N-acetyl cysteine induces quiescent-like pancreatic stellate cells from an active state and attenuates cancer-stroma interactions

Pancreatic stellate cells (PSCs) occupy the majority of the pancreatic cancer microenvironment, contributing to aggressive behavior of pancreatic cancer cells (PCCs). Recently, anti-fibrotic agents have proven to be an effective strategy against cancer, but clinical trials have shown little efficacy, and the driving mechanism remains unknown. N-acetyl-cysteine (NAC) is often used for pulmonary cystic fibrosis. Pioglitazone, an agonist of peroxisome proliferator-activated receptor gamma, was habitually used for type II diabetes, but recently reported to inhibit metastasis of PCCs. However, few studies have focused on the effects of these two agents on cancer-stromal interactions. We evaluated the expression of α-smooth muscle actin (α-SMA) and the number of lipid droplets in PSCs cultured with or without NAC. We also evaluated changes in invasiveness, viability, and oxidative level in PSCs and PCCs after NAC treatment. Using an indirect co-culture system, we investigated changes in viability, invasiveness, and migration of PSCs and PCCs. Combined treatment effects of NAC and Pioglitazone were evaluated in PSCs and PCCs. In vivo, we co-transplanted KPC-derived organoids and PSCs to evaluate the effects of NAC and Pioglitazone’s combination therapy on subcutaneous tumor formation and splenic xenografted mouse models. In vitro, NAC inhibited the viability, invasiveness, and migration of PSCs at a low concentration, but not those of PCCs. NAC treatment significantly reduced oxidative stress level and expression of α-SMA, collagen type I in PSCs, which apparently present a quiescent-like state with a high number of lipid droplets. Co-cultured PSCs and PCCs mutually promoted the viability, invasiveness, and migration of each other. However, these promotion effects were attenuated by NAC treatment. Pioglitazone maintained the NAC-induced quiescent-like state of PSCs, which were reactivated by PCC-supernatant, and enhanced chemosensitivity of PCCs. In vivo, NAC and Pioglitazone’s combination suppressed tumor growth and liver metastasis with fewer stromal components and oxidative stress level. NAC suppressed activated PSCs and attenuated cancer-stromal interactions. NAC induces quiescent-like PSCs that were maintained in this state by pioglitazone treatment.


Introducion
Pancreatic ductal adenocarcinoma (PDAC), which is the ninth most common cancer, causes fourth common in cancer-related deaths in America [1]. Although great strides have been made to explore the mechanism of pathogenesis in the past decade, the 5-year survival rate of PDAC is still less than 7% because of the high rate of resistance to chemotherapy and metastasis [2], which indicates that a better treatment strategy should be developed to improve the prognosis of PDAC.
There is a growing consensus has been accepted that the tumor microenvironment may be a significant contributing factor to the aggressive behavior of cancer and its resistance to chemotherapy [3]. PDAC is notable for its excessive desmoplasia that is caused predominately by pancreatic stellate cells (PSCs) [4], which are usually in a quiescent state containing a large amount of vitamin A stored in lipid vacuoles [5]. They are activated and become myofibroblast-like cells by various growth factors secreted from pancreatic cancer cells (PCCs), which include platelet-derived growth factor (PDGF) and vascular endothelial growth factor (VEGF) [6]. As a result, activated PSCs express α-smooth muscle actin (α-SMA) and secrete several cytokines and chemokines, such as IL-6, PDGF, TGF-β, and connective tissue growth factor (CTGF), which promote the invasion ability of PCCs [7]. Activated PSCs also secrete an abundant extracellular matrix (ECM), inducing immunosuppression and chemoresistance [3,8]. Thus, targeting cancer-stroma interactions (PCCs-PSCs) appears to be a promising therapeutic strategy. It has been recently reported that inhibiting Hedgehog signals in tumor stromal cells retards the growth and metastasis of PCCs [9]. However, some research revealed that targeting tumor stroma can also promote cancer cell aggressiveness [10,11]. Therefore, owing to the controversial nature of stromal-targeting therapeutic approaches, PSCs' precise characteristics are still under investigation.
N-acetyl-cysteine (NAC) is an aminothiol and synthetic precursor of de novo Glutathione (GSH) synthesis [12]. NAC is also commonly applied to ameliorate inflammation under pathological conditions such as chronic obstructive pulmonary disease, influenza, and idiopathic pulmonary fibrosis [13,14]. NAC also inhibits the activation of transcription factor activities like JNK, p38 MAPK, and NF-κB that regulate numerous genes' expression [15]. Besides, several reports indicate that NAC induces apoptosis in colon carcinoma cells [16]. We have previously shown that some anti-fibrosis agents, such as Pirfenidone and calpeptin, have the ability to suppress pancreatic cancer by disrupting cancer-stromal interactions [17,18]. Enhanced desmoplastic responses by coadministration of Pirfenidone and NAC have been observed in HapT1-derived orthotopic tumors in a hamster model [19]. However, few studies explored the effects of NAC on cancer-stromal interactions or the functional alterations of PSCs in PDAC.
In this study, we hypothesized that NAC might be an effective anti-fibrosis against pancreatic cancer and alter the function of PSCs. The present data revealed that low concentrations of NAC suppressed activation of PSCs and reduced its oxidative stress level. NAC-treated PSCs had a quiescent-like state that was maintained by coadministration of NAC and Pioglitazone, a ligand of peroxisome proliferator-activated receptor gamma (PPARγ) which used for type II diabetes previously [20]. Activated PSCs significantly increased the cell viability and invasiveness abilities of PCCs. Additionally, PCCs also promoted the cell viability, migration ability and oxidative stress levels of PSCs, but these mutual promotion effects were effectively attenuated by NAC treatment. The effects of combined of NAC and PLZ were showed superiority of treatment than Gemcitabine on subcutaneously transplanted mouse model, which were co-implanted with PDAC drived organoids and PSCs. Moreover, we used a splenic xenografted model to evaluate the effects of this combined treatment. Combined treatment with NAC and Pioglitazone may be a novel therapy that targets cancer-stromal interactions.

