Transcriptome analysis unveils survival strategies of Streptococcus parauberis against fish serum

Streptococcus parauberis is an important bacterial fish pathogen that causes streptococcosis in a variety of fish species including the olive flounder. Despite its importance in the aquaculture industry, little is known about the survival strategy of S. parauberis in the host. Therefore, the objective of this study was to produce genome-wide transcriptome data and identify key factors for the survival of S. parauberis SPOF3K in its host. To this end, S. parauberis SPOF3K was incubated in olive flounder serum and nutrient-enriched media as a control. Although S. parauberis SPOF3K proliferated in both culture conditions, the transcriptomic patterns of the two groups were very different. Interestingly, the expression levels of genes responsible for the replication of an S. parauberis plasmid in the presence of olive flounder serum were higher than those in the absence of olive flounder serum, indicating that this plasmid may play an important role in the survival and proliferation of S. parauberis in the host. Several ATP-binding cassette transporters known to transport organic substrates (e.g., biotin and osmoprotectants) that are vital for bacterial survival in the host were significantly up-regulated in S. parauberis cultured in serum. In addition, groEL, dnaK operon, and members of the clp protease family, which are known to play important roles in response to various stressors, were up-regulated in S. parauberis incubated in serum, thus limiting damage and facilitating cellular recovery. Moreover, important virulence factors including the hyaluronic acid capsule (has operon), sortase A (srtA), C5a peptidase (scp), and peptidoglycan O-acetyltransferase (oatA) were significantly upregulated in S. paraubers in serum. These results indicate that S. paraubers can resist and evade the humoral immune responses of fish. The transcriptomic data obtained in this study provide a better understanding of the mode of action of S. parauberis in fish.


Introduction
Streptococcosis is one of the most important bacterial diseases in a number of fish species worldwide [1]. In South Korea, streptococcosis caused by Streptococcus parauberis is the dominant bacterial disease in the olive flounder (Paralichthys olivacus) [2]. In an effort to reduce the devastating impact of the disease on the aquaculture industry, studies have addressed diagnostic sensitivity [3], antibiotic efficacy [4], serotype variations [5], and development of an inactivated whole-cell vaccine [6]. Also, it has been revealed that the heart and brain are the major pathological target organs of S. parauberis; thus, the microbe can cause pericarditis and/or meningitis, leading to mortality, in the olive flounder [7]. Despite these academic findings and the many practical applications of research in the field, Streptococcosis remains the most common and threatening bacterial disease in South Korean aquaculture farms [2]. Recently, several genomic studies have been done on S. parauberis derived from a variety of fish species including the olive flounder and the data is being made publicly available [8,9]. However, there is a limited amount of experimental data on functional characterization of genetic components. In addition, no efforts have yet been made to identify the pathogenic mechanism or key virulence factors of S. parauberis.
In recent years, transcriptomic analysis using massively parallel cDNA sequencing (RNAseq) techniques has commonly been used to understand the pathogenesis of various types of pathogens. Ex vivo models are used for this purpose, which study bacterial cultures in various environments such as milk [10], saliva [11], heparinized blood [12], and host serum [12][13][14][15][16]. Of these, serum may be the best option for furthering our understanding of microbial pathogenesis as it contains various humoral immune factors including complements, lysozymes, antitoxins, bacteriolysins, bacterial agglutinins, and antimicrobial peptides [17]. Previous transcriptomic data from bacterial cultures in host serum have revealed key factors involved in bacterial survival and adaptation to the humoral immune system [12][13][14][15][16]. For example, Huja et al. [14] showed that Fur is a major regulatory protein involved in pathogenic Escherichia coli resistance against human serum by screening for transcriptomic and proteomic alterations after exposure of the bacteria to the serum.
In this study, therefore, global transcriptome analysis was performed to investigate how S. parauberis survives in the presence of host immunity by systemically monitoring gene expression alterations after exposure to olive flounder serum, and to identify key factors mediating its survival and adaptation strategy in the host. The novel findings of this study will aid in better understanding the pathogenicity of this organism and developing prophylactic strategies.

