Putative positive role of inflammatory genes in fat deposition supported by altered gene expression in purified human adipocytes and preadipocytes from lean and obese adipose tissues

Background Obesity is a chronic low-grade inflammatory disease that is generally characterized by enhanced inflammation in obese adipose tissue (AT). Here, we investigated alterations in gene expression between lean and obese conditions using mRNA-Seq data derived from human purified adipocytes (ACs) and preadipocytes (preACs). Results Total mRNA-seq data were generated with 27 AC and 21 preAC samples purified from human visceral AT collected during resection surgery in cancer patients, where the samples were classified into lean and obese categories by BMI > 25 kg/m2. We defined four classes of differentially expressed genes (DEGs) by comparing gene expression between (1) lean and obese ACs, (2) lean and obese preACs, (3) lean ACs and lean preACs, and 4) obese ACs and obese preACs. Based on an analysis of comparison 1, numerous canonical obesity-related genes, particularly inflammatory genes including IL-6, TNF-α and IL-1β, i.e., the genes that are expected to be upregulated in obesity conditions, were found to be expressed at significantly lower levels in obese ACs than in lean ACs. In contrast, some inflammatory genes were found to be expressed at higher levels in obese preACs than lean preACs in the analysis of comparison 2. The analysis of comparisons 3 and 4 showed that inflammatory gene classes were expressed at higher levels in differentiated ACs than undifferentiated preACs under both lean and obese conditions; however, the degree of upregulation was significantly greater for lean than for obese conditions. We validated our observations using previously published microarray transcriptome data deposited in the GEO database (GSE80654). Conclusions Taken together, our analyses suggest that inflammatory genes are expressed at lower levels in obese ACs than in lean ACs because lean adipogenesis involves even greater enhancement of inflammatory responses than does obese adipogenesis.


Abstract Background
Obesity is a chronic low-grade in ammatory disease that is generally characterized by enhanced in ammation in obese adipose tissue (AT). Here, we investigated alterations in gene expression between lean and obese conditions using mRNA-Seq data derived from human puri ed adipocytes (ACs) and preadipocytes (preACs).

Results
We de ned four classes of differentially expressed genes (DEGs) by comparing gene expression between 1) lean and obese ACs, 2) lean and obese preACs, 3) lean ACs and lean preACs, and 4) obese ACs and obese preACs. Based on an analysis of comparison 1, numerous canonical obesity-related genes, particularly in ammatory genes including IL6, TNF-and IL-1 , i.e., the genes that are expected to be upregulated in obesity conditions, were found to be expressed at signi cantly lower levels in obese ACs than in lean ACs. In contrast, some in ammatory genes were found to be expressed at higher levels in obese preACs than lean preACs in the analysis of comparison 2. These two results indicate that (1) up-/downregulation of genes in ACs and preACs is inversely controlled during the fat deposition process and (2) preACs rather than ACs have increased in ammatory response genes in comparisons of lean and obese conditions for each of these cell types. Analysis of comparisons 3 and 4 showed that in ammatory gene classes were expressed at higher levels in differentiated ACs than undifferentiated preACs under both lean and obese conditions; however, the degree of upregulation was greater for lean than for obese conditions.

Conclusions
Taken together, our analyses may suggest that lean fat differentiation involves even greater enhancement of in ammatory responses than does obese fat differentiation.