Cell and reagents
Human primary activated PSCs were harvested from surgical specimens of pancreatic cancer, established and maintained as described by Bachem et al. [21][22][23] . The identity of the isolated PSCs was confirmed by their fibroblast-like morphology, positive staining for α-SMA and negative for cytokeratin 19 (CK19) (Fig. S1), though there are a tiny fraction of the fibroblast-like cells express very low levels of α-SMA. Three primary cultures of PSCs were used within eight passages in this study. Three human PCC lines were also used in this study: PANC-1 (Riken Bioresource Center, Ibaraki, Japan), SUIT-2, and BxPC-3 (National Kyushu Cancer Center, Fukuoka, Japan). All cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum and maintained at 37°C with 10% CO 2 .
Agents included NAC, chloroquine (CQ), JQ-1, Pioglitazone (PLZ), Cryptotanshinone (CTS), and Butylated hydroxyanisole (BHA) were purchased from Sigma-Aldrich (Tokyo, Japan). We also used Anti-IL-6 neutralizing antibody (69001-1-IG, Proteintech Group, Inc.) and human recombination IL-6 proteins (AF-200-06, PeproTech) in this study. For in vitro experiments, NAC was dissolved in PBS at 100 mM and other agents were dissolved in DMSO at 1 mM. All stock solutions were stored at − 20°C until use. Additionally, 400 μL of 100 mM NAC-containing PBS was added to 16 mL DMEM for a final concentration of 2.5 mM (pH =7.55). There was no noticeable change in pH after adding NAC to the culture medium, which confirmed that the buffering was sufficient to maintain a stable pH.

PDAC organoid culture
To establish PDAC organoids, we harvested and digested surgical specimens of PDAC patients with a Tumor Dissociation Kit (human, Cat#130-095-929; Miltenyi Biotec, CA, USA) and cultured them using the method as described previously [24]. Briefly, the digested tumor tissue was integrated into a growth factor-reduced Matrigel (Cat# 356231; BD Bioscience, CA, USA), and finally cultured in a human complete medium 37C°for 2 weeks.

Production of conditioned media (CM)
To exclude the effects of growth factors in FBS, we prepared conditioned media using supernatants (SNs) from PSCs and PCCs cultured in serum-free DMEM. SNs were passed through an Amicon Ultra-15 30 K Centrifugal Filter Device (Millipore, Billerica, MA, USA) in accordance with the manufacturer's protocol. In brief, subconfluent (75% confluence) SUIT-2 cells, PSCs, or 2.5 mM NAC-treated PSCs were cultured in serum-free DMEM for 48 h. Then, 12 mL of culture supernatant was collected in the filter device, centrifuged at 5000 g in a fixed-angle rotor for 30 min followed by recovery with 10 mL of 11.2 mM NaCl. The conditioned media [CM; also defined as supernatant (SN)] was collected after removal of salts and stored at − 20°C until use. SN from SUIT-2 (SUIT2-SN) or NAC-treated SUIT-2 (NAC-SUIT2-SN) were used to determine the effects on PSCs in cell viability and migration assays. SN from PSCs (PSC-SN) or NAC-treated PSCs (NAC-PSC-SN) was used in cell viability assays, migration assays, and gemcitabine resistance experiments. DMEM was centrifuged and concentrated as a control.

Cell viability assays
We performed a CellTiter-Glo viability assay (Promega; Madison, WI) to measure the viability of PSCs and PCCs in accordance with the manufacturer's protocol. Cells were seeded in 96-well plates (Greiner Bio-One, Kremsünster, Austria) in triplicate at 1 × 10 3 cells/well in 100 μL fresh DMEM containing 10% FBS. The background was subtracted using values from wells containing only culture medium. To assess stimulation effects, the medium was replaced with medium containing NAC, PBS, or each SN as indicated in the figure legends at 24 h after seeding. For all cell viability assays, data are representative of three independent biological experiments.

