If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Developmental Biology, Stanford University School of Medicine, Stanford, CaliforniaStanford Diabetes Research Center, Stanford University School of Medicine, Stanford, CaliforniaDepartment of Medicine, Endocrinology Division, Stanford University School of Medicine, Stanford, California
Center for Digital Health, Berlin Institute of Health and Charité - Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, Berlin, GermanyHealth Data Science Unit, Medical Faculty and BioQuant, University of Heidelberg, Heidelberg, Germany
Molecular evidence of cellular heterogeneity in the human exocrine pancreas has not been yet established because of the local concentration and cascade of hydrolytic enzymes that can rapidly degrade cells and RNA upon pancreatic resection. We sought to better understand the heterogeneity and cellular composition of the pancreas in neonates and adults in healthy and diseased conditions using single-cell sequencing approaches.
We innovated single-nucleus RNA-sequencing protocols and profiled more than 120,000 cells from pancreata of adult and neonatal human donors. We validated the single-nucleus findings using RNA fluorescence in situ hybridization, in situ sequencing, and computational approaches.
We created the first comprehensive atlas of human pancreas cells including epithelial and nonepithelial constituents, and uncovered 3 distinct acinar cell types, with possible implications for homeostatic and inflammatory processes of the pancreas. The comparison with neonatal single-nucleus sequencing data showed a different cellular composition of the endocrine tissue, highlighting the tissue dynamics occurring during development. By applying spatial cartography, involving cell proximity mapping through in situ sequencing, we found evidence of specific cell type neighborhoods, dynamic topographies in the endocrine and exocrine pancreas, and principles of morphologic organization of the organ. Furthermore, similar analyses in chronic pancreatitis biopsy samples showed the presence of acinar-REG+ cells, a reciprocal association between macrophages and activated stellate cells, and a new potential role of tuft cells in this disease.
Our human pancreas cell atlas can be interrogated to understand pancreatic cell biology and provides a crucial reference set for comparisons with diseased tissue samples to map the cellular foundations of pancreatic diseases.
A comprehensive transcriptomic characterization of human pancreatic cell types has been lacking, mostly due to technical limitations.
Novel protocols facilitated the generation of single-nucleus RNA-seq data and the identification of unprecedented cell type heterogeneity in the healthy adult, neonatal and in the diseased human pancreas.
This study made use of snap-frozen biopsies of human pancreata; further perturbation studies are needed using animal or in vitro models.
These data lay the foundation for future studies and comparisons of cell composition, transcriptomic and spatial alterations in pancreatic diseases.
This study generated the first comprehensive human pancreas cell atlas and revealed the existence of novel cell types and states in this organ.
Pancreatic exocrine tissues contain among the highest levels of hydrolytic enzyme activities in the human body,
hindering the preparation of intact RNA from this organ. Therefore, previous single-cell RNA–sequencing (scRNA-seq) studies investigating tissue heterogeneity of the human pancreas have been restricted to the islets of Langerhans after removal of the exocrine compartment, constituting the source of digestive enzymes. After their isolation, endocrine islets were cultured in vitro, dissociated, and processed on microfluidics devices before next-generation sequencing.
Although this strategy proved to be successful in generating a draft of the endocrine human pancreas cell atlas, it has distinct disadvantages. For example, the in vitro culture and dissociation steps are known to introduce technical artifacts in gene expression measurements.
Moreover, only a small number of exocrine cells from single-cell studies have been reported, leading to underrepresentation of acinar and ductal cells. As a consequence of this underrepresentation, the presence, extent, or quality of heterogeneity in pancreatic exocrine cells is yet not established.
Here, we innovated methods for rapidly processing tissue biopsy samples isolated from freshly isolated human donor pancreata, followed by single-nucleus sequencing (sNuc-seq), thereby avoiding in vitro expansion and dissociation procedures. This approach produced an index draft atlas of human pancreatic cells, including epithelial and nonepithelial cells, from both neonatal and adult samples and showed previously undetected heterogeneity within pancreatic exocrine cells. Application of in situ sequencing combined with computational approaches enabled us to elucidate spatial relationships and signaling pathways connecting cell types in the pancreas of both previously healthy and chronic pancreatitis human donors and showed dynamic cellular constitution and spatial arrangements during postnatal pancreas development.
