Multicentric bloodstream infection cohort study reveals new potential Staphylococcus aureus virulence factors influencing in-hospital mortality

Multicentric bloodstream infection cohort study reveals new potential Staphylococcus aureus virulence factors influencing in-hospital mortality | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Multicentric bloodstream infection cohort study reveals new potential Staphylococcus aureus virulence factors influencing in-hospital mortality Kristina Schmauder, Ulrich Schoppmeier, Jennifer Müller, Baris Bader, and 22 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9449451/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Mortality in Staphylococcus aureus bloodstream infections (BSI) is high. While clinical scores and host risk factors have been evaluated in large clinical cohorts, the relevance of the plethora of virulence factors produced by S. aureus for BSI mortality has remained elusive. By combining a comprehensive, standardized clinical data set with whole genome sequencing analysis of 643 S. aureus isolates from a multicentric BSI study and applying logistic regression analyses, we identified proteins associated with both increased and decreased in-hospital mortality. Adjustment with two clinical severity scores and sex revealed a diverse set of 116 staphylococcal proteins involved in immune modulation, metal homeostasis, adhesion, transcription, and translation (among other functions) as prognostic markers for in-hospital mortality. But also several uncharacterized proteins were associated with in-hospital mortality. Nine predictive proteins were confirmed with an orthogonal statistical approach. Most of the identified proteins have not previously been linked to infection outcomes and represent promising candidates for predictive biomarkers. Health sciences/Diseases/Infectious diseases/Bacterial infection Health sciences/Pathogenesis/Infection Health sciences/Risk factors Health sciences/Medical research/Clinical trial design/Clinical trials Staphylococcus aureus bloodstream infection virulence prognostic markers risk factors in-hospital mortality Figures Figure 1 Figure 2 Figure 3 Figure 4 BACKGROUND Staphylococcus aureus is among the most important human pathogens causing a large variety of infections. In 2019 alone, over 1 million deaths were associated with S. aureus infections globally. Among the clinical manifestations with a deadly outcome due to S. aureus , bloodstream infections (BSI) are at the forefront [1]. Clinical risk factors for mortality in BSI have been studied extensively revealing sex, age, high count in comorbidity scores ( e.g., Charlson Comorbidity Score (CCI)), acquisition (community vs. hospital-acquisition) and focus as prominent determinants of mortality risk in S. aureus BSI [2,3,4]. Additionally, disease severity at presentation represents an independent variable influencing patient outcome, which has resulted in the establishment of BSI risk scores based on clinical parameters. One well-validated, discriminative score is the Pitt bacteremia score (PBS) which assesses body temperature, mental status, presence of hypotension, mechanical ventilation and cardiac arrest during the initial clinical evaluation [4,5]. Given its clinical relevance and high burden of disease, S. aureus has been the subject of intensive research for decades. Numerous virulence factors enable S. aureus to adapt to different niches and metabolic environments. Surprisingly, even in an environment like the human bloodstream, S. aureus can overcome host immune responses and spread metastatically to various body sites. In the bloodstream, bacteria are confronted with the complement system, phagocytosing and antibody-producing immune cells and their armory of reactive oxygen species, proteases and antimicrobial substances (like defensins), as well as therapeutic antibiotics. Moreover, the battle for nutrients (macro- and especially micronutrients like iron, manganese, zinc) challenges bacterial survival [6]. To ensure survival in harsh environments, S. aureus has evolved immune-evasion strategies ( e.g. , staphylococcal protein A (spA), staphylococcal superantigen-like proteins (SSLs), chemotaxis-inhibitory protein of S. aureus (CHIPS), leucocidins, phenol-soluble modulins (PSMs), proteases), the ability to produce toxins ( e.g. , hemolysins, exfoliative toxins ETA-ETE, staphylococcal enterotoxins), adhesion molecules, biofilms, and the ability to induce abscesses ( e.g. , fibronectin-binding proteins, polysaccharide intercellular adhesin (PIA), coagulases, fibrinogen-binding proteins, Panton-Valentine-Leucocidin (PVL)), and metabolic singularities ( e.g. , metal capturing strategies, nucleotide and amino acid prototrophy). Moreover, flexible gene regulation of virulence factors under cell stress and antimicrobial pressure is ensured by two component systems such as AgrAC (belonging to the quorum-sensing accessory genome regulator agr) and diverse transcription factors [2,6,7,8,9,10,11]. The relevance of specific virulence factors for human infections has remained largely unclear because many of them are strictly human-specific and cannot be assessed in animal models [12]. Yet, some pathogenicity factors like the superantigen TSST-1 encoded by tst and an undisrupted beta-hemolysin gene ( hlb ) have been linked to severe disease progression during systemic S. aureus infections in humans [2,7,9]. Dysfunction in Agr signaling correlated with an adverse outcome in pneumonia and MRSA (methicillin-resistant Staphylococcus aureus ) infections in a big meta-analysis but not in endocarditis or central-line associated BSI, where a dysfunctional Agr was even beneficial [13]. A certain capA gene polymorphism was found to influence mortality in a bacteremia study investigating two common S. aureus clonal complexes (CC20 and CC30) after taking some clinical factors into account [14]. PVL producing strains promote an increased mortality in necrotizing pneumonia, but data remain inconclusive for S. aureus BSI [2,15]. Overexpression of hla (α-hemolysin) and the presence of the staphylococcal enterotoxin C (Sec) were linked to an adverse outcome in animal models of pneumonia and sepsis [9]. Overall, many of the pathogen factors found to be associated with clinical outcomes were derived from rather small case series, laboratory experiments, often involving only a limited number of strains and animal models, or were not adjusted to relevant host factors known to affect the prognosis. The BLOOMY study (BLOodstream Infection due to multidrug-resistant and susceptible Organisms Multicenter studY) and the follow-up project BLOOMY-PREDICT investigated the epidemiological, clinical and microbiological aspects of BSI due to S. aureus and other bacteria in hospitalized adult patients [16]. Here, 643 clinical S. aureus isolates from monomicrobial index blood cultures were analyzed, and the prognostic value of presence/absence of genetic factors for in-hospital mortality was assessed to potentially serve as biomarkers in S. aureus BSI. RESULTS Analysis of cohort composition Within the BLOOMY and BLOOMY-PREDICT studies, 970 cases of S. aureus BSI were documented, of which 928 (95.7%) were monomicrobial BSI. Only isolates from index blood cultures were sequenced and matched with the clinical data set of their infected host. The final cohort comprised 494 isolates from the BLOOMY study and 149 isolates from the PREDICT study phase (total = 643). The initial and final composition of the cohort is shown in Fig. 1. Study site and phase specific distribution of sequenced isolates are shown in Supplementary Fig. 1. Patient demographics, including age, sex, PBS, CCI, discharge status, appropriateness of therapy, acquisition type of BSI (community-acquired vs. hospital-acquired; defined as ≤72h and >72h after hospital admission), and BMI (body mass index), additionally to S. aureus specific features, are displayed in Supplementary Table 1. To correct for possible genetic variation during the recruitment span, isolates were randomly assigned to training and validation cohorts (75% and 25% of total). Of note, the validation data was handled as true holdout data set and only used to assess the quality of our model in the final ANOVA (analysis of variance). Comparison of the epidemiological features (Supplementary Table 1) between the two cohorts did not show significant differences. The overall in-hospital mortality rate was 24.7% (95% CI 21.4-28.3%). The Odds ratio of female vs. male in-hospital mortality was 1.68 (95% CI 1.16-2.43). Identification of clinical confounders for prediction of in-hospital mortality To account for confounding clinical variables on in-hospital mortality in S. aureus BSI, a multiple logistic regression analysis was performed. The following clinical and epidemiological factors as known confounding predictors of in-hospital mortality were preselected: age, sex, BMI, CCI, PBS, BSI focus at discharge and appropriateness and duration of antibiotic therapy. Among those, only sex, CCI and PBS contributed significantly to the final clinical model. Performance of the clinical multiple logistic regression analysis with these three factors resulted in a AUROC (area under the receiver operating characteristic) of 77% for the training and 82% for the validation cohort (Supplementary Table 2). Genomic features of S. aureus isolates A phylogenetic tree including all S. aureus isolates with their local origin, MLST type, the associated survival, study cohort and their assignment to the training vs. validation cohort, is shown in Fig. 2. In total, 61 distinct MLST types (most commonly ST45, ST7 and ST22) and 232 unique spa types (predominantly t091, t084 and t002) were identified (Supplementary Fig. 2). ST22 and t002 are known successfully circulating lineages in Europe [17]. Altogether, we postulate a genetically heterogeneous S. aureus cohort in this study. Genomic factors were identified by annotating the protein-based UniRef100 and UniRef90 trait assigned by the UniProt database. 