Fetuin B Is Related to Cytokine/Chemokine and Insulin Signaling in Adipose Tissue and Plasma in Humans

Abstract

Context

Fetuin B is a steatosis-responsive hepatokine that induces glucose intolerance in mice. Recently, we found that fetuin B in white adipose tissue was positively associated with peripheral insulin resistance in mice and a small study population, possibly through a fetuin B–induced inflammatory response in adipocytes.

Objective

This translational study aimed to investigate the link between plasma fetuin B and the adipose tissue transcriptome and plasma proteome in a large cohort of humans.

Methods

Continuous linear regression analysis in R was applied to investigate the link between plasma fetuin B and the adipose tissue transcriptome (n = 207) and plasma proteome (n = 558) in humans, after adjustment for sex, age, and study center (model 1); model 1 + BMI (model 2); and model 2 + insulin sensitivity (Matsuda index) (model 3).

Results

Plasma fetuin B was associated with more than 100 genes in white adipose tissue, belonging to pathways related to cytokine/chemokine signaling (models 1 and 2) and insulin signaling (all models), and with more than 146 plasma proteins involved in pathways related to metabolic processes and insulin signaling (all models).

Conclusion

Plasma fetuin B is related to adipose tissue genes and plasma proteins involved in metabolic processes and insulin signaling. Our findings provide evidence for the involvement of white adipose tissue in fetuin B-induced insulin resistance.

The importance of interorgan crosstalk in the regulation of whole-body metabolism is well recognized (1). Interorgan crosstalk encompasses the secretion of products into the circulation and the transport of these products to target organs, thereby influencing metabolic processes in these organs. The liver plays a major role in maintaining whole-body glucose homeostasis via interorgan crosstalk, in part through the release of hepatokines. These hepatokines exert autocrine, paracrine, and/or endocrine effects, influencing lipid and glucose metabolism in the liver, as well as in peripheral tissues such as skeletal muscle and adipose tissue (1). In 2015, we demonstrated in mice that the secretion of hepatokines changes with the development of liver steatosis compared with a lean liver, resulting in insulin resistance in skeletal muscle (2). We also identified fetuin B as a steatosis-responsive hepatokine that is increased in individuals with type 2 diabetes and that causes glucose intolerance in mice (2). Since then, several studies have reported that fetuin B is increased in individuals with disturbances in glucose homeostasis and insulin resistance, including adults with obesity and type 2 diabetes, women with gestational diabetes, and women with polycystic ovary syndrome (PCOS) (3-6).

Recently, we performed a series of in vivo and in vitro experiments in cells, mice, and humans to obtain a better insight into the mechanisms of action of fetuin B. Interestingly, our findings suggested that circulating fetuin B is taken up in white adipose tissue, and that fetuin B in adipose tissue was positively associated with peripheral insulin resistance both in mice and humans (7). Furthermore, we found that fetuin B induced an inflammatory response in cultured adipocytes, which may contribute to fetuin B–induced peripheral insulin resistance (7). In the present study, we aimed to further investigate the putative role of fetuin B in relation to glucose homeostasis in humans, and the role of abdominal subcutaneous adipose tissue herein. Specifically, we performed continuous linear regression analysis in R to investigate the link between plasma fetuin B concentration with alterations in the adipose tissue transcriptome profile and the plasma proteome profile in a large cohort.

Materials and Methods

Study Participants

Cohort 1

To investigate the relationship between plasma fetuin B and abdominal subcutaneous adipose tissue fetuin B content, we used data from two studies that were previously performed in our laboratory (8, 9). Only participants from whom we had paired data of plasma and abdominal subcutaneous adipose tissue fetuin B were included in the analysis (n = 54). Fetuin B levels in plasma and adipose tissue infranatant were determined using a fetuin B Human enzyme-linked immunosorbent assay (ELISA; RD191172200R, Biovendor; discontinued) according to the manufacturer's instructions, as explained previously (7). This cohort included men and women with a wide range in age (21-68 years) and body mass index (BMI) (25-39), and 2-hour glucose levels that could be categorized as normal glucose tolerant or impaired glucose tolerant. Study participants provided written informed consent for participation in this study, and experimental protocols were approved by the local medical ethical committee of Maastricht University (https://ClinicalTrials.gov, identifier: NCT02241421 and NCT02381145). All procedures were performed according to the Declaration of Helsinki (revised version, October 2008). Participant characteristics are reported in Table 1.

