Abstract

   As a significant infectious disease in livestock, porcine reproductive
   and respiratory syndrome (PRRS) imposes substantial economic losses on
   the swine industry. Identification of diagnostic markers and
   therapeutic targets has been a focal challenge in PPRS prevention and
   control. By integrating metabolomic and lipidomic serum analyses of
   clinical pig cohorts through a machine learning approach with in vivo
   and in vitro infection models, lysophosphatidic acid (LPA) is
   discovered as a serum metabolic biomarker for PRRS virus (PRRSV)
   clinical diagnosis. PRRSV promoted LPA synthesis by upregulating the
   autotaxin expression, which causes innate immunosuppression by
   dampening the retinoic acid‐inducible gene I (RIG‐I) and type I
   interferon responses, leading to enhanced virus replication. Targeting
   LPA demonstrated protection against virus infection and associated
   disease outcomes in infected pigs, indicating that LPA is a novel
   antiviral target against PRRSV. This study lays a foundation for
   clinical prevention and control of PRRSV infections.

   Keywords: antiviral approach, diagnostic biomarker, immune suppression,
   lysophosphatidic acid (LPA), machine learning, multi‐omics, porcine
   reproductive and respiratory syndrome virus
     __________________________________________________________________

   This study analyzes metabolomic and lipidomic profiles of
   PRRSV‐infected pigs using a machine learning approach, identifying LPA
   as a serum diagnostic biomarker. PRRSV augments LPA synthesis by
   upregulating ATX expression, leading to RIG‐I and IFN‐β suppression,
   and increases virus replication. Targeting LPA alleviates clinical
   symptoms, reduces viral loads, and mitigates viral shedding.

   graphic file with name ADVS-11-2402025-g006.jpg

1. Introduction

   Porcine reproductive and respiratory syndrome (PRRS), stemming from
   infection with porcine reproductive and respiratory syndrome virus
   (PRRSV), manifests as a highly contagious swine infectious ailment
   known for its capacity to induce host immune suppression. This malady
   exerts a substantial economic burden, resulting in annual production
   losses approaching one billion USD within the swine husbandry sector.^[
   [56]^1 ^] Clinical diagnosis of PRRS in pig populations is initially
   based on the manifestation of symptoms. However, variations in the
   rearing environments across different swine farms and the diverse
   health conditions of pig herds contribute to varying clinical
   presentations following PRRSV infection. Moreover, infections by other
   pathogens, such as pseudorabies virus (PRV), Actinobacillus
   pleuropneumoniae, and others, can induce similar clinical symptoms,
   which hinders accurate clinical diagnosis.^[ [57]^2 , [58]^3 , [59]^4
   ^]

   Currently, conventional laboratory diagnostic methods remain
   predominant in the clinical diagnosis of PRRSV. Egli et al. established
   a fluorescence reverse transcription‐polymerase chain reaction (RT‐PCR)
   method capable of rapid discrimination between PRRSV type 1 and type 2
   strains with high specificity and sensitivity.^[ [60]^5 ^] Sorensen
   et al. introduced a blocking enzyme linked immunosorbent assa (ELISA)
   for high‐throughput PRRSV antibody detection in serum samples.^[ [61]^6
   , [62]^7 ^] Additionally, oral fluid can be used for antibody detection
   with high sensitivity and specificity compared to serum.^[ [63]^8 ,
   [64]^9 , [65]^10 , [66]^11 ^] It has been widely employed for
   population‐level PRRSV antibody monitoring in sow herds and growing pig
   populations.^[ [67]^9 , [68]^12 , [69]^13 ^] However, to date, there
   has been a lack of development and application of diagnostic biomarkers
   specific to PRRSV.

   Presently, the clinical management of PRRS primarily relies upon
   palliative interventions aimed at symptomatic relief, given the absence
   of effective pharmacological treatments. Moreover, vaccine‐based
   immunization strategies face formidable challenges attributable to the
   marked variability exhibited by PRRSV.^[ [70]^14 ^] Therefore, there
   exists an urgent need for the identification of precise therapeutic
   targets endowed with the capacity to combat PRRSV infection in a highly
   specific manner.^[ [71]^15 , [72]^16 ^]

   During viral infections, viruses can induce metabolic alterations in
   infected host cells. Manipulating these changes holds promise for
   altering the outcomes of viral infections, thereby offering potential
   targets for therapies aimed at inhibiting viral infection or
   replication. For instance, influenza virus, dengue virus, and severe
   acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) have been
   observed to induce altered glucose and lipid metabolism in infected
   cells.^[ [73]^17 , [74]^18 , [75]^19 , [76]^20 ^] In the case of
   hepatitis B virus, the generation of lactate, dependent on lactate
   dehydrogenase A, suppresses RIG‐I‐like receptors (RLR) signaling, which
   in turn leads to immune evasion.^[ [77]^21 ^] Targeting these metabolic
   pathways has been found as an effective strategy to restrain viral
   replication. Clinical investigations of coronavirus disease 2019
   (COVID‐19) have unveiled close associations between circulating serum
   metabolites, inflammatory cytokines, and the immune competence of
   patients, potentially identifying them as promising therapeutic
   targets.^[ [78]^22 ^]

   During PRRSV infection, it has been shown that PRRSV infection actively
   promotes glycolytic metabolism that contributes to the maintenance of
   the tricarboxylic acid (TCA) cycle.^[ [79]^23 ^] Analysis of serum
   samples obtained from PRRSV‐infected pregnant sows suggests that
   certain molecules, namely alpha‐AAA, kynurenine, and lysoPCs, hold
   promise as crucial biomarkers associated with PRRSV infection.^[
   [80]^24 ^] These findings strongly advocate for the potential of
   metabolic modulation as a pivotal avenue for therapeutic interventions
   against PRRSV.

   In this study, by integrating clinical cohorts, multi‐omics analyses,
   and molecular mechanistic approaches, we identified and subsequently
   validated lysophosphatidic acid (LPA) as a prospective target for both
   diagnostic and therapeutic purposes in the context of PRRSV.
   Importantly, we provided compelling evidence confirming the efficacy of
   LPA‐targeting as a novel treatment of PRRSV. This study may provide new
   opportunities for the diagnosis and eradication of PRRSV in swine
   population.

2. Results

2.1. Utilizing Machine Learning for the Identification of Metabolic and Lipid
Serum Biomarkers in a Clinical Cohort of PRRSV‐Infected Pigs

   We employed a swine cohort to investigate the metabolic biomarkers in
   the serum samples of pigs infected with PRRSV. To do so, a cohort of a
   total of 110 animals, including PRRSV‐positive (n = 55) and negative
   matched controls (n = 55), were selected based on a quantitative
   reverse transcription polymerase chain reaction (RT‐qPCR) detection
   method, presence of PRRSV‐specific clinical signs, and absence of other
   relevant and common pig diseases. All animals were kept under
   monitoring for a change in body temperature (Figure [81] 1A,B) and the
   appearance of PRRSV‐specific clinical symptoms before the collection of
   serum samples (Figure [82]1B). Subsequently, serum samples were
   subjected to untargeted metabolomic and lipidomic analyses separately
   by using a mass spectrometry approach (Figure [83]1C). A protocol for
   developing a random forest machine learning‐based classifier for PRRSV
   detection is shown in Figure [84]1D. A training cohort (C1; n = 35) was
   established for initial sample acquisition and untargeted metabolomic
   and lipidomic profiling. The classifier's performance was subsequently
   assessed in an independent test cohort (C2; n = 8), followed by further
   validation in a validation cohort (C3; n = 12). A random forest
   algorithm of machine learning and principal component analysis (PCA)
   revealed a notable distinction among the tested groups
   (Figure [85]1E,F; Figure [86]S1A,C, Supporting Information). The
   efficiency of machine learning in analyzing untargeted metabolomic and
   lipidomic data was confirmed by generating the receiver operating
   characteristic (ROC) curves, the model reached an area under curve
   (AUC) of 1.00 both in the untargeted metabolic and lipidomic training
   set (Figure [87]S1B,D, Supporting Information). The significance of the
   predicted results was exceptionally high, with p‐values of 2.0e‐9 and
   3.2e‐9 for the untargeted metabolomic and lipidomic prediction models,
   respectively, as determined by the Chi‐square test. The utilization of
   this machine learning approach led us to identify biomarkers related to
   metabolites (Figure [88]1G) and lipids (Figure [89]1H) that showed a
   significant alteration in the serum of PRRSV‐infected pigs than
   uninfected controls. Moreover, the pathway enrichment analyses of
   differentially expressed serum biomarkers revealed their association
   with cellular pathways related, but not limited, to glucose and lipid
   metabolisms (Figure [90]S1E,F, Supporting Information). Overall, these
   analyses identify serum metabolic and lipid biomarkers that undergo
   differential alteration in a clinical cohort of PRRSV‐infected pigs.