Materials and Methods
Human and Pig Pancreas Samples
Deidentified human pancreata were procured from previously healthy, nondiabetic donors with less than 12-hour cold ischemia time through the Center for Organ Recovery and Education, International Institute for the Advancement of Medicine, and National Diabetes Research Institute, as reported previously
(in accordance with Stanford University institutional review board guidelines). Within minutes of removal from cold transportation media, tissue blocks of 2 cm × 1 cm × 0.2 cm were excised from 3 or 4 anatomic locations (ie, head, body, midbody, and tail) and then immediately transferred into liquid nitrogen to snap freeze. Samples from Technische Universität München were procured from nondiseased pancreatic tissue from patients undergoing partial pancreatectomy (ethics committee number 403/17S). Tissue blocks of 0.5 cm × 0.5 cm × 1 cm were collected immediately after removal of the pancreas, placed into cryotubes, and transferred into liquid nitrogen. Pancreatic tissue from healthy pigs killed for other reasons (approved by the local authorities [AZ .3-8-07, Regierung von Oberbayern, München, Germany]) was sampled under completely standardized conditions. The pancreas was removed after the heartbeat had stopped, and tissue blocks (0.5 cm × 0.5 cm × 1 cm) were sampled at different timepoints (cold ischemia time of 7 minutes and 30 minutes) and transferred into liquid nitrogen.
Isolation of Nuclei
A full step-by-step protocol has been deposited in the protocols.io repository.
Briefly, snap-frozen pancreatic tissue samples were cut into pieces of <0.3 cm and homogenized with 1 stroke of loose pestle in 1 mL citric acid–based buffer (sucrose 0.25 mol/L, citric acid 25 mmol/L, Hoechst 33342 1 μg/mL) by using a glass Dounce tissue grinder. The tissue was incubated on ice for 5 minutes and then homogenized with 5–10 more strokes. After 5 minutes of incubation, tissue was homogenized with 3–5 strokes with the loose pestle and 5 more strokes with the tight pestle. Homogenate was filtered through a 35-μm cell strainer and centrifuged for 5 minutes at 500g at 4°C. Supernatant was carefully removed, nuclei were resuspended in 1 mL of citric acid buffer, and the centrifugation step was repeated. Nuclei were then resuspended in 300 μL of cold resuspension buffer (KCl 25 mmol/L, MgCl2 3 mmol/L, Tris buffer 50 mmol/L, Recombinant ribonuclease inhibitor (Takara) 0.4 U/μL, dithiothreitol 1 mmol/L, SUPERaseIn RNase Inhibitor (Thermo Fisher Scientific) 0.4 U/μL, Hoechst 33342 1 μg/mL). Nuclei were counted on a Countess II FL Automated Cell Counter, diluted to the desired concentration, and immediately loaded on the 10x Genomics Chromium controller for a target capture of 10,000 nuclei.
10x Genomics Sample Processing, Library Preparation, and Sequencing
Samples were prepared according to the 10x Genomics Single Cell 3′ v2 and 10x Genomics Single Cell 3′ v3 Reagent Kit user guides with small modifications. Final libraries were sequenced with the NextSeq 500 system (Illumina) in high-output mode (paired end, 75 base pairs). Further details regarding downstream RNA-sequencing analyses are included in the supplementary materials.
Histology and RNA Fluorescence In Situ Hybridization
Human pancreatic snap-frozen samples (thickness: 2 mm) were fixed in 10% formalin at 4°C for 14–16 hours and paraffin embedded. Sections (4 μm) were processed for RNA in situ detection by using the RNAscope Multiplex Fluorescent Reagent Kit v2 according to the manufacturer’s instructions (Advanced Cell Diagnostics). The RNAscope human probes used were Hs-CPB1 (no. 569891-C3), Hs-RBPJL (no. 581131), Hs-AMY1A (no. 503551-C2, targeting also AMY1B, AMY1C, AMY2A, and AMY2B), and Hs-REG3A (no. 312061). RNA fluorescence in situ hybridization (FISH) images were acquired on a Leica SP8 confocal laser scanning microscope equipped with a 40×/1.30 oil objective (Leica HC APO CS2).