Experimental design and culture conditions
The SPOF3K, virulent strain of Streptococcus parauberis used in this study was isolated from the kidney of a diseased olive flounder (Paralichthys olivaceus) in Geojedo, Korea in 2013. The genome of strain SPOF3K was sequenced, annotated, and demonstrated in our previous study [9]. Healthy one-year-old olive flounders weighing approximately 200 g were purchased from a commercial fish farm in Busan, South Korea. Olive flounder blood was drawn from three individual fish that had been anesthetized with Ethyl 3-aminobenzoate methanesulfonate (MS-222; Sigma-Aldrich, St Louis, MO, USA). This study was approved by Ethics Committee of Pukyong National University (approval number: 2017-10) according to the Bioethics and Safety Act of the South Korean Ministry of Health and Welfare. Serum was separated from the blood by centrifugation at 6500 rpm at 4˚C. Bacteria were cultured in brain heart infusion broth (Becton Dickinson, Franklin Lakes, New Jersey, USA) supplemented with 1% NaCl (BN) at 26˚C for 18 hours with shaking at 160 rpm. Bacterial cells were harvested at the earlystationary phase and washed with phosphate-buffered saline (PBS, pH 7.4 at 25˚C) by centrifugation at 6000 rpm for 10 min at room temperature. The bacterial culture was then resuspended in PBS, mixed with an equal volume of BN medium or olive flounder serum (final concentration, 10 9 CFU ml -1 ) and incubated at 26˚C for four hours with gentle shaking. All samples were prepared in triplicate. At 0, 1, 2 and 4 hours post-incubation, bacterial cells were sampled for RNA extraction and viable bacterial cells were counted by the plate counting method. Student's t-test was used to identify significant differences (p-value < 0.05) in viable cell counts under different culture conditions.

RNA-sequencing analysis
Total RNA was extracted from bacterial pellets using the RiboPureTM Bacteria kit (Ambion Life Technologies, Grand Island, NY, USA) according to the manufacturer's instructions. Samples were treated with DNase I (Ambion Life Technologies) to remove trace amounts of genomic DNA. RNA concentration and quality were determined using a Qubit 3 Fluorometer and RNA High-Sensitivity Assay kit (Invitrogen, Carlsbad, CA, USA) and an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA), respectively. Ribosomal RNA in the samples was depleted using a Ribo-zero-rRNA Removal Kit (Epicentre, Madison, WI, USA). cDNA libraries were constructed using a TruSeq RNA Sample Prep Kit (Illumina, Sand Diego, CA, USA) (insertion size > 250 bp) and sequenced using an Illumina Hiseq 2500 (1 × 50 nucleotide read length). Raw Illumina sequence reads were evaluated and trimmed using FastQC [18] to remove low-quality reads. The remaining reads were mapped onto a reference SPOF3K genome [9] (accession numbers: CP025420.1, CP025421.1) using Bowtie2 [19]. DESeq2 [20] was used for data normalization and differential gene expression (DGE) analysis by uploading read counts for each gene into iDEP.92 [21]. The statistical significance of DGE was calculated based on fold change (significance at |fold change| > 1.5) and the false discovery rate (FDR, significance at < 1e -5 ) according to Benjamini and Hochberg [22]. Principle component analysis (PCA), hierarchical and K-means clustering analysis, Venn diagram analysis, and visualization were accomplished using iDEP.92 and Morpheus (https://software. broadistitute.org/morpheus). Transcripts of broth-cultured samples were used as controls for the serum-cultured samples at every sampling time point. To measure gene expression profiles based on eggNOG functional categories [23], z-score was calculated by dividing the number of significantly up-and down-regulated genes (with consideration of FDR only, significance at < 1e -5 ) by the square of the total number of differentially expressed genes (DEGs) [

Bacterial survival in serum
Bacterial growth in culture medium and fish serum is shown in Fig 1. While bacteria continued to grow in the culture medium (~2-fold increase at 4 hpe), the viability was maintained in the serum for 4 h. This result shows that S. parauberis SPOF3K is able to survive and persist in fish serum containing various antimicrobial components [17].

