Whole-genome sequencing GWAS reveals bovine genomic effects on enteric methane emissions in beef cattle

Abstract

The world has recognized the significance of sustainable animal production, especially in terms of mitigating methane emissions. Developing strategies to mitigate methane without compromising productivity presents a significant challenge for nutritionists and breeders. However, measuring methane emissions at the individual level can be expensive and laborious. Therefore, the use of genomic approaches combined with whole-genome information may be an alternative to overcome these challenges. This study aimed to use sequencing data to carry out GWAS to identify genomic regions and candidate genes involved in biological processes and metabolic pathways of enteric methane emission-related traits (ME: daily methane emission, RME: residual methane emission, MY: methane yield, MI: methane intensity, and MM: methane metabolic). For this, 1042 Nellore animals with phenotypic information and 2744 imputed for sequence genotypes belonging to three breeding programs from Brazil were used. The SNP significance was estimated through frequentist statistics using the single-step GBLUP approach. For ME, a total of 27 significant SNPs were deemed significant (p < 3.55 × 10 ^− 6), harboring 89 positional candidate genes. For RME, 21 SNPs showed significant association, and 48 genes were mapped. Regarding MY, 20 SNPs were deemed significant and surrounded 76 candidate genes. For MI, 5 significant SNPs mapped 15 potential candidate genes, while in MM, 10 significant SNPs were located near 50 positional candidate genes. Various statistically significant SNPs and genomic regions on BTA 5, 6, 8, 10, 11, 13, 19, and 27 were shared between methane emission-related traits. Comparing QTL regions affecting methane-related traits showed common genomic regions with QTL previously related to feed efficiency, growth, and enteric methane emission. In general, the potential candidate genes (DUOX1, DUOX2, FRMD4A, NOS2, CHRNB3, CHRNA6, CALM2, EPCAM, MSH2, MSH6, KCNK12, MUC4, MUC20, LDHAL6B, SLC20A2, LIPC, EDNRA, ACOXL, MAP4K4, IL1R1, IL1R2, PLCB3, ESRRA, and BAD) are involved in several biological processes and signaling pathways related to gastrointestinal motility, salivary secretion, enteric nervous system, mucosal barrier integrity, epithelial transport, olfactory receptors, lipid metabolism, oxidative stress, cAMP, cGMP-PKG, MAPK cascade, among others. Our results highlight the complexity of methane emission as a polygenic phenotype, suggesting that bovine genetics can modulate methane emissions by controlling the ruminal ecosystem. These findings may serve as a basis for future research focused on developing selection strategies for more sustainable beef cattle production.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-32815-z.

Introduction

Sustainable animal production is a topic that has been gaining more and more notoriety due to the current global environmental and geopolitical scenario. The livestock sector contributes to climate change with greenhouse gas (GHG) emissions estimated at 6.2 gigatons of carbon dioxide equivalents (CO₂-eq) per year, corresponding to 12% of anthropogenic emissions, caused mainly by feed production, enteric fermentation, animal waste, and changes in land use¹. Cattle represent 84.3% of enteric fermentation emissions in livestock production². The contribution of livestock to the atmospheric GHG pool is mainly in the form of enteric methane³. Methane (CH₄) is a GHG that is highly harmful to the environment since one unit of this gas corresponds to 27 equivalent units of carbon dioxide (CO₂-eq). So, CH₄ is 27 times more harmful than CO₂ in the global warming potential⁴. Given this reality, GHG emissions are expected to increase as demand for livestock products increases due to projected population growth⁵.

Reducing CH₄ production from beef cattle is desirable both as a mitigation strategy to reduce global GHG emissions and to improve efficiency in the production chain³. However, individual-level CH₄ emission measurements are expensive and labor-intensive, making large-scale phenotyping unfeasible in most breeding programs. Faced with these difficulties, approaches based on molecular-level methods appear as viable alternatives for effective genetic evaluation⁶.

Advances in next-generation sequencing (NGS) technologies have revolutionized the field of animal genetics by enabling whole-genome sequencing (WGS) at increasingly affordable costs. Unlike SNP (single single nucletotide polymorfisms) arrays or low-density genotyping panels, WGS provides a comprehensive view of genetic variation, including rare and regulatory variants that are often missed by conventional approaches⁷. This high-resolution data significantly enhances the power to detect associations with complex traits and facilitates the identification of potential causal mutations⁸. By capturing nearly all variations across the genome, NGS-based studies provide a more comprehensive understanding of trait architecture and support the development of more accurate and effective genomic selection strategies.

Genome-wide association studies (GWAS) use molecular information to explore genetic variants across the genome. The GWAS aims to identify genomic regions associated with traits by identifying SNPs significantly linked to phenotypes, providing a better understanding of biological functions and genetic influence on phenotypic expression⁹. A few GWAS have been conducted to identify quantitative trait loci (QTL) and genes associated with CH₄ emission in cattle^10–17. However, most of these studies were carried out in Taurine breeds and/or using genotypic data, which limits the scope for broadly identifying variants and QTL across the entire genome associated with traits of interest. Recently, Souza et al.¹⁸ performed GWAS analyses for CH₄ emission traits, considering part of the Nellore database considered in the present study. However, the authors used only a subset of the available data, and the study was conducted exclusively based on genotypic data from SNP arrays. Thus, the present study aims to use whole-genome sequencing data to carry out GWAS to identify genomic regions and potential candidate genes acting in biological processes and metabolic pathways regulating enteric methane emission-related traits in Nellore cattle.

Materials and methods

Ethics approval

All animal procedures used in this study were approved by the Institutional Animal Care and Use Committee of the São Paulo State University (UNESP), School of Agricultural and Veterinary Science Ethical Committee (protocol number 007757/18), and by the Committee on the Ethics of Animal Experiments of the Institute of Animal Science (protocol number 278 -9). Additionally, all experimental procedures were performed in accordance with the Guidelines for Animal Welfare and Humane Slaughter (State of São Paulo, Brazil, Law number 11.977) and the Regulations for the Administration of Affairs Concerning Experimental Animals (Ministry of Science and Technology, Brazil). In addition, we confirmed the statement that the study was conducted following the ARRIVE guidelines.

Animals and phenotypic data

Individual methane emission measurements were obtained from 1042 Nellore animals (994 males and 48 females), born between 2010 and 2021, from 122 sires and 658 dams. The animals came from an experimental breeding program at the Institute of Animal Science (Beef Cattle Research Center-Sertãozinho, Brazil, n = 759) and two commercial breeding programs, Qualitas (n = 111) and Cia de Melhoramento (n = 172). The Institute of Animal Science maintains three Nellore herds under different selection strategies^19,20: NeC (Control), NeS (Selection), and NeT (Traditional). NeC selects animals with yearling body weight (YBW) near the group average. NeS and NeT select for maximum YBW, and since 2013, NeT also includes selection for lower residual feed intake (RFI). On the other hand, animals from the breeding program Qualitas and Cia de Melhoramento are selected based on a selection index that combines multiple traits, with economic relevance, correlations (favorable or antagonistic), and genetic parameters²¹.

Methane emissions were measured using the sulfur hexafluoride (SF₆) tracer gas technique during performance tests. Each animal was evaluated for different measurements of enteric CH₄ emissions, such as daily methane emission in grams per day (ME), residual methane emission (RME), methane yield (MY), methane intensity (MI), and methane metabolic (MM). Performance tests were conducted to evaluate individual feed intake and weight gain after weaning, and the average test duration (± standard deviation) was 81.9 ± 3.1 days for the Institute of Animal Science, 56 ± 0 days for Qualitas, and 55.9 ± 1.9 days for Cia de Melhoramento. The animals started the test at an average age (± standard deviation) of 283.4 ± 40.8 (Institute of Animal Science), 658.4 ± 36.9 (Qualitas), and 547.6 ± 31.9 (Cia de Melhoramento) days and were kept in individual pens or collective paddocks equipped with electronic feed bunks (GrowSafe^®, Airdrie-AB, Canada; or Intergado^®, Contagem-MG, Brazil) for automated recording of individual daily feed intake, with access to diet and water ad libitum.

