Potential Metabolite Markers for Pancreatic Cancer Identified by Metabolomic Analysis of Induced Cancer-Associated Fibroblasts

Simple Summary Fibroblasts in normal tissues conduct energy metabolism via oxidative phosphorylation (OXPHOS). However, cancer-associated fibroblasts (CAFs) produce energy (i.e., ATP) via glycolysis. Nonetheless, whether intracellular metabolism transitions from OXPHOS to glycolysis when normal tissue fibroblasts differentiate into CAFs remains to be determined. Here, we established an experimental system and induced the in vitro differentiation of mesenchymal stem cells to CAFs and performed detailed metabolomic and RNA sequencing analyses. We found that the intracellular metabolic pathway was reprogrammed to the glycolytic pathway when mesenchymal stem cells were co-cultured with pancreatic cancer cells. Furthermore, we identified CAF-specific metabolites that were expressed post reprogramming. These metabolites have also been observed in pancreatic cancer mouse models, suggesting their potential as cancer biomarkers. Abstract Cancer-associated fibroblasts (CAFs) in the tumor microenvironment perform glycolysis to produce energy, i.e., ATP. Since the origin of CAFs is unidentified, it is not determined whether the intracellular metabolism transitions from oxidative phosphorylation (OXPHOS) to glycolysis when normal tissue fibroblasts differentiate into CAFs. In this study, we established an experimental system and induced the in vitro differentiation of mesenchymal stem cells (MSCs) to CAFs. Additionally, we performed metabolomic and RNA-sequencing analyses before and after differentiation to investigate changes in the intracellular metabolism. Consequently, we discovered that OXPHOS, which was the primary intracellular metabolism in MSCs, was reprogrammed to glycolysis. Furthermore, we analyzed the metabolites in pancreatic tumor tissues in a mice model. The metabolites extracted as candidates in the in vitro experiments were also detected in the in vivo experiments. Thus, we conclude that normal tissue fibroblasts that differentiate into CAFs undergo a metabolic reprogramming from OXPHOS to glycolysis. Moreover, we identified the CAF-specific metabolites expressed during metabolic reprogramming as potential future biomarkers for pancreatic cancer.


Introduction
The most common solid tumors, such as pancreatic cancer, have a large number of stromal cells. Indeed, the stroma forms a thickened extracellular matrix that provides a unique environment, such as a hypoxic environment by inhibiting blood vessel formation penicillin at 37 • C in a humidified atmosphere containing 5% CO 2 . Additionally, the KPC cell line derived from the KPC (LSL-KRAS G12D/+ ; LSL-TP53 R172H/+ ; PDX1-CRE) mice [19] was maintained in DMEM supplemented with 10% FBS and 1% streptomycin-penicillin at 37 • C in a humidified atmosphere containing 5% CO 2 . To establish the KPC cell line isolated from KPC mice, pancreatic tumor was minced, and two or three 2 mm tumor pieces were plated onto a dish in the medium. The dishes were then incubated under 5% CO 2 , 20% O 2 , at 37 • C. After a couple days of incubation, cancer cells grew around the tumor fragments, and the tumor fragments were removed. When a sufficient number of cells were observed, the cells were passaged or stored. Under these culture conditions, cancer cells grew selectively, while the other cells including CAFs were depleted after a few passages. We confirmed that these cells contained only KPC cells and no other cells by flowcytometry before further analysis. Subsequently, the AD-MSCs and Capan-1 cells were labeled with green fluorescent protein (GFP) and red fluorescent protein (RFP) by lentiviral transduction, respectively. For this purpose, we seeded 293LTV cells (LTV-100; Cell Billabs, Inc., San Diego, CA, USA) into a 6-well plate and co-transfected them with pLKO.1puro eGFP (Sigma-Aldrich, St. Louis, MO, USA), psPAX2 was a gift from Didier Trono (Addgene plasmid # 12260; http://n2t.net/addgene:12260, accessed on 17 January 2022; RRID:Addgene_12260), and pMD2.G was a gift from Didier Trono (Addgene plasmid # 12259; http://n2t.net/addgene:12259, accessed on 17 January 2022; RRID:Addgene_12259) at a ratio of 2:1.5:1.2 µg/well. Post 24 h, the viral supernatant was collected and filtered through 0.45 µm membranes. Thereafter, AD-MSCs were infected with this filtrate in the presence of polybrene (10 µg/mL) for 24 h. Thereafter, puromycin was added to the medium for the selection of GFP-positive cells. Once we identified AD-MSCs that expressed GFPs, we confirmed that these cells still retained their original characteristics, such as auto-differentiation, senescence, or weak stemness.