Figure 1.

   Figure 1
   [91]Open in a new tab

   Machine learning‐based discrimination of PRRSV‐infected and uninfected
   pigs using serum metabolomic and lipidomic features of a clinical
   cohort of pigs. A,B) Clinical profile of each animal recruited into the
   study. The Y‐axis represents individual animals, labeled N1–N55 for
   PRRSV‐negative (A) and P1–P55 for PRRSV‐positive (B) animals. The
   X‐axis denotes the age of the animals in days. The color of each bar
   indicates the body temperature of the pigs. Gender is represented by a
   “+” (male) or “–” (female) symbol, while clinical symptom scores are
   denoted by numerical values to the right of each bar. C) Overview of
   metabolic and lipidomic workflow. Clinical serum samples from pigs were
   collected and assessed for PRRSV positivity using RT‐qPCR, followed by
   classification into PRRSV‐positive or negative groups (n = 55 per
   group). Subsequent analysis included untargeted metabolomic and
   lipidomic analyses of each serum sample with the implication of data
   for functional study and machine learning. D) Protocol for developing a
   machine learning‐based classifier for PRRSV detection. A training
   cohort (C1) was established for initial sample acquisition and
   untargeted metabolomic and lipidomic profiling. The classifier's
   performance was subsequently assessed in an independent test cohort
   (C2), followed by further validation in a validation cohort (C3). E,F)
   A random forest analysis in machine learning to globally analyze
   untargeted metabolomic (E) and lipidomic (F) data. G,H) Top 30
   candidate biomarker metabolites (G) or lipids (H) identified through
   machine learning.

2.2. PRRSV Infection Induces a Host Reprogramming of Glucose and Lipid
Metabolism

   To further elucidate the differential serum metabolites and lipids in
   clinical cohort samples, we then classified enriched pathways related
   to untargeted metabolites and lipids, which were found to converge on
   lipid metabolism, glucose metabolism, carbon metabolism, and their
   corresponding biological functions (Figure [92] 2A,B; Table [93]S1,
   Supporting Information). To corroborate the clinical cohort findings,
   we employed a PRRSV challenge animal model, categorized into infected
   and uninfected groups. Serum samples were collected at 3, 7, 10, and 21
   days post‐infection (dpi) for untargeted metabolomic and lipidomic
   analyses using mass spectrometry (Figure [94]2C). Concurrently, ongoing
   monitoring of body temperature, body weight, clinical symptoms, and
   nasal viral shedding in the infected and uninfected groups revealed
   significant disparities. The infected animals exhibited pronounced
   clinical symptoms, including high fever and weight loss
   (Figure [95]2D,E; Figure [96]S2A, Supporting Information) together with
   high viral loads in nasal swabs (Figure [97]S2B, Supporting
   Information).

Figure 2.

   Figure 2
   [98]Open in a new tab

   Metabolic and lipidomic analysis in animal virus‐challenged experiments
   reflecting similarities to data in a clinical cohort. A,B)
   Classification of KEGG pathways related to metabolites (A) and lipids
   (B) detected in untargeted metabolomic (A) and lipidomic (B) data from
   the clinical cohort. C) Overview of study design for PRRSV challenge in
   a piglet model (n = 5 for each group). D,E) Changes in piglet body
   weight (D) and temperature (F) post‐infection. In D, the initial weight
   at 0 dpi was set as 100%, and subsequent changes were calculated
   relative to this value. F,G) Venn diagrams illustrate an overlap of
   differentially expressed metabolites (F) and lipids (G) between
   virus‐challenged animal experiment and clinical cohort, calculated and
   generated using online tool Jvenn.^[ [99]^48 ^] H) Composite diagram
   highlighting shared differentially expressed metabolites and lipids in
   both animal experiment and clinical cohort. Expression values are
   depicted in terms of Log2 fold‐change. The integrated figure was
   plotted using Adobe Illustrator.

   Upon analyzing the metabolomic data, PCA result demonstrated robust
   inter‐group distinctiveness and inner‐group reproducibility (Figure
   [100]S2C,D, Supporting Information). In addition, a statistical
   enumeration of differentially expressed metabolites and lipids was
   conducted (Figure [101]S2E–J, Supporting Information) and compared with
   the findings from the clinical cohort which showed a small overlap for
   both metabolites and lipids (Figure [102]2F,G). Under controlled PRSSV
   challenge settings, infected animals showed obvious differences in the
   levels of metabolites and lipids when compared to the animals belonging
   to the uninfected group (Figure [103]S2G,H, Supporting Information).
   Notably, the smallest number of differentially expressed metabolites
   was observed at 3 dpi, whose numbers increased markedly over time
   (Figure [104]S2E,F, Supporting Information). The number of differential
   metabolites showing an intersection at different infection time points
   is negligible (Figure [105]S2I, Supporting Information); however, that
   intersection proportion increased significantly for the differentially
   expressed lipids (Figure [106]S2J, Supporting Information).
   Furthermore, a temporal cluster enrichment analysis was performed to
   examine the variation in the expression of metabolites and lipids over
   time (Figure [107]S3, Supporting Information). Both types of molecules
   were clustered into eight distinct clusters by considering their
   expression tendencies along with the time. The metabolites (Figure
   [108]S3A, Supporting Information) and lipids (Figure [109]S3B,
   Supporting Information) that clustered together are more likely to be
   classified into the same functional set of molecules. The roles of
   differentially expressed metabolites and lipids were further elucidated
   through functional enrichment analysis (Figure [110]S2K,L, Supporting
   Information), highlighting their respective significance. Presentation
   of the common differentially expressed metabolites and lipids shared
   between the clinical cohort and the challenge model, including pathway
   annotations and relative expression levels, revealed a homology of
   these metabolites in lipid droplet metabolism, glucose metabolism, and
   amino acid metabolism (Figure [111]2H), aligning with the previously
   observed enrichment results in the clinical cohort.

