1. Introduction
Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal cancers due to its devastating metastatic nature. It is characterized by remarkably dense and firm desmoplasia composed of cancer-associated fibroblasts (CAFs), extracellular matrix (ECM), leukocytes, and endothelial cells [1]. CAFs are the major contributors to desmoplasia, producing ECM and soluble factors that drive tumor progression [1–3].Angiogenesis is indispensable for tumor growth and metastasis. Although PDAC is rather hypovascular, pro-angiogenic factors, such as VEGF and CXCL12, are overexpressed in PDAC. Further, blood vessel density is positively correlated with PDAC progression [4]. Previous studies have demonstrated that CAFs are involved in angiogenesis by releasing pro-angiogenic factors [5,6]. Genetic fate mapping revealed that fibroblasts in a murine myocardial infarction model were reprogrammed into endothelial cells, a process known as the mesenchymal-to-endothelial transition (MEndoT) [7,8]. Brumm et al.reported that astrocytes could also undergo MEndoT under serum deprivation in vitro [8]. Through lineage tracing and single-cell transcriptome sequencing, Menget al. suggested that MEndoT contributes to peripheral circulation perfusion recovery in a mouse hindlimb ischemia model [9]. However,whether PDAC CAFs undergo a similar phenotype transition and directly contribute to tumor angiogenesis remains unclear.Dense and firm desmoplasia creates a hypoxic, ischemic, and acidic tumor microenvironment (TME), compromising protein homeostasis and inducing endoplasmic reticulum (ER) stress within tumors [10]. Cell differentiation and transdifferentiation involve considerable protein synthesis, exerting enormous stress on the ER [11]. Consequently, cells initiate the unfolded protein response (UPR) in an attempt to maintain protein homeostasis. ER stress and UPR activation have been described in PDAC and contribute to tumorigenesis, angiogenesis, progression, cell transdifferentiation, metastasis, and treatment resistance [9–11]. However, the role of MEndoT in PDAC and its association with UPR remain unclear. Herein, we provide evidence that CAFs can act as a source of endothelial cells in PDAC by undergoing MEndoT regulated via PERK-eIF2α-ERK1/2 signaling.
Fig. 1. CAFs undergo endothelial transdifferentiation in vitro A, Representative images show capillary-like structures of CAFs under 10% and 0% serum on Matrigel at 6–8 h. Green, Calcein AM staining; Scale bars, 200 μm. B and C, Quantification of tube junctions (B) and vessel length (C). D, Representative images show DiI-acLDL uptake by CAFs under 10% and 0% serum for 48 h. Red, intracellular DiI-labeled acLDL; Blue, nuclear staining. Scale bars, 100 μm. E, Quantification of Dil-acLDL uptake. F and G, qPCR and western blotting of endothelial markers in CAFs under 10% serum and 0% serum for 48 h. H and I, qPCR and western blotting of mesenchymal markers in CAFs under 10% and 0% serum for 48 h. β-actin and β-tubulin were used as endogenous controls. J, Fluorescence micrographs show endothelial marker expression in CAFs cultured in 10% and 0% serum for 48 h. Red,VECAD, CD31, and Endoglin; green, α-SMA; blue, nuclear staining. Scale bars, 20 μm. K, Quantification of CAFs (%) expressing endothelial markers. Data arepresented as the mean ± s.e.m; *P < 0.05, **P < 0.01, and ***P < 0.001. . (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
2. Materials and methods
2.1. Cell culture
With approval from the Ethics Committee, all human PDAC samples were provided by the pancreatic cancer department of Tianjin Medical University Cancer Institute & Hospital and collected from the donors with informed consent. The clinical features of patients whose tumors were used for CAFs isolation were provided in Table S1. We isolated human pancreatic CAFs from fresh PDAC surgery specimens using a culture outgrowth method [12,13].In brief, fresh human PDAC surgical samples were cut into 1-mm3 blocks with a sharp blade. The blocks were then seeded in 6-cm culture dishes cultured with DMEM medium containing 10% fetal bovine serum (Fig. S1A). After 7– 15 days, when cell confluence reached 90%, CAFs were trypsinized and replanted into another culture plate and cultured with total DMEM medium (Fig. S1A). We used immunofluorescence and FCM to identify the isolated cells. They were positive for the following mesenchymal-specific markers: Desmin, Collagen I and α-SMA (Fig. S1B), and negative for other cell linage markers: CD326 for epithelial cells, CD31 for endothelial cells and CD45 for immune cells (Fig. S1C). The first to fifth passages of primary CAFs were used in our experiments. To label CAFs, CAFs were transduced with lentiviral-GFP (SyngenTech, Beijing, China). The human PDAC cell line L3.7 was a gift from Prof. Keping Xie (MD Anderson Cancer Center, Houston, TX) and cultured in RPMI-1640 with 10% FBS.
