Imaging Human Pancreatic Endocrinogenesis During Early Prenatal Life

The pancreas is a mixed gland mainly composed of exocrine tissue. The endocrine component is organized into small clusters, named islets of Langerhans, constituting about 1% of the pancreatic mass. Each adult islet contains a mean of 1,500 cells, including β-, α-, and δ-cells that produce and secrete insulin (INS), glucagon (GCG), and somatostatin (SST), respectively.

During development, the pancreas arises from the foregut endoderm and undergoes differentiation into exocrine and endocrine tissues. The majority of information related to pancreas development has been derived from rodents, using in vitro and in vivo experimental systems (1). For instance, GCG⁺ cells appear first at embryonic day 10.5 in mouse, followed by INS⁺ cells 1 day later and, finally, SST⁺ cells at embryonic day 13.5 (2). In humans, far less is known, but there exist obvious differences between rodent and human adult islets. For instance, in rodents, islets are organized with a core of β-cells surrounded by α- and δ-cells, whereas cell organization in human islets is more stochastic (3). This suggests differences in the spatiotemporal dynamics of endocrinogenesis and islet morphogenesis between mouse and human, which formed the basis of the present study.

The combination of light-sheet fluorescence microscopy (LSFM) and tissue clearing has enabled three-dimensional (3D) visualization of tissue architecture with unprecedented detail (4). We recently used LSFM to investigate β-cell development in the early human pancreas (5). In the present study, we compare, for the first time, the spatiotemporal dynamics of INS, GCG, and SST cell differentiation during the first trimester of pregnancy, highlighting interspecies differences between human and mouse. Additionally, we uncover morphometric differences between fetal and adult human islets, suggesting a novel mechanism for human adult islet formation through the fusion of human fetal islets.

Human embryonic and fetal pancreatic tissues were collected through the INSERM HuDeCA program, following French bioethics legislation and INSERM guidelines (5). Sample selection and age determination, given in Carnegie stages (CSs) and postconceptional weeks (PCWs) for fetal stages, were determined as in the work of Villalba et al. (5). Age and sex of specimens used in each figure are presented in Supplementary Table 1.

Human adult pancreatic tissues were provided by the Human Islet Core facility of St. Louis Hospital (Assistance Publique – Hôpitaux de Paris, France) with signed informed consents (registered under no. PFS12-006 by the French Agency of Biomedicine).

Embryonic and fetal samples were fixed, stained, and cleared following an iDISCO plus modified protocol (4,5). Image acquisition, processing, and visualization were performed as previously described (5). Samples were scanned at ×4 with a digital zoom of ×1.66, providing a resolution of 0.979 µm of pixel size. The primary and secondary antibodies used are listed in Supplementary Table 2.

The measurements of endocrine clusters number, volume, and distribution were performed as described (5). To quantify the percentage of intercalated and nonintercalated endocrine clusters, the shortest distance was measured from each endocrine cluster to the SOX9⁺ epithelium.

For two-dimensional (2D) imaging, human pancreases were fixed in formalin, embedded in paraffin, sectioned, and stained as previously described (5). The antibodies used are included in Supplementary Table 2.

The number of bihormonal INS⁺GCG⁺ clusters at fetal stage (PCW 10) was manually counted using FIJI (6), with a minimum of 100 INS⁺ cells per sample (n = 3). The distribution of endocrine cell types within fetal and adult islets was measured with the software PyCreas (7), and a minimum of 25 islets per sample was quantified (n = 3 per age). These values are measured in relative radius (arbitrary units), in which 0 represents the center and 1 the edge of the islet. Wilcoxon matched-pairs signed-rank tests were used to calculate the P values for comparison between cell types.

The data sets generated and/or analyzed during this study are available from the corresponding author on request. No applicable resources were generated during this study.

Previously, we demonstrated that the first INS⁺ clusters appear in the embryonic pancreas at approximately CS16. Here, we observe that the first GCG⁺ and SST⁺ clusters are also detected at CS16, present only in the dorsal bud, indicating that all three cell types appear at a similar age (Fig. 1A and Supplementary Video 1). Interestingly, we also detect specimens containing bihormonal clusters at early stages (Fig. 1B and Supplementary Video 2). From CS16 to CS18, the number of endocrine clusters remains low (Fig. 1C), with each cluster containing 1–5 cells (Fig. 1D–F).