Table 1.

Clinical characteristics of the study participants of cohort 1

	Mean ± SD	Range
Sex, male/female	44/10	—
Age, y	52.5 ± 12.2	21-68
BMI	30.8 ± 3.1	25.4-38.5
Fasting glucose, mmol/L	5.8 ± 0.7	4.5-7.5
Fasting insulin, mU/L	17.2 ± 6.1	8.9-30.2
HOMA-IR, AU	3.7 ± 2.0	0.8-8.8
Fasting FFAs, µmol/L	613.8 ± 151.4	343-960
Triglycerides, µmol/L	1086.0 ± 449.2	454-2343
2-h glucose, mmol/L	6.9 ± 2.0	3.7-11.2
Plasma fetuin B, µg/mL	1.8 ± 0.6	0.5-3.8
White adipose tissue fetuin B, AU	2330 ± 823	1038-4106

Abbreviations: AU, arbitrary units; BMI, body mass index; FFA, free fatty acid; HOMA-IR, homeostatic model assessment of insulin resistance.

Cohort 2

For the remainder of the study, baseline data was used from the Diet, Obesity, and Genes (DiOGenes) study, a large European, multicenter, dietary intervention study (NCT00390637) (10). The original study aimed to enroll approximately 450 families across 8 European centers, with all members younger than 65 years, and with at least 1 adult member having overweight or obesity. We included only adult study participants for whom plasma proteome data were available. This cohort consisted of 558 participants (209 men, 349 women) with a broad range in age (16-63 years), BMI (25.6-52), and glucose tolerance (2-hour glucose levels 2.2-16.6 mmol/L), representing a fairly typical cross-section of the general population, albeit slightly skewed toward individuals with higher weight and poorer health. Participant characteristics are reported in Table 2, and more details about this study have been described previously (10). The ethics committee of each center/country approved the protocol. The committees included the medical ethical committee from Maastricht University, the Netherlands; Copenhagen ethical research committee, Denmark; Bedfordshire local research ethics committee, Luton and Dunstable Hospital NHS Trust, UK; ethics committee of the Faculty Hospital, Prague University, Czech Republic; ethical committee by NMTI, Sofia, Bulgaria; ethical committee University Potsdam, Germany; ethical committee Medical University, Navarra, Spain; scientific council Heraklion general university hospital, Heraklion, Greece; and Commission Cantonale d’ éthique de la recherche sur l’ être humain, Canton de Vaud, Switzerland. Furthermore, the protocol was in accordance with the Declaration of Helsinki. All study participants provided written consent.

Table 2.

Clinical characteristics of the study participants of cohort 2

	Entire cohort (n = 558)		Subgroup (n = 207)
	Mean ± SD	Range	Mean ± SD	Range
Sex, male/female	209/349	—	97/110	—
Age, y	41.6 ± 6.2	16-63	42.1 ± 6.2	24-63
BMI	34.4 ± 4.8	25.6-52.0	34.5 ± 4.7	27.1-47.7
Fasting glucose, mmol/L	5.1 ± 0.7	3.4-9.4	5.1 ± 0.7	2.7-8.7
Fasting insulin, mU/L	11.9 ± 9.3	2.2-141	11.6 ± 10.8	2.2-141.0
HOMA-IR, AU	3.2 ± 2.5	0.6-37.1	3.2 ± 3.0	0.6-37.1
Fasting FFAs, µmol/L	642 ± 344	160-2614	663 ± 389	162-2536
Triglycerides, µmol/L	1.4 ± 0.7	0.4-3.9	1.4 ± 0.6	0.4-3.7
2-h glucose, mmol/L	6.7 ± 2.2	2.2-16.6	6.7 ± 2.2	2.5-15.7
Matsuda index	5.1 ± 3.2	0.4-19.9	5.0 ± 3.1	0.4-19.9
CRP, AU	4.1 ± 3.8	0.1-25.0	4.0 ± 3.5	0.3-22.0
Leptin, AU	12.7 ± 0.8	10.4-14.4	12.7 ± 0.8	10.7-14.3
MCP1, AU	7.8 ± 0.7	6.7-14.3	7.7 ± 0.5	6.7-10.9
Adiponectin, AU	10.3 ± 0.5	8.6-12.3	10.4 ± 0.5	8.9-12.3
PAI1, AU	9.6 ± 1.0	6.6-15.6	9.6 ± 0.9	6.6-11.5
TNF-α, AU	7.7 ± 0.8	6.8-15.7	7.6 ± 0.6	6.8-12.8
Plasma fetuin B, AU	11.9 ± 0.4	9.9-13.8	11.9 ± 0.4	10.7-13.8

