Autophagy loss impedes cancer-associated fibroblast activation via downregulating proline biosynthesis

ABSTRACT Cancer-associated fibroblasts (CAFs) are considered one of the most critical stromal cells that interact with pancreatic ductal adenocarcinoma (PDAC) and promote tumor growth, metastasis, and treatment resistance. Previous studies illustrated macroautophagy/autophagy contributes to CAF activation during tumor progression. Here in our study, we found that autophagy deficiency in CAFs impedes CAF activation by inhibiting proline biosynthesis and collagen production. Furthermore, we uncovered that autophagy promotes proline biosynthesis through mitophagy-mediated regulation of NADK2 (NAD kinase 2, mitochondrial), an enzyme responsible for production of mitochondrial NADP(H). Using an orthotopic mouse model of PDAC, we found that inhibiting mitophagy by targeting PRKN (parkin RBR E3 ubiquitin protein ligase) in the stroma reduced tumor weight. Thus, inhibition of CAFs mitophagy might be an attractive strategy for stroma-focused anti-cancer intervention. Abbreviations: ACTA2/α-SMA: actin alpha 2, smooth muscle, aorta; ACTB/β-actin: actin, beta; ALDH18A1/P5CS: aldehyde dehydrogenase 18 family, member A1; ATG3: autophagy related 3; ATG5: autophagy related 5; BNIP3L: BCL2/adenovirus E1B interacting protein 3-like; CAFs:cancer-associated fibroblasts; COL1A1: collagen, type I, alpha 1; DES: desmin; ECM: extracellular matrix; FABP4: fatty acid binding protein 4, adipocyte; FAP/FAPα: fibroblast activation protein; IHC: immunohistochemical staining; LAMP1: lysosomal-associated membrane protein 1; NADK2: NAD kinase 2, mitochondrial; PC1: pro-collagen 1; PDAC: pancreatic ductal adenocarcinoma; PDGFR: platelet derived growth factor receptor; PDPN: podoplanin; PRKN: parkin RBR E3 ubiquitin protein ligase; PSCs: pancreatic stellate cells; VIM: vimentin; WT: wild-type


Introduction
Pancreatic cancer (PC) is one of the most malignant tumors which is characterized by its early metastasis and poor prognosis with 5-year survival rate less than 9% [1]. The tumor microenvironment of pancreatic ductal adenocarcinoma (PDAC) is composed of extracellular matrix (ECM) proteins, tumor vasculature, fibroblasts and immune cells [2]. The microenvironment of tumor has long been considered as a contributor to malignant development [3]. Stromaassociated pancreatic stellate cells (PSCs) contribute in cancer progression as they are of great abundance in PDAC and can cause intense fibrotic stromal response [4][5][6]. Notably, there is a transition from quiescent PSCs to cancer-associated fibroblasts (CAFs), which is a hyper-activated myofibroblast-like state [7] that exhibits increased deposition of collagens [8] and secretes excessive amounts of ECM proteins, cytokines, chemokines, and growth factors to support tumor metabolism [2,9,10]. These activated CAFs are characterized by increased expression of some markers, including ACTA2/αSMA (actin alpha 2, smooth muscle, aorta), FAP/FAPα (fibroblast activation protein), and PDGFR (platelet derived growth factor receptor) [11].
Autophagy, a conserved process of lysosome-dependent protein degradation [12,13], has been reported to regulate the transition process from PSCs to CAFs, during which IL6 and ECM molecules are produced to promote tumor proliferation and metastasis [14]. Similarly, in stroma of squamous cell carcinoma, autophagy causes the downregulation of RBPJ protein which negative control the normal fibroblast conversing to CAFs [15]. Recently, a study using polyoma middle T antigen-driven (PyMT) mammary tumorigenesis model demonstrated that genetic ablation of stromal fibroblast autophagy significantly impedes the stromal desmoplastic response [16]. however, further exploration is still needed to elucidate the mechanism of CAF activation. Upon nutrition starvation, autophagy is required for maintenance of amino acid levels through constitutive turnover of cytopolasmic components [17]. Here in our study, we found that autophagy deficiency in CAFs disrupts collagen production by inhibiting proline biosynthesis. Furthermore, we uncovered that autophagy in CAFs promotes proline biosynthesis through mitophagy, which supports CAF activation and tumor growth in vivo.