Background
A widely accepted notion about obesity is that in ammatory responses are elevated in the serum as well as adipose tissue (AT) of obese organisms, as a so-called low-grade in ammatory disease [1]. AT is a primary organ that maintains homeostasis between energy uptake and energy expenditure, in which excess energy is stored in the form of triacylglycerols, whereas free fatty acids are released during fasting [2,3]. AT is also an endocrine organ that secretes various bioactive factors, namely, adipokines, that regulate the whole-body level of immune and in ammatory responses [4][5][6]. At the cellular level, obesity is de ned as accelerated AT expansion and remodeling that induces either AT hypertrophy (i.e., adipocyte expansion due to excessive fat storage) or hyperplasia (i.e., increased adipogenesis from preadipocytes) through ECM remodeling and angiogenesis [7,8]. Various ECM proteins, including MMP2, ADAM, TIMP, CTSK, and CTSS, are altered in obese AT [9][10][11][12]. Angiogenic genes such as VEGF and ANGPT2 are upregulated in response to activated HIF1A (i.e., hypoxia-related transcription factor) [13,14]. LEP and ADIPOQ are another genes that have potential adipokine functions involved in AT remodeling [8,15].
Several studies have shown that the accumulation of excess fat in AT leads to the release of in ammatory mediators, such as TNF-α and IL-6, and the reduction of anti-in ammatory cytokines, such as adiponectin, is associated with chronic in ammation in obese individuals. It is also known that excess fat that is over owed from AT can deposit in other organs such as liver, pancreas, and muscle, causing insulin resistance [16]. In addition, the oxidative stress due to excessive nutrients intake can contribute to increased in ammation associated with obesity [17]. The increased serum level of C-reactive protein (CRP) is one of the markers of chronic in ammation in obesity [18]. In fact, numerous studies strongly support the idea that the in ammatory responses mounted in AT and the accompanying extensive molecular and cellular changes are responsible for excess fat deposition associated with the metabolic pathogenicity of obesity such as diabetes and atherosclerosis [19,20].
In contrast, a few recent studies provided an interesting view on the positive role of in ammatory responses in controlling adipogenesis. For instance, Ye and McGuinness [21] described that in ammatory responses are required for the maintenance of a healthy AT microenvironment for AT remodeling and expansion. By constructing three mouse models with adipocyte (AC)-speci cally attenuated in ammatory responses, Asterholm et al. [22] showed that proper AT remodeling and expansion are executed by in ammatory responses at the level of ACs, so that reduced or impaired local in ammatory responses in the AC cause pathogenic obesity-related conditions, such as hepatic steatosis or metabolic dysfunction due to ectopic lipid accumulation [23,24]. Interestingly, a recent AT-derived RNA-seq analysis of pigs showed that pigs with thicker backfat tended to express signi cantly lower levels of immune and in ammatory genes than pigs with thinner back fat, which indicates that expression of high in ammatory genes may be associated with lower fat accumulation in AT [25]. Few studies have addressed the positive role of the in ammatory response which may have a role in fat deposition in lean individuals, although fat cell differentiation and fat deposition must be regulated appropriately to maintain healthy fat homeostasis.
We think that the contradictory conclusions regarding the positive or pathogenic role of in ammation in obesity are largely due to the use of differential experimental designs among different studies. As expected, retrieving lean (or healthy) AT samples is much more di cult than retrieving obese AT samples in humans. Puri cation of cells such as ACs, preadipocytes (preACs), macrophages, and endothelial cells residing in AT is even more di cult. For this reason, most transcriptome-based studies in humans have been unable to identify so-called differentially expressed genes (DEGs) by directly comparing lean and obese ATs from cohorts of lean and obese individuals. Instead, some studies have investigated DEGs in AT samples during weight loss induced by bariatric surgery for the same individuals [26], and others have analyzed DEGs between subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT) derived from the same obese individuals [27]. Gene expression has rarely been analyzed between obese and lean ATs even in model organisms such as mice, where a few studies have pursued identi cation of DEGs between mice fed normal chow and mice fed a high-fat diet (HFD) [28]. Most of these transcriptome-based studies have simply concluded that in ammatory genes in ATs are decreased as organisms lose weight and increased as organisms become obese, without asking which cell type in AT was responsible for the increase/decrease in in ammatory genes during weight gain or loss.
It seems that two independent questions are often intermingled in the studies related to obesity and obesity-associated genes described above, (1) whether increased in ammation is involved in fat accumulation and (2) whether increased in ammation is associated with the development of pathogenicity in obese patients. In fact, it is not simple to study relationships between obesity-associated gene expressions and pathogenicity because obesity is not necessarily connected to pathogenic diseases. Here, we investigate gene expression of obese AC samples in comparison with that of lean AC samples, particularly for the expression of in ammatory genes, in order to answer the question #(1). Further, we also investigate genes that are altered between lean ACs and lean preACs and between obese ACs and obese preACs. Transcriptome-based analysis was conducted using mRNA-Seq data generated from highly puri ed ACs and preACs obtained from visceral (omental) AT (i.e., VAT) collected during resection surgery from cancer patients.

