Transcriptomic analysis of Streptococcus agalactiae periprosthetic joint infection

Abstract Although Streptococcus agalactiae periprosthetic joint infection (PJI) is not as prevalent as staphylococcal PJI, invasive S. agalactiae infection is not uncommon. Here, RNA‐seq was used to perform transcriptomic analysis of S. agalactiae PJI using fluid derived from sonication of explanted arthroplasties of subjects with S. agalactiae PJI, with results compared to those of S. agalactiae strain NEM316 grown in vitro. A total of 227 genes with outlier expression were found (164 upregulated and 63 downregulated) between PJI sonicate fluid and in vitro conditions. Functional enrichment analysis showed genes involved in mobilome and inorganic ion transport and metabolism to be most enriched. Genes involved in nickel, copper, and zinc transport, were upregulated. Among known virulence factors, cyl operon genes, encoding β‐hemolysin/cytolysin, were consistently highly expressed in PJI versus in vitro. The data presented provide insight into S. agalactiae PJI pathogenesis and may be a resource for identification of novel PJI therapeutics or vaccines against invasive S. agalactiae infections.

Understanding transcript profiles of bacteria under physiological or pathological conditions may help identify genomic elements that contribute to disease processes (Croucher & Thomson, 2010;Wang et al., 2009). Massive parallel sequencing can be used to analyze transcriptomes via complementary DNA (cDNA) sequencing-so-called, RNA-seq (Kukurba & Montgomery, 2015;Wang et al., 2009), providing all transcriptomic data in an unbiased manner and at a higher resolution than microarray or individual gene or gene panel analysis (Croucher & Thomson, 2010).
Here, a transcriptome study based on RNA-seq analysis of in vivo S. agalactiae RNA from samples derived from sonication of explanted arthroplasties is presented. S. agalactiae PJI RNA-seq data were compared to previously generated RNA-seq data from S. agalactiae strain NEM316 grown in vitro (Rosinski-Chupin et al., 2015), to explore PJI-specific gene expression profiles.

| Materials
Sonicate fluid samples collected between April 2005 and August 2016 from six patients who underwent hip or knee arthroplasty revision for S. agalactiae PJI were studied. A publicly available RNA-seq transcriptome data set from S. agalactiae NEM316 (a serotype III  reference strain from the blood of a neonate with early-onset S. agalactiae disease [Glaser et al., 2002]) grown to mid-exponential phase in Todd Hewitt medium (three replicates), was used to compare gene expression patterns with RNA-seq data from sonicate fluid samples (BioProject accession number PRJEB8097: https://www. Bacterial genomes were assembled from raw reads using a de novo assembler SKESA v2.4.0 (Souvorov et al., 2018) and annotated by Prokka v1.14.5 using the S. agalactiae 2603V/R genome as the reference genome for purposes of gene annotation (Seemann, 2014).
A pan-genome was constructed with Roary v3.13.0 using annotated fragmented de novo assemblies to identify core and accessory genes (Page et al., 2015). This pan-genome served as a common reference for transcript quantification and outlier analysis between the RNA-seq data from sonicate fluid and NEM316.  (Rutherford et al., 2000).

| Phylogeny and virulence gene profiling
Extracted capsular locus sequences from each isolate were used to assign serotype, based on the highest identity using a BLAST query. The sequence type of each S. agalactiae isolate was determined by comparing allelic profiles of housekeeping genes adhP, pheS, atr, glnA, sdhA, glcK, and tkt (Jones et al., 2003)
RNA-seq reads were pseudoaligned to the pan-genome constructed as described above (Tettelin et al., 2005) and transcript abundances were quantified using Kallisto version 0.42.4 (Bray et al., 2016) and converted to transcripts per million (TPM). For the external NEM316 data, data from the three replicates were aggregated by calculating the mean TPM. Outlier expression analysis was performed by calculating a modified z-score for each gene g ij in a sample I with j genes present in the core genome such that: z = [log 2 (g ij ) − median (log 2 (g i )]/[1.4826 × MAD(log 2 (g i )], with a pseudocount added, if necessary. For this study, any gene with |z|>3 was considered an outlier.

| Homology modeling of protein structure
To predict protein homology, protein structures of selected genes were generated from amino acid sequences derived from RNA-seq data using a web-based bioinformatics server, Phyre2 (Kelley et al., 2015). Predicted structural models were retrieved for selected sequences, queried and templated in Phyre2, with a representation of structures drawn using Chimera (Pettersen et al., 2004).

