Isolation and immortalization of human PD-MSCs
Human adult pancreatic tissue was enzymatically digested, and the resulting islet subpopulation was separated using Ficoll density gradient centrifugation. The fraction comprising islets, along with a contingent of exocrine cells, was cultured under standard conditions in DMEM supplemented with 10% FBS (Figure 1A). Following several days of cultivation, exocrine cells failed to attach to the culture surface resulting in cell death, while islets adhered to the culture surface giving rise to fibroblast-like cells in the form of a monolayer with elongated spindle-shaped cells predominantly located at the periphery (Figure 1B, C). Cells were cultivated until reaching 90% confluency, and subsequently, were passaged by trypsin-EDTA treatment. The spindle shape morphology was preserved throughout numerous passages (Figure 1D). After a couple of passages, the cells were transduced with lentivirus harboring hTERT construct to ensure stable expression of hTERT cDNA (Supplementary Figure S1), subsequent positive selection based on puromycin resistance was employed over a period of two passages. Both immortalized and non-immortalized cells were cultured under uniform conditions. Non-hTERT-transduced cells exhibited a gradual cessation of proliferation, leading to cell death around 80-100 days (at passage 15), while hTERT-transduced cells sustained proliferative activity, persisting up to 60 passages at the time of manuscript preparation (Figure 1E). The presence of hTERT cDNA was additionally validated using RT-qPCR, hTERT mRNA was significantly higher in immortalized cells compared to non-immortalized control (Figure 1F).
Human pancreas-derived stromal cells display classical mesenchymal features
Isolated cells, whether immortalized or non-immortalized, attach to the culture surface, assuming a fibroblast-like morphology across passages from early to late stages. Subsequently, the expression of established mesenchymal stromal cell markers in hPD-MSCs was investigated through flow cytometry analysis conducted on both immortalized and non-immortalized cell populations. hPD-MSCs showed strong positive expression of classical MSC markers like CD90 (95%), CD106 (99%), CD105 (94%), CD44 (100%), and CD73 (95%) (Figure 2A-E). There was no significant difference between immortalized and non-immortalized hPD-MSCs (Figure 2F). Furthermore, in line with the characteristic attributes of mesenchymal stromal cells, hPD-MSCs demonstrated the capacity for differentiation into lineage-specific cell phenotypes under defined growth conditions. Specifically, these cells were cultured in adipogenic and osteogenic media, and subsequently stained with Oil Red O for adipocytes and Alizarin Red S for osteocytes (Figure 2G, H). These collective findings affirm the classical characteristics of MSCs, establishing hPD-MSCs as a member of the mesenchymal stromal cell category.
PD-MSCs share markers of stem cells and activated pancreatic stellate cells
Mesenchymal stromal cells express pluripotency-related genes that are implicated in self-renewal and cell proliferation, notably, this expression tends to diminish over prolonged culture periods 14. Key genes implicated in these processes include KLF4, OCT4, NANOG, SOX2, REX1, CD44, and VCAM1 14–17. To further elucidate the stemness characteristics of PD-MSCs, we conducted a comparative analysis using RT-qPCR to assess the expression levels of these markers. Our data indicate no significant disparity in the expression levels of SOX2, OCT4, and NANOG between BM-MSCs and PD-MSCs. BM-MSCs exhibit heightened expression of REX1, VCAM1, and CD44 compared to PD-MSCs, underscoring potential differences in their functional properties. Conversely, PD-MSCs demonstrate elevated expression of KLF4 relative to BM-MSCs. Moreover, the immortalization process of PD-MSCs leads to enhanced expression of OCT4 and KLF4, suggesting a potential augmentation of stemness-associated pathways upon immortalization. In contrast, AT-MSCs exhibit higher expression levels of SOX2, CD44, REX1, OCT4, and KLF4 compared to both BM-MSCs and PD-MSCs (Figure 3G-M). These results suggest that PD-MSCs, AT-MSCs, and BM-MSCs display similar expression levels of specific pluripotency-related genes. However, variations exist among them, underscoring distinct stem cell attributes inherent to cells derived from diverse tissue origins.
Moreover, our investigation revealed shared characteristics between PD-MSCs and pancreatic stellate cells (PSCs). In vivo, PSCs can exist in two states, activated and quiescent; accordingly, their morphological appearance and expression profile change 18. Isolated PSCs initially exist in a quiescent state characterized by the presence of cytoplasmic lipid droplets rich in vitamin A, along with the expression of intermediate filament proteins like Desmin and glial fibrillary acidic protein (GFAP). Upon culturing for a few days, these cells transition into an activated state, marked by the absence of lipid droplets, the upregulation of α-smooth muscle actin (α-SMA), and the secretion of ECM proteins 19,20. The activation of PSCs is governed by multiple signaling pathways 21 which can explain the complex heterogeneity of PSCs in vitro. In our study, we observed positive expressions of α-SMA, Nestin, Desmin, and ECM components in PD-MSCs and a lack of GFAP expression (Figure 3A-F) indicating a notable resemblance between PD-MSCs and activated PSCs. Moreover, we have also observed positive staining for CXCL12 which plays a major role in the maintenance, mobilization, and migration of stromal cells 22, and positive staining of neural progenitors’ markers ASCL1 and FOXA2 (Supplementary Figure S2).
