model of type 2 diabetes (HIP rat) compared with wild-type (WT) controls, with the goal of identifying the signaling networks and pathways altered in PDGs in the context of diabetes.

MATERIALS AND METHODS

Rats.

The generation of the human IAPP transgenic rats (HIP) has been described in detail previously (5). Wild-type rats were littermates of HIP rats. Rats were bred at the University of California, Los Angeles (UCLA) animal housing facility, subjected to a standard 12:12-h light-dark cycle, and fed Rodent Diet 8604 (Harlan Teklad, Madison, WI) ad libitum. All experimental procedures were approved by the UCLA Institutional Animal Care and Use Committee. HIP rats develop overt diabetes between 9 and 12 mo of age, with islet pathology similar to humans with type 2 diabetes, specifically ongoing β-cell apoptosis with a progressive defect in β-cell mass (5). HIP rats used for this study were ~6 mo of age, and so prediabetic with ongoing β-cell apoptosis but without the confounding secondary actions of hyperglycemia on gene expression (15). WT rats were age matched.

Tissue procurement and LCM.

On the day of study, animals were anesthetized by inhalation of isoflurane (Abbott Laboratories, Chicago, IL). Rat pancreas was rapidly dissected, divided into two portions (head and body of pancreas, and tail of pancreas), and cryopreserved in optimal cutting temperature compound (6). Sections (10–15) were cut from the head of the pancreas for each LCM experiment. In brief, 8-μm sections were mounted on UV-irradiated polyethylene naphthalate membrane slides and stored at −80°C. Right before use, slides were fixed in alcohol and stained with hematoxylin (solutions contained RNA inhibitors).

The PDG compartment was identified morphologically. As described previously (20, 24), the PDG compartment is readily identifiable in mouse and human pancreas because of its unique architecture distinct from the ductal epithelium. We confirmed previously that, similar to humans and mice, PDGs in rat pancreatic tissue sections appear as coiled structures embedded in the mesenchyme surrounding the main duct, are readily stained with Alcian blue and p-aminosalicylic acid, and also characterized by the increased proliferation rate relative to normal ductal cells (12). For the current study, we set out to collect only PDGs in the substantial layer of mesenchyme where, based on extensive experience (6, 12, 17, 19), they are readily and rapidly identified, ensuring specificity and quality of collected RNA.

The PDG compartment was cut into a 0.5-ml tube cap (Axygen Scientific, Union City, CA) filled with 10 µl of extraction buffer (PicoPure RNA Isolation Kit, KIT0204) and 0.5 µl RNase inhibitor (1 U/µl) (SUPERase• IN, AM2694; Ambion, Carlsbad, CA). To avoid RNA degradation, slides were processed one at a time, and staining and dissection from each slide were finished within 20 min. The LCM procedure was performed using a LMD7000 Laser Microdissection system (Leica, Wetzlar, Germany) at the California Nano Systems Institute Advanced Light Microscopy/Spectroscopy Shared Resource Facility at UCLA.

Quality control and LCM selectivity.

RNA quality was tested with a 2100 Bioanalyzer using a RNA 6000 Pico LabChip Kit (Agilent Technologies, Santa Clara, CA). To validate the LCM selectivity, RNA was isolated from LCM-derived samples from PDGs and islets and tested by RT-PCR for the abundance of transcripts known to be expressed in PDGs or islets. mRNA expression of PDG/ductal cell protein cytokeratin-19 (CK-19) and islet hormones insulin and glucagon were analyzed (6). By this approach, CK-19 was highly expressed in PDGs but not islets. Insulin and glucagon transcripts were abundantly expressed in islets and at low levels (~1,000-fold lower than islets) in PDGs, consistent with occasional endocrine cells in this compartment. In this study, we ranked expression levels of all 15,679 rat probes included in the RNA-seq annotation and calculated the corresponding expression percentile. CK-19 (Krt19) ranked 6/15,679 (<0.1th percentile), insulin (Ins1) ranked 19 (0.1% percentile), Ins2 ranked 126 (0.7th percentile), and glucagon ranked 13,682 (87th percentile). Whereas CK-19 and glucagon levels were consistently at the top and bottom of the list, respectively, insulin values were more variable across replicates, suggesting variability across our samples. Because we did not run RNA-seq on islet tissue for this experiment, we cannot assess expression relative to islet tissue.

RNA-seq.