62.472 unique proteins (99.8% coding, 0.02% non-coding) were annotated with 51.639 unique UniRef100 and 11.221 unique UniRef90 traits. When counting every single annotated genetic feature in all S. aureus isolates, a total of 1.736.046 traits (1.624.814 coding (cds), 5.859 short coding (sorf) and 105.373 non-coding genes ( e.g., tRNA, rRNA, ncRNA) were annotated. Among the proteins, 93.4% (n=1.622.156) had a UniRef90 while 89.3% (n=1.549.756) had a UniRef100 annotation. UniRef traits associated with S. aureus in-hospital mortality We next incorporated annotated coding and non-coding proteins in two distinct statistical models (Fig. 3). First, UniRef100 traits served as a unique protein-based trait for our rigid model ( Model A ). UniRef100 includes identical sequences at the amino acid level, and hereby represents a single protein variant. Second, to obtain prognostic traits on a protein group level we utilized the UniRef90 traits ( Model B ) comprising protein variants with at least 90% amino acid sequence similarity. Depending on the degree of protein conservation the amount of protein variants summarized in the UniRef90 trait differed from 1 – 45 variants. The final results of the significant traits in both models after correcting the significant UniRef traits for sex, CCI and PBS are shown in Fig. 4. Sorting of the significant traits and clinical factors was achieved by multiple repetitions of the ANOVA analysis each reduced by the traits/factors contributing least to the previous cycle of ANOVA. This sorting of the traits provided insight into the most impactful features in the predictive model. The top 10 traits for each model, together with their given protein names, associated in-hospital mortality rate and genetic abundance, are listed in Table 1 (A) and (B). The full list of significant traits (77 traits in Model A; 39 traits in Model B) with their suggested mode of action in S. aureus pathogenicity, the decisive clinical factors (three each in Model A and B), and the AUROC value of each cumulative feature set from the highest to the lowest rank of the training and the validation cohort as a marker for the quality of the prediction model are displayed in Supplementary Table 3. Eleven traits, all associated with increased in-hospital mortality, were concordant in both models highlighting the overlapping consistency of the two annotation approaches (Table 1 (C)). Those eleven traits showed a fairly conserved amino acid sequence reflected by maximum five different protein variants (UniRef100) within the UniRef90 trait. The commonly identified traits included four functionally unknown proteins, one uncharacterized phage protein, the staphylococcal enterotoxins type J and R, two replication proteins and two transcriptional regulators. To validate the relevance of the uncovered traits in a second independent genome-wide approach, we fed the same input used for model A and our phylogenetic tree in the Scoary2 pipeline (for results see Supplementary Table 3 and 4). Scoary2 is a tool for genome-wide association studies reflecting also the phylogenetic background of the strain collection but does not incorporate clinical data. Here, nine of the 77 UniRef100 traits were significant, including all five highest ranked traits from our Model A. Those five traits included an uncharacterized protein and protein variants of the ABC transporter substrate-binding protein FhuD1, the proline—tRNA ligase ProS (all three prognostically unfavorable), the ATPase component of the ATP binding cassette (ABC) superfamily transporter MntA and a lipoprotein with unknown function (both favorable). Additional commonly identified traits in Model A and Scoary2 were variants of a HTH transcriptional regulator, an unknown amino acid proton symporter, the fibronectin-binding protein A and a staphylococcal tandem lipoprotein. In summary, future studies should focus on those nine traits as promising and independently confirmed candidates influencing S. aureus virulence in human BSI. Co-dependency of relevant traits To evaluate patterns of co-occurrence in single S. aureus lineages, traits were displayed in an absence/presence matrix (Supplementary Table 5) sorting the isolates according to the phylogenetic tree as a backbone. Most traits are distributed across several MLST types, indicating fairly lineage-independent virulence proteins. Upon inspection, three clusters became apparent in which various traits frequently occurred in close proximity on the contig, suggesting either a functional or local dependency: cluster I., comprising two uncharacterized lipoproteins, cluster II., comprising a phage protein and integrase, and cluster III., comprising staphylococcal enterotoxins R and J, a MarR transcriptional regulator, a site-specific DNA recombinase, a phage protein, three replication-associated proteins and two uncharacterized proteins (see Supplementary Table 3). Interestingly, all proteins found in cluster III. are located on the pIB485-like plasmid. For each cluster, it has yet to be determined whether the corresponding proteins all actively influence in-hospital mortality, which we consider rather unlikely, or whether they are only surrogates for mortality, at least in part. To detect protein dependencies beyond their close proximity in the genome, UniRef100 traits were further investigated by in silico functional network analysis using the STRING database. A number of networks became apparent linking the traits by associations in curated databases, experimentally determined functional connections, gene neighborhood and co-occurrence across genomes (Supplementary Figure 3). Indexing of the most important traits identified in the analysis The most important traits with a known mode of action in our analysis were two protein variants involved in micronutrient transport (FhuD1 (UniRef100_Q93PN3) for iron and MntA (UniRef100_Q99VY2) for manganese and zinc transport) and a variant of the proline—tRNA ligase ProS (UniRef100_A7X1P3), involved in protein translation. Known virulence factors in S. aureus pathogenicity identified in our study were protein variants of the fibrinogen-binding protein (UniRef100_P68799) and fibronectin-binding protein A (UniRef100_UPI0002CA386D), both crucial for adhesion processes. Moreover, two staphylococcal enterotoxins (Sej (UniRef100_D2J5X0) and Ser (UniRef100_O85217)) were found among the prognostic traits. Contrarily, proteins that have been closely intertwined with S. aureus pathogenicity like hemolysins, coagulase, PSMs or superantigens like TSST-1 did not appear as significant risk factors in our models. The gene abundances of eta , etd and lukF/lukS (coding for PVL) were below 5%, therefore failing the criterion to be incorporated. DISCUSSION Pathogenicity of Staphylococcus aureus has been studied extensively over the last few decades, addressing an important clinical need. We sought to illuminate the impact of bacterial genetic factors and link them to outcome of BSI in hospitalized patients. Selected variables incorporated in the clinical model were age, sex, BMI, CCI, PBS, ascertained focus at discharge, and appropriateness and length of administered therapeutic substances. Our pragmatic definition for appropriate antibiotic therapy reflects antimicrobial recommendations in international guidelines [18] and was applied to account for in-hospital mortality. However, this definition does not claim to fully consider all recommended therapeutic strategies and interdisciplinary approaches, particularly in cases of deep-seated infections, endocarditis or bone and joint infections. PBS, CCI and sex turned out to be the most impactful clinical variables in predicting in-hospital mortality and are commonly known risk factors for lethal outcome [5,19,20]. However, PBS was used here not as an indicator for in-hospital mortality risk as its actual intended use, but to account for the patient’s state at the time of blood culture sampling on day 0. Irrespective of the genetic equipment of the diverse S. aureus isolates and the elapsed time until BSI was diagnosed, PBS served as a benchmark to level the initial clinical status. Thus, additional potential predictors (particularly specific laboratory values) were not included as confounders. Our data support the importance of gathering clinical scores to estimate in-hospital mortality risk in BSI. We created a rigid statistical Model A in which a vast number of proteins were annotated based on their 100% identity at the amino acid sequence level. In contrast, Model B was fed with proteins sharing 90% similarity at the amino acid sequence, allowing to associate orthologous or paralogous protein groups. Importantly, we correlated entirely novel factors with in-hospital mortality, but also well-described proteins linked to S. aureus virulence, highlighting the biological plausibility of our method. Of note, our approach did not consider transcriptional regulation of the underlying genes but focused on the influence of absence or presence of proteins and their amino acid variants. There are further explanations why virulence factors might not have been detected in our study: 1. Low-abundancy of the trait ( e.g., PVL). 