In Situ Sequencing by CARTANA and H&E Staining
Sections (4 μm) were cut from formalin-fixed, paraffin-embedded pancreatic tissue prepared as described for RNA-FISH. Four DNA probes for each target gene were designed and supplied by CARTANA, and samples were processed according to the manufacturer instructions (Neurokit 1010-01, CARTANA). To reduce lipofuscin autofluorescence, 1× Lipofuscin Autofluorescence Quencher (Promocell) was applied for 30 seconds before fluorescence labeling. Samples were then shipped to CARTANA for the sequencing step. The resulting table of the spatial coordinates of each molecule for the 98 targets, together with the reference 4′,6-diamidino-2-phenylindole (DAPI) image per sample, were provided by CARTANA. Further details regarding downstream analyses of in situ sequencing (ISS) data are included in the supplementary materials.
For H&E staining, the slides were unmounted in distilled water for 1 hour, stained with hematoxylin according to Mayer (Morphisto 10231) for 4 minutes, blued in tap water (6 minutes), and then stained with eosin 1% (Morphisto 11503). After rinsing again with distilled water, the slides were mounted with aqua mounter (Bio SB, 0091).
However, the RNA extracted from isolated nuclei was highly degraded compared to the RNA in the original bulk tissue (Supplementary Figure 1A–C). Several modifications to the original protocol were systematically applied, including the use of dithio-bis(succinimidyl propionate), methanol fixation, or the addition of RNAse inhibitors like ribonucleoside vanadyl complexes, but these failed to improve RNA quality. However, on the basis of protocols first described in the 19th century
we discovered a citric acid–based buffer that reduced RNA degradation during nuclei isolation and increased complementary DNA yields by 40–50-fold (compared to standard protocols) from human pancreatic samples (Supplementary Figure 1D). We isolated nuclei from flash-frozen human pancreas biopsy samples with short cold ischemia times collected from 3 male and 3 female deceased donors, spanning the age range from 1.5 to 77 years (13 samples in total) (Figure 1A, Supplementary Table 1, and Supplementary Figure 1E), generating—to our knowledge—the largest, most comprehensive extant human pancreas cell transcriptome data set.
To aid comprehensive identification of the different pancreatic cell types, we applied canonical correlation analysis. This achieved (1) the reduction of batch effects and (2) the integration of data with previously annotated human pancreas scRNA-seq data sets (Supplementary Figure 1F). Our results confirmed that independent sNuc-seq data sets could be merged and fully integrated with scRNA-seq data sets and that we could capture all previously reported pancreatic cell types (Figure 1B and C). The majority of sNuc-seq data derived from acinar or ductal cell nuclei (Supplementary Figure 2A and B) but also included important nonepithelial cell types (endothelial, stromal, immune) not comprehensively characterized in prior work focused on islet biology.
Characterization of Adult Human Pancreatic Cell Types
The comprehensive cell representation in our data aligns well with the known composition of the healthy human pancreas (Figure 2A). Acinar cells (70% of the nuclei) were identified based on the expression levels of digestive enzymes such as CPA1/2 and PRSS1 and hallmark transcription factors (TFs) RBPJL or FOXP2 (a BHLHA15/MIST1 target). Strikingly, our analysis showed unanticipated heterogeneity in this cell type (discussed later). Ductal cells (18.5% of the nuclei) expressed cardinal regulators or markers like CFTR, ANXA4, and SLC4A4 (Figure 2B). Unlike in prior studies,
we identified 2 distinct ductal subtypes (Figure 2A), and visual inspection of the principal component loadings confirmed that they were separated along the third principal component (Figure 2C). The smaller subtype (accounting for 1% of the total ductal cells) was characterized by higher expression levels of genes linked to mucous secretion such as MUC5B (hereafter, MUC5B+ ductal cells); the trefoil factor genes TFF1, TFF2, and TFF3; and the cysteine rich secretory protein 3 CRISP3 (Figure 2C and D). The other ductal subtype, by contrast, showed higher expression levels of classical ductal markers such CFTR, SLC4A4, and SCTR (Figure 2C and D).
Thus, our study provides evidence for unsuspected molecular, and possible functional, heterogeneity in human pancreatic ductal cells.
