Genetic lineage tracing reveals stellate cells as contributors to myofibroblasts in pancreas and islet fibrosis

Summary Pancreatic stellate cells (PSCs) are suggested to play an important role in the development of pancreas and islet fibrosis. However, the precise contributions and solid in vivo evidence of PSCs to the fibrogenesis remain to be elucidated. Here, we developed a novel fate-tracing strategy for PSCs by vitamin A administration in Lrat-cre; Rosa26-tdTomato transgenic mouse. The results showed that stellate cells give rise to 65.7% of myofibroblasts in cerulein-induced pancreatic exocrine fibrosis. In addition, stellate cells in islets increase and contribute partly to myofibroblasts pool in streptozocin-induced acute or chronic islet injury and fibrosis. Furthermore, we substantiated the functional contribution of PSCs to fibrogenesis of pancreatic exocrine and islet in PSCs ablated mice. We also found stellate cells’ genetic ablation can improve pancreatic exocrine but not islet fibrosis. Together, our data indicates that stellate cells are vital/partial contributors to myofibroblasts in pancreatic exocrine/islet fibrosis.


INTRODUCTION
Fibrosis is a common pathological outcome of several etiological conditions, resulting in organ dysfunction and eventual failure. 1 Excessive accumulation of extracellular matrix (ECM) components is a major feature of fibrosis. 2 In addition, current evidences show that myofibroblasts are key players in ECM synthesis and can be derived from several cell types in organ fibrogenesis. 3 Hepatic/pancreatic stellate cells, pericytes, fibroblasts, and endothelial cells (Endo-MT) have all been suggested as contributors to the myofibroblast pool. 3,4 Pancreatic fibrosis has been recognized as a key determinant of the pathogenesis and progression of chronic pancreatitis and pancreatic cancer. 5 Likewise, endocrine islet fibrosis in diabetes may affect insulin secretion and accelerate disease progression. 6,7 Pancreatic stellate cells (PSCs), firstly discovered by Watari et al. in 1982, are traditionally considered as pancreatic counterpart cells of hepatic stellate cells. 8 In 1998, two groundbreaking reports described the separation, culture, and characteristic expression of this type of cell, significantly promoting PSCs research progress. 9,10 Then most subsequent studies focused on PSCs to explore the pathogenesis and therapeutic strategies of pancreatic fibrosis despite the lack of in vivo evidence. 11 Besides, PSCs may migrate into islets and participate in the fibrotic process. 12 However, little progress has been made to establish and precisely quantify the relative contribution of PSCs to the pancreatic myofibroblast pool and pancreatic fibrosis in vivo.
Vitamin A (VA) storage is the main physiological function of HSCs/PSCs. 13,14 In addition, Lecithin:retinol acyltransferase (LRAT) is the key enzyme responsible for retinyl-esters formation from retinol in HSCs/ PSCs. 15 Previous studies employing Cre driven by Lrat promoter have traced HSCs successfully. 4 Here, we develop a novel fate-tracing strategy for stellate cells by VA administration in Lrat-cre transgenic mice. In addition, we demonstrated that stellate cells are the dominant contributors to collagen-producing myofibroblasts in cerulein-induced pancreatic exocrine fibrosis, and partial participants in islet fibrosis induced by streptozocin (STZ). Furthermore, we substantiated the functional contribution of stellate cells to pancreatic fibrogenesis in genetic cell ablation mice. Consistently, we found that stellate cells' genetic ablation can improve pancreatic exocrine fibrosis.
The storage of vitamin A (retinol) is the main physiological function of the stellate cell system. 16 Retinol is esterified by Lecithin retinol acyltransferase (LRAT, the key enzyme responsible for retinyl ester synthesis) and then stored in stellate cells. 15,[17][18][19] To genetically label stellate cells, we crossed Lrat Cre mice with Rosa26-LSL-tdTomato reporter mice ( Figure 1A). Lrat Cre transgenic mice marked hepatic stellate cells (HSCs) well as demonstrated by the nearly complete overlap of vitamin A autofluorescence (HSCs marker) with Lrat Cre-induced tdTomato reporter by confocal microscopy ( Figure 1B). However, only few PSCs were labeled with tdTomato reporter in Lrat Cre transgenic mice ( Figure 1C). This may be because of the fact that vitamin A is primarily stored in HSCs rather than PSCs. 18,20 ( Figure S1A). Therefore, we tried to improve the labeling efficiency based on vitamin A anabolic function of LRAT protein. 18,19 We up-regulated the transcription of Lrat gene, thereby increasing Cre expression by Vitamin A loading (5000 IU/day). In addition, the number of tdTomato labeled cells in pancreas gradually increased (before 20 days) and reached stability (after 20 days) when mice were intragastrically loaded with vitamin A palmitate ( Figures 1D and 1E). We also found that vitamin A accumulation in PSC in mice treated with vitamin A palmitate ( Figure S1B). The increase of marked PSC was not because of PSC proliferation as demonstrated by the nearly no overlap of ki67 with tdTomato ( Figure S1C). The tdTomato-labeled cells were stellate cells as demonstrated by immunofluorescence, showing nearly all the tdTomato+ cells expressed stellate cell markers (desmin and PDGFRb) by confocal microscopy ( Figures S2A and S2B). The percentage of tdTomato+ cells in Desmin+PDGFRb+ cells increased (before 20 days) and reached stability (after 20 days) when mice were intragastrically loaded with vitamin A palmitate ( Figures 1F and 1G). Desmin and PDGFRb are also expressed in pericytes, so we evaluated the proportion of stellate cells or pericytes in the pancreas by staining for Crbp1 (a stellate cell marker) and NG2 (pericytes marker) ( Figures S2C and S2D). The proportion of labeled cells was similar to the proportion of actual stellate cells in total cells, so we believed that 20-day vitamin A administration could obtain a high stellate cell labeling efficiency. Pancreatic stellate cells were mainly distributed in the exocrine pancreas, whereas a small portion was observed in the islets, as demonstrated by the tdTomato+ cells accounting for 4.78 G 0.74% of the total cells in the exocrine part and 0.14 G 0.015% of total cells in islets ( Figures 1H and 1I). To exclude the possible influence of bone marrow-derived cells to our tracing mice, we examined the bone marrow smears and bone marrow tissue from our cell tracing mice, and found no cells were labeled with tdTomato fluorescence (Figures S2E and  S2F). This implies that the stellate cell which we focused on in this study was not affected by BM-derived cell in our tracing mice. Therefore, by vitamin A administration for 20 days, we developed a novel fate-tracing method for stellate cells in the Lrat Cre; Rosa26-tdTomato mice.

