Transglutaminases and Obesity in Humans: Association of F13A1 to Adipocyte Hypertrophy and Adipose Tissue Immune Response

Transglutaminases TG2 and FXIII-A have recently been linked to adipose tissue biology and obesity, however, human studies for TG family members in adipocytes have not been conducted. In this study, we investigated the association of TGM family members to acquired weight gain in a rare set of monozygotic (MZ) twins discordant for body weight, i.e., heavy–lean twin pairs. We report that F13A1 is the only TGM family member showing significantly altered, higher expression in adipose tissue of the heavier twin. Our previous work linked adipocyte F13A1 to increased weight, body fat mass, adipocyte size, and pro-inflammatory pathways. Here, we explored further the link of F13A1 to adipocyte size in the MZ twins via a previously conducted TWA study that was further mined for genes that specifically associate to hypertrophic adipocytes. We report that differential expression of F13A1 (ΔHeavy–Lean) associated with 47 genes which were linked via gene enrichment analysis to immune response, leucocyte and neutrophil activation, as well as cytokine response and signaling. Our work brings further support to the role of F13A1 in the human adipose tissue pathology, suggesting a role in the cascade that links hypertrophic adipocytes with inflammation.


Results
Our first aim was to examine potential links of TGM family members to excess acquired weight in the 12 weight discordant MZ twins. The significant weight and adipocity discordance of the twins are presented in Table 1. Differential (∆Heavy-Lean twin) gene expression in adipose tissue and adipocyte-enriched fraction is reflective of cellular changes in the tissue upon its expansion during weight gain. As presented in Figure 1A,B, F13A1 is the only TGM family gene that reacts to excess weight in a significant manner and shows increased expression in both adipose tissue and adipocyte-enriched fraction of the heavier twin [30].

Results
Our first aim was to examine potential links of TGM family members to excess acquired weight in the 12 weight discordant MZ twins. The significant weight and adipocity discordance of the twins are presented in Table 1. Differential (Heavy-Lean twin) gene expression in adipose tissue and adipocyte-enriched fraction is reflective of cellular changes in the tissue upon its expansion during weight gain. As presented in Figure 1A,B, F13A1 is the only TGM family gene that reacts to excess weight in a significant manner and shows increased expression in both adipose tissue and adipocyteenriched fraction of the heavier twin [30].   Our previous work linked increase in F13A1 expression to increase in adipocyte size/volume [30]. This, with previous mouse data, suggests that F13A1 may be increasingly produced by larger adipocytes or may regulate the cell size. In the same study, we identified 182 F13A1and weight-associated genes which showed significant (p < 0.05) and strong Pearson correlation (r > 0.7) for Heavy-Lean ∆F13A1 [30]. Gene enrichment analysis showed that the 182 genes linked to cell stress, inflammation, tissue remodeling, hemostasis, and angiogenesis [30]. To further examine the link of F13A1 to adipocyte size, we applied the following strategy ( Figure 2). First, the pool of 182 genes was further screened for genes that also showed correlation to Heavy-Lean ∆Adipocyte Diameter in the 12 MZ twin pairs. This resulted in a total of 283 genes from which 42 were found as shared with the original 182 gene set. Second, we screened the set for adipocyte size correlating (positive and negative) genes that were identified from 15 MZ twins, some of which were shared with this study [17]. This resulted in no additional genes (Supplemental Tables S1 and S2). Third, we screened 14 reported 'large adipocyte genes' from a study where adipocytes, isolated from human SAT, were separated via 70 µm mesh into small (57.8 ± 0.23 µm) and large (95 ± 0.38 µm) adipocytes, followed by mRNA expression analysis and identification of genes that were expressed more than four-fold in large adipocytes versus small ones [18,31] (Supplemental Table S3). From this set, three additional F13A1and weight-correlated genes were identified (SELE, IL8, and LPCAT1). Fourth, we searched the literature for genes that have been published to regulate adipocyte size. A literature search ('adipocyte size' and 'regulation' and 'gene') resulted in 835 papers, from which 17 reported on genes with strong links to regulation of adipocyte size (associations were not included) (Supplemental Table S4). Only one out of these genes was found in the set of 182 genes (REPIN1). Thus, a total of 383 'adipocyte size' genes were screened, which resulted in a final set of 47 gene transcripts that showed significant correlation to ∆F13A1 and were differentially expressed in the Heavy vs. Lean twin. Classical regulators of adipogenesis and lipid storage, PPARG, CEBPA, CEBPB, and PLIN1, did not show correlation. The set of 47 genes are listed in Table 2. From this set, 12 genes showed strong negative correlation with ∆Weight-, ∆F13A1-, and ∆Adipocyte Diameter. The remaining 35 genes were positive correlators.