Genetic inhibition of autophagy changes the phenotype of CAFs
It has been reported that autophagy is required for the transition process from PSCs to CAFs [14]. An important question is whether control of fibroblast autophagy and CAF activation can be genetically associated. To validate whether autophagy promotes CAF activation in vivo, we orthotopically transplanted 15,376 T (mouse PDAC, from C57Bl/6 J isogenic tetO_LKRAS G12D TRP53/p53 L/ + PTF1A/p48-Cre, termed 15,376 T) cells into C57Bl/6 J mice without (Atg5 +/+ ) or with (atg5 Δ/Δ ) conditional whole-body ATG5 deficiency. (Orthotopic 15,376 T tumors faithfully recapitulated histological and molecular features of the primary tumors, suggesting that the cultured tumor cell line may serve to complement tumor studies in vivo [18].) Consistent with previous reports [4], tumors were significantly smaller when grown in autophagy-deficient hosts compared to Atg5 +/+ hosts (Figure 1(a,b)), demonstrating that host autophagy promoted tumor growth. Next, we investigated histological sections of the resected orthotopic tumors. We found that the number of ACTA2/α-SMA-or COL1A1 (collagen, type I, alpha 1)-positive cells was significantly decreased by autophagy deficiency (Figure 1(c-f)). A recent study reported that pancreatic stellate cells (PSCs) show very high expression of the adipocyte marker FABP4 (fatty acid binding protein 4, adipocyte) [19]. By double immunofluorescent staining of FABP4 and ACTA2/α-SMA in wild-type (WT) mice and autophagy-deficient (Ubc-Cre-Esr/ERT2; atg5 fl/fl ) mice, we found that the decreased expression of α-SMA and induction of FABP4 in autophagy-deficient hosts (Figure 1(g)), indicating that the transition from quiescent PSCs to CAFs is inhibited in autophagy-deficient hosts. Consistently with the in vivo data, both the mRNA and protein levels of FABP4 were increased in both atg3-KO and atg5-KO CAFs (Figure 1 (h,i), Figure S1A). It has been reported that the level of collagen is decreased in quiescent PSCs compared with activated CAFs [20,21]. Therefore, we speculated that autophagy inhibition attenuates CAF activation, leading to reduced collagen deposition. In addition, in both WT hosts and autophagy-deficient hosts, CASP3 (caspase 3) was negative in most ACTA2/α-SMA-positive CAFs and FABP4-positive PSCs [19] indicating that loss of autophagy did not affect cell death in the tumor stroma ( Figure S1B).
It is well-known that cancer cells primarily rely on glucose and therefore take up large amount of glucose, leading to a competition with surrounding nonmalignant cells (such as CAFs and immune cells) for glucose [22]. To mimic situations that might occur in vivo, we subjected CAFs to low-glucose treatment in vitro. We then tested the mRNA levels of activated CAFs markers including Pdpn (podoplanin), Pdgfra, Vim (vimentin), Des (desmin), Fap and Acta2. As shown in Figure 1(j), compared to WT CAFs, these markers expressions were further reduced in both atg3-KO and atg5-KO CAFs under low glucose starvation. Consistently, the protein level of VIM, FAP and ACTA2/α-SMA were also reduced in atg3-KO or atg5-KO CAFs (Figure 1(k)). However, no obverse change was detected in response to low serum treatment ( Figure 1(l-m)), suggesting that autophagy is critical for CAF activation, particularly under glucose starvation.