Fig. 2. CAFs acquire endothelial features in vivo A, L3.7 cells alone or mixed with GFP-labeled CAFs were subcutaneously injected into both flanks of BALB/cnude mice (n = 6 per group). B, Tumor volume images. C, Tumor growth curves. Repeated measures two-way ANOVA (time × tumor volume) and post hoc analysis were used to compare mouse tumor growth between groups. D,
Representative images of IHC for CD31 in mouse tumor tissues. Scale bars, 100 μm. E, Quantification of blood vessel counts per field. F, Representative images of mIHC (upper panel) and HE staining (lower panel) in L3.7+CAF-GFP tumors. White arrowheads indicate GFP+CD31+ cells; black arrowheads indicate erythrocytes; red, CD31; green, GFP; purple, CK19; blue, nuclear staining; Scale bars, 20 μm. Data arepresented as the mean ± s.e.m; **P < 0.01 and ***P < 0.001. . (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
2.2. Animal experiments
Before initiation, all proposed animal experiments were approved by the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital and were performed in accordance with NIH guidelines. 1 × 106 L3.7 alone or with an equal number of CAF-GFP were injected subcutaneously into two flanks of female 5-week-old BALB/cnude mice (SiPeiFu, Beijin, China) (n = 6 per group). Tumor volume was measured weekly. For treatment experiments, GSK2606414 (GSK, 50 mg/kg) [14] or corn oil (vehicle) was administered orally twice a day, gemcitabine (Gem, 15 mg/kg) [4] or saline solution (vehicle) was intraperitoneal injected into the corresponding mice twice a week for 4 weeks.
2.3. Tube formation assay
This assay was performed as previously reported [15]. For each assay, 6 × 104 cells/well were plated on Matrigel-coated 48-well cell culture plate and cultured for 6–8 h in DMEM without or with 10% serum. Tunicamycin (2.5 μg/mL), tauroursodeoxycholic acid (TUDCA, 4 mg/mL), GSK2606414 (GSK, 0.5 μM), CCT020312 (CCT, 0.5 μM), AEBSF (0.3 mM), STF-083010 (STF, 40 μM), and U0126 (10 μM) were added to the cells cultured in the above culture medium. Once tube formation was observed, cells were labeled using a live-cell labeling dye-Calcein AM (2 μM) and photographed using a Leica microscope.Images were analyzed using AngioTool 0.6.
2.4. Acetylated low-density lipoprotein (acLDL) uptake assay
CAFs were cultured in 10% or 0% serum with PBS, Tunicamycin, and TUDCA. After 48 h, DiI-labeled acLDL (25 μg/mL,Yeason, 20606ES76) was added, and 4 h later, cells were subjected to nuclear staining using Hoechst 33342 and analyzed with a Leica microscope.
2.5. Flow cytometry(FCM) and apoptosis assays
For FCM analysis, primary CAFs were stained with anti-CD31-PE (BioLegend, 303106), anti-CD326-FITC (BioLegend, 324203) and antiCD45-PE/Cy5 (BioLegend, 304010). Isotype controls were used as negative controls. Apoptosis was determined using Annexin-V Apoptosis Detection Kit (BioLegend, 640932) according to the manufacturer ’s protocol. Samples were assayed by BD FACSCanto™ II (BD Biosciences). Data were analyzed using FlowJo 10.4.