2. Redundancy of the trait ( e.g., PSMs which are conserved proteins of the core genome). 3. High amino acid variability ( e.g., α-hemolysin [21]) potentially reducing the statistical power of the trait. Scoary2 was applied as an orthogonal approach to validate our results and to minimize the phylogenetic background as a surrogate for clonal virulence [22]. The sheer number of significantly associated proteins compared to other genome-wide approaches underlines the importance of correcting not only for the phylogenetic background but also incorporating clinical data to untangle BSI mortality markers, the relevance of which could otherwise be under- or overestimated by those confounders. The 116 relevant proteins can be grouped by their mode of action and contribute to S. aureus pathogenicity and resilience at different levels: immune modulation, metal homeostasis, gene transcription and translation, recombination and genome editing, oxidative stress reduction, adhesion, abscess formation, metabolic functions, heavy metal resistance and environmental sensing. All these factors can influence bacterial survival chances in the harsh, contested bloodstream niche. The function of other proteins remains obscure. To date, it is unclear if they singularly contribute to high assessment scores like the PBS over time of ongoing BSI. Furthermore, since S. aureus lineages and their virulence factors are distributed in globally inhomogeneous patterns, the abundance and, thus, impact of the identified virulence factors are not per se generalizable to other patient cohorts [23]. In the following part, we discuss a selection of the identified proteins to contextualize their importance in S. aureus BSI. UniRef100_Q93PN3, a variant of the ABC transporter substrate-binding protein FhuD1, is an extracellular surface lipoprotein binding ferric hydroxamate and staphyloferrin, which is a siderophore extracting iron from transferrin [10]. Iron restriction has challenging consequences on both bacteria and the host, as shortage for this valuable co-factor causes deterioration of essential enzymatic functions. Since the host restricts access to iron during acute and chronic infection, in a process known as nutritional immunity, bacteria have evolved efficient iron-capturing strategies to outperform the hosts own capacity [24]. FhuD1 is selectively upregulated under iron restriction to enable uptake of iron-bound siderophores, and also xenosiderophores, a process known as iron piracy [25,26]. Patients with iron overload who are administered the chelator desferroxamine B are at risk of S. aureus infections. Whether this increased risk can be attributed to FhuD1/2 remains controversial [25,27,28]. It is also unclear whether this protein variant of FhuD1 confers advantages in iron scavenging compared to other variants, resulting in better bacterial survival in the bloodstream and thereby increasing human in-hospital mortality. The presence of UniRef100_Q99VY2, an ATPase variant of the ABC superfamily transporter MntA and component of a Mn 2+ /Zn 2+ transport system, correlated with decreased in-hospital mortality. Zinc is indispensable for enzyme activity, in humans and microorganisms alike, and is thus fiercely competed for in the host [29,30]. Manganese is an essential co-factor for the superoxide dismutase, which eliminates reactive oxygen species and thereby protects the bacterium from oxidative cell stress [31]. The importance of these two metals for bacterial survival is undisputed and the correlation of this protein variant with a beneficial outcome needs further investigation. Differences in the enzymatic activity of this variant might explain the observed association and further studies should focus on comparing the catalytic activity of the various variants. Staphylococcal enterotoxins J (Sej) (UniRef90_Q76LS7) and R (Ser) (UniRef100_O85217/UniRef100_D2J5X0) are known enterotoxins and act as superantigens [7,9]. Both were associated with increased in-hospital mortality and co-detected with eight other traits in our cohort, including a MarR family transcriptional regulator and other uncharacterized protein-encoding genes on the pIB485-like plasmid [32]. The presence of Sed (staphylococcal enterotoxin D), also located on this plasmid, and other enterotoxin-encoding genes was previously associated with the manifestation of endocarditis, although in another cohort, absence of sed/j/r correlated with embolizing endocarditis [33,34]. It is yet unclear whether the identified factors are drivers of in-hospital mortality or simply correlate with other unfavorable factors that are located on the plasmid and were not identified with our approach. Transcription factors play a crucial role in bacterial adaptation and operate in ubiquitous processes, including biofilm formation, chemotaxis, secretion mechanisms, adhesion and many more. They can either function as inhibitors or activators of transcription and interact with multiple effector proteins ( e.g., RNA polymerase, DNA, RNA) [35]. We found protein variants of different transcription regulator families often but not exclusively, correlating with elevated death rates. As an example, MarR (Multiple antibiotic resistance regulator), which is represented with two protein variants (UniRef100_D2J5X2/UniRef90_A0A1W5ISY9), comprises a family of transcription factors responding to antibiotics, chemical and oxidative stressors and contributing to virulence in S. aureus [35-37]. Whether each identified protein variant of those transcription factors has repressive or activating functions and how this is associated with a beneficial or unfavorable BSI outcome remains to be further elucidated. Finally, among the relevant traits were also several phage-originating proteins mostly associated with increased in-hospital mortality. Temperate bacteriophages harbor a substantial amount of virulence determinants and significantly contribute to S. aureus pathogenicity [38,39]. Moreover, several uncharacterized proteins like domains of unknown function (DUFs) or the highest ranked trait UniRef100_A0A6C0L8I7, also having the highest trait specific in-hospital mortality rate of 60.6%, appeared as relevant factors for outcome prediction. Origin and function of these uncharacterized proteins remain enigmatic requiring further investigation. Recently, a study was published searching for predictive factors discriminating S. aureus strains based on origin and clinical manifestation (colonization vs. different severity stages of infection). By combining Random Forrest and two genome-wide association methods, annotated genes in coding regions, intergenic regions (IGR) and sRNA from human and animal isolates were associated with the mentioned endpoints. mecA and an adjacent IGR were identified as discriminative factors for (severe) S. aureus infections versus colonization [40]. We could not confirm this finding, possibly due to different study endpoints and a MRSA rate of 5.1%, which is lower than the 43% in the study of Sassi et al .. In comparison, our analysis stands out for its large, clearly structured, multi-centric cohort and exclusive focus on S. aureus blood culture isolates, and for accounting and correcting for a tremendous number of clinical parameters. Nonetheless, a known limitation of our study is the use of short read sequencing, resulting in assembly issues especially affecting genomic low-complexity regions or plasmid assembly. This necessitates confirmation of the traits in vitro and in vivo , particularly those that have not yet been characterized. To the best of our knowledge, this is the first large BSI study on S. aureus combining a vast, standardized clinical data set with whole genome sequencing analysis to evaluate bacterial traits as predictors of in-hospital mortality in a congruent cohort. These findings provide new evidence that pathogen genomic factors may add clinically relevant prognostic information beyond host characteristics and disease severity, and uncover a set of biologically plausible candidates for translational development. Validation in independent cohorts and functional studies will be important next steps, but the present results offer a valuable framework for improving risk stratification and for guiding the future development of predictive biomarkers and innovative preventive or therapeutic strategies. MATERIAL AND METHODS Study population Patient recruitment for the BLOOMY study took place between 2017 and 2018, while recruitment for the BLOOMY-PREDICT cohort was carried out between 2019 and 2020. Participating hospitals consisted of six German