Autophagy supports proline biosynthesis and collagen production in CAFs
Tumor stroma generally exhibit increased deposition of collagens compared with quiescent fibroblasts [8]. Therefore we performed Masson's Trichrome staining in orthotopical samples from WT and autophagy-deficient mice and the results demonstrated that there was less collagen content in autophagy-deficient hosts (Figure 2  (a,b)). Consistently, decreased expression of COL1A1 and collagen secretion in atg3-KO CAFs were also observed upon glucose deprivation (( Figure 2(c,d)). Glycine and proline are highly abundant in collagen, comprising approximately 44% of the collagen molecule [23]. As we know, autophagy is an important process of replenishing amino acids when cells faced nutrientdeprived environments [17]. Therefore, we next tested the amino acid levels in WT CAFs and autophagydeficient CAFs. As shown in Figure 2(e), several amino acids, particularly proline, were lower in atg3-KO CAFs compared with that in WT CAFs under glucose deprivation condition. Interestedly, we also observed that ornithine and citrulline, which are closely associated to proline synthesis, were also declined in autophagydeficient CAFs after glucose depletion (Figure 2(f-h)). Therefore, we predicted that the decrease of proline induced by autophagy-deficiency may partially result from reduced proline synthesis. To test this hypothesis, we performed [U-13 C]-glutamine tracing experiments to assess whether autophagy affects diversion of glutamine carbons into proline biosynthesis (Figure 2(i)). As shown in Figure 2(j), glutamine-derived proline was reduced in atg3-KO CAFs upon glucose deprivation. In addition, the fractions of ornithine and putrescine derived from [U-13 C]-glutamine were also decreased in autophagydeficient CAFs (Figure 2(k,l)). Collectively, these data suggested that autophagy has an impact on proline biosynthesis in CAFs.

Autophagy deficiency reduces proline biosynthesis by repressing NADK2
We next investigate how autophagy influences proline biosynthesis in CAFs. Recent studies have shown that mitochondrial NADPH pool is essential to enable proline biosynthesis [24,25]. Therefore, we examined the role of autophagy in regulating NADPH levels in whole cells and immunopurified mitochondria in CAFs. As shown in Figure 3(a,b), although the total NADP(H) abundance was slightly changed at whole cell level upon autophagy deficiency, mitochondrial NADP(H) abundance was obviously reduced in atg3-KO CAFs. Moreover, NADK2, an enzyme responsible for production of mitochondrial NADP(H) and proline biosynthetic [24,25] (Figure 3(c)), was decreased in autophagy-deficient CAFs after glucose depletion ( Figure 3 Mitophagy is a selective form of autophagy targeting dysfunctional or redundant mitochondria for degradation [26], we then test whether mitophagy is involved in regulating mitochondrial NADP(H) abundance. We generated mitophagy-deficient CAFs by using CRISPR-Cas9 technology to knock-out PRKN or BNIP3L (BCL2/adenovirus E1B interacting protein 3-like), which are essential for mitophagy [27]. Consistently, mitochondrial NADP(H) abundance and NADK2 level were also reduced in mitophagy-deficient CAFs under glucose depletion ( Figure 3(e-j)). These results demonstrated that autophagy deficiency reduces mitochondrial NADPH poor in a mitophagy-dependent manner.
Since mitochondrial NADPH poor is essential for proline biosynthesis [24,25], we next performed [U-13 C]-glutamine tracing experiments to assess whether mitophagy affects the diversion of glutamine carbons into proline biosynthesis. As shown in Figure 3(k-m), glutamine-derived proline, ornithine, and putrescine were reduced in prkn-KO CAFs upon glucose deprivation, suggesting that mitophagy has an impact on proline biosynthesis in CAFs. In addition, we found that both the protein level of COL1A1 and collagen secretion were also decreased in prkn-KO CAFs after glucose depletion ( Figure 3(n-o)), indicating that loss of mitophagy could lead to collagen secretion deficiency. Previous work demonstrated that stroma derived from autophagy-deficient CAFs exhibited reduced levels of multiple immune and inflammatory populations in addition to reduced collagen deposition [16]. Consistently, IHC staining revealed that ADGRE1/F4/80-, CD3-, and CD4-positive cells were decreased in autophagydeficient (Ubc-Cre-Esr/ERT2; atg5 fl/fl ) mice compared with WT mice ( Figure S2A,B), suggesting that stroma autophagy influences the recruitment of certain immune cell populations in vivo. However, no obverse change was detected in the mitophagy-deficient (prkn −/− ) hosts ( Figure S2A,B).