2.3. LPA and IFN‐β Exhibit a Strong Inverse Correlation in Their Regulation

   Given that the primary target cells of PRRSV are the porcine alveolar
   macrophages (PAMs), we sought whether there is a concordance in the
   metabolic reprogramming outcomes between the target cells and the
   serum. To address this, we employed a PRRSV infection animal model,
   followed by utilizing PAMs isolated from porcine lungs for RNA
   sequencing (Figure [112] 3A). PCA result of sequencing data confirmed
   consistent reproducibility within groups and obvious distinctions
   between infected and non‐infected groups (Figure [113]S4A, Supporting
   Information). The differentially expressed genes (DEGs) were
   subsequently identified and illustrated in a volcano plot wherein the
   majority of DEGs showed an upregulated expression pattern (Figure
   [114]S4B,C, Supporting Information). Pathway enrichment analyses
   revealed metabolism as one of the highly enriched pathways related to
   DEGs (Figure [115]S4D, Table [116]S3, Supporting Information). A
   further characterization of the highly enriched metabolism pathway
   indicated its relevance primarily to lipid, amino acid, glycan, and
   glucose metabolism (Figure [117]3B). Consistently, the gene set
   enrichment analysis (GSEA) also suggested the DEGs enrichment in lipid
   and amino acid metabolism (Figure [118]3C,D), mirroring the serum
   enrichment results. We next assessed the cytokine profile in the serum
   of the PRRSV‐challenged animal model at multiple infection time points
   (Figure [119]3E) and evaluated the correlation between cytokines and
   metabolites or lipids. The cytokines belonging to anti‐inflammatory,
   pro‐inflammatory, and anti‐viral categories indicated notable changes
   in their expression levels upon infection (Figure [120]3F). Among them,
   interferon beta (IFN‐β) exhibited the most significant downregulation
   at all infection time points (Figure [121]3F), which was further
   confirmed to be consistently lowered in the serum of infected animals
   individually compared to non‐infected animals (Figure [122]3G).
   Correlation analysis of cytokines with metabolites (Figure [123]3H) or
   lipids (Figure [124]3I) indicated the strongest negative correlation of
   IFN‐β with lysophosphatidic acid (LPA) (Figure [125]3J). Subsequently,
   the ROC curve analysis of a clinical pig cohort data by machine
   learning unveiled LPA as a potential clinical serum biomarker
   (Figure [126]3K), exhibiting a significant upregulation in the infected
   group (Figure [127]3L). LPA, as well as phosphatidylinositol (PI), were
   also implicated as the highest number of pathways related to lipid
   metabolism (Figure [128]S4E, Supporting Information), which highlights
   a multifaceted role of LPA in the PRRSV pathogenesis. The elevated
   levels of serum LPA were also assessed by recruiting an additional
   cohort of PRRSV‐infected subjects (n = 200) which revealed a
   significant upregulation in LPA content in the serum of PRRSV‐positive
   subjects compared to the negative animals (Figure [129]3M). However, no
   significant differences were observed in the serum of clinical cohorts
   of African swine fever virus (ASFV)‐ and porcine epidemic diarrhea
   virus (PEDV)‐ positive subjects (Figure [130]3N,O).

Figure 3.

   Figure 3
   [131]Open in a new tab

   Correlation analysis reveals LPA as a potential biomarker and immune
   modulator during PRRSV infection of piglets. A) Schematic of the study
   design for RNA‐sequencing (n = 3 per group). B) KEGG pathway analysis
   of differentially expressed genes (DEGs) in metabolic processes.
   Diagram was generated by utilizing Proteomaps.^[ [132]^49 ^] C,D) Gene
   set enrichment analysis (GSEA) of RNA‐sequencing data highlighting the
   significance of lipid (C) and amino acid (D) metabolism during PRRSV
   infection. The analysis was performed with GSEA database.^[ [133]^50 ^]
   E) Schematic of cytokine detection in serum samples obtained from
   animal virus‐challenged experiments at 3, 7, 10, and 21 days
   post‐infection (dpi). F) Comparative fold change of cytokines showing
   differential expression in infected animals compared to mock‐infected
   animals. G) Measurement of IFN‐β protein levels in the serum samples of
   PRRSV‐infected and mock‐infected individual animals. H,I) Heatmaps
   showing correlations between cytokines and metabolites (H) or lipids
   (I). J) Targeted correlation analysis between IFN‐β and both
   metabolites and lipids. K) ROC curve analysis of LPA as identified by
   machine learning. L) Relative expression levels of LPA in a in
   lipidomic data of the clinical pig cohort (n = 110) related to
   Figure [134]1. M–O) LPA concentrations in the serum of PRRSV‐infected
   (M) pigs compared to ASFV‐ (N) or PEDV‐infected (O) pigs recruited in
   expanded clinical cohorts. Significant comparisons between two groups
   are made using Mann‐Whitney test.

   Taken together, these data indicate that PRRSV infection increases the
   serum level of LPA which negatively correlates with the cytokines
   production in vivo.

2.4. LPA Promotes PRRSV Replication by Suppressing IFN‐β Production In Vitro

   Considering that LPA and IFN‐β show significantly altered expressions
   upon PRRSV infection in vivo and their expressions are inversely
   related to each other, we further investigated their impact on PRRSV
   replication using an in vitro culture system. To this end, immortalized
   porcine alveolar macrophage (iPAM) cells were infected with PRRSV at a
   multiplicity of infection (MOI) of 0.1, and the production of LPA was
   determined using ELISA. PRRSV infection of cultured cells led to
   markedly increased levels of LPA in a time‐dependent manner when
   compared to non‐infected cells (Figure [135] 4A). We next examine the
   effect of LPA supplementation and its inhibition on IFN‐β production
   and virus replication. It was found that the treatment of infected
   cells with non‐cytotoxic concentrations of LPA supplement
   (Figure [136]4B,C; Figure[137]S5A, Supporting Information) resulted in
   a significantly reduced production of IFN‐β at both RNA and protein
   levels and enhanced virus replication in an LPA supplement
   dose‐dependent manner as assessed by quantifying viral protein
   (Figure [138]4D,E), genome copies (Figure [139]4F), and titers
   (Figure [140]4G). However, the non‐cytotoxic concentrations of LPA
   inhibitor HA130, inhibiting LPA production in the culture supernatants
   effectively (Figure [141]S5B,C, Supporting Information), caused an
   opposite effect on both IFN‐β levels (Figure [142]4B,C) and virus
   replication (Figure [143]4H–K) in treated cells. Similar effects on
   IFN‐β levels and virus replication were observed when a single
   non‐cytotoxic concentration of LPA supplement and inhibitor HA130 were
   tested in an infection time‐dependent manner (Figure [144]4L–Q; Figure
   [145]S5D–F, Supporting Information). Furthermore, the effect of LPA
   supplementation and inhibition on pro‐inflammatory, anti‐inflammatory,
   and antiviral cytokines was evaluated. None of the cytokines tested,
   except IL‐1β, at both RNA and protein levels, demonstrated a strong
   correlation with the LPA under infection conditions (Figure [146]S6,
   Supporting Information). Altogether, these findings indicate that LPA
   enhances PRRSV replication by dampening the host IFN‐β response in
   infected cells.

Figure 4.

   Figure 4
   [147]Open in a new tab

   LPA promotes PRRSV replication by suppressing the host IFN response. A)
   Confirmation of LPA upregulation in PRRSV‐infected iPAMs cell culture
   supernatants. Cells were infected with PRRSV at an MOI of 0.1, followed
   by LPA detection by ELISA in the supernatants at indicated time‐points
   post‐infection. B,C) Alteration of IFN‐β mRNA (B) and protein (C)
   expressions in PRRSV‐infected iPAM cells upon supplementation or
   inhibition of LPA. Cells were treated with LPA supplement or LPA
   inhibitor HA130 at indicated concentration for a period of 1 h,
   following infection with PRRSV at indicated MOI. At 24 h
   post‐infection, IFN‐β mRNA and protein expression were detected by
   RT‐qPCR and ELISA, respectively. D–G) Exogenous LPA increases PRRSV
   proliferation in iPAM cells evaluated at the level of viral protein (D
   and E), viral copy number (F), and viral titers (G) in a dose‐dependent
   manner. H–K) Inhibition of LPA reduces PRRSV proliferation in iPAM
   cells evaluated at the level of viral protein (H and I), viral copy
   number (J), and viral titers (K) in a dose‐dependent manner. L–Q)
   Exogenous LPA enhances PRRSV proliferation in iPAM cells evaluated at
   the level of viral protein (L and M), viral copies (P), and viral
   titers (Q) in a time‐dependent manner. LPA inhibition reduces PRRSV
   replication in iPAM cells evaluated at the level of viral protein (N
   and O), viral copy number (P), and viral titers (Q) in a time‐dependent
   manner. In D–Q, cells were treated with indicated concentration of
   exogenous LPA or HA130 for a period of 1 h, followed by PRRSV infection
   at indicated MOI and time. The biological experiments were conducted in
   triplicate. Data are represented as mean ± SEM. Significant comparisons
   between two groups are made using Mann‐Whitney test, significant
   comparisons among three or more groups are made by one‐way ANOVA, with
   significance levels indicated as ns (not significant), * (p < 0.05), **
   (p < 0.01), and *** (p < 0.001).