university hospitals. Hospitalized patients ≥18 years of age suffering from bloodstream infections due to ESKAPEE microorganisms ( Enterococcus faecium/faecalis , Staphylococcus aureus , Klebsiella species , Acinetobacter baumannii , Pseudomonas aeruginosa , Enterobacter species and Escherichia coli ) were eligible for study participation and were asked for consent. Clinical departments with low incidence of bloodstream infections (ophthalmology, psychiatry) were excluded. Clinical record data were reviewed and entered in a database structure (Research Electronic Data Capture, REDCap). A full list of recorded epidemiological and clinical variables, including the PBS and the Charlson comorbidity index (CCI) for each patient, was published previously [16]. To account for antimicrobial therapy as a confounding factor for survival criteria for the appropriateness of therapeutic antimicrobials and duration of therapy were selected in a fairly strict manner reflecting antimicrobial recommendations in international guidelines [18]. In patients with survival of ≥14 days after BSI onset, appropriate in-hospital therapy was defined as IV administration of one of the following antibiotics for at least 14 days: flucloxacillin, cefazolin, piperacillin/tazobactam (only if not exclusively administered for the duration of treatment due to know elevated mortality risk for piperacillin/tazobactam monotherapy [41]), ampicillin/sulbactam (all four antimicrobials for treatment of MSSA (methicillin-susceptible Staphylococcus aureus) cases only), vancomycin, daptomycin or linezolid (last three drugs for treatment of both MSSA and MRSA). In patients who died or were transferred before day 14, adequate therapy duration was set to x−1 days, where x is the time from BSI onset to death or transfer. We are aware that this definition does not perfectly reflect the demand for individually tailored therapeutic strategies and interdisciplinary approaches particularly in complicated cases of S. aureus BSI. It also does not distinguish between empirical or targeted therapy or consider the infection focus or diagnostic procedures like negative follow-up blood cultures. We tried to establish a pragmatic definition to purposefully account for appropriate therapy as a potential confounder in our large cohort and the endpoint ‘in-hospital mortality’. For the current study, isolates and clinical metadata (Supplementary Table 6) from five study sites of the BLOOMY cohort and two study sites of the BLOOMY-PREDICT cohort were available and included. We only included monomicrobial BSI, defined as no isolation of pathogens other than S. aureus from a blood culture obtained within 24 hours after the first positive blood culture with S. aureus (day 0). Common skin contaminants (coagulase-negative staphylococci, Corynebacterium spp. , Bacillus spp. , and Cutibacterium spp. ) detected in one of several blood cultures were not considered as polymicrobial infections. The study endpoint of interest in this S. aureus specific sub-cohort was in-hospital mortality due to BSI. The study was conducted under the ethics approval no. 765/2106BO1 and 584/2019BO1, issued by the local ethics committee of the University Hospital Tübingen. Cultivation and isolation of Staphylococcus aureus Isolation of S. aureus from blood cultures was performed using commercially available automated blood culture incubator systems following standard protocols at the respective study sites. Blood cultures were streaked on agar growth media and plates were incubated at 37 °C in different atmospheric environments following state of the art standard diagnostic protocols. For identification of microorganisms, rapid diagnostic molecular testing as well as MALDI-TOF (Matrix-assisted laser desorption/ionization-Time of flight) mass spectrometry using the Microflex LT instrument (Bruker Daltonics, Bremen, Germany) were used. Antimicrobial susceptibility testing was performed using commercial semi-automated systems ( e.g., VITEK 2, bioMérieux, Nürtingen, Germany) and interpreted according to the breakpoints published by EUCAST (European Committee of Antimicrobial Susceptibility Testing; https://www.eucast.org/). Suspected methicillin resistance was confirmed by commercially available assays for molecular mecA/C gene detection. Sequencing of isolates Isolates of S. aureus were frozen at -80 °C until further sequencing analysis. Genomic bacterial DNA was extracted from cultures grown on Columbia agar supplemented with 7% sheep blood (Thermo Fisher Scientific, Schwerte, Germany) using the Qiagen DNeasy® UltraClean® Microbial Kit (Qiagen, Hilden, Germany), following the manufacturer's instructions with minor modifications. DNA concentrations were quantified using the Qubit™ dsDNA BR Assay Kit (Thermo Fisher Scientific, Massachusetts, USA). Whole genome sequencing libraries were prepared with the Illumina DNA Prep Kit (Illumina, San Diego, USA) using a standard protocol, and samples were barcoded using Illumina® DNA/RNA UD Indexes. The barcoded libraries were subsequently quantified using the Qubit™ dsDNA BR Assay Kit (Thermo Fisher Scientific). Normalized libraries were pooled equimolarly and sequenced on a NextSeq 500 platform (Illumina, San Diego, USA) using a Mid Output Cartridge v2.5 (2x150 bp), aiming at approximately 100x genome coverage. Whole Genome Sequences are accessible under the accession number ERP191105 (BioProject No. PRJEB110420) at the European Nucleotide Archive (ENA). Bioinformatic analysis Genome assembly and annotation The sequencing files were demultiplexed with our in-house Nextflow pipeline ncct-mibi/nxf-bcl, which uses bcl2fastq (v2.19.0.316) for demultiplexing, fastp (v0.23.4) for quality check and visualized the results with MultiQC (v1.7). The pipeline was executed with Nextflow (v20.10.0) to allow parallel processing of files. The generated data were analyzed with the Nextflow based Bactopia (v3.0.1) pipeline, which allows a standardized processing of the data in a parallel fashion [42]. After a basic quality check with fastp (v0.23.4) and FastQC (v0.12.1), the raw reads were assembled using unicycler (v0.5.0) [43]. Afterwards, annotation was performed using bakta (v1.9.3) with the database available at that time (zenodo.10522951), which was downloaded on 17.04.2024 [44]. The bakta output included annotation of the protein-based UniRef100 and UniRef90 traits assigned by the UniProt database (https://www.uniprot.org/) on each gene predicted within all samples. UniRef100 and UniRef90 traits were further used as input for the statistical analysis. UniRef100 represents a reference cluster of proteins with a 100% identity at the amino acid sequence level. UniRef90 further groups the UniRef100 clusters with an amino acid sequence identity of 90% allowing to analyze orthologous or paralogous protein families. Finally, an additional taxonomic classification was performed with gtdb-tk (v2.4.0, database release 220, downloaded 04.06.2024) to validate the species and obtain average nucleotide identity (ANI) to the closest reference genome. Additional classification for MLST (Multilocus Sequence Typing), spa and agr type The resulting assembly was used for automated classifications of the MLST, spa and agr type. All of them are included in the tools collection (Bactopia Tools) of Bactopia (v3.0.1). The MLST type was determined using mlst (v2.23.0) and the mlst-database (v2.23.0-20240325), the spa type using spatyper (v0.2.1) and the agr type using agrvate (v1.0.2). Phylogenetic analysis Core genome analysis was performed using 645 S. aureus whole genomes (including 642 isolates of the analyzed cohort) in SPINE (version 0.2) [45], setting LCB (locally collinear blocks) to 1000, permissible SNP variation to 90% and number of included isolates to 95% for core genome identification. The core genome consisted of 2.347.879 bases (2.3 Mb). To check for possible phage insertion in the core genome, the computed core genome was uploaded to PHASTEST.ca (Version 3.0; upload date 27.09.2024 ) [46] where no prophage sequences could be assigned. For SNP calling and further filtering, GATK (The Genome Analysis Toolkit (GATK) v3.2-2), SAMtools v.0.1.19 and VCFtools - v0.1.11 were used (settings for minimal mapping score: 30; minimal read coverage: 10; maximal read coverage: 3; minimal SNP mapping quality: 30; minimal SNP base quality: 30; SNP percentage 0.8; no indels) [47-49]. For final maximum likelihood (ML) tree construction, we used IQ-TREE multicore version 1.6.3. (settings: UFboot and bootstraps 1000) [50]. Visualization of phylogenetic trees was performed in iTOL v6. Statistical analysis Two goals were to be achieved: First, we aimed to find a predictive model for in-hospital mortality due to S. aureus BSI. The prediction ought to be based on a suitable selection of clinical data and genomic information about the S. aureus bloodstream isolates. The second goal was to identify bacterial traits that were potentially responsible for the outcome. The data structure in the BLOOMY and the PREDICT study were different with regard to coding of some parameters and naming of data fields. Therefore, we first harmonized the data fields and incorporated missing information in single cases in a random fashion. This incorporation of missing information