Invasion and migration assays
The ability of migration and invasion in PSCs and PCCs were performed by counting the number of cells migrated or invaded through a Transwell System (8-mm pore size; Becton Dickinson, Franklin Lakes, NJ). For monoculture or co-culture migration and invasion assays, we seeded cells and treated them as previously described [25]. For a collagen-coated invasion assay, membranes were coated with 100 μl collagen type I (354, 236; Corning, America), type IV collagen (354,233; Corning), or Matrigel at 20 mg per well. To assess the effects of PSC and PCC supernatant stimulation on migration and invasion, separate batches of cells were cultured with each SN indicated in the figure legends. At 24 h (for migration) and 48 h (for invasion) after cell seeding, migrated or invaded cells were fixed with 70% ethanol, stained with hematoxylin and eosin, and counted in five random fields (× 100 magnification). In some cases, migration and invasion abilities were evaluated by normalization to total cell numbers of each cell type to avoid the effects of proliferated cell numbers on migration and invasion. For all migration and invasion assays, three experiments were independently conducted in triplicate.

Assessment of oxidative stress response
The levels of total glutathione (GSH) and intracellular ROS (Reactive oxygen species) were measured using an Oxiselect Total Glutathione Assay Kit (STA-312, Cell Biolabs Inc., San Diego, CA, USA) and an Oxiselect intracellular ROS Assay Kit (STA-342, Cell Biolabs) according to the manufacturer's instructions. The total glutathione concentration in the samples was measured by recording the absorbance at 405 nm with a microplate reader and analyzed by comparison to a standard curve. Intracellular ROS levels were measured in terms of relative fluorescence units (RFUs) at excitation and emission wavelengths of 485 nm and 535 nm. All experiments were independently performed three times.

Lipid droplet accumulation assay
Lipid droplet accumulation in PSCs was analyzed by staining with bodipy (#D-3922; Life Technologies) as described previously [26]. After staining, images were captured in three random fields using a fluorescence microscope, and we counted the number of bodipypositive punctures per cell in 20 cells.

qRT-PCR analysis
Total RNA of cells was extracted using a High Pure RNA Isolation kit (Roche, Mannheim, Germany). We performed qRT-PCR using a QuantiTect SYBR Green RT-PCR kit (Qiagen, Tokyo, Japan) and the CFX96 Real-Time PCR System (Bio-Rad Laboratories, Hercules, CA). 18S rRNA was used to normalize mRNA expression. All primers were purchased from Sigma-Aldrich. For details, see Supplementary Table S1. Three experiments were independently conducted in triplicate.

Western blotting analysis
Total cellular proteins of PSCs and PCCs were prepared with PRO-PREP Protein Extraction Solution (InTron Biotechnology, Seongnam, Korea). Supernatant proteins were collected as described previously [26]. Protein samples of cell lysates (20 μg) or supernatants (30 μg) were separated by 10% SDS PAGE on Mini-PROTEAN TGX Precast Gels (Bio-Rad Laboratories), followed by transfer to Trans-Blot Turbo Mini PVDF Transfer Packs (Bio-Rad Laboratories). The membrane was incubated at 4°C overnight with primary antibodies (1:1000; supplementary Table S2) and then probed with secondary antibodies (1:2000; supplementary Table S2). Immunoblots were developed by ECL western blotting detection reagent (Bio-Rad) and analyzed with the ChemiDoc XRS System (Bio-Rad Laboratories).

Statistical analysis
For in vitro experiments, results are expressed as means ± SD. Significances between two groups were assessed by one-way analysis of the variance or the Student's t-test using GraphPad Prism 7.0 (GraphPad Software). Values of P < 0.05 were considered statistically significant in all analyses. Survival analyses were conducted using the Kaplan-Meier method. All experiments were performed at least three times except animal experiments.

Results
NAC inhibits the cell viability, migration, and invasion of PSCs rather than PCCs.
First, we investigated the cell viability of both PSCs and PCCs by NAC treatment. In monoculture, NAC inhibited the viability of PSCs in a dose-dependent manner, and the IC 50 of each primary culture of PSCs was calculated (  Student's t-tests (unpaired, unequal variance) were used for comparisons. g Expression of Nrf2 and HMOX-1 in PCCs and PSCs treated with or without NAC by western blotting.*P < 0.05; **P < 0.01; ***P < 0.001; n.s, no significance Because NAC is commonly used as an antioxidant, we also examined the levels of total glutathione (GSH) and ROS, which are oxidative stress response markers, in PSCs and PCCs. As expected, NAC increased the level of GSH and decreased the ROS level in both PSCs and PCCs (Fig. 1e, f). However, when co-treated with H 2 O 2 , an oxidative stress inducer that causes oxidative stress in PSCs and PCCs, NAC still effectively inhibited oxidative stress in PSCs rather than PCCs. Besides, the expression of Nrf2, the master regulator of oxidative stress, as well as its downstream targets HMOX-1, NQO1and GCLC were also decreased by NAC treatment in PSCs but not in PCCs (Fig. 1g, Supplementary Fig. 2 J).
To confirm the effectiveness and safety of NAC for anti-tumor application, we explored the effects of various concentrations on PSCs. The high concentration of NAC had an apparent toxic effect on PSCs (Supplementary Fig. S2C). However, 1 mM NAC decreased the expression of α-SMA significantly. When PSCs were cocultured with supernatant from PCCs (SUIT2-SN), 1 mM NAC was not sufficient to decrease the expression of α-SMA ( Supplementary Fig. S2F). Additionally, 1 mM NAC had little effect on the migration ability of PSCs with or without SUIT2-SN ( Supplementary Fig. S2G). Although in flow cytometer the apoptosis rate of PSCs was increased after 48 h of 2.5 mM NAC treatment ( Supplementary Fig. S2D), which might be due to the suppression of growing active cells and normal programmed cell death mechanism was activated; the expression of cleaved caspase-3 evaluated by western blotting, one of the best-known markers of apoptosis, showed no significant difference between normal PSCs and NAC-treated PSCs ( Supplementary Fig. S2E). Thus, we used 2.5 mM to investigate its specific effects on PSCs in the following in vitro experiments in which the concentration inhibited cell viability and migration effectively (Fig. 1e).