One major group of sNuc-seq clusters contained endocrine cells (approximately 6% of the total) characterized by the expression of specific hormone genes—namely, glucagon (GCG, alpha cells), insulin (INS, beta cells), pancreatic polypeptide (PPY, gamma cells), and somatostatin (SST, delta cells) (Figure 3A). Other clusters included endothelial cells (1.9% of total nuclei), characterized by the expression of FLT1, PLVAP, VWF, CD36, and SLCO2A1, and macrophages (0.7% of the total nuclei), expressing CD74, PTPRC, ZEB2, HLA-DRA, HLA-DRB1, and HLA-DPA1 (Figure 3B). We identified pancreatic stellate cells (PSCs) and distinguished 2 states of PSCs called quiescent (qPSCs) and activated pancreatic stellate cells (aPSCs).3 qPSCs expressed higher levels of SPARCL1 messenger RNA (mRNA), similar to hepatic stellate cells, and also PDGFRB and FABP4, which likely regulate retinoid storage (Figure 3B and Supplementary Figure 3). Moreover, qPSCs showed higher levels of desmin (DES) and integrin (ITGA1), both encoding for known regulators of cell structure (Figure 3B and Supplementary Figure 3), and CSPG4 (also known as NG2) and MCAM (also known as CD146), suggesting that qPSCs also constitute pericytes in the pancreas. When PSCs activate, they acquire a myofibroblast-like morphology and are able to migrate and remodel the extracellular matrix. Both qPSCs and aPSCs express COL4A1 and COL4A2, but aPSCs showed higher levels of mRNAs encoding other collagens such as COL5A2, COL6A3, and components of the basement membrane such as laminin proteins LAMA2 and LAMB1 (Figure 3B and Supplementary Figure 3). Furthermore, in aPSCs, we detected higher levels of SLIT2 and LUM, known mediators of fibrogenesis and migration in hepatic stellate cells
(Figure 3B and Supplementary Figure 3). Importantly, fibroblasts are normally present around blood vessels and larger ducts of the pancreas, but the lack of specific markers hinders our ability to distinguish them from cells annotated as aPSCs. We also detected a cluster of Schwann cells (80 nuclei, 0.02% of total) expressing CDH19, S100B, CRYAB, PMP22, and SCN7A (Figure 3B). Overrepresentation analysis showed the enrichment of specific terms such as “axonogenesis,” “synapse organization,” and “synapse assembly” (Figure 3C). Notably, we did not detect transcripts encoding for genes associated with dedifferentiation and reduced myelin sheath formation, which can be up-regulated by cell extraction and culture.3
Heterogeneity of Acinar Cells in the Adult Human Pancreas
sNuc-seq data for acinar cells provided an unprecedented opportunity for rigorous assessment of acinar cell heterogeneity. One population of acinar cells (acinar-REG+) expressed higher levels of mRNAs encoding the regenerating (REG) protein family members such as REG3A, REG3G, and REG1B (Figure 4A and Supplementary Figure 4). Acinar-REG+ cells were reported in a previous scRNA-seq study
Strikingly, we detected 2 additional subtypes of acinar cells not previously identified in human scRNA-seq experiments. These 2 clusters had distinct unique molecular identifier (UMI) levels per nucleus but a similar number of expressed genes, denoting a distinct complexity of their transcriptomes (Supplementary Figure 5A). The acinar cell subtype with higher numbers of UMIs was characterized by higher expression levels of 21 genes (Supplementary Table 2) encoding for digestive enzymes and accounting for 50% of their transcriptome (Supplementary Figure 5A), confirming previous reports estimating that the majority of the mRNA molecules in a pancreatic acinar cell encode for fewer than 30 proteins.
We named this subset secretory acinar cells (hereafter, acinar-s). Gene overrepresentation analysis showed the enrichment of the “ribosome”, “protein processing in the endoplasmic reticulum” and “protein export” terms (Figure 4B), consistent with the view that acinar cells have the highest rate of protein synthesis of any human cell.
The other acinar cell type also expressed digestive enzyme genes but at markedly lower levels compared to acinar-s cells (<4% vs >50%) (Figure 4A and Supplementary Figure 5A). Gene overrepresentation analysis showed enrichment of terms including “protein processing in the endoplasmic reticulum,” “insulin signaling pathway,” “endocytosis,” and “glucagon signaling pathway” (Figure 4B). Thus, these acinar cells appear less robust in their protein secretion and, instead, enriched for responsiveness to external stimuli, like islet signals, and activation of the endocytic pathway. We named this subset idling acinar cells (hereafter, acinar-i).
