Association of an ACSL1 gene variant with polyunsaturated fatty acids in bovine skeletal muscle

Background The intramuscular fat deposition and the fatty acid profiles of beef affect meat quality. High proportions of unsaturated fatty acids are related to beef flavor and are beneficial for the nutritional value of meat. Moreover, a variety of clinical and epidemiologic studies showed that particularly long-chain omega-3 fatty acids from animal sources have a positive impact on human health and disease. Results To screen for genetic factors affecting fatty acid profiles in beef, we initially performed a microsatellite-based genome scan in a F2 Charolais × German Holstein resource population and identified a quantitative trait locus (QTL) for fatty acid composition in a region on bovine chromosome 27 where previously QTL affecting marbling score had been detected in beef cattle populations. The long-chain acyl-CoA synthetase 1 (ACSL1) gene was identified as the most plausible functional and positional candidate gene in the QTL interval due to its direct impact on fatty acid metabolism and its position in the QTL interval. ACSL1 is necessary for synthesis of long-chain acyl-CoA esters, fatty acid degradation and phospholipid remodeling. We validated the genomic annotation of the bovine ACSL1 gene by in silico comparative sequence analysis and experimental verification. Re-sequencing of the complete coding, exon-flanking intronic sequences, 3' untranslated region (3'UTR) and partial promoter region of the ACSL1 gene revealed three synonymous mutations in exons 6, 7, and 20, six noncoding intronic gene variants, six polymorphisms in the promoter region, and four variants in the 3' UTR region. The association analysis identified the gene variant in intron 5 of the ACSL1 gene (c.481-233A>G) to be significantly associated with the relative content of distinct fractions and ratios of fatty acids (e.g., n-3 fatty acids, polyunsaturated, n-3 long-chain polyunsaturated fatty acids, trans vaccenic acid) in skeletal muscle. A tentative association of the ACSL1 gene variant with intramuscular fat content indicated that an indirect effect on fatty acid composition via modulation of total fat content of skeletal muscle cannot be excluded. Conclusions The initial QTL analysis suggested the ACSL1 gene as a positional and functional candidate gene for fatty acid composition in bovine skeletal muscle. The findings of subsequent association analyses indicate that ACSL1 or a separate gene in close proximity might play a functional role in mediating the lipid composition of beef.