NAC decreases activation of PSCs and induces a quiescent-like state
Activated PSCs expressed a high level of α-SMA. We used Chloroquine (CQ) and JQ-1 as the positive control, which were reported to inhibit activation of PSCs by our previous research and Kumar K [26,28]. Consistent with CQ and JQ-1, after NAC treatment, the expression of α-SMA and collagen type I were decreased markedly (Fig. 2a, b). However, NAC did not decrease vimentin expression, a marker commonly used to characterize the cells of fibroblast origin [29]. Moreover, treatment with NAC increased the number of lipid droplets in PSCs (Fig. 2c), which indicated a transition from activation to a quiescent-like state. These data indicate that NAC decreases the activity of PSCs.
To examine the characteristics of PSCs by NAC treatment, we established NAC-treated PSCs used in the following experiments, which were treated with NAC for 2 weeks and then cultured without NAC for another 2 weeks. After 4 weeks of culture, the number of PSCs with spindle-like morphology was decreased significantly, and the number of PSCs with star-like morphology was generally increased (Fig. 2d- In cell viability assay, PCCs treated with PSC-SN and 5 mM NAC demonstrated no significant difference from PCCs treated with only PSC-SN (Fig. 3e). SN derived from NAC-treated PCCs still enhanced the viability of PSCs (Fig. 3b). These results indicate that NAC attenuates cancer-stroma interactions by targeting PSCs. Thus, we next examined if NAC-treated PSCs were attenuated their promotion effects on PCCs.
It has been well reported that PSCs promote migration and invasion of PCCs as well as EMT processes and resistance to chemotherapeutic agents [30]. E-cadherin expression was decreased and Vimentin was increased in PCCs cocultured with PSC-SN, but not in PCCs cocultured with NAC-PSC-SN (Fig. 3g). Additionally, PSC-SN increased the resistance of PCCs to gemcitabine. However, NAC-PSC-SN showed slightly increased resistance (Fig. 3h).
IL-6 is a central mediator of cancer cell-stromal cell interactions in PDAC. PSC-derived IL-6-activated STAT3 signaling in PCCs has been well reported in previous studies [31]. SN from PSCs, but not NAC-PSCs, increased the activation of STAT3 (Supplementary Fig. S5B). Additionally, an anti-IL-6-neutralizing antibody (IL-6 neu) and human recombination IL-6 (hr IL-6) were used to confirm activation of this pathway. To verify the effectiveness of these agents, we analyzed c-Myc (a transcriptional target of STAT3) and p21 (negatively regulated by STAT3 signaling) (Supplementary Fig. S5B). We also used a specific STAT3 inhibitor, cryptotanshinone (CTS), which significantly decreased the phosphorylation level of STAT3 in PCCs ( Supplementary Fig. S5A). In PCCs, the promotive effects of PSC-SN on migration and invasion were inhibited by IL-6 neu and STAT3 inhibitor CTS. However, PSC-SN had no further promotive effects when added with human recombination IL-6 ( Fig. 3i, Supplementary  Fig. S5C). IL-6 neu and CTS showed no further inhibitive  Taken together, these results indicate that NAC treatment has the potential to target activated PSCs and attenuate cancer-stroma interactions.