Four-way Identi cation Of Degs
The owchart of our analysis is depicted in Fig. 1. After production of mRNA-seq data from the puri ed ACs and preACs of lean and obese individuals, we investigated two different questions: 1) which genes are signi cantly upregulated or downregulated under obese conditions in comparison with lean conditions, and 2) which alterations in gene expression detected for obese AC differentiation (i.e., the information extracted from the DEGs obtained by comparing ACs and preACs from obese individuals) are signi cantly different from those detected for lean AC differentiation (i.e., the information extracted from the DEGs obtained by comparing ACs and preACs from lean individuals). As a validation, we compared our conclusion with that of a previously published paper that provided a list of DEGs obtained by comparing gene expression between lean and obese ACs derived from lean and obese SAT samples; the original dataset produced by the microarray platform was downloaded from the GEO database (GSE80654). How similar that list of DEGs is to our results discussed later in the DISCUSSION section. Various thresholds were tested to select DEGs (Additional le 2: Table S1), and DEGs were identi ed for the four classes mentioned above. Class I and II DEGs were investigated to answer question #1 described above, i.e., to determine the differences between obese ACs and lean ACs and between obese preACs and lean preACs. Class III and IV DEGs were chosen to answer question #2 described above, i.e., to determine how gene expression is altered during obese and lean AC differentiation.
De ning Intermediate Obesity Samples In 'class I: Ac-degs' A total of 1,198 genes and 314 genes were classi ed as 'Class I: AC-DEGs', with P < 0.01 and Q < 0.05, respectively (Additional le 2: Table S1). 'L-AC' samples could not be differentiated from 'O-AC' samples by either of these DEGs in the analysis of unsupervised clustering processed with a heatmap; the samples in the middle of the heatmap did not show gene expression patterns pertinent to the 'L-AC' and 'O-AC' categories ( Fig. 2A). Notably, Gene ontology (GO) analysis showed that canonical obesity-related genes involved in in ammation and ECM were downregulated in obese ACs rather than in lean ACs (Fig. 2B). The ambiguity of sample classi cation by these DEGs was also con rmed in principal component analysis (PCA) (Fig. 2C). Thus, we subcategorized these ambiguous samples separately  Table S2).
Subsequently, for these rede ned three groups of samples, some obesity-related clinical information was investigated, including BMI, waist circumference (WC), fasting plasma glucose (FPG) level, C-peptide (Cpep), high-density lipoprotein (HDL) level, and low-density lipoprotein (LDL) level (Fig. 2D). Interestingly, 'I-AC' samples were special in that the BMIs of 'I-AC' were expectedly positioned between 'L e -AC' and 'O e -AC'; however, the WCs and FPGs of 'I-AC' were comparable to those of 'L e -AC'. In addition, except for BMI, WC, and FPG, all the other clinical levels showed no signi cant difference among these three groups, although 'I-AC' was located between 'L e -AC' and 'O e -AC'. Notably, FPG levels seemed to be associated with WC rather than with BMI.
The existence of a third group, 'I-AC' samples, may not be surprising, considering the complexities of molecular etiologies causing obesity involved with various genetic and epigenetic alterations and the unclear association between obesity and obesity-related metabolic diseases.
In ammatory genes are expressed at lower levels in obese ACs than in lean ACs  Table S3). To understand alterations in gene expression related to obesity, we applied different thresholds to produce similar numbers of DEGs for these three categories (indicated '*' in Additional le 4: Table S3). Consequently, a total of 2,657 (Q < 0.01), 1,474 (P < 0.01), and 1,324 (Q < 0.05) DEGs were selected for 'LO-DEGs', 'LI-DEGs', and 'IO-DEGs', respectively; the heatmap of each group was constructed to visualize gene expression differences (Fig. 3A). Notably, the highest number of genes was allocated in 'LO-DEGs', despite the stringent threshold applied. Note that we focused on collecting similar numbers of DEGs rather than on determining a single criterion or a single most important gene (although all the thresholds were chosen in ranges considered statistically signi cant) to reveal trends in gene expressions between two different conditions. A striking observation emerged from GO analysis. Speci cally, a total of 1,874 genes of the 2,657 'LO-DEGs' (70.5%), i.e., genes largely assigned to the in ammatory response, cell adhesion were expressed at signi cantly lower levels in 'O-AC' than in 'L-AC' (Fig. 3B, in the wide blue box). A similar result was observed in Fig. 2B, and the downregulation of these genes seems to be ampli ed when 'I-AC' samples are excluded in the DEG analysis. Moreover, 'IO-DEGs' revealed the same pattern as did 'LO-DEGs' (Fig. 3B, in the wide blue box). This observation is striking because these classes of genes are all canonical obesity-related genes that are well-known to be expressed at higher levels in obese AT [19,20].
By contrast, other DEGs involved in fat metabolism in 'LO-DEGs' and 'IO-DEGs' were consistent with the previous ndings, i.e., upregulation under obese conditions. For instance, a total of 783 genes of the 2,657 'LO-DEGs' (29.5%), including LEP, CES1, and NQO1, that were expressed at higher levels in 'O e -AC' than in 'L e -AC', were largely assigned to mitochondrial metabolism and the oxidation-reduction process (ROS) (Fig. 3B, in the wide red box).
'LI-DEGs' were also assigned to distinctive GO functions; RNA metabolism genes were expressed at lower levels in 'I-AC' than in 'L e -AC' (i.e., downregulated; narrow blue box in Fig. 3B), and genes involved in centrosome organization and protein phosphorylation were expressed at higher levels in 'I-AC' than in 'L e -AC' (i.e., upregulated; narrow red box in Fig. 3B), con rming that 'I-AC' is distinct and not comparable to 'L e -AC' or 'O e -AC'.
Gene set enrichment analysis (GSEA), i.e., a tool designed to see whether an a priori de ned set of genes shows signi cant differences in expression between two different biological conditions, led to the same conclusion as did GO analysis. Speci cally, genes belonging to in ammatory or angiogenesis functions were signi cantly upregulated in lean ACs rather than obese ACs, whereas genes belonging to cellular respiration functions were signi cantly upregulated in obese ACs (Additional le 5: Figure S2).