was not applied for BMI (body mass index) due to the high number of missing values. Preparation of the genomic information was straight forward given the results from the bioinformatic preprocessing. Second, we split the data into two cohorts (training and validation cohort) by using a random mechanism. Third, we analyzed the distributions of demographic, clinical and isolate specific parameters (age, sex, infection focus, resistance pattern, etc.). We provide a descriptive overview of the cases and evaluated whether training and validation data were comparable (Supplementary Table 1), using Fisher’s exact and Chi Square tests with a significance level of p≤0.05. Lastly, we built a predictive model using ANOVA and logistic regression, described in detail in the following part. The workflow of the predictive model is displayed in Figure 3. Regarding the clinical data, we applied multiple logistic regression analysis with likelihood tests and three out of eight clinical factors with known influence on mortality in S. aureus BSI were significant (p≤0.05) (Step 1): PBS, CCI and sex. For the genomic content of isolates, we used the annotated protein-based UniRef100 traits (resulting in Model A) and UniRef90 traits (resulting in Model B) as input (Step 2 and 3). Coding and non-coding genomic regions were included. We applied univariate logistic regression analysis on protein absence/presence for the outcome survival vs. death during hospitalization to extract p-values (Step 4). Traits were incorporated in the multiple logistic regression analysis when their p-value was less or equal to 5% (Step 5). Of note, these p-values were used as condensed information for sieving purposes and not for hypothesis testing. Next, highly and rarely abundant traits (<5% and ≥95%) were removed (Step 6). Then, we ran a multiple logistic regression analysis in sets of preselected clinical data (PBS, CCI, sex) and each remaining trait to find out about the surplus of that specific trait to the clinical data (Step 7). This resulted in another list of p-values used in a sieving step (Step 8). Only data from the training cohort were used from Step 1-8. Only statistically significant traits (p≤0.05) were considered eligible for inclusion in the final analysis of variance (ANOVA). The validation cohort was not used for cross validation purposes but was handled as a true holdout data set and used to assess the quality of our model in the ANOVA analysis. By performing repeated cycles of ANOVA, we reduced the number of UniRef traits and clinical factors (traits and factors further on summarized as ‘features’) step by step (Step 9 and 10). At each step, the traits/clinical factor contributing least to the model was removed and were sorted accordingly to reflect their impact in the predictive model. The resulting cascade of ANOVA, including the initial set of features down to the smallest, was each submitted to an AUROC computation (Step 11). We calculated the AUROC value for each feature set from the ANOVA analysis on the training as well as on the validation data. Notably, some UniRef traits co-occurring in the exact same isolates had to be collapsed to one vicarious trait to enable ANOVA analysis leading to a supposedly reduced number of traits in Figure 4 compared to Supplementary Figure 2. Hidden traits were de-collapsed after ANOVA. All codes used in the data analysis were organized in a sequence of distinct R scripts. Each script processes the data from the preceding step and passes its results to the next. Parameter and data paths were stored in a separate data file so that flexibility and transparency are guaranteed. The code is available on GitHub in repository [http://github.com/UliSchopp/BloomyStaphAu]. We provide detailed description of the code including information about the R packages in the repository. To compare our method with a second independent approach we used Scoary2 (version scoary-2:0.0.15), a software for genome-wide association studies with, however, limitations to incorporate clinical data [22]. The calculations were based on all annotated UniRef100 traits of the training cohort isolates to proceed with the same input. Moreover, the phylogenetic tree prepared in a preceding step was incorporated to account for the phylogenetic background in our cohort. The values shown in output column with title fq*ep calculated and used by the Scoary2 pipeline as criteria for significance (significance level p≤0.05) were utilized to identify relevant traits. Visualization of figures and analyses were performed in R, GraphPad Prism (10.0.0), GIMP (3.0) and Microsoft Excel (version 2408). Absence-presence matrix of significant UniRef100 trait To determine the absence and presence of significant UniRef100 traits in all samples a custom python3 script was used. The script was created with the help of ChatGPT and requires the python packages argparse (v1.1), pandas (v2.2.2) and pathlib. The script was validated for functionality by the bioinformatician (JM). The input of the script is a gff3-like file with the annotations of all samples and a list of UniRef100 traits. The script extracts the samples, which contain the UniRef100 trait and creates an absence-presence matrix with UniRef100 traits as columns and samples as rows. For each sample-UniRef100 trait pair the location of the UniRef100 trait on the assembly is tracked as well. Protein network prediction To detect protein dependencies beyond their close localization in the genome, traits were further investigated by in silico functional network analysis using the STRING database (accessed on February, 2026, [51]). Within this database, protein-protein dependencies can be predicted from curated databases, experimentally determined functional connections, gene neighborhood and co-occurrence across genomes. Amino acid sequences of all significant UniRef100 traits were loaded into the database. Sequence identity to other homologues in the database was >90%, and names were assigned based on the UniProt name. Networks with at least three homologues and with minimum required confidence score set to 0.2 (between medium to low confidence) were retrieved and analyzed. Declarations Funding: The BLOOMY (TTU 08.810) and BLOOMY-PREDICT study (TTU 08.821) were funded by the German Center for Infection Research (DZIF) and supported by the DZIF Clinical Research Unit (TTU 08.701), TK is funded by the German Center for Infection Research (DZIF, TTU06.716 and TTU08.716). Authors contributions: Conceptualization of study (SP, MW, and KS), development of study design (SP, KS, US, and JM), sequencing of study isolates (BB and SP), data verification, bioinformatic and statistical analysis (JM, US, TK, and KS), review of study progress (KS and SP), development of the study concept, design and protocol of the BLOOMY/BLOOMY-PREDICT study (ET, WVK, SE, SP, HS, MJGTV, JR, TC, and CG), data collection, coordination and interpretation of the BLOOMY/BLOOMY-PREDICT study (ET, WVK, SE, BPG, SG, SR, MJGTV, HS, JR, TC, CG, LMB, EK, CI, NK, KS, KX and PGH), administration and validation of the clinical data base (BPG), writing of the first draft (KS, SP, LLP, and AP), contribution to the manuscript and approval of the final version of the article (all authors), submission of the manuscript (KS and SP). CONFLICTS OF INTEREST Maria JGT Vehreschild Grants or contracts from any entity: MSD, Heel, Roche, Tillotts, Pfizer Payment made to institution Consulting fees GILEAD, Tillotts, Pfizer, Bioaster, GSK, Ecraid, EUMEDICA, Bactolife, PAION Payment made to myself Payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing or educational events Akademie für Infektionsmedizin, Astra Zeneca, bioMerieux, Biotest, DGI, European Society of Neurogastroenterology, Falk Foundation, FomF GmbH, GFO Kliniken Bonn, GILEAD, GSK, Helios Kliniken, Hessisches Landessozialgericht, Infektio Forum, Janssen Cilag GmbH, Klinikum Kassel, Klinikverbund St. Antonius & St. Josef GmbH, Landesärztekammer Hessen, LMU Kliniken, MSD, Pfizer, Streamed up, St. Vincent Hospital, Tillotts, Vivantes Payment made to myself References GBD 2019 Antimicrobial Resistance Collaborators. Global mortality associated with 33 bacterial pathogens in 2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet. 2022 Dec 17;400(10369):2221-2248. doi: 10.1016/S0140-6736(22)02185-7. Epub 2022 Nov 21. PMID: 36423648; PMCID: PMC9763654. van Hal SJ, Jensen SO, Vaska VL, Espedido BA, Paterson DL, Gosbell IB. Predictors of mortality in Staphylococcus aureus Bacteremia. Clin Microbiol Rev. 2012 Apr;25(2):362-86. doi: 10.1128/CMR.05022-11. PMID: 22491776; PMCID: PMC3346297. Westgeest AC, Lambregts MMC, Ruffin F, Korn RE, Webster ME, Kair JL, Parsons JB, Maskarinec SA, Kaplan S, Dekkers OM, de Boer MGJ, Fowler VG Jr, Thaden JT. Female Sex and Mortality in Patients with Staphylococcus aureus Bacteremia: A Systematic Review and Meta-analysis. JAMA Netw Open. 2024 Feb 5;7(2):e240473. doi: 10.1001/jamanetworkopen.2024.0473. PMID: 38411961; PMCID: PMC10900971. Varon B, Palacios-Baena ZR, de Kraker MEA, Rodríguez-Baño J, Leibovici L, Paul M; UNIFORM research group. Universal Risk Factors for Mortality in Bloodstream Infections (UNIFORM): a systematic review and Delphi survey. Clin Microbiol Infect. 