Background
In recent decades, the continuing accumulation of knowledge and the increasing number of reports providing evidence regarding the beneficial health effects of polyunsaturated fatty acids (PUFA) have attracted the attention of the medical and public community. Consumers are becoming increasingly aware of the relationships between diet and health and also of the importance of the diet for general physical and mental wellbeing [1,2]. Many clinical and epidemiologic studies have indicated a positive impact of long-chain omega-3 fatty acids (n-3 long-chain polyunsaturated fatty acids, n-3 LC-PUFA) on human health and disease. Beneficial effects of n-3 LC-PUFA are described in infant development, cancer, and cardiovascular diseases (e.g., [3][4][5][6]), lipid and glucose metabolism (e.g., [7][8][9][10]), inflammation (e.g., [11,12]), and more recently, in various mental illnesses including depression, attention-deficit hyperactivity disorder, and dementia (e.g., [13]). It has been demonstrated that diets containing higher levels of n-3 LC-PUFA [namely DHA (docosahexaenoic acid; C22:6n-3) and EPA (eicosapentaenoic acid; C20:5n-3)], may reduce cardiovascular risk in diabetes by inhibiting platelet aggregation, improving lipid profiles, and reducing cardiovascular mortality. Thus, n-3 LC-PUFA were particularly recommended to people with diabetes and metabolic disorders associated to obesity [5,14]. Their beneficial health effects may be mediated through multiple distinct mechanisms, including alterations in cell membrane composition and function, gene expression, or eicosanoid biosynthesis [15,16]. It is known that n-3 LC-PUFA can exert important metabolic effects due to their ability to modulate the transcription of regulatory genes with function in lipid metabolism [17][18][19][20][21].
The n-3 LC-PUFA, like DHA and EPA, are particularly abundant in oily cold-water fish and seafood, however, they are also present in other animal products (e. g., ruminant meat and milk) but in lower concentrations. Increases of n-3 LC-PUFA content in the human diet can be achieved by dietary supplementation, but there is also a potential to alter the natural fatty acid (FA) profile in food from animals. FA composition of meat and milk reflects both, FA biosynthesis in the respective animal tissue and FA composition of ingested nutrients. A recent study showed that cattle and lambs fed grass-diet in the period before slaughter had an increased content of beneficial FAs in meat, and that subsequent moderate consumption of the respective meat had resulted in increased plasma and platelet n-3 LC-PUFA concentrations in healthy human individuals [22]. A ruminant diet on grass, which is rich in α-linolenic acid (C18:3n-3, ALA) compared to cereal-based concentrate diet can influence the FA profile of meat in the desired direction and improve its nutritional value [23][24][25]. However, the link between nutritional intake of FAs and its subsequent concentration in skeletal muscle is stronger in monogastric animals (pigs, poultry) than in ruminants due to hydrogenation of dietary FAs in the rumen (e.g., [26]).
In addition to the environmental conditions, genetic factors may also have a substantial effect on the variability of FA composition in animal products, especially for ruminants [27]. Consequently, genetic selection and breeding of animals with favorably enriched n-3 LC-PUFA content in skeletal muscle can provide a rich source of the desired beneficial FAs for the human diet. Therefore, it is necessary to elucidate the moleculargenetic background of fatty acid composition in bovine skeletal muscle for identifying the genes or gene variants favorable for human nutrition.
However, the biochemical processes and the molecular background affecting the genetic variability of the complex polygenic trait of FA composition are not yet completely understood, particularly with regard to European cattle breeds, because the majority of recent studies have been performed on the very specific genetic background of Japanese Black cattle.
Therefore, the aim of this study was to identify genetic factors affecting the variation of FA composition in bovine skeletal muscle. For our study, we took advantage of a unique F 2 resource population generated from the major European cattle breeds Charolais and German Holstein by means of embryo transfer and foster mothers [42]. In previous studies, this population had been shown to segregate for two major loci (NCAPG and MSTN) associated with prenatal and pubertal growth, postnatal body composition and general lipid deposition [43,44].

Results and discussion
The animals from our resource cross population were kept and fed at standardized uniform conditions and slaughtered at the same age. Therefore, we can exclude exogenous factors due to differences in herd, age, feeding and gender. Consequently, differences in skeletal muscle fatness or FA composition should be due to differences in endogenous factors of the animals like the genetic background. The primary focus of our study was to discover phenotypic differences of FA composition in skeletal muscle between the individual animals of the resource population due to genetic variation.
QTL analysis and identification of ACSL1 as a positional and functional candidate gene An initial QTL analysis in the Charolais × German Holstein cross population identified a QTL for FA composition on bovine chromosome 27 (BTA27) as exemplified for n-3 LC-PUFA in Figure 1. In our study, the trait n-3 LC-PUFA represents n-3 PUFA exceeding a carbon chain length of C18. The QTL interval corresponded to a region, where previously QTL affecting marbling had been detected in a Bos indicus × Bos taurus cross and two commercial US Angus populations [45,46]. The QTL explained 20.5% variance in the model calculated as the relative reduction of the residual variance due to including the QTL in the model [47].
Furthermore, QTL for FA composition, myristic acid, (C14:0) and oleic acid (C18:1) content, have been reported in this chromosomal region in a Jersey × Limousin back-cross cattle population [48]. In our study, the QTL interval affecting FA composition in skeletal muscle displayed a peak between 15 and 16 cM on our genetic map of BTA27 corresponding to a genomic position at approximately 16 Mb on the current bovine genome assembly of the chromosome (NCBI mapviewer, build 5.2, http://www.ncbi.nlm.nih.gov/projects/mapview/map_search.cgi?taxid = 9913).
Based on its chromosomal position and integration in biochemical pathways of lipid metabolism, we identified the acyl-CoA synthetase long-chain family member 1 (ACSL1) gene as the most plausible positional and functional candidate gene underlying the QTL with effect on FA composition on BTA27. The ACSL1 gene is located exactly under the peak of the QTL interval. Its protein, the ACSL1 enzyme, is known to catalyze the first step of activation of long-chain (LC) FAs by converting them into LC acyl CoA thioesters for channeling towards chain elongation, triacylglyceride synthesis or FA oxidation [49]. ACSL1 has a key function in both the synthesis of cellular lipids and FA degradation, and is also necessary for phospholipid remodeling [50]. Due to its physiological biochemical function, it can be suggested that ACSL1 plays an important role in lipid metabolism, insulin resistance and obesity. Recently, a study in humans reported that a gene variant located in intron 1 of the ACSL1 gene can influence the metabolic syndrome risk (characterized by insulin resistance, dyslipidaemia, abdominal obesity and hypertension associated to type 2 diabetes), and that this ACSL1 genotype-dependent effect can be modulated by dietary PUFA intake suggesting a genenutrient interaction [51].