In Vitro Co-Culture Assay
Co-culturing of AD-MSCs and Capan-1 cells was performed as described previously [18]. In this regard, AD-MSCs and Capan-1 cells were co-cultured under two different conditions: direct and indirect transwell co-culture. In the direct co-culture method, 4 × 10 5 AD-MSCs and 4 × 10 5 Capan-1 cells were mixed and seeded in 6-well culture plates. In contrast, in indirect transwell co-culture, 4 × 10 5 AD-MSCs were seeded in the lower compartment of the transwell membrane, whereas 4 × 10 5 Capan-1 cells were seeded in the upper compartment (Falcon Permeable Support for 6-well plates with 3.0 µm translucent high-density PET membrane #353092; Corning Inc., Corning, NY, USA).

In Vitro Macropinocytosis Assay
We performed an in vitro macropinocytosis assay as previously described [20]. Briefly, 1 × 10 5 cells were plated onto glass coverslips in 6-well plates for five days. We then incubated the cells for 30 min at 37 • C with 70-kDa fluoresceine isothiocyanate (FITC)dextran (Sigma) directly added to the culture media at a final concentration of 1 mg/mL. Subsequently, we assessed the macropinocytic uptake of cells, and rinsed the cells five times on ice with ice-cold PBS. Thereafter, the cells were fixed with 3.7% formaldehyde and their nuclei were counterstained with DAPI. Eventually, the coverslips were mounted onto glass slides using an aqua-poly/mount (Polysciences, Inc., Warrington, PA, USA). Fluorescent images were captured using a fluorescence microscope (BZ-710; Keyence). Notably, each experimental condition was performed in triplicates.

Mouse Models and In Vivo Experiments
We purchased eight-week-old female C57BL/6J wild mice from Japan CLEA Inc. (CLEA Japan, Tokyo, Japan) for our experiments. We bred and housed the mice under specific pathogen-free conditions at the Animal Center of AIST and the University of Tsukuba. We developed a xenograft model by subcutaneously transplanting 2 × 10 6 KPC cells into the mice (KPC xenograft model). Thereafter, we sacrificed the mice on day 28 and excised all subcutaneous tumors.
All invasive procedures were performed under inhalation anesthesia using isoflurane. Mice were euthanized by cervical dislocation following inhalation of the anesthesia. All animal experiments and procedures were approved by and performed in compliance with the guidelines of Institutional Animal Care and Use Committee of the respective institutes of AIST (A2020-310) and the Ethics Committee of the University of Tsukuba (19)(20)(21)(22)(23)(24)(25)(26)(27)(28). The study was conducted in accordance with the Animal Research Reporting in vivo Experiments (ARRIVE) guidelines [21].

Immunohistochemical Tissue Staining
All the staining protocols were performed on 2 µm thick mouse tissue sections. Hematoxylin and eosin (HE) and Masson's trichrome (MT) staining were performed according to standard protocols [10]. We performed immunohistochemistry (IHC) by first deparaffinizing the sections, following which we performed antigen retrieval at 121 • C in an autoclave for 10 min in a 10 mM sodium citrate buffer (pH 6.0). We then treated the sections with a 3% H 2 O 2 solution (Envision Plus System; Dako, Santa Clara, CA, USA) to inhibit any endogenous peroxidases. Rabbit polyclonal LIF antibody (1:500, ab113262; Abcam) was used for IHC. The labeled antigens were visualized by chromogen 3,30-diaminobenzidine tetrahydrochloride; hematoxylin was used as a nuclear counterstain. Eventually, the slides were observed under a fluorescence microscope (BZ-710; Keyence).