2024 Apr;30(4):453-461. doi: 10.1016/j.cmi.2023.12.030. Epub 2024 Jan 3. PMID: 38182050. Roth JA, Tschudin-Sutter S, Dangel M, Frei R, Battegay M, Widmer AF. Value of the Pitt Bacteraemia Score to predict short-term mortality in Staphylococcus aureus bloodstream infection: a validation study. Swiss Med Wkly. 2017 Aug 10;147:w14482. doi: 10.4414/smw.2017.14482. PMID: 28804865. Lê-Bury P, Echenique-Rivera H, Pizarro-Cerdá J, Dussurget O. Determinants of bacterial survival and proliferation in blood. FEMS Microbiol Rev. 2024 May 8;48(3):fuae013. doi: 10.1093/femsre/fuae013. PMID: 38734892; PMCID: PMC11163986. Cheung GYC, Bae JS, Otto M. Pathogenicity and virulence of Staphylococcus aureus . Virulence. 2021 Dec;12(1):547-569. doi: 10.1080/21505594.2021.1878688. PMID: 33522395; PMCID: PMC7872022. Jenul C, Horswill AR. Regulation of Staphylococcus aureus Virulence. Microbiol Spectr. 2019 Apr 5;7(2):10.1128/microbiolspec.gpp3-0031-2018. doi: 10.1128/microbiolspec.GPP3-0031-2018. PMID: 30953424; PMCID: PMC6452892. Ahmad-Mansour N, Loubet P, Pouget C, Dunyach-Remy C, Sotto A, Lavigne JP, Molle V. Staphylococcus aureus Toxins: An Update on Their Pathogenic Properties and Potential Treatments. Toxins (Basel). 2021 Sep 23;13(10):677. doi: 10.3390/toxins13100677. PMID: 34678970; PMCID: PMC8540901. Mlynarczyk-Bonikowska B, Rudnicka L. The Pathogenicity Mechanisms of Staphylococcus aureus . Int J Mol Sci. 2025 Dec 6;26(24):11803. doi: 10.3390/ijms262411803. PMID: 41465232; PMCID: PMC12733317. Aravind L, Anantharaman V, Balaji S, Babu MM, Iyer LM. The many faces of the helix-turn-helix domain: transcription regulation and beyond. FEMS Microbiol Rev. 2005 Apr;29(2):231-62. doi: 10.1016/j.femsre.2004.12.008. PMID: 15808743. Spaan AN, Surewaard BG, Nijland R, van Strijp JA. Neutrophils versus Staphylococcus aureus: a biological tug of war. Annu Rev Microbiol. 2013;67:629-50. doi: 10.1146/annurev-micro-092412-155746. Epub 2013 Jul 3. PMID: 23834243. Lee SO, Lee S, Lee JE, Song KH, Kang CK, Wi YM, San-Juan R, López-Cortés LE, Lacoma A, Prat C, Jang HC, Kim ES, Kim HB, Lee SH. Dysfunctional accessory gene regulator (agr) as a prognostic factor in invasive Staphylococcus aureus infection: a systematic review and meta-analysis. Sci Rep. 2020 Nov 26;10(1):20697. doi: 10.1038/s41598-020-77729-0. PMID: 33244173; PMCID: PMC7691521. Recker M, Laabei M, Toleman MS, Reuter S, Saunderson RB, Blane B, Török ME, Ouadi K, Stevens E, Yokoyama M, Steventon J, Thompson L, Milne G, Bayliss S, Bacon L, Peacock SJ, Massey RC. Clonal differences in Staphylococcus aureus bacteraemia-associated mortality. Nat Microbiol. 2017 Oct;2(10):1381-1388. doi: 10.1038/s41564-017-0001-x. Epub 2017 Aug 7. PMID: 28785103. Campbell AJ, Mowlaboccus S, Coombs GW, Daley DA, Al Yazidi LS, Phuong LK, Leung C, Best EJ, Webb RH, Voss L, Athan E, Britton PN, Bryant PA, Butters CT, Carapetis JR, Ching NS, Francis J, Hung TY, Nourse C, Ojaimi S, Tai A, Vasilunas N, McMullan B, Bowen AC, Blyth CC; Australian and New Zealand Pediatric Infectious Diseases (ANZPID) clinical research network (CRN) of the Australasian Society of Infectious Diseases (ASID) and the Australian Group on Antimicrobial Resistance (AGAR). Whole genome sequencing and molecular epidemiology of paediatric Staphylococcus aureus bacteraemia. J Glob Antimicrob Resist. 2022 Jun;29:197-206. doi: 10.1016/j.jgar.2022.03.012. Epub 2022 Mar 25. PMID: 35342022. Tacconelli E, Göpel S, Gladstone BP, Eisenbeis S, Hölzl F, Buhl M, Górska A, Cattaneo C, Mischnik A, Rieg S, Rohde AM, Kohlmorgen B, Falgenhauer J, Trauth J, Käding N, Kramme E, Biehl LM, Walker SV, Peter S, Gastmeier P, Chakraborty T, Vehreschild MJ, Seifert H, Rupp J, Kern WV; DZIF BLOOMY study group. Development and validation of BLOOMY prediction scores for 14-day and 6-month mortality in hospitalised adults with bloodstream infections: a multicentre, prospective, cohort study. Lancet Infect Dis. 2022 May;22(5):731-741. doi: 10.1016/S1473-3099(21)00587-9. Epub 2022 Jan 19. Erratum in: Lancet Infect Dis. 2022 Apr;22(4):e101. doi: 10.1016/S1473-3099(22)00160-8. PMID: 35065060. Asadollahi P, Farahani NN, Mirzaii M, Khoramrooz SS, van Belkum A, Asadollahi K, Dadashi M, Darban-Sarokhalil D. Distribution of the Most Prevalent Spa Types among Clinical Isolates of Methicillin-Resistant and -Susceptible Staphylococcus aureus around the World: A Review. Front Microbiol. 2018 Feb 12;9:163. doi: 10.3389/fmicb.2018.00163. PMID: 29487578; PMCID: PMC5816571. Delgado V, Ajmone Marsan N, de Waha S, Bonaros N, Brida M, Burri H, Caselli S, Doenst T, Ederhy S, Erba PA, Foldager D, Fosbøl EL, Kovac J, Mestres CA, Miller OI, Miro JM, Pazdernik M, Pizzi MN, Quintana E, Rasmussen TB, Ristić AD, Rodés-Cabau J, Sionis A, Zühlke LJ, Borger MA; ESC Scientific Document Group. 2023 ESC Guidelines for the management of endocarditis. Eur Heart J. 2023 Oct 14;44(39):3948-4042. doi: 10.1093/eurheartj/ehad193. Erratum in: Eur Heart J. 2023 Dec 1;44(45):4780. doi: 10.1093/eurheartj/ehad625. Erratum in: Eur Heart J. 2024 Jan 1;45(1):56. doi: 10.1093/eurheartj/ehad776. Erratum in: Eur Heart J. 2025 Mar 13;46(11):1082. doi: 10.1093/eurheartj/ehae877. PMID: 37622656. GBD 2013 Mortality and Causes of Death Collaborators. Global, regional, and national age-sex specific all-cause and cause-specific mortality for 240 causes of death, 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet. 2015 Jan 10;385(9963):117-71. doi: 10.1016/S0140-6736(14)61682-2. Epub 2014 Dec 18. PMID: 25530442; PMCID: PMC4340604. D'Hoore W, Sicotte C, Tilquin C. Risk adjustment in outcome assessment: the Charlson comorbidity index. Methods Inf Med. 1993 Nov;32(5):382-7. PMID: 8295545. Xiao M, Zhao R, Zhang Q, Fan X, O'Sullivan MV, Li DF, Wang XY, Wu HL, Kong F, Xu YC. Genotypic Diversity of Staphylococcus aureus α-Hemolysin Gene (hla) and Its Association with Clonal Background: Implications for Vaccine Development. PLoS One. 2016 Feb 11;11(2):e0149112. doi: 10.1371/journal.pone.0149112. PMID: 26866483; PMCID: PMC4750931. Roder T, Pimentel G, Fuchsmann P, Stern MT, von Ah U, Vergères G, Peischl S, Brynildsrud O, Bruggmann R, Bär C. Scoary2: rapid association of phenotypic multi-omics data with microbial pan-genomes. Genome Biol. 2024 Apr 11;25(1):93. doi: 10.1186/s13059-024-03233-7. PMID: 38605417; PMCID: PMC11007987. Jolley KA, Bray JE, Maiden MCJ. Open-access bacterial population genomics: BIGSdb software, the PubMLST.org website and their applications. Wellcome Open Res. 2018 Sep 24;3:124. doi: 10.12688/wellcomeopenres.14826.1. PMID: 30345391; PMCID: PMC6192448. Sebulsky MT, Speziali CD, Shilton BH, Edgell DR, Heinrichs DE. FhuD1, a ferric hydroxamate-binding lipoprotein in Staphylococcus aureus: a case of gene duplication and lateral transfer. J Biol Chem. 2004 Dec 17;279(51):53152-9. doi: 10.1074/jbc.M409793200. Epub 2004 Oct 8. PMID: 15475351. Endicott NP, Lee E, Wencewicz TA. Structural Basis for Xenosiderophore Utilization by the Human Pathogen Staphylococcus aureus. ACS Infect Dis. 2017 Jul 14;3(7):542-553. doi: 10.1021/acsinfecdis.7b00036. Epub 2017 May 30. PMID: 28505405. Sebulsky MT, Heinrichs DE. Identification and characterization of fhuD1 and fhuD2, two genes involved in iron-hydroxamate uptake in Staphylococcus aureus. J Bacteriol. 2001 Sep;183(17):4994-5000. doi: 10.1128/JB.183.17.4994-5000.2001. PMID: 11489851; PMCID: PMC95374. Mishra RP, Mariotti P, Fiaschi L, Nosari S, Maccari S, Liberatori S, Fontana MR, Pezzicoli A, De Falco MG, Falugi F, Altindis E, Serruto D, Grandi G, Bagnoli F. Staphylococcus aureus FhuD2 is involved in the early phase of staphylococcal dissemination and generates protective immunity in mice. J Infect Dis. 2012 Oct 1;206(7):1041-9. doi: 10.1093/infdis/jis463. Epub 2012 Jul 24. PMID: 22829645. Arifin AJ, Hannauer M, Welch I, Heinrichs DE. Deferoxamine mesylate enhances virulence of community-associated methicillin resistant Staphylococcus aureus. Microbes Infect. 2014 Nov;16(11):967-72. doi: 10.1016/j.micinf.2014.09.003. Epub 2014 Sep 22. PMID: 25251026. Xia P, Lian S, Wu Y, Yan L, Quan G, Zhu G. Zinc is an important inter-kingdom signal between the host and microbe. Vet Res. 2021 Mar 4;52(1):39. doi: 10.1186/s13567-021-00913-1. PMID: 33663613; PMCID: PMC7931793. Grim KP, San Francisco B, Radin JN, Brazel EB, Kelliher JL, Párraga Solórzano PK, Kim PC, McDevitt CA, Kehl-Fie TE. The Metallophore Staphylopine Enables Staphylococcus aureus To Compete with the Host for Zinc and Overcome Nutritional Immunity. mBio. 2017 Oct 31;8(5):e01281-17. doi: 10.1128/mBio.01281-17. PMID: 29089427; PMCID: PMC5666155. Kehl-Fie TE, Zhang Y, Moore JL, Farrand AJ, Hood MI, Rathi S, Chazin WJ, Caprioli RM, Skaar EP. MntABC and MntH contribute to systemic Staphylococcus aureus infection by competing with calprotectin for nutrient manganese. Infect Immun. 2013 Sep;81(9):3395-405. doi: 10.1128/IAI.00420-13. Epub 2013 Jul 1. PMID: 23817615; PMCID: PMC3754211. Shearer JE, Wireman J, Hostetler J, Forberger H, Borman J, Gill J, Sanchez S, Mankin A, Lamarre J, Lindsay JA, Bayles K, Nicholson A, O'Brien F, Jensen SO, Firth N, Skurray RA, Summers AO. Major families of multiresistant plasmids from geographically and epidemiologically diverse staphylococci. G3 (Bethesda). 