Multi-omics integration of F_st, GWAS, TWAS and transcriptomics revealed candidate genes and pathways associated with pneumonia lesion scores in pigs

Abstract

Background

Respiratory diseases significantly impair pig growth performance by reducing average daily gain and feed conversion efficiency. Pneumonia lesion score (PLS) is a reliable phenotypic indicator for detecting respiratory issues in pig herds, including subclinical infections that often remain undiagnosed. In this study, PLS data was collected from 327 Large White pigs, and whole-genome resequencing (WGS) was performed. Lung tissue samples from pigs with extremely high and low PLS values were subjected to transcriptomic sequencing and Fixation Index (F_st) analysis. Additionally, a transcriptome-wide association study (TWAS) was conducted using FarmGTEx prediction models based on blood and lung tissues.

Results

By integrating F_st, Genome-Wide Association Study (GWAS), TWAS, and transcriptomic data, we identified key genomic regions and pathways associated with PLS variation, including immune regulation, inflammation, and metabolic adaptation. Candidate genes, RAB11FIP2 and SLC18A2 are implicated in immune modulation: RAB11FIP2 enhances TLR4 signaling and phagocytosis, whereas reduced SLC18A2 expression may compromise monoamine-mediated immune regulation. FBXL13 and ALPK2, identified via F_st and transcriptomic analysis, are linked to macrophage migration and Wnt/β-catenin signaling. Similarly, Genes like TRAF3IP1 and ASB1, highlighted by F_st and GWAS, contribute to ciliary transport, epithelial defense, and regulation of NF-κB and MAPK signaling. Moreover, GWAS and TWAS pinpointed RAMP1, as a novel candidate gene influencing immune modulation and airway responses through CGRP signaling, vascular tone regulation, and neuro-immune interactions, these findings have not been previously reported in PLS studies.

Conclusions

Overall, this study provides novel insights into the genetic mechanisms underlying porcine lung lesions and offers a foundation for breeding strategies aimed at improving respiratory disease resistance and herd health.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12864-026-13203-x.

Background

Respiratory health remains a major challenge in modern swine production worldwide, largely due to infectious agents such as Mycoplasma hyopneumoniae, porcine reproductive and respiratory syndrome virus (PRRSV), swine influenza virus (SIV), and bacterial pathogens including Pasteurella multocida. These infections cause extensive pulmonary damage, which can be quantitatively assessed through porcine lung score (PLS) a key indicator of pathological lesions in the respiratory tract and lung parenchyma. Impaired lung function severely compromises production efficiency, leading to reduced growth performance, poor feed conversion ratios, and increased mortality rates, collectively imposing significant financial burdens on the global swine industry [1, 2]. For instance, PRRS alone is estimated to cause annual losses exceeding 1.2 billion USD in the U.S. swine industry, ranking among the most economically devastating swine diseases [1].

Despite the significant economic and welfare implications, no major loci or genes have yet been conclusively associated with PLS, limiting the application of molecular breeding strategies for enhancing disease resilience. Previous studies have predominantly employed genome-wide association studies (GWAS) to explore lung lesion traits; however, such approaches are often confounded by factors including population stratification and environmental variability, which can obscure true genetic associations [3, 4].

In the present study, we investigated the genetic basis of PLS in American Large White pigs, a modern commercial breed characterized by rapid growth, high lean meat yield, and relatively robust pathogen tolerance. These traits make Large White pigs a valuable genetic resource, particularly in crossbreeding programs aimed at improving both productivity and disease resistance. To detect genomic regions under selection, we employed F_st analysis, a widely applied metric of population differentiation effectively identifies signatures of selection in domesticated species [5]. By comparing allele frequency divergence between high- and low-PLS groups, F_st offers a powerful means of pinpointing adaptive genetic variation.

To our knowledge, few PLS studies have integrated F_st based selection scans with transcriptomic profiling. Here, we combined whole-genome resequencing with RNA-seq–based transcriptome analysis to identify candidate loci associated with pulmonary resilience. Furthermore, GWAS was performed using a mixed linear model to uncover single nucleotide polymorphisms (SNPs) linked to lung lesion traits, while transcriptome-wide association study (TWAS) was conducted using the FarmGTEx platform (https://twas.farmgtex.org/ ). Integrating results across F_st, GWAS, TWAS, and transcriptomics enabled us to detect convergent signals and assess their functional relevance to porcine lung health.

The overarching objective of this work is to leverage a multi-omics framework to identify novel genes and pathways that govern pulmonary resilience in American Large White pigs. Our findings not only enhance understanding of the genetic underpinnings of PLS but also highlight promising targets for molecular breeding, with potential applications in genetic improvement, disease prevention, and sustainable pig production.

