Introduction

Rates of diabetes in individuals with the life-limiting genetic disease, cystic fibrosis (CF), are incredibly high, i.e., ~ 20% of adolescents and 40–50% of adults1,2. As highly effective modulator therapies (HEMT) are increasingly used in this population, lung and nutritional aspects of this multi-organ disease are much improved, resulting in a greater life expectancy. As the CF population ages, however, age-related components are expected to become even more prevalent, including CF-related diabetes (CFRD) where rates are approaching 100% in individuals with common severe CFTR mutations3. Current data, albeit limited, suggest HEMT do not significantly improve glucose metabolism or pancreatic endocrine function4. Also, mortality for individuals with CFRD > 30 years remains higher than for individuals with CF without diabetes3. Therefore, additional therapies are urgently needed, and for these to be developed, fundamental knowledge of pancreas pathology in CF must be improved.

Clinical data suggest that incretin-based therapies, widely used for treatment of type 2 diabetes, may also be of benefit in CFRD5,6,7, but the literature remains inconclusive. The incretin hormone, glucagon-like peptide-1 (GLP-1), is an alternate product of the proglucagon gene generated by the enzymatic action of prohormone convertase (PC) 1/3 on proglucagon. GLP-1 is a well-known insulin secretagogue whose insulinotropic action has classically been linked to its production and secretion from intestinal L cells. Once released from L cells, circulating GLP-1 exerts its insulinotropic effects via the GLP-1 receptor (GLP-1R), which is predominantly expressed on the β cell8. GLP-1Rs are also present on afferent neurons within submucosal and myenteric nervous plexi, which transmit signals to the brain and pancreas through the vagus nerve9,10. Since GLP-1 is rapidly degraded by the proteases dipeptidyl peptidase-4 (DPP-4) and neprilysin11,12,13,14,15, it is thought that gut-derived GLP-1 mediates its insulinotropic action mainly via neuronal, rather than hormonal, pathways. However, emerging evidence from studies of mice deficient in pancreas or gut proglucagon-derived peptides suggests that GLP-1 produced in the gut minimally contributes to insulin secretion16. Maintenance of normal insulin secretion by β cells is dependent, in part, on paracrine signaling from islet α cells. This is supported by observations that glucagon can stimulate insulin release17,18,19,20, and that α cells also produce glucagon-like peptide-1 (GLP-1)21,22,23,24,25. Thus GLP-1’s local effects within the islet are likely to impact insulin release to a greater extent than GLP-1 coming from the intestine via the peripheral circulation.

To date, levels of GLP-1 within α cells have not been examined in CF, and it is unknown whether diabetes affects islet GLP-1 levels in CF subjects. Indeed, whether α-cell production of GLP-1 and/or GLP-1R signaling is impaired in CFRD is an important gap in our knowledge. If such impairments do exist, strategies to enhance levels of GLP-1 and/or GLP-1R signaling could be effective in increasing insulin release in CFRD. Thus, in the present study, we utilized paraffin-embedded pancreas sections from humans with CF or CFRD and age-matched non-CF/non-diabetic controls to quantify immunoreactivity of GLP-1 and GLP-1R via immunohistochemistry. One aim of this analysis was to identify differences between CF subjects with versus without diabetes.

Another aim of our study was to better understand CFRD pathogenesis, focusing on the islet since it is thought to play a critical role yet very little is known about underlying mechanisms. To this end, we also performed the first spatial transcriptomics analysis of CFRD pancreas. In an exploratory analysis, we identified genes and pathways that may be dysregulated in α and β cells from CFRD pancreas. We also examined differentially expressed α- and/or β-cell genes involved in GLP-1 production and signaling, when compared to control pancreas. Together, new knowledge gained from our study has the potential to illuminate avenues for improved therapies that restore pancreatic endocrine function in individuals with CFRD.

Results

Subject characteristics

We examined human pancreas specimens obtained from autopsy (Cohort 1) or organ donors (Cohort 2). For Cohort 1 tissue, subjects in the CF group were age-matched to non-CF/non-diabetic control subjects, whereas subjects in the CFRD group were older than non-CF/non-diabetic control subjects (Table 1). BMI was not different among the three groups (Table 1). For Cohort 2 tissue, subjects in the CFRD group were age-matched to non-CF/non-diabetic control subjects, whereas BMI was lower in CFRD subjects versus non-CF/non-diabetic control subjects (Table 1). In both groups of CFRD subjects, diabetes duration was ~ 4 years (Table 1).