Structure analysis and screening for polymorphisms of the ACSL1 gene
Although sequences for the ACSL1 gene and protein were deposited in the bovine genome databases, we found inconsistencies regarding the structural annotation of the gene in the bovine genome assemblies. A correct and conclusive structural gene annotation is a prerequisite for subsequent screening for gene variants and analysis of their functional relevance. Therefore, the first step of our study focused on the experimental confirmation of the structure of the ACSL1 gene on the genomic and cDNA level. Experimental verification by RT-PCR, re-sequencing and comparative sequence analyses confirmed the genomic annotation of the bovine ACSL1 gene in the alternate UMD_3.1 genome assembly (Figure 2), which is in contrast to the reference genome assembly Btau4.2 (http://www.ncbi.nlm.nih.gov/projects/ mapview/map_search.cgi?taxid = 9913).
Re-sequencing of DNA from pools and individuals differing in IMF content and Δ 9 desaturase activity index included a total of 8.5 Kb of genomic DNA. Comparative sequence analysis revealed a total of 19 single nucleotide polymorphisms (SNPs) in the targeted gene regions (Table 1). Three synonymous exonic (exons 6, 7 and 20), six intronic (introns 5, 6, 9, 13, 16 and 20), six SNPs in the promoter region and four SNPs in the 3'UTR of the bovine ACSL1 gene were detected ( Figure  2, Table 1). Eleven out of these SNPs identified in our study were novel and not previously represented in the SNP database (version 133) at NCBI.