Materials and methods

Experimental animals and phenotype measurements

A total of 327 castrated male American Large White pigs were obtained from the nucleus herd of Guangxi Haihe Swine Breeding Co., Ltd. All animals were raised under standardized husbandry conditions without intentional pathogen challenge. Pigs of similar age were humanely slaughtered in two consecutive batches using electrical stunning followed by exsanguination, in accordance with the Chinese National Standard GB/T 17,236 − 2019.

Following slaughter, lungs were photographed, and lesions were assessed using the porcine lung score (PLS) system developed by Madec and Kobisch (1982). This method, widely adopted in abattoir-based health surveillance, is recognized for its balance between simplicity and diagnostic accuracy [6]. Moreover, it has been favorably compared with other systems such as Blaha’s method in terms of detail versus practicality trade-offs [7]. Each of the seven lobes (left apical, cardiac, and diaphragmatic; right apical, cardiac, diaphragmatic, and accessory) was assigned a maximum of 4 points, yielding a total score of 28. To increase scoring precision, every 12.5% of lesion coverage corresponded to 0.5 points.

Based on the PLS evaluation, pigs were ranked according to lesion severity, and 34 individuals with high-PLS and 34 individuals with low-PLS were selected for subsequent analyses. Lung tissues were collected from lesion sites in the high-PLS group and from healthy regions in the low-PLS group, immediately snap-frozen in liquid nitrogen, and stored at − 80 °C. The same 68 pigs were used for transcriptomic profiling and F_st analysis, ensuring consistency across multi-omics comparisons.

Phenotype data preprocessing

Raw PLS values were normalized using the Yeo–Johnson transformation [8] to ensure compliance with GWAS analysis requirements. This method was selected over the Box-Cox transformation due to the presence of zero values in the original PLS data, as Yeo-Johnson effectively accommodates non-positive observations while preserving interpretability. The transformation is defined as:

1

where y represents the original observed value and λ is the transformation parameter determined by maximum likelihood estimation.

WGS data processing

Genomic DNA was extracted from porcine ear tissue samples using the Megi Universal Nucleic Acid Extraction Kit, followed by quantitative analysis and quality assessment using the NanoDrop 2000 instrument (Thermo Fisher Scientific). High-quality DNA samples with OD260/280 ratios between 1.7 and 2.1 and concentrations ≥ 50 ng/µL were retained for sequencing by Kangpusen Company (Beijing, China). PE sequencing libraries with insert sizes of 350 bp were prepared using the CWseq Universal Direct Fast DNA Library Prep Kit. Libraries passing quality inspection were subjected to PE150 sequencing on the MGISEQ-T7 sequencer, and raw sequencing data were converted to fastq files. Raw sequencing reads were first quality-controlled and adapter-trimmed with fastp [9]. Clean reads were mapped to the porcine reference genome (Sscrofa11.1) with BWA-MEM [10], sorted and converted to BAM with SAMtools (version 1.7), deduplicated with Picard (version 2.26.2), and genotyped with GATK [11]. Hard filtering was carried out with GATK VariantFiltration using the expression QD < 2.0||FS > 60.0||MQ < 40.0||SOR > 3.0||MQRankSum < 12.5||ReadPosRankSum<-8.0, and variants failing any criterion were tagged with the filter name “LowQualSNP”. Quality control of the genotype data was performed with PLINK (version 1.90) [12], excluding SNPs with call-rate < 90% or minor allele frequency < 0.01 and removing individuals with a call-rate < 90%. Genotype imputation was conducted using Beagle (version 5.2) to eliminate missing values (NA) in the genotype data. After filtering, 15,041,991 SNPs from 327 individuals were retained for subsequent analysis. On average, the sequencing depth per individual was 5.84×, and the mean mapping rate to the reference genome was 90.41%.

Identification of selection signatures

In this study, we utilized vcftools (version 0.1.16) to compute Weir and Cockerham’s F_st, aiming to detect selection signals between high- and low-phenotype pig groups. The analysis was performed on quality-controlled Single Nucleotide Polymorphism (SNP) data in VCF format, with high- and low-phenotype groups specified using the --weir-F_st-pop parameter. A sliding window approach was employed, with a window size of 500 kb and a step size of 50 kb, to calculate genome-wide F_st values, enabling the identification of regions with elevated genetic differentiation. The output included both per-site and window-based F_st estimates. Negative F_st values, which lack biological significance and typically arise from statistical noise, were excluded from further analysis. To identify significantly differentiated regions, we used the quantile function in R (version 4.2.1) to select the top 1% of window-based F_st values as the threshold for significance. For visualization, Manhattan plots were generated using the R package CMplot (version 4.0.0), with weighted F_st values plotted on the y-axis and genomic positions on the x-axis.