Figure 2.
Transcriptome-wide association study (TWAS) strategy to investigate the association of F13A1 with acquired excess weight and adipocyte size in weight-discordant monozygotic twins. To examine what weight gain and adipocyte size-associated pathways F13A1 is linked to, we performed a transcriptome-wide association study (TWAS) of differentially expressed and correlated genes (Heavy-Lean ΔF13A1 vs. Heavy-Lean Δtranscriptome). The initial screen identified 182 genes whose differential expression showed significant linear correlation with F13A1 (r 2 > 0.5 and p < 0.05) and significant differential expression between the Heavy and Lean co-twins (p < 0.05) [32]. This set was screened for previously published adipocyte size correlating (positive and negative) genes in the MZ twins, some of which were shared with this study. This resulted in no additional genes. The set was then screened for 14 reported 'large adipocyte genes' and 17 genes from the literature that have been published to regulate adipocyte size (PubMed search ('adipocyte size' and 'regulation' and 'gene')). All screened genes are listed in Supplemental Tables S1-S4. The resulting final set of 47 genes that correlated with F13A1, adipocyte diameter, and were differentially expressed in the co-twin pairs are listed in Table 2. To examine what weight gain and adipocyte size-associated pathways F13A1 is linked to, we performed a transcriptome-wide association study (TWAS) of differentially expressed and correlated genes (Heavy-Lean ∆F13A1 vs. Heavy-Lean ∆transcriptome). The initial screen identified 182 genes whose differential expression showed significant linear correlation with F13A1 (r 2 > 0.5 and p < 0.05) and significant differential expression between the Heavy and Lean co-twins (p < 0.05) [32]. This set was screened for previously published adipocyte size correlating (positive and negative) genes in the MZ twins, some of which were shared with this study. This resulted in no additional genes. The set was then screened for 14 reported 'large adipocyte genes' and 17 genes from the literature that have been published to regulate adipocyte size (PubMed search ('adipocyte size' and 'regulation' and 'gene')). All screened genes are listed in Supplemental Tables S1-S4. The resulting final set of 47 genes that correlated with F13A1, adipocyte diameter, and were differentially expressed in the co-twin pairs are listed in Table 2.  Gene enrichment analysis for the 47 genes, with GOnet and Panther gene ontology tools, resulted in a total of 61 GOterms for Biological Process (BP), 13 Cellular Component (CC), and no terms for Molecular Functions. Figure 3 shows a selection [28] of the most significant, over-represented Biological Process GOterms and all 13 Cellular Component terms. The full list of GOterms is listed in Supplemental Tables S5 and S6. Further cluster analysis ( Figure 4) for the GOterms using the GOnet tool revealed three clear functional clusters related to immune response, inflammation, and cytokine response and signaling. A detailed examination of each of the clusters revealed central GOterms, suggesting inflammatory response to external factors via neutrophil and leucocyte activation as well as cytokine secretion (Figures 5-7). Each of the Biological Process and Cellular Component terms and some relevant genes are discussed below in more detail.