The animals’ diet consisted basically of silage (corn or sorghum), Brachiaria hay, sugarcane bagasse, flour (cotton, soybean, or peanut), corn (ground or wet grain), citrus pulp, mineral premix, salt, ammonium sulfate, and urea. The animals were fed a single diet in each test, which varied according to the selection criteria of the breeding program. In the Institute of Animal Science, selection is based on yearling weight, and diets were formulated for gains of 0.8 kg/day (2010 to 2012) and 1.2 kg/day (2018 to 2021). Animals from Qualitas and Cia de Melhoramento breeding programs are selected based on empirical selection indexes including growth, visual scores, and precocity indicator traits, and the diets were formulated for an average gain of 1.7 and 1.1 kg/day, respectively. The nutritional composition of the diets used in each test is provided in Supplementary Material Table S1.

Animals were weighed either after a 14-h fasting period at the beginning and end of the test or at predetermined intervals without prior fasting (ranging from 2 to 19 records per animal during the test years). The body weight (BW) records were subsequently used to obtain average daily gain (ADG) and the mid-test metabolic weight (BW^0.75). Feed efficiency-related measurements were obtained for posterior determination of RME, MY, MI, and MM. Dry matter intake (DMI) during the feeding trial was estimated as the average individual feed intake (from GrowSafe^® and Intergado^® information), adjusted for the diet’s dry matter content and expressed in kg/day. The average daily gain (ADG) was calculated as the BW linear regression coefficient as a function of days in test (DIT) following as: , where is the animal BW in the ith observation, is the intercept corresponding to the initial BW, is the linear regression coefficient corresponding to ADG, and is the random error. The mid-test metabolic weight (BW^0.75) was calculated as: where is the intercept corresponding to the initial BW.

The experimental design for methane collection and measurement is detailed in Sakamoto et al.²². In summary, to measure the enteric CH₄ emission, the technique described by Berndt et al.²³ of SF₆ was adopted. In this technique, the animal administered a permeation capsule that releases a constant, small, and known rate of SF₆ tracer gas. It is assumed that the SF₆ emission pattern simulates the CH₄ emission pattern for the subsequent calculation of methane quantification. CH₄ gas was collected for five consecutive days during the performance test, with one sample taken per animal per day, totaling five measurements per animal. Sampling was performed using polyvinyl chloride canisters or stainless-steel cylinders, which were changed every 24 h. The expelled gas was aspirated under vacuum with a tube fixed in a halter close to the animal’s mouth and nostrils, and connected to the collector cylinder, which was attached to a saddle on the animal’s back. In the case of polyvinyl chloride canisters, they were mounted in the neck region of the animal. After each CH₄ sampling period, the cylinders/canisters were sent for analysis by gas chromatography and their content was diluted with pure nitrogen to determine the amounts of SF₆ and CH₄ gases. The daily CH₄ emission (ME in g/day) of each animal was obtained from the arithmetic mean of the emissions sampled on the five consecutive collection days. RME was calculated within each test group (farm and year) as the difference between the observed and expected methane amount using a linear regression of CH₄ on DMI as: , where is the animal’s CH₄ in the ith observation, is the intercept of the regression, is the linear regression coefficient of CH₄ on DMI of animal i, and is the random error in the ith observation representing RME. MY refers to the amount of CH₄ emitted per unit of dry matter intake and was calculated as . MI expresses the methane emitted amount per unit of product produced and was obtained as . MM adjusts CH₄ production to the animal’s metabolic weight, reflecting the relationship between methane production and the animal’s energy needs, and it was calculated as .

The contemporary groups (CG) were defined by the test group class, which was based on sex, year of birth, breeding program, and CH₄ sampling date. The test group encompassed animals from the same environment and consumed the same diet during the performance test. The data quality control removed observations with measurements of 3.5 standard deviations above or below the CG mean and CG with less than 5 animals. The number of animals and descriptive statistics for each trait after data quality control are shown in Table 1.

Table 1.

Descriptive statistics of phenotypic information for the different enteric methane emission measures in Nellore cattle.

Trait*	n	Mean (± sd)	Min	Max	CG	CV (%)
ME (g/day)	1042	176.19 ± 63.42	54.67	444.44	26	35.99
RME (g/day)	1042	-0.09 ± 27.36	-126.07	160.64	26	95.38
MY (g/kg)	1042	20.87 ± 6.15	6.76	42.53	26	29.49
MI (g/kg)	1039	157.12 ± 52.30	55.95	438.50	26	33.29
MM (g/kg)	1042	0.49 ± 0.13	0.18	1.04	26	26.77

n, number of observations; sd, standard deviation; Min and Max, minimum and maximum values; CG, number of contemporary groups; CV, coefficient of variation.

*ME, daily methane emission; RME, residual methane emission; MY, methane yield; MI, methane intensity; MM, methane metabolic.

Sequencing data

A total of 171 bulls representing the Brazilian Nellore population were sequenced using the Illumina HiSeq X™ Ten or Illumina NovaSeq 6000 platforms. The criteria used to define the most representative bulls of the breed were based on a k-means cluster analysis using the genomic relationship matrix in a sample of 100,000 genotyped Nellore animals. The sire with the largest number of genotyped progenies within each cluster was chosen for sequencing. Genome coverage by sample ranged from 7.81× to 23.96×. The quality control, alignment, and variant calling processes were performed according to guidelines suggested by the 1000 bull Genomes Project²⁴. The detailed processes and parameters used for genome sequencing can be found in Fernandes Júnior et al.²⁵.

For GWAS analyses, a total of 1994 animals were genotyped with medium density (50 and 75k, GeneSeek Genomic Profiler) and 780 were genotyped with a high-density panel (770k, Illumina Inc., San Diego, CA). Prior to imputation, markers situated in non-autosomal regions and SNPs with a GenCall score below 0.60 were removed to ensure data quality. Animals genotyped with medium-density panels were imputed to high-density genotypes using a reference population of 6862 animals genotyped with the HD SNP chip²⁶. Subsequently, 2774 animals were imputed to the autosomal whole-genome sequence level using a reference panel of 171 sequenced Nellore bulls. The imputation was performed using FImpute v3²⁷ considering the ARS-UCD1.3 bovine as a reference map²⁸, and its expected accuracy was higher than 0.94²⁵. A total of 29,706,777 autosomal sequence SNPs in autosomal regions remained after imputation. Due to computational limitations, variants in high linkage disequilibrium (LD > 0.95) were pruned using PLINK 2.0 software²⁹, thereby eliminating redundant markers and improving computational efficiency.

Quality control (QC) of SNP markers and samples was conducted using the QCF90 software³⁰. SNPs located at the same genomic position, with MAF (Minor allele frequency) < 0.05, Call rate < 0.90, and monomorphic SNP were removed. Hardy–Weinberg equilibrium (HWE) was assessed based on the maximum deviation of observed from expected heterozygosity, and SNPs with a deviation greater than 0.15 were excluded. In addition, samples with Call rate < 0.90 were also removed from the analyses. After QC filtering, 2,591,217 SNPs and 2744 animals remained for further analysis.

Genome-wide association statistical analysis

The animal model applied in GWAS analyses included the fixed effect of CG; the linear and quadratic effects of animals’ age (in days) at the beginning of the performance test as covariates; and as the random effect the additive and residual effects. The matrix representation of the model is:

where, is an observation vector for each trait; is the vector of fixed effects; is the vector of genetic additive effects, assumed as , where H is the matrix that combines the pedigree (A) and genomic (G) relationship matrices, and is the additive genetic variance; is the vector of residual effects, assumed as , where I is an identity matrix, and is the residual variance; X and W are incidence matrices related to and , respectively. The pedigree file used included 14,219 animals across 18 generations, including both genotyped and non-genotyped individuals. A total of 748 founders were identified, with pedigree depths ranging from 1 to 17 generations. For the construction of the H matrix, only the last three generations were considered, comprising 4778 animals, which were the offspring of 515 sires and 2235 dams.

