Single Cell Analysis of Cultivated Fibroblasts from Chronic Pancreatitis and Pancreatic Cancer Patients

Cancer-associated fibroblasts (CAFs) play a major role in the progression and drug resistance of pancreatic cancer. Recent studies suggest that CAFs exhibit functional heterogeneity and distinct transcriptomic signatures in pancreatic cancer. Pancreatic fibroblasts also form an integral component in pancreatic diseases such as chronic pancreatitis named disease-associated fibroblasts (DAFs). However, intra-tumoral heterogeneity of CAFs in pancreatic cancer patients and their pivotal role in cancer-related mechanisms have not been fully elucidated. Further, it has not been elucidated whether CAF subtypes identified in pancreatic cancer also exist in chronic pancreatitis. In this study, we used primary isolated fibroblasts from pancreatic cancer and chronic pancreatitis patients using the outgrowth method. Single-cell RNA sequencing (scRNA-seq) was performed, and bioinformatics analysis identified highly variable genes, including factors associated with overall survival of pancreatic cancer patients. The majority of highly variable genes are involved in the cell cycle. Instead of previously classified myofibroblastic (myCAFs), inflammatory (iCAFs), and antigen-presenting (ap) CAFs, we identified a myCAFs-like subtype in all cases. Most interestingly, after cell cycle regression, we observed 135 highly variable genes commonly identified in chronic pancreatitis and pancreatic cancer patients. This study is the first to conduct scRNAseq and bioinformatics analyses to compare CAFs/DAFs from both chronic pancreatitis and pancreatic cancer patients. Further studies are required to select and identify stromal factors in DAFs from chronic pancreatitis cases, which are commonly expressed also in CAFs potentially contributing to pancreatic cancer development.


Introduction
Pancreatic cancer is currently the fourth leading cause of cancer deaths and is projected to become the second most common cancer death in the United States by 2030 [1,2]. The dense stromal tumor microenvironment, which accounts for up to~90% of the pancreatic tumor mass, is a major cause of resistance to chemotherapy and radiotherapy [3][4][5]. The major stromal compartment consists of CAFs, which are a source of extracellular matrix proteins and potential therapeutic targets [6]. Fibrotic stroma and generally termed diseaseassociated fibroblasts (DAFs) form an integral component in pancreatic diseases including pancreatic cancer (CAFs) and chronic pancreatitis [7]. CAF depletion can also have tumorpromoting effects, suggesting that CAFs exhibit distinct functional heterogeneity [5]. CAFs, which are derived from several cell types such as pancreatic stellate cells or mesenchymal stem cells, play a major role in the progression and drug resistance of pancreatic cancer [8,9].

Cell Cycle-Associated Factors Are Majority of Highly Variable Genes
Next, to identify which factors are co-expressed together in CAFs/DAFs, we performed heat map analyses and show the top differentially expressed genes in 3 panels for each patient ( Figure 5). The analyses revealed that the variance was mostly covered by cell-cycle genes including MKI67, CENPF, and CCNB1 ( Figure 5). To mitigate the effects of cell cycle heterogeneity in scRNA data, we performed cell cycle regression analysis as previously demonstrated [17]. Prior to cell cycle regression, we observed 9 clusters of cells with similar expression profiles produced by the t-distributed stochastic neighbor embedding (t-SNE) algorithm in CAFs from patient1, 10 clusters in CAFs from patient2, and 11 clusters in fibroblasts from patient3 ( Figure 6A-C). After cell cycle regression, we did not observe clearly distinct clusters ( Figure 6D-F), suggesting that (1) signals derived from cell cycle-associated factors strongly impact on cluster heterogeneity presentation, and (2) primary isolated and cultured cells are not unique population but exhibit highly similar expression profiling between the clusters. We were still able to divide clusters 0 to 3 in CAFs from patient1, clusters 0 to 3 in CAFs from patient2, as well as clusters 0 to 4 in DAFs from patient3 ( Figure 6D-F). We summarize highly variable features after cell cycle regression in Table 2 (patient1), Table 3 (patient2), and Table 4 (patient3). Among 25 highly variable genes identified before cell cycle regression, two factors namely IGFBP3 and IGFBP5 were identified as highly variable genes even after cell cycle regression ( Table 2, Cluster 0 in Figure 2D), suggesting that these two genes may contribute to CAF heterogeneity independent of cell cycle stage. Interestingly, a number of factors, cell cycle-associated but also non-cell-cycle-associated factors, are identified as highly variable genes shared between CAFs/DAFs isolated from chronic pancreatitis and pancreatic cancer patients (Table 5).                         Table 5, markers were selected from the publications Öhlund et al. [10] for (A,C), Elyada et al. [11] for (B,D,E).