Detailed examination of expression alterations between lean and obese ACs
The regrouped 'L e -AC', 'I-AC', and 'O e -AC' were assumed to re ect different degrees of obesity based on BMI, as shown in Fig. 2D. Thus, changes in gene expression can be examined further by investigating 'LI-DEGs' and 'IO-DEGs'. For instance, if a gene that is upregulated both in 'LI-DEGs (i.e., genes expressed at higher levels in 'I-AC' than in 'L e -AC') and 'IO-DEGs (i.e., genes expressed at higher levels in 'O e -AC' than in 'I-AC'), it can be concluded that the gene is 'progressively upregulated (named 'progressive-up') because it is upregulated in 'L e ' compared with that in 'I' and upregulated again in 'I' compared with that in 'O e ' (refer to Additional le 6: Figure S3). Similarly, if a gene is downregulated in both 'LI-DEGs' and 'IO-DEGs', the gene is de ned as progressively downregulated (named 'progressive-down'). Using this scheme, genes were allocated into eight different categories as shown in Fig. 4 and Additional le 6: Figure S3. As a result, only seven genes, including ACAA1, KLHL22, and AKR1C3, were identi ed as 'progressive-up', while a total of 66 genes were categorized as 'progressive-down'. Notably, genes assigned to cell migration, cell adhesion, and angiogenesis were allocated to the 'progressive-down' category.
Cell cycle and metabolic process genes were upregulated in 'I-AC' compared with those in 'L-AC' and sustained their expression in 'O-AC', i.e., upregulated genes in the 'LI-DEGs' category but not in the 'IO-DEGs' category (named 'initial-up'). By contrast, RNA processing, angiogenesis, and signal transduction genes were downregulated in 'I-AC' and sustained their expression in 'O-AC', i.e., downregulated genes in the 'LI-DEGs' category but not in the 'IO-DEGs' category (named 'initial-down') ( Fig. 4). This result indicates that angiogenesis alteration and cell proliferation may start in the early stage of obesity. Genes involved in ROS metabolism and fatty acid biosynthetic process genes were upregulated in the later stage in obesity, i.e., genes not in the 'LI-DEGs' category but upregulated in the 'IO-DEGs' category (named 'laterup'). By contrast, genes involved in in ammation, cell adhesion, and ECM organization were downregulated in a later stage in obesity, i.e., genes not in the 'LI-DEGs' category but downregulated in the 'IO-DEGs' category (named 'later-down') ( Fig. 4). Notably, genes in the 'later-down' category showed that most canonical obesity genes were downregulated in in the later stage of obesity.
Integration Of Degs Between 'ac-degs' And 'preac-degs' Next, we investigated similarities and differences between 'Class I: AC-DEGs' and 'Class II: preAC-DEGs' by constructing a graph integrating the two classes of DEGs (Fig. 6). Note that 'LO-DEGs' were used here for the intersecting procedure with 'preAC-DEGs'. Genes were divided into ve groups depicted in the graph as blue (representing downregulation in obese samples) or pink (representing upregulation in obese samples) dots connected either to ACs or preACs: 'AC-speci c-up', 'AC-speci c-down', 'preAC-speci c-up', 'preAC-speci c-down', and 'Common'.
Two conclusions have been made from this analysis: 1) most DEGs were assigned to the 'AC-' or 'preACspeci c' categories, whereas only a few DEGs overlapped with the 'Common' category, indicating that alterations in gene expressions are distinctive for ACs and preACs; 2) alterations in gene expression re ected in 'AC-DEGs' were largely in the opposite direction of those in 'preAC-DEGs', as noted above. Again, the inverse relationship of alterations in gene expression was particularly prominent for genes involved in the in ammatory response (Fig. 6). In ammatory genes were expressed at lower levels in obese ACs (i.e., 'AC-speci c-down' category), but they were expressed at higher levels in obese preACs (i.e., 'preAC-speci c-up').
Additionally, genes involved in ROS were upregulated in obese ACs, as assigned to the 'AC-speci c-up' category. A few DEGs were 'Common' but expressed in the opposite directions between 'AC-DEGs' and 'preAC-DEGs', depicted as the two-colored balls in the middle of the graph (Fig. 6).