To determine whether proline played a role in autophagy or mitophagy deficiency-reduced collagen production, we cultured atg3-KO and prkn-KO CAFs in the low-glucose media with or without exogenous proline. CAFs lacking ATG3 or PRKN showed decreased COL1A1 expression and collagen secretion, which were rescued by proline supplementation to the low-glucose medium ( Figure 3(p-s)). NADK2 or ALDH18A1/P5CS (aldehyde dehydrogenase 18 family, member A1) are two important enzymes in the mitochondrial proline biosynthetic pathway [24,25,28]. To further validate the role of proline, we stably knocked out NADK2 or ALDH18A1/P5CS in CAFs (Figure 4(a,b)). Cell viability was not decreased in nadk2-KO or aldh18a1/p5cs-KO CAFs compared with WT CAFs ( Figure S2C), indicating that reduced proline could not cause cell death. Next, we orthotopically implanted PDAC cells with the same amount of WT CAFs, nadk2-KO CAFs or aldh18a1/p5cs-KO CAFs into the pancreas of C57Bl/6 J mice. As shown in Figure 4(c-f), both desmoplastic reaction and COL1A1-positive cell populations were reduced in tumor samples which were co-injected with nadk2-KO or aldh18a1/p5cs-KO CAFs. Consistent with the in vivo results, the protein level of COL1A1 and collagen secretion was also reduced in nadk2-KO or aldh18a1/p5cs-KO CAFs, further demonstrating that proline synthesis plays a critical role in collagen production ( Figure 4(g-h)). We next tested whether impaired synthesis of proline in CAFs influenced tumor cell proliferation. As shown in Figure 4(i-j), tumor-promoting activity of CM from CAFs was impaired when proline biosynthesis was inhibited upon glucose depletion. Then we tested the amino acid levels in CAFs CM and found that, besides proline, alanine was also declined in nadk2-KO or aldh18a1/p5cs-KO CAFs after glucose depletion ( Figure 4(k-l)), consistent with a previous report [4]. Therefore, we speculated that impaired proline synthesis has an impact on alanine secretion from CAFs, which might result from CAFs inactivation.

Mitophagy is required to maintain the CAFs phenotype
Next, we investigated whether mitophagy deficiency changes the phenotype of CAFs. As shown in Figure 5(a,b), the mRNA levels of activated CAFs markers including Pdpn, Pdgfra, Vim, Des, Fap, Acta2 were reduced in prkn-KO and bnip3l-KO mCAFs upon low-glucose depletion. To further confirm the effect of mitophagy on CAF activation in vivo, we orthotopically transplanted 15,376 T cells into WT or mitophagydeficient prkn −/− mice. Consistently with the in vitro data, there was less collagen content in mitophagy-deficient hosts ( Figure 5(c,d)). In addition, both ACTA2/α-SMA-positive and COL1A1-positive cell populations were reduced in prkn-KO   mice, demonstrating that mitophagy in stromal promoted CAF activation ( Figure 5(e-h)). We then tested the effect of mitophagy-deficient CAFs on supporting PDAC growth under glucose starvation. As shown in Figure 5(i,j), tumorpromoting activity of conditioned medium from CAFs was impaired when mitophagy was inhibited under glucose starvation. Moreover, in the orthotopic model of PDAC, tumors were significantly smaller when grown in prkn −/− mice compared to Prkn +/+ mice ( Figure 5(k)), which highlights the importance of stromal mitophagy in PDAC.
In summary, the findings we present in this report provide mechanism insights into the role of autophagy in CAF activation. It has been reported that impaired collagen secretion in autophagy-deficient cells due to defects in procollagen degradation via the lysosome [16,29,30]. Consistently, our data indicated that autophagy, but not mitophagy, is required for the pro-collagen 1 (PC1) degradation in CAFs ( Figure S3A,B). As shown in Figure S3A, B, LAMP1-PC1 colocalization was increased in prkn-KO CAFs upon treatment with CHQ (at 25 mM glucose). However, the colocalization was not increased in prkn-KO CAFs upon CHQ treatment after glucose starvation (at 1 mM glucose) ( Figure S3C-E). Especially, our study showed that mitophagy in CAFs could promote collagen secretion by regulating the synthesis of proline, but not degradation of procollagen ( Figure S3C-E). Therefore, our findings uncover a novel pathway that autophagy in CAFs has an impact on desmoplasia by influencing proline biosynthesis.