Comparison of transcriptomic profiles of obtained mesenchymal stromal cells
To further characterize PD-MSCs, we performed RNA-seq experiments on 5 original PD-MSCs samples, 3 immortalized PD-MSCs, 3 AT-MSCs, and 3 BM-MSCs samples. We obtained on average 5.5 million (minimum 4.3 million) mapped reads for the analysis. The direct comparison of the original PD-MSCs with the iPD-MSCs (Supplementary Table S3) revealed an expected enrichment in differentially expressed genes of mitotic processes (such as “DNA-dependent DNA Replication”, p-value = 4.4e-08), as well as genes involved in cell response to foreign DNA (such as “Negative Regulation of Viral Genome Replication”, p-value = 1.0e-07), likely caused by the transduction method (Supplementary Table S4). On the other hand, both original and immortalized cells seem to be more similar to each other than to both AT-MSCs and BM-MSCs (Figure 4A). The expression of established mesenchymal cell markers indicated that PD-MSCs exhibit similar expression profiles for mesenchymal markers as AT-MSCs and BM-MSCs (Figure 4B).
Moreover, we conducted a more precise comparison with the tissue of origin, the pancreatic scRNA-seq dataset from the Tabula Sapiens experiment 23 via a digital cytometry method, CIBERSORTx 24. This comparison allowed us to pinpoint a significantly higher similarity to the pancreatic stellate cells (Figure 4C, Supplementary Table S5). The Tabula Sapiens pancreatic dataset contains three clusters of stellate cells: quiescent, activated, and an unspecified additional cluster (Supplementary Figure S4A). Notably, the CIBERSORTx analysis predominantly identifies PD-MSCs within the unspecified stellate cell cluster, while for AT-MSCs and BM-MSCs, pancreatic fibroblasts seem to be the most similar cell type. However, across all cell types, the correlation with reference data is weak, averaging 0.15 for PD-MSCs and even lower (0.08) for other AT-MSCs and BM-MSCs.
Next, we tried to find genes whose expression would be characteristic of PD-MSCs. To minimize false positives, we tested original and immortalized PD-MSCs against AT-MSCs and BM-MSCs separately in four analyses for differentially expressed genes (DEG) (Supplementary Table S6). We found that there are 26 differentially expressed genes with effect log2FC > 2 and a p-value (BH-adjusted) < 0.05 (Figure 4D) in all four analyses. Among them are genes of 1) transcription factors and transcription regulation (FOXE1, FOXF1, HOXB-AS3, TCF21, ZNF804A), 2) enzymes and enzyme regulators (PLAT, PPP1R14, ALDH1A1, EDN1), 3) channels and other membrane proteins (ANO1, CHRM2, COL4A5, SCN9A, SYT1), 4) proteins involved in signal transduction (CORIN, GPR37, IGF2BP1, KIT, NPTX1, RGS7), and 5) with other or undefined function (COLEC10, C8orf34, CD163L1, LRRC2, RTN1, and a gene without the HGNC symbol: ENSG00000286190, also known as LOC728392) (Figure 4E). Note, that the novel gene was added in the Ensembl annotation files since release 96 (2019), and before that the corresponding transcript was attributed to a nearby gene NLRP1 (Supplementary Figure S5). Most of these genes whose counts are present in the pancreatic Tabula Sapiens dataset, are found in stellate cells, however, not exclusive for any of its subtypes, which is in line with our CIBERSORTx analysis (Supplementary Figure 4B). The selective expression of some of these genes was also evaluated by RT-qPCR, the results are in agreement for all of the tested genes (Figure 5).
Additionally, we verified the distinctive expression in PD-MSCs of ISL1. Despite its overall low expression, it remains significantly higher in PD-MSCs than in AT-MSCs and BM-MSCs. It is important to note that, in our DEG analysis, we were unable to demonstrate statistical significance for ISL1 when comparing immortalized PD-MSCs with AT-MSCs (P.adj = 0.1), other comparisons being significant (P.adj < 0.05) and an increase in expression was observed for all models (see Supplementary Table S6).
System analysis did not yield much enrichment, probably due to the limited number of samples. Gene set enrichment analysis (GSEA) on DEG for the four tested models (Supplementary Tables S7-10) revealed three common ontology sets with adjusted p-value < 0.05: “Skeletal System Development'', “Bone Morphogenesis'', and “Bone Development''. According to the GSEA’s normalized enrichment score, all of these processes are suppressed in PD-MSCs.