Stellate cells are vital contributors to the myofibroblast pool in cerulein-induced pancreatic exocrine fibrosis
To determine the contribution of stellate cells to cerulein-induced pancreatic exocrine fibrosis, we treated 8-week-old Lrat-cre; R26-tdTomato mice with cerulein for 6 weeks and collected pancreatic tissues at 1 week after injury (Figure 2A). We performed Masson staining and found severe fibrosis in the exocrine tissues of the pancreas (Figures 2B and 2C). Immunostaining for tdTomato and Col I on pancreatic sections revealed that 65.7% of collagen-producing cells were tdTomato-positive PSCs ( Figures 2D and 2E). Immunostaining for tdTomato and a SMA revealed that tdTomato+ cells express a SMA ( Figure 2F). Together, these data indicate that PSCs constitute a major myofibroblast population in cerulein-induced pancreatic exocrine fibrosis.

stellate cells in islets is increased in high-dose streptozocin-induced T1DM mice
To determine the contribution of stellate cells to pancreatic endocrine islet fibrosis, we treated 8-week-old Lrat-Cre; R26-tdTomato mice with a single high-dose injection of beta-cell specific toxin streptozocin (STZ), followed by analyses at 1 or 2 weeks after injury ( Figure 3A). Severe elevation of blood glucose indicated a successful construction of T1DM model ( Figure 3B). Next, we performed Masson staining of pancreatic tissue sections and found that islet structural destruction and fibrosis at these two time points (Figures 3C and  3D). Immunohistochemical staining also showed the increased expression of Col1a1, Fn and a-SMA in the Together, these data indicate that stellate cells in islets are increased in high-dose streptozocin-induced islet acute injury and fibrosis.