Table 1 Subject characteristics from cohorts 1 and 2.
Full size table

Islet GLP-1 immunoreactivity tends to be increased in CFRD but not CF

We first determined whether islet GLP-1 immunoreactivity differed in CF and CFRD versus non-CF/non-diabetic control subjects from Cohort 1. Representative GLP-1 staining is shown in Fig. 1A-C. The proportion of GLP-1 area per islet area did not significantly differ among the three groups (Fig. 1D). Mean islet area was similar among control, CF and CFRD pancreas (Fig. 1E). We also determined glucagon area per islet area and found it to be significantly increased in CFRD versus non-CF/non-diabetic control subjects (Control 15.1±2.3, CF 20.8±3.1 and CFRD 29.5±3.1%; p = 0.53 Control vs. CF, p = 0.01 Control vs. CFRD). When expressed per glucagon-positive area, GLP-1-positive area did not differ among the three groups (Control 67.4±12.3, CF 61.2±7.3 and CFRD 48.8±7.0%; p = 0.46 by one-way ANOVA).

Figure 1
figure 1

Immunohistochemistry for GLP-1 in subjects from Cohort 1. Representative staining of pancreas specimens from no-CF/no-diabetes control (A), CF-no diabetes (B) and CFRD (C) subjects, showing islet GLP-1 in brown. Quantitation of the proportion of GLP-1 area per islet area (D) and mean islet area (E) in CF and CFRD subjects compared with control subjects. Scale bar = 100 μm.

Full size image

To validate the islet GLP-1 data from Cohort 1, we performed GLP-1 staining in independent non-CF/non-diabetic control and CFRD pancreas samples from Cohort 2. Representative GLP-1 staining is shown in Fig. 2A and B. We observed a trend for increased GLP-1 area per islet area in CFRD versus non-CF/non-diabetic control pancreas (Fig. 2C; p = 0.05, without Bonferroni correction). Also, mean islet area was similar between the two groups (Fig. 2D), as was glucagon area per islet area (Control 35.6±2.6 vs. CFRD 26.5±6.3%; p = 0.43). When expressed per glucagon-positive area, GLP-1-positive area did not differ between the two groups (Control 79.1±17.9 and CFRD 84.9±21.4%; p = 0.76).

Figure 2
figure 2

Immunohistochemistry for GLP-1 in subjects from Cohort 2. Representative staining of pancreas specimens from no-CF/no-diabetes control (A) and CFRD (B) subjects, showing islet GLP-1 in brown. Quantitation of the proportion of GLP-1 area per islet area (C) and mean islet area (D) in CFRD subjects compared with control subjects. Scale bar = 100 μm.

Full size image

Islet GLP-1R immunoreactivity is reduced in CFRD

We next analyzed islet GLP-1R immunoreactivity in pancreas sections from Cohort 2. Representative immunohistochemistry shows a significant reduction of islet GLP-1R protein levels in pancreas from CFRD subjects, compared to non-CF/non-diabetic control subjects (Fig. 3A and B). Quantification of GLP-1R area per islet area revealed that levels were ~ 77% lower in CFRD versus control pancreas (Fig. 3C). Mean islet area did not differ between the two groups (Fig. 3D). We also determined insulin area per islet area and found it to be significantly reduced in CFRD versus non-CF/non-diabetic control subjects (Control 52.5±3.9 vs. CFRD 36.2±4.9%; p = 0.03). When expressed per insulin-positive islet area, GLP-1R-positive islet area remained significantly reduced in CFRD subjects (Control 36.2±2.7 vs. CFRD 13.5±3.3%; p = 0.004).

Figure 3
figure 3

Immunohistochemistry for GLP-1R in subjects from Cohort 2. Representative staining of pancreas specimens from no-CF/no-diabetes control (A) and CFRD (B) subjects, showing islet GLP-1R in brown. Quantitation of the proportion of GLP-1R area per islet area (C) and mean islet area (D) in CFRD subjects compared with control subjects. **p < 0.01 vs. Control. Scale bar = 50 μm.

Full size image

Spatial transcriptomics reveals differentially expressed genes in α and β cells between CFRD and control subjects

We performed exploratory spatial transcriptomics analysis on pancreas from 3 female CFRD and 3 female non-CF control subjects (all from Cohort 2). Transcripts from 7 to 5 glucagon-positive (α cell) ROIs were assessed in control and CFRD pancreas, respectively. Additionally, transcripts from 7 to 9 insulin-positive (β cell) ROIs were assessed in control and CFRD pancreas, respectively. Representative images are shown in Fig. 4A. A combined total of 9035 genes were identified in glucagon- and insulin-positive ROIs. Supplemental Table S1 lists the number of cells within each glucagon- or insulin-positive ROI analyzed.

