Self-reported dietary omega-3 polyunsaturated fatty acids are associated with adipose tissue markers and glucose metabolism in apparently healthy subjects

Abstract Background Plasminogen activator inhibitor 1 (PAI-1) and resistin are associated with dysfunctional adipose tissue (AT)-related metabolic complications. The role of dietary eicosapentaenoic (EPA) and docosahexaenoic (DHA) fatty acids in this relationship is unknown. Aim To investigate the association of EPA and DHA with PAI-1 and resistin, as well as the role of this association on the glucose metabolism of apparently healthy subjects. Subjects and methods Thirty-six healthy individuals were included. Validated food frequency questionnaires were used to analyse dietary habits. Inflammatory and glucose metabolism markers were quantified. Subcutaneous AT samples were obtained, and adipocyte number, area, and macrophage content were assessed. Results In 36 subjects aged 56 ± 8 years and with a body mass index of 26 ± 4 kg/m2, logEPA, and logDHA showed significant association with logresistin and a marginal association with PAI-1. Adipocyte number, area, and lognumber of macrophages per adipocyte significantly correlated with PAI-1 but not with logresistin. Although logEPA and logDHA were independently associated with loginsulin, loginsulin resistance, and C-Peptide, the addition of logresistin, but not of PAI-1, into the multivariable model, abolished the associations. Conclusions EPA and DHA could modulate glucose metabolism across AT functional states. Our data indicate that this association is independent of other metabolic risk factors.


Introduction
In the last three decades, the transition towards urbanisation and rapid technological advances has led to a global substitution of work-related physical activity and diets with energy-rich products that, in turn, lead to increased excess body weight rates (Thota et al. 2018). Unhealthy dietary patterns, characterised by a high intake of ultra-processed foods and saturated fats, are directly associated with excess body weight and the development of cardio-metabolic complications such as fatty liver, hypertension, diabetes mellitus, and cardiovascular diseases (CVD) (Park et al. 2017). Conversely, traditional diets based on fruits, vegetables, whole grains, and oily fish, such as the Mediterranean diet, are associated with health benefits and disease prevention (Mirabelli et al. 2020).
The beneficial effects of the Mediterranean diet are attributed to its high content of plant polyphenols, fermentable fibre, and especially long-chain omega-3 polyunsaturated fatty acids (n-3 PUFAs). Several studies have shown that the consumption of n-3 PUFAs is associated with a lower incidence of cardio-metabolic diseases (Mirabelli et al. 2020). Eicosapentaenoic (EPA) and docosahexaenoic (DHA) acids are n-3 PUFAs mainly obtained from oily fish with anti-inflammatory, anti-atherogenic, anti-diabetic, and cardioprotective properties (Rizos et al. 2012). These two fatty acids exert their beneficial metabolic activities. They have been described to activate G protein-coupled receptor 120 (GPR120) and peroxisome proliferator-activated receptorgamma (PPAR c) (Chistyakov et al. 2020).
White adipocytes and adipose tissue-resident macrophages express GPR120 and PPARc, making them possible targets for the beneficial metabolic effects of EPA and DHA (Chistyakov et al. 2020). Determining the effect of dietary EPA and DHA on subcutaneous white adipose tissue (SWAT) function is relevant since the dysfunction of this tissue is directly associated with insulin resistance and inflammation, even in apparently healthy normal-weight individuals (Reyes- Barrera et al. 2021). Dysfunctional white adipose tissue (WAT) is characterised by a reduced preadipocyte proliferative capacity (hypoplasia), altered adipokines secretion, and impaired extracellular matrix remodelling (Choe et al. 2016). These alterations derive from excessive lipid accumulation in existing adipocytes (hypertrophy) and inflammatory M1 macrophage activation (Goossens 2017). Recent studies have suggested that measuring morphological features and adipokines secretion could help evaluate WAT functionality and the risk of developing cardio-metabolic disorders (Choe et al. 2016;Åkra et al. 2020). In line with this, experimental and epidemiological studies have shown that hypertrophic and hypoplasic adipose tissue expansion are clear indicators of dysfunctional SWAT and metabolic derangements (Su et al. 2020).
Two adipokines overexpressed by dysfunctional adipocytes are resistin (Steppan et al. 2001) and PAI-1 (Åkra et al. 2020), which have also been associated with low-grade chronic inflammation and insulin resistance (Jonas et al. 2017). Although these observations suggest a direct role of both adipokines on the progression of dysfunctional adipose tissue-related metabolic complications, it is unclear whether dietary intake of EPA and DHA influences SWAT morphofunctional characteristics and the presence of cardio-metabolic abnormalities. Therefore, the objective of the present study was to analyse the relationship between self-report dietary intake of EPA and DHA with morpho-functional markers of SWAT, as well the possible impact of this association on the glucose metabolism of apparently healthy subjects.