terms for Molecular Functions. Figure 3 shows a selection [28] of the most significant, overrepresented Biological Process GOterms and all 13 Cellular Component terms. The full list of GOterms is listed in Supplemental Tables S5 and S6. Further cluster analysis (Figure 4) for the GOterms using the GOnet tool revealed three clear functional clusters related to immune response, inflammation, and cytokine response and signaling. A detailed examination of each of the clusters revealed central GOterms, suggesting inflammatory response to external factors via neutrophil and leucocyte activation as well as cytokine secretion (Figures 5-7). Each of the Biological Process and Cellular Component terms and some relevant genes are discussed below in more detail. . Gene enrichment analysis. Selected over-represented GOterms among the 47 genes whose differential expression correlated with F13A1, adipocyte diameter, and showed significantly altered expression between heavy and lean co-twins. Gene ontology analysis was done using GOnet and confirmed with Panther. A total of 61 GOterms were found for Biological Process, and 28 highly relevant are represented here. All 13 Cellular Component terms are shown. No terms for Molecular Functions arose in the search. Significance of the terms is represented as -log10 (p-value) and number of genes in each GOterm is indicated with the bar. Full set of the GOterms with p-values, as well as false discovery rate adjusted p-values and list of genes can be found in Supplemental Tables S5 and S6. Figure 3. Gene enrichment analysis. Selected over-represented GOterms among the 47 genes whose differential expression correlated with F13A1, adipocyte diameter, and showed significantly altered expression between heavy and lean co-twins. Gene ontology analysis was done using GOnet and confirmed with Panther. A total of 61 GOterms were found for Biological Process, and 28 highly relevant are represented here. All 13 Cellular Component terms are shown. No terms for Molecular Functions arose in the search. Significance of the terms is represented as -log 10 (p-value) and number of genes in each GOterm is indicated with the bar. Full set of the GOterms with p-values, as well as false discovery rate adjusted p-values and list of genes can be found in Supplemental Tables S5 and S6.

Discussion
Adipose tissue expansion through hypertrophy, where large adipocytes predominate in the expanding tissue, correlates with adipose tissue low-grade systemic inflammation and worse metabolic consequences of obesity [33][34][35][36]. The mechanisms and pathways that precede and follow the hypertrophic expansion and inflammation are numerous. Accumulating data points to sequela that involves free fatty acid (FFA)-mediated initiation of adipocyte hypertrophy, which is followed by initiation of innate immune response by the hypertrophic adipocytes [34]. Previous studies have

Discussion
Adipose tissue expansion through hypertrophy, where large adipocytes predominate in the expanding tissue, correlates with adipose tissue low-grade systemic inflammation and worse metabolic consequences of obesity [33][34][35][36]. The mechanisms and pathways that precede and follow the hypertrophic expansion and inflammation are numerous. Accumulating data points to sequela that involves free fatty acid (FFA)-mediated initiation of adipocyte hypertrophy, which is followed by initiation of innate immune response by the hypertrophic adipocytes [34]. Previous studies have linked two TGM family members, TGM2 and F13A1, to adipose tissue biology and inflammation [25,26,[28][29][30][37][38][39]. In this study, we have investigated the differential expression of the whole TGM family in adipose tissue of MZ twins that are discordant in weight (Heavy-Lean twin pairs). The MZ twins are a unique model of acquired weight gain where background involving genetic, age, sex, and early environmental influences can be controlled. The twins' adipocity parameters and metabolism are well characterized and include adipocyte morphology, mitochondrial, fibronolytic and coagulatory, and inflammatory status [40][41][42][43][44][45][46][47]. We report that F13A1 is the sole TGM family member that shows significantly altered, and in this case increased, expression in the heavier twin. Despite robust data from mouse studies linking TG2 to adipose tissue function in obesity [25,26], TGM2 showed no altered expression between the Heavy-Lean co-twins. Further studies in larger cohorts of more severely obese individuals could shed more light into the role of TG2 in human adipose tissue pathology.