Figure 4
figure 4

Spatial transcriptomics analysis of pancreas specimens from subjects in Cohort 2. (A) Representative images of islets immunostained for glucagon and insulin that were imaged on the GeoMx Digital Spatial Profiler. (B) Principal component analysis for the entire transcriptome of 7 (green) and 5 (pink) glucagon-positive regions of interest in control and CFRD pancreas respectively, and 7 (blue) and 9 (orange) insulin-positive regions of interest in control and CFRD pancreas respectively. Up- (red) and down- (blue) regulated genes in glucagon- (C) and insulin- (D) positive regions of interest in CFRD vs. control pancreas.

Full size image

We initially assessed the transcriptomics dataset derived from 28 ROIs by applying PCA on the entire data. PCA showed segregation of ROIs among the four conditions: glucagon-positive CFRD, glucagon positive controls, insulin-positive CFRD and insulin-positive controls (Fig. 4B), implying global gene expression signals based on disease status and cell type. We observed that the ROIs from control pancreas were more tightly clustered together whereas the glucagon- and insulin-positive ROIs from CFRD pancreas exhibited greater spread, indicating greater transcriptional variability.

Next, we identified differentially expressed genes between CFRD and control subjects using both nominal (p-value < 0.01) and adjusted (FDR < 0.05) significance thresholds. We found 216 genes to be differentially expressed (114 up, 102 down, p-value < 0.01) in α cells between CFRD vs. control subjects, and 272 genes differentially expressed (119 up, 153 down, p-value < 0.01) in β cells between CFRD vs. controls (Fig. 4C and D, Supplemental Tables S2 and S3).

Gene set enrichment analysis highlights islet cell-specific pathway dysregulation in CFRD

To gain insight into how CFRD alters specific biological processes in α and β cells, we performed GSEA on the entire transcriptional dataset. Given the small sample size, we used a nominal (unadjusted) p-value < 0.01 to identify enriched pathways, but also calculated FDR. As summarized in Fig. 5, many gene sets were altered in α and β cells in CFRD relative to controls. Specifically, immune, inflammatory and complement pathways were upregulated in CFRD vs. control α cells, whereas cholesterol homeostasis, ER stress and proteasome gene sets were downregulated (Fig. 5A, Supplemental Tables S4 and S5. We found that some of these pathways were similarly enriched in β cells, but multiple gene sets mapping to extracellular matrix remodeling/fibrotic programs were distinctly upregulated in CFRD vs. control β cells, whereas cell transport, ER stress, proteasome and mitochondrial processes were suppressed (Fig. 5B, Supplemental Tables S6 and S7). Of note, some of the observed differentially enriched pathways reached FDR-adjusted significance of < 0.05.

Figure 5
figure 5

Gene set enrichment analysis (GSEA) of pancreas specimens from subjects in Cohort 2. GSEA was performed on the entire transcriptional dataset from glucagon- (A) and insulin- (B) positive regions of interest in CFRD and control pancreas. Up- and down-regulated pathways in CFRD vs. control pancreas are denoted in red and blue, respectively. Selected gene sets are labeled.

Full size image

α cell genes involved in regulating GLP-1 levels are altered in CFRD

We did not identify GLP-1R signaling as a differentially enriched pathway in CFRD using GSEA. However, a limitation of this approach is its underlying assumption that enrichment of gene sets is based on uniform up- or down-regulation of its member genes. Furthermore, GSEA may not capture the granular and cell type-specific role played by genes in a pathway. Thus, to assess whether the GLP-1R signaling pathway is dysregulated in CFRD islet cells, we first performed a gene subset analysis in which we selected genes known to mediate GLP-1R signaling in both α or β cells and that were reliably expressed in our dataset (Fig. 6A). Based on an FDR-adjusted p value of 0.05, PCSK1 was found to be upregulated in α cells (Fig. 6A) and INS was downregulated in β cells (Fig. 6B) in CFRD vs. control subjects. Using a nominal (p-value < 0.01) significance threshold, an additional three genes (GCG, PCSK2, PCSK1N) were found to be downregulated in α cells in CFRD vs. control subjects (Fig. 6A). Next, we performed a single gene set test (ROAST26) that is suited to our limited sample size. This test revealed that genes involved in regulation of GLP-1 levels in α cells (GCG, PCSK1, PCSK2, PCSK1N, DPP-4) significantly changed in mixed directions (up- or down-regulated; p = 0.0002), and that the active proportion of genes was 0.75. No such perturbation was seen in GLP-1-associated genes in β cells (INS, GLP1R, DPP-4, PLCG1, PIP4P2, ITPR3, ADCY1, ADCY6, PRKACA, RAPGEF4; p = 0.4).

Figure 6
figure 6

GLP-1R-specific pathway analysis in pancreas specimens from subjects in Cohort 2. Mean expression pattern of selected GLP-1R pathway genes among all (A) glucagon- and (B) insulin-positive regions of interest (ROIs) in CFRD and control pancreas, with heatmaps of z-scores depicting changes in gene expression in each ROI. Up- and down-regulated genes in CFRD vs. control pancreas are denoted in shades of red and blue, respectively. *p < 0.05 (FDR-adjusted) vs. Control; #p < 0.01 (unadjusted) vs. Control.

