Site-Dependent Lineage Preference of Adipose Stem Cells.

Adult stem cells have unique properties in both proliferation and differentiation preference. In this study, we hypothesized that adipose stem cells have a depot-dependent lineage preference. Four rabbits were used to provide donor-matched adipose stem cells from either subcutaneous adipose tissue (ScAT) or infrapatellar fat pad (IPFP). Proliferation and multi-lineage differentiation were evaluated in adipose stem cells from donor-matched ScAT and IPFP. RNA sequencing (RNA-seq) and proteomics were conducted to uncover potential molecular discrepancy in adipose stem cells and their corresponding matrix microenvironments. We found that stem cells from ScAT exhibited significantly higher proliferation and adipogenic capacity compared to those from donor-matched IPFP while stem cells from IPFP displayed significantly higher chondrogenic potential compared to those from donor-matched ScAT. Our findings are strongly endorsed by supportive data from transcriptome and proteomics analyses, indicating a site-dependent lineage preference of adipose stem cells.


INTRODUCTION
Adult stem cells, found in varied tissues such as bone marrow, adipose, and synovium, have the capacity for proliferation and differentiation and are important cell sources for clinical therapy (Brown et al., 2019). However, adult stem cells from different sources have different capacities for proliferation and tissue-specific differentiation (Pizzute et al., 2015). This discrepancy even exists in a single tissue, for example, adipose tissue. Recent evidence indicates that, compared with subcutaneous adipose tissue (ScAT), infrapatellar fat pad (IPFP) is a unique adipose tissue with some distinct features (Sun et al., 2018). ScAT, a fat pad beneath the skin, is becoming an attractive stem cell source for tissue regeneration due to its ease of harvest in large quantities with a minimally invasive procedure (Chu et al., 2019). IPFP, located inside the knee joint and outside the synovial membrane, is becoming a promising stem cell source for cartilage engineering and regeneration (Sun et al., 2018).
In osteoarthritis, IPFP exhibits both anti-inflammatory and pro-inflammatory characteristics (Alegre-Aguarón et al., 2012). Increasing evidence indicates the superiority of IPFP-derived stem cells (IPFSCs) over subcutaneous adipose stem cells (ScASCs) in chondrogenic differentiation capacity (Alegre-Aguarón et al., 2012;Lopa et al., 2014); however, most reports are not donor-matched studies, leading to an uncertain conclusion unless large-scale samples are used. In this study, donormatched IPFSCs and ScASCs isolated from four New Zealand white rabbits were used for comparison in cell proliferation and tri-lineage differentiation capacities. We hypothesized that adipose stem cells have a depot-dependent lineage preference. Transcriptome and proteome analyses were used to facilitate an in-depth understanding of site-dependent variations of lineage preference.

Cell Culture and Proliferation and Surface Phenotype Evaluation
Donor-matched inguinal ScAT and IPFP were obtained from four 4-month-old male New Zealand white rabbits. This project was approved by our Institutional Animal Care and Use Committee (IACUC). The harvested tissues were finely minced and digested using 0.1% collagenase P (Roche, Indianapolis, IN, United States) for 90 min at 37 • C in a shaking water bath. The collected cells were cultured in growth medium containing alpha-minimum essential medium (αMEM), 10% fetal bovine serum, 100 U/mL penicillin, 100 µg/mL streptomycin, and 0.25 µg/mL fungizone (Invitrogen, Carlsbad, CA, United States) in a 37 • C, 5% CO 2 , humidified incubator. After reaching 80% confluence, ScASCs and IPFSCs were detached and counted (n = 4) using a hemocytometer, then seeded on T175 flasks at a density of 3000/cm 2 for expansion. The cells were counted for three passages after harvesting. Cell population doubling time (PD time) was then calculated based on the formula "PD time = T * log (2) / (log (N 1 ) -log (N 0 ))". T stands for incubation time, N 1 stands for harvesting cell number, and N 0 stands for seeding cell number.
Passage 1 ScASCs and IPFSCs from four donors were used to perform EdU (5-ethynyl-2 -deoxyuridine) cell proliferation assay (cat no. C1024; Invitrogen) according to manufacturer's protocols. Briefly, the cells were seeded at a density of 3000/cm 2 , grew to 40% confluence, and then were incubated with 10 µM EdU solution for 3 h before detaching, fixing, permeabilizing, and staining the cells. EdU fluorescence was detected and analyzed by FACSCalibur (BD Biosciences, San Jose, CA, United States) using FCS Express 5 software (De Novo Software, Los Angeles, CA, United States).
A CD146 assay was performed on passage 1 ScASCs and IPFSCs from four donors. Briefly, 3 × 10 5 cells (n = 4) were incubated with CD146 antibody conjugated with phycoerythrin (cat no. 12-1469-42; Thermo Fisher Scientific, Milford, MA, United States) in darkness for 30 min. The fluorescence was analyzed by FACSCalibur using FCS Express 5 software (De Novo Software, Los Angeles, CA, United States).