Abbreviations: AU, arbitrary units; BMI, body mass index; CRP, C-reactive protein; FFA, free fatty acid; HOMA-IR, homeostatic model assessment of insulin resistance; MCP1, monocyte chemoattractant protein 1; PAI1, plasminogen activator inhibitor-1; TNF-α, tumor necrosis factor α.

Oral Glucose Tolerance Test

Participants from cohort 2 underwent a 5-point oral glucose tolerance test (OGTT). Following an overnight fast, venous blood samples were collected before ingesting a 75-g glucose drink and at time point 30, 60, 90, and 120 minutes. Plasma was stored at −80 °C until analysis to assess concentrations of glucose and insulin.

Adipose Tissue Biopsy

In a subset of cohort 2, fasted abdominal subcutaneous adipose tissue biopsies were taken under local anesthesia. Biopsies were obtained 6 to 8 cm laterally from the umbilicus under local anesthesia. Subsequently, the biopsy specimens were rapidly frozen in liquid nitrogen and stored at −80 °C until further analysis. Total RNA was extracted and analyzed from these subcutaneous adipose tissue biopsy specimens, following the procedures outlined in detail previously (11). We included participants only for whom we had both plasma fetuin B levels and adipose tissue transcriptome profiles (n = 207). Participant characteristics are reported in Table 2. There were no statistically significant differences in patient characteristics between the entire cohort 2 and the subset.

Proteomics and Transcriptomics Analysis

Plasma proteomics analysis was previously performed in cohort 2 (n = 558). Specifically, a panel of 1128 plasma proteins was measured, using the SOMAscan aptamer-based technology from the SomaLogic proteomics platform (12). Proteomics data were log₂-transformed with an offset of 1 before downstream analysis. The white adipose tissue transcriptomics data were obtained in a subset of cohort 2 (n = 207) using Illumina HiSeq 2000. A total of 54 043 genes were measured in abdominal subcutaneous adipose tissue. Only protein coding genes were included in the analysis. In total, 18 821 protein-coding genes were detected above background level. RNA expression data are available from the Gene Expression Omnibus under accession number GSE95640. Other data are unsuitable for public deposition due to ethical restrictions and privacy of participant data. Data are available from the corresponding author for any interested researcher who meets the criteria for access to confidential data and on reasonable request.

Statistical Analysis

Plasma fetuin B levels of cohort 2 were derived from the proteomic analysis. To investigate the link between plasma fetuin B and white adipose tissue fetuin B concentration in cohort 1, the Pearson correlation coefficient was calculated using SPSS for Windows version 27.0. In cohort 2, first linear regression analysis was used to relate plasma fetuin B concentration with measures of glucose homeostasis and inflammation. Since fetuin B expression is higher in women compared with men and increases with age (13), we included sex and age as covariates in the statistical analyses. Specifically, we created two models with increasing complexity to evaluate the contribution of plasma fetuin B to the dependent variables. In model 1, data were adjusted for sex, age, and study center; model 2 included an additional adjustment for BMI (model 1 + BMI). Standardized β coefficients (stdβ) and 95% CIs are reported in a forest plot. The data were analyzed using SPSS for Windows version 27.0 and statistical significance was set at P less than .05. Participant characteristics are reported as mean ± SD (see Table 2). Figures were generated using GraphPad Prism (version 10).