Association of ACSL1 gene variants with PUFA profile in skeletal muscle
The association analysis included all exonic and intronic ACSL1 gene variants (except for the one in intron 16) and one SNP in the promoter region, which were identified by re-sequencing and validated by genotyping in the Charolais × German Holstein resource population. The nine SNPs analyzed in the Holstein × Charolais cross bred population showed a minor allele frequency ≥ 0.2 in the analyzed data set (Table 1). Intragenic linkage disequilibrium (LD) analysis revealed a strong LD between the SNPs in intron 20, exon 20, intron 13 and intron 9 (r 2 >0.9), whereas there was only a moderate LD (0.5 < r 2 <0.6) between these SNPs and the one in intron 5.  The association analysis with intragenic ACSL1 SNPs revealed that the SNP located in intron 5 of the ACSL1 gene (c.481-233A>G) showed the most significant associations with FA composition in skeletal muscle. The gene variant ACSL1 c.481-233A>G was significantly associated with the relative content of distinct fractions of unsaturated FAs, n-3 FA, PUFA, n-3 LC-PUFA and docosapentaenoic fatty acid (DPA) as well as with the absolute content of total FA, MUFA, and trans vaccenic acid (C18:1trans-11) in M. longissimus dorsi ( Table 2). The results revealed that the c.481-233A allele of this gene variant is strongly associated with a higher relative level of n-3 FA, PUFA, DPA, and n-3 LC-PUFA. In contrast, the c.481-233A allele showed a decreasing effect on content of C18:1trans-11, total FA, and MUFA, and tended to be associated with a lower IMF content in skeletal muscle compared to the c.481-233G allele. The c.481-233A allele had a higher freqeuncy (73%) in the analyzed population compared to the alternative allele (27%).
Although the c.481-233A allele tends to be associated with a slightly lower total IMF content, the relative content of the FA fractions, n-3 FA, PUFA, DPA, and n-3 LC-PUFA, known to exert health-beneficial effects in humans is highly increased indicating a higher nutritional value for beef originating from animals with the favorable ACSL1 allele.
The strongest allelic effect of the ACSL1 c.481-233A>G locus was observed for n-3 FA content. This trait also includes the polyunsaturated C18 fatty acids, α-linolenic acid (ALA, C18:3n-3) and stearidonic acid (C18:4n-3). The n-3-FA content is different to the trait n-3 LC-PUFA, which exclusively comprises n-3 FA with a chain length > C18. As an essential FA, ALA cannot be synthesized by mammalian species and must be obtained from the diet. The ALA concentration in skeletal muscle, therefore, could be linked to the dietary absorption. However, the standardized concentratebased feeding regimen in our study provides uniform feeding conditions for the animals. ALA is the precursor for the n-3 FA pathway [52] by serving as parent FA for the synthesis of stearidonic acid and n-3 LC-PUFA (EPA, DPA, and DHA) via sequential steps of desaturation and/or chain-elongation. The association of ACSL1 c.481-233A>G with DPA and with n-3 LC-PUFA (containing n-3 FA exceeding a chain length of C18) could suggest that a substantial proportion of their precursor ALA might be activated and channeled to chain elongation processes.
In contrast to the increasing effect associated with the c.481-233A allele on the relative content of the FA fractions, n-3 FA, PUFA, DPA, and n-3 LC-PUFA, and the PUFA/SFA and P/S ratios, we observed a decreasing effect of this allele on the absolute content of the trans vaccenic acid C18:1trans-11 in skeletal muscle in our study. This effect was in concert with the associated parallel decrease in total FA and MUFA content in the tissue. The effect on C18:1trans-11 is of particular interest, because trans vaccenic acid is a precursor of conjugated linoleic acid (CLAcis-9, trans-11) generation. CLAs are believed to have several important physiological functions, including anti-carcinogenic, anti-atherogenic, immunomodulating, growth and lean body mass promoting effects [55]. Thus, targeted selection of cattle carrying the homozygous c.481-233A/c.481-233A genotype in the ACSL1 gene would possibly be accompanied by detrimental effects on the CLA profile in skeletal muscle.
There is the open question, whether the significant effects of the ACSL1 gene variant c.481-233A>G on FA composition were due to general fatness differences in skeletal muscle, which is supported by several QTL for marbling in the targeted chromosomal region, or whether the effects were associated with the ACSL1 gene variant c.481-233A>G. Alternatively, the effects of this gene variant might modulate the accumulation of specific FAs in skeletal muscle. To address this issue, we extended our association analysis and fitted IMF as a covariate in the model. When adjusting for IMF (Table  2), the association of the ACSL1 gene variant c.481-233A>G with absolute content of trans vaccenic acid in skeletal muscle remained significant, whereas the other associations dropped below a stringent threshold of statistical significance (Bonferroni q < 0.1) and were only tentatively significant (e.g., for relative content of n-3 FA and DPA). Thus, we cannot exclude that variants in the bovine ACSL1 gene may exert a substantial effect on total intramuscular fat content, which indirectly affects intramuscular composition of specific FA fractions. However, as the results for trans vaccenic acid demonstrate, it is suggested that there are also direct effects associated with the ACSL1 gene variant c.481-233A>G on intramuscular content of specific FAs.

Conclusions
Due to our observation that the c.481-233A>G SNP in intron 5 of the ACSL1 gene cannot fully explain the QTL variance (Figure 1), we conclude that this gene variant is presumably not causal, but in LD to another not yet detected polymorphism in its close vicinity affecting FA composition in bovine skeletal muscle. Presumably, these effects are not exclusively the consequence of variation in intramuscular fat content, but due to effects on specific FA. Prior to selective breeding of cattle carrying the desired genotype of the ACSL1 gene variant c.481-233A>G in order to produce meat with specific FA profiles, the association between c.481-233A>G and FA composition has to be confirmed in the particular target cattle population.
Nevertheless, our results indicate that the ACSL1 gene might play a functional role in mediating the FA composition in bovine skeletal muscle and provide a basis to further elucidate the function of the ACSL1 gene and its coordinated network with genes integrated in FA metabolism to dissect the molecular background of lipid composition of beef.