2.5. LPA Synthesis is Promoted by the Upregulation of ENPP2 (ATX) Following
PRRSV Infection

   To elucidate the underlying mechanisms of PRRSV‐mediated upregulation
   of LPA, we analyzed metabolomic data from the PRRSV‐challenged in vivo
   model pertaining to metabolites associated with LPA decomposition and
   synthesis (Figure [148] 5A). Concurrently, we conducted the RT‐qPCR
   analysis of genes related to LPA decomposition and synthesis. In
   addition to several genes showing a significantly perturbed expression,
   the levels of ectonucleotide pyrophosphatase 2 (ENPP2, also called
   autotaxin or ATX), one of the genes involved in LPA synthesis, showed a
   significant increase at all three infection time points following PRRSV
   infection (Figure [149]5B). Protein‐level quantification corroborated
   these findings, demonstrating consistency with the mRNA expression
   levels (Figure [150]5C–E). Furthermore, the treatment of cells with an
   ATX inhibitor (HA155) resulted in a significant reduction in LPA levels
   in the cell supernatants, concomitant with the restoration of IFN‐β
   levels and pronounced inhibition of PRRSV replication
   (Figure [151]5F–K). Similar outcomes were observed when ATX expression
   levels were knocked down using the ATX‐specific siRNAs
   (Figure [152]5L–R). In contrast, the overexpression of ATX exhibited a
   modest promotion of LPA levels in the cell supernatant, albeit
   insignificantly, and demonstrated no significant impact on IFN‐β level
   and viral replication (Figure [153]S7, Supporting Information). Thus,
   these data demonstrate that PRRSV augments LPA synthesis via
   upregulating the ATX production in infected cells.

Figure 5.

   Figure 5
   [154]Open in a new tab

   PRRSV promotes LPA synthesis via upregulation of ENPP2 (ATX). A)
   Heatmap showing the expression levels of lipids, associated with LPA
   synthesis and degradation, in animal virus‐challenged experiment data.
   B) RT‐qPCR analysis of mRNA expression of genes related to LPA
   synthesis and degradation in PRRSV‐infected iPAM cells. C–E)
   Confirmation of ATX upregulation at protein level following PRRSV
   infection in iPAM cells. F–K) HA155‐mediated inhibition of ATX
   decreases LPA synthesis, alleviates IFN‐β suppression and reduces the
   PRRSV proliferation in iPAM cells. L–R) siRNA‐mediated inhibition of
   ATX decreases LPA synthesis, alleviates IFN‐β suppression and reduces
   the PRRSV proliferation in iPAM cells. In F–R, cultured iPAM cells were
   subjected to treatment with HA155 or siATX for either 1 hour or 24
   hours, along with their respective controls, followed by 0.1 MOI
   infection of PRRSV for a period of 24 h. LPA and IFN‐β levels were
   determined by ELISA. The viral replication at the level of protein,
   genome, and titers was determined by Western blot, RT‐qPCR, and plaque
   assay, respectively. The biological experiments were conducted in
   triplicate. Data are represented as mean ± SEM. Significant comparisons
   between two groups are made using Mann‐Whitney test, significant
   comparisons among three or more groups are made by one‐way ANOVA.

2.6. LPA Inhibits IFN‐β Response by Suppressing RIG‐I Expression Following
PRRSV Infection

   It is reported that LPA regulates interferon regulatory factor 3 (IRF3)
   via the LPA1 receptor, thereby inhibiting the levels of IFN‐β.^[
   [155]^25 ^] To confirm whether this phenomenon also occurs in the case
   of PRRSV infection, iPAM cells, treated with or without an LPA1
   inhibitor (Ki16425), were subjected to PRRSV infection followed by
   quantification of IFN‐β levels. Inhibition of LPA1 caused no effect on
   IFN‐β production at both mRNA and protein levels (Figure [156] 6A,B)
   without altering cell viability (Figure [157]S8A, Supporting
   Information) and virus replication (Figure [158]S8B,C, Supporting
   Information). To elucidate the mechanism by which LPA negatively
   regulates IFN‐β response under PRRSV infection, the LPA inhibitor HA130
   treated or untreated iPAM cells were infected with PRRSV, and protein
   samples were harvested at 24 hpi for proteomic analysis
   (Figure [159]6C). PCA analysis of proteomics data showed notable
   distinctions between the groups and similarities within the groups
   (Figure [160]6D). Differentially expressed protein (DEPs) analysis,
   shown as volcano plots, revealed the perturbation of several cellular
   proteins dependent on tested experimental conditions (Figure [161]6E;
   Figure [162]S8D–F, Supporting Information), wherein the HA130 treated
   and infected cells predominantly showed a differential downregulated
   DEPs (Figure [163]6E,F). The pathway enrichment analysis of DEPs in
   HA130‐treated and infected cells indicated their enrichment primarily
   in metabolism pathways (Figure [164]6G). Further analysis of the IFN
   pathway revealed indicated the marked downregulation of retenoic
   inducible gene‐I (RIG‐I) and myeloid differentiation primary response
   protein 88 (MyD88) levels upon PRRSV infection, whose expression
   rescued in HA130‐treated cells (Figure [165]6H), indicating that the
   expression of RIG‐I and MyD88 is regulated by LPA. This phenotype was
   confirmed in vitro where the exogenous treatment of LPA suppressed
   RIG‐I expression at both RNA and protein levels in cultured cells,
   whereas LPA inhibition had the opposite effect (Figure [166]6I–L).
   Similar results were observed for the MyD88, though the regulatory
   effect of LPA on MyD88 was not as pronounced as on RIG‐I (Figure
   [167]S8G–I, Supporting Information). Subsequently, silencing RIG‐I
   expression or modulating LPA at the cellular level resulted in no
   significant regulatory effect on IFN‐β (Figure [168]6M–O), suggesting
   that LPA regulates IFN‐β by suppressing RIG‐I expression. The mechanism
   by which LPA suppresses RIG‐I at the mRNA and protein levels remains
   unclear. Interferon regulatory factor 1 (IRF1) is known to activate the
   RIG‐I promoter, thus regulating RIG‐I transcription.^[ [169]^26 ^] To
   investigate this, we isolated and analyzed proteins from the nucleus
   and cytoplasm. Our results showed that PRRSV infection prevents IRF1 to
   translocate from the cytoplasm to the nucleus. LPA exacerbates this
   effect, whereas inhibition of LPA production restores IRF1's nuclear
   translocation (Figure [170]S8J–L, Supporting Information).
   Collectively, these findings suggest that LPA probably suppresses RIG‐I
   expression by blocking IRF1 nuclear translocation, thereby regulating
   the IFN‐β response.

Figure 6.

   Figure 6
   [171]Open in a new tab

   LPA suppresses IFN‐β response by dampening the RIG‐I expression during
   PRRSV infection. A,B) LPAR1 inhibition in iPAM cells does not affect
   IFN‐β suppression during PRRSV infection. C–H) Proteomic analysis of
   IFN‐β inhibition by PRRSV in iPAM cells: iPAMs treatment and collection
   for proteomic sequencing (C); PCA analysis showing strong internal
   group replication (D); identification of differentially expressed
   proteins (DEPs) (E); heatmap of DEPs in PRRSV+HA130 group relative to
   PRRSV group (F); KEGG functional enrichment analysis of DEPs (G);
   expression levels of proteins associated with IFN generation (H). I–L)
   LPA‐mediated regulation of RIG‐I expression. Cultured iPAM cells were
   treated with indicated concentrations of LPA or HA130, followed by
   PRRSV infection at an MOI of 0.1 for 24 h. The protein and mRNA levels
   of viral N‐protein and RIG‐I were determined by Western blot and
   RT‐qPCR. M–O) Effect of RIG‐I inhibition on IFN‐β production. Cultured
   iPAM cells, treated with siRIG‐I, LPA, HA130, or their controls for
   1 h, were subjected to PRRSV infection at MOI of 0.1 for 24 h. The
   effect of siRNAs on RIG‐I protein expression was assessed by Western
   blot. The proteins and mRNA levels of IFN‐β were quantified by ELISA
   and RT‐qPCR, respectively. The biological experiments were conducted in
   triplicate. Data are represented as mean ± SEM. Significant comparisons
   between two groups are made using Mann‐Whitney test, significant
   comparisons among three or more groups are made by one‐way ANOVA, with
   significance levels indicated as ns (not significant).