Differential Gene Expression and Protein Abundance Analysis

To identify (i) the genes in white adipose tissue and (ii) proteins in plasma that are associated with plasma fetuin B levels (P < .01, no adjustment for multiple testing), continuous linear regression analysis was performed in R (version 4.3.1), using the R-packages “DESeq2” and “Limma,” respectively (14, 15). Again, data were corrected for sex, age, and study center (model 1), with further adjustment for BMI (model 2: model 1 + BMI). Notably, to investigate whether plasma fetuin B was linked with genes and proteins independently of insulin resistance, we created a third model that included correction for insulin sensitivity (Matsuda index) (model 3: model 2 + Matsuda). An overrepresentation analysis was then performed on the proteins and genes that were significantly associated with plasma fetuin B, using the ClusterProfiler package in R (version 4.8.2) (16, 17). To perform the pathway analysis on the proteomics data set, the SomaLogic identifiers were translated to Entrez IDs by using the aptamers data from SomaLogic. Transcriptome and proteome pathways were considered significantly enriched when the unadjusted P value was less than .05 and with a differential gene or differential protein ratio greater than 10%. For the pathway analysis, both the KEGG (Kyoto Encyclopedia of Genes and Genomes) and WikiPathways database have been consulted (18, 19). Pathways with similar names and contents were considered redundant, so only one representative pathway was selected for downstream analyses. All data related to differential gene expression and protein abundance analysis has been added as supplemental tables to figshare (DIO: 10.6084/m9.figshare.27993833) (20).

Transcriptome and Proteome Data Overlap

To investigate overlap between fetuin B-associated genes in white adipose tissue and fetuin B–associated proteins in plasma, we studied the genes and proteins that were obtained with model 2, using a P value threshold of .05. Additionally, we restricted our selection to genes that are known to translate into “secreted proteins.” We labeled genes as “secreted proteins” when they had either the GO term annotation “GO:0005576” or “GO:0005615,” using the Biomart database, or when the gene was labeled in the Uniprot database as Uniprot proteins with “secreted” as the official key word. We then converted all gene IDs and protein IDs in our data sets to their corresponding Ensembl Gene ID using Biomart. This enabled us to match the list of potentially secreted genes with our proteomics data.

Results

Plasma Fetuin B Concentration Is Correlated With White Adipose Tissue Fetuin B Concentration

Based on our previous findings (7), preferentially abdominal subcutaneous adipose tissue fetuin B protein levels should be used to investigate the association with different metabolic parameters. However, since abdominal subcutaneous adipose tissue fetuin B protein levels were not available in the large study population (cohort 2), we used cohort 1 to investigate whether plasma fetuin B is as good a surrogate marker for fetuin B levels in adipose tissue. Indeed, abdominal subcutaneous adipose tissue fetuin B was found to be strongly associated with plasma fetuin B (r = 0.64; P < .0001; Fig. 1), independent of sex, age, and BMI (data not shown), allowing us to use plasma fetuin B levels for the analysis in cohort 2.

Figure 1.

Plasma fetuin B concentration is associated with abdominal subcutaneous adipose tissue fetuin B concentration. For correlative analysis between plasma fetuin B and abdominal subcutaneous adipose tissue fetuin B, data are used from two previous cohort studies performed in our laboratory (8, 9).

Plasma Fetuin B Concentration Is Negatively Associated With Glucose Tolerance and Insulin Sensitivity in Humans

We used linear regression analysis to examine whether plasma fetuin B concentration is related to glucose homeostasis in a large cohort of well-phenotyped individuals with overweight or obesity. Fetuin B was not associated with most of the measures taken in the fasted condition, including fasting glucose, fasting insulin, homeostatic model assessment of insulin resistance (HOMA-IR), or fasting free fatty acids (FFAs) (Fig. 2). In contrast, plasma fetuin B was associated with measures related to postprandial glucose homeostasis. Specifically, fetuin B was positively associated with 2-hour glucose levels, area under the curve (AUC) glucose, and AUC insulin, and negatively associated with the Matsuda index, a measure of whole-body insulin sensitivity. These results were obtained after adjustment for sex, age, and study center (model 1), and also after additional adjustment for BMI (model 2) (see Fig. 2). Furthermore, we found a positive association between plasma fetuin B levels and fasting triglyceride levels as well as several inflammatory parameters, including C-reactive protein (CRP) and monocyte chemoattractant protein 1 (MCP-1) (see Fig. 2). Together, these data demonstrate that plasma fetuin B concentration is negatively associated with glucose tolerance and whole-body insulin sensitivity and positively related to systemic low-grade inflammation in humans.

Figure 2.

Forest plot illustrating the association of plasma fetuin B with fasting and postprandial parameters (n = 558). Data are corrected for sex, age and study center (model 1; indicated by ●) and model 1 + BMI (model 2; indicated by ■). Reported is the standardized β coefficient with the 95% CI. Abbreviations: AUC glucose, area under the curve for glucose during oral glucose tolerance test; AUC insulin, area under the curve for insulin during oral glucose tolerance test; CRP, C-reactive protein; FFA, free fatty acid; HOMA-IR, homeostatic model assessment of insulin resistance; MCP1, monocyte chemoattractant protein 1; PAI1, plasminogen activator inhibitor-1; TNF-α, tumor necrosis factor α. *P less than .05; **P less than .01; ***P less than .001.