The genomic estimated breeding values (GEBV) for all traits were computed using single-step GBLUP (ssGBLUP) and estimated through the BLUPF90+ software³¹, considering the imputed whole-genome sequences. In the ssGBLUP procedure, the inverse of the numerator of the kinship matrix is replaced by the inverse of the combined matrix of the genomic-pedigree relationship³²:

where H^{− 1} is the inverse combined pedigree-genomic relationship matrix, A^{− 1} is the inverse of the relationship matrix based on the pedigree of all animals, is the inverse matrix of the relationship coefficients for the genotyped animals, and G^{− 1} is the inverse of the genomic relationship matrix³³, which is described as:

where, Z = (M – P), in which M is the SNP incidence matrix, with m columns representing the number of markers and n lines representing the number of sequenced animals. Each element in M was set to 0, 1, or 2, for genotypes AA, AB, and BB, respectively. P is the matrix containing the allele frequencies expressed in 2p_i, where p_i is the MAF for the ith SNP.

The effects of SNP and p-values were obtained based on the animals’ GEBV, using the POSTGSF90 software³⁴. The equation for computing the effect of SNP can be described as Wang et al.³⁵:

where, u is the effect vector of each SNP; Z is the matrix of SNP marker content adjusted for allele frequency and Z’ is its transposed; a is the vector with the predicted genetic values for genotyped animals, which is represented by a function of the SNP effects (a = Zu); λ represents the ratio of the SNP marker effect variance and the breeding value variance; D is a diagonal matrix of weights for SNP markers. Assuming an infinitesimal model, unweighted ssGBLUP was performed with the assumption that each SNP had an equal allele substitution effect variance. So, in this case, D was equal to an identity matrix.

The p-values for the SNP effects were computed according to Aguilar et al.³⁶. Estimated SNP effects were standardized and p-values that test whether the allele substitution effect differs from 0 were obtained as:

where, Φ is the cumulative density function of the normal distribution, is the estimated effect for the jth SNP, and is the standard deviation of .

Multiple testing correction and significance levels

False discovery rate (FDR) was calculated using the Benjamini–Hochberg method. However, no SNP reached the significance threshold of q < 0.05. This reflects the conservative nature of FDR correction in WGS studies with many markers and a limited sample size. To balance false positive control and statistical power, we applied a Bonferroni correction based on the number of independent chromosomal segments (Me), instead of the total number of markers. This method accounts for genomic structure and has been used in previous WGS-GWAS studies³⁷ providing a more appropriate threshold to detect true associations. Me was calculated as a function of the effective population size (Ne) and chromosome length (L, in centi-Morgans-cM) as proposed by Goddard et al.³⁸:

Considering the absence of accurate cattle-specific conversion values and based on reports in other mammalian species, we assumed 1 cM ~ 1 Mb, as previously adopted in the literature³⁹. The effective population size (Ne) used was 100, the most conservative value reported for Nellore cattle⁴⁰. Thus, SNPs were considered statistically significant if their -log₁₀(p-value) was equal to or higher than the threshold of 5.45 (p < 3.55 × 10^{− 6}).

Functional enrichment

Manhattan plots containing SNPs distributed according to -log₁₀(p-value) were constructed using the qqman package v0.1.9⁴¹ for R. Candidate genes were annotated considering an upstream and downstream interval of 0.25 Mb (0.50 Mb windows) from the significant SNP using the Ensembl annotation release 113⁴² considering the ARS-UCD1.3 bovine assembly genome. The GALLO package v1.5⁴³ in R was used to identify annotated QTL near each significant SNP considering the cattle database from Animal QTLdb⁴⁴. The candidate genes list obtained from each trait GWAS results were subjected to functional enrichment analysis covering terms related to biological processes and Kegg (Kyoto Encyclopedia of Genes and Genomes) pathways^45,46. The DAVID Functional Annotation Tool 2021 update⁴⁷ was used to conduct the analyses based on Gene Ontology (GO terms) and Kegg, considering the Bos taurus as background.

Results and discussion

Shared genomic regions and candidate genes among methane emission-related traits

The common regions and genes found between the studied traits are listed in Table 2. In total, 47 genes located on Bos taurus autosomes (BTA) 5, 6, 8, 10, 11, 13, 19, and 27 were found to be shared between CH₄ traits. The GWAS results indicated loci with pleiotropic effects influencing multiple aspects related to methane emissions. This genetic interconnection facilitates the design of breeding strategies, allowing us to take advantage of the loci pleiotropic effects to elaborate ways to reduce GHG emissions through selection. The functional analysis results for common genes between the methane-related traits are detailed in the Supplementary Material (Table S2).

Table 2.

Common regions and candidate genes for enteric methane emission-related traits in Nellore cattle.

Traits	Chr	Region (Mb)	SNPs	Genes
ME/RME	5	33.35–33.92	rs382240843 rs520653350	SLC38A4, SLC38A2
ME/RME	5	50.55–51.05	rs435268290	PPM1H, MIRLET7I, MON2
RME/MY	6	103.18–103.69	rs523102089 rs517228679	CRMP1, EVC, EVC2
RME/MY	8	16.07–16.57	rs452831134	LINGO2
RME/MY	10	62.21–65.89	rs454823945 rs381764436 rs377970965 rs477459494	SHF, DUOX1, DUOXA1, DUOXA2, DUOX2, SORD, FERMT2, DDHD1
ME/RME/MY	11	19.67–30.17	rs211452369	CALM2, EPCAM, MSH2, KCNK12, MSH6, FBXO11
RME/MY/MM	13	28.50–29.00	rs444287144 rs451465441 rs467164799	FRMD4A
ME/MM	19	19.15–19.62	rs109673061 rs110927664	KSR1, LGALS9, NOS2, LYRM9
RME/MY	27	15.56–16.10	rs133349420 rs441563320	SNX25, LRP2BP, ANKRD37, UFSP2, CFAP96, CCDC110, PDLIM3, SORBS2, TLR3
RME/MY	27	37.24–37.75	rs464491058	VDAC3, SLC20A2, SMIM19, CHRNB3, CHRNA6, THAP1, RNF170, HOOK3, FNTA, POMK

ME, daily methane emission; RME, residual methane emission; MY, methane yield; MM, methane metabolic.

The dual oxidase genes (DUOX1, DUOX2, DUOXA1, and DUOXA2), potential candidates on BTA 10 for MY and RME, were enriched for response to oxidative stress, hydrogen peroxide metabolism process, superoxide anion generation, thyroid hormone generation, and others biological processes. In addition, these genes participate in the thyroid hormone synthesis pathway. DUOX genes catalyze reactive oxygen species (ROS) production and are implicated in innate immune defense and antimicrobial function⁴⁸. ROS, including hydrogen peroxide (H₂O₂) and superoxide anion (O²⁻), cause chemical damage to the cells’ organic components, promoting oxidative stress⁴⁹. This can lead to disruptions in normal metabolism and physiology⁵⁰, negatively impacting animal performance, such as reduced growth development⁵¹. Furthermore, H₂O₂ tends to accumulate in the rumen, which can be toxic to ruminal microbiota and epithelial cells, compromising efficient fermentation and fatty acid absorption⁵². Therefore, the response to oxidative stress is essential to promote redox regulation ensuring adequate animal physiology and performance. In the thyroid, DUOX1 and DUOX2 play a role in H₂O₂ production, which acts as a co-substrate in the thyroid hormones’ biosynthesis⁵³. In sheep, it was observed that increasing the plasma concentration of thyroid hormones reduces digesta retention time and leads to a reduction in enteric CH₄ production⁵⁴. In a transcriptome analysis, Sun et al.⁵⁵ found the DUOX1, DUOX2, and DUOXA2 genes highly expressed in rumen and backfat tissues in beef cattle. Xiang et al.⁵⁶ found DUOX2 and DUOXA2 preferentially expressed in the rumen of ovine, while DUOX1 had rumen-biased expression and DUOXA1 was highly expressed in epithelial tissues. These findings suggests that DUOX genes may play an important role in the epithelium and controlling rumen microbial colonization.

The FRMD4A gene (BTA 13) was identified within a significant region for RME, MY, and MM. This gene was annotated to the regulation of protein secretion and epithelial cell polarity establishment GO biological process terms. FRMD4A encodes a protein containing a PERM domain that regulates epithelial polarity. It binds to a PAR protein complex to ensure accurate ADP ribosylation factor 6 (ARF6) activation, playing a central role in actin cytoskeleton dynamics and membrane traffic in epithelial cells⁵⁷. In this sense, it might play a crucial role in volatile fatty acid (VFA) absorption in rumen. During microbial fermentation, ingested carbohydrates are converted into VFAs, such as acetate, propionate, and butyrate⁵⁸, and the absorption efficiency of these VFAs directly influences the ruminal pH and H₂ concentration available for methanogens⁵⁹. Exploring genes and genomic regions under selection, Ma et al.⁶⁰ reported that FRMD4A gene may play an important role in growth and development of goats.