Study population
To the present analysis, 350 healthy individuals were invited (through phone calls and personal interviews) to participate in the study to investigate the morpho-functionality of their adipose tissue. Fifty-two subjects without cardiovascular disease, dyslipidaemia (triglycerides >280mg/mL), infectious disease, diabetes, cancer, or any autoimmune disease, and with ages from 35 to 70 years, voluntarily agreed to participate and were enrolled from February 2018 to May 2019. Nevertheless, 16 subjects were excluded from the analyses due to inadequate subcutaneous adipose tissue biopsy or incomplete clinical information. The clinical information and the different biological samples of the participants were collected in a single visit. The study was approved by the Ethics and Research Committee of the Instituto Nacional de Cardiolog ıa Ignacio Ch avez (protocol number: 17-1040) based on the guidelines from the Declaration of Helsinki. Written informed consent was obtained from all studied subjects.

Inclusion criteria
Subjects of both sexes, aged !18 years, with BMI from 19 to 35 kg/m 2 , and those participants who were willing and able to comply with the scheduled visit that included laboratory, clinical, and biopsy procedures. Additionally, subjects had to accept signed informed consent.

Exclusion criteria
Medical conditions including diabetes, hypertension, cancer, autoimmune diseases, active infections, any pharmacological treatment, history of cardiovascular disease, bleeding, or refusal of the subject to any procedure during a scheduled visit.

Clinical and anthropometrical evaluation
The trained research staff interviewed participants and fulfilled a clinical assessment questionnaire detailing medical history, demographic characteristics, cardiovascular history, medications, and tobacco use. Height, weight, and waist circumference (WC) were measured. With the patient's standing position, WC was measured at the midpoint between the top of the iliac crest and the lower margin of the last palpable rib in the midaxillary line. Body mass index (BMI) was calculated based on weight (kg) divided by height (m 2 ).

Dietary assessment
The dietary habits assessment of each participant was performed using a 116-item semiquantitative food frequency questionnaire (FFQ) previously validated for the Mexican population by the Instituto Nacional de Salud P ublica (Jacobo-Albavera et al. 2015). The FFQs were designed to estimate the usual dietary intake over the previous 12-month period and collect information regarding carbohydrates, proteins, and fats consumption, including detailed EPA and DHA intake. Daily energy intake and proportion of macronutrients and micronutrients were estimated using the system evaluation of nutritional habits and food intake (SNUT) (Medina-Urrutia et al. 2015). The validity and reproducibility of this questionnaire to assess dietary intake of nutrients have been previously reported for this population (Jacobo-Albavera et al. 2015).