| Statistical analysis
GraphPad Prism (ver. 8.0; GraphPad Software) was used to perform a Fisher's exact test to access functional enrichment of differentially expressed genes belonging to the specific clusters of orthologous genes (COGs) category.

| Description of subjects
Six subjects with S. agalactiae PJI (mean age: 62 years, range: 42-73 years) who underwent surgery at Mayo Clinic from 2006 to 2016 were studied, four (67%) of whom were male. All had local pain at the involved site, with fevers and/or chills. Two had undergone hip and four knee arthroplasty. The age of implanted material at the time of surgery ranged from 29 days to 5.4 years, including two cases of "early" (less than 3 months after placement), 1 of "delayed" (3 months to 1-2 years after placement) and 3 of "late" (more than 1-2 years after implantation) infection (Tande & Patel, 2014). S. agalactiae was isolated from sonicate fluid culture from all subjects (Table 1). There was no obvious coinfection.

| Genomic description of S. agalactiae isolates
Whole-genome sequencing of the cultured isolates showed the isolates to have diverse characteristics. Three isolates displayed serotype V, two serotype Ia, and one serotype II (Table 1). There was also diversity in isolate multilocus sequence types (Table 1)

| S. agalactiae expression profiles by outlier expression analysis
Total read counts of transcripts from sonicate fluids ranged from 31,291 to 522,023. Since Samples 1 and 4 had low read counts of non-rRNA transcripts, they were excluded from expression analysis studies (Table A1). Read counts of non-rRNA transcripts from S. agalactiae strain NEM316 RNA-seq are shown in Table A2. The S. agalactiae pan-genome constructed from the associated PJI isolates, comprised of 2738 genes, of which 1683 were identified as core genes, was used to quantify bacterial transcripts from PJI and the in vitro NEM316 strain. There were 227 genes identified as strong outliers (in vitro |z| > 3, where z is a modified z score-see Section 2) in expression between PJI (in vivo) and in vitro conditions. Of these, 164 were upregulated (in vitro z < −3), and 63 downregulated (in vitro z > 3) in sonicate fluid compared to in vitro. Genes with detected outlier expression are listed in Table S2 (https://doi.org/10. 5281/zenodo.5717630) and whole core gene lists are shown in Table S3 (https://doi.org/10.5281/zenodo.5717630), identified by the matching locus tag in S. agalactiae 2603V/R used as an annotation reference.
nik operon genes (sag1514-1518), cop operon genes (sag0384-0386), and lmb-phtD operon genes (sag1233-1234), encoding known or predicted metal transport systems, were highly expressed in sonicate fluid. sag1514-1518 comprise an operon putatively involved in intake of nickel/cobalt via nikA-E (ATP-binding cassette [ABC] transporter), which are homologous with genes involved in a nickel/ cobalt uptake system in S. aureus, cntA-D, and cntF, which has been shown to contribute to S. aureus virulence (Remy et al., 2013). The function of these genes in S. agalactiae disease has not, however, been demonstrated. The predicted structure of the SAG1518 protein is homologous to the structure of NikA from Brucella suis (protein data bank ID: 4OER), with 93% coverage and 41% identity based on Phyre2 prediction (Kelley et al., 2015) and Chimera (Pettersen et al., 2004) alignment of the two structures (Table A3, Figure A2). agalactiae, also known as an adhesin that binds a human extracellular matrix component called laminin, is involved in zinc uptake (Moulin et al., 2016). Two Lmb homologs, AdcA (SAG0535) and AdcAII (SAG1938), redundant binding proteins that combine with the AdcCB translocon (SAG0155 and SAG0156) form a zinc-ABC transporter, with their expression controlled by zinc-dependent regulator AdcR (SAG0154) (Moulin et al., 2016). In this study, sag0535 and sag0154-0156 were highly expressed in sonicate fluid compared to the in vitro strain, and sag1938 was not detected in the in vitro strain. Genes labeled as being involved in energy production and con- 3.5 | Genes associated with S. agalactiae adhesion and biofilm formation