Transcriptome data
RNAseq analysis generated an average of 25,249,899 reads in the respective samples (S1 Table). After removal of ribosomal RNA, intergenic reads, and low-quality reads through quality trimming, the final mapping rate of filtered transcript reads into the reference genome was 87.1-98.2% (S1 Table). The RNA sequencing data can be found in the NCBI sequence Read Archive (SRA) under accession number of PRJNA689581. PCA (Fig 2a) and a hierarchical clustering heatmap (Fig 2b) showed that the transcriptome data obtained in this study was well-clustered depending on culture condition and sampling time point. In addition, the transcriptome profiles of samples taken from the medium and serum at 2 and 4 h post-exposure (hpe) were more similar than those of samples collected at 1 hpe. This indicates that there was a systemic alteration after 1 h of exposure to each condition in terms of gene expression ( Fig  2). According to the DGE analysis, 610 (28.5%), 440 (20.6%), and 411 (19.2%) genes were significantly up-regulated at 1, 2, and 4 hpe, respectively, in the fish serum compared to the culture medium, while 547 (25.6%), 446 (20.9%), and 460 (21.5%) genes were significantly downregulated at 1, 2, and 4 hpe, respectively (Fig 3a). Three-way Venn diagram analysis of the identified DEGs showed that 191 (8.9%) and 174 (8.1%) genes were significantly up-and down-regulated, respectively, regardless of sampling time point (Fig 3b). A plasmid of S. parauberis SPOF3K contains 13 genes [9]. Of these genes, inlJ (SPSF3K_02212) [25] and tetS (SPSF3K_02213) [26] have been implicated in the pathogen's virulence and antibiotic resistance, respectively. Interestingly, all genes in this plasmid were significantly up-regulated in the presence of olive flounder serum at 1 hpe (Table 1). Genes responsible for plasmid replication (repB; SPSF3K_02216-17) and mobilization (mobC; SPSF3K_02219-21) showed the greatest DGE (fold change of 14.7-34.5) ( Table 1). The Rep and Mob proteins are multifunctional elements that are essential for the initiation of both plasmid replication and conjugation in the rolling circle replication mechanism [27]. This result indicates that S. parauberis may increase its plasmid copy number when it encounters fish serum. Several previous studies [28][29][30] have demonstrated that plasmid copy number is correlated with bacterial virulence. For example, a previous study on a total of 32 Bacillus anthracis strains showed that strains harboring higher plasmid copy numbers tend to be more pathogenic to guinea pigs than those with lower copy numbers [28]. Also, animal-and human-derived B. anthracis isolates have higher plasmid copy numbers and are more pathogenic than those from environmental sources [30]. Plasmid copy number and the expression level of a plasmid replication-related gene, repA, in Yersinia pseudotuberculosis are increased in mice infected with the pathogen, while infectivity is reduced when the plasmid copy number is genetically restricted [29]. Taken together, our results indicate that this plasmid might be important for bacterial survival in fish serum at the very early stage of infection.