The presence of large versus small adipocytes in adipose tissue in obesity has been associated with worse metabolic outcome [18]. Our previous work linked increase in F13A1 in adipose tissue to increased adipocyte size [30]. The aim of this current study was to further examine through which pathways F13A1 associates to large adipocytes. This was done by screening the pool of the previously identified 182 F13A1and weight-correlating genes [30] in the adipocyte-enriched fraction for: (1) genes that have been reported to associate positively or negatively to adipocyte size [17], and (2) genes that are expressed significantly more in larger adipocytes or demonstrated in the literature to regulate adipocyte size [18,31]. Here, we demonstrate tight association of F13A1 to 47 'adipocyte size' genes that via gene enrichment analysis linked to Biological Processes involving response to external stimulus, secretion, exocytosis, vesicle-mediated transport, cell surface receptor signaling, neutrophil activation, and response to interleukin 1. The GOterms for Cellular Components suggested association of F13A1 in lysosome, secretory vesicles, plasma membrane rafts, and extracellular space. Network and cluster analysis of the GOterms link these to immune response/inflammation and cytokine response (Figure 4), which jointly suggest that F13A1 may be involved in pro-inflammatory response from large adipocytes. In our previous study, we discussed several genes that are also among the current 47 genes and they were associated with adipocyte oxidative stress, extracellular matrix (ECM) organization and remodeling, angiogenesis, and macrophages [30]. Associations to all the above processes suggests that F13A1 may be involved throughout in molecular events of the tissue expansion.
The adipocyte-enriched fraction used in this study was prepared from SAT biopsies by a method involving collagenase digestion followed by collecting of the released, buoyant adipocytes. This method is considered to separate adipocytes from the stromal vascular fraction that contain pre-adipocytes, mesenchymal stem cells, endothelial and vascular smooth muscle cells, and immune cells [48], however, other cell types do co-purify and contribute to the transcripts observed. Thus, the preparation enriches adipocytes, but is not void of other cell types. It has become apparent that pure adipocyte preparations require flow cytometry sorting and is more difficult than previously considered [49]. The other cell types have likely co-purified and may include macrophages whose marker ITGAM (Integrin αM; CD11b) is present also in this final set of 47 genes. The preparation also likely includes endothelial cells, which as suggested by the expression of genes like SELE (Selectin E) [30], is a product of cytokine-stimulated endothelial cells and mediates endothelial cell adhesion to leucocytes [50]. Macrophages and endothelial cells also both produce FXIII-A enzyme [3,[51][52][53][54][55][56][57][58][59]. Our major future aims include detailing the cell types that are responsible for F13A1/F13a1 production at different phases of adipose tissue expansion, which is critical to understanding where in the inflammatory sequela F13A1/F13a1 may be involved.  Of the 47 gene transcripts that correlate with weight, F13A1, and adipocyte size, we have discussed four in our previous study (CYBB, CTSS, ITMG, and VCAN) [30]. Below, we will focus mostly on the new ones identified here. The most numerous GOterms for Biological Processes were associated with neutrophil and leucocyte activation ( Figure 5). One of the highest F13A1 correlating genes in this set was PYCARD (r 2 = 0.89) (Figure 5B), which encodes for ASC (apoptosis-associated speck-like protein containing a CARD). PYCARD/ASC is made up by macrophages, where it activates inflammasome via acting as an adaptor protein between Nlrp3 and caspase-1 [60]. Caspase-1 is critical for proteolytic activation and secretion of pro-inflammatory cytokines IL1β and IL-18 [61,62]. The presence of FFAs in adipose tissue can induce activation of inflammasome [61] and its essential role in metabolic disturbances has become clear [61,63,64]. A very recent work demonstrated that the absence of these three critical inflammasome components, Nlrp3, ASC, and caspase-1, in mice protects mice from HFD-induced obesity, insulin resistance, and adipocyte hypertrophy [64]. The Heavy-Lean increase in ∆IL1B also shows strong linear correlation to ∆F13A1 (r 2 = 0.62) ( Figure 5B). IL1β is one of the major pro-inflammatory cytokines whose role in metabolic dysfunction is well documented [65].   