Biochemical measurements
Blood samples were obtained from an antecubital vein of each patient after a 12-h overnight fast and 20 min in a sitting position. Plasma glucose, total cholesterol, triglycerides, and high-density lipoprotein cholesterol were measured using standardised enzymatic procedures in a COBAS c311 autoanalyzer (Roche Diagnostics, Mannheim, Germany). Lowdensity lipoprotein cholesterol was estimated using the De Long et al. formula (DeLong et al. 1986). The accuracy and precision of lipid measurements in our laboratory are under periodic surveillance by the Centre for Disease Control and Prevention service (Atlanta, GA, USA). Inter-assay coefficients of variation were <6% for all these assays. Resistin, PAI-1, Cpeptide, glucose-dependent insulinotropic polypeptide (GIP), glucagon, and insulin were quantified using a Bio-Plex 200 system and analysed in Bio-Plex Manager software 6.1 (Bio-Rad Inc, USA) with intra-and inter-assay variation coefficients below 4% and 5% (respectively). The homeostatic model assessment of insulin resistance (HOMA-IR) index was calculated using the formula:

Adipose tissue biopsies
In the 36 apparently healthy volunteers enrolled in the present study, a subcutaneous white adipose tissue sample was obtained from periumbilical fat, with surgical technique under local anaesthesia (2% lidocaine) and after an overnight fast (Chachopoulos et al. 2017). Biopsies were immediately rinsed with sterile saline, and visible blood vessels were removed. Adipose tissue biopsies were immediately fixed in PBS-buffered 4% paraformaldehyde for histological analyses.

Adipose tissue histological analysis
After 24 h, fixed tissues were dehydrated in ethanol, cleared in xylene, embedded in paraffin, sectioned at 4 mm, and stained with haematoxylin and eosin. Digital images were obtained using a digital camera (Leica ICC50 HD) coupled to a Leica DM750 microscope using a 20X lens at a resolution of 2048 Â 1536 pixels using LAS EX V 3.0 software (Leica Microsystems, Switzerland). Mounted sections were divided into 20 fields of 1mm 2 . Five random fields from each sample were analysed. Adipocyte number and area were measured using the automated analysis software Adiposoft (ImageJ), with the following parameters: minimum diameter 10 mm, maximum diameter 1000 mm, and microns per pixel 0.439 as previously described (Reyes-Barrera et al. 2021). Each value was verified manually after completing the measurements to prevent errors that may have occurred during automated analyses (Reyes- Barrera et al. 2021). The macrophages (M1, proinflammatory type) were identified by staining sections with an antibody against inducible nitric oxide synthase (iNOS) (Santa Cruz-7271). Macrophage content per slide was normalised by dividing the number of positively stained cells per 100 adipocytes (Jia et al. 2020).

Statistics
Clinical, biochemical, and dietary parameters are presented as means and standard deviation if normally distributed, or median (interquartile range) if presented with the asymmetric distribution. Qualitative variables are presented as the number of subjects (percentages). The normality of the distribution of the analysed variables was assessed with the skewness and kurtosis tests. Due to their non-parametric distribution, EPA, DHA, GIP, HOMA-IR, insulin, resistin, and the number of macrophages per adipocytes were log-transformed. All correlations between log EPA and log DHA, PAI-1, and log resistin with morpho-functional adipose tissue characteristics and glucose metabolism markers were analysed, and Pearson's correlation coefficient values (r s ) were calculated. The independence of these associations was performed through a multiple stepwise linear regression analysis, adjusted by potential confounders (sex, age, body mass index, and total kcal intake). All p values <0.05 were considered statistically significant. The statistical analyses were performed using SPSS 15.0 software (Chicago, IL, USA).