Plasma Fetuin B Concentration Is Associated With Genes in White Adipose Tissue That Are Involved in Cytokine/Chemokine and Insulin Signaling

Next, the link between plasma fetuin B concentration and the transcriptome profile in abdominal subcutaneous adipose tissue was investigated. We found that 101 genes were significantly associated with plasma fetuin B using model 1 (P < .01; correcting for sex, age, and study center); 109 genes with model 2 (P < .01; model 1 + BMI); and 101 genes with model 3 (P < .01; model 2 + Matsuda) (supplemental tables (20)). To gain insight into the biological function of the genes that were significantly associated with plasma fetuin B, pathway enrichment analysis was performed. Pathways that were significantly enriched in both model 1 and model 2 include “viral protein interaction with cytokine and cytokine,” “phospholipase D signaling pathway,” “cytokine-cytokine receptor interaction,” and “chemokine signaling pathway” (Fig. 3 and supplemental tables (20)). Notably, there were 2 pathways that were significantly enriched in all 3 models, which include “focal adhesion,” “PI3K-Akt-mTOR-signaling pathway,” and “PI3K-Akt signaling pathway” (see Fig. 3 and supplemental tables (20)). All pathways that were significantly enriched both in model 1 and model 2 and their associated proteins are shown in Fig. 3. A complete overview of all pathways identified is presented in Table 3.

Figure 3.

Visualization of adipose tissue genes and pathways that were significantly enriched with both model 1 and model 2, indicating links between genes and pathways. Each square represents a separate pathway. Genes are depicted as circles, in which red indicates a positive link with plasma fetuin B and blue indicates a negative association. Log fold change of individual genes are encoded in color intensity.

Table 3.

Enriched pathways-based sets of plasma fetuin B-associated genes in adipose tissue

	Model 1	Model 2	Model 3
Viral protein interaction with cytokine and cytokine receptor	x	x
Focal adhesion: PI3K-Akt-mTOR-signaling pathway	x	x	x
G coupled–protein receptor, other	x
PI3K-Akt signaling pathway	x	x	x
Calcium signaling pathway	x		x
Phospholipase D signaling pathway	x	x
Cytokine-cytokine receptor interaction	x	x
Chemokine signaling pathway	x	x
Adrenergic signaling in cardiomyocytes		x
Cell lineage map for neuronal differentiation			x
Hypertrophic cardiomyopathy			x
Dilated cardiomyopathy			x
Neuroinflammation and glutamatergic signaling			x
MAPK signaling pathway			x
Focal adhesion			x

Data were analyzed in models with increasing complexity: model 1: adjustment for center, sex, and age; model 2: model 1 + adjustment for BMI; model 3: model 2 + adjustment for Matsuda.

Abbreviations: Akt, protein kinase B; BMI, body mass index; MAPK, mitogen-activated protein kinase; mTOR, mechanistic target of rapamycin.

Plasma Fetuin B Concentration Relates to Plasma Proteins Involved in Metabolic Pathways

To provide further insight into the link between plasma fetuin B and metabolic impairments, we investigated the link between plasma fetuin B and the plasma proteome profile. In total, 1128 plasma proteins were measured, of which 158 proteins were significantly associated with plasma fetuin B using model 1 (correcting for sex, age, and study center); 162 proteins using model 2 (model 1 + BMI); and 146 proteins using model 3 (model 2 + Matsuda) (supplemental tables (20)). Pathway enrichment analysis showed 14 distinct pathways that were significantly enriched both in model 1 and 2 (supplemental tables (20)), of which 13 pathways remained statistically significant after additional correction for insulin resistance (model 3) (supplemental tables (20)). Among others, these pathways were related to the complement system and coagulation cascades, PI3-Akt signaling, cytokine signaling, and a number of inflammatory pathways (Fig. 4). Relevant pathways and their associated proteins are shown in Fig. 4. A complete overview of all pathways identified is presented in Table 4.

Figure 4.