Full size image

Discussion

Signaling from islet α-cell secretory products, including GLP-1, is required for maintenance of normal insulin secretion17,18,19,20,21. To our knowledge, no studies have examined whether this paracrine signaling from islet α-to-β cells is defective in CFRD, a disease in which impaired insulin secretion is a critical component. Here we show for the first time that in pancreas from humans with CFRD, intra-islet GLP-1 protein levels do not differ significantly whereas GLP-1R protein levels are reduced, when compared to pancreas from humans without CF or diabetes. Further, in a transcriptomics analysis of CFRD pancreas, we find that changes in α-cell gene expression favor increased GLP-1 production/levels, and that β-cell genes mediating GLP-1 action downstream of GLP-1R are unchanged. Together, these data suggest that although intra-islet GLP-1 levels are not reduced in CFRD, α-to-β cell signaling of GLP-1 (and its ability to enhance insulin secretion) may be impaired as a result of reduced GLP-1R availability.

Our study focused on the pancreas, as loss of β-cell secretory function is known to be important in the development of diabetes. In CF in particular, several studies have demonstrated that impaired insulin secretion, rather than insulin resistance, is likely the major contributor to progression to CFRD. As an example, in CF subjects with normal glucose tolerance, β-cell function is substantially reduced compared to non-CF controls, even when measured as the disposition index, which normalizes the insulin response to the prevailing insulin sensitivity27,28. Indeed, insulin secretory defects are present very early in the development of abnormal glucose tolerance in CF, and it has been shown that amongst CF subjects with normal glucose tolerance vs. impaired glucose tolerance vs. CFRD, there is no difference in insulin sensitivity29.

Several studies have demonstrated that α-cell mass is increased in CF and CFRD30,31,32,33, including in our Cohort 1 population31. Based on this, we expected that α-cell derived GLP-1 levels would be similarly increased. Moreover, we anticipated that α-cell GLP-1 levels would be greatest in CFRD because the diabetic state would be expected to increase GLP-1-positive α cells, as in human type 2 diabetes34, as well as levels of the proglucagon processing enzyme PC1/3 that generates GLP-1, as documented in other hyperglycemic models35,36,37,38,39. Our data show that there are no statistically significant changes in GLP-1 immunoreactivity in CF or CFRD pancreas relative to non-CF/non-diabetic control pancreas. In CFRD pancreas from Cohort 2, islet GLP-1 levels tend to be increased (p = 0.05), though this is seen only when expressed relative to total islet area and not glucagon-positive islet area. The latter suggests the trend for increased GLP-1 in CFRD pancreas is driven by an increase in glucagon-expressing cells. Of note, islet area was unchanged in CFRD pancreas, which is consistent with some but not all literature that report islet area in CF/CFRD pancreas. For example, in one study, islet area was comparable in pancreas from age-matched CF/CFRD and control subjects30. However, when assessed by presence of diabetes or “lipoatrophic” pattern in CF pancreas, islet area was decreased30. In another study, islets of both diabetic and non-diabetic subjects with CF were found to be hypertrophic40. These two examples indicate that reports of islet area in CF/CFRD are mixed, which is likely due to several factors such as differences in subject age, BMI and severity of disease amongst studies. Related to GLP-1, our transcriptomics data indicate that expression of the gene encoding PC1/3 (PCSK1) is significantly increased in α cells from CFRD pancreas. Given that upregulation of GLP-1 production has been suggested to be a compensatory mechanism aimed at enhancing β-cell function and survival under diabetic conditions34,38,39,41,42, we postulate the lack of reduced GLP-1 combined with enhanced PCSK1 expression represent an attempt to compensate for defective β-cell function in CFRD.