Stemness and Senescence Gene Expression
Expanded cells were evaluated using real-time quantitative polymerase chain reaction (qPCR) for differences between ScASCs and IPFSCs in stemness and senescence-related gene expression. Total RNA was extracted from passage 1 cells of all four donors using TRIzol (Invitrogen). Then cDNA was synthesized from mRNA by reverse transcriptase using a High-Capacity cDNA Archive Kit (Thermo Fisher Scientific). Primers of stemness-related genes, senescence-related genes, and an endogenous reference gene (Table 1) were customized from Integrated DNA Technologies (IDT, Coralville, IA, United States) as Sybr green gene expression assay using their qPCR tool. qPCR was performed with iCycler iQ TM Multicolor RT-PCR Detection and calculated by computer software (PerkinElmer, Wellesley, MA, United States).
Total RNA from ScASCs and IPFSCs (n = 4) was extracted from pellets using an RNase-free pestle in TRIzol, and other Reverse: GGGCACCAGCATCACCCCAC qPCR procedures were the same as described above. Primers for chondrogenic-related genes and the endogenous control gene (Table 1) were customized as TaqMan R gene expression assay from Applied Biosystems (Foster City, CA, United States). ACAN primer was customized as Sybr green gene expression assay from IDT (Table 1).

Adipogenic Induction and Evaluation
When passage 2 cells grew to 90% confluence, they were cultured in adipogenic medium consisting of growth medium supplemented with 1 µM dexamethasone, 10 µM insulin (Biovendor, Asheville, NC, United States), 0.5 mM 3-isobutyl-1-methylxanthine, and 200 µM indomethacin in a 37 • C, 5% CO 2 , humidified incubator for as long as 21 days. Primers for qPCR analysis (n = 4) of adipogenic-related genes ( Table 1) were customized as TaqMan R gene expression assay from Applied Biosystems. LPL primer was customized as Sybr green gene expression assay from IDT ( Table 1). Adipogenic induced cell samples were dissolved in lysis buffer with protease inhibitors. Total protein was quantified by a NanoDrop Spectrophotometer. Twenty micrograms of proteins (n = 4) were separated on a 12% polyacrylamide gel and transferred to nitrocellulose membrane (30 V, overnight at 4 • C). Membranes were blocked with 5% non-fat milk in TBS (Tris-buffered saline) and probed with primary adiponectin (ADIPOQ) monoclonal antibody (cat no. MA1-054; Thermo Fisher Scientific) followed by secondary antibody conjugated with HRP (horseradish peroxidase) (cat no. RK244131; Thermo Fisher Scientific). Next, a chemiluminescence kit (GE Healthcare, Chicago, IL, United States) was used for developing; the blot was imaged using a GE gel documentation. Anti-β actin antibody (Invitrogen) was used to normalize the loading amounts.

Osteogenic Induction and Evaluation
When passage 2 cells grew to 90% confluence, they were cultured with osteogenic medium consisting of growth medium supplemented with 0.01 µM dexamethasone, 50 mg/L ascorbic acid-2-phosphate, and 10 mM β-glycerophosphate in a 37 • C, 5% CO 2 , humidified incubator for as long as 21 days. Primers for qPCR analysis (n = 4) of osteogenic-related genes ( Table 1) were customized as TaqMan R gene expression assays from Applied Biosystems.
Protein extracts were assayed for osteocalcin using Osteocalcin (OCN) assay kit (cat no. MBS2019743; MyBioSource, San Diego, CA, United States) following manufacturer's instructions. Briefly, protein samples (n = 4) were added to wells with pre-coated osteocalcin antibody and incubated at room temperature for 1 h. Upon incubation, 100 uL of detection reagent A was added to the microwells and incubated at 37 • C for 1 h. Microwells were washed and incubated with 100 µL of detection reagent B at 37 • C for 30 min. Upon washing, 90 µL of 3,3' ,5,5'-tetramethylbenzidine substrate solution was added and the sample was incubated at 37 • C for 30 min. The reaction was stopped by the addition of 50 µL of stop solution. This enzyme-linked immunosorbent assay (ELISA) was quantitatively measured using a microplate reader at 450 nm.