RNA-seq was performed in the UCLA Neuroscience Genomics Core (UNGC, https://www.semel.ucla.edu/ungc). Between 5 and 20 ng of total RNA were extracted per tissue. After quantification and quality check, 5 ng of total RNA were amplified at UNGC using the NuGEN Ultralow Library System kit (NuGEN), which is optimized for downstream Illumina library preparation. We extracted RNA from six samples as follows: three from HIP and three from WT animals. Illumina RNA-seq libraries were then prepared according to the manufacturer's instructions. Sequencing was performed using the Illumina HiSeq 2500 sequencer and the version 3 Illumina chemistry. We barcoded multiple samples and ran them over multiple lanes, to minimize batch effects (2). We ran the equivalent of three samples per lane, with paired-end 100-bp read length, corresponding to two HiSeq 2500 lanes.

Between 83 and 187 million, 100-bp-long, paired-end reads were obtained and aligned to the rat genome (rn5) using the STAR spliced read aligner (10). Reads (65–72%) mapped uniquely to the rat genome, and ~50% of the genes in the rat genome were detected as present by at least 100 mapping reads. Samples were clustered using hierarchical clustering and multidimensional scaling (MDS), and no outliers were detected.

Data analysis.

An RNA-Seq pipeline is established in the UCLA Informatics Center for Neurogenetics and Neurogenomics (https://github.com/icnn/RNAseq-PIPELINE). Initial analysis steps included: 1) quality analysis, alignment to reference genome using STAR (10), and filtering of reads not uniquely mapping or mapping to repetitive regions; 2) mapping of reads to exons, untranslated regions (UTRs), and intron-exon junctions using STAR, and generation of RefSeq isoform counts; 3) normalization and differential expression analysis by tissue and condition using the software DEseq (1) and edgeR (18); and 4) data upload on our web-based gene expression database.

MDS was used to cluster PDG samples with samples obtained from an atlas of gene expression in human tissues. Briefly, the human GNF database was downloaded from the bioGPS website (http://biogps.org), probes targeting transcripts shared across the two platforms (Affymetrix, used in the GNF database and RNA-seq, used in this study) were normalized jointly using quantile normalization, and MDS plots were generated to organize samples in a two-dimensional space, based on the expression of the top 1,000 most variable genes. Differential expression analysis was performed to compare HIP and WT samples, using the edgeR package and setting the P value threshold of 0.005. Gene ontology and pathway analysis were performed using DAVID (http://david.abcc.ncifcrf.gov/) and Ingenuity Pathway Analysis (ingenuity.org).

RESULTS

Identification of the PDG compartment.

PDGs share many properties with human ileal crypts (Fig. 1). Both are crypt-like structures embedded in mesenchyme. Ileal crypts and PDGs are composed primarily of gut or pancreas epithelium, respectively, but with occasional endocrine cells (Fig. 1, A and B). Both ileal crypts and PDGs have zones of increased replication (Fig. 1, C and D). In human and rat pancreas, PDGs were readily identified in hematoxylin-stained tissue sections based on their unique anatomical location as invaginations off main pancreatic ducts embedded as coiled structures within the mesenchyme surrounding main pancreatic ducts (Figs. 1 and 2). In keeping with prior descriptions (12, 19, 20, 24), PDGs were notable for abundant mucins as detected by Alcian blue, increased frequency of proliferating cells (Fig. 1, A and B), and the presence of occasional insulin-positive cells (Figs. 1B and 2B).

Fig. 2.

Pancreatic duct gland histology. A: section through a large duct in pancreas from a wild-type (WT) rat demonstrating the pancreatic duct gland (PDG) compartment present in the mesenchyme surrounding the large duct, with PDGs connecting directly with the large duct lumen (arrows). B: section through a large duct in pancreas from a human IAPP transgenic (HIP) rat demonstrating the extensive PDG compartment relative to WT rats. Sections are stained for insulin (DAB) with hematoxylin counterstain. Images were taken at ×10 (×100 magnification). ★Large duct lumen. Inset in B shows a PDG epithelial cell staining for insulin. Scale bar = 200 μm.

RNA sequencing-based gene expression analyses in dissected PDG samples.

Having previously established a method to obtain RNA by LCM from PDGs of a suitable quality and confirmed to reflect the PDG compartment based on RT-PCR of sentinel genes (6), we applied this approach here to obtain RNA from PDGs to perform RNA-seq so as to obtain an unbiased genetic expression profile of the PDG compartment (GSE-117679). With these data, we then compared the PDG transcriptome with an atlas of transcriptional data including 84 tissues. MDS analysis (see materials and methods and the legend for Fig. 3) clustered related tissues (e.g., brain regions or blood cell types) together because of similarity of their transcriptome. Our PDG samples were placed in a region including other pancreatic samples (both exocrine and endocrine), further supporting their source tissue of origin, but also testis (including seminiferous tubules, which host actively replicating cells with pluripotent potential) and pituitary gland (Fig. 3).