From CS16 to CS18, 7 of 9 samples showed bihormonal clusters. All types of double-positive hormonal clusters were detected (Fig. 2A, Supplementary Figs. 1A and 2A). Three of 9 samples contained INS⁺SST⁺ clusters (Fig. 2A and Supplementary Video 3), and 2D immunostaining at CS18 demonstrated the existence of bihormonal cells (Fig. 2B and B′), which were not detected at fetal (Fig. 2C and C′) or adult (Fig. 2D and D′) stages. Similarly, INS⁺GCG⁺ clusters were detected in 2 of 9 samples from CS16 to CS18 (Supplementary Fig. 1A and Supplementary Video 4), and single-cell resolution highlighted bihormonal cells (Supplementary Fig. 1B and B′). At fetal stages, only 5% of INS⁺ cells remained INS⁺GCG⁺ (Supplementary Fig. 1C and C′, Supplementary Fig. S3), and no double-positive cells were detected in the adult pancreas (Supplementary Fig. 1D and D′). Finally, GCG⁺SST⁺ clusters were detected from CS16 to CS18 in 2 of 9 samples (Supplementary Fig. 2A and Supplementary Video 5), and 2D immunostaining validated the existence of bihormonal cells (Supplementary Fig. 2B and B′). These cells were not detected at PCW10 (Fig. 2C and C′) nor in the adult pancreas (Fig. 2D and D′). We also performed some triple staining (INS, GCG, and SST) but did not observe triple-positive cells.

We next determined the spatial location of the different endocrine cell populations. INS⁺, GCG⁺, and SST⁺ clusters were found preferentially aligned to the longitudinal axis of the primordium rather than the periphery (Fig. 3A–C and Supplementary Video 6). We quantitatively measured these 3D distributions by mapping them to the centroid of the SOX9⁺ pancreatic epithelium in z-slices from three pancreases. In all, the vast majority of endocrine clusters were located closer to the center and almost absent in the periphery (Fig. 3D–F), suggesting the existence of a niche in the center of the pancreas promoting the endocrine differentiation.

In rodent embryos, endocrine cells are first observed in the epithelium and next delaminate to reach the parenchyma (8). Less is known in humans and we used our 3D approach to determine the dynamic profile of each endocrine cell type in relation to the epithelial tree. We observed that most of them are located within the SOX9⁺ tree at early stages, before CS23 (Fig. 3G–I, Supplementary Fig. 4A and C). On the other hand, at fetal stages, half of the fetal endocrine clusters—either INS⁺, GCG⁺, or SST⁺—are found outside of the tree (Fig. 3G–I, Supplementary Fig. 4B and D), demonstrating parallel migration patterns for all endocrine cell types.

Quantification in 23 samples from CS16 to fetal stages demonstrated a sharp increase from CS22 for both INS⁺ (Fig. 4A, D, and G) and SST⁺ (Fig. 4C, F, and I) clusters. This pattern was different for GCG⁺ clusters, whose number increases at fetal stages (Fig. 4B, E, and H). Unexpectedly, the number of INS⁺ clusters was lower than SST⁺ at the fetal stage, which does not match with the β- to δ-cell ratio observed in the adult pancreas. However, 3D measures of the fraction of each cell type revealed that INS⁺ clusters account for half of the endocrine volume (Fig. 4J). Thus, INS⁺ clusters, even if fewer than SST⁺ clusters, are bigger (Fig. 4K–M).

We compared the volumes of INS⁺ clusters at early (CS20) and late (fetal) stages with those of adult human islets by incorporating recently published 3D data (9). At CS20, most of the INS⁺ clusters (75.33% ± 2.30%) fit into the smallest islet size category (<25,000 µm³) and at fetal stage, the vast majority (88.52% ± 0.66%) display a volume smaller than 100,000 µm³. These volumes are considerably smaller than the mean volume of an adult INS⁺ cluster (between 3,200,000 and 6,400,000 µm³). This reveals that fetal islets are much smaller than adult ones (Fig. 4N–P).

When we compared the distribution of the different endocrine cell types within the islets at fetal versus adult stages, we observed that the different endocrine cell types display different locations. In fetal islets, INS⁺ cells located in the center, surrounded by both GCG⁺ and SST⁺ cells, whereas in the adult, these three endocrine cell types are distributed stochastically, with no preferential cell distribution (Fig. 4Q–S).