Metabolomic Analysis
Monocultured and co-cultured Capan-1 cells and AD-MSCs (each group n = 3) and KPC xenograft mice tumors (n = 3) were analyzed by metabolomics (Human Metabolome Technologies (HMT) Inc., Tsuruoka, Japan) [22][23][24]. Firstly, frozen cells or mouse tumor samples were transferred into 500 µL of methanol containing 50 mM of an external standard. Thereafter, we homogenized the cells five times at 200× g for 120 s by BMSM10N21 (BMS, Tokyo, Japan). Following this, we added and mixed 500 µL of chloroform and 200 µL of ultrapure water to the homogenate, and subsequently centrifuged the mixture at 2300× g for 5 min at 4 • C. The resultant aqueous phase was subjected to ultrafiltration using a Millipore Ultrafree-MC PLHCC HMT Centrifugal Filter Device, 5 kDa (Millipore, Billerica, MA, USA). The filtrates were then dried and dissolved in 50 µL of ultrapure water. Subsequently, the samples were subjected to capillary electrophoresis time-of-flight mass spectrometry (CE-TOFMS) in the Agilent CE-TOFMS system (Agilent Technologies, Santa Clara, CA, USA) at 4 • C. The detected peaks were aligned according to the m/z values and normalized migration times. Furthermore, the peaks were mean-centered and scaled using their standard deviations on a per-peak basis as a pretreatment. Following the application of autoscaling, we conducted principal component analysis (PCA) and hierarchical clustering analysis using SampleStat v3.14 (HMT Inc., Tsuruoka, Japan) and PeakStat v3.18 (HMT Inc.). In the PCA, score plots of the first and second principal components were generated. Additionally, we generated heat maps by coloring the data values across their value ranges. The relative area of each peak was calculated and used for comparison among the four groups.

RNA-Sequencing (RNA-Seq)
We evaluated the difference in the gene expression levels of AD-MSCs and AD-CAFs by performing RNA-seq as previously reported [10,18]. Briefly, total RNA was extracted from the cells using the TRI Reagent (Molecular Research Center, Inc., Cincinnati, OH, USA). Thereafter, library preparation and sequencing were conducted using the Truseq library prep kit and NovaSeq 6000 (Illumina, San Diego, CA, USA). The data acquired from two biological replicates for each cell type were analyzed using STAR (2.7.1a, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA) [25], RSEM (1.3.1, University of Wisconsin-Madison, Madison, WI, USA) [26], and edgeR (3.30.3, Walter and Eliza Hall Institute of Medical Research, Victoria, Australia) [27]. Subsequently, we obtained the normalized counts (trimmed mean of M values) and identified the differentially expressed genes that satisfied the condition |log2 (fold-change)| ≥ 1 and false discovery rate < 0.05. Raw sequences in the FASTQ format were deposited at the DNA Data Bank of Japan (DDBJ; accession numbers DRR231745-DRR231748).

Statistical Analysis
Data are represented as the mean ± standard deviation (SD) unless otherwise noted. We analyzed the data among groups of three or more groups by one-way analysis of variance followed by post hoc Tukey tests with two-tailed distribution. Furthermore, the student's t-test was used to compare data between the control and experimental groups. Statistical significance was set at p < 0.05. All calculations were performed using the Graph-Pad Prism software or EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), which is a graphical user interface of the R software (R Foundation for Statistical Computing, Vienna, Austria) [28]. Of note, EZR is a modified version of R Commander designed for statistical functions that are frequently used in biostatistics. Additionally, the SD of the data are represented as error bars in the figures.