2.7. Targeting LPA Exhibits Promising Therapeutic Efficacy Against PRRSV
Infection

   To validate whether targeting LPA can suppress PRRSV replication and
   ameliorate clinical symptoms, we determined the activity of LPA
   inhibitor HA130 in an in vivo PRRSV‐challenged model system. All
   animals (n = 5 per group) were classified into the following groups:
   blank (control group), HA130 (treatment group), PRRSV (infection
   group), and PRRSV+HA130 (infection‐treatment group) (Figure [172] 7A).
   Clinical assessments, including clinical scores, body temperature, and
   weight were conducted. The average body temperature of piglets in PRRSV
   infection group markedly increased from 38.7 °C at 0 dpi to 41.2 °C at
   3 dpi and remained above 40 °C until the observation endpoint. However,
   treatment with HA130 resulted in a decrease in body temperature to
   below 40 °C after 7 dpi (Figure [173]7B). HA130 treatment also
   significantly promoted body weight gain and mitigated clinical symptoms
   induced by PRRSV infection (Figure [174]7C,D). These findings indicated
   that the treatment with HA130 effectively alleviated the severe fever,
   weight loss, and clinical disease symptoms in piglets induced by PRRSV
   infection (Figure [175]7B–D), which correlated with the limited virus
   production in the upper respiratory tract as assessed by quantifying
   virus RNAs in nasal swab samples (Figure [176]7E). Gross and
   histopathological analysis of animal tissues, including lungs, spleen,
   and inguinal lymph nodes, demonstrated a significant amelioration of
   PRRSV‐mediated pathological changes and apoptosis upon treatment of
   animals with HA130 (Figure [177]7F; Figure [178]S9, Supporting
   Information). Moreover, the treatment of HA130 significantly decreased
   the levels of LPA and its various subtypes in the serum of treated
   animals, while markedly restoring the serum IFN‐β levels
   (Figure [179]7G–I). The regulatory effect of LPA inhibition on the
   other cytokines remained modest: interleukin 8 (IL‐8) and IL‐4 showing
   an upregulation, whereas IFN‐α, IL‐1β, IL‐10, IL‐6, and tumor necrosis
   factor alpha (TNF‐α) demonstrated a trend of downregulation. (Figure
   [180]S10, Supporting Information). We finally examined if the observed
   outcomes are dependent on ATX and RIG‐I expression in lung tissues.
   Indeed, PRRSV infection significantly enhanced ATX, but reduced RIG‐I,
   proteins abundance in lung tissues. In contrast, the HA130 treatment of
   animals led to the rescue of the inhibitory effect of PRRSV on RIG‐I,
   which ultimately caused a significant suppression of viral loads in the
   lungs, spleen, and inguinal lymph nodes (Figure [181]7J–L). Overall,
   these data suggest that the inhibition of LPA during PRRSV infection
   exerts an antiviral effect which results in the amelioration of disease
   outcomes in infected piglets.

Figure 7.

   Figure 7
   [182]Open in a new tab

   Therapeutic efficacy of HA130 in limiting PRRSV replication and
   mitigating disease pathogenesis in piglets. A) Overview of experimental
   design indicating the HA130 treatment and PRRSV infection of piglets.
   B–E) Effect of HA130 treatment on body temperature (B), body weight
   (C), clinical symptoms (D), and viral replication (E) in PRRSV‐infected
   piglets. F) Effect of HA130 treatment on morphology, histology, and
   apoptosis of lung tissues of infected animals. Histological and
   apoptotic features were visualized by H&E staining and TUNEL assay,
   respectively. G–H) ELISA‐based measurement of serum LPA (G) and IFN‐β
   (H) concentrations at 5, 7, 9, and 14 days post‐infection (dpi). I)
   Lipidomic sequencing for serum LPA quantification at 5, 7, 9, and 14
   dpi. J–K) Western blot analysis of ATX and RIG‐I expression in lung
   tissues at 14 dpi. L) RT‐qPCR assessment of viral loads in lungs,
   spleens, and inguinal lymph nodes at 14 dpi. Data are represented as
   mean ± SEM. Significant comparisons between two groups are made using
   Mann‐Whitney test, significant comparisons among three or more groups
   are made by one‐way ANOVA, with significance levels indicated as ns
   (not significant).

3. Discussion

   We analyzed metabolomic and lipidomic profiles of clinical cohorts of
   PRRSV‐infected pigs with a machine learning approach and identified LPA
   as a serum diagnostic biomarker for PRRSV infection. Mechanistically,
   PRRSV augmented LPA synthesis by upregulating the ATX expression which
   led to RIG‐I and IFN‐β suppression and consequently increased virus
   replication in vitro and in vivo. Targeting LPA proved effective in
   alleviating clinical symptoms induced by PRRSV, reducing viral loads in
   various organs, and mitigating viral shedding (Figure [183] 8 ).

Figure 8.

   Figure 8
   [184]Open in a new tab

   Graphical depiction of the study design and the underlying mechanism by
   which PPRSV elevates LPA levels to downregulate RIG‐I, thereby
   inhibiting the production of IFN‐β and promoting the proliferation of
   PRRSV.

   With the emergence of high‐throughput technologies, a novel approach to
   investigate viral diseases and facilitate diagnosis involves multiomic
   assessment of clinical cohorts for virus‐specific diagnostic markers.
   Shen et al. have suggested various distinctive molecular alterations in
   the serum of SARS‐CoV‐2‐infected patients, which resulted in the
   identification of several specific biomarkers. Such approaches hold
   promise for predicting the progression from mild to severe
   conditions.^[ [185]^27 ^] Similarly, the use of machine learning has
   yielded a remarkable 94% accuracy in identifying critical biomarkers
   related to long COVID‐19.^[ [186]^28 ^] Another recent study has also
   uncovered 22 long COVID‐19 protein markers, providing valuable insights
   for the early identification of high‐risk individuals.^[ [187]^29 ^] In
   this study, we pioneered the use of metabolomics and lipidomics with
   machine learning and discovered serum metabolites and lipids as
   biomarkers that are differential and distinct to PRRSV infection. Of
   them, we confirmed that LPA serves as a decisive biomarker of PRRSV
   infection, thus presenting implications for the clinical diagnosis of
   PRRSV infections and providing guidance to discover biomarkers for
   other viral infections.

   An aspect of the host‐virus interaction is the tendency of viruses to
   reprogram the metabolic activity of the host cells, providing
   opportunities to develop novel antiviral drugs.^[ [188]^17 , [189]^22 ,
   [190]^30 , [191]^31 ^] In the context of PRRSV infection, studies
   suggest that PRRSV hijacks tryptophan‐polyamine metabolism to promote
   viral proliferation,^[ [192]^32 ^] which is in agreement with the
   pathway enrichment data derived from our in vivo infection model. PRRSV
   activates the AMP activated protein kinase (AMPK)‐ acetyl‐CoA
   carboxylase 1(ACC1) signaling pathway to regulate fatty acid synthesis
   in cultured cells, which is unfavorable for PRRSV proliferation.^[
   [193]^33 ^] However, no significant changes in fatty acid metabolism
   were observed in our study involving clinical pig cohorts and in vivo
   infection models. Such discrepancies may occur due to the fact that
   PRRSV‐infected cells may not accurately reflect the metabolic
   alterations in the infected pig's body in clinical settings. To the
   best of our knowledge, our study presented for the first time a
   comprehensive metabolite and lipid profile changes within
   PRRSV‐infected pigs. These findings may provide novel insights into
   PRRSV replication biology and could lead to the development of
   potentially new antiviral approaches.