2.6. Cell viability assay
Cell viability were assessed using CCK8 (Bimake, B34302). CAFs were seeded in 96-well plates and cultured in DMEM with or without serum. At the specified time point, 10% CCK8 solution was added to each well, and then the cells incubated in darkness for 2h. The absorbance was measured at 450 nm using absorbance microplate reader.
2.7. Immunofluorescence, immunohistochemistry (IHC), and multiplex fluorescent IHC (mIHC) For immunofluorescence, primary CAFs were seeded onto poly-L-lysine-coated glass slides for different treatments. Paraffin sections were used for IHC and mIHC analysis. The following primary antibodies were added for immunostaining: anti-VECAD (1:400, Abcam, 33168), CD31 (1:50, Abcam, 28364), C-SMA (1:100, Sigma, A5228), Endoglin (1:100, Abcam, 69772), Collagen I (1:200, Abcam, 260043), Desmin (1:400, Abcam, 32362), BiP (1:200, Abcam, 21685), CK19 (1:100, Abcam, 52625) and GFP (1:200, Abcam, 183734). DAB substrate kit (ZSGB-BIO) was used for IHC analysis. OpalTM 4-color manual IHCkit (PerkinElmer, NEL810001KT) was used for mIHC analysis. Slides were viewed with Zeiss microscopy.
Fig. 3. Serum deprivation activates UPR signaling in CAFs.A, Representative images of BiP IHC staining in a normal human pancreas and PDAC. Scale bars, 50 μm. B, Representative electron microscopy images of ER morphological changes in CAFs cultured in 10% and 0% serum for 48 h. Tunicamycin (ER stress agonist, 2.5 μg/mL) was used as a positive control. Scale bars, 2 μm. C and D, qPCR and western blotting of BiP were performed in CAFs cultured in 10% and 0% serum for 48 h. E-G, Western blotting of UPR markers. H, XBP1s/XBP1u (spliced XBP1/unspliced XBP1) ratio of CAFs cultured in 10% and 0% serum. Data are presented as the mean ± s.e.m; *P < 0.05 and ***P < 0.001.
2.8. Western blot analysis
Target proteins were detected by Western blot with primary antibodies as follows: C-SMA (1:1000, Sigma, A5228), VECAD (1:1000, Abcam, ab33168), CD31 (1:500, Abcam, 28364), Desmin (1:5000, Abcam, ab32362), Collagen I (1:1000, Sigma, c2456), BiP (1:1000, Abcam, 21685), PERK (1:1000, CST, 5683), phospho-PERK (1:1000, CST, 3197), eIF2C (1:1000, CST, 5324), phospho-eIF2C (1:1000, CST, 3398), ATF6 (1:1000, Abcam, 227830), IRE1C (1:1000, Abcam, 37073), phospho-IRE1C (1:1000, Abcam, 124945),ERK1/2 (1:1000,CST, 9102) and phospho-ERK1/2 (1:1000, CST, 3398).
2.9. RNA isolation and quantitative PCR (qPCR)
The total RNA of the cells was isolated from CAFs with TRizol reagent (Invitrogen) and reverse transcription was performed using PrimeScript似 RT Master Mix (Takara). Then, qPCR was conducted to analyze the cDNA levels. ACTB was used as a loading control. Each experiment was repeated at least three times independently. qPCR primers used are listed in Supplementary Table S2.
2.10. Statistical analysis
Statistical analyses were performed using GraphPad Prism 9.0.0 software. Each experiment was performed in triplicate, and the values are presented as the mean value ± standard error of the mean (s.e.m.) unless otherwise stated. The student,s t-test was used for comparison between the two groups. One-way ANOVA with Bonferroni post-test analysis was used for multiple comparisons. All probability values had a statistical power level of 90% and a two-sided level of 5%. P < 0.05 was considered significant.