Characteristic changes in PSCs induced by NAC treatment in cancer-stroma interaction
To determine the secreted factors related to the inhibitory effects of NAC on cancer-stroma interactions, we evaluated mRNA and protein expression levels of those factors in PSC-SN and whole PSC cell lysates (Fig. 4a). Treatment with NAC significantly reduced the mRNA and protein levels of most secreted factors, especially PDGF-A, PDGF-B, TGF-β, CTGF, VEGF, α-SMA, FN, and collagen type I, but not collagen IV. Exposure to PCC-SN considerably increased the expression of PDGF-A, IL-6, CTGF, MMP2, α-SMA, FN, but NAC still effectively inhibited their expression. Additionally, the expression of oxidative stress response markers Nrf2 and HMOX-1 was increased by PCC-SN, but NAC inhibited their expression (Fig. 4b). In PSCs, the GSH level was increased and the ROS level was decreased by PCC-SN and NAC inhibited these changes induced by PCC-SN (Fig. 4c, d). PCC-SN dramatically decreased collagen IV in supernatants from PSCs, although there was no significant collagen IV change in the whole-cell lysate. NAC inhibited the decrease of collagen IV in supernatants of PSCs treated with PCC-SN (Fig. 4a). These inhibitory effects were also observed by immunofluorescence staining (Fig. 4e). Compared with normal Matrix-and collagen type I-coated gels, the collagen type IV-coated gel showed a reduced number of invaded PCCs (supplementary Fig. S5E), which indicated that collagen type IV might have an anti-invasion function that was not affected by NAC. These findings suggest that NAC is an effective agent that targets tumor-stroma interactions.
To assess the mechanism mediated by NAC, we investigated several signaling pathways associated with the activation of PSCs. Activities of PI3K-AKT and NF-κB signaling pathways in PSCs were significantly decreased after NAC treatment. However, there was no change in ERK pathway activity (Fig. 5a). Exposure to PCC-SN increased the activities of PI3K-AKT and NF-κB signaling pathways in NAC-treated PSCs (supplementary Fig.S7A). We also found that the expression of peroxisome proliferator-activated receptor gamma (PPARγ), a nuclear receptor that regulates fatty acid storage and glucose metabolism [32], was significantly decreased (Fig. 5a). To investigate whether PPARγ plays a crucial role for inducing the in-active state in NAC-treated PSCs, we conducted small interfering RNA-mediated silencing of PPARγ in PSCs. In NAC-treated PSCs, but not untreated PSCs, the decreased expression of α-SMA was rescued by siPPARγ ( Supplementary Fig. S6A). In cell viability and migration assays of PSCs, the effects were decreased in NAC-treated PSCs. However, knockdown of PPARγ attenuated these effects ( Supplementary  Fig. S6B, C). Because untreated PSCs expressed little PPARγ, the knockdown effects in PSCs were minimal. Furthermore, we examined the correlation between the expression of PPARγ and the oxidative stress response by cbioportal, an online tool in which RNA-seq data of 184 PDAC samples from TCGA were analyzed (Supplementary Fig. S8). High expression of PPARγ was positively associated with low expression of ACOX2, ICAM1, ICAM family, MPO, PRIMPOL, NANOS1, NOS3, and NOS1.  These genes are well-reported as oxidative stress markers. These results indicated that the oxidative stress level and activation level in PSCs were decreased by NAC treatment.

Combined NAC with PLZ suppresses reactivation of PSCs and enhances chemotherapy sensitivity of PCCs
We evaluated the therapeutic efficiency of the combination of NAC and Pioglitazone (PLZ; a ligand of PPARγ). When treated with NAC, PLZ, or both, a morphological change was observed in PSCs (supplementary Fig.S7B). PLZ (20 μM) inhibited neither the cell viability of NAC-treated PSCs directly (Fig. 5c) nor the migration and invasiveness of PSCs (supplementary Fig.S7C). NAC-treated PSCs were reactivated by SUIT-2-SN (Fig. 5b, c). However, when cotreated with 20 μM PLZ, PSCs maintained their inactive state. The expression of PPARγ was also increased in PCCs after NAC treatment (Fig. 5d). Although low concentrations of PLZ or NAC alone did not inhibit the viability of PCCs, their co-treatment remarkably restricted the cell viability, migration, and invasiveness of PCCs (Fig. 5e, f; supplementary Fig.S7D). We also assessed downstream of PPARγ by quantitative RT-PCR. Both NAC and PLZ treatment increased the transcriptional expression of PLIN1, FABP4, Adipo, CD36, and LPL ( Supplementary Fig. S8B), but not RXRα that forms a transcription factor heterodimer with PPARγ to bind to DNA elements [33]. Cotreatment with NAC and PLZ had an additive effect on these downstream targets, which promoted lipid metabolism in PSCs. To confirm whether these effects were dependent or independent of intracellular GSH levels, we used another antioxidant, butylated hydroxyanisole (BHA). BHA at 50 μM also decreased the oxidative stress level in PSCs to a similar level as 2.5 mM NAC, but its effect was less than that of NAC when cocultured with SUIT-SN ( Supplementary  Fig. S3F, Fig. 4d). Moreover, both BHA and NAC decreased the expression of α-SMA and increased PPARγ expression, but NAC more effectively reduced the expression of α-SMA (Supplementary Fig. S3G). These results indicate that co-treatment of NAC and PLZ may be a promising new approach for cancer treatment.