We further investigated whether these DEGs were signi cantly more enriched with obesity-associated genes than were non-DEGs by mapping them to the obesity-associated genes identi ed by genome-wide association studies (GWAS). We collected a total of 614 obesity-associated genes from GWASdb2 (http://jjwanglab.org/gwasdb) [30] and intersected them with the DEGs as speci cally depicted gene symbols in the graph (Fig. 6). As expected, obesity-associated genes were signi cantly enriched in 'ACspeci c-down' (P < 0.05) and 'preAC-speci c-down' (P < 0.05) (Additional le 8: Table S4). This result may con rm the importance of the functional roles of these DEGs in understanding the molecular alterations leading to obesity.
Both lean and obese AC differentiation are required for the enhancement of in ammatory genes Comparing expression between preACs and ACs was assumed to reveal the changes in gene expression during AC differentiation from preAC. Under this assumption, we investigated how obese AC differentiation is differentiated from lean AC differentiation by obtaining Class III and Class IV DEGs; Class III: Lean_Ag-DEGs were estimated by comparing 'L-preAC' and 'L e -AC'; and Class IV: Obese_Ag-DEGs were obtained by comparing 'O-preAC' and 'O e -AC' (Additional le 1: Figure S1 and Additional le 2: Table   S1). For the two classes, 'upregulation' or 'downregulation' was determined based on the expression between ACs and preACs, i.e., genes that were expressed at higher levels in ACs than in preACs were upregulated genes, and genes that were expressed at lower levels in ACs than in preACs were downregulated genes.
First, the degree of alterations in gene expression between preACs and ACs was extremely large, regardless of lean or obese samples; a total of 8,448 genes and 10,234 genes were identi ed as 'Class III: Lean_Ag-DEGs' and 'Class IV: Obese_Ag-DEGs', respectively at Q < 0.01. Thus, we subcategorized these DEGs into four subcategories by considering the log 2 fold change in gene expression along with the Q < 0.01 threshold (Fig. 7). Subsequently, for each of these categories, Class III and Class IV intersected, which led to three subcategories of DEGs, i.e., 'lean AC differentiation-speci c (LS)', 'obese AC differentiation-speci c (OS)', and 'commonly altered for both (CA)', to investigate how lean AC differentiation and obese AC differentiation are different from each other.
As a result, most in ammatory genes, such as leukocyte migration, cell chemotaxis, and complement activation genes, were allocated in the highest upregulation category (Q < 0.01 and |log 2 fold change| > 4) for both lean and obese AC differentiation ('CA') (genes in pink boxes in Fig. 7A). However, the extent of upregulation of these in ammatory genes in 'LS' was signi cantly higher than that of those in 'OS' (Fig. 7B), indicating that both lean and obese AC differentiation are coupled with increased expression of in ammatory genes and that lean AC differentiation, rather than obese AC differentiation, requires stronger upregulation of these in ammatory genes.
In addition, RNA metabolism genes were downregulated for both 'LS' and 'OS' (Fig. 7A, genes boxed in yellow) but with small fold changes, indicating that both lean and obese AC differentiation are required for dynamic alterations in gene expression machineries. Interestingly, genes involved in mitochondrial metabolism, including oxidative phosphorylation or electron transport, were upregulated with small fold changes in either 'LS' or 'OS', but no genes were altered in the 'CA' category ( Fig. 7A, genes in green boxes), indicating that different kinds of mitochondrial metabolism genes are slightly but signi cantly altered during lean and obese AC differentiation. In contrast, fatty acid metabolism and catabolic process genes were mostly allocated to the 'CA' category with a wide range of fold changes (Fig. 7A, genes boxed gray), indicating that some common metabolic processes underlie both lean and obese AC differentiation.