RNA-Seq Analysis
The total RNA samples collected from donor-matched passage 1 IPFSCs (n = 4) and ScASCs (n = 4) underwent an initial quality control (QC) check. When samples were deemed high quality [RIN (RNA Integrity Number) > 8], a Next Generation Library was built followed by a final QC check, which involved Qubit quantification and size estimation on the Agilent Bioanalyzer using a High Sensitivity DNA chip. The completed libraries were sequenced on their HiSeq 2500 (Illumina, Inc., San Diego, CA, United States). FastQC 1 and multiQC (Ewels et al., 2016) were used to confirm that the read quality was good and no trimming was required. We mapped the pair-end RNA-Seq reads to the reference genome of Oryctolagus cuniculus (OryCun2.0) from Ensembl (Zerbino et al., 2018) using the Subread aligner (Liao et al., 2013). The number of read counts corresponding to gene transcripts annotated by Ensembl (OryCun2.0.98) were summarized by the featureCounts function from the Rsubread package (Liao et al., 2019). Gene expression levels were quantified by Reads Per Kilobase of transcript, per Million mapped reads (RPKM) with in-house script. We applied Edger3 (McCarthy et al., 2012) to call differentially expressed genes with the following criteria: fold change (FC) > 2, false discovery rate (FDR) < 0.05, and an average expression (RPKM) higher than one in at least one condition.

Proteomics Analysis of Donor-Matched Cells and Their Deposited Matrices
Donor-matched passage 1 IPFSCs (n = 4; C-IPFSC) and ScASCs (n = 4; C-ScASC) as well as the decellularized extracellular matrices (dECM) deposited by passage 1 IPFSCs (n = 4; E-IPFSC) and ScASCs (n = 4; E-ScASC) were used for this analysis. dECM was prepared following our previous protocol (Li and Pei, 2018). Briefly, 0.2% gelatin solution was used to coat a tissue culture plastic (TCP) flask for 1 h, followed by treatment with 1% glutaraldehyde and 1M ethanolamine for 30 min each. Both cells were cultured in the above-treated TCP flasks using growth medium until 100% confluence followed by supplementation with 250 µM L-ascorbic acid phosphate magnesium salt for 10 days and then treatment with Extraction Buffer (0.5% Triton X-100 and 20 mM NH 4 OH in phosphate buffered saline) (Pizzute et al., 2016). Cell lysates (50 µg) were diluted in 8 M urea in 100 mM ammonium bicarbonate (ABC) pH 8.5 for digestion according to the Filter Aided Sample Preparation (FASP) protocol. dECMs were processed as previously described (Barrett et al., 2017). Briefly, samples were lyophilized, and digested with 100 mM cyanogen bromide (CNBr). The resultant digest was solubilized in 8 M urea and FASP digested with sequencing grade trypsin (Wiśniewski et al., 2009). Liquid chromatography tandem mass spectrometry (LC-MS/MS) was performed on an Eksigent 2D nano LC coupled to an LTQ-Orbitrap Velos mass spectrometer. MS acquisition parameters were detailed previously (Hill et al., 2015). Raw files were directly loaded into Proteome Discoverer 2.2 3 (Thermo Fisher Scientific) and searched against mouse SwissProt database using an in-house 1 https://www.bioinformatics.babraham.ac.uk/projects/fastqc/ Mascot TM server (Version 2.5, Matrix Science). Mass tolerances were +/-10ppm for MS peaks, and +/-0.6 Da for MS/MS fragment ions. CNBr /Trypsin specificity was used allowing for 1 missed cleavage. Met oxidation, proline hydroxylation, protein N-terminal acetylation, and peptide N-terminal pyroglutamic acid formation were allowed for variable modifications while carbamidomethyl of Cys was set as a fixed modification. Label Free Quantifications were done using the Minora Feature detector for peak intensity based abundance. Protein thresholds of 1% FDR were used to filter for high confidence identifications. Results were directly exported into Microsoft Excel. Statistical enumeration of the data was achieved through Metaboanalyst (v3.6) (Xia and Wishart, 2016) and pathway analysis was done using the Reactome (Fabregat et al., 2018).

Statistical Analyses
For the data from flow analysis, ELISA, and qPCR, results are presented as the mean and the standard deviation of the mean; the t-test was used to assess data between two groups. All statistical analyses were performed with SPSS 13.0 statistical software (SPSS, Inc., Chicago, IL, United States). p < 0.05 was considered statistically significant.

