Since 1975 the prevalence of obesity has nearly tripled worldwide (1), and it is estimated that up to 2030, 38% of the world adult population will be overweight, while 20% will be obese (2). The main culprit for the present obesity pandemic is the modern lifestyle, which includes the lack of physical activity and increased availability and consumption of energy-dense (high-fat/sugar), processed, “western-style” food (2). The increasing prevalence of overweight and obesity is an emerging public health concern not only in the western, industrialized, high-income countries, but also (and even more) in the low-income and middle-income countries (3). Particularly the countries in economic transition are on the increased burden, since their economic transition is also followed by corresponding lifestyle and dietary changes (so called “nutritional transition”) (4). Serbia is one of the European countries in transition, and during the last 50 years an increased trend in prevalence of overweight and obesity among adults, adolescents and children has been observed (5). According to the latest national survey in 2019, even 57.1% of citizens 15 years and more old in Serbia were with overweight or obesity (6), which is much higher than the average prevalence of overweight and obesity worldwide (1).

Even though the modern lifestyle is the main culprit for development of obesity and related co-morbidities, there are additional factors which can contribute to their development, including hereditary factors (7).

Overweight and obesity are among the main factors responsible for the development of the metabolic syndrome (MetS), which is the key element in development of their metabolic and cardiovascular co-morbidities. Metabolic syndrome is characterized by abdominal and visceral adiposity, insulin resistance, hyperinsulinemia, hyperglycemia, dyslipidemia [increased triacylglycerols (TAG) and decreased high-density lipoproteins - cholesterol (HDL-C) levels], hypertension, non-alcoholic fatty liver disease (NAFLD) and proinflammatory and prothrombotic state (8, 9). However, not all obesity phenotypes are connected with development of MetS, and there is also a metabolically healthy phenotype of obesity (MHO), which does not have the adverse cardiometabolic components of MetS, despite increased adiposity (10). Vice versa, MetS, associated with substantial insulin resistance and increased cardiometabolic risk, is also often observed in normal-weight or underweight subjects, despite normal (11) or even decreased adiposity (e.g., in lipodystrophic syndromes, where in particular peripheral adipose tissue is absent, while central and visceral adiposity can exist) (12, 13). Therefore, increased adiposity and components of metabolic syndrome do not have to be necessary connected (14, 15). Nevertheless, increased adiposity itself certainly worsens all the MetS components (16). In accordance, there is considerable evidence that even in the subjects with MHO, with time the MetS components and increased risk for cardiovascular disease and diabetes develop, and this is mostly determined by the increased visceral adiposity (17, 18). Vice versa, even moderate weight loss (5–10%) significantly reduces almost all of the MetS components in subjects with overweight or obesity, deepening on the initial body weight (BW), with the greater weight loss being associated with the greater MetS components reduction (19).

In the last several decades, there is substantial evidence that long chain fatty acids (FAs) compositions in various body compartments (e.g., the liver, muscle, adipose tissue, whole blood, plasma, erythrocytes, thrombocytes, cerebrospinal fluid, follicular fluid, seminal plasma, etc.) and activities of enzymes involved in their metabolism – desaturases and elongases, are connected with development of obesity (20–24), abdominal adiposity (25–27), insulin resistance and hyperglycemia (28–30), dyslipidemia (31), MetS (32–39), NAFLD (40), diabetes (41–46), and various cardiovascular diseases (hypertension, atherosclerosis, stroke, myocardial infarction, heart failure, etc.) (47–49).

The activities of enzymes involved in metabolism of FAs and profiles of FAs in erythrocytes, thrombocytes and various plasma lipids – phospholipids (PL), cholesterol-esters (CE), TAG and free fatty acids (FFA) are influenced by many factors, including genetics (50–52), gender (53–55), ethnicity (55–57), age (53, 58, 59), hormonal status (60–63), long-term and short-term dietary intakes of specific macronutrients (51, 64–67) and micronutrients (minerals: Fe, Zn; vitamins: D, B9, B12, B6, A, C; polyphenols, etc.) (64, 68–74), physical activity (53, 75), smoking (53, 74), but also (as mentioned above) by the presence of some metabolic conditions (including obesity, dyslipidemia, metabolic syndrome, NAFLD and diabetes), or by the use of certain drugs (e.g., statins) (76).

Delta-9 desaturase (D9D, also known as stearoyl-CoA desaturase 1, SCD-1), delta-6 desaturase (D6D) and delta-5 desaturase (D5D) are enzymes involved in desaturation process of, correspondingly, saturated (D9D) or polyunsaturated (D6D and D5D) long chain FAs. They are encoded by genes Scd1, Fads2 and Fads1, respectively. To directly measure the activities of desaturases in different body compartments (specific tissues, whole blood, plasma, erythrocytes, etc.) is difficult and demanding, requires biopsy of tissues (e.g., the liver, fat tissue, muscle) and complicated techniques for the microsomal activity and expression measurements, or includes very costly and laborious methods with stable isotope administration (D₂O, ¹³C-acetate or ¹³C-labeled FAs), which are impractical for large samples (77–79). For all of these reasons, desaturase activities are the most often estimated through their product/precursor FAs ratios in the circulation.