Genome-wide association study

GWAS was performed for PLS using the LDAK software [13]. The mixed linear model for the GWAS is as follows:

 y=Xb+Wa+Zu+e 2

 y is the vector of phenotypic values, and  b is the vector of fixed effects, including the first three principal components.  a denotes the fixed effect of SNPs.  u denotes the random polygenic effects, which are assumed to be normal distribution as . Here,  G is the kinship matrix built from the imputed SNP data, and is the additive genetic variance.  X,  W and  Z are the respective association matrices. The residual vector  e is also assumed to be normally distributed, described by , where  I represents the identity matrix, and is the residual variance. Due to slight population stratification, the population PCA values are calculated using the PLINK software (version 1.9) with the command --pca 10, and the first three principal components are used to adjust for population structure.

In this study, the Bonferroni correction was applied to adjust for multiple comparisons. The imputed dataset was pruned using PLINK (version 1.9) with the parameter --indep-pairwise 50 5 0.3, resulting in 900,185 effective SNPs. Accordingly, the genome-wide significance threshold was set at 0.05 / 900,185 = 5.55 × 10⁻⁸, while the chromosome-wide significance threshold was defined as 1 / 900,185 = 1.11 × 10⁻⁶ [14]. The Manhattan and quantile-quantile (Q-Q) plots were drawn using the CMplot R software package [15]. The phenotypic variation explained by each SNP can be calculated using the formula , where is the minor allele frequency for SNPs,  b is the allelic substitution effect of SNPs, and is the phenotypic variance of the trait.

Transcriptomic sequencing

Lung tissue samples were sent to Shanghai Majorbio Bio-pharm Technology Co., Ltd. for RNA extraction and transcriptomic sequencing. Total RNA was extracted using TRIzol Reagent (Invitrogen, USA). RNA concentration and purity were assessed using Nanodrop2000, RNA integrity was evaluated by agarose gel electrophoresis, and RNA Integrity Number (RIN) values were determined using Agilent2100. Library construction criteria included total RNA ≥ 1 µg, concentration ≥ 50 ng/µL, and OD260/280 between 1.8 and 2.2. mRNA was enriched using magnetic beads conjugated with Oligo(dT), randomly fragmented to approximately 300 bp using fragmentation buffer, and reverse transcribed to synthesize first- and second-strand cDNA. Final sequencing libraries were obtained after end repair, A-tailing, adapter ligation, and PCR amplification. Libraries were sequenced on the DNBSEQ-T7 platform with 2 × 150 bp PE sequencing.

Raw reads were adapter-trimmed and quality-filtered with Trimmomatic software (version 0.39). The resulting clean reads were aligned to the Sscrofa11.1 reference genome using STAR (version 2.7.9a), after which gene-level read counts per individual were quantified with featureCounts. Differential expression analysis between the high- and low-PLS groups was performed using the DESeq2 package (version 1.34.0) [16]. Genes with FDR < 0.05 and |log2 fold change| > 1 were defined as differentially expressed genes.

Functional enrichment analysis

In this study, GO and KEGG enrichment analyses were performed separately for F_st candidate genes, upregulated differentially expressed genes, and downregulated differentially expressed genes. All enrichment analyses were conducted using the R package clusterProfiler [17]. For each gene set, the top ten enriched GO terms and KEGG pathways were mainly considered to summarize the major biological functions and signaling pathways involved in different analytical layers. To evaluate whether genetic differentiation signals and transcriptomic changes showed functional convergence at the pathway level, we further compared the significantly enriched KEGG pathways among different gene sets and focused on the overlapping significant KEGG pathways for further interpretation.

Transcriptome-wide association study

We leveraged the FarmGTEx TWAS-server (https://twas.farmgtex.org/), an interactive and user-friendly web platform developed to perform transcriptome-wide association studies (TWAS) in farm animals [18]. The server integrates expression and eQTL (Expression Quantitative Trait Locus) data across 34 pig tissues (from 5,457 individuals), and supports TWAS by providing prediction models for 21,037 pig genes and offering 657 tissue–gene–trait associations relevant to pig traits .

For this study, we focused on blood and lung tissues in TWAS, as both are directly related to immune responses and respiratory health, which are fundamental to PLS phenotypes. GWAS summary statistics were uploaded to the FarmGTEx TWAS-server, where gene expression prediction and association testing were performed using the MetaXcan software package with Elastic Net-based imputation models. Significant associations were determined using a Bonferroni-corrected p-value threshold, with additional false discovery rate (FDR < 0.05) control to account for multiple testing. Genes that passed both criteria were designated as candidate hits for further investigation.

Results

Descriptive statistics

Table 1 provides a detailed summary of the maximum, minimum, mean, and standard error for PLS in Large White pigs after Yeo-Johnson transformation. The maximum and minimum PLS values were 2.97 and 0, respectively, with a mean of 1.29 and a standard error of 0.05. The estimated transformation parameter λ was − 0.076. For the extreme groups used in F_st and transcriptomic analyses, the high-PLS group (n = 34) had a mean transformed PLS score of 2.57 ± 0.04, while the low-PLS group (n = 34) had a mean of 0. A t-test revealed significant differences between the two groups (P < 2.2E-16), confirming the effectiveness of selecting extreme phenotypes for downstream analyses.