Adaptation of Transcription Factors During Differentiation Induction
To determine the adaptation of transcription factors in ScASCs and donor-matched IPFSCs, qPCR was used to FIGURE 1 | Comparison of proliferation capacity between ScASCs (n = 4) and donor-matched IPFSCs (n = 4) in (A) cell morphology; (B) cell population doubling (PD) time calculated by cell counting; (C) relative EdU incorporation; (D) CD146 surface marker expression level evaluated by flow cytometry analysis; (E) other surface marker gene expression levels evaluated by RNA-seq, in which a positive value in log2FC means higher expression in IPFSCs (in red) and a negative value means higher expression in ScASCs (in blue); and (F) stemness-related genes NANOG, REX1, NESTIN, and SOX2 and senescence-related genes CDKN1A and TP53 measured by qPCR. Data are shown as bar charts. * indicates a significant difference compared to the corresponding control group (p < 0.05).
evaluate SOX9 for chondrogenesis, PPARG for adipogenesis, and RUNX2 for osteogenesis during differentiation induction. Before induction, only SOX9 expression was significantly higher in IPFSCs compared with ScASCs (Figure 2A). After an 18-day chondrogenic induction, SOX9 expression was maintained at a higher level in IPFSCs similar to day 0 while PPARG expression was significantly lower than ScASCs despite a remarkable increase of PPARG in both cells (Figures 2B,C). After 21day adipogenic induction, the difference in SOX9 expression between groups was diminished; interestingly, both PPARG and RUNX2 expression in IPFSCs were significantly lower than donor-matched ScASCs (Figure 2D). After a 21-day osteogenic induction, the difference in SOX9 expression decreased but IPFSCs still had a higher level than ScASCs; intriguingly, IPFSCs exhibited a higher expression level in PPARG than ScASCs ( Figure 2E).

Development of Differentiation Factors During Induction
To further explore the differentiation preference of donormatched adipose stem cells, qPCR was used to evaluate COL2A1 and ACAN as markers of chondrogenesis, ADIPOQ and LPL for adipogenesis, and BGLAP and SPP1 for osteogenesis during induction. Before induction, IPFSCs had significantly higher expression of COL2A1 and ACAN than ScASCs (Figure 3A), in line with SOX9 expression (Figure 2A). The advantageous expression of COL2A1 and ACAN was further strengthened FIGURE 2 | Expression of transcription factors in ScASCs (n = 4) and donor-matched IPFSCs (n = 4) measured by qPCR. SOX9, PPARG, and RUNX2 were used to represent chondrogenesis, adipogenesis, and osteogenesis, respectively. Expanded cells were evaluated before induction (A) and after induction (B-E): (B) 24 h after centrifugation to form a pellet but before chondrogenic induction; (C) 18 days after chondrogenic induction; (D) 21 days after adipogenic induction; and (E) 21 days after osteogenic induction. Data are shown as bar charts. Expression of each target gene in undifferentiated IPFSCs (A) and differentiated ScASCs/IPFSCs (B-E) is plotted against undifferentiated ScASCs (A), which is set as "1". * indicates a significant difference compared to the corresponding control group (p < 0.05).
during chondrogenic induction ( Figure 3B). No significant difference of COL10A1 expression was observed between groups; however, we found that MMP13 expression was more than 1000 times higher in IPFSCs compared with ScASCs before induction, which became comparable between groups after chondrogenic induction ( Figure 3C).
Before induction, ScASCs exhibited higher expression of ADIPOQ and LPL than IPFSCs (Figure 3D), despite no significant difference in transcription factor PPARG expression (Figure 2A). The dominant expression of ADIPOQ and LPL became dramatically increased during adipogenic induction along with differentiation gene LEP and transcription factor PPARG expression ( Figure 3E). Moreover, higher expression of adiponectin (ADIPOQ) in ScASCs than donor-matched IPFSCs was also confirmed in Western blot analysis ( Figure 3F).
For osteogenic differentiation, despite no difference in BGLAP expression before induction, IPFSCs displayed a higher expression of SPP1 than ScASCs (Figure 3G), which became comparable after 21-day osteogenic induction along with other differentiation genes BGLAP, DCN, and SPARC expression ( Figure 3H). Furthermore, comparable expression of OCN between ScASCs and IPFSCs was also confirmed in ELISA analysis ( Figure 3I).