Association of circulating levels of PAI-1 and resistin with SWAT morphology and glucose metabolism
Dysfunctional SWAT is characterised by an altered adipokine secretion, such as resistin and PAI-1. It can be evaluated morphologically by increased adipocyte size, reduced adipocyte number, and macrophage infiltrates (Choe et al. 2016;Jonas et al. 2017;Åkra et al. 2020). Figure 1 shows that adipocyte number, mean area, and log number of macrophages per adipocyte significantly correlate with PAI-1 but not with log resitin. Furthermore, PAI-1 also significantly correlates with log insulin levels and log HOMA-IR values. In contrast, log resistin correlates with C-peptide and glucagon serum levels ( Figure 2). To investigate the independence of these correlations, a multiple linear regression analysis adjusted by sex, age, BMI, and total energy intake was carried out. Table 3 shows that PAI-1 was marginally associated with log insulin and log HOMA-IR. In contrast, log resistin was significantly associated with log insulin, log HOMA-IR, C-Peptide, and glucagon. These results indicate that even in apparently healthy individuals, inflammatory adipokines increase as a function of adipose tissue hypertrophy and are associated with insulin resistance.
Association of self-report dietary EPA and DHA intake with circulating levels of PAI-1, resistin, and glucose metabolism Diet is a major independent variable that regulates an organism's metabolic and inflammatory status (Mirabelli et al. 2020;Chistyakov et al. 2020). The results of the univariate correlation in the present study indicates that log EPA correlates inversely with log resistin (r ¼ À0.581; p < 0.001), PAI-1 (r¼ À0.331; p ¼ 0.04), and mean area of adipocytes (r¼ À0.304; p ¼ 0.07), and directly with number of adipocytes per field (r ¼ 0.389; p ¼ 0.01). Similar trend correlations were found for log DHA (r¼ À0.518; p ¼ 0.001, r¼ À0.308; p ¼ 0.06, r¼ À0.344; p ¼ 0.01, and r ¼ 0.390; p ¼ 0.01; respectively). In addition, the multivariable linear regression analysis indicates that independent of sex, age, BMI, and total energy intake, log EPA, and log DHA have a significant association with log resistin and a marginal association with PAI-1 (Table 4). A similar analysis was performed to know the role of self-reported dietary EPA and DHA intake on glucose metabolism. Table 5 shows that log EPA and log DHA were independently associated with log insulin, log HOMA-IR, and C-Peptide. Interestingly, adding log resistin, but not PAI-1, into the model abolished the associations. Altogether, these results could indicate that frequent intake of foods rich in EPA and DHA is related to adipose tissue markers and metabolic alterations in apparently healthy subjects.