Table 1.

Descriptive statistics of PLS after Yeo-Johnson transformation in the large white pig population

Pheno	λ	Group	Max	Min	Mean	SE	p-Value
PLS	-0.076	all (n = 327)	2.97	0	1.29	0.05	/
		high (n = 34)	2.97	2.19	2.57	0.04	p < 2.2E-16
		low (n = 34)	0	0	0	0

PLS values are presented for all pigs and for the two extreme phenotype groups selected for downstream analyses. λ denotes the transformation parameter, and SE denotes standard error. The p value refers to the significance of the difference between the high-PLS and low-PLS groups

Selection signature analysis

To detect genomic regions under selection in Large White pigs, we performed F_st analysis using whole-genome resequencing data. Candidate regions were defined as genomic segments with F_st values in the top 1% of the genome-wide empirical distribution, and genes within these high-F_st regions were considered potential candidate genes. As shown in Fig. 1A, prominent F_st peaks were observed across multiple chromosomes, with several surpassing the significance threshold (F_st ≥ 0.0745178), indicating potential selection pressure. This analysis identified 220 candidate genes (Supplementary Table S1), representing likely targets of selection within these regions.

Fig. 1.

Genomic selection signals and functional enrichment in Large White pigs. A Manhattan plot of F_st analysis based on whole-genome resequencing comparing the high-PLS group with the low-PLS group, with the red dashed line indicating the 1% significance threshold. B GO and KEGG enrichment analysis of candidate genes identified in regions under selection in the PLS high group compared with the PLS low group

The strongest signal was detected on Sus Scrofa Chromosome (SSC) 15, with a maximum F_st value of 0.173. Other chromosomes exceeding the threshold included SSC1, SSC2, SSC7, SSC9, SSC11, SSC12, and SSC13. Supplementary Table S2 details the maximum F_st values and nearby genes for the strongest signals on each chromosome. These results underscore genomic regions potentially associated with the PLS trait in Large White pigs, particularly the prominent peak on SSC15 and elevated signals on other chromosomes, which may contain variants under selective pressure.

Functional enrichment analysis of the F_st-identified candidate genes revealed significant enrichments in various categories using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases. GO analysis indicated enrichment in biological processes, including response to light stimulus, cellular response to abiotic stimulus, and negative regulation of cell growth; cellular components such as extracellular matrix and phosphatidylinositol 3-kinase complex; and molecular functions like enzyme regulator activity, kinase regulator activity, and G-protein beta/gamma-subunit complex binding. KEGG pathway analysis highlighted significant enrichment in Apelin signaling, Oxytocin signaling, and Chemokine signaling pathways (Fig. 1B, Supplementary Table S3).

Genome-wide association study

Principal component analysis (PCA) indicated mild population stratification in the Large White pig cohort (Supplementary Figure S1A). To identify genetic variants associated with the PLS trait, we performed GWAS using PLS scores from 327 pigs and 15,041,990 high-quality filtered SNPs. SNP distribution across the genome was assessed in 1 Mb windows, as illustrated in Supplementary Figure S1B, providing insights into marker density. GWAS identified SNPs with p-values below predefined significance thresholds as associated with PLS, revealing 72 SNPs at chromosome-wide significance and 1 at genome-wide significance. These SNPs mapped to SSC9, SSC14, and SSC15. The Manhattan plot (Fig. 2A) shows the genome-wide distribution of SNP associations, while the Q-Q plot (Fig. 2B) evaluates model fit. Functional annotation of the 73 significant SNPs showed that 38 reside in intronic or exonic regions, with the remainder in noncoding RNA or intergenic areas. Additionally, 39 candidate genes were identified within 500 kb upstream and downstream of these SNPs (Supplementary Table S4).

Fig. 2.

GWAS results of PLS in Large White pigs. A Manhattan plot of GWAS based on 327 pigs and 15,041,990 high-quality SNPs. B Q-Q plot of GWAS analysis for PLS

Transcriptomic analysis

To further explore the molecular mechanisms underlying the PLS trait in Large White pigs, we performed RNA Sequencing (RNA-seq) analysis on lung tissues from high- and low-PLS groups. Comparative gene expression analysis identified 1,096 differentially expressed genes (DEGs), comprising 909 up-regulated and 187 down-regulated genes in the high-PLS group relative to the low-PLS group. Of these, 475 DEGs were successfully annotated with gene names. For transparency and to support future research, a complete list of all 1,096 DEGs, including annotated and unannotated entries, is provided in Supplementary Table S5. Functional analyses, including GO and KEGG pathway enrichment, were conducted using the annotated DEGs, as these genes can be reliably mapped to established biological processes and pathways. The volcano plot (Fig. 3A) illustrates the distribution and expression differences of these DEGs.