Transcriptome Difference Between ScASCs and Donor-Matched IPFSCs
We applied RNA-seq data analysis to explore the difference in gene expression levels between ScASCs and IPFSCs. Principal coordinate analysis on gene expression across all genes revealed FIGURE 3 | Expression of differentiation genes in ScASCs (n = 4) and donor-matched IPFSCs (n = 4) before and after induction. For chondrogenic differentiation, typical markers, COL2A1 and ACAN, were evaluated using qPCR before (A) and after chondrogenic induction (B). Hypertrophic markers COL10A1 and MMP13 were also measured after chondrogenic induction (C). For adipogenic differentiation, typical markers, ADIPOQ and LPL, were evaluated using qPCR before (D) and after adipogenic induction (E). LEP expression was also measured after induction (E). Adiponectin (ADIPOQ) expression was confirmed using Western blot analysis with β-actin as an internal control (F). For osteogenic differentiation, typical markers, BGLAP and SPP1, were evaluated using qPCR before (G) and after osteogenic induction (H). Expression of DCN and SPARC were also measured after induction (H). Osteocalcin (OCN) expression was confirmed using ELISA analysis (I).
Expression of each target gene in IPFSCs is plotted against the corresponding ScASCs, which is set as "1", but for those in chondrogenic induction (B/C), expression in day 0 is plotted as "1". Data are shown as bar charts. * indicates a significant difference compared to the corresponding control group (p < 0.05).
that samples from ScASCs and IPFSCs were from two distinctive clusters on the first dimension ( Figure 4A). By applying EdgeR (McCarthy et al., 2012) with stringent thresholds [FC > 2 and FDR < 0.05], we identified 359 and 277 protein-coding genes upregulated and downregulated from the ScASC samples to the IPFSC samples, respectively ( Figure 4B). Intriguingly, differentially expressed genes were enriched in those encoding transcription factors (Figure 4C), including the Homeobox (HOX) gene family related to stem cell differentiation (Seifert et al., 2015; Figure 4D).
We next examined the expression changes on particular groups of genes known to function in chondrogenesis and/or adipogenesis. The data were presented as Log2FC of expression and FDR. Functional annotation clustering was categorized (

Proteome Difference Between ScASCs and Donor-Matched IPFSCs
Both cells and deposited proteins were interrogated with bottom-up mass spectrometry-based proteomics to monitor differences in protein abundance ( Figure 6A). The pathway analysis highlighted that the matrix proteins deposited by ScASCs mapped to more cell cycle regulators than traditional ECM proteins (Figures 6B,C) These proteins did not seem to be laying down as much ECM in general as those deposited by IPFSCs as shown by both the pathway analysis and the relatively lower amount of structural ECM. In contrast, the matrix proteins enriched from IPFSCs include numerous basement membrane components (Figures 6B,C), such as nidogen 1 (NID1) (FC = 0.27, p = 0.008), heparin sulfate proteoglycan 2 (HSPG2)/perlecan (FC = 0.54, p = 0.001), laminin subunits including LAMA5 (FC = 0.10, p = 0.030), LAMB1 (FC = 0.40, p = 0.033), LAMB2 (FC = 0.28, p = 0.004), and LAMG1 (FC = 0.34, p = 0.005). Of relevance, and consistent with the mRNA level, the integrin A6 subunit (ITGA6) that forms a laminin receptor was also significantly increased (FC = 0.03, p = 0.005). In addition, various isoforms of fibronectin (FN) were significantly higher. Interestingly, compared to those secreted by ScASCs, collagen modifying enzymes, such as LOX and tissue