Discussion
Subcutaneous white adipose tissue serves as a sink for surplus energy intake during positive energy balance, preventing lipotoxic damage to non-adipose tissues such as the liver, skeletal muscle, kidneys, and pancreas (Chait and den Hartigh 2020). However, hypertrophic SWAT growth results in adipocyte dysfunction and inflammatory M1 macrophage activation (Choe et al. 2016;Jonas et al. 2017;Åkra et al. 2020). Dysfunctional WAT is closely related to cardiometabolic diseases, even in normal-weight individuals (Fan et al. 2021). During the last decades, several studies have shed light on the detrimental effect of unhealthy dietary patterns on WAT functionality and its association with cardio-metabolic complications (Park et al. 2017). Conversely, other studies have shown the beneficial effects of dietary bioactive compounds, such as n-3 PUFAs, on WAT function and their cardio-protective effects (Mirabelli et al. 2020). Although n-3 PUFAs play a pivotal role in metabolic processes, there are controversial findings on the association between dietary EPA and DHA intake, WAT function, and glucose metabolism (Zhuang et al. 2019). Moreover, the analysis of the participation of n-3 PUFAs on WAT and metabolism has been addressed separately (Thota et al. 2018). Additionally, it is unknown if dietary EPA and DHA prevent early SWAT morpho-functional alterations and, thus, the risk of developing metabolic abnormalities. The results of the present study indicate that self-reported dietary EPA and DHA are associated with morphological (size and number of adipocytes) and functional (circulating levels of PAI-1 and resistin) features of SWAT from apparently healthy subjects. Furthermore, our data highlight that dietary intake of EPA and DHA is independently associated with glucose metabolism markers and directly associated with circulating resistin levels. These findings suggest that EPA and DHA could regulate whole-body glucose metabolism directly affecting SWAT morpho-functional characteristics. White adipose tissue fatty acid composition has been considered a direct reflection of dietary fatty acids. Although consuming foods rich in EPA and DHA has been associated with a lower incidence of overweight and obesity, the debate continues on whether these fatty acids promote fat loss (Rosenwald et al. 2013). Nevertheless, some investigations have shown that n-3 PUFA is inversely related to white adipocyte size and inflammation markers (Itariu et al. 2012). Consistently, our data confirm that dietary intake of EPA and DHA is inversely correlated with hypertrophic subcutaneous white adipocytes (SWAd) and adds to the body of knowledge that the n-3 PUFA intake is directly related to the number of SWAd. It is essential to underline that a high number of SWAd is an indicator of healthy expansion of SWAT, whereas hypertrophic SWAd is an indicator of the unhealthy process of expansion of SWAT associated with cardio-metabolic diseases (Choe et al. 2016;Jonas et al. 2017;Åkra et al. 2020). In line with this, an additional analysis in the present work indicates that self-reported dietary EPA and DHA intake was inversely and independently associated with the SWAT inflammatory markers PAI-1 and resistin.
PAI-1 is a serine protease inhibitor that hinders the activity of tissue plasminogen activator and urokinase, the activators of plasminogen and hence fibrinolysis. This protein favours recruitment and polarisation to proinflammatory M1 macrophages. It is secreted by white adipocytes in response to proinflammatory cytokine signalling, such as tumour necrosis factor-a and transforming growth factor-b, but also in response to high glucose and insulin levels connecting inflammation with metabolic derangements (Åkra et al. 2020). PAI-1 also inhibits PPAR-c activity, impairing adipocyte differentiation, glucose uptake, and insulin sensitivity in adipocytes (Zhang et al. 1996). Accordingly, it has been shown that PAI-1 correlates directly with adiposity measurements and white adipocyte hypertrophy (Cesari et al. 2010;Åkra et al. 2020). Results of the present study confirm that circulating levels of PAI-1 are higher in subjects with enlarged SWAd and inversely correlated with the number of SWAd ( Figure 1). Furthermore, our data show a direct correlation of PAI-1 with the number of macrophages in SWAT and are supported by Wang L. et al. who demonstrated that in an animal model, PAI-1 deficiency or inhibition reduces SWAT macrophage infiltration (Wang et al. 2018). In addition, the authors of this study suggested that abnormal increases in the number of macrophages within SWAT could result in local inflammation. In the present study, those individuals consuming a higher amount of EPA and DHA presented lower circulating PAI-1 along with smaller SWAd and reduced macrophage content in SWAT. On the contrary, in a study in which insulin-resistant subjects were intervened with n-3 PUFAs, no reduction in PAI-1 concentrations was observed (Spencer et al. 2013). On the other hand, it has been demonstrated that EPA and DHA inhibit PAI-1 synthesis and secretion through a GPR120-dependent mechanism and increase white adipocyte differentiation and functionality through a ligand-induced transcriptional activation of PPAR c (Kalupahana et al. 2011). WAT growth requires massive extracellular matrix remodelling in order to expand efficiently. However, inadequate fibrinolytic activity induces WAT fibrosis and dysfunction (Catal an et al. 2012). Accordingly, Akra S. et al. designed a study similar to the present study, which observed in healthy subjects that PAI-1 impairs the remodelling process of the extracellular matrix of WAT associated with the development of glucose intolerance (Åkra et al. 2020). Conversely, the correlation analysis in the present study shows a positive relationship between circulating PAI-1 and HOMA-IR or insulin serum levels (Figure 2), suggesting that PAI-1 could be directly associated with the metabolic profile.
