Integrative gene network analysis identies TIMP1 as key factors for Haemophilus parasuis infection

Haemophilus parasuis (H.parasuis), an important swine pathogen, causes Glässer's disease leading to pulmonary brosis, polyserositis, meningitis, and arthritis. However, the common molecular response and reaction from the host remain unknown. In this study, to uncover novel host factors involved in H.parasuis infection, we identied the global transcriptomics of porcine lung, spleen, blood, alveolar macrophages (PAM), peripheral blood mononuclear cell (PBMC) and aortic vascular endothelial cells (PAVECs) after infection of H.parasuis (Hps0165 strains) using microarray data and high throughput sequencing from Gene Expression Omnibus (GEO), respectively. The results showed that fteen overlapped genes were signicantly regulated in H.parasuis infected porcine lung and spleen, and then were compared with the data from porcine blood, revealing RETN, TIMP1 and C4BPA play potentially an important role for H.parasuis invasion. Furthermore, through analysing porcine cells infected with H.parasuis, we uncover the only overlap gene TIMP1 remarkably upregulated in all assembled data, indicating that TIMP1 could function as key target for the treatment of H.parasuis infection.


Introduction
Haemophilus parasuis is a signi cant pathogen in contemporary swine production systems and cause Glässer's disease characterized by acute systemic in ammation of brinous polyserositis, meningitis and polyarthritis, leading the devastating losses to the pig industry 1,2 . Moreover, H.parasuis can be frequently isolated from the upper respiratory tract of healthy pigs. Therefore, once the virulence strains emerge, H.parasuis has become an increasing threat in pig herds of high health status, especially early-weaned pigs 3 . Currently, although successful vaccination has been achieved to control mortality, multiple different genotypes and serotypes of H.parasuis frequently result in poor cross-protection of vaccine 4,5 . H.parasuis infection requires adhesion to and invasion of host cells, resistance to phagocytosis by macrophages, resistance to serum complement and induction of in ammation. Host-pathogen interactions are of great importance in understanding the pathogenesis of infectious microorganisms and host factors play key role for the microorganisms' invasion, and have the potential to become novel broad-spectrum targets for antibacterial drugs 6,7 . However, there is lack of systematic comparison of responses from those diverse sorts of tissues and cells invaded with H.parasuis.
With the development of microarray and high throughput sequencing technologies, genome-wide molecular expression pro ling has been adopted to identify key genes over the decades, providing the chances to identify the potential genes involved in H.parasuis infection. In this study, we aim to investigate the potential crucial genes in H.parasuis infection and invasion through an integrative bioinformatic analysis of gene expression pro ling in public datasets. We identify TIMP1 upregulated consistently in the diverse sorts with H.parasuis infection and predicted that TIMP1 may function pivotally in H.parasuis infection and be as a potential drug targets for anti-bacterial therapy.

Data collection and processing
We downloaded the gene expression pro les related with H.parasuis infection from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) and exploited as discovery datasets to identify DEGs (Table 1). Then we screened the differentially expressed genes (DEGs) between H.parasuis infection and controls by using the "limma" R package. Genes with |log2 fold change (FC)| larger than 1 and adj P-value < 0.01 were statistically considered as the cut-off criterion. At the same time, genes with multiple probes were collapsed to keep probes with the most delity P value within DEGs. In addition, mRNA, miRNA and LncRNA expression pro les data of PAVECs were acquired from the published papers 8,9 . Pheatmap, ggplot2 and vennDiagram packages of R were applied to generate heatmap, volcano plot and Venn diagram, respectively, for the visualization of the identi ed and overlapped DEGs.