3. Results
3.1. CAFs transdifferentiate into the endothelial lineage in vitro
MEndoT induction was reported in fibroblasts of the heart, brain, and hindlimb after serum deprivation in vitro [7一9]. Thus, we seeded CAFs on Matrigel, subjecting them to different serum concentrations [15]. After 6一8 h of seeding, in contrast to 10% serum, low-concentration serum (≤2%) induced the formation of capillary-tube-like structures (Figs. S1D一F). CAFs exhibited pronounced tube formation, especially under 0% serum (Fig. 1A一C, S1D-F). We then subjected CAFs to serum starvation for 48h [8] and assessed acetylated LDL (acLDL) uptake,a characteristic of mature endothelial cells [16]. As shown in Fig. 1D一E, 23.45% ± 1.81% of starved CAFs acquired acLDL uptake ability. We also assessed cell viability after serum deprivation. Viability was not impaired within 12 h of serum deprivation, but was reduced after 24 h (Fig. S1G). As shown in Figs. S1H一I, apoptosis rate slightly increased after starvation for 12 h, remaining below 10%.To confirm whether starved CAFs underwent the MEndoT, we performed qPCR,western blotting, and cell immunofluorescence staining to assess the expression of endothelial-specific (VECAD, CD31, and Endoglin) and mesenchymal-specific (α-SMA, Desmin, and Collagen I) markers. Serum deprivation resulted significantly upregulated of endothelial-specific genes while downregulating mesenchymal-specific markers at the mRNA and protein levels (Fig. 1F–I). Starved CAFs exhibited intense fluorescence for VECAD, CD31, and Endoglin (Fig. 1J–K). These results indicated that the starved CAFs lost fibroblast features and acquired endothelial cell characteristics after serum deprivation in vitro.
Fig. 4. The UPR mediates CAF MEndoT.A, Representative images show capillary-like structures of CAFs subjected to 10% serum or 0% serum with PBS, TUDCA (ER stress inhibitor, 4 mg/ml), and tunicamycin (ER stress agonist, 2.5 μg/ml) on Matrigel at 6–8 h. Green, Calcein AM staining; Scale bars, 200 μm. B-C, Quantification of tube junctions (B) and vessel length (C). D, Representative images show DiI-acLDL uptake of CAFs cultured in 10% or 0% serum with PBS, TUDCA, and tunicamycin for 48 h. Red, Dil-labeled acLDL; Blue, nuclear staining; Scale bar, 100 μm. E, Quantification of acLDL uptake. F and G, qPCR and western blotting of BiP, endothelial, and mesenchymal marker expression in CAFs under 10% or 0% serum with PBS, TUDCA, and tunicamycin for 48 h. Data arepresented as the mean ± s.e.m; *P < 0.05, **P < 0.01, ***P < 0.001 versus 10% serum; †P < 0.05, ††P < 0.01 versus 0% serum. . (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
3.2. CAFs adopt endothelial cell-like features in vivo
To determine whether CAFs acquire endothelial features and directly contribute to tumor angiogenesis in vivo, we established a xenograft model wherein L3.7 tumor cells mixed with the same number of GFPlabeled CAFs were subcutaneously injected into the left flanks of mice. L3.7 tumor cells were into the right flank of mice as a control (Fig. 2A). The average tumor volume was significantly greater in the mixed-cell group relative to L3.7 alone (Fig. 2B–C). Mixing with CAF-GFP resulted in a substantial increase in bloodvessel density of xenograft tumors (Fig. 2D–E). Our results were consistent with previous studies [5,17], suggesting that CAFs promote tumor growth and angiogenesis.Next, we detected GFP-labeled CAFs selleckchem expressing endothelial marker CD31 in vivo via mIHC. As anticipated, 25.33% ± 2.03% of GFP-labeled CAFs expressed CD31. Furthermore, the vessel lumen was lined with GFP− CD31+ endothelial cells (Fig. 2F, upper panel). To confirm that CD31+GFP+ CAF-containing vessels were functional, we performed HE staining on serial sections and found that erythrocytes were also present in the vessels (Fig. 2F, lower panel). eggshell microbiota Taken together, CAFs acquired features of endothelial cells and directly contributed to angiogenesis within the TME.