Neutrophils are the first responders of acute inflammatory reaction and are the first cells to be recruited to the site of inflammation [66][67][68][69]. They are the most populous but short-lived cell type in early inflammation and are capable of producing large quantities of cytokines and chemokines, including TNF-α, IL-1β, and IL-8, which can initiate the second wave of immune cells, such as infiltration of macrophages and lymphocytes [66][67][68][69]. The role of neutrophils in adipose tissue inflammation has also become apparent [33,70,71]. The links of F13A1 to both neutrophil and leucocyte activation ( Figure 5) supports a role in early inflammation as well as perpetuation of the pro-inflammatory milieu. The genes contributing to GOterms that link F13A1 to neutrophil activation are, among others, CYBB (NADP oxidase) (Nox2), which is made by neutrophils and macrophages and responsible for generation of reactive oxygen species and oxidative stress, which in turn can activate neutrophils and neutrophil extracellular trap formation [72]. In adipose tissue, CYBB/Nox2 contributes to adipose tissue inflammation and development of insulin resistance [73]. The presence of TNFSF13B (TNF Superfamily Member 13b) (also known as BAFF) in the set of 47 weight and F13A1 correlating genes also supports neutrophil links ( Figure 5B). TNFSF13B/BAFF is a neutrophil-derived Neutrophils are the first responders of acute inflammatory reaction and are the first cells to be recruited to the site of inflammation [66][67][68][69]. They are the most populous but short-lived cell type in early inflammation and are capable of producing large quantities of cytokines and chemokines, including TNF-α, IL-1β, and IL-8, which can initiate the second wave of immune cells, such as infiltration of macrophages and lymphocytes [66][67][68][69]. The role of neutrophils in adipose tissue inflammation has also become apparent [33,70,71]. The links of F13A1 to both neutrophil and leucocyte activation ( Figure 5) supports a role in early inflammation as well as perpetuation of the pro-inflammatory milieu. The genes contributing to GOterms that link F13A1 to neutrophil activation are, among others, CYBB (NADP oxidase) (Nox2), which is made by neutrophils and macrophages and responsible for generation of reactive oxygen species and oxidative stress, which in turn can activate neutrophils and neutrophil extracellular trap formation [72]. In adipose tissue, CYBB/Nox2 contributes to adipose tissue inflammation and development of insulin resistance [73].
The presence of TNFSF13B (TNF Superfamily Member 13b) (also known as BAFF) in the set of 47 weight and F13A1 correlating genes also supports neutrophil links ( Figure 5B). TNFSF13B/BAFF is a neutrophil-derived cytokine that supports B lymphocyte survival. B lymphocytes can tune adipose tissue inflammation via secreting cytokines and immunoglobulins, and they contribute to metabolic failure of adipose tissue [63,74,75]. BAFF is also produced by adipocytes [76]. The role of F13A1 in inflammation in many tissues is supported by vast literature and it is expressed by macrophages and dendritic cells [51,52,[77][78][79][80][81][82]. The mechanisms by which FXIII-A contributes to inflammation involve its fibrin-stabilizing function [83,84] and monocyte/macrophage adhesion [79,85]. It may also crosslink other substrates such as plasma fibronectin [28,29]. It is not known if F13A1 is produced by neutrophils themselves.