   LPA is characterized as a small and structurally simple biologically
   active phospholipid. LPA is involved in various biological processes,
   including the regulation of wound healing, blood clotting, blood
   pressure maintenance, cell survival, proliferation, differentiation,
   migration, morphology, and adhesion.^[ [194]^34 , [195]^35 , [196]^36
   ^] A recent study has indicated that the binding of LPA to one of its
   receptors LPAR1 induces Rho‐associated coiled‐coil‐forming protkinase
   1/2 (ROCK1/2), leading to the inactivation of IRF3 and dampening of IFN
   responses against viruses like Zika virus and SARS‐CoV‐2.^[ [197]^25 ^]
   However, the role of LPA in PRRSV infection remained unknown so far. In
   this study, we elucidated that PRRSV reprogrammed host metabolites and
   lipids by ATX‐mediated upregulation of LPA level which subsequently
   induces a proviral effect. Targeting LPA effectively inhibited viral
   replication at both in vitro and in vivo levels. The LPA‐mediated
   proviral effect was independent of LPAR1, indicating the engagement of
   other upstream mechanisms. One possible mechanism was found through
   increased ATX expression; however, the precise mechanism by which PRRSV
   modulates the ATX‐LPA axis remained unprobed in this study and requires
   further investigations. Since several genotypes of PRRSV contribute to
   pig mortality in the natural environment, additional studies are needed
   to determine whether targeting LPA has a universal effect on all PRRSV
   genotypes.

   PRRSV exerts significant immunosuppressive effects in pigs. During
   PRRSV infection, the virus undergoes prolonged replication within the
   host without triggering pronounced innate immune responses,
   particularly the type 1 IFN response. To date, most studies have
   predominantly focused on the suppression of the IFN pathway from the
   perspective of PRRSV infection in general or viral protein
   specifically. For instance, PRRSV infection can suppress the IFN‐β
   production by interfering with the mitochondrial antiviral signaling
   protein (MAVS) activation in the RIG‐I signaling pathway.^[ [198]^37 ^]
   PRRSV non‐structural protein 4 (Nsp4) and Nsp11 proteins selectively
   target RIG‐I and MAVS to inhibit type I IFN signaling cascade.^[
   [199]^38 , [200]^39 ^] However, the regulation of IFN or other innate
   immune pathways by the host metabolites or lipids is unknown. We
   unveiled that a phospholipid LPA, increased upon PRRSV infection,
   negatively regulates the IFN‐β production via suppressing the RIG‐I.
   Our findings shed light on a novel immunosuppressive mechanism employed
   by PRRSV, guiding further exploration into the molecular intricacies of
   PRRSV replication.

   In summary, by utilizing a machine learning approach and integrating
   clinical pig cohorts with in vivo and in vitro infection models, we
   identified serum biomarkers of metabolic and lipid origins that are
   capable of distinguishing PRRSV infection. This led to unveil LPA as an
   inducer of PRRSV replication and associated disease pathogenesis which
   upon targeting showed a promising therapeutic effect. This work could
   serve as a pivotal reference for the clinical diagnosis of PRRSV
   infection and the establishment of novel therapy to treat PRRSV
   infections.

4. Experimental Section

Detection of Untargeted Metabolome and Lipidome

   Metabolomic detection was performed using the Q Exactive HF
   high‐resolution mass spectrometer at BGI Genomics (Shenzhen, China),
   employing liquid chromatography‐tandem mass spectrometry (LC‐MS/MS).
   Data were collected in both positive and negative ion modes to enhance
   metabolite coverage. Untargeted metabolomic data were processed with
   Compound Discoverer 3.1, including peak extraction, alignment, and
   compound identification. Lipidomic data were analyzed using LipidSearch
   4.1 for intelligent peak extraction, lipid identification, and peak
   alignment. Compound identification utilized multiple databases,
   including the BGI Library (a proprietary standard product library),
   mzCloud, and ChemSpider (encompassing HMDB, KEGG, LipidMaps analyses).

Data Preprocessing and Quality Control

   Results from Compound Discoverer 3.1 and LipidSearch 4.1 were imported
   into metaX for preprocessing, which involved: 1) Data normalization
   using the Probabilistic Quotient Normalization (PQN) method to obtain
   relative peak areas; 2) Batch effect correction with the QC‐RLSC
   (Quality Control‐based Robust LOESS Signal Correction); 3) Removal of
   compounds with a Coefficient of Variation (CV) over 30% in all QC
   samples. Data quality was evaluated through the reproducibility of
   quality control (QC) samples, including chromatogram overlap, PCA, peak
   numbers, and peak response intensity differences.

Differential Metabolites and Lipids Screening

   Differential metabolites and lipids were screened using the first two
   principal components' VIP values from the PLS‐DA (Partial Least Squares
   Method‐Discriminant Analysis) model, combined with univariate analysis
   of fold‐change and Student's t‐test. Criteria for differential
   metabolites and lipids were: VIP ≥ 1 in the first two PLS‐DA
   components, Fold‐Change ≥ 1.2 or ≤ 0.83, and q‐value < 0.05.

Functional Enrichment Analysis of Differential Metabolites and Lipids

   Differential metabolites and lipids were subjected to functional
   identification and pathway enrichment analysis using the KEGG
   database^[ [201]^40 ^] and MetaboAnalyst website
   ([202]https://www.metaboanalyst.ca).^[ [203]^41 ^] The information of
   software and databases used in functional analysis were listed in Table
   [204]S5 (Supporting Information).

Correlation Analysis Between Differential Metabolites/Lipids and Cytokine
Levels

   The correlation between differential metabolites/lipid molecules and
   cytokine expression changes was calculated using the Spearman
   correlation coefficient in the R package corrplot.^[ [205]^42 ^] The
   information of software used for correlation analysis were listed in
   Table [206]S5 (Supporting Information).

Cell Culture

   MARC‐145 cells were cultured in Dulbecco's modified Eagle's medium
   (DMEM) supplemented with 10% of fetal bovine serum (FBS) and 1%
   penicillin‒streptomycin (PS). iPAM cells were cultured in RPMI 1640
   medium supplemented with 10% FBS and 1% PS. All cells were incubated at
   37 °C with 5% of CO[2] in an incubator. The information of regents used
   for cell culture could be found in Table [207]S5 (Supporting
   Information).

Propagation and Measurement of PRRSV

   MARC‐145 cells were grown in T175 cell culture flasks to full
   confluency. Before infection, cultured cells were washed thrice with 1×
   PBS gently. Subsequently, cells were infected with highly pathogenic
   PRRSV (HP‐PRRSV) Li11 strain,^[ [208]^43 ^] diluted in an infection
   medium (DMEM supplemented with 2% FBS), for a period of 1 h at 37 °C
   with 5% CO[2]. Post‐viral incubation, the cells were washed with 1× PBS
   and maintained in a fresh infection medium. Upon observing significant
   cytopathy, cells were subjected to three freeze‐thaw cycles at −80 °C
   and centrifuged at high speed (10,000 rpm) at 4 °C for 10 min to
   collect the supernatant, which was then stored at −80 °C. A 50% tissue
   culture infectious dose (TCID[50]) of PRRSV was determined by seeding
   MARC‐145 cells in a 96‐well plate at a density of 1×10^4 cells/well,
   incubating them until they become fully confluent, and then treating
   them with PRRSV (serially diluted in the infection medium) at 37 °C and
   5% CO[2]. After 1 h of virus absorption, cells were washed with 1× PBS
   and continued cultivation with a maintenance medium (DMEM supplemented
   with 2% FBS) at 37 °C with 5% CO[2] for 96–144 h. Cell cytopathy for
   each dilution was recorded daily, and the TCID[50] of PRRSV was
   calculated using the Reed‐Muench method. The information of regents
   used for PRRSV propagation were listed in Table [209]S5 (Supporting
   Information).

CCK‐8 Assay for Drug Cytotoxicity

   iPAMs were seeded in a 96‐well plate at a density of 1×10^4 cells/well
   and cultured at 37 °C with 5% CO[2] until they reach full confluency.
   Different concentrations of LPA (5 µM to 500 µM) were prepared in RPMI
   1640 medium and added to the PBS‐washed cells, followed by incubation
   at 37 °C for 48 hours. Later, the cell supernatants were replaced with
   a CCK‐8 solution, and the cells were incubated for an additional 1 h.
   Absorbance was measured at 450 nm by using the ELISA microplate reader
   (BioTek, USA), and cell viability was calculated. This procedure was
   repeated independently three times for averaging. The cytotoxicity
   assessment of HA130 and Ki16425 followed a similar procedure, ensuring
   the DMSO concentration in the cell culture medium was kept below 0.1%.
   The information of regents used for CCK‐8 were listed in Table [210]S5
   (Supporting Information).