The activity of D9D can be estimated through the ratios 16:1n-7/16:0 and 18:1n-9/18:0 (for D9D-16 and D9D-18 activities, respectively). Similarly, the ratios 18:3n-6/18:2n-6 and 18:4n-3/18:3n-3 can both be used as the estimations of D6D activity, while the ratios of 20:4n-6/20:3 n-6 and 20:5n-3/20:4n-3 as the estimations of D5D activity, as n-3 and n-6 FAs compete for the same enzymes. However, the levels of 16:1n-7, 18:3n-6 and 20:3n-6 are more often used for the estimation of enzymatic activity, since they are less dependent on FAs dietary intakes, given that they are scarce in habitual diets, and therefore more reflect the endogenous synthesis and metabolism (20, 77, 80). Additionally, because of very rapid conversion of 18:3n-6 by elongase 5 (Elov5) to 20:3n-6 (80), the ratio of 20:3n-6 to 18:2n-6 is more often used to estimate D6D activity (49).

Erythrocyte membrane PL FAs composition is determined by a combination of diet and endogenous metabolism (61). Even though each lipid pool has a distinctive pattern of FAs profiles, there is a significant correlation between erythrocyte PL and TAG FAs composition, and hepatic, adipose tissue, muscle and plasma lipids FAs composition, because there is an exchange of FAs between different lipid pools (21, 77, 81–89). It is commonly thought that the profiles of FAs in erythrocyte membranes are more steady long-term and less influenced by the temporary dietary changes in intakes of FAs and carbohydrates, compared with the FAs profiles in plasma lipids (90, 91), even though there are some studies which disagree with this (85, 87, 89). Moreover, they are less influenced by the plasma TAG levels, compared with the whole plasma FAs profiles (92). Therefore, they can be viewed as indicators of average macronutrient intakes during several proceeding weeks or months, as well as indicators of endogenous metabolism in other tissues, particularly in the liver (90, 93, 94). Besides, there are also membrane-bond desaturases in erythrocytes, and their activity can be stimulated by oxidative stress and inflammation, caused by exercise, smoking, aging or other pro-oxidative and pro-inflammatory conditions (including obesity), in order to improve erythrocyte-membrane flexibility and protect erythrocytes from the free radical–induced hemolysis (95, 96).

Even though there is enough evidence that FAs composition (as well as the activities of desaturases and elongases) in different lipid compartments are associated with development of obesity, MetS, dyslipidemia, NAFLD, diabetes, atherosclerosis and cardiovascular diseases, the majority of the studies usually did not distinguish the associations with biochemical and clinical cardiometabolic risk markers from the associations with the increased adiposity (i.e., overweight and obesity). It is well known that the level of adiposity is in a direct association with biochemical and clinical markers of cardiometabolic risk, including levels of glucose (GLU), lipids, liver enzymes and blood pressure (BP), so it is unclear if there are some additional, adiposity-independent associations of FAs profiles and desaturase activities with biochemical and clinical indicators of cardiometabolic risk. Just a few studies controlled these associations, and only for body mass index (BMI) as a measure of adiposity (32, 97–99), while the other indicators of adiposity (100) were not considered. Some of these studies included only subjects with obesity (98), while the associations in normal-weight and underweight subjects were rarely examined. Additionally, in some studies the influence of other possible confounding factors, including sex, ethnicity, age, diet, smoking, physical activity, and alcohol consumption was not controlled. Many studies included either a mixed sample of men and women (without a clear separation of the results by sex) (39, 101) or only men (even elderly men) (32, 99), while it is known that FAs profiles and activities of desaturases can be influenced by sex (53, 54, 98), and the studies in women are particularly lacking. There are also huge racial and ethnic differences in FAs profiles and estimated activities of desaturases, which are not necessarily related to different cultural and dietary influences (55–57, 102–105). Moreover, many studies were performed in subjects with present cardiometabolic conditions (diabetes, cardiovascular diseases), while subtle changes in healthy subjects were not often examined.

This study showed that product/precursor estimated desaturase activities of D9D, D6D and D5D, and the FAs profiles of 18:0, 16:1n-7, 20:3n-6 and 20:5n-3 in erythrocyte membranes PL are significantly associated with the examined indicators of cardiometabolic risk among non-diabetic otherwise healthy Serbian women. The positive association of desaturase activities with adiposity was crucial and explained the association with other (biochemical and clinical) markers of cardiometabolic risk (together with age), since in our study we did not find age- and adiposity- independent associations of the mentioned FAs levels and desaturase activities with the levels of lipids, GLU, transaminases or BP (except for the associations of 20:5n-3 with ALT/AST ratio). Body mass index (together with age and smoking) was a significant predictor for almost all of the above-mentioned FAs (except 20:5n-3) and desaturases, and explained up to 25% of their variance. Vice versa, the activities of desaturases (particularly D9D and D6D) were significant predictors of anthropometric indices, together with age and physical activity, while (interestingly) dietary factors, postmenopausal status, smoking and educational level did not have significant influence in multiple regression models.