The NOS2 gene (nitric oxide synthase 2) on BTA 19, was associated with ME and MM. This gene encodes inducible nitric oxide synthase (iNOS), which is responsible for producing nitric oxide (NO) in organisms⁶¹. NO is a crucial molecule in the defense against infections, having antimicrobial properties⁶² that may influence the microbial population in the rumen, particularly methanogens. In addition, functional analysis revealed that NOS2 is involved in the relaxin signaling pathway. Relaxin hormone affects gastrointestinal motility by acting on enteric nerves and smooth muscle, primarily through a mechanism involving NO⁶³. Increased gastrointestinal peristalsis leads to shorter rumen retention time, impacting the digestion process and CH₄ emissions in ruminants⁶⁴. It has been suggested that an increase in ruminal passage rate may decrease the abundance of ruminal methanogens, contributing to reduced CH₄ emissions⁶⁵. In an immunohistochemical study, Castro et al.⁶⁶ observed traces of iNOS expression in sheep’s rumen epithelium, suggesting that NO may help preserve the integrity of the rumen mucosa barrier, protecting it against the passage of antigens and bacteria that can damage the mucosa and compromise its function. Using RNA-seq technology, Tan et al.⁶⁷ found the NOS2 gene upregulated in tissues from pigs pooled for high feed conversion rates, suggesting that this gene may positively influence feed efficiency.

Cholinergic receptor nicotinic subunits (CHRNB3 and CHRNA6–BTA 27) were associated with both MY and RME traits. The functional annotation analysis revealed that these genes are involved in the acetylcholine receptor signaling pathway. Gastrointestinal tract (GIT) motility is driven by the enteric nervous system (ENS), which is one of the three branches of the autonomic nervous system, composed of highly complex neural networks⁶⁸. Myenteric neurons control motility events by releasing specific combinations of neurotransmitters⁶⁹. Acetylcholine (ACh) is the main endogenous neurotransmitter released from ENS cholinergic neurons⁷⁰. ACh interacts with muscarinic acetylcholine receptors (mAChRs) triggering smooth muscle excitation and contraction, generating GIT peristaltic reflex⁷¹. In ruminants, have been reported that cholinergic myenteric neurons directly innervate the muscle layers of the rumen and abomasum, leading to contractions in response to ACh⁷². In addition, salivary secretion is also under the control of autonomic nervous activity. ACh released by the cholinergic system binds to the mAChRs of acinar cells in the salivary glands, stimulating salivary fluid secretion⁷³. In this context, the cholinergic system and the acetylcholine receptor signaling pathway may be intrinsically linked to enteric methane production through the regulation of intestinal motility and saliva flow to the rumen, thus influencing the digesta passage in GIT and contributing to ruminal environment homeostasis.

A significant SNP (rs211452369) was identified in association with ME, RME, and MY at a locus on chromosome 11 (Table 2). This locus (19.67–30.17 Mb) was previously identified for residual feed intake⁷⁴ and grazing behavior⁷⁵ in cattle. This suggests that this region may be a potential QTL for feed efficiency-related traits as well as enteric CH₄ emission since these traits are genetically correlated^76,77. Within this significant region, there are potential candidate genes (CALM2, EPCAM, MSH2, MSH6, and KCNK12) that may play an important role in CH₄ production. CALM2 (calmodulin 2) is involved in multiple signaling pathways related to the gastrointestinal system, including vascular smooth muscle contraction, insulin, glucagon, salivary secretion, and gastric acid secretion, among others. These suggest that this gene may have a significant function in the digestive system of ruminants and could potentially impact enteric CH₄ production. Saliva secretion in ruminants maintains the homeostasis of the rumen ecosystem by supplying buffers, such as bicarbonate and phosphate, which neutralize the acidic pH caused by events during fermentation⁷⁸. In this sense, saliva plays an important role in maintaining the pH of rumen fluid and cellulolytic microbial activity⁷⁹. Furthermore, lysozyme present in saliva can influence the selection of methanogenic microorganisms and affect the rumen environment, thus modulating methane emissions⁸⁰. Salivary secretion also acts in the control of rumen fluid hypertonicity by diluting its content⁸¹, leading to a faster passage rate through the rumen, which in turn can reduce CH₄ production by reducing digesta retention time⁸². Recently, Zhang et al.⁸³ observed that applying a saliva stimulant to cattle led to a significant reduction in CH₄ emissions concomitantly with a decrease in fluid retention time. EPCAM (epithelial cell adhesion molecule) is a cell adhesion molecule involved in cellular communication and tight junctions, which plays an important role in epithelium integrity by maintaining proper intestinal function⁸⁴. Tight junctions of epithelial cells regulate permeability and prevent translocation of toxins across the epithelial barrier⁸⁵. It was observed that EPCAM is highly expressed in bovine rumen epithelial cells⁸⁶ and has been associated with feed efficiency in chickens⁸⁷. MSH2 and MSH6 encode MutS homologue proteins that interact with each other to form the MutSα heterodimer, a complex involved in mismatch repair and mutagen-induced lesions in DNA⁸⁸. MSH6 has been associated with pathways related to digestion, antimicrobial response, and G protein-coupled receptor signaling⁸⁹, and its expression appears to occur in the duodenum, to help animals deal with toxic and mutagenic substances present in ingested feed⁹⁰. Increased MSH2 expression levels were observed in steers with a phenotype of low weight gain combined with high feed intake⁹¹. KCNK12 (potassium two pore domain channel subfamily K member 12) is a two-pore potassium-selective channel family member. These channels regulate smooth muscle tone and may contribute to the maintenance of relaxation in visceral organs⁹². Epithelial potassium (K⁺) channels control the passage of nitrogen (N) metabolism products across the ruminal wall. Urea, the main N source in ruminant diets, is hydrolyzed into ammonia (NH₃) for amino acid biosynthesis in the rumen⁹³. However, excess NH₃ can increase the risk of ammonia toxicity⁹⁴. Surplus NH₃ is absorbed by the rumen epithelium via K⁺ channels or passive diffusion and transported to the liver⁹⁵. In the liver, NH₃ is metabolized into urea, which can be excreted in urine or recycled into the rumen via saliva or epithelial influx⁹⁶. Once in the rumen, it can be hydrolyzed into microbial protein and used as a protein source for muscle synthesis or milk composition⁹⁷. These suggest that KCNK12 is related to urea recycling and plays an important role in protein synthesis in the rumen, ensuring better feed efficiency and productivity in ruminants.

QTL annotation for enteric methane emission-related traits

QTL annotation analyses were performed based on information from CattleQTLdb for all significant SNPs detected by GWAS analyses. The proportion of different class types (exterior, health, meat and carcass, production, and reproduction) and the different annotated production class QTL that overlapped with significant SNP regions are presented in Fig. 1 as a pie chart and bar chart, respectively. A total of 673, 413, 322, 75, and 700 previously reported QTL overlapped with the SNP regions identified for ME, RME, MY, MI, and MM, respectively (Table S2). Overall, the most frequent QTL types were Production, Meat and Carcass, and Reproduction (Fig. 1—pie charts). The most frequent traits related to Production QTL for all CH₄ measures analyzed (ME, RME, MY, MI, and MM) predominantly include growth and feed efficiency-related traits (Fig. 1—bar charts). All results of the QTL annotation analyses for each trait are detailed in Supplementary Material (Table S3).

Fig. 1.

Percentage of total annotated QTL in different classes (pie chart) and percentage of annotated QTL associated with Production QTL class (bar chart) for: (A) daily methane emission, (B) residual methane emission, (C) methane yield, (D) methane intensity, and (E) methane metabolic.