Fig. 5. Activation of PERK-eIF2“-ERK1/2 signaling induces MEndoT in CAFs.A, Representative images show capillary-like structures of CAFs subjected to 10% serum or 0% serum with DMSO, GSK (PERK inhibitor, 0.5 mM), and CCT (PERK agonist, 0.5 nM) for 6–8 h. Green, Calcein AM staining; Scale bars, 200 μm. B and C, Quantification of tube junctions (B) and vessel length (C). D, Western blotting of PERK branch, endothelial, and mesenchymal marker expression in CAFs cultured under 10% or 0% serum with DMSO, GSK (0.5 mM), or CCT (0.5 mM) for 48 h. E, Western blotting of ERK pathway members in CAFs cultured under 10% or 0% serum for 48 h. F, Western blotting of ERK pathway members in CAFs cultured in 10% or 0% serum with DMSO, GSK (0.5 mM), or CCT(0.5 mM) for 48 h. G, Representative images show capillary-like structures of CAFs subjected to 10% or 0% serum with DMSO, U0126 (ERK1/2 inhibitor, 10 μM) for 6–8 h. Green, Calcein AM staining; Scale bars, 200 μm. H and I, Quantification of tube junctions (H) and vessel length (I). J, Western blotting of ERK pathway, endothelial, and mesenchymal expression in CAFs under 10% or 0% serum with DMSO, U0126. GSK, GSK2606414; CCT, CCT020312. Data are presented as the mean ± s.e.m; **P < 0.01, ***P < 0.001 versus 10% serum; †††P < 0.001 versus 0% serum. . (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
3.3. Serum deprivation triggers ER stress in CAFs
Hypovascularity is common in PDAC and unfavorable for protein folding within the ER, thus causing ER stress [10,11,18]. In response, cells evoke the UPR to reestablish ER homeostasis. To assess whether these processes are involved in CAF endothelial transdifferentiation, we determined the expression of BiP, a global ER stress marker [10], in PDAC CAFs via IHC. ER stress was activated in CAFs but not in normal fibroblasts (Fig. 3A). We used transmission electron microscopy to observe ER morphological changes under serum deprivation. ER stress inducer tunicamycin was used as a positive control. The ER of CAFs was enlarged and dilated under serum-deprived conditions (Fig. 3B).Moreover, BiP expression was upregulated in starved CAFs at both the mRNA and protein levels (Fig. 3C–D).
ER stress activates the UPR, which is mediated by three signaling pathways (ATF6, PERK, and IRE1α) [11,19,20]. Thus, we further examined changes in the expression of pathway-associated factors upon serum deprivation. Under ER stress, full-length ATF6 (90 kDa) is cleaved to its active 50 kDa form [19]. Our data showed that generation of cleaved ATF6 was increased and was associated with lower levels of full-length ATF6 under the 0% serum condition (Fig. 3E). PERK activation depends on the phosphorylation of PERK and its downstream molecule, eIF2α [10]. The induction of PERK and eIF2α phosphorylation was observed in starved CAFs (Fig. 3F). Under ER stress, IRE1α undergoes autophosphorylation and selectively cleaves XBP1 mRNA to spliced XBP1 (XBP1s), an active transcription factor [10]. Phospho-IRE1α and XBP1s levels increased under serum deprivation, consistent with the decrease in unspliced XBP1 (XBP1u) (Fig. 3G). The XBP1s/XBP1u ratio increased approximately 2.18-fold in starved CAFs (Fig. 3H). These data indicated that the IRE1α branch was activated. Collectively, serum deprivation simultaneously triggered all three UPR signaling branches to varying degrees in vitro.