Our results on the associations of desaturase activities with the level of adiposity agree with almost all published data (20, 22, 27), even in children (25, 129, 130), but disagree with some studies concerning the adiposity-independent associations with clinical and biochemical biomarkers of cardiometabolic risk. For example, our results for adiposity-mediated D6D and D9D associations with metabolic profile are in agreement with a prospective study in elderly men (32), but disagree with D5D associations, since in that study the inverse association of D5D with adverse metabolic profile was independent from BMI. In a 10-year longitudinal study in Chinese women and men, higher serum D9D-18 activity, 16:1n-7, 18:1n-9, 20:3n-6 levels and lower D5D activity at baseline were predictors of development of metabolic abnormalities and transition of MHO to metabolically unhealthy obesity (MUO) phenotype, independently of sex, age, and BMI (97, 98). Nevertheless, in metabolically healthy normal-weight subjects who progressed to metabolically unhealthy normal-weight phenotype, these differences in desaturase activities were not observed (97). Similar results were obtained by the same authors in additional cross-sectional studies, where subjects with MUO had higher serum 16:1n-7, 18:1n-9, 20:3n-6 levels, higher D9D-16/D9D-18 activities and lower D5D activity, compared with subjects with MHO, despite similar BMI (97, 98). Similarly, serum ALT was independently positively associated with D9D activity in elderly men in one study (99). Risk for hypertension was associated with 16:0, 16:1n–7, 22:6n–3, 20:3n–6, D9D, D6D and D5D activity among Chinese cohort, independently of age, gender, BMI and other confounders (131). Possible reasons for discrepancy in these results with ours could be the differences in the studied populations and the analyzed FAs pools. In our study healthy women were studied, majority of them were of younger age, normal weight and only several of them had MetS. If the subjects with older age, higher level of adiposity and more adverse metabolic profile were included, or men (who are, e.g., more predisposed to NAFLD) (132), probably we would see also the adiposity-independent effects. Additionally, the ethnic differences may also contribute to the observed discrepancy in the results (57, 103). Apart from possible genetic differences between ethnic populations in Scd1, Fads2 and Fads1 polymorphisms (105, 133, 134), also dietary differences between ethnic populations may contribute (103). For example, in Caucasians on a high carbohydrate diet, 44% of variance in serum TAG was explained by D9D activity, while on high fat diet, only 11% of variance in serum TAG was explained by D9D activity (135). In our study, a high-fat/low-carbohydrate dietary pattern was identified, so the associations between desaturase activities and serum TAG levels may be less observable. Therefore, many differences between studied populations can contribute to contradictory results. Nevertheless, according to here presented data, in healthy, non-diabetic, predominantly younger and non-obese Serbian women, the connection of FAs profiles and desaturases activity with adverse metabolic profile was mostly conveyed by the effect of increased adiposity.

The association of adiposity and desaturase activities can be bidirectional: the increased adiposity produced by increased energy intake can change the expression and activities of desaturases in various tissues, and vice versa, the activities of desaturases may contribute to increased adiposity.

In support of the effect of adiposity on desaturases activities, weight loss in humans achieved by various dietary caloric reductions (with or without exercise) or metabolic surgery, significantly increased the D5D activity, while decreased the D6D and D9D activities and 16:1n-7, 20:3n-6, 18:3n-6 and 18:0 levels in erythrocyte membranes, whole plasma lipids, or separate plasma/serum lipid fractions (97, 98, 136–140). The reasons for the differences in the effect of weight loss on 18:0 in the cited studies compared with this study findings are unclear.

Likewise, animal studies with diet induced obesity (DIO) are in line with findings in humans, even though there can be some differences and discrepancies. For example, in the liver of mice on high-fat diets (HFD) or high sucrose diets (HSD), the increased, decreased or unchanged mRNA and protein expressions and estimated desaturases activities were shown (94, 141–146). The reasons for these discrepancies in animal results are not clear, and can be results of different experimental designs, animal strains, mode of expressing desaturase activity, type of diet or level of adiposity achieved.

The expression and activity of desaturases is controlled by many factors, including nuclear receptors: sterol regulatory element binding protein-1c (SREBP1c), liver X receptors (LXR), retionoid X receptor (rXR), carbohydrate regulatory element binding protein/ Max-like factor-X (ChREBP/MLX), PPAR-gamma peroxisome proliferator activated receptor (PPAR) alpha and gamma, and many others (64, 147, 148). In addition, epigenetic alterations (i.e., methylation of DNA), stability and degradation of the enzyme protein, as well as changes in the cytosolic NAD+/NADH ratio may contribute to changes in desaturases activities. In general, desaturases expressions and activities are stimulated by insulin, carbohydrates, and inhibited by PUFA, while the effects of leptin, SFA and MUFA can vary, depending on specific desaturase (64, 147–150).