NAC inhibits subcutaneous tumor growth in mice
To evaluate the effects of NAC on cancer-stroma interactions in vivo, we implanted SUIT-2 cells (1 × 10 6 ) alone into the left flank of mice and co-implanted SUIT-2 cells with PSCs (5 × 10 5 , respectively) into the right flank of the same mice (supplementary Fig. S9A). On day 7, NAC, or PBS (Control) were injected intraperitoneally into the mice for 4 weeks. In the group without treatment, tumors in the right flank (SUIT-2 + PSC) were much larger than the left flank (SUIT-2 alone) (supplementary Fig. S9A, B). Although the left flank showed few α-SMA-positive cells, probably originated from the host, the right flank's tumors showed much more numbers of α-SMA-positive cells and collagen fiber than the left flank (supplementary Fig. S9C, E). Treatment with NAC significantly suppressed tumor growth in both flanks compared with those of the control group (supplementary Fig. S9A, B). Compared to the control group, NAC decreased the collagen fiber areas and α-SMA positive index in both flanks (supplementary Fig. S9C-F). NAC also reduced the PCNA index in the right flank but not in the left flank than the control group (supplementary Fig. S9G). Moreover, NAC increased the PPARγ index and decreased HMOX-1 and Nrf2 index in both flanks. These findings suggest that NAC suppresses tumor formation and oxidative stress in tumor microenvironments in pancreatic cancer.

Combined treatment of NAC with PLZ inhibits tumor growth and metastasis in KPC cancer organoid cells coimplanted with PSCs
We previously established pancreatic cancer organoids for 2 weeks using PDAC cells derived from the KPC mouse (Fig. 6a). Compared with normal 2D-cultured cells, the implanted tumor showed more collagen fiber areas and more α-SMA-positive index in splenic xenograft models with 3D-organoids (supplementary Fig.S10). Thus, using this xenograft model was primely reflect tumor microenvironment in PDAC. We subcutaneously implanted KPC organoids with PSCs into nude mice treated with the control, gemcitabine (a first-line therapy for PDAC), and NAC with PLZ (Fig. 6b). After 4 weeks of treatment, the tumor volume in the cotreatment group (See figure on previous page.) Fig. 4 Characteristic changes in PSCs induced by NAC treatment in cancer-stroma interaction. PSCs were treated with 2.5 mM NAC, SUIT-2-SN, or SUIT-SN puls NAC, or untreated (Control). a The indicated expression of growth factors and ECM in PSCs were evaluated in qRT-PCR and western blotting. Protein secretion expression levels of growth factors were measured by western blotting of PSC-SN. Protein expression levels of collagen types I, collagen type IV, FN, and α-SMA was measured in whole-cell lysates with β-actin as the Control. b Western blotting of NRF2 and HMOX-1 (two oxidative stress response marker) levels of PSCs treated as above. c Glutathione (GSH) level in PSC cells treated as above. d Intracellular ROS levels in PSCs were treated as above. The values of RFUs (ROS level) were normalized to Control in each PSCs. e Representative microphotograph of α -SMA (Red) and collagen type IV (Green) immunofluorescence staining in PSC1 cells with indicated treatment. Original magnification, × 100. *P < 0.05; **P < 0.01; ***P < 0.001; n.s, no significance was significantly decreased compared with GEM and control groups (Fig. 6c, Supplementary Fig. S11A). Moreover, α-SMA-positive cells and collagen fibers were reduced by cotreatment with NAC and PLZ (Fig. 6e). The IL-6positive area index was also decreased in the cotreatment group. However, in cotreatment groups, expression of PPARγ was increased significantly and oxidative stress markers HMOX-1 and Nrf2 were decreased compared with the control group (Fig. 6e). To assess the toxicity and safety of this combination therapy, we monitored the weight change of each mouse and the average food intake of each group. Although there was no statistical difference in the average weight of each group, mice in the cotreatment group showed a weight gain trend (Fig. 6d, Supplementary Fig. S11B). The average food intake by mice in the cotreatment group showed no significant difference from the control group (Supplementary Fig. S11D). Nevertheless, mice in the GEM group showed significant  loss of appetite and weight gain cessation because of drug side-effects. Thus, NAC + PLZ co-treatment has excellent potential as a new approach to inhibit tumor growth. Next, we co-implanted KPC organoids with PSCs into the spleen of nude mice to investigate the effects of cotreatment on tumor metastasis. Mice were treated intraperitoneally with the vehicle, NAC, PLZ, or a combination for 3 weeks. On day 28, the mice were sacrificed, and their liver metastases were harvested and evaluated (Fig. 7a-c). Compared with control, the number of metastatic nodules (15 vs. two, average) in the liver (Fig. 7d) (liver volume: 2.61 vs. 1.41 cm 3 , average; liver weight: 2.12 vs. 1.13 g, average; Fig. 7e) was decreased significantly in the combined treatment group, although NAC or PLZ alone also decreased the number of metastatic nodules (Fig. 7d). The number of peritoneal disseminated nodules was also significantly decreased after the combined treatment (Supplementary Fig.S11E-G). We also performed immunohistochemical staining to evaluate metastasis nodules. NAC and NAC + PLZ groups showed significant decreases in α-SMA-positive areas, but metastasis nodules in the latter group showed smaller areas (Fig. 7e). Pathological studies on liver slices showed that the NAC + PLZ had not produced significant liver damage, which safety was furtherly confirmed. These findings indicate that NAC suppresses PDAC organoids' metastasis, and combined treatment of NAC with PLZ has cooperative effects on inhibiting tumor metastasis.