Two significant SNPs identified for ME are within a chromosomal region in BTA 10 that overlaps with a QTL associated with methane production in cattle (Table S2). Pszczola et al.¹⁴ also found an SNP within this locus associated with methane production in dairy cattle. This provides strong evidence that this is a potential candidate region regulating CH₄ production and can be investigated further in future works to explore CH₄ mitigation strategies.

For all variables studied, genomic regions controlling CH₄ emissions overlapped with QTL previously associated with growth traits such as body weight, body metabolic weight, body size, average daily gain, among others (Fig. 1 and Table S2). The proportional connection between intestinal volume and body weight can ground the relationship between body development and CH₄ emissions from enteric fermentation¹⁸. As animals grow and their body weight increases, there is a corresponding development in intestinal volume capacity, impacting the volume available for fermentation in the digestive system, which may affect CH₄ emissions⁹⁸. Furthermore, a higher body weight corresponds to a higher feed intake and a greater amount of substrate available for fermentation, leading to an increase in CH₄ production⁹⁹. Maciel et al.¹⁰⁰ observed a higher CH₄ production (g/day) in animals with better weight gain performance compared to animals with lower performance. However, despite higher daily methane emissions, these high-performing animals spent less time in confinement, resulting in total CH₄ emissions similar to those with lower weight gain¹⁰⁰. In this context, the use of higher-performance animals can indirectly lead to a reduction in overall GHG emissions.

Significant regions within previously reported QTL in association with feed efficiency and intake traits (e.g. residual feed intake, dry matter intake, and feed intake) were also found for all traits under study (Fig. 1 and Table S2). Methane production by enteric fermentation is affected by feed intake, diet composition, and amount of energy consumed¹⁰¹. The connection between feed intake and CH₄ production is linked to the amount of substrate ingested and available for enteric fermentation¹⁰², favoring the increase in H₂ concentrations in the rumen for methanogenesis. Previous studies have observed that dry matter intake and CH₄ production are strongly correlated, both genetically and phenotypically^13,103, emphasizing that methane emissions increase in response to increased feed intake. Furthermore, increased CH₄ production leads to losses of dietary energy¹⁰⁴, and therefore more efficient animals should produce less CH₄¹⁴. Studies have shown that selection of animals with lower RFI (i.e. those more efficient) led to a significant reduction in enteric CH₄ production^105,106.

Using a smaller subset (n = 743) of the population used in this study, Souza et al.¹⁸ also found several regions for ME and RME within QTL associated with residual feed intake and growth traits. Although there are nearby regions, none of those identified in this study correspond to those found by Souza et al.¹⁸. This inconsistency may be due to differences in the number of samples, animal variability in terms of age and phenotypic values, and mainly by the number of markers used between the studies. The increased sample size in our study (about 40%) may have resulted in greater variability in observed methane emissions in the evaluated population. This sample size allows us to capture a wider range of phenotypic differences than the previous study, which used a smaller data set.

A larger number of samples can significantly increase the statistical power¹⁰⁷, allowing the identification of more subtle associations between markers and phenotypes¹⁰⁸. Furthermore, using sequencing data in GWAS analyses allows the broad identification of SNPs across the entire genome, including those that explain a small fraction of trait variation¹⁰⁹. This is because the analysis covers both potential causal SNPs and SNPs with high linkage disequilibrium (LD) with causal variants, exploring all genetic markers while minimizing the influence of LD between SNPs and the underlying genes¹¹⁰. Comparing GWAS analyses with whole-genome sequence and SNP array data, Heidaritabar et al.¹¹¹ observed that the increase in sample size combined with a higher variant density, increased the detection power of QTL regions, supporting our hypothesis.

Enteric methane emission traits

The Manhattan plots containing SNPs associated with genetic variation in enteric CH₄ emission measurements are shown in Fig. 2. The GWAS results identified loci associated with daily emissions (ME) on BTA 1, 2, 3, 5, 7, 9, 10, 11, 12, 13, 15, 17, 19, 25, and 27 (Fig. 2A). These regions harbored 27 significant SNPs and 89 positional candidate genes associated with ME (Table 3). For RME, we identified 21 significant SNP markers located on BTA 5, 6, 8, 10, 11, 12, 13, 15, 17, 26, and 27 (Fig. 2B), mapping 48 genes (Table 4). A total of 20 SNPs were significantly associated with MY and covered regions on BTA 6, 8, 10, 11, 12, 13, 15, 17, 19, 23, and 27 (Fig. 2C), which surround 76 candidate genes (Table 5). For MI, we identified 5 significant SNP markers on BTA 1, 11, 12, and 26 (Fig. 2D), mapping 15 potential candidate genes (Table 6). For MM, 10 SNPs were found surrounding 50 positional candidate genes on BTA 7, 8, 11, 13, 15, 19, and 29 (Fig. 2E; Table 7). Of the candidate genes found for CH₄ variables, only known and unambiguous genes were kept for GO Term functional analysis. The Supplementary Material (Table S4) contains detailed information on the biological processes GO terms and metabolic pathways (Kegg) associated with each trait’s putative candidate genes.

Fig. 2.

Manhattan plots of genome-wide association analyses showing the significant SNPs (highlighted in green above the red significance threshold line, − log₁₀ = 5.45) associated with CH₄ emission traits in Nellore cattle: (A) daily methane emission, (B) residual methane emission, (C) methane yield, (D) methane intensity, and (E) methane metabolic.

Table 3.

SNP identification, chromosome (Chr), position, p-value, region (± 0.25 Mb from the significant SNPs), and gene identification for daily methane emission (ME).

SNP*	Chr	Pos (Mb)	p-value	Region (Mb)	Genes
rs210943249	1	68.71	7.74e−07	68.46–68.96	KALRN
rs455586928	1	70.29	1.70e−06	70.03–70.53	OSBPL11, LMLN, IQCG, RPL35A, LRCH3, FYTTD1, RUBCN, MUC20, MUC4
rs209662448	2	49.13	1.67e−06	48.88–49.38	–
rs721081177	3	46.43	3.31e−06	46.18–46.68	DPYD, PTBP2
rs382240843	5	33.6	2.68e−06	33.35–33.85	SLC38A4, SLC38A2
rs435268290	5	50.8	2.58e−06	50.55–51.05	PPM1H, MIRLET7I, MON2
rs523777463	7	91.32	2.83e−06	91.07–91.57	–
rs81180483	9	93.89	3.54e−06	93.63–94.13	ARID1B, TMEM242, LDHAL6B, ZDHHC14
rs210164177	10	17.62	5.68e−07	17.37–17.87	UACA, LARP6, THAP10, LRRC49
rs519897568	10	17.62	2.27e−06	17.37–17.87	UACA, LARP6, THAP10, LRRC49
rs111020934	11	29.66	3.25e−06	29.40–29.90	TTC7A, STPG4, CALM2, EPCAM, MSH2
rs211452369	11	29.92	1.59e−06	29.67–30.17	CALM2, EPCAM, MSH2, KCNK12, MSH6, FBXO11
rs210375589	12	1.95	2.18e−06	1.70–2.20	TDRD3, LOC100140262, LOC789865, DIAPH3
rs480765082	13	29.43	1.25e−06	29.17–29.67	FAM107B, CDNF, HSPA14, SNORD22, SUV39H2, DCLRE1C, MEIG1, OLAH, NMT2, RPP38
rs446405528	13	29.43	1.28e−06	29.18–29.68	FAM107B, CDNF, HSPA14, SNORD22, SUV39H2, DCLRE1C, MEIG1, OLAH, NMT2, RPP38
n/a	13	32.34	6.14e−07	32.09–32.59	STAM, SNORD62, TMEM236, MRC1, SLC39A12, CACNB2
rs209669442	15	50.82	1.47e−06	50.57–51.07	OR52P2, OR51R1, TRIM21, OR52B3, OR52B4K, OR52B4L, OR55B1, RRM1, STIM1
rs1116785308	17	46.26	6.45e−07	46.01–46.51	ADGRD1, RAN, STX2, RIMBP2
rs482932914	17	46.28	3.96e−07	46.03–46.53	ADGRD1, RAN, STX2, RIMBP2
rs714901264	17	46.3	1.38e−06	46.05–46.55	ADGRD1, RAN, STX2, RIMBP2
rs109673061	19	19.4	1.56e−06	19.15–19.65	KSR1, LGALS9, NOS2, LYRM9
rs136596602	25	23.71	9.39e−07	23.46–23.96	HS3ST4
rs380986960	25	27.79	2.43e−06	27.54–28.04	AHSP, OR7A53, OR7D4, SEPT14, ZNF713, MRPS17, NIPSNAP2, PSPH, CCT6A, SUMF2, PHKG1, CHCHD2, NUPR2, VKORC1L1, GUSB, ASL, CRCP
rs723162743	25	31.82	3.37e−06	31.57–32.07	–
rs522126892	25	31.83	2.43e−06	31.58–32.08	–
rs714968476	25	31.88	3.18e−06	31.63–32.13	–
rs452064505	27	31.94	3.38e−06	31.69–32.19	KCNU1

*Significant SNP (p < 3.55 × 10^− 6); n/a, not available.