Even though bio-mechanisms by which increased adiposity affects desaturase activities are still not well defined and can involve various pathways, it seems that the desaturase activity regulation by insulin and leptin can provide satisfying explanations.

Increased adiposity causes insulin resistance and leptin resistance. Insulin is one of the major stimulators of desaturase activities in many tissues, through induction of SREBP-1c, a nuclear factor which regulates lipogenesis and increases all three desaturases’ expressions and activities (151). Insulin increases the SREBP-1c gene transcription and mRNA stability, proteolytic maturation and nuclear transition of SREBP-1c protein, and inhibits its proteasomal degradation. Even though in obesity there is insulin resistance in many tissues for various actions of insulin, interestingly, there is no resistance of the effect of insulin on SREPB-1c induction in the liver (152, 153). Consequently, the compensatory increased insulin levels in insulin resistant states (including obesity) stimulate the SREPB-1c and increase expression and activity of all three desaturases, which was confirmed by studies in streptozocin induced diabetic animals (147, 154). However, it remains unclear which mechanism is responsible for different regulation of D5D by insulin in obesity, compared with D9D and D6D. Most studies show that D9D, D5D and D6D are regulated by the same factors (including insulin) in the same direction (64, 147). However, it seems that there is a differential effect of insulin on desaturases, depending on desaturase itself, the concentration of insulin and insulin sensitivity (155). Some studies suggested that regulation by insulin can be differential for D5D compared with D6D and D9D (156–158). For example, in streptozocin induced diabetes, the inhibition of D6D is much stronger than inhibition of D5D, indicating that D6D more strongly depends on insulin than D5D (156). Another study in animals showed no effect of insulin on the estimated D5D activity in hepatocytes or myocytes (159). Moreover, while in small, physiological doses insulin can have an increasing effect on the decreased D5D activity in insulin deficient diabetic rats or humans with type 1 diabetes (156, 160, 161), it was shown, in contrast, that in non-diabetic rats, insulin in higher doses inhibits the D5D activity in hepatocytes, measured by the labeled FAs incorporation (162). Unfortunately, we did not measure the insulin levels in our subjects, and there are no studies on the effect of insulin in non-diabetic subjects, to test the hypothesis of the biphasic effect of insulin on D5D activity (e.g., kinetic studies involving hyperinsulinemic-euglycemic-clamp and measuring desaturase activities in the blood by the use of stable isotopes).

With increased adiposity there is also leptin resistance. Leptin is known as potent inhibitor of D9D and D6D transcription and activation, mostly through its indirect central effects in the brain, but also some direct effects of leptin on the liver and adipose tissue can exist (163). In leptin deficient ob/ob animals, leptin administration centrally (i.e., intra-cerebroventricularly) decreased liver and adipose tissue Scd1 and Fads2 expression (unfortunately, Fads1 expression was not examined), through its effects on SREBP-1c, ChREBP, PPAR-gamma, PPAR-alpha, PGC-1alpha, AMP-activated protein kinase (AMPK), STAT3, and ERK1/2 MAPK pathway (163–166). Apart from influencing mRNA expression, leptin also decreased D9D half-life, which could additionally contribute to decreased D9D activity (147). Nevertheless, with increased adiposity, there is a central and peripheral resistance on leptin effects (caused by its diminished transport across the blood–brain barrier, decrease in the leptin receptor mRNA expression and density in the hypothalamus, hepatocytes, adipose tissue and muscles, as well as the deterioration of leptin-receptor function and signaling) (167), and therefore leptin cannot exert its inhibitory effect on Scd1 and Fads2 expression. On the other hand, to the best of our knowledge, no studies examined effect of leptin on Fads1 expression and D5D activity, even though in the mentioned study with central leptin infusion (163), also the increased liver 20:4n-6 values were shown, in parallel with increased 18:2n-6 values, which could indicate inhibited D6D activity, but stimulated D5D activity with leptin administration. All together, these data could suggest that in leptin resistant states (including simple obesity), there could be lack of both inhibitory effect of leptin on D6D and D9D activity and stimulatory effect of leptin on D5D activity.