Cluster analysis of GOterms also revealed that F13A1 links to cytokine-mediated signaling pathways ( Figure 6). Here, we demonstrate two indicators that show that F13A1 is also negatively associated with important insulin modifiers, which is apparent from its negative correlation to PTPRF and STAT5A. The ∆F13A1/∆PTPRF linear regression graph in Figure 6B shows that the high increase in F13A1 correlated with a smaller increase in PTPRF. PTPRF encodes for Protein Tyrosine Phosphatase Receptor Type F (also known as LAR), which inhibits the insulin receptor activation [86,87]. PTPRF/LAR knockdown in adipocytes increases adipogenesis [88] and in mice, disturbances in glucose homeostasis [89]. This data interestingly links high F13A1 to promotion to insulin signaling, which we demonstrated in 3T3-L1 adipocytes [28]; however, in mice, the F13A1 absence improves insulin sensitivity [29]. Also, negative association to STAT5A ( Figure 6B) (Signal Transducer and Activator of Transcription 5A), a known promoter of adipogenesis and adipocyte differentiation, supports our previous cell culture results, where FXIII-A was required for pre-adipocyte proliferation but was rapidly downregulated during adipogenesis [28]. The link to large adipocytes suggests that it is also made later during hypertrophy, however this could also mean that large adipocytes induce F13A1 in other adipose tissue cells. Interestingly, recent advances in understanding the phenotypes and functionality of large, hypertrophic adipocytes in adipose tissue in obesity has shown that they likely function as antigen-presenting cells to active innate immunity/immune cells via MHCII and play an important role in the initiation of adipose tissue inflammation [20,90]. Mature adipocytes have also been demonstrated to express leucocyte chemo-attractants that via direct adipocyte-leucocyte interaction stimulate T-cell proliferation [91]. The correlation of F13A1 to HLA-DRA ( Figure 5B), an HLA class II histocompatibility complex component, found on the surface of antigen presenting cells, could suggest a link to large adipocytes specialized to initiation of inflammation.
The stimuli initiating F13A1 production, cell types producing it, and cellular location/externalization and activation in adipose tissue are major questions arising from this study. Here, the increase of F13A1 is linked to GOterms for cellular response to external stimuli and response to lipopolysaccharide (LPS), supporting the idea that F13A1 could be induced by those factors that initiate the inflammatory reaction. The stimuli that can initiate adipose tissue inflammation are complex and numerous [33,34] and include fibrosis and alterations in extracellular matrix components and stiffness [83,[92][93][94][95][96], apoptosis and necrosis of hypertrophic adipocytes [97], hypoxia from defective angiogenesis [24,98], free fatty acids (lipotoxicity) [21], and exogenous, as well as gut microbiota-derived LPS [36]. The GOterms for this cluster of Biological Processes are presented in Figure 7, and there are two genes of particular interest PLA2G16 and MAPK1.PLA2G16-phospholipase 2A-a lipolytic enzyme made in hypertrophic adipocytes that produces FFAs and induces hyperlipidemia [99,100]. Coregulation between PLA2G16 and F13A1 could thus imply a link to FFAs. Positive association with MAPK1 ( Figure 6B) is also worthy of mention as we demonstrated that in osteoblasts, extracellular collagen type I, also produced by adipocytes, is a major regulator of FXIII-A expression and externalization [15]. MAPK1 is also needed for lipolysis in rat and mouse adipocytes [101,102].
Our original screen contained a set of regulatory genes demonstrated to modulate adipocyte size. From the final set of genes, 46 of the 47 were large adipocyte-associated genes and only one, REPIN1, was a gene that has been demonstrated to regulate adipocyte size and whole-body insulin sensitivity [103]. The association between F13A1 and REPIN1 is negative (r = −0.77; r 2 = 0.70; p = 0.0002), i.e., higher ∆REPIN1 levels link to lower ∆F13A1. This and the positive association of ∆F13A1 to ∆Adipocyte Diameter and ∆Adipocyte Volume [30] strongly suggests that majority of F13A1 is a result of adipocyte hypertrophy-induced events in weight gain, rather than it being a gene that influences the adipocyte size. Detailed weight gain studies in mice can address when F13a1 transcription and protein production is initiated, when and how it is activated during adipose tissue expansion, and what are the consequences of its absence or its overexpression to adipocytes and adipose tissue inflammatory cells and processes. These studies are ongoing in our laboratory.
In conclusion, our study brings further evidence to the involvement of F13A1 in the pathological response of adipose tissue to expansion in weight gain in humans. The concepts revealed in our study suggest novel pathways linked to F13A1 and inflammation providing many new hypotheses to be explored that will hopefully show if FXIII-A could be a target to control inflammation of adipose tissue.