AD-MSCs Co-Cultured with Capan-1 Cells Are CAF-Progenitors That Are Capable of Reacting with Cancer Cells
We generated CAFs derived from AD-MSCs (AD-CAFs) by two methods: direct contact and indirect non-contact transwell co-culture with the Capan-1 cell line. Our qPCR analysis confirmed that these two methods generated AD-CAFs by two distinct mechanisms of differentiation ( Figure 1a). For instance, AD-CAFs generated by contact coculture exhibited increased expression of the myoblast markers ACTA2 and CTGF. On the other hand, AD-CAFs generated by the non-contact co-culture had upregulated expression of cytokine-related genes, such as CXCL1, IL6, and LIF ( Figure S1). These alterations were verified by immunostaining ( Figure 1b). Moreover, the αSMA protein was more prominently expressed in the direct co-culture than in the non-contact co-culture. Notably, IL6-positive cells were rarely observed in this approach. In contrast, the number of IL6positive cells was greatly increased and αSMA-positive cells were not observed when cells were cultured via the non-contact transwell co-culture. However, vimentin, a fibroblast marker, had uniform expression in both the co-cultures, confirming that the analyzed cells were indeed CAFs. Next, we examined whether co-culturing GFP-transfected AD-MSCs with RFP-transfected Capan-1 cells resulted in cell-cell interactions. Indeed, we observed vesicles after three days of contact co-culture (Figure 1c), and the presence of green vesicles in Capan-1 cells or red vesicles in CAFs was confirmed under a strong magnification ( Figure 1d). Furthermore, we observed the presence of several FITC-positive vesicles when dextran-FITC was added to the co-culture media ( Figure 1e). These observations verified that AD-MSCs differentiate into AD-CAFs when co-cultured with Capan-1 cells. Therefore, we used AD-MSCs as progenitors of CAFs for subsequent metabolomic analysis. Notably, we used AD-MSCs maintained in non-contact co-cultures in consideration of the time required for sample collection and to prevent contamination between AD-MSCs and Capan-1 cells.

AD-CAFs Undergo a More Drastic Metabolic Transformation Than Capan-1, as Discovered by a Global Metabolomic Analysis
We performed metabolomic analyses of AD-MSCs (progenitors of CAFs), AD-CAFs, and Capan-1 cells before and after co-culture by CE-TOFMS (Figure 2a). The results of our PCA are presented in Figure 2a. Remarkably, we observed that the metabolite profile of AD-MSCs was significantly different from AD-CAFs. The metabolites involved in the PC2axis in Figure 2a, and their contribution rates are represented in Figure 2b. Interestingly, these include metabolites that are involved in polyamine metabolism, but not those directly involved in major intracellular metabolic pathways, such as glycolysis and OXPHOS. Since these identified metabolites have not been well-studied, they are possible novel biomarkers of CAF-abundant tumors; we have discussed this matter in detail in later sections. On the other hand, lesser metabolite changes were observed in Capan-1 than in CAF. A heat map of the metabolite changes in each cell type is shown in Figure 2c. Overall, metabolites that were abundant in Capan-1 cells were scarce in CAFs, and conversely, metabolites that were abundant in CAFs were scarce in Capan-1 cells. This held true for metabolites that were highly abundant in Capan-1 cells both before and after co-culturing but were almost absent in CAFs both before and after co-culturing. These results indicate that each cell type had its own specific metabolite profile. Remarkably, only 36 metabolites exhibited altered abundance in Capan-1 cells; while 12 (33%) were increased in number, 24 (24%) were decreased. In contrast, 86 metabolites were altered in CAFs, of which 67 (78%) were downregulated. Thus, the metabolite changes observed in CAFs were larger than those observed in Capan-1 cells; notably, these changes indicated a decrease in abundance of the metabolites.

AD-CAFs Undergo a More Drastic Metabolic Transformation Than Capan-1, as Discovered by a Global Metabolomic Analysis
We performed metabolomic analyses of AD-MSCs (progenitors of CAFs), AD-CAFs, and Capan-1 cells before and after co-culture by CE-TOFMS (Figure 2a). The results of our PCA are presented in Figure 2a. Remarkably, we observed that the metabolite profile of indicate that each cell type had its own specific metabolite profile. Remarkably, only 36 metabolites exhibited altered abundance in Capan-1 cells; while 12 (33%) were increased in number, 24 (24%) were decreased. In contrast, 86 metabolites were altered in CAFs, of which 67 (78%) were downregulated. Thus, the metabolite changes observed in CAFs were larger than those observed in Capan-1 cells; notably, these changes indicated a decrease in abundance of the metabolites.