Fig. 3.

Transcriptomic analysis of PLS high and low groups in Large White pigs. A Volcano plot showing DEGs between the PLS high and low groups, with 1,096 DEGs identified. B GO terms and KEGG pathways enriched among up-regulated DEGs. C GO terms and KEGG pathways enriched among down-regulated DEGs

GO enrichment analysis was conducted separately on up-regulated and down-regulated DEGs from RNA-seq of lung tissues in Large White pigs (Supplementary Table S6). Up-regulated DEGs showed significant enrichment in biological processes such as ATP synthesis coupled electron transport, energy derivation by oxidation of organic compounds, and cellular respiration (Fig. 3B). In contrast, down-regulated DEGs were enriched in NADH dehydrogenase (ubiquinone) activity and oxidoreductase-driven active transmembrane transporter activity (Fig. 3B). KEGG pathway analysis indicated that up-regulated DEGs were enriched in neuroactive ligand-receptor interaction, bile secretion, cell adhesion molecules, and chemical carcinogenesis - reactive oxygen species pathways, whereas down-regulated DEGs were enriched in metabolism of xenobiotics by cytochrome P450, chemical carcinogenesis - DNA adducts, and steroid biosynthesis pathways (Fig. 3C, Supplementary Table S6).

Transcriptome-wide association study

TWAS analysis was performed using GWAS summary statistics through the FarmGTEx TWAS-server. As shown in Fig. 4A and B, and detailed in Table 2, only two gene-tissue associations reached the predefined significance threshold: TTC29 in lung tissue and RAMP1 in blood tissue. These findings indicate that the genetically predicted expression of these genes may be associated with PLS variation, although further validation is required.

Fig. 4.

TWAS results of PLS in Large White pigs. A Manhattan plot of TWAS analysis. B FDR-adjusted TWAS results

Table 2.

Significant gene-tissue associations identified by TWAS for PLS

Gene toperform ID	Gene	CHR	Start	End	TWAS.P	TWAS.Z	FDR	Gene.tissue
ENSSSCG00000016331	RAMP1	15	137,490,087	137,535,555	5.76546E-06	-4.53481279	0.025742779	RAMP1:Blood
ENSSSCG00000040272	TTC29	8	81,746,489	81,982,760	7.12807E-06	4.489827908	0.012944582	TTC29:Lung

CHR indicates chromosome, Start and End indicate genomic coordinates based on the reference genome assembly, TWAS.P represents the nominal TWAS p value, TWAS.Z represents the TWAS Z-score and FDR indicates the false discovery rate-adjusted p value

Identification of overlapping candidate genes from F_st, GWAS, DEGs, and TWAS analyses

To elucidate the genetic basis of the PLS trait in Large White pigs, we integrated results from four multi-omics approaches: F_st selection signal analysis, GWAS, RNA-seq-derived DEGs, and TWAS. A Venn diagram (Supplementary Figure S2) illustrates overlaps among candidate gene sets: 220 genes in top 1% F_st regions (Supplementary Table S1), 39 genes within 500 kb of significant GWAS SNPs (Supplementary Table S4), 475 DEGs from lung tissues (Supplementary Table S5), and two TWAS genes (Table 2). The diagram highlights pairwise intersections, suggesting convergent evidence for genetic regulation of PLS from selection, expression, and association data. Although no genes overlapped across all four analyses, indicating partial rather than complete convergence, these pairwise overlaps represent strong candidates for roles in lung health and adaptation in Large White pigs (Table 3).

Table 3.

Overlapping candidate genes identified by integrating F_st, GWAS, RNA-seq, and TWAS analyses for PLS

Fst	DEGs	GWAS	TWAS	set	values	count
FALSE	FALSE	TRUE	TRUE	(GWAS∩TWAS)∖(Fst∪DEGs)	RAMP1	1
FALSE	TRUE	TRUE	FALSE	(DEGs∩GWAS)∖(Fst∪TWAS)	RAB11FIP2, RBM44, SLC18A2	3
TRUE	FALSE	TRUE	FALSE	(Fst∩GWAS)∖(DEGs∪TWAS)	TRAF3IP1, ASB1	2
TRUE	TRUE	FALSE	FALSE	(Fst∩DEGs)∖(GWAS∪TWAS)	ALPK2, CATSPERD, FBXL13	3

TRUE and FALSE indicate whether a gene was identified by the corresponding analysis. “set” denotes the overlap pattern among methods, “values” lists representative genes in each overlap category, and “count” indicates the number of genes in that category

Considering the limited number of candidate genes identified by GWAS and TWAS, we focused the pathway-level comparison on F_st candidate genes, upregulated DEGs, and downregulated DEGs. In addition to gene-level overlap, we further compared the KEGG enrichment results of F_st candidate genes, upregulated DEGs, and downregulated DEGs to evaluate potential convergence at the pathway level. The pathways Adrenergic signaling in cardiomyocytes and Cholinergic synapse were shared between F_st candidate genes and upregulated DEGs. In addition, Neuroactive ligand-receptor interaction was shared between upregulated and downregulated DEGs. These overlapping KEGG pathways, summarized in Supplementary Table S7, suggest that different omics layers may converge at the functional pathway level, even when overlap at the individual gene level is limited.