Table 4.

SNP identification, chromosome (Chr), position (Pos), p-value, region (± 0.25 Mb from the significant SNPs), and gene identification for residual methane emission (RME).

SNP*	Chr	Pos (Mb)	p-value	Region (Mb)	Genes
rs520653350	5	33.67	3.46e−06	33.42–33.92	SLC38A4, SLC38A2, SLC38A1
rs435268290	5	50.8	5.84e−07	50.55–51.05	PPM1H, MIRLET7I, MON2
rs523102089	6	103.43	1.76e−06	103.18–103.68	CRMP1, EVC, EVC2
rs517228679	6	103.44	2.64e−06	103.19–103.69	CRMP1, EVC, EVC2
rs452831134	8	16.32	7.50e−07	16.07–16.57	LINGO2
rs723234149	8	19.13	1.71e−06	18.89–19.38	–
rs1115375721	10	51.81	1.93e−06	51.56–52.06	ADAM10, LIPC
rs454823945	10	65.46	2.41e−06	65.21–65.71	SHF, DUOX1, DUOXA1, DUOXA2, DUOX2, SORD, FERMT2, DDHD1
rs381764436	10	65.47	8.91e−07	65.22–65.72	SHF, DUOX1, DUOXA1, DUOXA2, DUOX2, SORD, FERMT2, DDHD1
rs377970965	10	65.47	3.14e−07	65.22–65.72	SHF, DUOX1, DUOXA1, DUOXA2, DUOX2, SORD, FERMT2, DDHD1
rs477459494	10	65.64	1.09e−06	65.39–65.89	FERMT2, DDHD1
rs211452369	11	29.92	3.57e−07	29.67–30.17	CALM2, EPCAM, MSH2, KCNK12, MSH6, FBXO11
rs437245732	12	3.37	7.28e−07	3.12–3.62	–
rs467164799	13	28.75	2.05e−06	28.49–28.99	FRMD4A
rs451465441	13	28.76	8.25e−07	28.51–29.01	FRMD4A
rs42656812	15	10.74	2.91e−06	10.49–10.99	–
rs380730330	17	49.69	3.10e−06	49.44–49.94	–
rs209625845	17	49.82	3.47e−06	49.57–50.07	–
n/a	26	47.69	3.33e−06	47.44–47.94	PTPRE, MKI67
rs133349420	27	15.81	1.85e−06	15.56–16.06	SNX25, LRP2BP, ANKRD37, UFSP2, CFAP96, CCDC110, PDLIM3, SORBS2, TLR3
rs464491058	27	37.5	4.11e−07	37.25–37.75	VDAC3, SLC20A2, SMIM19, CHRNB3, CHRNA6, THAP1, RNF170, HOOK3, FNTA, POMK

*Significant SNP (p < 3.55 × 10^− 6); n/a, not available.

Table 5.

SNP identification, chromosome (Chr), position (Pos), p-value, region (± 0.25 Mb from the significant SNPs), and gene identification for methane yield (MY).

SNP*	Chr	Pos (Mp)	p-value	Region (Mb)	Genes
rs523102089	6	103.43	3.35e−06	103.18–103.68	CRMP1, EVC, EVC2
rs452831134	8	16.32	4.87e−07	16.07–16.57	LINGO2
rs720620539	8	19.09	3.54e−06	18.84–19.34	–
rs723234149	8	19.13	9.08e−07	18.88–19.38	–
rs381764436	10	65.47	3.43e−06	65.22–65.72	SHF, DUOX1, DUOXA1, DUOXA2, DUOX2, SORD, FERMT2, DDHD1
rs377970965	10	65.47	1.86e−06	65.22–65.72	SHF, DUOX1, DUOXA1, DUOXA2, DUOX2, SORD, FERMT2, DDHD1
rs43661069	11	7.68	3.27e−06	7.43–7.93	MFSD9, TMEM182
rs133167336	11	9.92	2.20e−06	9.67–10.17	POLE4, HK2, SEMA4F, M1AP, DOK1, LOXL3, HTRA2, AUP1, DQX1, TLX2, PCGF1, LBX2, TTC31, CCDC142, MRPL53, MOGS, WBP1, RTKN, INO80B
rs211452369	11	29.92	7.29e−07	29.67–30.17	CALM2, EPCAM, MSH2, KCNK12, MSH6, FBXO11
rs524522263	12	45.1	1.59e−06	44.85–45.35	–
rs451465441	13	28.76	3.22e−06	28.50–29.00	FRMD4A
rs42656812	15	10.74	2.13e−06	10.49–10.99	–
rs381458564	17	10.69	1.61e−06	10.43–10.93	PRMT9, TMEM184C, EDNRA
rs526104998	17	49.84	3.18e−06	49.58–50.08	–
rs43710071	19	24.81	2.95e−06	24.56–25.06	ATP2A3, ZZEF1, CYB5D2, ANKFY1, UBE2G1, SPNS3, SPNS2, MYBBP1A, GGT6, TEKT1
rs526823272	23	13.74	1.38e−07	13.49–13.99	KIF6, DAAM2, MOCS1
rs476300210	23	14.41	3.88e−07	14.15–14.65	LRFN2
rs133349420	27	15.81	1.52e−06	15.56–16.06	SNX25, LRP2BP, ANKRD37, UFSP2, CFAP96, CCDC110, PDLIM3, SORBS2, TLR3
rs441563320	27	15.85	1.92e−06	15.60–16.10	SNX25, LRP2BP, ANKRD37, UFSP2, CFAP96, CCDC110, PDLIM3, SORBS2, TLR3
rs464491058	27	37.5	1.61e−06	37.24–37.74	VDAC3, SLC20A2, SMIM19, CHRNB3, CHRNA6, THAP1, RNF170, HOOK3, FNTA, POMK

*Significant SNP (p < 3.55 × 10^− 6).

Table 6.

SNP identification, chromosome (Chr), position (Pos), p-value, region (± 0.25 Mb from the significant SNPs), and gene identification for methane intensity (MI).

SNP*	Chr	Pos (Mb)	p-value	Region (Mb)	Genes
rs525094679	1	78.52	1.05e−06	78.27–78.77	TPRG1, LPP
rs714224875	11	1.65	3.03e−06	1.39–1.89	ACOXL, BUB1, TPC3, MTLN, NPHP1, MALL, MAL
rs210997643	11	6.59	3.39e−06	6.34–6.84	MAP4K4, IL1R2, IL1R1
rs443360439	12	57.3	1.86e−06	57.05–57.55	–
n/a	26	50.56	2.81e−06	50.31–50,81	TCERG1L, KNDC1, ADGRA1

*Significant SNP (p < 3.55 × 10^− 6); n/a, not available.

Table 7.

SNP identification, chromosome (Chr), position (Pos), p-value, region (± 0.25 Mb from the significant SNPs), and gene identification for methane metabolic (MM).