AD-CAFs Exhibit Upregulated Glycolytic Metabolism
We examined the changes in intracellular metabolism by comparing the metabolite profiles between AD-CAFs and AD-MSCs. Figure 3a depicts changes in the glycolytic metabolites, including glucose. While the levels of glucose 6-phosphate were substantially

AD-CAFs Exhibit Upregulated Glycolytic Metabolism
We examined the changes in intracellular metabolism by comparing the metabolite profiles between AD-CAFs and AD-MSCs. Figure 3a depicts changes in the glycolytic metabolites, including glucose. While the levels of glucose 6-phosphate were substantially decreased upon differentiation of AD-MSCs to AD-CAFs, those of fructose 1,6-phosphate and phosphoenolpyruvic acid were increased by 1.9 and 1.4 folds, respectively. In addition, the aerobic glycolysis metabolite lactate was significantly increased (1.5 fold) in AD-CAFs, indicating that these cells undergo a metabolic shift towards glycolysis. We further validated this metabolic shift by analyzing altered gene expression levels by RNA-seq. The list of genes involved in glycolysis and their expression levels are displayed in Figure 3b. Consequently, we identified HK2, PFKL, PFKP, ALDOC, GAPDH, PGK1, ENO1, ENO2, and LDHA as differentially expressed genes (i.e., upregulated or downregulated gene expression). However, our pathway analysis did not reveal any significant changes, and the metabolic shift to glycolysis was not supported by our gene expression analysis. validated this metabolic shift by analyzing altered gene expression levels by RNA-seq. The list of genes involved in glycolysis and their expression levels are displayed in Figure  3b. Consequently, we identified HK2, PFKL, PFKP, ALDOC, GAPDH, PGK1, ENO1, ENO2, and LDHA as differentially expressed genes (i.e., upregulated or downregulated gene expression). However, our pathway analysis did not reveal any significant changes, and the metabolic shift to glycolysis was not supported by our gene expression analysis.

OXPHOS Is Downregulated in AD-CAFs
Changes in the metabolites associated with the tricarboxylic acid (TCA) cycle are shown in Figure 4a. In this cycle, acetyl-CoA is synthesized from pyruvate and is converted to citric acid, and ATP is produced during reactions involving cis-aconitic acid, 2-oxoglutaric acid, and malic acid. Notably, acetyl-CoA (−7.4 fold), citric acid (−1.5 fold), and malic acid (−1.3 fold) were observed to be considerably downregulated in AD-CAFs. Indeed, we determined that downregulation of OXPHOS was associated with upregulation of glycolysis in AD-CAFs. The list of genes involved in OXPHOS and their expression levels are presented in Figure 4b. Remarkably, we observed no significant changes in the mRNA expression levels of these genes. Thus, the metabolic shift towards glycolysis in the AD-CAFs was not verified by changes in the gene expression levels. and malic acid (−1.3 fold) were observed to be considerably downregulated in AD-CAFs. Indeed, we determined that downregulation of OXPHOS was associated with upregulation of glycolysis in AD-CAFs. The list of genes involved in OXPHOS and their expression levels are presented in Figure 4b. Remarkably, we observed no significant changes in the mRNA expression levels of these genes. Thus, the metabolic shift towards glycolysis in the AD-CAFs was not verified by changes in the gene expression levels.