Mouse models. prkn KO mice (prkn −/− ) were purchased from Cyagen Biosciences (Suzhou, China). The nuclear background of the mice was C57BL/6 N. Heterozygous mice were then interbred to obtain homozygous WT control and KO mice. Mice carrying two conditional alleles of ATG5 (Atg5 fl/fl ) were crossbred with UBC-Cre Esr/ERT2 mice (Model Organisms Center, Shanghai). Cre-mediated recombination was induced with intragastric administration of tamoxifen in corn oil (2 mg/day for 5 days). Control groups were intragastric administrated with corn oil. All genotyping was done by PCR.
Cell culture. All cell lines were grown in DMEM (MACGENE, CM15023) supplemented with 10% (vol:vol) fetal bovine serum and the appropriate amount of penicillinstreptomycin in a 37°C incubator with a humidified, 5% CO2 atmosphere. CAFs were generated from C57BL/6 J harboring mouse PDAC. These animals were pre-treated with doxycycline (Selleck, S5159) drinking water (2 g/L) and kept in doxycycline regimen for the duration of the experiment, and then were injected with10 5 15376T cells into the pancreas. Pancreatic tumors were extracted at 10 days, digested in collagenase and dispase (Roche, 10,269,638,001) and mechanically minced. Neoplastic tissues were dissociated at 37°C for 1 h. Cells were collected in cell culture dishes in DMEM with 10% FBS in the absence of doxycycline to limit the growth of iKRAS mouse PDAC cells. After 24 h, mCAFs were immortalized by transfection with SV40 (Addgene, 22,298; deposited by Dr. Eric Campeau) retro-viruses.
CAFs were verified by measuring ACTA2 expression.  Cell proliferation assay. Cells were seeded in 12-well plates at a density of 5 × 10 4 cells/well and treated with different treated medium after culturing for 24 h. Cell proliferation was evaluated after 24 h, the cells were collected, diluted with a trypan blue working solution and counted.
De-paraffined sections with 3-μm thickness were repaired by microwave. The sections were then treated with 0.5% Triton X-100 (Solarbio, T8200) for 30 min and blocked with goat serum (ZSGB-BIO, ZLI-9022) for 1 h at room temperature. Washing with PBS 3 times between each step. The sections were incubated with antibodies, overnight at 4°C in a humidified chamber, followed by incubation with secondary antibodies and DAPI. Anti-Fade Mounting Medium (Solarbio, S2100) was used to seal the sections. Immunofluorescent images were obtained using confocal microscope.
Cell death. Post-treatment cell viability was determined by staining with propidium iodide (PI, 2 μg/ml; MedChemExpress, HY-D0815) and flow cytometric analysis on a FACScan.
Rapid isolation of mitochondria. Rapid isolation of mitochondria was performed following the published methodology [32]. WT cells or cells with atg3 or prkn or bnip3l knockout were engineered to express the HA-tagged SYNJ2BP/OMP25 protein. 3XHA-EGFP-SYNJ2BP/OMP25 was amplified and cloned into pcDNA 3.1. For transfecting CAFs, Neon transfection system (Invitrogen, MPK5000) was used following the manufacturer's instructions. Cells were washed on ice and dounce homogenized in KPBS (136 mM KCl, 10 mM KH2PO4, pH 7.25). The homogenate was then centrifuged at 1000 g for 2 min at 4°C and the supernatant from the centrifuged sample was applied to anti-HA beads (Thermo Fisher Scientific, 88,837) and incubated with end-over-end rotator for 8 h. The resultant beads were then washed with KPBS. Then 80:20 methanol: water was used to elute mitochondria for NADP(H) measurements.