Visualization of plasma proteins and relevant pathways that were significantly enriched, indicating links between plasma proteins and pathways in abdominal subcutaneous adipose tissue. Each square represents a separate pathway. Plasma proteins are depicted as circles, in which red indicates a positive correlation with plasma fetuin B and blue indicates a negative correlation. Log fold change of individual genes are encoded in color intensity.

Table 4.

Enriched pathways-based sets of plasma fetuin B-associated proteins in plasma

	Model 1	Model 2	Model 3
Complement and coagulation cascades	x	x	x
Network map of SARS CoV 2 signaling	x	x	x
Endochondral ossification	x	x	x
Folate metabolism	x	x
Hippo signaling regulation	x	x	x
PI3K Akt signaling	x	x	x
Cytokine cytokine receptor interaction	x	x	x
Pleural mesothelioma	x	x	x
Focal adhesion PI3K-Akt-mTOR signaling	x	x	x
VEGFA VEGFR2 signaling	x	x	x
MAPK signaling pathway	x	x	x
Proteoglycans in cancer	x	x	x
Ras signaling pathway	x	x	x
Rap1 signaling pathway	x	x	x

Data were analyzed in models with increasing complexity: model 1: adjustment for center, sex, and age; model 2: model 1 + adjustment for BMI; model 3: model 2 + adjustment for Matsuda.

Abbreviations: Akt, protein kinase B; BMI, body mass index; MAPK, mitogen-activated protein kinase; mTOR, mechanistic target of rapamycin; VEGFA, vascular endothelial growth factor A; VEGFR2, vascular endothelial growth factor receptor 2.

Overlap Between Transcriptome and Proteome Data

To determine how many of the fetuin B–associated genes in abdominal subcutaneous adipose tissue were secreted proteins, and how many of these were also found in plasma to be significantly associated with fetuin B, we investigated the overlap between the two data sets. Using model 2, we identified 521 genes (P < .05) that were statistically significantly associated with plasma fetuin B, of which 106 were classified as secreted proteins. Of these, 29 were detected in plasma and 8 were found to be significantly associated with plasma fetuin B (Table 5).

Table 5.

Overlap between genes in white adipose tissue that are significantly associated with plasma fetuin B and that encode secreted proteins, and plasma proteins that are significantly associated with plasma fetuin B

Overlap between genes in white adipose tissue and plasma proteins
Carboxypeptidase B2
Carboxypeptidase E
Coiled-coil domain-containing protein 80
Dermatopontin
Low-affinity immunoglobulin γ Fc region receptor III-B
N-acetylglucosamine-6-sulfatase
Lipopolysaccharide-binding protein
Endoplasmic reticulum aminopeptidase 1

Discussion

Fetuin B is known to induce glucose intolerance in rodents but the underlying mechanism remained unclear. Recently, we found that specifically the amount of fetuin B in adipose tissue was strongly associated with peripheral insulin resistance in mice and in a small group of individuals, which may involve a fetuin B–induced inflammatory response in white adipocytes (7). In the present study, we aimed to investigate the link between plasma fetuin B concentration and alterations in the adipose tissue transcriptome and plasma proteome in a large cohort of individuals. We found that plasma fetuin B was related to more than 100 genes (P < .01) in white adipose tissue, with functions related to cytokine/chemokine and insulin signaling. With respect to the plasma proteome, we found a statistically significant association between plasma fetuin B and more than 146 proteins (P < .01) with functions related to the complement system and coagulation cascades, PI3-Akt signaling, cytokine signaling, and a number of inflammatory pathways.