Discussion
Obesity is often characterized as a low-grade in ammatory disease in which an enhanced in ammatory response in AT and the serum of obese organisms has been well established [31]. However, surprisingly, studies that addressed the possible enhancement of in ammatory genes by analyzing DEGs between obese fats and lean fats collected from obese and lean cohorts are extremely rare, which is primarily due to the di culties of obtaining fat samples derived from lean and healthy individuals.
By estimating DEGs between lean ATs and obese ATs respectively derived from the human lean and obese cohorts, we reached a contradictory conclusion regarding the up-or downregulation of in ammatory genes involved in obesity. We found that in ammatory genes such as IL-6 and TNF-α were actually expressed at lower levels in obese ACs than in lean ACs and at higher levels in obese preACs than in lean preACs, indicating that 1) in ammatory genes are downregulated in obese conditions for ACs but 2) upregulated in obese conditions for preACs. The downregulation of these in ammatory genes, including IL-6 and TNF-α, were partly con rmed by intersecting the list of our DEGs and the DEGs that were previously published [32] and deposited in the GEO database (GSE80654), as will be discussed below in detail.
These results led us to revisit the notion that the upregulation of in ammatory genes may be the primary cause of obesity. We thus designed another comparison of gene expression between ACs and preACs, i.e., a comparison that was expected to estimate gene expression differences that occur during AC differentiation. As shown in Fig. 7, we showed that ACs express higher levels of in ammatory genes than do preACs for both lean and obese conditions, indicating that both lean and obese AC differentiation require increased in ammatory response genes. Interestingly, we observed that the extent of enhancement of in ammatory genes in ACs compared with preACs was signi cantly higher for lean conditions than for obese conditions (Fig. 7A and B). We think that this result clearly supports Asterholm et al.'s (2014) view of the positive role of in ammation in fat deposition, i.e., an attenuated in ammatory response may be linked to more harmful fat deposition. We validated our work by comparing with similar previous studies that were conducted based on an experimental design similar to that in the present work. Unfortunately, only one paper was found where the authors reported DEGs that were signi cantly altered in obese ACs compared with lean ACs puri ed from human AT samples [32]; in total, 24 upregulated DEGs and 64 downregulated DEGs were reported.
Note that the ACs in this study were derived from SAT rather than VAT. Interestingly, we found signi cant overlap in the list of 'LO-DEGs' between our study and Ehrlund et al.'s study. For instance, genes, including NQO1, VLDLR, etc., were upregulated, whereas genes including IL-6, MMP2, and CD44 (i.e., in ammatory response genes) were downregulated in obese ACs compared with those in lean ACs. Approximately 21.6% (19/88) overlapped with the same direction (i.e., up or down) between our study and Ehrlund et al.'s study. We think that this degree of overlap is quite remarkable, considering that we used an mRNA-Seq platform whereas they used a microarray platform to produce transcriptome data, and that a signi cant gene expression difference has been reported between SAT and VAT [33,34]. In addition, Xing et al. (2015)'s RNA-Seq-based transcriptome study of pigs showed that pigs with thicker back fats (corresponding to obese ATs in the present work) showed a signi cant lower expression of in ammatory genes than pigs with thinner back fats (corresponding to lean ATs in the present work), which is consistent with our observation in the present work that in ammatory genes are downregulated in obese ACs compared to lean ACs.
Another important but often forgotten aspect is that AT is a complex organ with a residing mixture of highly heterogeneous cell types, including macrophages, other immune cells, preACs, endothelial cells, and lipid-lled ACs [35][36][37][38]. Compositional changes in these cells are associated with obesity and construct unique microenvironments within AT, entailing the synthesis and turnover of ECM components that lead to changes in adiposity accompanying limited/excessive nutritional supply [39]. A key player in regulating AT in ammation is macrophage in ammation. The involvement of macrophages in AT associated with obesity is known as M1 macrophage polarization. Unfortunately, due to the lack of puri ed macrophages, we were unable to investigate whether macrophages have a major role in enhanced in ammation in obese AT. Several studies have already noted that the source of enhanced in ammatory cytokines such as TNF-α and IL-6 is macrophages rather than ACs [40,41]. Instead, in the present work, we suggest that preACs are partly responsible for the enhanced in ammatory responses in obese AT.