To validate our sNuc-seq findings further, we combined experimental and computational approaches. First, we performed RNA-FISH on the same samples used for nuclei isolation using probes for CPB1 (carboxypeptidase) and AMY2A/B (amylase), and after quantification (Supplementary Figure 5C), we confirmed the presence of acinar cells characterized by differential RNA-FISH signal (Figure 4C and D).
Second, we applied single-cell regulatory network inference and clustering (SCENIC) to infer TF-target regulatory networks (regulons) from single-cell gene expression.
Both the acinar-s and acinar-i subtypes showed activation of the regulon CREB3L1, likely involved in the basal secretory activity of the cells (Supplementary Figure 5B). The XBP1 regulon shows high activation in acinar-s cells, in agreement with the role of this TF in the unfolded protein response pathway and consistent with the view that biosynthetic activity and the accompanying increase of endoplasmic reticulum stress are higher in acinar-s cells (Supplementary Figure 5B). Notably, only acinar-s cells showed activation of regulons associated with the maintenance of acinar cell identity such as GATA4, NR5A2, and MECOM (Supplementary Figure 5B).
Third, we elucidated relationships between acinar-s or acinar-i subtypes and other pancreatic cell types using partition-based graph abstraction (PAGA).
With PAGA representation, nodes represent distinct cell states, whereas edges indicate potential routes of cell transitions between them. Here, we included cell types deriving from a common multipotent progenitor during embryonic development, namely, acinar and ductal cells in the exocrine compartment and alpha, beta, gamma, and delta cells in the endocrine compartment. This unsupervised approach places acinar-i cells in a central position, showing similar connections to the majority of the other cell types such as ductal and endocrine cells (Figure 4E). Fourth, we exploited the principle of pseudotime analysis to order cells on the basis of the similarity of their transcriptome. Interestingly, acinar-i cells occupy an intermediate position between acinar and ductal cells, reflecting the known plasticity of acinar cells and their ability to convert toward the ductal lineage (Figure 4F).
but less is known about the postnatal development of this organ. We procured 2 1-day-old samples (Supplementary Table 3) and performed sNuc-seq (10,528 nuclei) (Figure 5A and B). The exocrine neonatal compartment accounts for approximately 50% of the organ, whereas it constitutes approximately 90% of the adult pancreas, in agreement with early studies of postnatal growth in rodents (Figure 5C).
Moreover, pancreatic endocrine cells account for 22% of detected cells in neonates compared to 6% in the adult (Figure 5C). Endocrine cells showed changes in cellular composition, with neonatal delta cells, the third major cell type of endocrine islets, accounting for 29% of the endocrine cells compared to 15% in adults (Figure 5D). Alpha and beta cells accounted for 21% and 48% of the neonatal endocrine cells, in line with findings in humans and pigs (Figure 5D).
(Figure 5B), characterized by high mRNA levels of ACSL1, encoding for an Acyl-CoA synthetase enzyme that catalyzes the addition of an octanoyl group to ghrelin, a modification that is essential for its optimal biological activity
In exocrine cells from neonates, we detected acinar-i and acinar-s cells (Figure 6A). Approximately 4% of the transcripts in acinar-i cells encode for digestive enzyme genes compared to 33% in acinar-s, similar to the levels found in cognate adult cells (Figure 6A). However, we did not detect AMY2A, AMY2B, and PNLIP in neonatal acinar-s cells, consistent with prior findings that little to no pancreatic amylase or lipase enzyme activity is detectable in newborns (Figure 6B).
Moreover, we did not detect a population of acinar-REG+ cells, suggesting a further layer of REG protein regulation during postnatal to adult development (Figure 5B). In the ductal compartment, we did not detect the MUC5B+ population, but this may reflect low duct cell yields from human neonatal pancreas (Figure 5B).