The number of stellate cells in islets is increased in HFD and low-dose streptozocin-induced T2DM mice
After having established the increase of stellate cells in streptozocin-induced islet acute injury and fibrosis, we next determined the contribution of stellate cells to islet chronic injury and fibrosis. We treated 8-weekold Lrat-Cre; R26-tdTomato mice with HFD and low-dose streptozocin, followed by analyses at 1, 2, and 3 months after injury ( Figure 4A). Ten-week HFD-feeding and low dose streptozocin made the mice glucose intolerant ( Figure 4B) and the blood glucose levels before sacrificing are shown in Figure 4C. Next, we stained pancreatic tissue sections for Masson and observed islet structural destruction and fibrosis at these three time points (Figures 4D and 4E). Immunostaining for tdTomato and insulin ( Figures 4F and 4G) or Glucagon ( Figures S4A and S4B) revealed that the numbers of stellate cells in diabetic islets were increased. Similar to high-dose streptozocin-induced islet acute injury and fibrosis, these data indicate that stellate cells are increased in the islets of T2DM mice induced by HFD-Fed and low-dose streptozocin.

Stellate cells in the islets contribute to the myofibroblast pool in both T1DM and T2DM
After having established the increase of stellate cells in the islets, we next determined the contribution of stellate cells to the myofibroblast pool during islet injury and fibrosis in type 1 or 2 diabetic mice. Immunostaining for tdTomato and Col I revealed that most of the stellate cells expressed Col I ( Figure 5A). However, less than 5% of collagen-producing cells were tdTomato-positive stellate cells ( Figure 5B). Immunostaining revealed that stellate cells express a SMA ( Figure 5C). Together, these data indicate that stellate cells in the islets constitute a small myofibroblast population in type 1 or 2 diabetic islet fibrosis.

Pancreatic stellate cells are genetically ablated when mice are injected with vitamin A palmitate
To further substantiate the functional contribution of stellate cells to fibrogenesis in pancreatic exocrine and endocrine tissues, we ablated stellate cells via Lrat Cre-induced diphtheria toxin (DTA) in stellate cells ( Figure 6A). We found most of hepatic stellate cells were ablated but pancreatic stellate cells were reserved in Lrat Cre; Rosa26-DTA mice ( Figure 6B). Then we attempted to eliminate pancreatic stellate cells by intraperitoneal injection of vitamin A (induce the expression of cre-recombinase in pancreatic stellate cells). In addition, results show that vitamin A palmitate injection (2500 IU/time, 3 times in 3 days intervals) strongly reduced stellate cell number determined by immunostaining of stellate cell markers Desmin and Crbp1 in the pancreas of mice treated with vitamin A palmitate ( Figures 6C and 6D). Therefore, by vitamin A injection, we developed a pancreatic stellate cells genetically ablated method in the Lrat Cre; Rosa26-DTA mice.

Stellate cells genetic ablation improves pancreatic exocrine fibrosis but not islet fibrosis
We treated stellate cells ablation mice with cerulein for 3 consecutive days (50 mg/kg, six times per day), and collected pancreatic tissues on day 6 after injury ( Figure 7A). Masson staining showed that the area percentage of blue-labeled collagen tissue significantly decreased in stellate cell ablated cerulein-treated mice ( Figures 7B and 7C). Western blotting showed that stellate cells ablation decreased cerulein-induced up-expression of Col 1, Fibronectin and a-SMA proteins in pancreatitis (( Figures 7D and 7E). We also treated stellate cells ablation mice with a single STZ injection after 3 times of vitamin A palmitate administration and then analyzed the pancreas samples at 7 days after injection ( Figure 7F). But no significantly decreased of the area percentage of blue-labeled collagen tissue in islet of ablated mice treated with STZ as determined by Masson staining (Figures 7G and 7H). Western blotting also showed that no significantly decreased of Col 1, Fibronectin and a-SMA proteins in STZ treated PSC ablation mice compared to WT-DM iScience Article mice ( Figures 7I and 7J). Together, these data indicate that stellate cells ablation improve pancreatic exocrine fibrosis but not islet fibrosis.