Animals and phenotypes
The generation of the Charolais × German Holstein resource cross population (SEGFAM), details regarding feeding and housing of the animals analyzed in our study, have been previously described [43,44]. The animals were kept under standardized environmental and feeding conditions in barn facilities at the Leibniz Institute for Farm Animal Biology (FBN). After birth, the calves were fed a milk/replacer/hay/concentrate diet ad libitum until day 121. Thereafter, the animals received a feed ration of concentrates and chaffed hay with a hay to concentrate ratio of 1:3 and an energy content of 12.7 MJ ME/kg dry matter fed ad libitum until slaughter. The animals were kept in a tight stall barn with individual daily feed recording. At the age of 18 months (547 days of age), the male animals were slaughtered, and a detailed dissection of the carcass was performed. A wide range of phenotypic data related to beef production and beef quality including FA composition were recorded including FA composition of selected skeletal muscles.
Analysis of FA composition of lipids involving 26 different FAs in skeletal muscle (M. longissimus dorsi) was determined for 156 F 2 bulls using capillary gas chromatography as described previously [56]. The absolute amount of FAs in skeletal muscle was determined from 2 g of skeletal muscle and calculated as mg/100 g tissue. The relative content of individual fatty acids was calculated as percentage of the total amount of FAs extracted. Based on the data obtained for individual fatty acids, sums of specific fatty acid fractions were calculated: saturated fatty acids (SFA), unsaturated fatty acids (UFA), monounsaturated fatty acids (MUFA), trans fatty acids (TFA), n-3 fatty acids (n-3 FA), n-6 fatty acids (n-6 FA), polyunsaturated fatty acids (PUFA) and n-3 longchain PUFA (n-3 LC-PUFA). Furthermore, the ratios n-6/n-3 FA, MUFA/SFA, PUFA/SFA, P/S and LA/ALA as well as four different Δ 9 desaturase indices [57] were calculated. Intramuscular fat (IMF) content (percentage in 100 g tissue) was ascertained in M. longissimus dorsi by FoodScan Lab (FOSS) as described previously [56]. The phenotypic traits for FA composition of IMF included in our study are summarized in Table 3.

QTL analysis
An initial QTL scan comprising 244 microsatellite markers [58] for variation of FA composition in skeletal muscle had pinpointed a region on bovine chromosome 27 (BTA27) with effect on n-3 PUFA content in skeletal muscle. Five microsatellite markers located on BTA27 (BM3507, RM209, BMS689, BM1857, BM203) had been genotyped in all 733 P 0 , F 1 , and F 2 individuals of the Charolais × German Holstein resource population.
The respective QTL interval pointed to a chromosomal region on BTA27 harboring the acyl-CoA synthetase long-chain family member 1 (ACSL1) gene according to the sequence assembly of the chromosome. Therefore, in a second step of our analysis, nine intragenic ACSL1 SNPs ( Figure 1) were added to the initial marker set. All microsatellite markers and all genotyped ACSL1 SNPs were included to calculate a genetic map using CRIMAP Version 2.50 [59], incorporating modifications by Ian Evans and Jill Maddox (University of Melbourne). The resulting genetic map was applied in the QTL analyses with a variance component QTL model as implemented in Qxpak [60] and essentially as described previously [43]: where y is a vector of phenotypes, b is a vector of the fixed effects (slaughter year, NCAPG I442M genotype), u is the vector of individual infinitesimal polygenic effects, g is a vector of the additive QTL effects not fixed within founder breeds; F, Z and Q represent the incidence matrices for the fixed, polygenic and the QTL effect, respectively, and e is the vector of random residuals. An MCMC algorithm was used to calculate identity-by-descent probabilities as implemented in Qxpak. The NCAPG I442M mutation was included in the model, because previous analyses had shown a major effect of the mutation on carcass lipid deposition and growth in the resource population [44].
Statistical significance of the QTL analyses was tested by a likelihood-ratio test (LRT). Significance thresholds for the LRT were determined according to [61], considering one chromosome with a length of 0.6 M and an average marker density of 0.1 M. The significance thresholds for false positive results with α = 0.05 and α = 0.01 correspond to LRT values > 7.2 and LRT > 10.2, respectively.

Structural analysis of the ACSL1 gene
The coding sequence of the bovine ACSL1 gene is represented by the reference mRNA sequence NM_001076085.1, which spans 3690 bp and is located on BTA27.