2011 Dec;1(7):581-91. doi: 10.1534/g3.111.000760. Epub 2011 Dec 1. PMID: 22384369; PMCID: PMC3276174. Rasigade JP, Leclère A, Alla F, Tessier A, Bes M, Lechiche C, Vernet-Garnier V, Laouénan C, Vandenesch F, Leport C; AEPEI Study Group. Staphylococcus aureus CC30 Lineage and Absence of sed , j , r -Harboring Plasmid Predict Embolism in Infective Endocarditis. Front Cell Infect Microbiol. 2018 Jun 8;8:187. doi: 10.3389/fcimb.2018.00187. PMID: 29938201; PMCID: PMC6003251. Nienaber JJ, Sharma Kuinkel BK, Clarke-Pearson M, Lamlertthon S, Park L, Rude TH, Barriere S, Woods CW, Chu VH, Marín M, Bukovski S, Garcia P, Corey GR, Korman T, Doco-Lecompte T, Murdoch DR, Reller LB, Fowler VG Jr; International Collaboration on Endocarditis-Microbiology Investigators. Methicillin-susceptible Staphylococcus aureus endocarditis isolates are associated with clonal complex 30 genotype and a distinct repertoire of enterotoxins and adhesins. J Infect Dis. 2011 Sep 1;204(5):704-13. doi: 10.1093/infdis/jir389. Erratum in: J Infect Dis. 2013 Feb 1;207(3):546. PMID: 21844296; PMCID: PMC3156104. Al-Tohamy A, Grove A. Targeting bacterial transcription factors for infection control: opportunities and challenges. Transcription. 2025 Feb;16(1):141-168. doi: 10.1080/21541264.2023.2293523. Epub 2023 Dec 21. PMID: 38126125; PMCID: PMC11970743. Chen PR, Nishida S, Poor CB, Cheng A, Bae T, Kuechenmeister L, Dunman PM, Missiakas D, He C. A new oxidative sensing and regulation pathway mediated by the MgrA homologue SarZ in Staphylococcus aureus. Mol Microbiol. 2009 Jan;71(1):198-211. doi: 10.1111/j.1365-2958.2008.06518.x. Epub 2008 Nov 5. PMID: 19007410; PMCID: PMC2698432. Deochand DK, Grove A. MarR family transcription factors: dynamic variations on a common scaffold. Crit Rev Biochem Mol Biol. 2017 Dec;52(6):595-613. doi: 10.1080/10409238.2017.1344612. Epub 2017 Jul 3. PMID: 28670937. Novick RP, Ram G. Staphylococcal pathogenicity islands-movers and shakers in the genomic firmament. Curr Opin Microbiol. 2017 Aug;38:197-204. doi: 10.1016/j.mib.2017.08.001. Epub 2017 Nov 1. PMID: 29100762; PMCID: PMC5884141. Ingmer H, Gerlach D, Wolz C. Temperate Phages of Staphylococcus aureus . Microbiol Spectr. 2019 Sep;7(5):10.1128/microbiolspec.gpp3-0058-2018. doi: 10.1128/microbiolspec.GPP3-0058-2018. PMID: 31562736; PMCID: PMC10921950. Sassi M, Bronsard J, Pascreau G, Emily M, Donnio PY, Revest M, Felden B, Wirth T, Augagneur Y. Forecasting Staphylococcus aureus Infections Using Genome-Wide Association Studies, Machine Learning, and Transcriptomic Approaches. mSystems. 2022 Aug 30;7(4):e0037822. doi: 10.1128/msystems.00378-22. Epub 2022 Jul 5. PMID: 35862809; PMCID: PMC9426533. Beganovic M, Cusumano JA, Lopes V, LaPlante KL, Caffrey AR. Comparative Effectiveness of Exclusive Exposure to Nafcillin or Oxacillin, Cefazolin, Piperacillin/Tazobactam, and Fluoroquinolones Among a National Cohort of Veterans With Methicillin-Susceptible Staphylococcus aureus Bloodstream Infection. Open Forum Infect Dis. 2019 Jun 6;6(7):ofz270. doi: 10.1093/ofid/ofz270. PMID: 31281864; PMCID: PMC6602898. Petit RA 3rd, Read TD. Bactopia: a Flexible Pipeline for Complete Analysis of Bacterial Genomes. mSystems. 2020 Aug 4;5(4):e00190-20. doi: 10.1128/mSystems.00190-20. PMID: 32753501; PMCID: PMC7406220. Wick RR, Judd LM, Gorrie CL, Holt KE. Unicycler: Resolving bacterial genome assemblies from short and long sequencing reads. PLoS Comput Biol. 2017 Jun 8;13(6):e1005595. doi: 10.1371/journal.pcbi.1005595. PMID: 28594827; PMCID: PMC5481147. Schwengers O, Jelonek L, Dieckmann MA, Beyvers S, Blom J, Goesmann A. Bakta: rapid and standardized annotation of bacterial genomes via alignment-free sequence identification. Microb Genom. 2021 Nov;7(11):000685. doi: 10.1099/mgen.0.000685. PMID: 34739369; PMCID: PMC8743544. Ozer EA, Allen JP, Hauser AR. Characterization of the core and accessory genomes of Pseudomonas aeruginosa using bioinformatic tools Spine and AGEnt. BMC Genomics. 2014 Aug 29;15(1):737. doi: 10.1186/1471-2164-15-737. PMID: 25168460; PMCID: PMC4155085. Wishart DS, Han S, Saha S, Oler E, Peters H, Grant JR, Stothard P, Gautam V. PHASTEST: faster than PHASTER, better than PHAST. Nucleic Acids Res. 2023 Jul 5;51(W1):W443-W450. doi: 10.1093/nar/gkad382. PMID: 37194694; PMCID: PMC10320120. Danecek P, Auton A, Abecasis G, Albers CA, Banks E, DePristo MA, Handsaker RE, Lunter G, Marth GT, Sherry ST, McVean G, Durbin R; 1000 Genomes Project Analysis Group. The variant call format and VCFtools. Bioinformatics. 2011 Aug 1;27(15):2156-8. doi: 10.1093/bioinformatics/btr330. Epub 2011 Jun 7. PMID: 21653522; PMCID: PMC3137218. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G, Durbin R; 1000 Genome Project Data Processing Subgroup. The Sequence Alignment/Map format and SAMtools. Bioinformatics. 2009 Aug 15;25(16):2078-9. doi: 10.1093/bioinformatics/btp352. Epub 2009 Jun 8. PMID: 19505943; PMCID: PMC2723002. McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, Garimella K, Altshuler D, Gabriel S, Daly M, DePristo MA. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 2010 Sep;20(9):1297-303. doi: 10.1101/gr.107524.110. Epub 2010 Jul 19. PMID: 20644199; PMCID: PMC2928508. Nguyen LT, Schmidt HA, von Haeseler A, Minh BQ. IQ-TREE: a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol Biol Evol. 2015 Jan;32(1):268-74. doi: 10.1093/molbev/msu300. Epub 2014 Nov 3. PMID: 25371430; PMCID: PMC4271533. Szklarczyk D, Kirsch R, Koutrouli M, Nastou K, Mehryary F, Hachilif R, Gable AL, Fang T, Doncheva NT, Pyysalo S, Bork P, Jensen LJ, von Mering C. The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res. 2023 Jan 6;51(D1):D638-D646. doi: 10.1093/nar/gkac1000. PMID: 36370105; PMCID: PMC9825434. Table Table 1 is available in the Supplementary Files section. Additional Declarations Yes there is potential Competing Interest. Maria JGT Vehreschild declares to have the following conflicts of interest: Grants or contracts from any entity: MSD, Heel, Roche, Tillotts, Pfizer Payment made to institution Consulting fees GILEAD, Tillotts, Pfizer, Bioaster, GSK, Ecraid, EUMEDICA, Bactolife, PAION Payment made to myself Payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing or educational events Akademie für Infektionsmedizin, Astra Zeneca, bioMerieux, Biotest, DGI, European Society of Neurogastroenterology, Falk Foundation, FomF GmbH, GFO Kliniken Bonn, GILEAD, GSK, Helios Kliniken, Hessisches Landessozialgericht, Infektio Forum, Janssen Cilag GmbH, Klinikum Kassel, Klinikverbund St. Antonius & St. Josef GmbH, Landesärztekammer Hessen, LMU Kliniken, MSD, Pfizer, Streamed up, St. Vincent Hospital, Tillotts, Vivantes Payment made to myself Supplementary Files SupplementaryFigS12Table.docx SupplTable3allUniRef90XUniRef100FINAL.xlsx Supplementary Table 3: Full list of significantly associated traits on in-hospital mortality in S. aureus bacteremia, including their abundance, type of association, their trait associated in-hospital mortality rate, suggested site of action, rank achieved in the Scoary2 pipeline and clustering with other traits in the absence/presence matrix. SupplTable4ResultsScoary2inclfinalphylogenyUniRef100Inputcompletelywithoutourmodel.xlsx Supplementary Table 4: Output of the Scoary2 analytic pipeline after incorporating the same input used for the training cohort in model A and feeding the maximum likelihood tree of the analyzed cohort for phylogenetic background correction. Significant UniRef100 factors identified in Scoary2 (represented by p≤0.05 in fq*ep) are colored in orange, 77 significant UniRef100 trait of our statistical model are colored in blue. SupplTable5presenceabsencematrixUniRef100inclMLSTMRSAsurvivalsortedfinal.xlsx Supplementary Table 5: Absence-presence matrix of UniRef100 traits across the study isolates. Columns are sorted traits associated with increased in-hospital mortality (left, red headings) and decreased in-hospital mortality (right, green headings). From left to right, traits associated with increased in-hospital mortality are arranged in ascending abundance, and traits associated with decreased in-hospital mortality in descending abundance. Arrows indicate the top 10 traits. The second spreadsheet displays contig localization of traits with start and end nucleotide. SupplTable6clinicalmetadata.xlsx Supplementary Table 6: Clinical metadata used for the analysis. The second spreadsheet lists the field names with their respective meaning. (NA = information not available) Table1.