In contrast to our data on islet GLP-1 levels in CFRD pancreas, we observed a significant reduction in islet GLP-1R protein levels in CFRD pancreas, regardless of whether we expressed GLP-1R area relative to total islet area or insulin-positive islet area. This is consistent with data from human type 2 diabetes where islet GLP-1R protein expression also appeared to be decreased43, though immunostaining was not quantified in that study and it is unclear whether a validated anti-GLP-1R antibody was utilized. In another study utilizing the same validated antibody we used44, islet GLP-1R protein levels were reduced in human type 1 diabetes, but not type 2 diabetes45. Together with our data from CFRD pancreas, these differences in islet GLP-1R protein expression amongst various forms of diabetes suggest underlying mechanisms may differ. In addition to GLP-1, glucagon can bind GLP-1R to elicit an increase in insulin secretion46. In CFRD, glucagon area per islet area is increased30,31,32,33. Thus, although there are no deficits in islet levels of either GLP-1 or glucagon in CFRD, their action to potentiate insulin secretion may be limited due to insufficient expression of islet GLP-1R. Of note, glucagon can also potentiate insulin secretion by binding glucagon receptors on β cells20,47,48,49. We observed no change in β-cell GCGR transcript levels between CFRD and control pancreas, but we cannot rule out that protein levels may be reduced in CFRD. That said, studies have shown that α- to β-cell communication is largely mediated by the GLP-1R, with significantly less contribution from the glucagon receptor17,20. Importantly, reduced islet GLP-1R levels may also impact the insulinotropic action of systemically circulating GLP-1. For example, in a study of humans with pancreatic insufficient CF (including CFRD), treatment with a dipeptidyl peptidase-4 inhibitor for 6 months increased circulating active GLP-1 levels without changing insulin secretory rates during a mixed-meal tolerance test5. The latter may be explained by reduced islet GLP-1R expression; however, it is important to acknowledge other data that suggest GLP-1R remains functional to a degree in CF/CFRD. For example, GLP-1 infusion increased second-phase insulin concentrations during a hyperglycemic clamp in pancreatic insufficient CF humans with abnormal glucose tolerance6. Also, the GLP-1R agonist semaglutide increased baseline plasma C-peptide and decreased HbA1c levels over time in a case report of a CFRD subject7.

To further explore potential defects in GLP-1R signaling components in CFRD, we performed spatial transcriptomics on islet cells from paraffin-embedded human pancreas sections. This technique enabled an analysis of gene expression in adjacent α and β cells within any given islet. In an analysis focused on a subset of genes known to regulate GLP-1 levels and GLP-1R signaling in α and β cells, we found that PCSK1 was increased whereas PCSK2 and PCSK1N were decreased in α cells from CFRD pancreas. These changes would be expected to promote production of GLP-1 and thus, may explain the lack of decrease in GLP-1 expression we observed by immunohistochemistry. Interestingly, we found no changes in the subset of GLP-1R signaling genes expressed in β cells, besides a decrease in INS. In a separate gene subset analysis (ROAST) better suited to our small sample size, we found similar results. Specifically, expression of genes involved in regulating GLP-1 levels in α cells was changed in CFRD; however, genes involved in GLP-1R signaling in β cells were unaltered. Importantly, these analyses revealed that β-cell expression of GLP-1R itself was unchanged in CFRD. A lack of change in GLP-1R gene expression has previously been reported in an RNA sequencing analysis of whole islets isolated from humans with CF/CFRD32. Together, these data would suggest that reductions in GLP-1R are not occurring at the level of gene transcription, but rather that there may be defects in translation, processing or even degradation of GLP-1R. Previously, it was shown that exposure of rodent insulinoma cell lines and isolated mouse islets to palmitate caused a reduction in GLP-1R protein expression50. Thus, we speculate that the fatty infiltration of the pancreas with CFTR mutation(s) in CFRD may similarly result in reduced GLP-1R protein levels. Additionally, lipids/cholesterol have been shown to modulate G protein-coupled receptors, including GLP-1R51. Given that our transcriptomics data revealed dysregulation of islet cell cholesterol homeostasis (discussed below), this may be a contributing factor to reduced GLP-1R protein levels.

Aside from genes involved in GLP-1R signaling, we found several other differentially expressed genes between CFRD and control pancreas. For example, SCG5 was downregulated in both α and β cells in CFRD pancreas. SCG5 encodes neuroendocrine protein 7B2, which functions as a chaperone of proprotein convertase 2 (PC2)52. The enzymatic action of PC2 is important for cleavage of proglucagon to glucagon in α cells53, as well as processing of pro-islet amyloid polypeptide in β cells54. Thus, decreased expression of α- and β-cell SCG5 in CFRD may suggest defects in glucagon and islet amyloid polypeptide processing. In α cells, it may also be related to a switch from PC2- to PC1/3-mediated cleavage of proglucagon in an attempt to increase GLP-1 production. Another gene found to be downregulated in α cells of CFRD pancreas was GC, which encodes a vitamin D binding protein shown to be essential for α-cell morphology, electrical activity and glucagon secretion55. Further, we also observed downregulation of ER-associated genes, including PDIA4 and HSPA5 in α cells, and GRINA and HSP90AA1 in β cells, some of which are involved in the unfolded protein response. This suggests there may be inadequate adaptation to accumulation of unfolded or misfolded proteins, such as islet amyloid polypeptide31, which could lead to ER stress and cell death in CFRD. All differentially expressed genes between CFRD and control pancreas are presented in Supplemental Tables S2 and S3.