DISCUSSION
In this study, we developed a novel fate-tracing strategy for stellate cells by vitamin A administration. Using chemicals-induced pancreatic exocrine or islet fibrosis models, we found that stellate cells contributed crucially or partly to myofibroblast pools in exocrine or islet fibrosis. Furthermore, stellate cells' genetic ablation improved pancreatic exocrine fibrosis. These data suggest that stellate cells are vital/partial contributors to pancreatic exocrine/islet fibrosis. Our findings provide solid in vivo evidence that stellate cells as important antifibrotic target cells in pancreatic fibrosis.
Because of the lack of specific promoters, there has been no breakthrough in genetic tracing techniques for PSCs in vivo in the past decade. The stellate cells can be detected using different markers: those with ectoderm origin [e.g., GFAP, nestin]; mesoderm origin [desmin, a-SMA]; and vitamin A related metabolic markers [e.g., LRAT, Crbp1]. 21,22 However, previous studies have shown that the ectoderm origin promoters, like GFAP, do not efficiently mark HSCs, 4 and that the mesoderm origin promoters are also expressed in other mesenchymal cells (like pericytes). 23,24 Therefore, in this study, we tried to put Cre under the mice Lrat promoter because this strategy has been successfully applied to HSCs lineage tracing. 4 Surprisingly, only a small fraction of PSCs was labeled with tdTomato by this strategy in physiological conditions. This may be because of the fact that PSCs rarely store vitamin A in normal dietary mice. 20 Therefore, we tried to improve the labeling efficiency based on vitamin A anabolic function of LRAT protein. 18,19 We up-regulated the transcription of Lrat gene, thereby increasing Cre expression by Vitamin A loading and the results were consistent with our expectations.
PSCs are generally recognized as the pancreatic counterpart cells of HSCs. Despite the lack of solid in vivo evidence that PSCs are the primary drivers of pancreatic exocrine fibrosis, much of the current research focuses on this cell type. Our results are in line with previous studies suggesting that targeted interventions on PSCs can improve pancreatic exocrine fibrosis. [25][26][27][28] Recently, Gen Yamamoto found that only a minority (21.8%) of collagen-producing cells were GFAP-positive. Then he concluded that myofibroblasts in the pancreas might be derived not only from PSCs but also from other fibrogenic cells. 20 However, this may be because of the fact that GFAP expression is nonexistent or significantly reduced in activated PSCs. 29,30 Our data also suggests that a subset of myofibroblasts is not derived from PSCs. As reported in the previous study, bone marrow-derived fibrocytes may contribute to pancreatic exocrine fibrosis. 31,32 In addition, the pericytes, abundant but long-neglected cells in the pancreas, similar to their role in fibrotic diseases in other tissues, may also differentiate into myofibroblasts. [33][34][35] In recent years, it has been found that endothelium can transdifferentiate into myofibroblasts through endothelial-to-mesenchymal transition and participate in organ fibrosis. 36,37 Therefore, more research is needed to identify those possible cell types in the future, such as cell lineage tracing and bone marrow transplantation experiment.
For endocrine islets, the primary cells responsible for fibrosis are unclear, but previous studies have suggested multiple cell types implicated in islet fibrogenesis. By genetic tracing, Mateus Gonç alves L found that islet pericytes contribute a substantial fraction (approximately 40%) of islet myofibroblasts in AktTg mice (a mouse model of islet vascular fibrosis). 6 Melvin R Hayden found that pericyte can differentiate into an islet pancreatic stellate-myofibroblast-like cell capable of synthesizing fibrillar-banded collagen in the islet-exocrine interface. 38,39 Advanced glycation end products have also been shown to induce islet endothelial cells endothelial-to-mesenchymal transition in vitro, which may contribute to islet fibrosis in diabetes. 40 We previously isolated stellate cells from islets and found that they were involved in the process of islet fibrosis. 41,42 Consistent with a previous study. 12 our data suggest that stellate cells are increased in islets of type 1 or 2 diabetes and contribute to a small subset of myofibroblasts in the process of islet fibrosis. iScience Article Taken together, our results show that stellate cells are mainly distributed in the pancreatic exocrine while a small portion was observed in the endocrine islets. Stellate cells are vital contributors to the myofibroblast pool in cerulein-induced pancreatic exocrine fibrosis, and genetic ablation of stellate cells can improve pancreatic exocrine fibrosis. Besides, during the process of islet fibrosis induced by streptozocin, stellate cells in the islets are increased and partly contribute to the myofibroblast pool. So, we conclude that stellate cells are vital/partial contributors to myofibroblasts in pancreatic exocrine/islet fibrosis. Our study provides the in vivo lineage tracing evidence for the cellular target in pancreatic fibrosis.