Discussion
In this study, we investigated the effects of NAC on PCCs and PSCs and its functional effect on pancreatic cancer-stroma interactions. We found that PCCs efficiently increased the cell viability, migration, and invasion of PSCs. PSCs also increased these abilities of PCCs. Through these tumor-stroma interactions, PCCs showed more rapid growth and higher malignancy. However, NAC strongly suppressed the activities of PSCs, decreased the production of growth factors and reduced cell oxidative stress level at a low concentration rather than those of PCCs and attenuated the protumorigenic effects promoted by cancer-stroma interactions. NAC-treated PSCs were reactivated after exposure to PCC-SN, but co-treatment with NAC and PLZ maintained their quiescent-like state. Thus, NAC enhanced the therapeutic effect of PLZ at low concentrations in PCCs. This is the first report showing that NACinduced quiescent-like PSCs from an activated state was maintained by combined treatment with PLZ. This new combination therapy may induce conversion from protumorigenic PSC to anti-tumorigenic PSCs, which leads to remodeling the tumor microenvironment of pancreatic cancer.
Current studies have focused mainly on suppressing pancreatic cancer desmoplasia or disrupting cancerstroma interactions. Previously, agents such as Pirfenidone and calpeptin have been used to suppress desmoplasia, making PCCs more sensitive to Gemcitabine [16,17]. Moreover, Sujit, et al. combined Pirfenidone and NAC to decrease tumor growth in orthotopic tumor models via inhibition of TGF-β and ROS in PSCs [19]. Thus, antifibrotic and anti-NAF therapeutic strategies appear to have a significant effect in anti-tumor therapy. However, the exact mechanism underlying their combined effect and the functional or state alteration of NAC-treated PSCs were yet to be investigated. In the present study, NAC treatment inhibited activated PSCs and significantly decreased the growth factors and prominent extracellular matrix (ECM) via alterations of multiple pathways. Our data also revealed that NAC not only decreased desmoplasia but also induced a quiescent-like state in PSCs. Quiescent or resting PSCs are star-shaped stellate cells that are activated when needed [34]. After external stimulation such as oxidative stress, ROS, chemokines or growth factors, quiescent fibroblasts can reversibly transfer to normally activated fibroblasts (NAFs), which express α-SMA and Vimentin and present spindle-shaped morphology [34]. These fibroblasts show high expression of α-SMA, which is why they are also called myofibroblasts [35]. However, upon encountering chronic stimuli, activated PSCs gain enhanced proliferative properties that induce a fibrotic tumor microenvironment. Consistent with previous studies, we observed PSCs' morphological changes after NAC treatment and decreased oxidative stress level and expression of α-SMA. Other antioxidants like BHA have also decreased the oxidative stress level in PSCs and decreased expression of α-SMA, thus, these effects were (See figure on previous page.) Fig. 6 co-treatment with NAC and PLZ inhibited subcutaneous tumor growth in the xenografted organoid model with PSC co-implantation. a Representative photomicrographs of cancer organoids derived from KPC mice. Scale bars = 100 μm. b Nude mice (n = 7/per group) were subcutaneously transplanted with KPC mouse-derived organoids (5 × 10 5 ) and mouse PSCs (5 × 10 5 ) into the flank. After 1 week, the mice were injected intraperitoneally with Gemcitabine (40 mg/kg, once a week, also named group2) or NAC (500 mg/kg) + PLZ (4 mg/kg) (twice a week, group3), or Control (100 μl PBS + 0.1 μl DMSO, twice a week, group1) for 4 weeks. Representative photograph showing significantly reduced tumor formation after 4 weeks of NAC + PLZ co-treatment. c Tumor volume was significantly decreased in GEM and NAC + PLZ co-treatment groups after 4 weeks of treatment. d The average weight of each group of mice. Also see supplementary figure S6.E for weight change of each mouse in each group. e H&E staining, Masson's staining trichrome and immunohistochemical staining of α-SMA, CK19, PPARγ, HMOX-1, Nrf2 and PCNA. Original magnification, × 100. Scale bar = 100 μm. *P < 0.05; **P < 0.01; ***P < 0.001; n.s, no significance seemly dependent on intracellular GSH levels, but NAC showed more superiority than BHA. Moreover, we found changes in signaling pathways of NAC-treated PSCs. These data also indicate that NAC altered the activated state of PSCs to a quiescent-like state. However, further study is needed to investigate whether quiescent PSCs induced by NAC are maintained when PSCs encounter continuous stimuli from PCCs in vivo. Fig. 7 Cotreatment with NAC and PLZ decreased liver metastasis in the xenografted organoid model with PSC co-implantation. a The spleens of nude mice were implanted with KPC mouse-derived organoids (1 × 10 5 ) and mouse PSCs (1 × 10 5 ). After 1 week, mice were intravenously injected with vehicle (100 μl PBS + 0.1 μl DMSO, n = 5), NAC (500 mg/kg, n = 5), PLZ (4 mg/kg, n = 5), or their combination (n = 5) once every 2 days for 21 days. On day 28, we sacrificed the mice and evaluated their hepatic metastasis. b Representative