Gene expression profiling based on functional categories
To understand the global transcriptome dynamics of S. parauberis grown in fish serum, the expression patterns of 1,432 genes in 18 functional categories were analyzed (Fig 4). As a result, three clusters were generated by hierarchical and k-means clustering analysis (z-score, shown in colored circle) according to sampling time point. In addition, the top twenty differentially regulated genes were listed with their functional categories and clusters in Table 2. A total of 726 genes were divided into 11 functional categories. These genes were designated cluster A (Fig 4). The expression levels of the genes in this cluster became elevated as time went on. Genes in all the categories in this cluster may be essential for S. parauberis survival and proliferation, especially in fish serum, given the gradual increase in their expression levels during the incubation period. Defense mechanism was one of the most over-represented categories in cluster A (z-scores of 2.3 and 3.0 at 2 and 4 hpe, respectively). This finding coincides with the results of previous study [12] on Staphylococcus aureus cultured in human serum and blood. Some major components of the defense mechanism category include transport-related proteins (S2 Table). In many bacterial pathogens, activation of various substrate transporters has been shown to be essential for survival, as it maintains physiological balance in response to environmental changes as well as full pathogenicity [31]. In agreement with previous findings,
Cluster C is composed of four functional categories including 341 genes. The expression levels of these genes decreased as time went by. Of these four, the categories of energy production and conversion and carbohydrate transport and metabolism sampled were significantly under-represented at 4 hpe. In the carbohydrate transport and metabolism category, the main genes identified were phosphotransferase system (PTS)-related genes responsible for various carbohydrate transport mechanisms; these were significantly down-regulated (Table 2 and S3  Table). The phosphorylation status of the PTS reflects the availability of carbohydrates as well as bacterial energy conditions [43]. In the SPOF3K genome, 33 genes involved in the PTS which target 12 carbohydrates were identified (KEGG KO number: ko02060) [44]. The majority of these genes were significantly down-regulated after four hours of exposure to fish serum ( Table 2 and S3 Table). For instance, a gene encoding PtsG (SPSF3K_00506), a glucose-specific PTS component, was down-regulated (~1.9-, 8.5-, and 6.4-fold change at 1, 2 and 4 h,

PLOS ONE
Transcriptome analysis of Streptococcus parauberis in fish serum respectively) in the presence of fish serum (Table 2 and S3 Table). The under-representation of these genes might be due to the limited availability of carbohydrates in fish serum compared to nutrient-enriched culture medium.

Stress response in olive flounder serum
Induction of stress responses often represents the bacterial response to environmental changes, specifically adaptation to the host [45]. In addition to playing a key role in protein folding, heat-shock proteins (Hsp) regulate transcription of various virulence-related factors in many bacterial species including those in the genus Streptococcus [46][47][48]. Remarkably, groESL (SPSF3K_00141-142) and dnaK-dnaJ-grpE (SPSF3K_002050-2052) chaperone (also known as Hsp60 and Hsp70) and their regulatory genes ctsR (SPSF3K_00139) and hrcA (SPSF3K_02054) were all significantly up-regulated at all sampling time points, showing increasing fold changes as time went on (~2-fold change at 1 hpe and~7-29-fold change at 2 and 4 hpe) (Tables 2 and  4). These chaperones can refold, disaggregate, and properly assemble abnormal or misfolded polypeptides resulting from cell damage induced by a hostile environment [45]. This indicates

PLOS ONE
that bacteria incubated in fish serum are under severe stress due to the antibacterial activity of various components of the serum. Therefore, these proteins may play an essential role in maintaining cellular structure and integrity. In addition, several genes encoding Clp chaperone proteins were significantly up-regulated by serum treatment (Table 4). This multigene chaperone family is able to efficiently disaggregate and refold a variety of protein aggregates [49]. Several previous studies have shown that Clp proteases are essential for bacterial stress tolerance and full virulence in the host [50,51]. Three Clp ATPase-encoding genes, clpC (SPSF3K_00140), clpL (SPSF3K_01110) and clpE (SPSF3K_01798), were significantly up-regulated at 2 and 4 hpe (~4-13-fold change), while another Clp ATPase-encoding gene, clpX (SPSF3K_01107), and an endopeptidase-encoding gene, clpP (SPSF3K_00735), were not significantly up-regulated (Tables 2 and 4). Previous studies [49,51] demonstrated that three ATPases, ClpC, ClpE, and ClpX, can recognize and bind ClpP protease to form a complex (e.g., ClpXP) that can cleave misfolded proteins. In contrast, streptococcal ClpL does not require any co-chaperones to exhibit proteolytic activity [51,52]. Therefore, ClpL seems to be more crucial to the S. parauberis stress response against fish serum than the other ClpP-dependent proteases. However, further experimental studies are needed to better understand the mode of action of Clp protease family members in S. parauberis.