Discussion
Increasing evidence shows that CAFs represent a highly heterogeneous subpopulations that can be both tumor-promoting and tumor-suppressing, highlighting the importance of the identification and characterization of the diversity of CAF subtypes [4]. In the current study, we isolated primary disease-associated fibroblasts by the outgrowth method from pancreatic cancer and chronic pancreatitis patients. By using single-cell RNA sequencing and bioinformatics analysis, we identified several highly variable genes including IFI27, KRT18, KRT19, MMP1, MMP3, and NEFL. Expression of these genes is associated with shorter overall survival of pancreatic cancer patients (Figure 3). IFI27 has been previously identified as a prognostic marker for pancreatic cancer [18], and involved in pancreatic cancer migration and invasion [19]. The above-mentioned genes are not exclusively expressed in CAFs/DAFs. Yet, our data support that factors expressed in CAFs/DAFs can contribute to pancreatic cancer migration, invasion, and patient outcome.
We selected the outgrowth technique that has been established for isolating pancreatic CAFs and has become a common CAF isolation strategy. In our scRNAseq analysis, we found that the majority of identified highly variable genes are associated with the cell cycle. It was therefore important to conduct cell cycle regression [17], which showed that signals originating from cell cycle-associated factors have a strong impact on cluster heterogeneity ( Figure 6). Interestingly, a number of highly variable features after cell cycle regression are commonly found in CAFs/DAFs isolated from chronic pancreatitis and pancreatic cancer patients (Table 5). These 135 factors still include a number of cell cycle-associated factors, but also cell cycle independent factors that are implicated in tumorigenesis. For example, KNSTRN promotes tumorigenesis and gemcitabine resistance in bladder cancer [20]. ZWINT supports pancreatic cancer proliferation [21]. Further studies are required to identify which stromal factors among these 135 genes play a key role in pancreatic cancer. It needs to be further clarified, whether pancreatic cancer-derived CAFs and chronic pancreatitis-derived DAFs have already similar profiling before isolation, or whether they become similar during culture. To that end, it is necessary in the future to establish chronic pancreatitis and pancreatic cancer mouse models with fibroblast lineage-tracing systems, to subsequently isolate CAFs/DAFs, and conduct single-cell RNA sequencing both before and after cell culture. Interestingly, our findings are consistent with the study by Barrera et al. Here, microarray experiments from DAFs isolated by the outgrowth method were performed demonstrating that pancreatic cancer-and chronic pancreatitis-derived fibroblasts share the greatest similarity [7].
One of the goals of the study was to see whether we identify previously described major CAF subtypes, myCAFs-, iCAFs-, and apCAFs. We did not observe any clear myCAFs-, iCAFs-, or apCAFs-specific clusters (Figure 7). Our data are in part consistent with a recent publication, where Grünwald et al. chose the outgrowth method and identified eleven sub-clusters from ten pancreatic cancer patients, but no clear myCAF versus iCAF subpopulation differences [22]. Initially, myCAFs and iCAF subtypes have been described by Öhlund et al. [10]. For the majority of their experiments, quiescent cells were isolated by Collagenase P and DNAse I digestion and co-cultured with pancreatic cancer organoids [10]. Murine late-stage tumors of KPC mice (Pdx1-Cre; lox-stop-lox-Kras G12D/+ ; Trp53 R172H/+ ) supported the presence of iCAFs and myCAFs, that were isolated by digesting tissue with Pronase, Collagenase P, and DNAse I [23]. Elyada et al. isolated single cells from human pancreatic cancer patients and KPC mice by digestion with Collagenase D, Liberase DL, and DNAse I [11]. Beside myCAFs and iCAFs, the study identified an additional subpopulation of CAFs expressing MHC class II and CD74 (apCAFs) [11]. Notably, in these studies enzymatic digestion methods were applied rather than the outgrowth method for CAF isolation. Whether methodological disparities have an influence on the outcome of these studies, need to be clarified in the future (e.g., via above-mentioned lineagetracing strategies).
Our study is limited by the small sample size but is the first study to conduct scR-NAseq and bioinformatics analyses combined for chronic pancreatitis and pancreatic cancer patients. In summary, our scRNAseq and bioinformatics analysis identified several highly variable genes, some of whose expression is associated with shorter overall survival of pancreatic cancer patients. We could not identify myCAF-, iCAF-, or apCAF-specific clusters, but the expression profiling resemble rather a myCAF-like subtype. Most interestingly, after cell cycle regression, we observed a large number of overlapping highly variable genes between samples from pancreatic cancer and chronic pancreatitis patients, which need to be verified and functionally characterized in both pancreatic cancer and chronic pancreatitis in vivo.