Additional analysis of our spatial transcriptomics data was performed using GSEA. In general, our findings are in keeping with previously published pathway analysis of data from whole islets isolated from humans with CF/CFRD32; however, we report for the first time to our knowledge, α and β cell-specific pathway dysregulation in CFRD. For example, we observed downregulation of pathways for cholesterol homeostasis (biosynthesis, metabolism, transport, etc.) in α cells, and mitochondrial processes (TCA cycle, oxidative phosphorylation) in β cells from CFRD pancreas. Also, there was upregulation of immune, inflammatory and complement pathways in α cells, and extracellular matrix remodeling (integrins, proteoglycans, cell adhesion, etc.) pathways in β cells. Common to both α and β cells in CFRD was a downregulation of multiple gene sets for ER stress and proteasome pathways. Many of these enriched pathways align with data showing that in CF/CFRD, protein misfolding (e.g., islet amyloid), inflammation and immune cell infiltration in islets is common, as is pancreatic remodeling with adipose and fibrotic tissue deposition31,32. Regarding dysregulation of α-cell/islet cholesterol homeostasis, this has not been studied in cystic fibrosis; however, accumulation of free cholesterol has been described in human CF lung tissue and non-islet cell models of CF56,57. Whether cholesterol also accumulates in the islet in CF warrants further investigation, especially since prior studies have shown elevated islet cholesterol to contribute to islet dysfunction and death58.

The present study was greatly facilitated by access to pancreatic tissue from humans with CF/CFRD; however, there are some limitations in utilizing this tissue. First, lack of availability or poor quality of tissue meant that the sample size was relatively small for some outcomes measures. Specifically, use of autopsy pancreas was not ideal, and better-quality donor pancreas from nPOD was limited to n of 5 for the CFRD group. This made it difficult to account for potentially confounding factors, such as sex or BMI in Cohort 2 or medication usage in all subjects, as well as reducing power for differential gene and pathway enrichment analysis. Furthermore, it limited our ability to make comparisons regarding islet GLP-1R protein levels and α or β cell gene expression between non-diabetic subjects with CF versus those without CF. Despite these shortcomings, in CFRD pancreas we demonstrated perturbations in GLP-1R signaling components and several other important genes and processes involved in islet function/survival, which will help guide future studies to better understand the pathogenesis of CFRD. Second, we cannot exclude the possibility that some genes identified in the transcriptomics analysis were derived from ‘contaminating’ cell types collected alongside the glucagon- and insulin-positive cells. For example, acinar cell contamination was particularly problematic since genes within this cell type were so strongly up- or down-regulated in CFRD pancreas due to pancreatic insufficiency. To minimize the impact of these contaminating cells, we filtered the entire transcriptomics dataset to exclude genes known to be expressed in acinar cells but not α or β cells. This resulted in populations that were enriched for α or β cells, though we acknowledge it is unlikely these populations were entirely pure. Third, the CFRD pancreas was obtained from humans with diabetes duration of approximately 4 years. Thus, it is unknown whether the changes we observed in islets from these subjects also manifest in CF with shorter or longer diabetes duration – such knowledge could impact treatment strategies in CFRD. Additionally, we had no information on glucose tolerance status of non-diabetic CF subjects; specifically, it is unknown whether these subjects had impaired glucose tolerance, which may have affected our findings. Similarly, we had no information on whether these subjects had suppression of GLP-1R-mediated amplification of insulin release – this may inform on the degree to which GLP-1R signaling is involved in development of CFRD. Finally, while immunohistochemical and gene expression data are informative, additional studies are necessary to discern the functional impact of our findings on insulin secretion in CF/CFRD. In order to perform such studies, better models of the β-cell defects in CFRD than those currently available are required.

In conclusion, this is the first study to investigate intra-islet GLP-1R signaling in CFRD using both immunohistochemical and spatial transcriptomics approaches. Our findings suggest that intra-islet GLP-1R signaling may be impaired in humans with CFRD due to a reduction in β-cell GLP-1R protein expression. Further, using spatial transcriptomics, we identify several differentially expressed genes in α and β cells between CFRD and control pancreas, and uncover α and β cell-specific pathway dysregulation in CFRD. These novel findings illuminate potential targets for therapeutic intervention to improve islet function in CFRD, including strategies to restore islet protein levels of GLP-1R and/or promote intracellular signaling beyond GLP-1R.

Methods

Subjects

Sections of formalin-fixed paraffin embedded pancreas specimens from de-identified human donors were obtained from two sources. First, for Cohort 1, samples came from an existing retrospective autopsy collection from Seattle Children’s Hospital (SCH) and University of Washington (UW) Medical Centers, as previously described31. These autopsy cases comprised 10 individuals with CF (no diabetes), 9 with CFRD and 9 with no history of CF or diabetes (controls). Second, for Cohort 2, pancreas samples prospectively collected from human organ donors were obtained from the JDRF-supported Network for Pancreatic Organ Donors with Diabetes (nPOD). From this repository, we obtained 5 CFRD cases and 6 controls with no history of CF or diabetes. Table 1 summarizes subject characteristics. Of note, all subjects with CF (with or without diabetes) had exocrine pancreatic insufficiency.