Measurement of NADP(H). NADP(H) measurements were performed using an NADP + /NADPH Assay Kit (Beyotime, S0179), with modifications as described in published study [33]. Briefly, metabolites from whole cells and isolation of mitochondria samples were extracted with 80:20 methanol: water. Then the metabolites was dried down in a SpeedVac concentrator for 2 h. Metabolites were then resuspended with NADPH extraction buffer and centrifuged for 5 min at 600 g. The supernatant was split in half after centrifued. One half was kept on ice for 30 min, the other half was exacted to 60°C incubation for 30 min to decompose NADP + . 100 μL of G6PDH mix (98 uL buffer and 2 μL G6PDH enzyme mix from the manufacturer) was added to each sample and incubated for 10 min at 37°C. 10 μL of manufacturer's NADPH developer was then added into each well and incubated for one hour at 37°C. Values were recorded with a microplate reader at 450 nm. The amount of NADPH was calculated from the standard curves and normalized for cell counts. media (25 mM glucose) for 24 h. (k) WT CAFs and nadk2-KO CAFs (Nadk2 knocked-out) were cultured in complete media (25 mM glucose) or low glucose media (1 mM glucose) for 24 h. The levels of intracellular free amino acids were quantified using mass spectrometry. (l) WT CAFs and aldh18a1/p5cs-KO CAFs (Nadk2 knocked out) were cultured in complete media (25 mM glucose) or low glucose media (1 mM glucose) for 24 h. The levels of intracellular free amino acids were quantified using mass spectrometry.
Extracellular matrix (ECM) extraction and collagen staining. ECM extraction and collagen staining were performed as previously described [28]. In brief, confluent CAFs were grown for two days on plates coated with 0.1% gelatin in the presence of 50 μM L-ascorbic acid 2-phosphate in the different media. Plates were decellularized with 0.5% Triton X-100 for 5 min on a rotating platform. Three times the volume of PBS was added, and ECM was equilibrated overnight at 4°C. To measure collagen abundance, extracted ECM was stained with picrosirius red (Freemore, D00041). The stain was extracted with 0.1 M NaOH, and collagen staining was measured at 550 nm using a microplate reader. Values were normalized for cell counts.
Xenografts. All experimental uses of mice in this study were housed and supplied by the Experimental Animal Center of Peking University Health Science Center (Beijing, China). Five-to-six-week-old C57BL/6 J mice were used for orthotopic tumorigenesis. These animals were given doxycycline-containing diet and kept in doxycycline regimen for the duration of the experiment for induction of tumor development. Mice were injected with10 5 15376T cells into the pancreas. Briefly, mice were fixed in a supine position and performed an abdominal incision. The spleen was identified and gently ripped out through the incision to expose the pancreas. Twenty μL of 15,376 T cells were injected into the pancreas. The injection site was clamped for about 30s after removing the syringe. The tissues were put back carefully in its original position and the wound was sutured with absorbable sutures. Ten days after inoculation, the mice were sacrificed when the tumors have grown. Then the tumor volume was measured with a vernier caliper. The drug was injected intraperitoneally, starting from the fifth day. All animal experiments were performed in accordance with a protocol approved by the Department of Laboratory Animal Science of Peking University Health Science Center and supervised by the institutional review board of Peking University.
Metabolite extraction and stable isotope labeling. Cells were washed with ice-cold PBS, quenched with 80% methanol and subjected to three rapid freeze-thaw cycles. The cell debris was removed by centrifugation (12,000 g, 15 min) at 4°C. After sample normalization based on total protein levels, the supernatant containing aqueous metabolites was collected and evaporated to dryness using a SpeedVac concentrator. For glutamine tracing experiments, cells were seed in 10-well plates and cultured for 24 h in DMEM supplemented with 10% FBS. And then medium was replaced with DMEM without l-glutamine and glucose (Gibco, A1443001) supplemented with 2 mM [U-13 C] l-glutamine (Cambridge Isotope Laboratories, CLM-1822-H-0.1), 10% dialyzed FBS and 25 mM glucose or 1 mM glucose. After 24 h, metabolism was quenched and metabolites were extracted using the above method.
Masson's trichrome staining. Masson's trichrome staining was used to detect collagen distribution. Deparaffined sections with 5-μm thickness were stained according to instructions (Applygen, B1130). Images were taken under upright microscopy for analysis.