Differentially expressed virulence-related genes
Hyaluronic acid capsule (M protein) and capsular polysaccharide (CPS) are major virulence factors in several streptococcal species; they can increase cell adhesion ability and reduce the phagocytic and opsonic activities of the host [53][54][55]. In this study, hasA (SPSF3K_00187), a gene encoding hyaluronan synthase, and hasB (SPSF3K_00188), a gene encoding UDP-glucose dehydrogenase, were significantly up-regulated in serum-incubated samples, although hasC (SPSF3K_00192), a glucose-1-phosphate uridyltransferase gene, did not show significant

PLOS ONE
Transcriptome analysis of Streptococcus parauberis in fish serum transcriptional changes during incubation in serum (Table 3). Similarly, a previous study [56] demonstrated that inactivation of hasA and hasB results in a significant decrease in encapsulation in group A Streptococcus (GAS), while inactivation of hasC does not affect the GAS encapsulation level. Ashbaugh et al. [56] suggested that enzymes other than HasC can produce UDP-glucose, a precursor of hyaluronic acid. A recent study [57] showed that UDP-glucose can be metabolically synthesized through the galactose pathway with the involvement of GalE, a UDP-glucose 4-epimerase-encoding gene, and GalT, a galactose-1-phosphate uridylyltransferase. Our strain harbors galT (SPSF3K_00522) and galE (SPSF3K_00523), which seem to be responsible for the production of UDP-glucose. These two genes were significantly up-regulated in fish serum, with very similar expression patterns at 1 and 2 hpe (~3-and 1.5-fold change, respectively). Antiphagocytic M protein (SPSF3K_00426) showed significant downregulation at 1 hpe (Table 3), in agreement with the results of a previous study [58], showing that hyaluronic acid capsule rather than M protein plays a crucial role in Streptococcus pyogenes survival in human serum. In this study, there were no significant differences in the expression of CPS operon (nineteen genes (SPSF3K_01554-1572)) between S. parauberis incubated in serum and culture medium (S1 File). These results indicate that the hyaluronic acid capsule of S. parauberis SPO3K may play an important role in its resistance against fish serum. The Sortase A-encoding gene srtA (SPSF3K_01301) was up-regulated at all sampling time points (~4-, 2.3-, and 1.5-fold change at 1, 2, and 4 hpe, respectively) ( Table 3). Sortase A is a membrane-associated transpeptidase responsible for covalent anchoring of many virulence factors of Gram-positive bacteria to cell wall peptidoglycan by recognizing and cleaving a signal peptide containing a C-terminal LPXTG (or LPXTA) motif [59]. Several studies have shown that inactivation of srtA can result in significant attenuation of bacterial pathogenicity (e.g. in the closely related bacterial fish pathogen Streptococcus iniae) [60]. Conversely, overexpression of srtA indicates enrichment of bacterial surface proteins in cell walls. Ten CDSs containing LPXTG (or LPXTA) motifs were found in the SPOF3K genome using a hidden Markov model (HMM) (S4 Table) [61]. These include M-like protein (SPSF3K_00426), C5a peptidase (SPSF3K_00380), and Internalin J (SPSF3K_02222), which are sortase-mediated cell wall-anchored proteins and virulence factors. C5a peptidase can specifically cleave complement component 5a (C5a) [62] and was significantly over-expressed in all serum samples (~2-fold changes) (Table 3 and S4 Table). The increase in C5a peptidase in the cell wall of S. parauberis in this study may correspond to a significant reduction in the serum level of C5a, which is an important chemoattractant molecule involved in innate immunity [17], such that the bacterium has a better chance to survive and multiply in the host. GAPDH (SPSF3K_00249) and an elongation factor (Ef-Tu; SPSF3K_00907), which have multiple biological functions, were up-regulated in fish serum at 1 and 4 hpe (~2 and 1.5-fold changes respectively) ( Table 3). The release of GAPDH in group B Streptococcus (GBS) can cause re-association of bacterial cells, stimulate host IL-10 production (thus further impairing neutrophil recruitment), and induce macrophage apoptosis [63]. Besides its role in translation as a GTPase, Ef-Tu is also involved in pathogenesis as it can serve as an adhesin and an immune evasion factor by binding to complement regulatory factors such as factor H, which further inactivates complement C3b [64]. Increased expression of these multifunctional genes in fish serum-exposed S. parauberis implicates them in S. parauberis pathogenesis.