Nevertheless, as already mentioned, the association of the increased adiposity and desaturase activities can be bidirectional. Vice versa, the increased activities of desaturases can predispose for adiposity development. For example, mice that lack the Scd1 gen activity are lean, resistant to DIO (despite increased food intake), have increased basal metabolism and fat oxidation, decreased de novo lipogenesis (DNL) and fat accumulation in the liver and adipose tissues, lower insulin resistance and more favorable metabolic profile (lower blood TG, GLU and insulin levels). On the other hand, they have increased oxidative and inflammatory liver damage and endoplasmic reticulum stress, due to increased SFA accumulation and oxidation in hepatocytes, as well as increased atherosclerosis (168–172). Liver-specific knock-out of Scd1 gen protects from the HSD induced adiposity in the same degree as the whole body Scd1 knock-out, but not from the HFD induced adiposity, indicating that in rodents, the liver D9D has a vital role for obesity development particularly in circumstances with high carbohydrate intakes, while in circumstances with high fat intakes, the adipose tissue D9D is more involved (171, 173). Especially liver D9D-18 activity (but not D9D-16 activity) contributes to DIO development, as shown in transgenic animals on HSD (174). Pharmacological inhibition of D9D showed similar effects as the genetic knock-out of Scd1 (169). The similar (but not identical) effects were also was shown with knock-out of Fads1 and Fads2 genes (175–181) or pharmacological inhibition of D5D and D6D in mice (94, 145, 182): protection from HFD- DIO, insulin resistance, unfavorable metabolic and pro-inflammatory profiles in the case of D5D deficiency, but only protection from DIO with partial protection from unfavorable metabolic profiles in case of D6D deficiency. In accordance, even in humans genetic polymorphisms of Scd1 (183–188), Fads2 and Fads1 (189–202) genes were related to increased predisposition to weight gain, increased adiposity, abdominal adiposity and related traits (insulin resistance, high GLU, TAG and LDL-C, low HDL-C, and pro-inflammatory adipokine/cytokine/chemokine profiles), with the effects varying depending on dietary, gender and ethnic interactions (102, 184, 185, 187, 188, 194, 197, 203–205) Therefore, the evidence supports both directions in the associations of increased adiposity with desaturases activities.

The more significant associations of D6D and D5D with increased adiposity compared with D5D found in this study can be causal (as explained above), but could also reflect the methodological issues in estimation of enzymatic activity. Desaturase indices, calculated as product/substrate ratios, are only an indirect measure of desaturase activity and do not have to represent the real activity of the enzyme (206). For example, there could be also a process of retro-conversion of 20:4n-6 to 20:3n-6 (207). Moreover, the product of D5D, 20:4n-6, is further metabolized to 22:4n-6 by Elov2, so the activity of this elongase could additionally affect the estimated D5D activity. In our study, 22:4n-6 values were about 25% of 20:4n-6 values, which indicates that significant proportion of 20:4n-6 is metabolized to 22:4n-6. However, in our study we did not find the associations of the estimated Elov2 activity (calculated as 22:4n-6/20:4n-6 or 22:5n-3/20:5n-3 ratios) with anthropometric indices, nor with D5D activity (data not shown), so the effect of Elov2 did not significantly influence results for D5D, which is in line with the literature data (97). Similarly, membrane bond 20:4n-6 in erythrocytes is known to be non-enzymatically oxidized/peroxidized or metabolized to eicosanoids and endocannabinoids by activity of phospholipases, cyclooxygenase, lipoxygenase, cytochrome P450-dependent enzymes, phosphatases and diacylglycerol-lipase, and their activity can be changed in obesity and MetS, which are considered as pro-inflammatory states (208, 209). Unfortunately, the eicosanoid and endocannabinoid levels in the blood were not assessed in this study. Furthermore, while the products of D6D and D9D (i.e., 20:3n-6 and 16:1n-7) are rarely found in the diet, and mostly reflect endogenous synthesis, 20:4n-6 is also provided by animal dietary sources (including meat, poultry, eggs, fish, and dairy products) (210). Therefore, its proportions in erythrocytes PL can be increased by dietary intakes and can erroneously inflate the estimated D5D activity with the increased dietary consumption of animal sources food. Nevertheless, it still seems that 20:4n-6 much more derives from endogenous conversion of 18:2n-6, 18:3n-6 and 20:3n-6 (211, 212), since large dietary intakes of 18:2n-6 much overcome low dietary intakes of 20:4n-6, and it is estimated that dietary 20:4n-6 contribution is less than 1% of total 20:4n-6 pool (213). In agreement, some studies showed only slightly lower 20:4n-6 levels in vegans compared with omnivores, while others even did not find the difference, despite huge differences in 20:4n-6 intakes (214, 215). In our study, we could not distinguish the intake of 20:4n-6 from different sources (animal vs. plant), and thus endogenous from exogenous 20:4n-6, due to methodological limitations, but 20:4n-6 levels were positively predicted by the levels of 18:2n-6, D6D and D5D, and that model completely explained the 20:4n-6 content, indicating that 20:4n-6 levels are principally determined by endogenous synthesis in our study. Interestingly, both D6D and D5D desaturases equally explained the variability in 20:4n-6 (together up to 71.6% of 20:4n-6 variability in our model). The addition of 18:2n-6 in the model explained 100% of variability of 20:4n-6. Finally, our data could also suggest that D5D activity could be more influenced by some intrinsic factors (e.g., genetic polymorphism) than by external factors (compared with D6D and D9D), which is in agreement with genetic polymorphism studies that underline genetic influence on D5D activity (216–219). For example, in a study which examined the FAs profiles in weight-discordant monozygotic twin pairs, the estimated activity of D5D was not different with obesity, while the estimated activities of D6D and D6D were increased in the obese twin (220). Nevertheless, the results of genetic studies on Fads1 polymorphisms and obesity are often contradictory, with more effects observed in children compared with adolescents and adults (192, 193, 198, 202), and it appears that other factors can still overcome the genetic influences.