SNP*	Chr	Pos (Mb)	p-value	Region (Mb)	Genes
rs523777463	7	91.32	2.85e−07	91.06–91.56	–
rs516073638	7	91.33	3.30e−06	91.08–91.58	–
rs380285035	7	92.15	3.26e−06	91.89–92.39	–
rs208250950	8	17.41	3.08e−06	17.16–17.66	IFT74, LRRC19, PLAA, CAAP1, ATP6V0C
rs43677944	11	31.35	3.45e−06	31.10–31.60	FSHR
rs444287144	13	28.75	2.77e−06	28.50–29.00	FRMD4A
rs42656812	15	10.74	9.81e−07	10.49–10.99	–
rs110927664	19	19.37	2.41e−06	19.12–19.62	KSR1, LGALS9, NOS2, LYRM9
rs522915763	19	34.93	4.56e−07	34.68–35.18	RAI1, PEMT, RASD1, MED9, NT5M, COPS3, FLCN, PLD6, MPRIP, TNFRSF13B, USP22, DHRS7B
rs109873504	29	42.47	3.03e−06	42.22–42.72	MARK2, RCOR2, NAA40, COX8A, COX8L, OTUB1, MACROD1, FLRT1, STIP1, FERMT3, TRPT1, NUDT22, DNAJC4, VEGFB, FKBP2, PLCB3, BAD, GPR137, PPP1R14B, KCNK4, CATSPERZ, ESRRA, TRMT112, UPF0315, PRDX5, CCDC88B, RPS6KA4

^*Significant SNP (p < 3.55 × 10^− 6).

Daily methane emission

MUC4 and MUC20 found for ME on BTA 1 are members of the transmembrane mucin family (mucus glycoproteins) expressed in epithelial cells and crucial components of GIT mucosal barrier¹¹². Mucins form a protective layer that adheres to the mucosal surfaces of GIT, protecting the epithelium from mechanical damage caused by the passage of food and against pathogenic organisms¹¹³. In a previous study on cattle, the MUC20 gene was found to be transcribed in several tissues of the bovine GIT, including the rumen¹¹⁴. Additionally, MUC4 transcripts were detected in the esophagus, abomasum, and hindgut¹¹⁴. SNP markers located in or near mucin-encoding genes, including MUC4 and MUC20, have been identified as playing a role in shaping the GIT microbiota in pre-weaning calves, particularly before the rumen is fully developed¹¹⁵. Assessing the influence of the host genome on the abundance of functional microbial genes, Martínez-Álvaro et al.¹¹⁶ found a gene in the rumen that encodes the sensor for host-microbiome crosstalk mediator fucose. Fucose is a component of mucins produced by the GIT mucosa and in saliva¹¹⁴, and as a mediator of crosstalk between the host and the microbiome, it can lead to an increase in saliva production, influencing rumen pH and, consequently, CH₄ production¹¹⁶.

LDHAL6B (BTA 9) is a homolog of lactate dehydrogenase A (LDHA), an enzyme that catalyzes the reversible interconversion of pyruvate and lactate, a key metabolic step in anaerobic glycolysis and other metabolic pathways¹¹⁷. Evidence suggests increased LDHA expression results in higher lactate production in response to nutrient intake¹¹⁸. In this regard, a greater lactate availability can serve as a substrate for propiogenic microorganisms. Lactate is converted to propionate through microbial pathways, influenced by the host’s genome¹¹⁶. The production of propionate from lactate leads to a potential reduction in the availability of H₂ for methanogenesis and may prevent ruminal acidosis, resulting in improved metabolic efficiency¹¹⁶.

A significant SNP (rs209669442) on BTA 15 was identified within a genomic region that harbors genes encoding olfactory receptors (OR51R1, OR52B4K, OR52B4L, OR7A53, OR55B1, OR52P2, and OR52B3). Olfactory receptors contribute to the food hedonic evaluation, resulting in feed choice and possible consumption. The ruminants’ ability to select feed based on smell is crucial for maximizing essential nutrient intake¹¹⁹. This selection may influence fermentation and VFAs production in the rumen, since diet significantly affects the ruminal microbial community¹²⁰. Furthermore, olfactory receptors can bind to molecules such as nutrients and metabolites in organs outside the nasal cavity, triggering physiological responses such as changes in gut motility and nutrient absorption in GIT¹²¹. The amount of CH₄ produced is directly related to the diet of ruminants, especially the fiber and type of fermentable carbohydrates present¹²². Thus, the ability of animals to detect and choose foods based on odor may influence their preferences for certain types of feed that minimize CH₄ production. Previous studies have reported olfactory transduction genes associated with feeding behavior and feed efficiency in cattle^123–125. Additionally, Sarghale et al.¹⁵ found candidate genes significantly enriched in olfactory receptor activity for predicted CH₄ emission traits using Holstein cattle sequence data, elucidating the potential influence of olfactory genes and pathways on enteric methane emissions.

Residual methane emission

SLC20A2 is a member of the type III sodium-dependent phosphate cotransporters. The SLC20 gene family mediates inorganic phosphate (Pi) uptake into cells¹²⁶ and has been identified as the predominant phosphate transporter in rats¹²⁷ and human¹²⁸ vascular smooth muscle cells. Pi acts as a buffer for the rumen environment and serves as a nutrient for rumen microorganisms¹²⁹. Given prior information on SLC20A2 expression in the rumen¹³⁰ and parotid glands¹³¹, it is suggested that this gene participates in the endogenous Pi recycling, ensuring an adequate supply of this compound in the rumen for adequate microbial fermentation. Previously, SLC20A2 gene has been associated with carcass and meat quality traits in beef cattle^132,133.

LIPC and VDAC3 genes were enriched for cholesterol metabolism, a signaling pathway that is closely related to lipid metabolism. Lipid metabolism involves a series of processes that utilize lipids, including VFAs produced by enteric fermentation, for several essential bodily functions, such as maintenance, energy production, and synthesis of cellular components¹³⁴. The VFA mobilization from the rumen, to meet the physiological demands of the animal, affects the VFA stoichiometry in the ruminal environment¹³⁵. The change in the relative proportions of VFAs may impact ruminal fermentation dynamics, including H₂ production and, as a result, methanogenesis¹³⁶. Previous studies have shown that the milk fatty acid profile can be used as an indicator to predict individual methane emissions in ruminants^137,138. Moreover, Pszczola et al.¹⁴ found a gene involved in lipid metabolism in association with CH₄ production in dairy cattle. These findings establish a possible relationship between lipid metabolism and the regulation of methane emissions, providing support for a better understanding of the underlying mechanisms of enteric CH₄ production.

Methane yield

The EDNRA (endothelin receptor type A) gene found for MY on BTA 17 was enriched for the biological processes of blood pressure regulation and vasoconstriction, in addition to the vascular smooth muscle contraction signaling pathway. EDNRA is primarily expressed on vascular smooth muscle cells, where its activation leads to sustained vasoconstriction¹³⁹. Therefore, this gene may be involved in regulating blood flow in GIT capillaries, contributing to its proper functionality. Transcriptomic¹⁴⁰ and differential miRNA expression¹⁴¹ studies found the EDNRA gene differentially expressed in the liver tissue of beef steers with divergent phenotypes for feed efficiency, suggesting its involvement in the molecular control of feed efficiency.

The enrichment analysis revealed that EDNRA and TLX2 are involved in the enteric nervous system (ENS) development. The ENS is a complex and integrative system that controls motility, enteric secretions, local blood flow, and immune system regulation in GIT¹⁴². In addition to its local functions, the ENS mediates interactions between GIT microbiota and the nervous system through the microbiota-gut-brain axis. When ingested, food can be metabolized by microbiota into specialized metabolites, such as neurochemicals. These can be absorbed into the circulation or act locally in the ENS, where they transduce signals to the central nervous system, which in turn can cause changes in the enteric system which feedback on the microbiota¹⁴³.

In addition, a gene cluster (EDNRA, ATP2A3, and CALM2) was found to play a role in cyclic adenosine monophosphate (cAMP), cyclic guanosine monophosphate (cGMP-PKG), and calcium signaling pathways. cAMP and cGMP are types of second messengers involved in various biological functions, including relaxation of vascular smooth muscle and inhibition of smooth muscle contraction^144,145. Smooth muscle contraction is activated in response to an increase in cytoplasmic Ca²⁺ concentration, which allows interaction between contractile proteins¹⁴⁵. The role of these cyclic nucleotides in smooth muscle relaxation involves reducing Ca²⁺ levels via activation of cAMP- and cGMP-dependent protein kinases, or PKA and PKG, respectively¹⁴⁶. cGMP is also involved in vascular tone regulation by modulating cytosolic Ca²⁺ levels, myofilament Ca²⁺ sensitivity, and smooth muscle cell proliferation and differentiation¹⁴⁷. Li et al.¹⁴⁸ found differentially expressed genes enriched in the cGMP-PKG signaling pathway in calves’ ruminal epithelium, suggesting a significant role of this pathway in rumen development in response to diet. In turn, the cAMP pathway was associated with regulating feed efficiency in pigs¹⁴⁹ and Nellore cattle¹⁵⁰.