Discussion

Genetic architecture of PLS and rationale for multi-omics integration

Pulmonary lesion score (PLS) has emerged as an important indicator of respiratory health and disease resistance in pigs. To date, however, no single major gene has been unequivocally associated with this trait, suggesting that PLS is a complex quantitative trait shaped by the interplay of multiple genomic loci and environmental influences. Deciphering the genetic architecture of PLS is therefore critical for improving our understanding of respiratory pathophysiology and for developing breeding strategies aimed at enhancing disease resilience in swine.

Recent studies highlight the value of integrating F_st, GWAS, TWAS, and transcriptomic profiling to unravel the genetic basis of complex traits. For example, Liu et al. demonstrated that combining genotypic and transcriptomic data can identify regulatory variants linked to economically important traits [19]. Uemoto et al. employed GWAS and haplotype analysis to uncover novel quantitative trait loci (QTL) related to production, respiratory disease, and immune traits in pigs [20]. Similarly, Liu et al. showed that integrating GWAS with F_st analyses can reveal candidate genes for meat color [3]. TWAS has further strengthened trait-gene connections by integrating GWAS signals with gene expression prediction models. The recent FarmGTEx TWAS-server now facilitates such analyses across 34 pig tissues, enabling functional annotation of more than 21,000 genes [18]. Collectively, these studies underscore the power of multi-omics integration in prioritizing biologically relevant candidate genes.

It should be noted that these four analytical approaches capture different layers of biological information and are not expected to identify completely overlapping gene sets. F_st analysis detects genomic regions showing allele frequency differentiation between high- and low-PLS groups, which may reflect genetic divergence related to disease susceptibility or resistance [21]. GWAS identified genetic variants statistically associated with phenotypic variation in PLS [22]. In contrast, transcriptomic analysis reveals genes whose expression differs between high- and low-PLS individuals, reflecting downstream molecular responses in lung tissues [23]. TWAS further integrates GWAS summary statistics with tissue-specific expression prediction models to prioritize genes whose genetically regulated expression may be associated with the trait [24]. Therefore, limited gene-level overlap among these approaches is biologically plausible for a complex quantitative trait such as PLS. Rather than expecting complete convergence at the individual gene level, the integration of these complementary analyses provides a framework for identifying recurrent biological processes and pathways involved in PLS variation.

In the context of respiratory traits, GWAS has played a particularly crucial role. Tong et al. leveraged whole-genome sequencing in case-control and quantitative GWAS of swine lung lesions, identifying four significant loci. RNF216 and CALD1 were confirmed as susceptibility genes, while TMEM132B, TMEM132C, TMEM132D, and NCOR2 (on SSC14) and PID1 (on SSC15) emerged as novel candidates influencing lesion severity [25]. In sheep, Huang et al. demonstrated the effectiveness of integrating GWAS with transcriptomic analysis for dissecting lung pathology, further validating the value of multi-layered approaches [26]. These insights provide valuable reference points for our own investigation of the genetic mechanisms underlying the PLS trait in Large White pigs.

Against this background, the present study aimed to dissect the genetic determinants of PLS in Large White pigs by integrating F_st, GWAS, TWAS, and transcriptomic profiling. By synthesizing evidence across complementary methods, we sought to enhance the robustness of candidate gene discovery and to advance mechanistic understanding of respiratory disease resistance, ultimately providing genetic resources for molecular breeding.

Genomic signals associated with PLS revealed by F_st and GWAS

From the genome-wide selection scans revealed significant differentiation signals on several chromosomes, most prominently on SSC15, supporting the notion that PLS variation arises from polygenic contributions rather than single-gene effects. Functional enrichment analyses highlighted the chemokine and apelin signaling pathways. Chemokine signaling mediates immune cell recruitment and inflammation in porcine respiratory disease [27, 28], while apelin signaling has been implicated in vascular stability and attenuation of lung inflammation [29]. These findings suggest that both natural and artificial selection may have acted on immune- and inflammation-related genes influencing PLS.

The GWAS revealed notable loci and candidate genes linked to the PLS trait in Large White pigs, specifically on SSC14 and SSC15, distinct from those previously reported [25]. On SSC14, a 2.94 Mb region (124.9–127.8 Mb) containing 57 significant SNPs was identified, with the lead SNP at 14_126907688. On SSC15, novel associations included intronic SNP 15_137499570 and three SNPs located in the 3′untranslated region (UTR) (15_137706054, 15_137706069, 15_137706089). Candidate genes within these regions, such as GFRA1, RAMP1, PER2, TRAF3IP1, CADM1, and HES6, are implicated in immune regulation, inflammation, and lung tissue homeostasis [30–33]. Highlighting these loci as potential drivers of PLS variation.