Despite all these considerations, it seems that the most reasonable explanation for the less significant associations of increased adiposity with D5D activity is that relation of D5D activity with anthropometric indices was the inverse U-shaped. Both underweight and overweight subjects had lower estimated D5D activities, compared with normal weight subjects in our study, while obese subjects had the lowest D5D activities compared to all others. The non-parametric locally weighted regression showed an increasing trend for D5D activity up to BMI cut-off of about 23–24 kg/m², and after which there was a decreasing trend. Nevertheless, in the stratified analyses across BMI quartiles, only in overweight/obese subjects there was a significant association of BMI with D5D activity. However, due to the small number of subjects in quartiles, we could not perform proper stratified regression analyses, and these results should be interpreted with caution. In contrast, D6D and D9D activities (as well as 16:1n-7 and 20:3n-6 levels) showed a more linear (exponential) increase from underweight to overweight subjects, while for the 18:0 content, a linear decrease was shown. These findings can be explained by stimulation of D6D and D5D by insulin (which directly correlates with anthropometric indices, i.e., level of adiposity), and the biphasic effect of insulin on D5D activity. While in the insulin sensitive subjects (with lower BMI) insulin can have positive effect on D5D, in the insulin resistant subjects (with higher BMI), there can be an inhibitory effect, as suggested by studies in rats (156, 160–162). Unfortunately, we did not asses the insulin levels in our subjects to estimate insulin sensitivity at least by HOMA-IR index (221). Additionally, measuring leptin levels could be of interest, to assess the effect of both low and high leptin levels on D5D activity. As mentioned above, there are not published studies on the effect of leptin on D5D activity (only on the effect on D6D and D9D activity) (163).

The inverse U-shaped association was un unexpected finding, and to the best of our knowledge, there are no studies which reported such association. In the majority of the published studies, only normal weight and obese subjects were compared, and rarely the subjects with lower BMI were involved in research of FAs profiles. Our unexpected finding raises the need for more studies in this group of subjects, but interestingly, there are no studies on that topic. Additionally, subjects with eating disorders or subjects with lipodystrophic syndromes can be studied. Of note, the later ones represent insulin resistant subjects (222), while in eating disorders – e.g., anorexia nervosa (AN), both increased and diminished insulin sensitivity can be seen, depending on the muscle mass atrophy and abdominal fat accumulation (223, 224). Lipodystrophic subjects have disturbed lipid cell metabolism and they could have genetically determined changes in desaturases activities (225, 226), while subjects with AN can have completely different nutrient intakes (particularly of essential FAs) and disturbed metabolism and endocrine function (227–232), thus they make specific groups of underweight subjects, not suitable for comparison with general population.

As mentioned above, in our study three FAs were linearly correlated with anthropometric indices of adiposity, and both centripetal and peripheral fat accumulation: 20:3n-6, 16:1n-7 and 18:0. However, 18:0 correlated less with adiposity indices than 20:3n-6 and 16:1n-7. Our findings for 20:3n-6 and 16:1n-7 are in accordance with all other findings in the literature, but our data for 18:0 can differ from other studies, where the levels of 18:0 in both erythrocytes PL and plasma (whole plasma, plasma FFA, CE and PL) were positively related to indices of adiposity (20, 26, 98, 101, 233–235). However, more obese subjects and both men and women were included in the mentioned studies, and there could be also some dietary influences.

With increased adiposity, the levels of 18:0 could be decreased because of the increased activity of D9D (which transforms 18:0 to 18:1n-9) and the decreased activity of Elov6 (which produces 18:0 from 16:0) (97, 220). In our study, we also observed a significant negative correlation of the estimated Elov6 activity (calculated as 16:1n-7/18:1n-9 and 16:1n-7/18:1n-7 ratios) with the indices of adiposity and visceral fat accumulation (data not shown). This is in accordance with the mentioned study in weight discordant twins (220) and study of Zhao and associates (97).

