Results from whole genome sequencing combined with metabolomics facilitated a more complete analysis of the impact of genome-wide genetic variation (including common and low-frequency genetic variants) on the biologically-important structural and signaling lipid molecules. These data support the FADS locus, and particularly FADS1, as a central genetic control point for determining circulating levels of molecular species representing four distinct classes of physiologically-important lipids (phospholipids, lyso-phospholipids, free fatty acids and an endocannabinoid). Most recent studies have described the biochemistry and genetics of LC-PUFAs biosynthesis in the context of an oversimplified pathway of FADS1/2 and ELOVL5/2-encoded desaturation and elongation steps that produce unesterified LC-PUFAs (Fig. 1; blue arrows). In fact, it is typically LC-PUFA-containing complex lipids, intermediates and signaling metabolites such as PLs, lyso-PLs, endocannabinoids, and eicosanoids that have vital structural and signaling functions within physiological systems. Even when unesterified LC-PUFAs and their metabolites impact cellular function, they are not directly derived from this desaturation-elongation pathway, but several biochemical steps (Fig. 1; green arrows, purple arrows) downstream after they have been incorporated into PLs and then mobilized by phospholipases. Importantly, genetic association studies to date have largely utilized an analysis approach in which total lipid fractions, or in some cases phospholipids, are isolated, and then, all fatty acids including PUFAs and LC-PUFAs are removed from complex lipids (typically a glycerol lipid or cholesterol backbone) by a saponification step³⁹. This in turn leaves the PUFA or LC-PUFA for analysis but leaves no evidence as to the biologically-relevant lipid molecular species most impacted by variation in the FADS locus. Additionally, since fatty acids were removed from total lipid or phospholipid fractions in those studies, little information is available to determine whether variants in the genes that encode the numerous non-FADS encoded biochemical steps (such as those shown in Fig. 1) are important in regulating lipid molecular species levels. Gieger and colleagues⁴⁴ initially addressed this complex issue by examining associations between genetic variation and 363 metabolites in the serum of 284 male participants. Several SNPs in a linkage disequilibrium (LD) block containing the FADS1 gene were strongly associated with a number of LC-PUFA-containing phospholipid molecular species. The strength of these associations were shown to be dramatically increased when product/precursor flux (ratios) at the FADS1 enzymatic step were examined. This and a subsequent study⁴⁵ demonstrated large effect sizes for the FADS1 gene variant associations and showed common genetic variation in FADS1 have major effects in the “metabolic make-up” of individuals in human populations.

We have built upon these studies by utilizing large-scale published data where whole-genome sequencing and more complete metabolomic profiling was conducted in 1,960 adult participants⁴⁰. Metabolomics provided an opportunity for us to examine enzyme substrates and products in such a way that the metabolome serves as a surrogate measure of enzyme activities responsible for lipid molecular species⁴⁴. The most significant associations identified were located within the FADS locus. FADS1 encodes for the ∆-5 desaturase enzymatic step in the LC-PUFA biosynthetic pathway (Fig. 1). Individual minor allele variants were strongly associated with accumulation of LA and DGLA, precursors to the FADS1 and FADS2 enzymatic steps respectively, as found in phospholipid and lyso-phospholipid molecular species. In contrast, FADS minor allele variants were associated with lower levels of phospholipids, lyso-phospholipids, free fatty acids and an endocannabinoid containing n-6 LC-PUFAs (ARA and adrenic acid), n-3 LC-PUFAs (EPA and docosapentaenoic acid), and a n-9 LC-PUFA (mead acid). It is not possible to discern the impact of the FADS1 and FADS2 steps individually due to the genetic architecture of the FADS locus, which has long regions of LD throughout. However, the biochemical pattern demonstrating elevated levels of precursor PUFAs and LC-PUFAs upstream of the FADS1 (∆5 desaturase) step are associated with FADS minor allele variants and lower levels of LC-PUFAs downstream of the FADS1 step suggests that the FADS1 step is the major biosynthetic event impacted by variation in the FADS locus. This observation of FADS locus variation impacting the FADS1 biochemical step is consistent with other studies^15,44. Notably, there were a few exceptions to the observed pattern, including 1-arachidonyl-2-lyso-GPE in which an ARA-containing lyso-phospholipid was directly associated with some FADS minor alleles (e.g. rs174564, effect size, SE = 0.59, 0.24) and inversely associated with other FADS minor alleles (e.g. rs174529 effect size, SE = − 0.89, 0.24). Given the high LD of rs174564 and rs174529 (R² = 0.933 in 1,000 genomes (GBR) British in England and Scotland population), these bidirectional effect directions were unexpected. Additional studies will be necessary to better understand why some of the metabolites, including 1-arachidonyl-2-lyso-GPE, appear to have a bidirectional associations with FADS variants in high LD.

Perhaps the most important finding of this study, from a metabolism perspective, is that the FADS locus has the strongest genetic impact on the identified LC-PUFA-containing lipids including phospholipids, lyso-phospholipids, free fatty acids, and an endocannabinoid, with minimal to no association outside of this locus. For example, the FADS locus is the only region to associate at genome-wide significance with levels of biologically-important and diverse forms of ARA-containing molecular species including unesterified ARA, 2-ARA glycerol, 1-acyl-2-ARA-GPC, and 1-ARA-2-lyso-GPC (Fig. 3). This is unexpected because a molecule such as 2-ARA glycerol is formed from ARA-containing phospholipids (catalyzed by several steps after the FADS1 enzyme step; green arrows) utilizing phospholipase D + phosphatase steps or after PIP kinase + phospholipase C + diglyceride lipase steps (Fig. 1; orange arrows). In other words, the biosynthesis of this endocannabinoid requires 6–8 additional biochemical steps beyond the FADS1 enzymatic step that initially formed the ARA from DGLA. While Fig. 1 outlines the metabolism of only ARA-containing lipids, similar pathways (utilizing many of the same enzymes) could be drawn for the n-3 (eicosapentaenoic and docosapentaenoic acids) and n-9 (mead acid) LC-PUFAs.