Polyamine Metabolism Is Altered in AD-CAFs
Polyamines are involved in various physiological functions, such as cell division and proliferation, and nucleic acid and protein synthesis; studies have also determined their roles in cancer cells [29][30][31]. Since we identified several polyamine metabolites in the PC2 axis in the PCA analysis (Figure 1), we also analyzed the polyamine synthesis pathway. Figure 5a presents changes in ornithine synthesized from arginine and those observed in metabolites involved in the polyamine synthesis pathway. We observed that ornithine, the starting point of this pathway, was significantly downregulated by 1.3-fold in the AD-CAFs.
Moreover, while putrescine, N8-acetylspermidine, N-acetylputrescine, and 5 -deoxy-5methylthioadenosine were not detected in AD-MSCs, they were detected in AD-CAFs. A list of genes involved in the polyamine synthesis pathway and their expression levels are listed in Figure 5b. Since we observed no significant changes in the gene expression levels, the metabolic shift towards glycolysis was not supported by our gene expression analysis.
proliferation, and nucleic acid and protein synthesis; studies have also determined their roles in cancer cells [29][30][31]. Since we identified several polyamine metabolites in the PC2 axis in the PCA analysis (Figure 1), we also analyzed the polyamine synthesis pathway. Figure 5a presents changes in ornithine synthesized from arginine and those observed in metabolites involved in the polyamine synthesis pathway. We observed that ornithine, the starting point of this pathway, was significantly downregulated by 1.3-fold in the AD-CAFs. Moreover, while putrescine, N8-acetylspermidine, N-acetylputrescine, and 5′-deoxy-5′-methylthioadenosine were not detected in AD-MSCs, they were detected in AD-CAFs. A list of genes involved in the polyamine synthesis pathway and their expression levels are listed in Figure 5b. Since we observed no significant changes in the gene expression levels, the metabolic shift towards glycolysis was not supported by our gene expression analysis.

Certain Metabolites Are Uniquely Present in AD-CAFs
Spermidine is acetylated at the N8 position by N-acetyltransferase in the cell nucleus. Subsequently, it is transported to the cytoplasm, where it is deacetylated by metaldependent N8-acetylspermidine deacetylase, also called polyamine deacetylase, and N8acetylspermidine is synthesized [32,33]. Alterations in the levels of N8-acetylspermidine in all the samples, including Capan-1 cells, are presented in Figure 6a. Interestingly, we detected this metabolite only in AD-CAFs. Therefore, we deduced that it is possibly a specific biomarker for CAF-rich cancers. On the other hand, the metabolites that were not detected in CAF progenitors were detected in AD-CAF, although they were not changed in Capan-1 before and after co-culture, as shown in Figure 6b. One such metabolite was the polyamine putrescine. Indeed, putrescine was detected in both AD-CAF and Capan-1. However, given that tumors in vivo are a mixture of cancer cells and CAFs, and the total amount of putrescine was increased after co-culture, putrescine may be a specific biomarker. Similarly, N-acetylcysteine, O-succinylhomoserine, and butyrylcarnitine could be regarded as specific metabolites of pancreatic cancer, as they were detected specifically in AD-CAFs.
Subsequently, it is transported to the cytoplasm, where it is deacetylated by metal-dependent N8-acetylspermidine deacetylase, also called polyamine deacetylase, and N8-acetylspermidine is synthesized [32,33]. Alterations in the levels of N8-acetylspermidine in all the samples, including Capan-1 cells, are presented in Figure 6a. Interestingly, we detected this metabolite only in AD-CAFs. Therefore, we deduced that it is possibly a specific biomarker for CAF-rich cancers. On the other hand, the metabolites that were not detected in CAF progenitors were detected in AD-CAF, although they were not changed in Capan-1 before and after co-culture, as shown in Figure 6b. One such metabolite was the polyamine putrescine. Indeed, putrescine was detected in both AD-CAF and Capan-1. However, given that tumors in vivo are a mixture of cancer cells and CAFs, and the total amount of putrescine was increased after co-culture, putrescine may be a specific biomarker. Similarly, N-acetylcysteine, O-succinylhomoserine, and butyrylcarnitine could be regarded as specific metabolites of pancreatic cancer, as they were detected specifically in AD-CAFs.