Conclusions
Using transcriptome-based analyses of mRNA-Seq data derived from human puri ed ACs and preACs between lean and obese conditions, we found that the canonical obesity-related upregulated genes, particularly in ammatory response genes, were expressed at signi cantly lower levels in obese ACs than in lean ACs. Moreover, the levels of these classes of genes increased in both lean ACs and obese ACs compared to the respective lean preACs and obese preACs; however, the levels of enhancement of these genes were even greater for lean ACs than obese ACs.
To our knowledge, our study is the rst to investigate alterations in gene expression using the human transcriptome from puri ed preACs and ACs in lean and obese VAT. We believe our present work will help to resolve some of the unanswered questions regarding the molecular alterations that occur in lean and obese fat accumulation.

Methods
Preparation of transcriptomes derived from human AC and preAC samples All the transcriptome samples used in the present work were produced by one of the out-sourced studies performed by the Korea National Institute of Health (KNIH). KNIH made efforts to collecting epigenomes and transcriptomes as a participating institute of the International Human Epigenome Consortium (IHEC), granting research funds to recruited research groups (selected by an evaluation process) for collecting tissue samples and their epigenome data. The collected data were also strictly regulated and distributed by the KNIH to the research groups who had proposed to analyze them after evaluation. We were one of the research groups that were selected to access and analyze the raw data in the KNIH server called the open access (OA) system under limited permission.

Ethics Statement
This study was performed in accordance with the principles of the Declaration of Helsinki and was approved by the Kangwon National University Hospital (Chuncheon, Korea) Institutional Review Board (IRB) (KWNUIRB-2017-11-003).

Puri cation Of Ac And Preac Samples
Retrieving VAT from healthy individuals is extremely di cult, and puri cation of ACs or preACs from the small amount of lean VAT is even more di cult, so that we puri ed ACs and preACs from AT that was isolated from VAT in the abdominal region during surgical resection of human cancer patients. To exclude a possibility that gene expression in AT can be affected by tumors residing in locations distant from the AT we collected, we rst con rmed that there was no cancer type bias between lean and obese samples (Additional le 9: Table S5). Second, we con rmed that the CRP levels, i.e., an indicator of systemic in ammation, of the samples that we analyzed were all less than 1 except one for 1 outlier (#ob50), indicating that no systemic cachexia response affected gene expression for both lean and obese ATs (or ACs/preACs) from the tumors (Additional le 10: Table S6). In addition, we found no bias in sex, age, cancer grades between lean and obese samples (Additional le 10: Table S6).
Then, AT with blood vessels and connective tissue removed was washed with PBS to remove blood, including white and red blood cells. The 100 ~ 200 g of fat that was collected from each patient was minced and treated with collagenase I for 1 hour at 37 °C, in which the samples were washed three times with PBS every 20 minutes. The digested samples were ltered with 350 µm mesh to remove undigested tissue. Fetal bovine serum (10%) was added to stop the collagenase I reaction.
After centrifugation at 400x g for 10 minutes, the supernatant fraction, i.e., the fraction containing mature ACs, and the stromal vascular fraction (SVF) pellet were collected separately (Additional le 11: Figure  S5). Mature ACs washed with PBS and medium were then used for RNA isolation. The SVF pellet was incubated in red blood cell lysis buffer for 15 minutes to remove red blood cells, and the SVF pellet was retrieved again after centrifugation. The cells in the SVF pellet were ltered through 100 µm mesh, and 40 µm nylon mesh was used for MACS/FACS sorting to purify preadipocytes (CD45−/CD34+/CD31 − cells) (Additional le 11: Figure S5) [42].