Methane intensity

ACOXL (BTA 11) was found as a potential candidate gene for methane intensity. Acyl-CoA oxidase is a key enzyme in the β-oxidation pathway, involved in the catabolism of fatty acids for acetyl-CoA biosynthesis, a molecule used as an intermediate in several metabolic processes, including mitochondrial energy production¹⁵¹. ACOXL gene has a well-documented relationship with fat metabolism^152,153 and has already been associated with reproductive¹⁵⁴ and feed efficiency¹²³ traits in Nellore cattle.

MAP4K4 on BTA 11 encodes for a member of the serine/threonine protein kinase family and is involved in MAPK signaling pathway. MAPK cascade is a critical pathway for regulating several physiological processes, such as cell growth and differentiation, adaptation to environmental stress, and inflammatory responses^155,156. MAPK also plays a key role in regulating energy balance and thus, has been associated with feed efficiency in different production species^150,157. Moreover, MAPK signaling pathway is related to the development and proliferation of ruminal epithelia in small ruminants¹⁵⁸.

IL1R1 and IL1R2 (BTA 11) are interleukin-1 receptors, members of the TIR domain receptor superfamily (Toll–IL-1-receptor). TIR family is involved in innate immunity activation, the defense response against pathogenic microorganisms, and other insults. Specifically, interleukin-1 receptors trigger cytokine-induced immune and inflammatory responses to maintain barriers’ integrity or cell fitness¹⁵⁹. A study conducted in goats demonstrated that variations in the innate immune function of the host mucosa can affect the composition of the rumen microbial community¹⁶⁰. Additionally, recent studies have identified differences in the expression of IL1R2 mRNA and protein in the ruminal epithelium of different ruminant species^161,162. The IL1R2 gene has also been previously reported to be associated with the gut microbiota profile in pigs¹⁶³ and humans¹⁶⁴. These evidences suggest that host genetically mediated immune responses may interact with the ruminal microbiota and indirectly affect enteric methane production. This assumption is supported by a study that reported a differential abundance of proteins involved in cellular defense, including IL1R2, in the rumen epithelium of sheep with phenotype for high CH₄ emission¹⁶⁵.

Methane metabolic

The PLCB3 gene on BTA 29 encodes phospholipase C β3, a versatile enzyme involved in various signaling pathways, including calcium signaling pathway, vascular smooth muscle contraction, salivary secretion, taste transduction, gastric acid secretion, carbohydrate digestion and absorption, among others. This enzyme stimulates phosphoinositide hydrolysis to generate inositol 1,4,5-triphosphate (IP3), a product involved in intracellular Ca²⁺ control. IP3-dependent Ca²⁺ release promotes gastrointestinal smooth muscle contraction¹⁶⁶ and salivary secretion¹⁶⁷. Therefore, PLCB3 gene might indirectly influence methane emissions by controlling ruminal motility and pH. In addition, PLCB3 was enriched for the Wnt signaling pathway. A recent study showed that ruminal epithelial and muscle development is accompanied by activation of the Wnt and Ca²⁺ signaling pathways¹⁶⁸. This gene is also involved in Neutrophil extracellular trap formation and NOD-like receptor signaling pathways, which play an important role in regulating the immune system in the GIT¹⁶⁹.

The estrogen-related receptor alpha (ESRRA), on BTA 29, is a transcription factor highly expressed in the GIT, essential in regulating genes responsible for oxidative phosphorylation and lipid absorption¹⁷⁰. ESRRA is also involved in maintaining energy balance and metabolism and plays a key role in the stress-induced response to fasting, caloric restriction, or overnutrition^171,172. Connor et al.¹⁷³ observed high ESRRA mRNA expression in the rumen epithelium of newborn calves and reported that roughage intake at weaning appeared to induce the activation of ESRRA-mediated gene targets, suggesting that this gene may be involved in the function and energy metabolism of the developing rumen. Furthermore, studies have provided evidence attributed to the function of ESRRA in the regulation of food intake^174,175. This suggests that this gene may indirectly affect CH₄ emissions, through the control of the amounts of substrate ingested for enteric fermentation.

BAD (BTA 29) is one of the proapoptotic proteins of BCL-2 family, a gene group known to regulate programmed cell death. When dephosphorylated, BAD can interact with antiapoptotic proteins and promote cell death¹⁷⁶. Previous studies have provided strong clues that the BAD gene plays a role in rumen development in response to non-genetic factors. Lin et al.¹⁷⁷ reported that promoting microbiome-driven generation of ruminal VFA could decrease BAD gene expression, suggesting that when downregulated, BAD may inhibit apoptosis to accelerate physiological growth. Additionally, Xu et al.¹⁷⁸ also found the BAD gene downregulated in the rumen of sheep fed a high-grain diet, indicating that its downregulation may reduce the apoptotic rate of the ruminal epithelium.

Conclusion

The genome-wide association (GWAS) using DNA sequencing analyses allowed us to identify genetic variants spread across the entire genome associated with methane emission-related traits in Nellore cattle. The main genomic regions harbor promising genes that are involved in a series of biological processes and metabolic pathways related to gastrointestinal motility, salivary secretion, enteric nervous system, mucosal barrier integrity, epithelial transport, olfactory transduction, lipid metabolism, oxidative stress, cAMP, cGMP-PKG, MAPK cascade, among others. These results support the hypothesis that the genetic control of methane emissions in cattle may be mediated by physiological mechanisms that influence ruminal fermentation and digestion efficiency. The identification of consistent genomic signals, particularly those overlapping previously reported QTL for methane, feed efficiency, and production traits, reinforces the robustness of the results and highlights their potential applicability in future research and breeding programs. Although the results highlight the polygenic nature of methane emissions, further functional studies are needed to validate these candidate genes and better understand their roles in modulating this complex phenotype. Overall, this study contributes to a better understanding of the genetic architecture of methane-related traits and provides a basis for future research focused on developing selection strategies for more sustainable beef cattle production.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

L.M.A.: conceptualization; methodology; formal analysis; investigation; writing-original draft. L.F.M.M.: conceptualization; methodology; formal analysis; writing-review and editing. L.F.S.F.: collection and preparation of DNA samples; writing-review and editing. G.A.F.J.: sequence alignment and file preparation; writing-review and editing. S.L.C.N.: investigation; writing-review and editing. J.P.S.V.: methodology; data collection; writing-review and editing. T.L.S.S.: methodology; data collection; writing-review and editing; M.S.B.: methodology; data collection. J.A.S.: methodology; data collection; writing-review and editing. A.M.P.: methodology; data collection; writing-review and editing. M.E.Z.M.: resources; conceptualization; supervision; project administration; writing-review and editing. L.G.A.: conceptualization; funding acquisition; supervision; project administration; writing-review and editing. All authors read, revised, and approved the final manuscript.

Funding

This work was supported by the Foundation for Research Support of the State of São Paulo (FAPESP-Grants #2017/10630-2 and #2018/20026-8), and the scholarship was granted to L. M. Arikawa by FAPESP (Grants #2023/17818-8 and #2024/16663-3) and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior-Brazil (CAPES) - Financial Code 001.

Data availability

The sequencing data that support the findings of this study are available from the Gensys breeding program (https://gensys.com.br/), while the phenotypic data were provided by the Instituto de Zootecnia breeding program (http://www.iz.sp.gov.br/); however, their availability is restricted, as they were used under license for the present study and therefore cannot be publicly released. Nonetheless, the data can be obtained upon reasonable request and with authorization from the programs by contacting the corresponding authors (Dr. Lucia G. Albuquerque: galvao.albuquerque@unesp.br or Dr. Maria Eugênia Z. Mercadante: mezmercadante@gmail.com).

Declarations

Competing interests

The authors declare no competing interests.