| Adhesion factors, pilus islands, and sortases
Expression of genes encoding adhesion factors important in S. agalactiae biofilm formation was analyzed (Table S4, (Tande & Patel, 2014). Since bacteremia accompanies S. agalactiae PJI in up to 50% of cases, hematogenous spread is thought to be an important source (Everts et al., 2004). To establish infection, bacteria theoretically first colonize the gastrointestinal tract, genitourinary tract, and/or skin, where they form biofilms, and then spread hematogenously, adhere to prostheses, and again form biofilms on prosthesis surfaces. Studies on this complex pathological process are, however, limited.
In this study, the most enriched functional gene category in PJI sonicate fluid was the inorganic ion transport and metabolism category, of which genes involved in nickel, zinc, and copper transport were highly expressed. Metals play a role in life processes of microorganisms, with organisms having developed processes for their uptake. Pathogenic bacteria encounter metal restriction when placed in the metal-poor environment of their host (Hammer & Skaar, 2012). "Nutritional immunity" set up by hosts to prevent bacterial growth presumably extends to many, if not all "essential" micronutrients, with mechanisms having been described for sequestering zinc, iron, and manganese (Grim et al., 2017;Kehl-Fie & Skaar, 2010). The synovial space and surrounding tissues in which PJI occurs are limited spatially and in terms of nutrients (Jackson & Gu, 2009). The action of micrometallic molecules on surrounding human tissues, prostheses, and causative bacteria is an interesting topic; this study provides insight into this process. Although means of metal acquisition are well-known for iron, manganese, and zinc (Corbin et al., 2008), others metals in trace amounts may be important under specific conditions (Remy et al., 2013). Nickel is a cofactor of bacterial enzymes potentially involved in a myriad of cellular processes (Mulrooney & Hausinger, 2003). For Helicobacter pylori, for example, nickel, a cofactor of urease, is essential for survival and successful colonization of human gastric mucosa (Molnar et al., 2010). Recently, a nickel/cobalt uptake system (CntA-D and F/NikA-E) in S. aureus has been shown to contribute to virulence of this species (Remy et al., 2013).
In a murine bacteremia model, mortality was lower in S. aureus cnt mutant infection compared to wild-type strain infection. Bladder and kidney colonization in a urinary tract infection model were reduced with the cnt mutant versus the wild-type strain (Remy et al., 2013). In this study, sag1514-1518 (nikA-E), genes putatively involved in nickel uptake, were highly expressed in sonicate fluid. Although the roles of these genes have not been demonstrated in S. agalactiae, gene orthology suggests that they may function similarly to the S. aureus CntA-D and F system and play a role in PJI pathogenesis. The findings in S. agalactiae are novel and reported here for the first time. In addition, a transcriptome study revealed that cnt genes were upregulated in S. aureus PJI sonicate fluid compared to corresponding isolates grown in vitro (Le Masters et al., 2021). The finding of upregulation of nik and cnt genes shown in S.
aureus and S. agalactiae PJI, respectively, suggests a potential role of nickel/cobalt uptake systems in the pathogenesis of PJI.
Copper is an essential metal element in bacterial cells. However, excessive copper is hazardous to cells due to free-radical damage (Ladomersky & Petris, 2015). Keeping a balance of copper at human pathogen interfaces is needed for bacterial survival and pathogenesis.
copA encoding the copper-transporter ATPase CopA mediates control of copper efflux in several human pathogens (Johnson et al., 2015;Ladomersky et al., 2017;Macomber & Imlay, 2009;White et al., 2009). A recent study showed the role of this mechanism on survival, growth, and virulence of S. agalactiae in the mammalian host (Sullivan et al., 2021). Although copper levels in sonicate fluid were not determined in this study, they are known to be elevated in inflamed tissue (Djoko et al., 2015). As copper levels in infected tissues are increased, this may be related to increased expression of cop operon genes, which regulate copper efflux for virulence and survival of bacteria, in sonicate fluid.