Unlike in the adult pancreas, we observed evidence of endothelial cell heterogeneity in neonatal samples (Figure 6C). In addition to an adult-like endothelial signature, we observed an angiogenetic endothelial type enriched for mRNA previously linked to programs of blood vessel morphogenesis and extracellular matrix remodeling (Figure 6E). Furthermore, on the basis of the expression of canonical markers such as LYVE1, PDPN, FLT4, and PROX1, we identified a population of lymphatic endothelial cells (Figure 6G) that most likely function to collect interstitial fluid containing cell debris.
Few immune cells were detected in the human adult healthy pancreas, but in the neonatal pancreas, we identified at least 3 immune populations, including follicular B cells (CD19, CR2, CD22, FCER2), T lymphocytes (CD247/CD3Z, CD3G, IL7R), and macrophages (CD14, CD86, CSF2RA) (Figures 5B and 6F). To clarify the potential interactions of macrophages with other cell types in the neonatal pancreas, we applied CellPhoneDB to predict ligand-receptor interactions from single-cell transcriptomics data.
We also noted specific sets of ligands and receptors, such as the TAM receptors (AXL, MERKT, and the ligand GAS6), usually active in tissues subject to remodeling and involved in the phagocytosis of apoptotic cells, and members of the Notch pathway, including the NOTCH2 receptor and the antagonistic ligands DLL4, JAG1, and JAG2. Moreover, a strong interaction was predicted between CD74 and APP, suggesting an important role for APP in pancreatic neonatal angiogenesis (Figure 6H).
Beta cells are not fully functional at the perinatal stage, but their functions, including glucose-regulated insulin secretion, mature with age.
To investigate changes across the 2 different age groups (neonatal and adult), we combined the data sets and performed diffusion pseudotime analyses to investigate age-dependent development. Pseudotime ordering recapitulated donor age (Figure 6I), thereby permitting us to model gene expression using a generalized additive model and to identify highly dynamic genes. We identified groups of genes showing increasing or decreasing expression levels across pseudotime (Supplementary Figure 6A). Some of the genes expressed at lower levels in adult samples are involved in beta cell proliferation such as PDZD2 and IGFBP5
also had increased mRNA levels in adult human beta cells (Supplementary Figure 6A and B). In human adults, we also detected higher levels of RASD1 and SYT16, genes previously reported to be up-regulated in human islets exposed to relatively high glucose.
Genes encoding members of the secretogranin-chromogranin family like SCG2, SCG5, and CHGB, known for their essential role within the insulin secretory granule, were also more highly expressed in adult beta cells (Supplementary Figure 6A and B). Thus, by applying sNuc-seq to tissues procured at specific developmental stages, our work reveals markers and unrecognized possible regulators of human beta cell functional maturation.
In Situ Sequencing of the Human Pancreas Localizes Messenger RNA
To elucidate the spatial relation of the cell types identified in our study, we performed ISS
using matching tissues processed for nucleus isolation (Figure 1A and Supplementary Figure 7A). We selected 98 marker genes (Supplementary Table 4) identified from sNuc-seq that distinguish pancreatic cell types, then applied ISS to tissue from 1 juvenile (1.5 years old) and 2 adult (30 and 53 years old) donors. ISS-based mRNA localization was used to generate spatial cell maps by applying spot-based spatial cell–type analysis by multidimensional mRNA density estimation (SSAM), a segmentation-free algorithm that identifies cell type signatures from spatially resolved in situ transcriptomics data (Figure 7A and Supplementary Figure 7B–D).
SSAM cell maps contained all the cell types identified by sNuc-seq and permitted ready recognition of multicellular tissue features (Figure 7A, magnified views). The proportion of cell types detected by SSAM cell maps also corresponded well with those detected with sNuc-seq (Figure 7B). To probe spatial relationships among pancreatic cell types, we performed empirical statistical modeling of cell type proximity. To confirm the validity of this approach, we first quantified spatial relations within unambiguous multicellular structures, confirming the mutual proximity of alpha and beta cells (Figure 7C and D). We then looked at nonepithelial cell types in close proximity to the endocrine islets and found that endothelial cells are the closest ones (Figure 7E) because they support high oxygen demand, and the glucose-sensing and endocrine functions of islets.