Transcriptomic and TWAS insights into the biological basis of PLS

Transcriptomic profiling provided complementary insights into molecular mechanisms. In high-PLS pigs, genes involved in mitochondrial respiration and ATP production were upregulated, suggesting heightened metabolic demand in response to respiratory stress [34]. Conversely, downregulation of detoxification and steroid biosynthesis pathways, including cytochrome P450 genes, implied reduced metabolic resilience and greater susceptibility to tissue damage [34, 35]. Upregulated genes associated with neuroactive ligand-receptor interactions and cell adhesion pointed to immune modulation and tissue repair processes, consistent with responses to Mycoplasma hyopneumoniae infection [28].

TWAS analysis using the FarmGTEx server revealed tissue-specific associations after multiple-testing correction (Bonferroni correction and False Discovery Rate < 0.05), Several notable tissue-gene associations were identified. In lung tissue, TTC29 showed a strong association with PLS, implicating its role in epithelial integrity, whereas in blood, RAMP1 emerged as a significant candidate, consistent with its involvement in immune modulation and systemic inflammatory responses. These findings highlight the interplay between lung-specific and systemic mechanisms underlying PLS.

Integrated candidate genes and study limitations

Although no single gene overlapped across all four approaches, pairwise intersections identified robust candidates. For example, RAB11FIP2 and SLC18A2 showed up in both GWAS and transcriptomic results, with RAB11FIP2 boosting TLR4 signaling and cell eating [36], and lower SLC18A2 levels possibly messing with monoamine-based immune control [37]. F_st–transcriptome overlaps implicated FBXL13 and ALPK2, involved in macrophage migration and Wnt/β-catenin signaling [38, 39]. F_st and GWAS intersections prioritized TRAF3IP1 and ASB1, where TRAF3IP1 handles ciliary movement and epithelial protection [40], and ASB1 manages NF-κB and MAPK paths in immune reactions [41]. Finally, GWAS and TWAS overlaps reinforced RAMP1, a core regulator of calcitonin gene-related peptide (CGRP) signaling and airway inflammation [42, 43]. RAMP1 signaling eases acute lung injury by dialing back macrophage action, highlighting its shielding role in inflammatory lung issues [33].

Taken together, convergent evidence across methods highlights a set of biologically plausible candidates for PLS. At the pathway level, enrichment of chemokine and apelin signaling complements these gene-level findings, offering a mechanistic framework linking immune recruitment, vascular homeostasis, and anti-inflammatory protection.

Convergent evidence across F_st, GWAS, TWAS, and transcriptomic analyses delineated a focused set of functionally plausible candidates underlying PLS. The GWAS–TWAS overlap highlighted RAMP1, a key regulator of immune modulation and airway physiology through its role in calcitonin gene-related peptide (CGRP) signaling, vascular tone, and neuro–immune interactions. Functionally, RAMP1 contributes to the suppression of airway inflammation, attenuation of hyperresponsiveness, and protection against acute lung injury by limiting macrophage activation. The GWAS–transcriptome overlap further prioritized RAB11FIP2 and SLC18A2 emerged from the GWAS–transcriptome overlap, implicating these genes in innate immune activation and neuro–immune regulation, respectively. From the F_st–transcriptome intersection, FBXL13 and ALPK2 were identified, linking macrophage migration and Wnt/β-catenin signaling to inflammatory responses and tissue remodeling in the lung. The F_st–GWAS overlap pointed to TRAF3IP1 and ASB1, both integral to ciliary transport, epithelial defense, and NF-κB/MAPK pathway regulation—mechanisms essential for mucociliary clearance and inflammation control. At the pathway level, enrichment of chemokine and apelin signaling pathways further supports these findings by providing a broader biological framework involving immune cell recruitment, vascular homeostasis, and anti-inflammatory regulation.

In addition to gene-level intersections, pathway-level overlap provided further evidence for functional convergence among different analytical layers. Because the numbers of GWAS- and TWAS-derived candidate genes were limited, pathway overlap analysis focused on F_st candidate genes, upregulated DEGs, and downregulated DEGs. Notably, Adrenergic signaling in cardiomyocytes and Cholinergic synapse were shared between F_st candidate genes and upregulated DEGs. Although these pathways are annotated based on cardiovascular or neuronal functions, their components are also involved in cAMP/PKA signaling, calcium regulation, airway responsiveness, mucus secretion, macrophage activity, and neuro-immune regulation, all of which are relevant to pulmonary inflammation and host defense [44–46]. This overlap suggests that genetic differentiation between high- and low-PLS pigs may converge with transcriptionally activated neuroendocrine and immune-regulatory processes in the lung. Moreover, Neuroactive ligand-receptor interaction was enriched in both upregulated and downregulated DEGs, indicating that this pathway may undergo bidirectional transcriptional remodeling rather than uniform activation or suppression. This pathway includes diverse neurotransmitters, neuropeptides, receptors, and ligand-encoding genes, which may participate in airway reactivity, immune cell recruitment, vascular regulation, and inflammatory signaling [46, 47]. Therefore, the recurrent enrichment of neuroactive signaling-related pathways supports the involvement of neuro-immune interactions in the development of pneumonia lesions. These findings indicated that, even when overlap at the individual gene level is limited, convergence at the pathway level can provide complementary insight into the biological mechanisms underlying PLS.