photographs showed the number of disseminated visible nodules in the liver, and liver volume was decreased after combined treatment with NAC and PLZ. c-d The metastasis nodules (c) and the tumor weight and volume (d) in mice liver were evaluated. e H&E staining, Masson's staining trichrome and Immunohistochemical staining of α-SMA and CK19 were performed. *P < 0.05; **P < 0.01; ***P < 0.001; n.s, no significance It has been well reported that multiple signaling pathways and molecules are involved in the activation and pro-tumor cell functions of PSCs, including PI3k-AKT, NF-κB, and ERK1/2 signaling pathways [36][37][38][39][40]. The change of these signaling pathways underlying NACtreated PSCs was also investigated in this study. We found that PCCs increased activation of PI3K-AKT and NF-κB signaling pathways in PSCs derived from tumor tissue via cancer-stroma interactions. The present data also revealed that NAC treatment significantly suppressed activated PI3K-AKT and NF-κB pathways, but not of the ERK1/2 signaling pathway in PSCs. Moreover, we found that the expression of PPARγ was increased in PSCs treated with NAC and that co-treatment with NAC and PLZ, a ligand of PPARγ, was significantly inhibited PI3K-AKT and NF-κB pathways, but also activation of ERK1/2 in PSCs. The expression level of PPARγ decreases markedly during activation of hepatic stellate cells [25]. Our knockdown experiments also confirmed that siPPARγ affected the effects of NAC in decreasing the activity of PSCs. These findings suggested when the expression of PPARγ was increased in quiescent-like state PSCs induced by NAC, administration of PPARγ ligand, like PLZ, seemly to be feasible to maintain this state.
Previously, PLZ was often used for type II diabetes but recently identified as a potential therapeutic agent to inhibit proliferation and metastasis of PCCs [41]. However, the treatment of a high concentration of PLZ was reported to increase bladder cancer [42]. In this study, we administrated PLZ with NAC at a safe concentration in co-culture or implantation of PSCs and PCCs and observed promising therapeutic effects. These data suggest that NAC increases sensitivity to PLZ in PCCs, although it cannot inhibit PCCs directly. Thus, co-treatment with NAC and PLZ may exert its therapeutic effects in both PSCs and PCCs.
We found that NAC inhibited most ECM proteins' secretion, but the expression of collagen type IV, one of the major components of the basement membrane (BM), was maintained even in NAC-treated PSCs. BM is a sheet-like structure that separates ductal epithelial cells from the surrounding stroma. PCCs also increase the secretion of various Matrix metalloproteinases (MMP) to decompose the BM structure for their invasion [43]. .The BM primarily consists of collagen IV degraded by extracellular proteases such as MMP2 and MMP9 [44]. We previously reported that PSCs secreted MMP2 and induced BM destruction in the cancer-stromal microenvironment [24]. The present results showed that NAC decreased the expression of MMP2 in PSCs, which possibly maintained the BM structure and impeded PCCs from invading into the stroma. These data indicate that the microenvironment in pancreatic cancer was altered to an anti-tumor microenvironment by NAC treatment.
To clarify the precise mechanism induced by the changes in collagen type IV and MMP-2 expression when PSCs were treated with NAC, further studies will be performed in the future.
Tumor organoid is considered as a brand-new model tool in biomedical research. In a previous study, we demonstrated that an ERK inhibitor reduces the number of metastasis in xenografts formed by tumor organoids [45]. In the present study, we performed in vivo xenograft experiments using KPC mouse-derived cancer organoids representing the microenvironment in PDAC, including PSCs. PSCs co-cultured with organoids from pancreatic cancer-induced excessive desmoplasia compared with co-culture with PCCs, which indicated that this organoid model is of great benefit to investigate the therapeutic effect of combined NAC and PLZ treatment. By this model, the effects of co-treatment therapy were confirmed in both subcutaneous tumor growth and splenic liver metastasis, as well as safety advantages.

Conclusion
In summary, the present data indicated that NAC treatment decreases the activity of PSCs and attenuates cancer-stroma interactions. Moreover, the combination of NAC and PLZ maintains the quiescent-like state of PSCs, which leads to an enhanced therapeutic effect on tumor stromal components. Furthermore, NAC makes PCCs more sensitive to PLZ treatment. Taken together, the present data suggest that combination therapy of NAC and PLZ is a promising treatment that targets both stromal and cancer cells in pancreatic cancer.  Fig. 7 treated with indicated reagents (arrow). (F) Disseminated visible nodules with a diameter larger than 1 mm are counted, NAC decreased the number of visible nodules, and combined treatment of NAC with PLZ further improved therapeutic efficacy. (G) Immunohistochemical staining of CK19 showed the amount of metastatic cell cluster. Original magnification, × 100. Scale bar = 100 μm. *P < 0.05; **P < 0.01; ***P < 0.001; n.s, no significance.