Weight-Discordant MZ Twins
This study involved 12 rare, healthy MZ twin pairs discordant for weight (intra-pair difference in acquired excess weight ∆BMI (body mass index) ≥ 3 kg/m 2 (males n = 2, females n = 10, aged 27, 7 ± 1.4 years)) which were identified from two population-based twin cohorts, FinnTwin16 (n = 2839 pairs) and FinnTwin12 (n = 2578 pairs) [104]. A detailed description of the twin material has been published previously [45,105]. The twins were healthy and were not on any regular medication, excluding contraceptives. Table 1 describes the clinical characteristics of the 12 twin pairs regarding weight and adipocity. This data table has been also presented in Reference [30]. All participants have given a written informed consent. The study design was approved by the Ethical Committee of the Helsinki University Central Hospital.

Adipose Tissue Biopsies, Adipocyte and Tissue Preparations, and RNA Extraction
Abdominal subcutaneous adipose tissue (SAT) biopsies were obtained surgically under the umbilicus under local anesthesia. Specimens were flash frozen in −80 • C liquid nitrogen. Adipocyte-enriched fraction was prepared by a previously reported protocol which involved adipose tissue digestion in 2% collagenase dissolved in DMEM/F-12 with supplements and 2% bovine serum albumin [17]. This was followed by adipocyte mRNA extraction using RNeasy Lipid Tissue Mini Kit (Qiagen, Nordic, Solletuna, Sweden) with a DNase I (Qiagen) digestion according to the manufacturer's instructions.

Adipocyte Morphology
Fresh SAT biopsy specimens were digested with 2% collagenase in DMEM/F12 with supplements [17] and 2% bovine serum albumin (as above) and were used for measurement and calculation of adipocyte size in all 12 discordant twin pairs, as before [17]. After collagenase treatment, the digests were washed and maintained in the above medium and images of live adipocytes were taken with a light microscope (Zeiss, Axioplan2) using 50× magnification. A minimum of 200 cells were measured for diameter using an image processing and analysis software ImageJ (ImageJ 1.42q/Java 1. 6.0_10 32-bit), as previously described [17].

Affymetrix Transcriptomics
A total of 500 ng of RNA from adipocyte-enriched fraction, isolated as described above, was used for gene expression analysis on an Affymetrix U133 Plus 2.0 array (Thermo Fischer, Santa Clara, CA, USA) according to the manufacturer's instructions and validated as described previously [45]. Hybridization, staining, and washing were performed using the Affymetrics Fluidics Station 450 and Hybridization oven 640 under standard conditions. Expression data was pre-processed with BioConductor software (open source, www.bioconductor.org) and the GC-RMA algorithm [106].

Transcriptome-Wide Association Study, Statistics, and Bioinformatics
Intra-pair values (∆Heavy-Lean) for twins' transcriptome (differential gene expression, ∆) were calculated and screened for correlation to intra-pair ∆F13A1 expression in adipocyte-enriched fraction using R Statistical Software (https://www.r-project.org/) [107]. Significance within the TGM family (TGM1-7 and F13A1) expression in twin pairs was calculated via analysis of variance (ANOVA) with Bonferroni post-test using Prism GraphPad software (release 8.0, San Diego, CA, USA). p-Values < 0.05 were considered significant. Pearson (r) correlation and r 2 parameters were obtained for ∆F13A1 and ∆Adipocyte Diameter, and those with r 2 > 0.5 and associated p-values < 0.05 were considered for further screening. Normal distribution of F13A1 mRNA expression values and adipocyte diameter values were assessed as before [30]. Significance of differential gene expression between the co-twins was tested with a paired T-test and associated p-values < 0.05 were considered significant. Further linear regression analysis was performed for selected genes with significant associations to ∆F13A1 with Prism GraphPad software (release 8.0, San Diego, CA, USA). Here, p-values were considered significant as follows: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Gene Ontology enrichment analysis and GOterm cluster analysis for over-represented genes from transcriptome screening was performed using Panther, GOnet, and Gorilla online tools. Significance for GOterms is represented in False Discovery Rate (FDR) adjusted p-values.