Nonetheless, several limitations should be acknowledged. The transcriptomic and F_st analyses were based on relatively small sample sizes (34 pigs per group), which may limit detection power, whereas the GWAS included 327 animals. Functional validation of prioritized genes such as RAMP1, RAB11FIP2, and SLC18A2 were not performed, and subclinical infections may have influenced PLS despite controlled husbandry. Moreover, the lack of a universally overlapping gene across all methods underscores the complexity of PLS genetics, reflecting its polygenic and multi-pathway nature. Larger cohorts, pathogen-free environments, and experimental validation are needed to further refine these findings.

In conclusion, our study identified novel genes and pathways associated with PLS in Large White pigs, including RAMP1, RAB11FIP2, SLC18A2, FBXL13, ALPK2, TRAF3IP1, and ASB1, together with enrichment of chemokine and apelin signaling. These results provided new mechanistic insights into respiratory disease susceptibility and offer valuable genetic resources for marker-assisted and genomic selection aimed at improving swine resilience to respiratory disorders.

Conclusions

By integrating F_st, GWAS, TWAS, and transcriptomic analyses in a Large White pig population, we identified multiple genomic regions, genes, and pathways associated with PLS. Our results demonstrated that PLS was not determined by a single major locus, but rather shaped by the combined effects of multiple genes involved in immune regulation, inflammatory responses, metabolic adaptation, and epithelial defense. Notably, convergent evidence across different approaches highlighted candidates such as RAMP1, RAB11FIP2, SLC18A2, FBXL13, ALPK2, TRAF3IP1, and ASB1, all of which represent novel candidates previously unreported in PLS studies, underscoring their potential contributions to respiratory health. What’s more, we found pathways like chemokine signaling and apelin signaling showing up prominently, which adds useful insights into how the host’s immune response ties in with managing inflammation and maintaining tissue strength. Taken together, our results help expand what we know about the genetic setup of PLS, while also supplying handy tools to move forward with breeding approaches that build up pigs’ ability to fight off diseases and stay tough overall.

Abbreviations

DEG

Differentially Expressed Gene

eQTL

Expression Quantitative Trait Locus

F_st

Fixation Index

GO

Gene Ontology

GWAS

Genome-Wide Association Study

KEGG

Kyoto Encyclopedia of Genes and Genomes

OD260/280

Optical Density at 260 nm and 280 nm (indicator of nucleic acid purity)

PLS

Pneumonia Lesion Scores

QTL

Quantitative Trait Locus

RIN

RNA Integrity Number

RNA-seq

RNA Sequencing

SNP

Single Nucleotide Polymorphism

SSC

Sus scrofa Chromosome

TWAS

Transcriptome-wide association study

WGS

Whole-Genome Resequencing

Authors’ contributions

G.Z, Y.Y, P.L and Q.Z conceptualized, supervised and investigated the article. P.L, R.H and Q.Z provided funding acquisition and methodology. G.Z and Y.Y analyzed the data. Y.Y, J.C and J.M installed software. G.Z, J.H, M.L, X.H, J.Y and J.Z collected lung tissue. G.Z and Y.Y wrote the original draft. Q.Z, P.L and C.C reviewed the article. All authors have read and agreed to the published version of the manuscript.

Funding

This work was financially supported by Biological Breeding-National Science and Technology Major Projects (2023ZD04070).

Data availability

The genotype datasets used and analyzed during this study are available in the Figshare repository, https://doi.org/10.6084/m9.figshare.30069106. The RNA-seq datasets used and analyzed during this study are available in the Genome Sequence Archive (GSA) https://ngdc.cncb.ac.cn/gsa/browse/CRA031493. The WGS datasets used and analyzed during this study are available in the European Variation Archive (EVA) https://www.ebi.ac.uk/eva/?eva-study=PRJEB101011.

Declarations

Ethics approval and consent to participate

All laboratory animals were raised following the guidelines for the care and use of laboratory animals prepared by the Institute of Animal Welfare and Ethics Committee of Nanjing Agricultural University. All experimental schemes have been approved by the Animal Care and Use Committee of Nanjing Agricultural University (certificate No. SYXK (Su) 2022-0031).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.