Another interesting question arises from our research: it is known that n-3 PUFA can ameliorate NAFLD, ALT and AST levels (236–239), but here we have found a positive correlation of 20:5n-3 with the ALT/AST ratio, as an indicator of liver steatosis, even after controlling for all possible confounders (including dietary n-3 PUFA intakes and n-3 PUFA supplements usage), or after excluding the subjects with n-3 PUFA supplementation. This is the opposite from what one would expect, considering that people with NAFLD have the lower n-3 PUFA content in their livers and that n-3 PUFA can ameliorate NAFLD, by increasing beta oxidation and decreasing DNL and liver TAG synthesis (146, 236, 240–243). One possible explanation is that in NAFLD there can be a lower conversion rate of 20:5n-3 to 22:6n-3. Indeed, we observed in this study a significant negative correlation of estimated “putative D4D” activity (calculated as ratio 22:6n-3/20:5n-3) with ALT/AST ratio (r = −0.536, p < 0.001) and ALT (r = −0.353, p = 0.002) (unpublished observations). The linear regression models also confirmed the negative association of “putative D4D” activity with ALT/AST ratio (data not shown). Even though instead of putative D4D enzyme there are actually several different steps during process of conversion of 20:5n-3 to 22:6n-3, which include Elov2/5, D6D and peroxisomal beta-oxidation by acyl-CoA oxidase and 17β-hydroxysteroid dehydrogenase 4 (64, 146), the ratio 22:6n-3/20:5n-3 can show the rate of this conversion, which in humans is quite low (240), and can be further reduced in states of hepatic steatosis and steatohepatitis (244–246), so there could be accumulation of 20:5n-3 and 22:5n-3 in such states. In agreement, also 22:5n-3, the intermediate Elov2/5 product during this conversion, positively correlated with ALT/AST ratio, although less than 20:5n-3. In addition, we also observed the positive correlation of 22:5n-3 with GGT (r = 0.469, p = 0.012). However, there could be other possible mechanisms to explain the positive association of 20:5n-3 with ALT/AST ratio. For example, there could be a possibility of increased desaturation and elongation of 18:3n-3, increased retroconversion of 22:6n-3 to 20:5n-3, reduced beta-oxidation of 20:5n-3 or reduced transformation by cyclooxygenase and lipoxygenase to eicosanoids in hepatic steatosis. However, quite the opposite was described in NAFLD (142, 239, 240, 243, 247), and both the retroconversion of 22:6n-3 to 20:5n-3 and conversion of 18:3n-3 to 20:5n-3 and 22:6n-3 in humans are relatively inefficient (207, 248–251). Furthermore, ALT is more a marker of liver steatosis and gluconeogenesis, while AST is more a marker of cellular damage, and it could be possible that the anti-inflammatory and anti-oxidative 20:5n-3 properties overcome its influence on hepatic lipid accumulation (243, 252, 253). In agreement, supplementation with n-3 FAs (fish oil) increased ALT/AST ratio in rats on a hypercholesterolemic HFD (254), and in some studies in humans, the more potent effect on AST than ALT reduction was observed, but not in all, and in some even more potent effect on ALT reduction was shown (236, 240, 243, 255, 256). Therefore, the observed positive association probably cannot be explained by the possible effect of 20:5n-3 on ALT/AST ratio. Moreover, none of our subjects had severe liver damage, and the association was present across different BMI-categories and BMI-quartiles, although the association was stronger in higher BMI quartiles (data not shown).

The positive associations of 16:1n-7 and VFL with ALT/AST ratio is not surprising, bearing in mind that 16:1n-7 is the one of the best indicators of the increased DNL, which is a hallmark of NAFLD (77, 240). Even though many other FAs are also the products of DNL, including SFA from C12:0 to C18:0 and other MUFA (18:1n-9 and 18:1n-7) (42, 48), 16:1n-7 is probably the best indicator of DNL, due to its low abundance in common dietary sources and high correlation of D9D activity with activity of other enzymes involved in DNL, since they are all regulated by the same regulatory elements, SREBP-1c and ChREBP (77, 257, 258). In accordance, D9D-16 is much better indicator of DNL compared with D9D-18 or ratio 16:0/18:2n-6 (also known as DNLIndex) (40, 77). Our data also confirm the association of D9D with ALT and ALT/AST ratio, and very strong association of D9D with VFL.

The found associations of indicators of cardiometabolic risk as well as desaturase activities and FAs profiles with age, postmenopausal status, smoking, moderate alcohol consumption, physical activity and educational level are more/less in accordance with previously published data (20, 53, 54, 59, 259). Smoking, age, and physical activity all increase oxidative stress, which can affect desaturase activities (74, 95, 96, 260, 261). Age was associated with 16:1n-7 and D9D, because of age-related increases in visceral fat mass (262–264). When VFL was also included in the model, VFL became the only significant predictor of 16:1n-7 and D9D (data not shown). Data on menopausal status are scarce, and one study did not find the differences in FAs profiles (106). However, some other studies have shown some differences (61), but the study pre- and post-menopausal groups were very different regarding age and BMI, so the effect of menopause could not be clearly distinguished from the effects of age and BMI. In one large-sample study (265), only the levels of 16:0 were negatively associated with menopause, which in accordance with our study results. In our study, after adjusting for multiple confounders (including age), only the negative association of 16:0 with menopause remained significant (r_pb = 0.256, p = 0.033). According to literature data, smoking has negative associations with D5D (and much less D6D) activity, while has positive associations with D9D activity (20, 74, 76, 96, 266–268). In our study we also have found negative associations with D5D and positive associations with D9D activity, but after adjustments for multiple confounders only negative associations with D5D activity remained. In majority of the studies, physical activity was in positive association with D5D activity, and negative association with D9D and D6D activities (269–273), what is similar with our results, even though in linear regression models the physical activity level was not a significant predictor, indicating that other factors (e.g., BMI, smoking, age) confounded this association, which is in agreement with results of Warensjö and associates (20).