Fig. 3.

Multidimensional scaling plot of tissue samples human expression atlas (http://biogps.org/). Each dot represents the relative location of gene profile from one of the 176 samples from 84 surveyed tissues. Dots are color coded by tissue group or system. Samples cluster based on similarity. Validity of the analytical approach is shown by the clustering of expression profiles obtained for brain or blood (shaded areas). In red, RNA sequencing (RNA-seq)-derived genetic profiles obtained from pancreatic duct glands (PDGs) are reassuringly close to each other. Also, PDG gene expression profiles are placed between a classical stem cell profile (germ cells) and stem cell niche profile (Leydig cells) in one dimension and then close to the two pancreas compartments (islet and exocrine pancreas). The [networks, functional analyses, etc.] were generated through the use of IPA (QIAGEN, https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis).

We then proceeded to identify the markers most specific to the PDG compartment. Briefly, after normalization, we compared all six PDG samples with the average of all other tissues in the GNF database. This analysis identified 22 transcripts as putatively overexpressed in PDG cells relative to the average expression levels of 84 tissues [P < 0.005, Supplemental Table 1 (Supplemental data for this article may be found on the journal website.)]. Finally, we compared PDG expression profiles between HIP and WT rats (Supplemental Table 2). Standard differential expression analysis at a permissive threshold (P < 0.005) identified 245 genes (Fig. 4, A and B) that were differentially expressed across the three replicates. Gene Ontology analysis (Fig. 4C) highlighted an overrepresentation of transcripts involved in acute inflammatory response, regulation of cell proliferation, and tissue development. Pathway analysis pointed to enrichment of cell movement-related pathways, and the top two networks (Fig. 5, A and B) included ADIPOQ, playing an important role in type 2 diabetes, and the known therapeutic target PPARG.

Fig. 4.

Differential expression analysis comparing pancreatic duct gland (PDG) samples from human IAPP transgenic (HIP) and wild-type (WT) rats. A: number of upregulated (red bar) and downregulated (green) transcripts when comparing HIP vs. WT. B: heatmap representing ratios of the 245 dysregulated transcripts. Individual genes are in rows, and samples are in columns. Each cell represents a ratio (each HIP sample vs. the average of WT). Shades of red, upregulation; shades of green, downregulation. C: overrepresentation of gene ontology categories within the differentially expressed gene set. Within each category, green represents the proportion of downregulated and red the portion of upregulated transcripts when comparing HIP vs. WT samples.

Fig. 5.

Top two networks (A and B) identified by Ingenuity Pathway Analysis as overrepresented within the list of 245 genes (P < 0.005) dysregulated in human IAPP transgenic (HIP) vs. wild-type (WT) rats. Symbols in shades of green or red denote up- and downregulated transcripts, respectively. Symbols in gray are not significantly differentially expressed in this data set, but known to be part of the pathway or network. Solid lines denote direct interaction and dotted lines indirect interaction, such as an alteration in expression levels, or posttranslational modification.

In conclusion, unbiased transcriptional analyses support the notion that the PDG compartment includes a unique transcriptional niche, similar to both the tissue of origin (pancreas) and to replicating cells, and that a signal related to diabetes is detectable in a rat model of type 2 diabetes.

DISCUSSION

Although the existence of blind pouches from main pancreatic ducts bearing stem cell markers has been appreciated for many years, only recently has this anatomical compartment been named the PDG compartment and been proposed as a stem cell niche responsible for repair of pancreas following acute injury (20).

Features of the PDG compartment that are consistent with a stem cell niche include a tubular crypt-like structure embedded in mesenchyme, expression of stem cell markers such as Hes-1, and proliferation with expansion in response to inflammation and/or injury (20). Other foregut-derived tissues such as the duodenum and proximal ileum have well-characterized stem cell niches that are also crypt like, located at the base of the intestinal villi embedded in mesenchyme. These well-defined stem cell niches generate cells that transition through proliferative transit amplifying zones to generate sufficient cells to replace the short-lived epithelial cells that migrate up the villi and are discarded after several days. A small subset of cells (~1%) derived from the intestinal crypt are transdifferentiated into endocrine cells under the induction of Ngn-3 signaling (9). This raises the possibility that the PDG compartment might not only serve to repair exocrine ductal tissue as already reported but also be a potential source of pancreatic endocrine cells, although no effective β-cell formation from the PDG compartment was identified in humans with type 1 or 2 diabetes (17, 19).