DISCUSSION
Adult stem cells have the ability to differentiate into varied lineages. Increasing evidence indicates that adult stem cells from different tissues do not have equivalent ability toward a specific lineage differentiation (Jones and Pei, 2012;Pizzute et al., 2015). However, there is a lack of comparable studies using donormatched adult stem cells and their matrix microenvironment. In this study, we found that, despite both deriving from adipose tissue, ScASCs and IPFSCs isolated from different depots of four donor-matched rabbits exhibited significant divergence in both proliferation and differentiation potential. Transcriptome and proteomics data provide robust evidence to support our hypothesis that there are site-dependent variations in lineage potential and preference of stem cells from adipose tissue.
Our proliferation data showed that ScASCs grew faster than donor-matched IPFSCs, which might result from higher levels of COL12A1, COL14A1, THBS1, THBS2, and VCAN expression in the matrix proteins deposited by ScASCs. These matrix proteins trend higher in tumor microenvironments and tissues to support proliferative or structurally rigid phenotypes (Wight et al., 2014;Subramanian and Schilling, 2015;Belotti et al., 2016). We also found that IPFSCs expressed significantly higher CD146 than donor-matched ScASCs, which was in line with the gene expression level in RNA-seq data (Log2FC from IPFSCs to ScASCs in CD146/MCAM was 2.62, FDR = 5.7E-07). Regardless of the role as a perivascular marker for adult stem cells (Crisan et al., 2008), a higher level of CD146 expression in adult stem cells might be associated with an advantage in chondrogenic differentiation (Hagmann et al., 2014;Su et al., 2015). Despite the failure to measure expression of surface markers CD90 and CD105 using flow cytometry on both ScASCs and IPFSCs (data not shown), our RNA-seq data indicated that the genes (NT5E, THY1, ENG, and CD44) encoding surface marker proteins (CD73, CD90, CD105, and CD44) were expressed in both cells with NT5E preferentially expressed in ScASCs and CD44 advantageously expressed in IPFSCs. We also found no statistical difference in measured stemness-related genes (NANOG, REX1, NESTIN, and SOX2), indicating that these genes expressed in the donor-matched IPFSCs and ScASCs might not contribute to the discrepancy of a specific lineage differentiation found in this study.
Our differentiation data indicated that IPFSCs exhibited a significantly higher level of chondrogenic markers compared with donor-matched ScASCs. RNA-seq data showed that advantageous genes in IPFSCs can be categorized into several clusters based on functional annotation clustering. For example, the HOX gene, an evolutionarily highly conserved gene family, determines lineage specific differentiation of adult stem cells once tissue damage occurs (Seifert et al., 2015). In this study, HOXD genes, including HOXD9-11 and HOXA10/11, were only dominantly expressed in IPFSCs over donor-matched ScASCs. HOXD genes, one of four grouped HOX clusters (HOXA, HOXB, HOXC, and HOXD) in vertebrates, are suggested to play a role in stem cell based chondrogenesis (Jung and Tsonis, 1998;González-Martín et al., 2014). HOXA11 and HOXD11 double mutant mice had shorter limb formation (Boulet and Capecchi, 2004). Compared with IPFSCs, however, donor-matched ScASCs exhibited robust expression of other HOX genes, such as HOXA4, HOXA5, HOXC4, and HOXC8, which are closely associated with adipogenesis rather than osteogenesis (Levi et al., 2010). An exception is HOXC8 which does not inhibit but retards chondrogenesis, resulting in a boost of proliferating precursor cells and a negative influence on chondrogenesis (Yueh et al., 1998;Cormier et al., 2003).
Another group of conservative transcription factors, TBX family genes, also plays crucial roles throughout development. Our RNA-seq data showed that TBX5, TBX15, and TBX18 were highly expressed in IPFSCs compared to donor-matched ScASCs. TBX5 was reported to activate WNT and fibroblast growth factor (FGF) signaling in upper limb development (Takeuchi et al., 2003). Overexpression of TBX15 in 3T3-L1 preadipocytes impeded adipocyte differentiation and reduced triglyceride content (Gesta et al., 2011). TBX15 and TBX18 co-expressed in the limb mesenchyme indicate their functions during limb development (Tanaka and Tickle, 2004;Singh et al., 2005). Recently, TBX18 was reported to co-express with SOX9 in chondrogenic progenitors and differentiating chondrocytes (Haraguchi et al., 2015). In contrast, TBX1 was reported to be preferentially expressed in the inducible pool of brown adipocytes (beige cells) present in white fat of the inguinal depot (Wu et al., 2012). The above findings indicate that two fat depots, ScAT and IPFP, exhibit distinct HOX and TBX conservative gene expression profiles that likely influence their differentiation preference.
The TGFβ superfamily is composed of two subfamilies; the TGFβ subfamily mainly includes TGFβs (1, 2, and 3) and the BMP subfamily mainly includes BMPs (2, 4-10) (Wang et al., 2014). A considerable amount of in vitro data indicate that TGFβ signaling pathways promote mesenchymal condensation (Barry et al., 2001;Tuli et al., 2003) and the early stages of chondrocyte differentiation, but restrain terminal hypertrophy (Mueller and Tuan, 2008;Shintani et al., 2013). In this study, we found that all TGFβ isotypes (TGFB1, TGFB2, and TGFB3) and latent-TGFB-binding protein 1 (LTBP1) were preferentially expressed in IPFSCs rather than donor-matched ScASCs. However, the protein level showed an opposite trend for TGFB1 and LTBP2, indicating that, while ScASCs have more available bound ligand, IPFSCs may have more free ligand, which needs to be further confirmed. This finding is in line with the advantageous expression in IPFSCs of the combination of SOX5, SOX6, and SOX9 (the SOX trio) that supplies signals sufficient to induce permanent cartilage (Ikeda et al., 2004) and of SOX11 that contributes to the regulation of growth/differentiation factor 5 (GDF5) in joint maintenance (Kan et al., 2013). Our data also showed that only BMP4 exhibited a statistically higher level of expression in IPFSCs compared with donor-matched ScASCs. BMP4 plays an important role in retaining chondrogenic phenotype, through both increasing matrix production and suppressing the production of collagen type X (Reddi, 2001;Steinert et al., 2003). Another important BMP isotype in chondrogenesis, BMP2, was expressed at a higher level in IPFSCs than in donor-matched ScASCs despite no significant difference (FDR = 5.9E-01). Interestingly, BMP3, which was reported to inhibit bone formation by hindering BMP2 activity (Bahamonde and Lyons, 2001), was significantly expressed in ScASCs compared with donor-matched IPFSCs. These findings indicate that IPFSCs have an expression profile consistent with endochondral bone formation over donor-matched ScASCs. The above data support that TGFβ signaling cooperated with BMP signaling in stimulating articular cartilage regeneration (Wang et al., 2014).