Surprisingly, in our study we did not obtain the significant associations of both desaturase activities and biochemical indices with dietary factors, except for negative associations of anthropometric indices with total energy and macronutrients intakes (except for n-3 PUFA), negative associations of PUFA intakes (particularly n-6 PUFA) with biochemical indices of cardiometabolic risk, and positive associations of 20:5n-3 with n-3 PUFA intake and with n-3 PUFA or n-3 PUFA/Zn supplementation. Nevertheless, only the associations of 20:5n-3 with n-3 PUFA intake and n-3 PUFA and Zn supplementation did remain significant in linear regression models. The associations of total n-3 PUFA dietary intakes (from both plant and animal sources) with 20:5n-3 levels, but not 22:6n-3 levels, can be explained by low conversion rates of 18:3n-3 to 22:6n-3 (much lower than 20:5n-3) (215, 250, 274). In contrast to what could be expected from the literature (51, 64, 65, 275), no other significant associations of desaturase activities or FAs levels were found with dietary data obtained from 24 h-recalls, related to total energy, carbohydrate, protein, total fat, and (particularly) SFA, MUFA and PUFA intakes. One of the reasons for that is that only two 24 h-recalls were used for dietary assessments, while usually much longer dietary surveys (up to 15 days), or validated FFQ were used to precisely estimate different dietary fats intakes (210, 276, 277). We did not have FFQ data (only FPQ) to estimate dietary intakes over a year, and 24 h-recalls can reflect recent dietary intakes, while erythrocytes FA profiles more represent the long-term dietary intakes compared with plasma FA profiles (67, 278, 279). Nevertheless, also other authors noted a week correlation between macronutrient nutritional intakes and desaturase activities in erythrocytes, except for n-3 PUFA (22:6n-3, 20:5n-3) and TFA (278). There is also the problem with self-reported data, including under-reporting, which is particularly common among obese subjects (109, 280). In line with that, the possible under-reporting could also explain the observed negative associations of total energy intake and intakes of almost all macronutrients with anthropometric indices indicative for obesity.

This study confirmed that the levels of certain FAs (16:1n-7, 20:3n-6, and 18:0) and estimated desaturase activities in erythrocytes PL were significantly associated with anthropometric, clinical and biochemical cardiometabolic risk indicators in healthy, non-diabetic Serbian women. However, the associations of clinical and biochemical indicators of cardiometabolic risk were not independent from the associations with the level of adiposity. After controlling for multiple confounding factors (including age, postmenopausal status, smoking, physical activity, educational level, alcohol consumption, use of supplements and dietary macronutrient intakes), the level of adiposity was the most significant predictor of the estimated desaturase activities and 16:1n-7, 20:3n-6, and 18:0 levels, and mediated their association with biochemical and clinical indicators. Vice versa, desaturase activities predicted the level of adiposity, but not other components of cardiometabolic risk (if the level of adiposity was also accounted). While the associations of 16:1n-7, 20:3n-6, 18:0, D9D and D6D activities with anthropometric indices were linear (positive with 16:1n-7, 20:3n-6, D9D and D6D, and negative with 18:0), the associations of D5D activity with anthropometric indices were the inverse U-shaped. The only adiposity-independent association of FAs profiles with indicators of cardiometabolic risk was the association of 20:5n-3 with ALT/AST ratio, which requires further exploration. More studies are needed to explore the mechanisms of the observed associations of FAs profiles and desaturase activities with anthropometric, clinical, and biochemical cardiometabolic risk indicators, including larger-scale confirmatory studies.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Institute of Occupational Health Niš Ethics Board. The patients/participants provided their written informed consent to participate in this study.

IŠ wrote the manuscript and performed anthropometric measurements, nutritional interviews, software-exploration of nutritional data, and all statistical analyses. JDM, MT, NKV, TP, and JJ performed chromatographic and biochemical analyses. VS, JM, and MZ were involved in anthropometric and dietary assessment. All the authors contributed to the manuscript writing.

The authors acknowledge financial support from the Ministry of Education, Science and Technological Development of the Republic of Serbia (project number III41030 2011-2017 and contract number: 451-03-68/2022-14/200015) and Ministry of Science, Technological Development and Innovation of the Republic of Serbia (contract number: 451-03-47/2023-01/200015).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.