Metabolites Unique to AD-CAFs Are Also Expressed in Mouse Pancreatic Cancer Models
Next, we examined whether these AD-CAF-specific metabolites could also be detected in mouse tumor tissues (Figure 7a). Figure 7b displays the tumor tissue section. In this mouse model, mouse cells were spontaneously induced to differentiate into CAFs, as confirmed by the MT staining. This model also contained a population of LIF-positive CAFs. Tumor samples were collected from three independent mice, and metabolomic analyses were performed (Figure 7c,d). N8-acetylspermidine was indeed detected, as in the in vivo analysis, and other metabolites were also detected in living tumor tissues; however, some of these metabolites were close to the detection limit. this mouse model, mouse cells were spontaneously induced to differentiate into CAFs, as confirmed by the MT staining. This model also contained a population of LIF-positive CAFs. Tumor samples were collected from three independent mice, and metabolomic analyses were performed (Figure 7c,d). N8-acetylspermidine was indeed detected, as in the in vivo analysis, and other metabolites were also detected in living tumor tissues; however, some of these metabolites were close to the detection limit.

Metabolic Transformation in CAFs
In this study, we used an original experimental approach to induce cell differentiation into CAFs in vitro and performed metabolomic analysis to distinguish cancer cells from CAFs. Since the TME contains CAFs in addition with various cell types, such as blood or immune cells, it is difficult to understand the metabolic changes that occur in each cell type. However, in this study, we accurately determined the metabolic transformations that occur in CAFs. Previously, molecular analyses have revealed changes in metabolic genes expressed in CAFs [34][35][36]. These results suggested that CAFs within tumors have an active glycolytic metabolism. In this study, we verified AD-MSCs as the progenitors of CAFs. Furthermore, we compared AD-MSCs and AD-CAFs to precisely elucidate the metabolic changes that occur during differentiation of AD-MSCs into CAFs. It has been shown that CAFs differentiate from MSCs and pancreatic stellate cells (PSCs) by interaction with cancer cells [4,7,8]. Koikawa et al. collected adipose tissue from mice and co-cultured it with mouse pancreatic cancer [37]. Ohlund et al. also co-cultured mouse-or human-derived PSCs with mouse or human pancreatic cancer-derived organoids [7]. On the other hand, we utilized immortalized AD-MSC cell line and pancreatic cancer cell line, and these cell lines could recapitulate CAF heterogeneity and its important functions. Proliferation of cells and tissues derived from living organisms and their properties tend to be unstable; therefore, CAFs derived from such sources are also unstable and challenging to replicate. However, the use of immortalized cells allowed CAFs to maintain their heterogeneity and important functions in the tumor [10,18], making the investigation of CAF-specific metabolites possible in this study. The cancer cell line used in this study was a human pancreatic cancer cell line. However, the robustness and homology of this data needs to be verified in various types of cancer cells, including those derived from various solid tumors, such as gastric cancer and colon cancer. In addition, it is necessary to validate the results of this study using 3D in vitro assays that consider the presence of ECM components that are known to be crucial players in CAF biology.

Polyamine Metabolism in CAFs
Putrescine and N8-acetylspermidine were detected as metabolites specific to CAFs. Putrescine is synthesized from ornithine and is converted to spermidine. In turn, spermidine is converted to N8-acetylspermidine by the histone acetyltransferase P/CAF (KAT2B) in the presence of acetyl CoA. Since the synthesized N8-acetylspermidine is deacetylated by polyamine deacetylase, decreased activity of this enzyme causes accumulation of N8-acetylspermidine in CAFs. However, owing to the high substrate specificity of polyamine deacetylase, it does not deacetylate other acetylspermidines, such as cytoplasmic N1-acetylspermidine or N1-acetylspermine [38]. In addition, selective inhibition of polyamine deacetylase activity in HeLa cells has been demonstrated to increase N8-acetylspermidine levels but not acetylated histone levels. Additionally, polyamine deacetylase has a different function than histone diacetylase (HDAC) [39]. Recently, studies have reported that two Zn 2+ -dependent HDACs, HDAC6, and HDAC10, function as polyamine deacetylases [40,41]. Therefore, we also examined the expression of HDAC6 and HDAC10 in AD-MSCs and AD-CAFs but found no changes in their mRNA levels. However, to investigate whether N8-acetylspermidine accumulates in CAFs because of decreased activity of polyamine deacetylase, it is necessary to measure the amount and activity of the enzyme products. Notably, putrescine was upregulated in AD-CAFs, suggesting that N8-acetylspermidine may have been synthesized more by P/CAF via spermidine in the CAFs than progenitors. However, we observed no changes in the mRNA levels of KAT2B or HDAC. Further studies are needed to determine the pathway of N8acetylspermidine synthesis and metabolism by performing detailed enzyme activity and metabolite flux analyses.