Improving the development of stem cell–derived human islet cells is crucial for advancing novel therapies for type 1 diabetes, with current protocols focusing on mimicking endocrinogenesis (10). However, most of our understanding of islet cell differentiation comes from research on mouse pancreas development, mainly due to the scarcity of human samples. We accessed a significant collection of developing pancreases, thanks to the HuDeCA consortium (https://hudeca.com). Here, we used an experimental pipeline comprising in toto staining, tissue clearing, and LSFM imaging to unravel human islet endocrinogenesis during the first trimester of pregnancy.

By 3D analyses, we previously revealed that INS⁺ cells emerge by CS16 (5), 2 weeks earlier than observed in previous studies (11). In the present work, we demonstrate that GCG⁺ and SST⁺ cells also appear earlier than detected by conventional staining techniques (12–14). Interestingly, in the mouse embryonic pancreas, α-cells appear first, followed by β- and δ-cells (2). In zebrafish, β-cells are the first to appear, followed by δ- and α-cells (15). Here, we detect in the human embryonic pancreas the three endocrine cell populations at the same stage (CS16), underscoring notable interspecies differences. These findings indicate that the differentiation dynamics of INS⁺, GCG⁺, and SST⁺ cells in the human embryonic pancreas are more synchronized than previously thought. Our work highlights the importance of 3D analyses in identifying rare cell types during early development.

In mice, INS⁺GCG⁺ cells are present at early stages of endocrinogenesis, along with monohormonal cells, whereas few GCG⁺SST⁺ cells have been detected (16). In humans, the existence of bihormonal cells in the developing pancreas analyzed by single-cell transcriptomics or by immunostaining has been debated (12,14,17,18). Here, an intriguing aspect of our findings is that a significant number of samples contain bihormonal endocrine cells at CS16–18 with all possible combinations of bihormonal cells. We found a lot of variability among the samples with a mean of ∼30% of bihormonal clusters, but we believe that more samples are needed to get robust data. Finally, bihormonal cells are detected in a recurrent fashion in protocols aiming to produce pancreatic endocrine cells from stem cells (19). This observation underscores the complexity of endocrine cell differentiation, suggesting that the mechanisms involved may be more intricate than previously understood.

Islets are micro-organs dispersed in the pancreatic parenchyma. It is well established that in rodent, an islet does not form by clonal expansion but through the aggregation of endocrine cells differentiated from multiple endocrine progenitors (20), and a similar model has been proposed for human islets (21). At late fetal stages in mice, endocrine cells form a large area in the center of the pancreas that will give rise through fission to islets with a core of INS⁺ cells surrounded by GCG⁺ and SST⁺ cells (22). The mechanism of islet morphology seems different between rodent and human. In fact, the mean volume of each embryonic/fetal islet is significantly lower, by an order of magnitude, than in the adult pancreas. Additionally, these small human fetal islets resemble adult mouse fetal islets (a core of β-cells surrounded by α- and δ-cells), as described by others (14,23–25). To reach the size and the arrangement of α-, β-, and δ-cells of human islets, we propose a mechanism of fusion of preexisting small human islets, containing a core of β-cells surrounded by α- and δ-cells, which aggregate during development to give rise to bigger adult islets displaying a stochastic distribution of endocrine cell types (Supplementary Fig. S5). Altogether, this approach enabled us to elucidate the spatiotemporal dynamics of human pancreatic endocrinogenesis in an unprecedented manner.

Funding. This project has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement no. 874839 (R.S.), INSERM cross-cutting program HuDeCA 2018 (to R.S., A.C., and S.M.-G.), Fondation pour la Recherche Medicale EQU201903007793 (to R.S.), l'Agence Nationale de la Recherche (ANR-19-CE16-0021-01 to A.C.), the Foundation Bettencourt Schueller (to R.S.), and the Fondation Francophone pour la Recherche sur le Diabete (to R.S.). The R.S. laboratory belongs to the Laboratoire d’Excellence consortium Revive (Investissements d'Avenir ANR-10-LABX-73-01). A.V. is supported by postdoctoral grant from Laboratoire d’Excellence consortium Revive. A.C. is a CIFAR fellow in the McMillan multiscale human program. The A.C. laboratory is supported by the DIM C-BRAINS and funded by the Conseil Régional d’Ile-de-France.

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

Author Contributions. A.V. and V.A. performed and analyzed experiments. A.V., A.C., and R.S. designed the experiments, interpreted the data, and wrote the manuscript. M.T. and S.M.-G. collected the human embryo samples. R.S., Y.G., and A.C. supervised the study and obtained funding. R.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.