Limitations of the study
Our study has several limitations. First, the loads of vitamin A may affect the activation of PSCs. Several studies have shown that vitamin A inhibits the activation of PSCs or HSCs. [43][44][45] However, a study also showed that the absence of retinoid-containing lipid droplets (Lrat KO mice) does not promote HSCs activation. 46 In this study, after a 10-day gavage of vitamin A, the labeling efficiency of PSCs did not significantly improve. This may be because of the fact that vitamin A is primarily stored in HSCs but not PSCs. 47 In addition, the labeling efficiency rapidly increased from 15-to 18-day and maintained after 20-day gavage of vitamin A. Furthermore, the undetectable vitamin A autofluorescence in PSCs of the tracing mice implied a low intracellular amount of vitamin A, which is similar to the negative results in normal dietary mice in a iScience Article previous study. 20 So, we speculate that the PSCs in this strategy may not be very different from physiological conditions. Second, because of lifetime limitation, 2-4 weeks ablated mice were used to substantiate the functional contribution of PSCs to fibrogenesis. This is not representative of the adult stage, and superior ablation mouse models are needed in further studies. Third, we only adopted chemicals-induced fibrosis models, which may not be applicable to other pathological conditions. Fourth, as reported in previous studies, bone marrow derived cells contribute to pancreatic fibrosis. In addition, the role of BM cells in pancreatic fibrosis will be more fully verified if further bone marrow transplantation experiments are performed. Besides, pericytes and endothelium may also be involved in pancreatic fibrosis. So related cell tracing and ablation experiments can more completely and clearly show the cellular mechanism of pancreatic fibrosis, which is our follow-up research direction of pancreatic fibrosis research. Finally, our results provide solid in vivo evidence that stellate cells as important antifibrotic target cells in pancreatic fibrosis, but PSC ablation is unlikely to be a clinical application directly because of the other physiological functions of PSC (such as vitamin A storage and metabolism). So, compare to PSC ablation, future development of new drugs or inhibitors targeting PSC may be more beneficial to the clinical prevention and treatment of pancreatic fibrosis. In this study, we also mainly focused on fibrosis itself, and further functional tests will

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

AUTHOR CONTRIBUTIONS
J.W., T.L., and Y.Z. designed the study, performed experiments and analyzed the data. X.W. and C.N. bred the mice, performed experiments or provided valuable comments. V.C. and W.L. contributed to interpreting the data and editing the manuscript. Q.W., Z.S., and Y.C. performed immunohistochemistry and data analysis. S.Q. and Z.S. supervised the study, analyzed the data and wrote the manuscript.

DECLARATION OF INTERESTS
The authors declare no competing interests.