Fig. 6. GSK inhibits MEndoT and enhances gemcitabine efficacy in vivo.A, BALB/c nude mice were subcutaneously inoculated with L3.7 and CAF-GFP cells. Mice were intraperitoneally injected with Gem (15 mg/kg) or saline (vehicle) twice a week. GSK (50 mg/kg) or corn oil (vehicle) was administered orally twice per day (n = 6 per group). B, Images of tumors after sacrifice. C, Tumor growth curves. Repeated measures two-way ANOVA (time × tumor volume) and post hoc analysis were used to compare tumor growth between groups. D, Representative image of IHC staining for CD31 in mouse tumor tissues. Scale bars, 100 μm. E, Quantification of blood vessel counts IHC images (100 × ). F, Representative mIHC images of mouse tumors. White arrowheads indicate GFP+CD31+ cells. Red, CD31; green, GFP; purple, CK19; blue, nuclear staining. Scale bars, 20 μm. G, Quantification of CD31-expressing GFP+ CAFs. Gem, gemcitabine; GSK, GSK2606414. Data arepresented as the mean ± s.e.m; *P < 0.05, **P < 0.01, and ***P < 0.001; n. s., no significance. . (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
3.4. UPR activation is essential for MEndoT induction in CAFs
Differentiation involves significant ER expansion to support elevated protein production, activating the UPR [11]. Using gain and loss-of-function analysis, we determined whether the UPR affects serum deprivation-induced MEndoT. Serum-deprived CAFs formed tubes on Matrigel,but the addition of UPR inhibitor TUDCA [21] suppressed tube formation (Fig. 4A–C) as well as acLDL uptake (Fig. 4D–E). qPCR and western blotting revealed a consistent decrease in the expression of endothelial-specific markers (Fig. 4F–G). We subsequently investigated whether UPR activation alone could induce MEndoT. We treated CAFs with UPR agonist tunicamycin [22] under serum-fed conditions. As expected, tube formation and acLDL uptake were induced (Fig. .4A-E) along with a dramatic increase endothelial-specific marker expression (Fig. 4F–G), suggesting that UPR activation was the control switch of CAF MEndoT.
3.5. The PERK-eIF2“-ERK1/2 signaling branch regulates CAF MEndoT
Serum deprivation triggered three UPR signaling pathways (Figs. 3 and 4). We used specific inhibitors, including GSK2606414 [14] (GSK, PERK inhibitor), AEBSF [23] (ATF6 inhibitor), and STF-083010 [24] (IRE1α inhibitor), to determine which UPR pathway mainly contributions to the MendoT. The inhibition efficiency of each agent was confirmed via western blotting (Fig. 5D, Figs. S2A–B). Tube formation assays revealed that only GSK significantly inhibited tube-like structure formation by starved CAFs (Fig. 5A–C, Figs. S2C–E). Conversely, PERK activator
CCT020312 [25] (CCT) induced tube formation by CAFs under 10% serum conditions (Fig. 5A-C), suppressing mesenchymal-specific and upregulating endothelial-specific markers (Fig. 5D). These results suggested that MEndoT was mainly mediated via the PERK branch of UPR signaling.
Fig. 7. The mechanism of MEndoT in CAFs.
The hypovascular TME evokes ER stress in hepatitis and other GI infections CAFs and drives MEndoT via PERK-eIF2α-ERK1/2. GSK2606414 inhibits MEndoT by suppressing PERK phosphorylation levels.The ERK pathway is involved in the differentiation of endothelial cells from pluripotent stem cells [26]. ERK signaling was reported to be activated through PERK-eIF2α [27]. Therefore, we explored whether serum deprivation induces MEndoT through ERK activation via PERK-eIF2α . ERK signaling was activated by serum deprivation (Fig. 5E). PERK inhibitor GSK significantly suppressed ERK1/2 phosphorylation (Fig. 5F). Conversely, PERK activator CCT upregulated ERK1/2 activation (Fig. 5F). ERK1/2 inhibitor U0126 [28] suppressed tube formation by CAFs (Fig. 5G, H, and I) and blocked serum deprivation-induced changes in MEndoT marker expression (Fig. 5J). Altogether, our data suggested that CAFs undergo MEndoT through the PERK-eIF2α-ERK1/2 axis.