Among the virulence factors studied, lmb and cspA were highly expressed in PJI compared to in vitro. Lmb is an adhesin that binds to laminin in human tissue; it also promotes bacterial invasion in human brain microvascular endothelial cells (Spellerberg et al., 1999;Tenenbaum et al., 2007). Lmb is also involved in zinc uptake, showing homology with the zinc-binding protein AdcA of other streptococcal species (Bayle et al., 2011;Linke et al., 2009). Zinc is also a trace element that serves as a cofactor for a number of essential prokaryotic enzymes and transcriptional regulators. Pathogenic bacteria must adapt zinc transport mechanisms to accommodate these differences to both avoid toxicity and meet their requirements for this metal. In a zinc-deficient environment, zinc acquisition in streptococci is mostly performed by an ABC transporter, which is composed of one or several metal-binding proteins (AdcA, Lbp, or Lmb), an integral membrane component (AdcB), and an ATPase (AdcC) (Moulin et al., 2016). In contrast, in the presence of adequate intracellular zinc concentrations, the AdcR repressor regulator inhibits expression of adcABC and lmb. In this study, these zinc uptake genes, adcABC, and lmb were highly expressed in sonicate fluid compared to in vitro, suggesting a potential role of increased zinc uptake in the pathogenesis of S. agalactiae PJI.
Although biofilm formation by S. agalactiae may be associated with PI-2a pilus production (Rinaudo et al., 2010), expression of PI-2a pilus genes was downregulated in sonicate fluid compared to in vitro in this study. While some studies have suggested that nonpilus adhesion regulated by covR may be a contributor to bacterial adherence and biofilm formation (Park et al., 2012), covR (sag0416) expression was only weakly higher in sonicate fluid compared to in vitro conditions in this study. This suggests that biofilms formed on arthroplasty surfaces may be affected by the expression of nonpilus rather than pilus adhesins, or other mechanisms.
β-Hemolysin/cytolysin (β-HC, also CylE), is a surface-associated, pluripotent toxin crucial for S. agalactiae pathogenesis; it promotes S. agalactiae invasion of lung epithelial and endothelial cells and the blood-brain barrier (Rajagopal, 2009). Hemolytic activity is associated with S. agalactiae colonization and pathogenesis, with hemolysin-deficient S. agalactiae mutants being attenuated for virulence in a S. agalactiae arthritis murine model; while more joint inflammation and damage were observed with hyperhemolytic mutant-infected animals than in those infected with the parental strain, nonhaemolytic mutant-infected mice had mild and transient arthritis (Puliti et al., 2000). The cyl operon (cylX-K) is necessary for the synthesis of granadaene, the ornithine rhamnolipid pigment in S. agalactiae, which is hemolytic and cytotoxic to human amniotic CHO ET AL.
| 7 of 12 epithelial cells and innate immune cells (Armistead et al., 2020;Forquin et al., 2007;Gottschalk et al., 2006;Rosa-Fraile et al., 2014;Whidbey et al., 2013). In this study, all 12 genes belonging to cyl operon were highly expressed in PJI compared to in vitro, suggesting that they could contribute to the pathogenesis of PJI.
S. agalactiae is encapsulated by a sialic acid CPS. Since sialic acid is also present on glycans of eukaryotic cells, the host may not recognize S. agalactiae as nonself (Rajagopal, 2009). Accordingly, CPS prevents complement factor C3 deposition and phagocytosis of S.
agalactiae (Rajagopal, 2009). The genes required for CPS synthesis are part of a single cps locus, harboring a variable serotypedetermining region (cpsG-cpsK) flanked by other CPS genes (cpsA-cpsF and neuB-neuA) conserved among different serotypes . In this study, expression of cpsB, D, E, G, and K, neuC, and D was downregulated in PJI compared to in vitro, although not among outliers. Contrary to a recent study that reported that cps genes are conditionally essential for the survival of S. agalactiae in human blood (Hooven et al., 2017), the role of cps genes in PJI might be less significant.
There are several limitations to this study. Ideally, in vitro transcriptomic analysis of each isolate under conditions corresponding to each sample in vivo would have been helpful to understand the potential pathogenic role of the genes analyzed. NEM316 and the conditions under which it was grown in vitro may not be representative of the whole S. agalactiae population or the PJI isolates studied. That said, NEM316 is a human strain from invasive disease and biofilm-producing, which is also relevant to PJI. In this study, functional validation of novel genes identified was not performed.
Finally, gene expression may have been affected at least in part by the specimen processing used. The lack of an RNA stabilizer is also a limitation.
In conclusion, the data generated provides a glimpse into the transcriptomic landscape of S. agalactiae in the environment around prosthetic joints. Using outlier expression and functional enrichment analysis, the nik operon was upregulated in PJI, suggesting a role of nickel transport in PJI pathogenesis. Among known virulence factors, β-HC was consistently upregulated in PJI. The findings presented contribute to understanding of S. agalactiae PJI pathogenesis and provide molecular targets for the identification of novel PJI therapeutics or future vaccines against invasive infections caused by S. agalactiae.

ACKNOWLEDGMENTS
Research reported in this publication was supported by the National