ELISA for LPA Concentration Measurement

   LPA concentrations in cell culture supernatants and animal serum
   samples were determined using the Signalway Antibody (SAB) LPA ELISA
   kit. LPA standards provided in the kit were dissolved, shaken
   vigorously, and serially diluted following the manufacturer's
   instructions. These standards, along with the cell culture supernatants
   or serum samples, were added to a 96‐well plate. After adding the
   HRP‐conjugate to each well and incubating at 37 °C for 1 h, the wells
   were washed, and a substrate solution was added. The reaction was
   stopped upon color development, and absorbance was measured at 450 nm
   by using the ELISA microplate reader (BioTeck, USA) to calculate the
   LPA concentration. The information of regents used for ELISA assays
   were listed in Table [211]S5 (Supporting Information).

Western Blot

   Cells were lysed on ice for 10 min with a radio immunoprecipitation
   assay (RIPA) lysis buffer, and the lysates were centrifuged at 12,000 g
   and 4 °C for 5 min to collect the supernatant. Protein concentration in
   each sample was measured using the BCA protein assay kit. Protein
   samples were prepared by mixing the supernatant with Omni‐Easy Protein
   Sample Loading Buffer and were subjected to boiling at 100 °C for
   3 min. SDS‐PAGE electrophoresis was performed, and proteins were
   transferred onto the PVDF membranes using a Mini Trans‐Blot Cell
   (BioRad, USA). Membranes were blocked at room temperature for 2 h in a
   blocking buffer containing 1% bovine serum albumin and 0.05% Tween‐20
   in 1× PBS, followed by incubation with indicated primary antibodies
   overnight at 4 °C. After washing thrice with a 0.05% Tween‐20 solution
   in 1× PBS, membranes were incubated with indicated secondary antibodies
   at room temperature for 1 h. Membranes were visualized using the
   enhanced chemiluminescence (ECL) reagent. The information of antibodies
   and regents used for western blot were listed in Table [212]S5
   (Supporting Information).

Indirect Immunofluorescence Assay (IFA)

   iPAMs were seeded in a 24‐well plate at a density of 5×10^4 cells/well,
   and cultured until became full confluency, followed by drug and virus
   treatment as indicated. Treated cells were fixed with paraformaldehyde
   for 15 min at room temperature. Then, cells were permeabilized with
   Triton X‐100 for 15 min at room temperature., Subsequently, cells were
   washed thrice with 1× PBS and were blocked in 1% BSA solution in 1× PBS
   for a period of 1 h at room temperature. Thereafter, cells were
   incubated with primary mouse anti‐PRRSV‐N protein antibody for 2 h at
   room temperature, washed thrice with 1× PBS, and incubated with
   FITC‐conjugated goat anti‐mouse secondary antibody for 1 h at room
   temperature. Nuclei were stained with DAPI for another 10 min at room
   temperature, and cells were visualized using a confocal fluorescence
   microscope (Nikon, Eclipse 80i). The information of antibodies and
   regents used for IFA were listed in Table [213]S5 (Supporting
   Information).

Isolation of Porcine Alveolar Macrophages (PAMs) from Bronchoalveolar Lavage
Fluid

   Animals were subjected to anesthetize with Zoletil and exsanguinate.
   Lungs were carefully exposed, and a cannula was inserted into the
   trachea. A sterile PBS was gently infused into the lungs, ensuring even
   distribution across the lung lobes. The PBS was then slowly aspirated
   back, collecting the bronchoalveolar lavage fluid which contains a
   mixture of lung cells and alveolar contents. The collected lavage fluid
   was then transferred to sterile centrifuge tubes and centrifuged at
   300× g for 10 min at 4 °C to pellet the cells. The supernatant was
   carefully aspirated and discarded by without disturbing the cell
   pellet. The pellet, primarily consisting of PAMs, was gently
   resuspended in an RPIM 1640 medium. This suspension of PAMs was further
   purified by layering over a density gradient medium and centrifuging at
   400× g for 20 min at room temperature. The layer containing the PAMs
   was carefully aspirated and transferred to a new tube, washed with the
   culture medium, and centrifuged under the same conditions to obtain a
   purified PAMs pellet. The final PAMs pellet was finally resuspended in
   a fresh cell culture medium suitable for subsequent experimental
   procedures. The whole process was performed under aseptic techniques to
   ensure cell viability and prevent contamination.

Hematoxylin and Eosin (H&E) Staining of Tissue Sections

   Tissue samples were preserved and fixed in a 10% formalin solution to
   maintain cellular morphology at 4 °C. Then tissues were trimmed, washed
   under running water for 30 min, and dehydrated in a graded ethanol
   series (70%, 80%, 90%, 95%, and 100%), each step lasting 40 min,
   followed by clearing in xylene. The tissues were then embedded in
   paraffin wax in a warming oven. Once infiltrated with wax, they were
   embedded, cooled, and solidified into blocks. These blocks were
   sectioned into thin slices of 5–8 µm using a microtome. The sections,
   if crinkled, were flattened in warm water and mounted on slides for
   drying at 45 °C. The sections were subjected to dewaxing in a gradient
   of xylene and rehydrating through a series of alcohols. Subsequently,
   sections were stained with hematoxylin, rinsed, differentiated in
   hydrochloric acid alcohol, blued in ammonia water, and stained with
   eosin before final washing. After clearing in ethanol and xylene, the
   slides were sealed with neutral gum for observation. The information of
   regents used in H&E were listed in Table [214]S5 (Supporting
   Information).

TUNEL Staining of Tissue Sections

   The process of embedding and dewaxing tissue sections was similar to
   that described for H&E staining. TUNEL Staining was performed by
   following the manufacturer's protocol. In brief, the sections were
   treated with 1% Triton X‐100 for permeabilization for 5 min at room
   temperature, followed by washing with PBS. They were then treated with
   Proteinase K from the TUNEL kit, washed, and dried. TdT enzyme reaction
   solution was applied to each sample, which was then incubated in a
   humidified, light‐protected box. After washing, Streptavidin‐TRITC was
   added and incubated in the dark, followed by PBS washes. The nuclei
   were counterstained with DAPI, washed with PBS thrice and visualized
   utilizing a slide scanner (ZEISS Axioscan 7, Germany). The information
   of antibodies and regents used for TUNEL assay were listed in Table
   [215]S5 (Supporting Information).

Plasmids and small interference RNAs (siRNAs)

   The pig ENPP2 and RIG‐I gene was amplified through reverse
   transcription PCR from total RNA extracted from iPAM cells using
   gene‐specific primers and subsequently cloned into a pUC57 vector
   containing a C‐terminal HA tag. The plasmid was sequenced to confirm
   that no errors were introduced during PCR amplification. Cultured iPAM
   cells were transfected with 1 µg of plasmids or 100 pmol of siRNAs
   (small interference RNAs) using jetPRIME according to the
   manufacturer's instructions. The information of regents and
   oligonucleotides used for over‐expression and siRNAs were listed in
   Table [216]S5 (Supporting Information).

Detection of Cytokines and Chemokines

   Cytokine and chemokine levels were quantified using the Luminex 200
   high‐throughput liquid protein chip as described previously,^[ [217]^44
   ^] following a streamlined protocol. Initially, reagents were prepared,
   including the dilution of wash buffer to 1× concentration with
   deionized water, and the preparation of bead and detection antibody
   mixtures to the 1× concentrations. Standard samples were prepared by
   brief centrifugation and mixing with universal assay buffer, followed
   by a four‐fold serial dilution in PCR 8‐strip tubes. Beads were
   vortexed and allocated to a 96‐well plate, which then underwent
   magnetic separation for washing. Each well was subsequently filled with
   a mixture of universal assay buffer and added with experimental or
   standard samples. The plate was incubated with agitation at room
   temperature, then stored at 4 °C. The procedure continued with the
   addition of detection antibodies post‐washing, followed by incubation,
   and then the addition of streptavidin‐phycoerythrin (SA‐PE). After a
   final washing step, a reading buffer was added, and the Luminex 200
   instrument was used for data acquisition. Data analysis was conducted
   using a five‐parameter nonlinear regression method for standard curve
   fitting and concentration calculations. The information of regents used
   in Luminex‐based cytokines and chemokines detection could be found in
   Table [218]S5 (Supporting Information).