To investigate the link between plasma fetuin B concentration and measures of glucose homeostasis, we first performed linear regression analyses in a cohort of 558 individuals. The data showed that plasma fetuin B was not associated with measures taken in the fasted state, including fasting glucose, fasting insulin, HOMA-IR, or fasting FFAs, but there was a strong association between plasma fetuin B with measures related to the OGTT, including 2-hour glucose levels, AUC glucose, AUC insulin, and Matsuda index. This is in line with our previous findings in a smaller study population, showing a strong association between plasma fetuin B and 2-hour glucose levels but not fasting glucose levels or HOMA-IR (7). Other studies, however, did report a link with fasting indices, including fasting insulin, fasting glucose, and HOMA-IR (6, 21, 22). The reason for this discrepancy is not entirely clear, but may depend on the characteristics of the study population. The median and average 2-hour glucose levels in the present study were 6.3 and 6.7 mmol/L, respectively, indicating that, despite the presence of overweight and obesity, this was a relatively healthy population. Similarly, our previous study also was performed in individuals that were either healthy or impaired glucose tolerant (7). Notably, a study that investigated the relation between serum fetuin-B concentration and fasting glucose parameters in healthy women and women with PCOS reported positive associations between plasma fetuin B and insulin concentration, HOMA-IR, and HOMA-β in women with PCOS, but not in healthy volunteers (6). Another study in individuals with type 2 diabetes reported positive correlations between plasma fetuin B and triglycerides, fasting plasma glucose, 2-hour postprandial plasma glucose, HOMA-IR, fasting insulin, glycated hemoglobin A_1c, and high-sensitivity CRP (22). Thus, although there is no doubt that plasma fetuin B is closely linked to parameters of postprandial glucose homeostasis in a wide range of individuals, its link with fasting parameters might be obvious only in individuals with more pronounced insulin resistance or type 2 diabetes.

Previously, we reported that specifically the amount of fetuin B in adipose tissue was strongly associated with the degree of peripheral insulin resistance both in mice and humans, and we found that fetuin B affected inflammatory gene expression in 3T3-L1 adipocytes (7). This prompted us to investigate the link between adipose tissue fetuin B levels and alterations in the adipose tissue transcriptome in individuals of the DiOGenes study. Since adipose tissue fetuin B levels were not available for these individuals, we first investigated the correlation between plasma fetuin B and abdominal subcutaneous adipose tissue fetuin B concentration in samples from two previous studies performed in our laboratory. We observed a strong and highly significant correlation between plasma and adipose tissue fetuin B levels, on which we decided to use plasma fetuin B levels as a marker for fetuin B in adipose tissue.

We found that, depending on the model used, between 101 and 109 genes in adipose tissue were significantly associated with plasma fetuin B concentration. In models 1 and 2, plasma fetuin B was associated with genes related to cytokine signaling and chemokine signaling. This includes a negative association with CSF3R, PPBP, CVCR1, and PF4V1. Interestingly, cytokine and chemokine signaling pathways were also found to be enriched in our previous study, in which 3T3-L1 adipocytes were treated with fetuin B (7). In our present study, however, different genes were found to be associated with these pathways, and, due to the limited number of genes identified and the limited research available, it is difficult to assign a clear direction to these pathways.

We also found that fetuin B was associated with genes related to PI-3K-Akt signaling, and this pathway was consistently enriched across all models. Specifically we found a positive association between fetuin B and PDGFD, FGF9, and PPP2R2C, genes that are linked with obesity or decreased insulin signaling (23-26). Notably, it has previously been reported that fetuin B overexpression suppresses proliferation, migration, and invasion in prostate cancer by inhibiting the PI3K/AKT signaling pathway (27). Also in cardiomyocytes, fetuin B incubation has been shown to decrease insulin-induced glucose uptake via reduced tyrosine kinase activity of insulin resistance (28). Altogether, these data support our hypothesis that fetuin B is associated with genes in the adipose tissue related to cytokine/chemokine signaling and ineffective insulin signaling.

In a subsequent part of our study, we performed proteomics analysis in plasma to investigate which plasma proteins were associated with plasma fetuin B. Depending on the model used, we found between 146 and 162 proteins to be associated with plasma fetuin B. These include a number of proteins that are related to decreased insulin sensitivity or obesity-associated diseases such as RET (29), vascular endothelial growth factor (VEGF) (30), and JAK2 (31). Pathway analysis indeed revealed that the significant proteins were, among others, involved in pathways related to the complement system and coagulation cascades, PI3-Akt signaling, focal adhesion, cytokine signaling, and a number of inflammatory pathways.