Rna Extraction
We tried to collect ACs and preACs as a pair of samples obtained from the ATs from the same individual, but the success rate of extracting preACs was not very high, which is why the number of preACs is signi cantly lower than that of ACs, as described below.
Total RNA from 27 AC and 21 preAC samples was extracted with an RNeasy Lipid tissue kit (Qiagen, Hilden, Germany) and RNeasy Micro Kit (Qiagen, Hilden, Germany) using the manufacturer's recommendations. Samples from AC comprising 12 L-AC and 15 O-AC samples, whereas preAC samples consisted of 3 L-preACs and 18 O-preACs. Obesity among collected patient samples was diagnosed by BMI > 25 kg/m 2 rather than BMI > 30 kg/m 2 . Based on a report from the Korea National Institute of Health, Koreans are particularly troubled by a higher incidence of metabolic diseases coupled with obesity, even though they have a BMI lower than the worldwide average. This is why Korean obesity is diagnosed as BMI > 25 kg/m 2 rather than BMI > 30 kg/m 2 [43,44].

Availability Of Data And Materials
All RNA-Seq data are accessible on the KNIH OA system through gaining speci c permission by the evaluation of KNIH.  Figure 1 Overall schematic of the work ow. Work ow is depicted as a owchart. Abbreviations used in this owchart are as follows; AC, adipocyte; preAC, preadipocyte; DEG, differentially expressed gene; Le, lean extreme; Oe, obese extreme; Ag, AC differentiation; LS, lean AC differentiation-speci c; OS, obese AC differentiation-speci c; CA, commonly altered for both. The statistical test of clinical information among the groups was performed by the 'Wilcoxon rank-sum test'. '*' indicates that the statistical test is P < 0.05.  Table S3). The same notations used for Fig. 2A are also used in this heatmap (refer to Fig. 2A legend). B. Analysis of GO functional terms for each of the three classes of DEGs (see Materials and methods). 'LO', 'LI', and 'IO' represent 'LO-DEGs', 'LI-DEGs' and 'IO-DEGs', respectively. For each class of DEGs, genes are divided into upregulated genes (i.e., genes with higher levels in obese ACs than in lean ACs) and downregulated genes (i.e., genes with lower levels in obese ACs than in lean ACs). Wide red and blue boxes within the plot indicate upregulated and downregulated genes, respectively, for both 'LO-DEGs' and 'IO-DEGs'. Narrow red and blue boxes within the plot indicate upregulated and downregulated genes, respectively, for the 'LI-DEGs'.  Figure S3 for the detailed strategy for the subcategorization). A heatmap is generated for each category of DEGs with genes assigned to each category. The bars in the rst column represent the trend of gene expression levels in each category, and the colors, blue, gray, and red of the lines surrounding each bar represent genes in the Le, I, and Oe categories, respectively.
Construction of heatmap accompanied with unsupervised hierarchical clustering. A heatmap coupled with unsupervised hierarchical clustering is generated with a total of 213 'Class II: preAC-DEGs' (p < 0.01) (Additional le 2: Table S1). Refer to the notations in the Fig. 2A legend. B. GO analysis of 'preAC-DEGs'; red and green bars represent upregulated (i.e., genes with higher levels in obese preACs than in lean preACs) and downregulated genes (i.e., genes with lower levels in obese preACs than in lean preACs), respectively. Refer to the Fig. 2B legend for how to select the functional terms and the meaning of the scale on the bottom.

Figure 6
Integration of DEGs between 'AC-DEGs' and 'preAC-DEGs'. To compare DEGs obtained from ACs with those from preACs, a total of 2,657 'LO-DEGs' and 213 'preAC-DEGs' are intersected; from these, DEGs are divided into ve groups depicted in the graph; dots in the blue box represent genes that are expressed at lower levels in obese samples than in lean samples (i.e., downregulated genes), and dots in the pink box represent genes that are expressed at higher levels in obese samples than in lean samples (i.e., upregulated genes). Each dot representing each gene is connected to ACs or preACs accordingly; 'ACspeci c-up', 'AC-speci c-down', 'preAC-speci c-up', 'preAC-speci c-down', and 'Common'. The two colored (depicted as blue-red or red-blue) balls in the middle of the graph represent DEGs that are commonly found in both 'LO-DEGs' (i.e., AC-DEGs) and 'preAC-DEGs' but in the opposite directions (i.e., up for 'LO-DEGs' but down for 'preAC-DEGs' or its inverse). Genes depicted with the gene symbols and gene symbols with the bold face represent the genes that overlap with the genes previously identi ed as obesityassociated genes by GWAS with p < 5 x 10-6 and p < 5 x 10-8, respectively.