Importantly, only 9 of the 48 LC-PUFA metabolites in association with the peak-associated SNP rs174564 have genome-wide significant signals with genetic loci outside of the FADS locus. Of note, MBOAT7, which encodes for lyso-GPl acyltransferase that has specificity for arachidonoyl-CoA as an acyl donor in the remodeling of phosphatidylinositol (Fig. 1), was associated with linolenoyl- and arachidonoyl-containing GPI molecular species. These data, which utilize genome-wide common and low-frequency genetic variants, provide confidence that the original study upon which our inferences are made is robust. Furthermore, the current study considered a high proportion of the DNA sequence variation in the genes that could impact LC-PUFA lipid levels and validates the central role of FADS variation in the formation of biologically-important lipids.

With regard to the physiological importance of the lipids impacted by FADS variation, circulating endocannabinoids and particularly 2-AG plays a critical role in almost all physiological processes, including energy balance and metabolism, thermoregulation, appetite, immune function, stress responses, memory, physical activity, reproduction, and sleep (Fig. 4). The necessity to precisely regulate levels of 2-AG and orchestrate such a wide variety of functions must have been created by intense evolutionary pressures. Fumagalli and colleagues compared signals of selection in an Inuit population and European and Chinese populations to better understand how humans adapted to the Arctic environment¹⁹. Variants in the FADS cluster were the strongest targets of selection and were also strongly associated with multiple metabolic and anthropomorphic phenotypes (smaller body size and shorter stature). These investigators speculated that FADS variation led to changes in LC-PUFAs, which in turn affected levels of growth hormones resulting in anthropomorphic changes. While this may be the case, the current study suggests that the direct pathway association of FADS variation with 2-AG levels could also be a key molecular mechanism linking the FADS locus to several of the aforementioned metabolic and anthropomorphic phenotypes (Fig. 4).

The brain is highly enriched with n-3 and n-6 LC-PUFAs-containing phospholipid pools^46–48, and studies to date suggest the brain has a limited capacity to synthesize these n-6 and n-3 LC-PUFAs. Consequently, most LC-PUFAs are thought to move from circulation across the blood brain barrier. Recent studies reveal that much of the acquisition of LC-PUFAs by the brain is mediated by a transporter, Mfsd2a, which moves n-3 and n-6 LC-PUFAs as lyso-GPCs from circulation across the blood brain barrier^49–53. The current study shows that FADS variants are strongly associated with levels of n-6 and n-3 LC-PUFA-containing lyso-GPCs, including 1-arachidonoyl-2-lyso-GPC, 1-lyso-2-arachidonoyl-GPC, 1-docosapentaenoyl-2-lyso-GPC, and 1-eicosapentaenoyl-2-lyso-GPC, which are substrates for the Mfsd2a transporter. Taken together, these data suggest FADS variation has the capacity to alter the content of brain phospholipids (Fig. 4).

The other two lipid classes impacted by FADS variation are n-6 and n-3 LC-PUFAs within phospholipids or as unesterified fatty acids. Levels of LC-PUFA-containing phospholipids within membranes impact a wide range of structural and communication functions. This occurs through numerous mechanisms such as affecting membrane fluidity and serving as substrates for large families of phospholipases (such as PLA₂, PLA₁, PLC, and PLD) in order to generate metabolic intermediates for a wide assortment of signaling molecules (Fig. 1; orange, green and purple arrows). The current study confirms that FADS variation is associated with both levels of LC-containing PLs and unesterified LC-PUFAs, which is linked to levels of eicosanoids in stimulated human blood⁵⁴.

Dramatic ancestral-based differences in genetic variation within the FADS locus have been identified with an ancestral haplotype (minor alleles in this study) associated with reduced capacity to synthesize LC-PUFAs. This haplotype is nearly fixed in Native American and Greenland Inuit populations, elevated in Amerindian-Ancestry Hispanic populations, and virtually absent in African populations²⁰. In contrast, there is also a derived haplotype (major alleles in this study) that is associated with higher levels of LC-PUFAs, which is fixed in Africa, elevated in African Americans, and is observed at varying frequencies (25–50%) in Europe and East Asia^18–22,55. A rapidly emerging literature indicates that the FADS locus, and thus the efficiency of LC-PUFA biosynthesis, is an evolutionary ‘hotspot’ that has been targeted numerous times by positive selection during human evolution. The current study illustrates the genetic impact of the FADS locus and particularly the FADS1 ∆-5 desaturation step on the formation of molecular species from four biologically-crucial classes of LC-PUFA-containing lipids. These affect a wide diversity of functions including innate immunity, energy balance, and brain development (Fig. 4). This study also illustrates the potential for rapid nutritional transitions, such as modern diets that induce broad evolutionary discordance driven by numerous molecular mechanisms that in turn impact a wide range of diseases/disorders. These transitions may then provide the framework for a better understanding of the molecular underpinnings behind gene-diet interactions that contribute to the biology of racial/ethnic health disparities.