Increasing evidence indicates that the canonical Wnt signaling pathway inhibits adipogenesis while the noncanonical Wnt pathway promotes adipogenesis by antagonizing the Wnt/β-catenin pathway (Christodoulides et al., 2009). In accordance with the above findings, we found that preferential expression favoring Wnt/β-catenin signaling genes such as SPARC-related modular calcium binding protein 2 (SMOC2) (Nie and Sage, 2009), low-density lipoprotein receptor-related protein 6 (LRP6) (Peröbner et al., 2012), and Gremlin 2 (GREM2)  in IFPSCs might be responsible for the inhibition of adipogenesis and the enhancement of chondrogenesis. In contrast, expression inhibiting Wnt/β-catenin signaling genes such as WNT4 (Liu et al., 2016) and nuclear factor of activated T-cells 4 (NFATC4) (Yang et al., 2006) in ScASCs might be responsible for promoting adipogenesis and blocking chondrogenesis (Church et al., 2002).
Interestingly, we found FACIT collagen, such as COL14A1 and COL21A1, was expressed at a higher level in ScASCs rather than IPFSCs. We also found that, despite collagen modifying enzymes detectable in both cells, IPFSCs exhibited significantly higher expression of LOX isoforms and tissue transglutaminase (Log2FC from IPFSCs to ScASCs for TGM2 was 3.68, FDR = 4.9E-06) at both mRNA and protein levels than ScASCs, along with fibrillar collagens COL5A1 and COL5A3, which participate in the formation of a fibrillar collagen network and the regulation of fibrillogenesis (Imamura et al., 2000;Malfait et al., 2005;Lincoln et al., 2006). These findings indicate that IPFSCs are more suitable adipose stem cells for chondrogenesis compared to donor-matched ScASCs. During chondrogenesis, the contents of collagen type II and of non-reducible collagen cross-links were positively linked with tensile biomechanical properties and the functional integrity of cartilage tissues (Williamson et al., 2003). LOX family enzyme activities catalyze the eventual enzymatic conversion that is needed for the generation of biosynthetic collagen cross-links. For example, knockdown studies demonstrated that LOXL2 expression is required for mouse ATDC5 chondroprogenitors' differentiation toward chondrocytes through regulation of SNAIL and SOX9, two important transcription factors for chondrogenesis (Iftikhar et al., 2011). Interestingly, hypoxia increased LOX gene expression, which promoted pyridinoline cross-links in engineered cartilage (Makris et al., 2013).
Regarding the matrix components, we found that IPFSCs deposited more basement membrane proteins compared to ScASCs, including LAMA1, LAMA3, and HSPG2 at an mRNA level and LAMA5, LAMB1, LAMB2, LAMG1, NID1, and HSPG2 at a protein level. In contrast, NID2 and COL4A5 were significantly more highly expressed in ScASCs compared to donor-matched IPFSCs. Given the function of HSPG2 encoding perlecan as the "factotum" in cartilage (Melrose et al., 2008) and the role of laminins in cartilage regeneration (Sun et al., 2017), NID2 and COL4A5 expression might be associated with adipogenic differentiation, which remains to be elucidated.
Interestingly, matrix turnover enzymes were uniquely preferentially expressed in IPFSCs or ScASCs. For example, matrix metallopeptidase 13 (MMP13), known as collagenase 3, was found exclusively in IPFSCs. The enzyme's initial role is in tissue remodeling, especially during embryonic development, but it is also highly active in some pathological processes such as osteoarthritis and cancer (Inada et al., 2004;Tardif et al., 2004). In osteoarthritis, MMP13 was highly expressed in chondrocytes and synovial cells responsible for the degradation of cartilage matrix (Inada et al., 2004;Davidson et al., 2006;van den Berg, 2011). However, little is known about MMP13 expression in IPFP, either in healthy donors or in the pathological process. In this study, qPCR data showed that MMP13 was expressed more than 1000-fold higher in IPFSCs than donor-matched ScASCs during cell condensation (day 0 pellet), in accordance with RNA-seq data of cell samples (Log2FC from IPFSCs to ScASCs in MMP13 was 7.45, FDR = 1.4E-16). Interestingly, after an 18-day chondrogenic induction, IPFSCs exhibited a comparable expression of MMP13 to donor-matched ScASCs. From the proteomics analysis, A disintegrin and metalloproteinase with thrombospondin motifs 5 (ADAMTS5) and ADAMTS1 were identified as significantly higher levels in IPFSCs versus a much higher level of ADAMTSL1 (ADAMTS like 1) in ScASCs. The protease profile is likely important for key processing events associated with the physiological role of IPFP in support of the remodeling of articular cartilage, which requires further investigation.
Integrin, a transmembrane receptor, facilitates cell-ECM adhesion. Interestingly, more integrins were advantageously expressed in IPFSCs including ITGA3, ITGA4, ITGA6, ITGA7, and ITGB8 compared to ITGA8, which was superiorly expressed in donor-matched ScASCs. Another transmembrane protein, aquaporin, called the "the plumbing system for cells", was found to serve as a water channel to rapidly move water through cells. Chondrocytes from articular cartilage are unique due to their sensitivity to water transfer and ionic and osmotic adjustments from the extracellular environment and are responsible for the creation of synovial fluid. A recent report showed that the expression of AQP1 and AQP3 were upregulated during chondrogenic induction of human adipose stem cells, indicative of physiological modification of functionally mature chondrocytes to the local environment (Graziano et al., 2018). This finding was consistent with our data, in which all statistically significant AQP gene expression was advantageously found in IPFSCs rather than donor-matched ScASCs, not only AQP1 and AQP3 but also AQP4 and AQP11.
Our study did not find a significant difference in expression levels of osteogenic markers between ScASCs and donor-matched IPFSCs after induction. It has long been speculated that IPFSCs have a higher chondrogenic potential than other fat depots in the body (Sun et al., 2018), which might be due to the IPFP's anatomic site that is closer to synovium and cartilage; it is regarded as a special form of fibro-adipose tissue (Macchi et al., 2016), or a continuation of synovium. IPFSCs may share some common characteristics with synovium-derived stem cells, a tissue-specific stem cell for chondrogenesis (Jones and Pei, 2012). A study comparing IPFSCs with near knee joint ScASCs from donor-matched osteoarthritis patients found that IPFSCs had a superior chondrogenic potential; ScASCs were superior in osteogenesis while both were similar in adipogenic capacity (Lopa et al., 2014). In this study with donor-matched normal adipose stem cells, we observed clear superiority of IPFSCs for chondrogenesis. However, we obtained divergent results on adipogenic and osteogenic potential. There were two possible explanations for the discrepancy; first, species differences between humans and rabbits can lead to considerable diversity, and secondly, Lopa et al. harvested IPFSCs from aged patients affected by osteoarthritis (Lopa et al., 2014) while we harvested from young healthy rabbits. The adult stem cell profile could be greatly affected by disease and aging (Li and Pei, 2012;Lynch and Pei, 2014).
Taken together, by donor-matched comparison and high throughput assay of transcriptome and proteomics, this study investigated the variance of depot-specific adult stem cells from adipose tissues. The findings might shed some light on the hypothesis that adipose stem cells have a depot-dependent lineage preference. Here, we have profiled RNA and protein levels to identify unique transcriptional programs, signaling components, and matrix environments that reveal novel insights into the higher chondrogenic potential of IPFSCs.