3.6. PERK inhibition suppresses tumor MEndoT and enhances the chemotherapeutic efficacy of gemcitabine
We further investigated the therapeutic potential of MendoT suppression via PERK inhibitor administration in a mouse xenograft model (Fig. 6A). The L3.7 cell line and CAF-GFP were mixed equally and subcutaneously injected into the flanks of mice. Compared to the vehicle group, GSK2606414 (GSK) moderately suppressed tumor growth (Fig. 6B-C) and downregulated angiogenesis (Fig. 6D-E). Furthermore, the MEndoT of CAFs was significantly inhibited (Fig. 6F-G). Given the major role of tumor angiogenesis in chemotherapy resistance, we combined GSK with first-line chemotherapeutic gemcitabine (Gem) to assess whether MendoT-targeted therapy could enhance Gem efficacy. The combination of GSK and Gem significantly suppressed tumor volume, angiogenesis, and the MendoT of CAFs (Fig. 6). Collectively, our data suggested that targeting MEndoT via PERK inhibition represents a strategy for suppressing tumor angiogenesis and enhancing chemotherapeutic efficacy (see Fig. 7).
4. Discussion
Mounting evidence highlights CAFs as a heterogeneous and complex cell population [29]. In parallel to cancer progression, CAFs take on diverse functions that can support or suppress tumor growth [30]. Thus, a better understanding of their complex nature could help tailor therapy and improve patient prognosis [31]. The functional heterogeneity of CAFs maybe ascribed to their diverse cellular sources [32]. Studies have shown that endothelial cells can alter their morphological and functional characteristics within the TME in order to obtain CAF properties in a process called the endothelial-mesenchymal transition (EndMT) [33,34]. Approximately 40% of CAFs may originate from the EndMT process [35,36].Fibroblasts are recognized as a highly plastic cell type [3,37]. They can transdifferentiate into other fibroblast subtypes or even into entirely different lineages when exposed to certain external stimuli, such as tumor cues, culture conditions, and therapeutics [29]. In the heart and brain, fibroblasts exhibit native cellular plasticity that enables them to adopt endothelial cell-like features after ischemic injury [7,8]. Biffiet al. observed CAF subtype reprogramming in vivo following administration of JAK inhibitor AZD1480,which shifted α-SMA-low inflammatory CAFs to α-SMA-high myofibroblastic CAFs in PDAC mice [38]. Herein, we demonstrated that CAFs could undergo a conversion from mesenchymal to endothelial cells (MEndoT) both in vitro and in vivo. Primary CAFs were isolated from fresh PDAC samples and subjected to serum deprivation conditions. Robust capillary-like structures were formed by serum-deprived CAFs on Matrigel. After starvation for 48 h in 2D culture, CAFs exhibited endothelial marker upregulation at both the mRNA and protein level. Further, 23.45 ± 1.81% of starved CAFs took up acLDL. Cells undergoing this transition had downregulated mesenchymal marker expression. In mouse xenograft models, 25.33% ± 2.03% of GFP-labeled CAFs acquired endothelial features and contributed to tumor angiogenesis.Furthermore, we found that the UPR was the key switch regulating this phenotype transition. Previous studies have reported that UPR activation is tightly associated with cell fate [11,16,34,35]. In the current work, serum deprivation triggered ER stress and activated all three branches of the UPR in CAFs.Further, the addition of UPR inhibitor TUDCA decreased tubular structure formation and endothelial marker expression. CAFs generated tubules even without serum deprivation when UPR agonist tunicamycin was administered, suggesting that UPR activation was required for MEndoT induction. Moreover, we used specific inhibitors of the three UPR branches and determined that MEndoT was mainly mediated via the PERK-eIF2α-ERK1/2 branch. PERK pharmacological inhibition suppressed the MEndoT and angiogenesis, improving chemotherapy efficacy in mouse xenograft models.Although PDAC is generally hypovascular, blood vessel density is positively correlated with PDAC progression [4]. Previous evidence has suggested that CAFs regulate tumor angiogenesis by producing crucial angiogenic factors, such as VEGF, CXCL12, and CXCL8 [6]. Our data provide evidence for the involvement of CAFs in tumor angiogenesis via their transdifferentiation into endothelial cells, highlighting the MEndoT as a therapeutic target.