Screening for polymorphisms in the ACSL1 gene
Screening for polymorphisms was carried out by resequencing and covered the complete coding sequence, exon-flanking intronic regions, the 5' and 3' UTRs and 724 bp of the promoter of the ACSL1 gene. DNA primer pairs for PCR amplification and sequencing were designed based on genomic contig sequences (NW_001494406.2 and NW_930554.1) and the mRNA sequence (NM_001076085.1), respectively (Table 4).
Four genomic DNA pools consisting of selected animals from the Charolais × Holstein resource population differing in their intramuscular fat content and index of delta 9-desaturase were established and subjected to screening for gene variants by comparative re-sequencing. The IMF pools contained DNA from sampling time-and pedigree-matched animals with high (n = 5, 4.93 ± 1.73%) and low (n = 7, 1.78 ± 0.21%) IMF. The Δ 9 desaturase index pools consisted of DNA from sampling time-and pedigree-matched animals with a high (n = 7; 50.87 ± 0.89) or low (n = 6, 43.86 ± 0.95) Δ 9 desaturase index. Furthermore, two genomic DNA samples from control individuals and two individual DNA samples originating from extreme animals displaying the lowest (1.63%) and highest (6.09%) IMF were included to validate the results received from the pools.
Genomic DNA was isolated from blood leucocytes using standard methods. PCR with exon-flanking primers (Table 4) was performed with a total of 60 ng genomic DNA as described above. The generated PCR products were purified using the peqGOLD Cycle-Pure Kit (PEQLAB) according to the manufacturer's instructions and sequenced. Sequencing was performed on a capillary sequencer (MEGABACE, GE Healthcare) with primers used for targeted PCR amplification. To identify variable DNA positions, the sequences were analyzed meticulously by visual inspection of the sequencing profiles from DNA-pools and individuals' DNA and by sequence alignment to the reference cDNA sequence (NM_001076085.1) as well as to the respective bovine genome sequences. All SNPs identified by sequencing of DNA pools were verified by single sample resequencing.

Association analysis
Prior to association analysis, we tested whether the phenotypic data of the individual traits were normally distributed using the Shapiro Wilk test. For those data displaying distributions significantly different from normality (P < 0.01), we performed natural log (ln) transformation, and the log-transformed data were subjected to association analysis. The respective data are indicated in Tables 2 and 3. The BTA27 marker haplotypes of the individuals of the resource population were estimated by a Markov chain Monte Carlo (MCMC) algorithm implemented in Qxpak [60]. The corresponding haplotypes were submitted to pairwise LD analysis calculating r 2 values using PowerMarker V3.25 [66].
Subsequently to the QTL analyses, association analyses were performed between ACSL1 gene SNPs and the absolute and relative FA composition traits in M. longissimus dorsi. The following model testing for LD as implemented in Qxpak [60] was applied: where y i is the record of individual i, a p is the fixed effect of slaughter year p, λ ihk is an indicator variable for the NCAPG I442M locus, which is 1 when the allele at the h th haplotype (1 or 2) of the i th individual is 1 and otherwise 0, λ imn is a respective indicator variable for the specific ACSL1 SNP, u i is the infinitesimal genetic effect of individual i, g k and g m are the respective allelic effects for NCAPG I442M and the ACSL1 SNP, and e ihkmnp is the residual. Analogous to the QTL analyses, the NCAPG I442M mutation was included in the model, because previous analyses had shown a major effect of the mutation on carcass lipid deposition and growth in the resource population [44]. A likelihood-ratio test (likelihood of model with both loci vs. likelihood of model with NCAPG I442M) was applied to test for statistical significance. In order to dissect whether the association of the respective ACSL1 variant with intramuscular FA composition is solely due to indirect effects on IMF or a consequence of direct effects on the specific FA accumulation, we extended the model and fitted IMF as an additional covariate. A Bonferroni correction was calculated (q-value) to account for testing several SNPs in order to avoid false positive associations. The q-values thresholds of 0.05 and 0.1, respectively, indicate an experiment-wise significant or suggestive association, respectively. Finally, an additive fixed effect of the SNP in intron 5 was added in the QTL model described above to test whether this SNP might explain the QTL variance at the identified position on BTA27.