Results
Gene transcriptional pro les of porcine spleen and lung in response to H.parasuis infection Through comparative analysis, we found that 264 transcripts showed a level of expression that differed signi cantly from that of the control group with H.parasuis serovar 5 SH0165 (HPS0165) strain infected group, while a total of 89 genes were identi ed in porcine lung infected with HPS0165 strain compared with uninfected tissues ( Fig. 1A and 1B, Supplementary S1). At the same time, we integrated the DEGs of spleen and lung and found that There were 15 overlapping genes (ALAS2, SOD2, C4BPA, TCN1, CXCL2, LTF, PDK4, TGM3, TIMP1, CRABP1, NREP, RETN, DGAT2, UPP1 and CD163) between the two datasets (Fig. 1C). The volcano plot of the datasets was shown in Fig. 1D and E.
Gene transcriptional pro les of porcine alveolar macrophages in response to H.parasuis infection Porcine alveolar macrophages (PAMs) are important lung tissue-resident professional phagocytes and play a central role in in ammation and host defense. Hence, to uncover the key genes of PAM in response to H.parasuis, we collected two different gene expression pro les of GSE30172 and GSE34544 infected by HPS5 and HPS4, respectively. The results showed that the screening of GSE30172 identi ed 257 DEGs, 204 genes which were up-regulated and 53which were down-regulated ( Fig. 2A). Similarly, we obtained DEGs from GSE34544, only 20 genes which were up-regulated and 3 which were downregulated (Fig. 2B). 11 overlapping genes (C1H15orf48, RNF128, TIMP1, CXCR6, CXCL14, IGHA, IGG2B, S100A4, GBP1, CALHM6 and LTB) were found between these two datasets (Fig. 2C). In the meantime, the overlap genes were shown using the volcano plot in Fig. 2D

Discussion
Haemophilus parasuis infection is a constant threat to the swine industry and lead to severe acute systemic infection, characterized by brinous polyserositis, polyarthritis and meningitis. Therefore, early assessment and treatment are essential to control infection rate and the mortality. To identify the key factors specially involved in H.parasuis infection, Our study was designed to take advantages of the large collection of transcriptome data of different hosts infected with H.parasuis based GEO database. In our study, we found that H.parasuis infection could produce a mass of differential expressed genes in different hosts, but the ovlapped genes among different hosts were rarely thought integrated analyses.
TIMP1 was only uncovered to act as a potential key gene associated with the invasion of H.parasuis.
TIMP1 is a glycoprotein belonging to the member family of Tissue inhibitor of metalloproteinase and is an endogenous inhibitor of matrix metalloproteinases (MMPs) 14 to regulates the extracellular matrix (ECM) turnover and remodelling during normal development and pathogenesis. The virulence-associated trimeric autotransporters (VtaAs) in H.parasuis was found containing collagen domains and binding to extracellular matrix proteins for adherence to the host 15 . In addition, TIMP1 has been identi ed to be bene cial for vascular integrity and can interact with CD63/integrin β 1 complex and regulate FAK/RhoA signaling to protect blood-brain barrier function 16 . The virulence-associated trimeric autotransporters (VtaAs) in H.parasuis was found containing collagen domains and binding to extracellular matrix proteins for adherence to the host 17 . Moreover, TIMP1 also plays an important role of anti-in ammatory and antinociceptive 18 , indicating that TIMP1 may be a critical component of a signalling cascade involved in reducing in ammatory hypersensitivity. H.parasuis infection activated the in ammatory signaling molecules and produced several pro-in ammatory cytokines in porcine cells 19,20 . Additionally, highly virulent H. parasuis infection increased a proportion of CD163 + monocytes in pigs, which are able to produce high amounts of proin ammatory cytokines, such as TNF-α, IL-1 and IL-6 21 . Furthermore, Studies have shown that the expression of TIMP1 can be statistically signi cantly regulated by TGF-β1 22 . TGF-β1 plays an important role during the invasion of H.parasuis into cells by regulating the expression of extracellular matrix proteins 23 . We summarize the above and found that TIMP1 may be involved in the pathogenic mechanism of pathogens.
Prior works indeed showed that TIMP1 regulated pathogens infection. After P. aeruginosa infection, adequate endogenous expression of TIMP-1 in cornea protects against basement membrane and extensive corneal tissue destruction 24,25 .
However, the role of TIMP1 in virus infection may act as a contrary result compared to bacteria.
Coxsackievirus B3 (CVB3) infection induced a higher level of TIMP1 mRNA expression, TIMP1 knockout mice exhibited an increased survival and attenuation of myocarditis as well as reducing viral replication 26 . TIMP1 knockout mice showed considerably decreased in ammatory in ltrates in lungs compared to wild type after in uenza virus infection 27 . Although the differences role of TIMP1 between virus and bacterial so far is unclear and need to be further uncovered to understand the molecular characteristics, TIMP1 may act as key roles for regulating pathogen infection. In