Another adipokine that is secreted in dysfunctional white adipocytes is resistin. Although resistin was first described as expressed and secreted by white adipocytes (Steppan et al. 2001;Steppan and Lazar 2002), other studies have shown that immune cells are the primary source of this adipokine (Yang et al. 2003). The content of this peptide is proportional to the intensity of WAT macrophage infiltration (Bokarewa    et al. 2005;Jonas et al. 2017). Controversially, the present study's findings did not show a relationship between resistin and SWAT macrophage content ( Figure 1). Notwithstanding, it is essential to note that compared with data in the literature that included patients with a wide variety of cardiometabolic disorders, subjects in the present study were apparently healthy and could present an earlier stage of metabolic derangement. Despite this, and consistent with the literature, we found that resistin is strongly associated with impaired glucose metabolism (Steppan et al. 2001;Steppan and Lazar 2002;Jonas et al. 2017). Although the link between resistin and insulin resistance has been reported, this is debateable since other groups have failed to confirm this relationship (Mazaherioun et al. 2017). The mechanisms by which resistin exerts its biological effects are not yet fully understood (Park et al. 2017). Interestingly, we also found a direct association of resistin with C-peptide and glucagon. Although some studies have shown similar findings, these associations, mechanisms, and interactions have been poorly investigated (Su et al. 2019). C-peptide is widely used to measure pancreatic beta-cell function due to its lower degradation rate than insulin. Moreover, it has been suggested that C-peptide could be useful in assessing beta-cell function, glycemic control, response to hypoglycaemic agents, and risk of diabetes complications (Leighton et al. 2017). On the other hand, increased plasma concentrations of glucagon have been proposed as an important marker of hepatic glucagon resistance, leading to abnormal liver metabolism (Galsgaard 2020). Together, these results could partially explain the independent association of resistin with markers for glucose metabolism abnormalities in our group of apparently healthy subjects, in whom early-stage metabolic derangement could be present. Epidemiological studies examining the n-3 PUFA intake have supported the role of these fatty acids in promoting health benefits (Rizos et al. 2012). Although it has been postulated that EPA and DHA may have several beneficial effects on glucose metabolism and WAT morphology and function (Itariu et al. 2012;Zhuang et al. 2019), their possible effects have been approached separately. So, as far as we know, this is the first study suggesting the role of self-reported dietary EPA and DHA intake on glucose homeostasis through modulation of PAI-1 and resistin as SWAT functionality markers.
The present study has several strengths. First, subjects were thoroughly characterised, allowing us to evaluate selfreported dietary intake of n-3 PUFAs and their relationship with SWAT morpho-functional markers and glucose metabolism. Second, unlike other studies that include subjects with morbid obesity and other comorbidities, participants of the present study had a BMI more similar to the general population ($26 kg/m 2 ) (Barquera et al. 2020) and were apparently healthy, which could reflect the early stages of SWAT dysfunction that could occur in the general population. Third, unlike others, adipose tissue in the present study was obtained from SWAT biopsies, which act as an energy buffer to avoid lipotoxic tissue damage (Reyes- Barrera et al. 2021). Fourth, unlike other studies, the biopsies were not obtained during an emergency or scheduled surgery, a situation that alters SWAT function at the time of sampling. A potential limitation of the study is the small number of subjects included. Another weakness is that n-3 PUFA was not measured in the biopsy or plasma samples. In addition, the questionnaire used to estimate n-3 PUFA intake depends on the memory of the participant, which could be a possible limitation. However, it is essential to note that the relationship between n-3 PUFA biomarkers and self-reported dietary n-3 PUFA intake through FFQs, has been previously validated (Jacobo-Albavera et al. 2015). Regarding limitations, the cross-sectional design is one of the main disadvantages of this study. However, the development of this type of analysis allows for generating hypotheses that lead to longitudinal studies to know the causality of the findings.
The present study shows that, in apparently healthy subjects, the self-reported dietary intake of EPA and DHA is associated with glucose metabolism and SWAT functional state. Data also indicate that this association is independent of other metabolic risk factors. This evidence indicates that dietary EPA and DHA intake can represent a non-pharmacological approach to maintaining adequate SWAT functions to prevent the development of cardiovascular disease at a later age and stresses the importance of dietary intake assessment as a valuable diagnostic tool in the clinical setting to identify early abnormalities in glucose metabolism even in apparently healthy individuals. Further investigations will be required to clarify the present findings, including interventional and prospective studies.