Animal Experiment of PRRSV‐Challenged Model

   An animal infection model was established using 4‐week‐old commercial
   piglets, which all tested negative for PRRSV, porcine circovirus type 2
   (PCV2), pseudorabies virus (PRV), and classical swine fever virus
   (CSFV) antibodies and antigens via ELISA and RT‐qPCR methods. The
   piglets were randomly divided into two groups (n = 5 per group) and
   acclimatized in separate environments with free access to water and
   food for two days. Piglets in one group were intramuscularly injected
   with 2×10^6 TCID[50] of HP‐PRRSV Li11 virus, while the other group
   received an equivalent volume of DMEM as the mock control. Continuous
   monitoring of body temperature and weight was conducted post‐infection.
   Clinical symptoms in the infected group were recorded and scored as
   follows: sneezing (normal: 0 points, sneezing: 1 point); coughing
   (normal: 0 points, coughing: 1 point); spirit (normal: 0 points,
   lethargy/drowsiness: 1 point); appetite (normal: 0 points, reduced: 1
   point, severely reduced/not eating: 2 points); skin (normal: 0 points,
   cyanosis: 1 point); and breathing (normal: 0 points, mild difficulty: 1
   point, obvious abdominal breathing: 2 points, continuous/intermittent
   mouth breathing: 3 points). Animals were subjected to anesthetize with
   Zoletil and exsanguinate. Blood samples were collected on days 3, 7,
   10, and 21 post‐infection for untargeted metabolomic and lipidomic
   analysis. The experiment was conducted under supervision of the
   Institutional Animal Care and Use Committee (IACUC) of Sun Yat‐sen
   University (IACUC DD‐17‐0403), adhering strictly to the regulations and
   guidelines set forth by this committee. The information of regents and
   animals were listed in Table [219]S5 (Supporting Information).

Animal Experiment of HA130 Treatment Model

   About four‐week‐old commercial piglets tested negative for PRRSV, PCV2,
   PRV, and CSFV antibodies and antigens via ELISA and RT‐qPCR, were
   randomly divided into four groups (n = 12 per group). The groups were:
   (1) Mock control, each piglet injected with an equivalent volume of
   DMEM with concurrent injection of solvent carrier at the treatment
   initiation; (2) Drug control, injected with DMEM, followed by treatment
   with HA130 solution during the treatment phase; (3) Infection group,
   intramuscularly injected with virus and with solvent carrier during
   treatment; and (4) Infection plus treatment group, injected with virus
   and treated with HA130 solution intramuscularly for three consecutive
   days upon the onset of fever symptoms. After transportation to the
   disinfected experimental site, piglets were divided into four groups
   and acclimatized to their respective environments with free access to
   food and water for two days. On the third day, the third and fourth
   groups were infected by intramuscular injection of 2×10^6 TCID[50]
   HP‐PRRSV Li11 virus, while the first and second groups received
   equivalent volumes of DMEM. Daily monitoring of rectal temperature was
   conducted post‐infection, and treatment with HA130 was initiated in the
   second and fourth groups once fever symptoms (≥40 °C) were observed.
   HA130 solution was administered intramuscularly at a dosage of
   1 mg·kg^−1 body weight for three consecutive days. Continuous
   monitoring of rectal temperature and clinical symptoms, including
   weight, skin condition, feeding behavior, and mental state, was carried
   out. Animals were subjected to anesthetize with Zoletil and
   exsanguinate. Serum samples were collected on 1, 3, 5, and 7 days
   post‐drug treatment. Three piglets from each group were randomly
   selected for necropsy to collect lung, spleen, and inguinal lymph node
   samples for subsequent experimental analysis. The experiment was
   conducted under the same strict supervision as the previous experiment,
   ensuring compliance with all relevant regulations and guidelines. The
   information of regents and animals were listed in Table [220]S5
   (Supporting Information).

Preparation of Samples for RNA‐Sequencing

   Six four‐week‐old commercial piglets, confirmed negative for PRRSV,
   PCV2, PRV, and CSFV antibodies and antigens via ELISA and RT‐qPCR, were
   randomly allocated into two groups (n = 3 per group). One group was
   subjected to infection by intramuscular injection of 2×10^6 TCID[50] of
   the HP‐PRRSV Li11 virus, while the other group received an equivalent
   volume of DMEM as a control. Seven days post‐infection, all piglets
   were euthanized with Zoletil, and bronchoalveolar lavage fluid was
   collected for the isolation of porcine alveolar macrophages (PAMs). The
   isolated PAMs were then processed for RNA sequencing on the DNBSEQ‐T7
   platform. The experiment was conducted under the same strict
   supervision as previous description. The information of regents and
   animals were listed in Table [221]S5 (Supporting Information).

Preparation of Samples for Proteomics

   iPAM cells were seeded in 150 mm culture dishes at a density of 5×10^6
   cells/dish. The dishes were divided into four groups, each with three
   biological replicates, and allowed to grow until reaching 90%
   confluency. One group served as a blank control with no treatment, one
   group was pre‐treated with HA130 for 1 h before the growth medium was
   replaced, one group was infected with PRRSV at an MOI of 0.1, and the
   final group was pre‐treated with HA130 for 1 h before being infected
   with PRRSV at an MOI of 0.1. At 24 h post‐treatment, cells from each
   group were harvested for proteomic analysis using a mass spectrometry
   approach.

Data Analysis and Visualization

   Data analysis and visualization, including differential metabolites and
   lipids screening, volcano plots and correlation analysis were performed
   in R language^[ [222]^45 ^] within R Studio.^[ [223]^46 ^] The heatmaps
   were generated with TBtools.^[ [224]^47 ^] The bar plots were generated
   in GraphPad Prime. Finally, the integrated figures were plotted
   together using Adobe Illustrator. The information of software used for
   data analysis and visualization could be found in Table [225]S5
   (Supporting Information).

Statistical Analysis

   All data in this study were presented as the mean ± SEM. Statistical
   analyses were conducted using GraphPad Prism (Version 8.4.3). Error
   bars indicate the SEM. The Mann‐Whitney test was employed to assess the
   statistical significance of differences between two groups. For testing
   the significance of predicting results in machine learning models,
   Chi‐square test was used to calculate the p‐values. For correlation
   analysis, Spearman's correlation coefficient was calculated using
   Spearman's correlation test, and the p‐value was calculated using
   Pearson's Chi‐square test. Significance levels were denoted as follows:
   * for 0.01 < p‐value < 0.05, ** for 0.001 < p‐value < 0.01, and *** for
   0.0001 < p‐value < 0.001.

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   H.Z., F.H., and O.P. contributed equally to this work. H.Z. and Y.C.
   designed and supervised the study. F.H., Y.H., and G.H. performed the
   experiments. H.Z., O.P., F.H., Y.H., and C.Z. analyzed and visualized
   the data. H.Z., O.P., and U.A. wrote and revised the manuscript. Q.X.,
   B.Z., W.L., C.L., and C.X. provided the ideas and suggestions. M.C. and
   X.W. provided the technical supports. Y.W. and Q.L. helped to collect
   clinical samples. All authors have given approval to the final version
   of the manuscript.

Supporting information

   Supporting Information
   [226]ADVS-11-2402025-s001.pdf^ (2.2MB, pdf)

   Supplementary Table S1
   [227]ADVS-11-2402025-s003.xlsx^ (12.9KB, xlsx)

   Supplementary Table S2
   [228]ADVS-11-2402025-s004.xlsx^ (29.7KB, xlsx)

   Supplementary Table S3
   [229]ADVS-11-2402025-s005.xlsx^ (1.5MB, xlsx)

   Supplementary Table S4
   [230]ADVS-11-2402025-s002.xlsx^ (179.1KB, xlsx)

   Supplementary Table S5
   [231]ADVS-11-2402025-s006.xlsx^ (15.9KB, xlsx)

Acknowledgements