DATA AVAILABILITY STATEMENT
The RNA-Seq datasets generated for this study can be found in the GEO, under the accession number GSE142626, https://www. ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE142626. The data that support the findings of this study are available from the corresponding author upon reasonable request.

ETHICS STATEMENT
The animal study was reviewed and approved by IACUC of West Virginia University, protocol number 1608003840.

AUTHOR CONTRIBUTIONS
TW contributed to the conception and design, collection and/assembly of data, data analysis and interpretation, manuscript writing, and final approval of the manuscript. RH and MD contributed to the collection and/assembly of data, data analysis and interpretation, manuscript writing, and final approval of the manuscript. AI, GH, LZ, and KH contributed to the data analysis and interpretation, manuscript writing, and final approval of the manuscript. MP contributed to the conception and design, financial support, data analysis and interpretation, manuscript writing, and final approval of the manuscript.

FUNDING
This work was supported by Research Grants from the Musculoskeletal Transplant Foundation (MTF) and the National Institutes of Health (1R01AR067747-01A1) to MP. We also would like to acknowledge the WVU Flow Cytometry and Single Cell Core Facility and the grants that support the facility, TME CoBRE grant P20GM131322 and the WV CTSI grant GM104942. The work was also partially supported by the National Institute of General Medical Sciences Grant 5U54GM104942-04 to G.H. and P20 GM103434 to AI.