N8-Acetylspermidine Is a Potential Biomarker of CAF
N8-acetylspermidine was the only AD-CAF-specific metabolite identified in this study. Although the source of N8-acetylspermidine in tumors cannot be determined in vivo by metabolomic analysis, in this study, we speculate N8-acetylspermidine to be CAF-derived. The presence of this metabolite needs to be verified by metabolite analysis of KPC cells. However, since this metabolite was not detected in vitro in human pancreatic cancerderived cells such as Capan-1 cells, its level may be an indicator of the amount of CAF present in the tumor. Furthermore, we detected this metabolite in a pancreatic cancer mouse model. Since N8-acetylspermidine was not detected in Capan-1 cells, we assumed that N8-acetylspermidine is only biosynthesized in CAFs in vivo. Determination of the source of N8-acetylspermidine in tumors in vivo is challenging for the following reasons. First, there is no antibody available against N8-acetylspermidine. Second, accurate metabolite analysis is difficult in cells in vivo or in direct co-culture because metabolites change with the time required for FACS sorting; identification of CAF subtypes that release N8-acetylspermidine may facilitate identification of the source. Remarkably, N8-acetylspermidine has been detected more frequently in patients with early-stage pancreatic cancer than in that in healthy subjects, and thus is likely to be a promising CAF-specific metabolite [42]. The other metabolites presented in Figure 6 are either biosynthesized in CAFs or taken up by CAFs from Capan-1 cells. We observed macropinocytosis and secretion of extracellular vesicles, such as exosomes, in the TME. Given that putrescine could be synthesized by both Capan-1 and CAF, we presume that putrescine detected in CAFs was either synthesized in CAFs or transported to CAFs from Capan-1 through these intercellular transport mechanisms. At present, while the relationship between micropinocytosis and putrescine is not clear, it would be beneficial to perform block micropinocytosis using pharmacological inhibitors and investigate the impact on putrescine production in the future.
Cancer cells and CAFs generally grow under hypoxia and hypotrophic conditions, while the in vitro experiments were conducted in normoxia. However, the same metabolites were detected as in the in vivo experiments, which is considered to be a hypoxic environment. Therefore, the in vitro experiments may have mimicked some localized hypoxic environments. In the future, it will be necessary to examine the metabolites in more detail by reproducing the hypoxic environment in vitro.

Conclusions
In this study, we used a simple in vitro experimental system to analyze metabolites with high accuracy and identify several factors that could be potential biomarkers of pancreatic cancer. Although the CAF types produced in this simple experimental system may be limited, the potential biomarkers identified in this study are expected to prove to be important in pancreatic cancer in vivo.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/cancers14061375/s1, Figure S1: Global gene expression pattern showed that AD-MSCs were differentiated into CAF.  Institutional Review Board Statement: All animal experiments and procedures were approved by the Institutional Animal Care and Use Committee of the respective National Institute of Advanced Industrial Science and Technology (A2020-310) and the Ethics Committee of the University of Tsukuba (19)(20)(21)(22)(23)(24)(25)(26)(27)(28), and were carried out in accordance with the approved guidelines. The study was conducted in compliance with the Animal Research Reporting in vivo Experiments (ARRIVE) guidelines [20].

Informed Consent Statement: Not applicable.
Data Availability Statement: RNA sequencing data is available at the DDBJ Read Archive under accession number DRR231745-DRR231748. Raw data are available from the corresponding author upon request.