The origin of the proteins in the circulation is not known. It is well recognized that the adipose tissue is a major metabolic endocrine organ, and similar to other metabolic organs, including skeletal muscle and the liver, the adipose tissue secretes proteins into the plasma to influence whole-body metabolism. Next, we set out to explore the overlap between the genes in abdominal subcutaneous adipose tissue that were significantly associated with plasma fetuin B and the plasma proteins that were significantly associated with plasma fetuin B. We specifically focused on the genes within the adipose tissue that are known to encode secreted proteins, and matched the genes and proteins based on their Ensembl gene IDs. In total, the Ensemble gene IDs of 8 genes in abdominal subcutaneous adipose tissue that were significantly associated with plasma fetuin B levels and known to encode secreted proteins were also detected in plasma and found to be associated with plasma fetuin B levels. This includes an upregulation of several genes/proteins that have previously been linked to inflammation or insulin resistance, including dermapontin (32), lipopolysaccharide-binding protein, N-acetylglucosamine-6-sulfatase (33), and endoplasmic reticulum aminopeptidase 1 (34). Although the overlap is relatively low, this is not unexpected. Of the 106 genes that were significantly associated with fetuin B and that are suggested to translate into secreted proteins, only 29 were detected in plasma. This issue arises partly because the annotation process used to identify which genes are secreted is not tailored to specific tissues or conditions. As a result, some genes may have been incorrectly marked as “secreted” when in reality they are not. Also, to compare our transcriptome and proteome data, we had to convert them to a common identifier system (Ensembl Gene IDs), which may have introduced some conversion errors. From a physiological point of view, it is well known that posttranscriptional modifications can significantly alter the secretion profile of proteins in adipose tissue and furthermore, other tissues in the body also contribute to the plasma protein pool. If these tissues secrete proteins independently of fetuin B, it introduces a variability that can obscure possible associations. In line with the present findings, our previous research also found that the protein secretome of the liver could not be accurately predicted by assessing changes in the transcriptome (2).

One of the major strengths of this study is its use of data from the extensive DiOGenes cohort, which is based on a large, multicenter design and enhances the generalizability and robustness of our findings. The study provides valuable insights into the associations between plasma fetuin B, adipose tissue gene expression, plasma proteome profiles, and glucose homeostasis in humans. A notable limitation, however, is that our study is unable to establish causality. Fetuin B correlates with other secreted factors, as shown by the plasma proteome analysis, and these correlations likely reflect shared pathways or systemic processes, complicating the determination of direct cause-and-effect relationships. Nevertheless, while causality cannot be confirmed, our previous study (7) demonstrated that incubating adipocytes with fetuin B significantly increased genes involved in inflammatory pathways, providing mechanistic evidence linking fetuin B to adipose tissue biology. This supports the hypothesis that fetuin B may directly influence transcriptomic changes, reinforcing its role as an independent factor in this context.

In summary, the present study demonstrates that plasma fetuin B is related to adipose tissue genes and plasma proteins involved in metabolic processes and insulin signaling in participants of the large, European, multicenter DiOGenes study. While this study cannot establish causality, these findings strengthen our previous work, showing a strong association between fetuin B levels in adipose tissue and peripheral insulin resistance both in mice and a small-scale human study, as well as a causal link between fetuin B and inflammation in cultured adipocytes (7). Thus, the present translational study, which bridges the gap between findings in cellular and animal models and humans, provides more robust evidence for the role of fetuin B in metabolic regulation. Future studies are warranted to examine whether fetuin B may act as a biomarker for inflammation and insulin resistance and could be a potential therapeutic target to improve insulin sensitivity in humans.

Abbreviations

AU

arbitrary units

BMI

body mass index

CRP

C-reactive protein

FFA

free fatty acid

HOMA-IR

homeostatic model assessment of insulin resistance

MCP1

monocyte chemoattractant protein 1

OGTT

oral glucose tolerance test

PAI1

plasminogen activator inhibitor-1

PCOS

polycystic ovary syndrome

TNF-α

tumor necrosis factor α

Funding

R.C.R.M. was supported by an Aspasia Grant of The Dutch Research Council (NWO). The Diogenes project was funded by EU funding under the FoodQuality and Safety Priority of the Sixth Framework Programme for Research and Technological Development of the European Union (2005-2009) (FP6-2005-513946), through a grant from the Maastricht University Medical Center and Nestlé Institute of Health Sciences, Lausanne, Switzerland.

Author Contributions

E.J.K., G.H.G., E.E.B., M.A., and R.C.R.M. performed research and analyzed and interpreted data. E.J.K. and R.C.R.M. wrote the paper. G.H.G., E.E.B., and M.A. revised the manuscript. R.C.R.M. had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. All authors read and approved the final version of the manuscript.

Disclosures

The authors declare no conflicts of interest in relation to the work described.

Data Availability

The data presented in this manuscript are available from the corresponding author on reasonable request.

Clinical Trial Information

ClinicalTrials.gov identifier numbers NCT00390637, NCT02241421, and NCT02381145.