Immunohistochemistry

Four µm sections were obtained for both Cohort 1 and Cohort 2. One section per subject was utilized for GLP-1 immunohistochemistry and another for GLP-1R immunohistochemistry. Sampling from the body of the pancreas was routinely performed for Cohort 1 specimens, although autopsy records did not always indicate the specific location of the sample. Sampling from the head, body or tail of the pancreas was performed for Cohort 2 specimens. For both cohorts, given the severe pancreas pathology and fat replacement characteristic of CF, sections were selected where pancreatic and islet morphology was of sufficient quality to enable robust immunostaining and transcriptomics analyses.

Slides underwent automated immunohistochemistry (LeicaBond Rx; Leica Microsystems, Buffalo Grove, IL), including deparaffinization, antigen retrieval (EDTA at 100 °C for 20 min; only for GLP-1 and GLP-1R immunohistochemistry) and peroxide block. Sections were then incubated in the following primary antisera: mouse monoclonal anti-GLP-1 (8G9; reacts with the amidated C-terminus of GLP-1(7–36)amide; Abcam, Cambridge, MA) or the extensively validated mouse monoclonal anti-GLP-1R clone 3F52 (Developmental Studies Hybridoma Bank, IA)44,59. Anti-GLP-1 antisera was applied at a dilution of 1:1500 for Cohort 1 and 1:10,000 for Cohort 2. Anti-GLP-1R antisera was applied at a dilution of 1:100 for sections from Cohort 2 only. Attempts to achieve reliable GLP-1R staining in sections from Cohort 1 were unsuccessful, likely due to the known challenges in detecting G protein-coupled receptors60, especially in autopsy samples. Sections from both Cohorts 1 and 2 were also incubated with guinea pig polyclonal anti-insulin (A0564, 1:4,000; Dako, Carpenteria, CA) or rabbit monoclonal anti-glucagon (EP3070, 1:10,000; Epitomics, Burlingame, CA) antisera. Following incubation in primary antisera, linking IgG (Leica Post Primary) was applied (since the Leica polymer reagent recognizes rabbit IgG). Antibody binding was detected using a polymer horseradish peroxidase 3,3′-diaminobenzidene detection system (Bond Polymer Refine Detection kit, DS9800; Leica Biosystems) followed by hematoxylin-eosin counterstaining and coverslipping. Consecutive sections from each subject were used for GLP-1 and GLP-1R staining wherever possible.

Pancreas sections were digitized using a whole slide scanner (Nanozoomer Digital Pathology system; Hamamatsu Corporation, Bridgewater, NJ). Total islet and GLP-1-positive areas, as well as insulin-positive and glucagon-positive areas were determined automatically based on pixel value and density (Visiopharm software, Hoersholm, Denmark) and verified by manual examination of segmented images, as we have done previously31. For GLP-1, the average number of islets (mean ± SEM) quantified per subject in Cohort 1 was 559 ± 150, 980 ± 210 and 398 ± 60 for control, CF and CFRD respectively, and in Cohort 2 was 446 ± 54 and 186 ± 98 for control and CFRD respectively. For GLP-1R quantification, islets (29 ± 4 per subject, mean ± SEM) were hand-circled to generate regions of interest (ROIs). Binary thresholding was used to generate GLP-1R-positive areas and object counts. GLP-1 and GLP-1R areas were expressed as percentage of islet area. GLP-1 and GLP-1R areas were also expressed as percentage of glucagon and insulin areas, respectively. In calculating the GLP-1/glucagon areas, one CFRD case from Cohort 2 was identified as an outlier (by the ROUT method), likely due to the glucagon area value being very low, resulting in an elevated ratio of GLP-1 to glucagon; thus, this case was excluded from the GLP-1/glucagon area data only. Individuals performing immunohistochemistry and quantitative analyses were blinded to the group assignment of specimens.