Analysis of ISS provided further insights into the intraislet and interislet architecture. Quantification of islet size showed a higher frequency of small islets (radius smaller than 40 μm) in juvenile tissue compared to adult tissue (Supplementary Figure 8A). We then performed proximity analysis on cells located outside of the islets and discovered that, in the juvenile sample (1.5 years), alpha cells are the most proximal (in the first 40 μm), followed by beta cells, suggesting enrichment of alpha cells in the mantle of the islets, whereas beta cells preferentially locate in the core (Supplementary Figure 8B).
We then calculated the distance between the centroids of manually annotated endocrine islets; in both neonatal and adult samples, we found a minimum distance of 400 μm, providing quantification of pancreatic islet dispersion during pancreas morphogenesis (Supplementary Figure 8D).
Finally, we investigated the spatial relations of the new acinar cell states identified in this work. The acinar-s and acinar-i cells did not show a specific cell neighborhood (Supplementary Figure 9), in agreement with their vast abundance (approximately 80%) in the tissue. By contrast, acinar-REG+ cells appeared to localize significantly closer to islet cells,
like delta and gamma cells (Supplementary Figure 8C), and to macrophages. This latter finding is consistent with prior reports that REG3A/PAP protein modulates chemoattraction and activation of macrophages in pancreatic diseases.
Here, snap-frozen samples from 2 CP donors (Supplementary Table 5) were processed for sNuc-seq, and despite the extensive fibrosis and reduced cellular content (Supplementary Figure 10A), we successfully profiled 2726 nuclei. Cell types were identified by using gene markers established in the healthy tissue; moreover, we detected a chemosensory (tuft cells) and 2 immune (mast cells and monocytes) CP-specific cell types (Supplementary Figure 10B and C). The cellular composition of the pancreas is altered in CP, with a reduction of both endocrine and exocrine cells and simultaneous increase of activated stellate and immune cells (Supplementary Figure 10D). Importantly, the acinar cells detected by sNuc-seq belonged to the REG+ subtype, suggesting a previously unexplored role of these cells in the development of the disease.
Macrophages and T lymphocytes constitute the predominant immune cell infiltrates in the CP samples. Macrophages were CD163+ (Supplementary Figure 11A), indicating an alternatively activated or M2 status. Generally, M2 macrophages dampen inflammation and promote healing of tissues partially through the recruitment and activation of fibroblastic cells. Therefore, we investigated ligand-receptor pairs involved in the communication between M2 macrophages and aPSCs. CellPhoneDB analysis showed the involvement of different signaling axes, including transforming growth factor β, platelet-derived growth factor, vascular endothelial growth factor, and granulin. In turn, aPSCs stimulate macrophage proliferation and M2 polarization via the complement system and the CXCL12-CXCR4 axis (Supplementary Figure 11B). Importantly, we applied ISS to 1 CP sample and confirmed the presence of inflammatory infiltrates consisting of macrophages and aPSCs in close spatial relation (Supplementary Figure 10A and Supplementary Figure 11C, magnified views), as validated by unbiased ISS cell proximity analysis (Supplementary Figure 11D).
Among the other cell types, T cells were characterized by the expression of PRF1 and GZMH, supporting their involvement in cell-mediated cytotoxicity (Supplementary Figure 11E). Tuft cells, identified for the first time in human pancreatitis samples, were characterized by high expression of cytosolic phospholipase A2 (PLA2G4A) and prostaglandin D2 synthase (HPGDS), encoding for 2 enzymes involved in the generation of prostaglandin D2, a molecule responsible for anti-inflammatory signaling in pancreatic diseases
suggested that high-throughput sequencing could reveal undetected singularities in pancreatic exocrine cells. From the analysis of more than 120,000 pancreatic cells, we found 3 distinct acinar populations (acinar-i, acinar-s, and acinar-REG+), distinguished by differential expression levels of digestive enzyme genes, distinct activation of pancreatic gene regulatory networks, expression of specific protein family genes, and cell neighborhoods. Because in vitro systems that reconstitute mature human acinar cells in their native architecture have not yet been achieved, we used orthogonal approaches to validate our sNuc-seq results, including RNA-FISH, in situ sequencing, and computational approaches. Notably, the new methods were also applied and validated in pancreatic disease samples.