Bacterial lipoproteins have a variety of physiological functions, such as nutrient acquisition, adaptation to environmental changes, protein maturation, and adherence [65]. In Gram-positive bacteria, pro-lipoprotein diacylglyceryl transferase (Lgt) (SPSF3K_00871) and signal peptidase II (LspA) (SPSF3K_01057) are known to be involved in maturation of bacterial lipoproteins. In this study, we found that the expression of lgt was significantly up-regulated at 1 and 2 hpe (~2.3-and 1.5-fold change, respectively) ( Table 3). This indicates that bacterial lipoproteins are synthesized and mature after exposure to fish serum. In fact, a number of lipoproteins that work as substrate-binding proteins for ABC transporters are essential for bacterial survival as they maintain physiological balance in a hostile environment and sustain virulence throughout pathogen-host interactions [31]. Several significantly up-regulated ABC transporter-related genes were identified (KEGG KO number: ko02010, Table 5). These include the biotin transporter-encoding genes bioY (SPSF3K_00173-4), ecfT (SPSF3K_00021 and SPSF3K_00582), and ecfA1 (SPSF3K_00583), the osmoprotectant transporters-encoding genes, proVWX (SPSF3K_00166-7) and opuC (SPSF3K_01282), as well as the oligopeptide permease transport system-encoding opp operon (SPSF3K_01238-41). Many previous studies demonstrated that these genes are closely related to survival and virulence in several bacterial species under stressful conditions as well as in host species. For example, the significant upregulation of osmoprotectant transporter-related genes indicates that they are important for counterbalancing against osmotic stress, as previously described [66]. In addition, although opp operon activity is related to nutrient intake, it is also essential for bacterial pathogenicity, cytoadherence, and environmental adaptation in many different bacterial species [67].
In the SPOF3K genome, we identified a seven-gene cluster including BC synthase (bcsA; SPSF3K_02074), PNAG synthase (pga; SPSF3K_02077), and diguanylate synthase/phosphodiesterase (dgc/pdeA; SPSF3K_02080), which is known to modulate the concentration of cyclicdi-GMP, a positive regulator of bacterial cellulose (BC) synthesis [68]. Bacterial biofilm is formed with extracellular polymeric substances (EPS). BC is one of the major components of EPS [69]. In addition to BC, poly-N-acetylglucosamine (PNAG) is another well-known polysaccharide component of EPS. It is an adhesion molecule that is required for microbial biofilm formation in pathogenic bacterial species [70]. All of those biofilm-related genes were significantly up-regulated at 2 and 4 hpe in S. parauberis incubated in fish serum (Table 3). França and Cerca [71] reported that the transcription levels of biofilm-related genes including icaA, irgB, and capA in Staphylococcus epidermidis significantly increase when the bacterium is incubated with human blood and plasma rather than culture medium. Those authors also suggested that various proteins in plasma may interact with bacterial cells to modulate the transcriptional levels of these genes. In accordance with this idea, our results indicate that this gene cluster may also be important for bacterial biofilm formation.

Conclusions
In this study, we successfully explored the global transcriptomic dynamics of S. parauberis after exposure to olive flounder serum. Extensive remodeling of the bacterial transcriptome was identified in fish serum, which represents the early steps of S. parauberis adaptation in the host. The major genetic changes induced by exposure to fish serum include increased expression of stress resistance-related genes (e.g., groEL, dnaK operons and clp protease family genes), genes responsible for resistance to innate immunity in fish serum (e.g., hasAB, srtA, scp, oatA, etc.) and important substrate transporters, which may be key factors contributing to S. parauberis survival in the host (Fig 5). The data presented here provide fundamental background knowledge that will aid in future studies on pathogenesis and help to identify putative targets for development of new diagnostic and prophylactic strategies.