Digital spatial profiling

mRNA transcript expression profiles were analyzed using the GeoMx Digital Spatial Profiler (NanoString Technologies, Seattle, WA, USA). Formalin-fixed, paraffin-embedded pancreas Sect. (4 μm thick) from 3 female CFRD and 3 female control subjects (1–2 sections per subject, all from Cohort 2) were selected using the same criteria for preservation of pancreas/islet morphology described above. Sections were deparaffinized and underwent antigen retrieval using EDTA at 99 °C for 20 min followed by 1 ug/mL proteinase K at 37 °C for 15 min. Sections were then fixed in 10% neutral buffered formalin (NBF) for 5 min at room temperature and rinsed, after which NanoString human Whole Transcriptome Atlas (WTA) probes with synthetic DNA oligonucleotide barcodes attached were applied to the tissue. Sections were incubated at 37 °C overnight, then washed and treated with blocking buffer prior to incubation at room temperature with the following primary antisera conjugated to Alexa Fluor: mouse monoclonal anti-insulin clone ICBTACLS (1:100; Thermo Fisher, MA) and mouse monoclonal anti-glucagon clone C-11 (1:100; Santa Cruz, TX). Nuclei were stained with Syto83 and sections imaged on the GeoMx Digital Spatial Profiler.

Islets were distinguished morphologically, and areas of illumination (AOI) were selected based on insulin- and glucagon-positive immunofluorescence within islets. All intact islets on a given section were selected for sequencing. A total of 24 AOIs (each containing both insulin- and glucagon-positive cells) were identified across the 6 subjects – 15 from CFRD pancreas sections and 9 from control pancreas sections. Each AOI was exposed to a UV LED light to cleave the oligonucleotide barcodes, which were collected through microcapillary aspiration. Sequencing was performed on an Illumina NovaSeq 6000 system using an S2 flow cell.

Data analysis on the GeoMx digital spatial profiler

The data was analyzed on the GeoMx Digital Spatial Profiler Data Analysis Suite (version 2.5.0.145). Data initially went through several rounds of quality control and generation of new datasets for each AOI according to the GeoMx Digital Spatial Profiler Data Analysis User Manual (MAN-10154-01 for v2.5 software). Briefly, data from the sequencing run was assessed for any non-specific amplification or contamination of the libraries, in addition to accurate sequencing depth of the oligonucleotide barcodes. Then each AOI segment was screened for the quality of signal-to-noise of the probes over negative controls, area size, and nuclei count. All segments with an area size less than 10,000 μm2 were omitted due to a low dynamic range in the signal and poor transcript diversity. Datasets were then filtered to establish an expression threshold and a frequency at which the targets or segments are allowed to be below that threshold (set to 5%). Finally, data normalization was performed by using the geometric mean of the upper quartile (Q3) of all target probe counts for each dataset. In order to remove contaminating genes deriving from adjacent acinar tissue, the entire dataset (all AOIs) was filtered to remove genes known to be expressed in pancreatic acinar cells but not α or β cells. The latter was determined on the basis of previously published datasets61,62. Overall, 7 and 5 glucagon-positive (α cell) regions of interest (ROIs) were analyzed in control and CFRD pancreas, respectively. Additionally, 7 and 9 insulin-positive (β cell) ROIs were analyzed in control and CFRD pancreas, respectively. Supplemental Table S8 provides the complete gene expression data for all samples.

Bioinformatics analyses

Dimensionality reduction using principal component analysis (PCA) was applied to the entire transcriptome of all ROIs. Differential gene expression analysis in CFRD versus control data from glucagon- and insulin-positive ROIs was performed using DESeq263 with nominal p-value cutoff < 0.01 for significance, as well as Benjamini-Hochberg adjusted p-values with a false discovery rate (FDR) threshold < 0.05. Gene Set Enrichment Analysis (GSEA)64 was performed using molecular signature databases (MSigDBs): canonical biological pathways (e.g. Reactome Hallmark, KEGG, Wikipathways, etc.) and Gene Ontology (GO). For GSEA, glucagon- and insulin-positive CFRD samples were compared to their respective controls using all identified unique transcripts rank ordered based on their DESeq2 statistic and random gene set permutation (n = 5,000). To identify significantly enriched gene sets, a nominal (non-adjusted) p value of < 0.01 was applied, but an FDR < 0.05 threshold was also explored. For GLP-1 pathway-specific analysis, the ROAST function26 was applied within R package “limma” using 10,000 random rotations of the residuals to generate the null distribution.

Statistics

Subject characteristic data are presented as mean±SD. All other data are presented as mean±SEM. One-way ANOVA was used for overall statistical differences, with pairwise comparisons conducted with a Mann-Whitney test or Kruskal-Wallis test for multiple comparisons. Prism (version 9.5.1; GraphPad Software, USA) was used for all statistical analyses. Numbers of experimental replications are represented by individual data points in figures.

Study approval

The study was reviewed by the Seattle Children’s Hospital and University of Washington institutional review boards and did not meet the definition of human subjects research per the Office for Human Research Protections within the United States Department of Health and Human Services. As such, informed consent was deemed unnecessary for the study according to national regulations; however, broad consent to use the human biospecimens was originally obtained at autopsy or through the agreement between organ procurement organizations and nPOD. The research described was approved and performed in accordance with Seattle Children’s Hospital and University of Washington institutional review boards’ guidelines/regulations.