Our studies showed that acinar-REG+ cells were absent in the neonatal pancreas, suggesting that their function might be specific for the adult tissue. Previous scRNA-seq studies described a subset of REG3A+ acinar cells,
but the application of in situ sequencing analyses—for the first time, to our knowledge—to the human pancreas, showed their localization near macrophages, nominating acinar-REG+ cells as possible regulators of pancreatic inflammatory processes. Importantly, we detected acinar-REG+ cells in chronic pancreatitis samples, and these cells were also recently described in a single-cell study of pancreatic ductal adenocarcinoma.
In acinar-i cells, we find evidence suggesting that hydrolytic enzyme production may be reduced compared to acinar-s cells and speculate that this acinar cell state could be a protective adaptation to periods of intense zymogen production and increased endoplasmic reticulum stress. If so, acinar-i cells might be analogous to a subset of postulated metabolically stressed islet beta cells.
Acinar-i cells showed a decreased activation of acinar cell gene regulatory networks and occupied a central position in a PAGA lineage relation graph; hence, these cells might have the capacity to convert into other pancreatic cell types, including both ductal and endocrine cells. A previous single-cell study of the mouse pancreas identified acinar cell subtypes characterized by differential expression of TFs, partially recapitulating our findings in the human pancreas and suggesting that acinar cell heterogeneity is a conserved feature in the 2 species.47
In this work, we did not identify centroacinar cells, which have been postulated to include pancreatic progenitor cells.
Future studies could clarify the transcriptome of centroacinar cells in the human pancreas and their possible lineage or spatial relation with the acinar cell and ductal cell types captured in our data sets.
sNuc-seq data generated from neonatal tissue uncovered developmental dynamics in the pancreas, unique immune interactions (involving B cells, T cells, and macrophages), and different cellular composition compared to the adult pancreas. We identified distinct endothelial cell states, including angiogenetic endothelial cells in neonatal pancreata, that may reflect the increased supply of nutrients required by the rapidly replicating cells at this developmental stage.
Furthermore, we identified a previously uncharted landscape of cell crosstalk, including macrophages and endothelial cells, and possible ligand-receptor interactions involved in organ remodeling during growth. We also revealed aspects of human endocrine pancreas development and regulation that constitute the basis of future comparisons with human
The combination of adult and neonatal data sets allowed pseudotime analyses and nominated candidate regulators and effectors of beta cell maturation, including age-dependent restriction of beta cell proliferation and development of secretory activities.
ISS analyses confirmed differences in islet size and intraislet architecture between juvenile and adult pancreatic tissue and provided evidence for possibly stereotyped dispersion of islets throughout the human pancreas, like in rodents.
In summary, our studies combine technical innovations to produce a human pancreas cell atlas that provides conceptual advances and reveals cellular, genetic, signaling, and physiologic mechanisms regulating pancreatic cells in health and disease.
The authors would like to thank the organ donors and their families; David Ibberson (Heidelberg University), Ulrike Krüger ( Berlin Institute of Health [BIH], Berlin), and Marten Jager (BIH, Berlin) for next-generation sequencing services; the Biomaterial Bank (MTBIO) of the Technical University Munich for support; Katharina Jechow (BIH, Berlin), Lorenz Chua (BIH, Berlin), and Alison McGarvey (MDC, Berlin) for critically revising the manuscript; Jeongbin Park (BIH, Berlin) for his help with the analysis of ISS data; and all members of the Conrad laboratory for the constructive discussions. The graphical abstract was created with BioRender.com. This publication is part of the Human Cell Atlas (www.humancellatlas.org/publications).
Conflicts of interest The authors disclose no conflicts.
Funding This study was supported by the Human Cell Atlas pilot studies of the Chan Zuckerberg initiative (Charité and Technische Universität München: CZI grant 2017-174170), the European Marie-Skłodowska Curie Actions (EC no. 841755), the European Commission (ESPACE, no. 874710, Horizon 2020) and the Diabetes Genomics and Analysis Core and the Stanford Islet Research Core of the Stanford Diabetes Research Center (P30DK116074). The authors gratefully acknowledge the data storage service [email protected] supported by the Ministry of Science, Research, and the Arts Baden-Württemberg and the German Research Foundation through grants INST 35/1314-1 FUGG and INST 35/1503-1 FUGG.
Author names in bold designate shared co-first authorship.