The turnover of pancreatic duct epithelial cells is much less frequent than that of the cells that form intestinal villi. However, in common with intestinal crypts, PDGs have a zone of increased replication compared with that of the duct epithelia, and this is enhanced in response to injury or the known growth factor GLP-1 (12, 19). The unbiased gene expression studies by RNA-seq of PDGs presented in this manuscript reveal a molecular signature intermediate between the exocrine and endocrine pancreas as well as the well-defined stem cell niche in the testis, consistent with a pancreatic stem cell niche that may serve both the exocrine and endocrine pancreas.

There has been controversy as to whether endocrine cells arise from pancreatic duct epithelium in postnatal life (so called ductal neogenesis) (13). The postulate that they do was initially rendered based on the adjacency of islets and pancreatic duct epithelium commonly found in pancreas (4). However, most lineage studies have failed to demonstrate endocrine cells arising from pancreatic ductal or acinar cells. One possible explanation for this is that the lineage markers employed represent those of definitive ductal epithelium rather than that of the putative pancreatic stem cell niche. One of the purposes of the present study was to further characterize the molecular identity of the PDG compartment to offer investigators potential candidates to lineage trace the derivatives of the compartment. Several lines of investigation suggest that there is ongoing β-cell formation in the adult pancreas that cannot be attributed only to replication of existing β-cells (16).

Modeling studies applied to the HIP rat model of type 2 diabetes revealed an adaptive increase in β-cell formation not attributable to β-cell replication (14). Therefore, we selected this model of type 2 diabetes to investigate the PDG transcriptome for evidence of adaptive changes that might be expected in a relevant stem cell niche. Differential gene expression analysis between PDGs from HIP and WT rats were consistent with tissue stem cell response to injury in a relevant compartment. For example, the most significant alteration was in genes engaged in inflammatory responses. This implies communication between injured pancreatic islets and PDGs, presumably either through the known intrapancreatic portal venous system or through the rich intrapancreatic neural network. Also, given the well-recognized role that inflammatory pathways play in the induction of tissue repair, it is of interest that PDGs in HIP rats not only apparently sense and respond to pancreatic islet inflammation but also that regulation of cell proliferation and tissue development genes are also highly represented in the PDG transcriptome of HIP rats compared with nondiabetic WT rats. Pathway analysis also pointed to enrichment of cell movement-related pathways, and the top two networks (Fig. 5, A and B) included ADIPOQ, a gene with known linkage to obesity and type 2 diabetes, and the therapeutic target in type 2 diabetes, PPARG. The PDG compartment has been reported to be expanded with increased proliferation in humans with both type 1 and 2 diabetes, and this is reproduced in the HIP rat model (17, 19). Here we now report the transcriptome of the PDG is consistent with a tissue stem cell niche serving both the exocrine and endocrine pancreas, and undergoing anticipated adaptive changes in response to inflammatory signals arising from stressed β-cells.

If the PDG compartment is a potential source of new β-cells, the question arises, why are β-cells not restored in type 1 and 2 diabetes? Tissue stem cell niches recapitulate the development of the host tissue, and so endocrine cells would be expected to be a minority of new cells formed by a pancreatic stem cell niche. Moreover, in the face of ongoing β-cell autoimmunity in type 1 diabetes, and misfolded protein stress in type 2 diabetes, presumably β-cell loss would likely match any β-cell formation. On the other hand, if the PDG compartment can serve as a source of new β-cells, it is plausible that the relative fate of newly forming cells might be therapeutically manipulated toward an endocrine rather than exocrine fate to enhance new cell formation.

As with all studies, the present studies have limitations. The samples sizes are relatively small, constrained by the costs of the dissection protocol and RNA-seq. The studies are limited to rodent pancreas, since efforts to procure consistently high-quality RNA from human pancreas samples unfortunately were unsuccessful. Nonetheless, the data that have been established imply that the PDG compartment may indeed serve as a pancreatic stem cell niche and provide some insights that might be exploited to establish lineage dynamics in genetic models.

GRANTS

This research was performed with the support of the nPOD, a collaborative type 1 diabetes research project sponsored by the Juvenile Diabetes Research Foundation International. We acknowledge the support of the National Institutes of Health (NIH) Informatics Center for Neurogenetics and Neurogenomics (P30-NS-062691). The work was supported by NIH Grant DK-077967 and by the Larry L. Hillblom Foundation (2014-D-001-NET).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.E.B. and D.K. performed experiments; A.E.B., D.K., T.G., F.G., and G.C. analyzed data; A.E.B., F.G., and G.C. interpreted results of experiments; A.E.B. prepared figures; A.E.B. drafted manuscript; A.E.B., T.G., and P.C.B. edited and revised manuscript; A.E.B., D.K., T.G., F.G., G.C., and P.C.B. approved final version of manuscript; T.G. and P.C.B. conceived and designed research.