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9449451","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":632435209,"identity":"13b8e7c6-0ac9-418c-89f6-bee0b598026e","order_by":0,"name":"Kristina Schmauder","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0002-1705-0253","institution":"University Hospital Tübingen","correspondingAuthor":true,"prefix":"","firstName":"Kristina","middleName":"","lastName":"Schmauder","suffix":""},{"id":632435210,"identity":"5cc70189-97c4-463b-98e8-d3beaed72c7e","order_by":1,"name":"Ulrich Schoppmeier","email":"","orcid":"","institution":"University Hospital Tübingen","correspondingAuthor":false,"prefix":"","firstName":"Ulrich","middleName":"","lastName":"Schoppmeier","suffix":""},{"id":632435211,"identity":"3b142b8f-09a3-4140-b993-b65ff6e13802","order_by":2,"name":"Jennifer Müller","email":"","orcid":"https://orcid.org/0000-0002-7907-2669","institution":"University Hospital Tübingen","correspondingAuthor":false,"prefix":"","firstName":"Jennifer","middleName":"","lastName":"Müller","suffix":""},{"id":632435212,"identity":"a7c35c9b-55da-4942-b545-6e62c0820728","order_by":3,"name":"Baris Bader","email":"","orcid":"","institution":"University Hospital Tübingen","correspondingAuthor":false,"prefix":"","firstName":"Baris","middleName":"","lastName":"Bader","suffix":""},{"id":632435213,"identity":"1685206f-3e35-40ae-9b77-711e14570fa9","order_by":4,"name":"Thales Kronenberger","email":"","orcid":"https://orcid.org/0000-0001-6933-7590","institution":"University Hospital of Tübingen","correspondingAuthor":false,"prefix":"","firstName":"Thales","middleName":"","lastName":"Kronenberger","suffix":""},{"id":632435214,"identity":"28bbbe9c-a27d-4440-8e6f-cf325e36a176","order_by":5,"name":"Beryl Gladstone","email":"","orcid":"","institution":"University Hospital Tübingen","correspondingAuthor":false,"prefix":"","firstName":"Beryl","middleName":"","lastName":"Gladstone","suffix":""},{"id":632435215,"identity":"af1343b1-5a2e-451d-83f0-3380851e3afe","order_by":6,"name":"Kyriaki Xanthopoulou","email":"","orcid":"","institution":"University Hospital Cologne","correspondingAuthor":false,"prefix":"","firstName":"Kyriaki","middleName":"","lastName":"Xanthopoulou","suffix":""},{"id":632435216,"identity":"783bb60f-7a03-4bd8-b8a0-524a491d0bac","order_by":7,"name":"Can Imirzalioglu","email":"","orcid":"","institution":"Institute of Medical Microbiology, German Centre of Infection Research, Site Giessen-Marburg-Langen, Justus-Liebig-University Giessen, Giessen, Germany.","correspondingAuthor":false,"prefix":"","firstName":"Can","middleName":"","lastName":"Imirzalioglu","suffix":""},{"id":632435217,"identity":"7d147d05-8827-49af-b065-3e50104dbf7f","order_by":8,"name":"Nadja Käding","email":"","orcid":"","institution":"University of Lübeck","correspondingAuthor":false,"prefix":"","firstName":"Nadja","middleName":"","lastName":"Käding","suffix":""},{"id":632435218,"identity":"21e8ba37-bd6e-4098-9f0e-78d3fed66d54","order_by":9,"name":"Simone Eisenbeis","email":"","orcid":"","institution":"University Hospital Tübingen","correspondingAuthor":false,"prefix":"","firstName":"Simone","middleName":"","lastName":"Eisenbeis","suffix":""},{"id":632435219,"identity":"6e5f6adc-d689-4fc7-b039-4765846ff566","order_by":10,"name":"Siri Göpel","email":"","orcid":"https://orcid.org/0000-0002-7666-4634","institution":"University Hospital Tübingen","correspondingAuthor":false,"prefix":"","firstName":"Siri","middleName":"","lastName":"Göpel","suffix":""},{"id":632435220,"identity":"dc8f50a2-0f38-4f06-9415-3c13d241fdb4","order_by":11,"name":"Christine Geffers","email":"","orcid":"","institution":"Charité-Universitätsmedizin Berlin","correspondingAuthor":false,"prefix":"","firstName":"Christine","middleName":"","lastName":"Geffers","suffix":""},{"id":632435221,"identity":"5dcfbbff-c757-4d72-8a5d-78922214a2d4","order_by":12,"name":"Harald Seifert","email":"","orcid":"","institution":"University Hospital Cologne","correspondingAuthor":false,"prefix":"","firstName":"Harald","middleName":"","lastName":"Seifert","suffix":""},{"id":632435222,"identity":"02380b7d-e32f-4bb9-9f8e-2723fa8c257e","order_by":13,"name":"Trinad Chakraborty","email":"","orcid":"https://orcid.org/0000-0002-1452-4759","institution":"Justus Liebig University Giessen","correspondingAuthor":false,"prefix":"","firstName":"Trinad","middleName":"","lastName":"Chakraborty","suffix":""},{"id":632435223,"identity":"941e66c9-be6f-41a4-9b5b-038ce864a630","order_by":14,"name":"Jan Rupp","email":"","orcid":"https://orcid.org/0000-0001-8722-1233","institution":"University of Luebeck","correspondingAuthor":false,"prefix":"","firstName":"Jan","middleName":"","lastName":"Rupp","suffix":""},{"id":632435224,"identity":"76a70956-9e9a-4bca-9ed3-3030ad444009","order_by":15,"name":"Maria Vehreschild","email":"","orcid":"","institution":"University of Cologne, Faculty of Medicine and University Hospital of Cologne, Department I of Internal Medicine","correspondingAuthor":false,"prefix":"","firstName":"Maria","middleName":"","lastName":"Vehreschild","suffix":""},{"id":632435225,"identity":"5f986b12-e68f-4cc3-9e0e-d0ef714a54b4","order_by":16,"name":"Siegbert Rieg","email":"","orcid":"","institution":"University of Freiburg","correspondingAuthor":false,"prefix":"","firstName":"Siegbert","middleName":"","lastName":"Rieg","suffix":""},{"id":632435226,"identity":"37c6026d-078c-4272-bad3-18ee428e7399","order_by":17,"name":"Paul Higgins","email":"","orcid":"","institution":"University Hospital Cologne","correspondingAuthor":false,"prefix":"","firstName":"Paul","middleName":"","lastName":"Higgins","suffix":""},{"id":632435227,"identity":"bd811dba-fd7c-4259-a18a-df0c97cef548","order_by":18,"name":"Evelyn Kramme","email":"","orcid":"","institution":"University Hospital Lübeck","correspondingAuthor":false,"prefix":"","firstName":"Evelyn","middleName":"","lastName":"Kramme","suffix":""},{"id":632435228,"identity":"17ee4ce9-7d3c-4fa0-948b-ef1c6be82cf4","order_by":19,"name":"Lena Biehl","email":"","orcid":"","institution":"University of Cologne, Faculty of Medicine and University Hospital of Cologne, Department I of Internal Medicine","correspondingAuthor":false,"prefix":"","firstName":"Lena","middleName":"","lastName":"Biehl","suffix":""},{"id":632435229,"identity":"b2414fed-3795-40d5-823c-2645a77cc2ae","order_by":20,"name":"Libera Lo Presti","email":"","orcid":"","institution":"University Hospital Tübingen","correspondingAuthor":false,"prefix":"","firstName":"Libera","middleName":"Lo","lastName":"Presti","suffix":""},{"id":632435230,"identity":"41bad9b8-76fb-4c75-9b43-de86c22698e8","order_by":21,"name":"Matthias Willmann","email":"","orcid":"","institution":"SYNLAB MVZ Leinfelden GmbH","correspondingAuthor":false,"prefix":"","firstName":"Matthias","middleName":"","lastName":"Willmann","suffix":""},{"id":632435231,"identity":"a7f2ad05-24d0-40cf-a16a-ab025742a1e1","order_by":22,"name":"Andreas Peschel","email":"","orcid":"https://orcid.org/0000-0002-3209-8626","institution":"Interfaculty Institute ofUniversity of Tübingen, Tübingen, Germany; Cluster of Excellence 'Controlling Microbes to Fight Infections', University o","correspondingAuthor":false,"prefix":"","firstName":"Andreas","middleName":"","lastName":"Peschel","suffix":""},{"id":632435232,"identity":"9a1ef6f9-a8ee-4092-9f6d-0d65492375e3","order_by":23,"name":"Winfried Kern","email":"","orcid":"","institution":"University Hospital Freiburg","correspondingAuthor":false,"prefix":"","firstName":"Winfried","middleName":"","lastName":"Kern","suffix":""},{"id":632435233,"identity":"8c107c46-d6bf-40c2-b2cd-95cc48d13bcb","order_by":24,"name":"Evelina Tacconelli","email":"","orcid":"","institution":"Verona University","correspondingAuthor":false,"prefix":"","firstName":"Evelina","middleName":"","lastName":"Tacconelli","suffix":""},{"id":632435234,"identity":"a35d611a-b90f-4715-b860-f2bd6883a36c","order_by":25,"name":"Silke Peter","email":"","orcid":"","institution":"Institute of Medical Microbiology and Hygiene, University of Tübingen, Tübingen, Germany","correspondingAuthor":false,"prefix":"","firstName":"Silke","middleName":"","lastName":"Peter","suffix":""}],"badges":[],"createdAt":"2026-04-17 13:02:37","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9449451/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9449451/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108410835,"identity":"4e051bd2-dbb8-4363-a419-61487009fa02","added_by":"auto","created_at":"2026-05-04 10:11:02","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":101565,"visible":true,"origin":"","legend":"\u003cp\u003eComposition of cases of \u003cem\u003eStaphylococcus aureus\u003c/em\u003e bloodstream infections within the BLOOMY and BLOOMY-PREDICT study. Of 970 documented cases, 327 cases had to be excluded due to unavailability of isolates or polymicrobial bloodstream infection. This resulted in a total of 643 sequenced \u003cem\u003eS. aureus \u003c/em\u003eisolates from monomicrobial bloodstream infections to analyze further.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/f5fdf8eafb83bebc9e508c6f.png"},{"id":108410834,"identity":"8a578c81-b021-4ab4-b83f-0fb6e8ab9895","added_by":"auto","created_at":"2026-05-04 10:11:02","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":262963,"visible":true,"origin":"","legend":"\u003cp\u003eRooted phylogenetic tree of all included \u003cem\u003eStaphylococcus aureus\u003c/em\u003eisolates. Sub-circles and charts illustrate the assignment of isolates to the training vs. validation cohort, the study cohort, the patient’s survival status, study site and MLST type, demonstrating a heterogeneous distribution of MLST\u003cem\u003e \u003c/em\u003etypes among different study sites. Tree scales are projected as 0.01.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/812a8b49224e38f5d20bd2ed.png"},{"id":108492741,"identity":"908e6562-b916-4c8c-bdf6-774f6116e59d","added_by":"auto","created_at":"2026-05-05 09:58:30","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":164206,"visible":true,"origin":"","legend":"\u003cp\u003eStepwise process used for statistical analysis. BMI = Body Mass Index; CCI = Charlson Comorbidity Index; PBS = Pitt Bacteremia Score; ANOVA = Analysis of Variance; AUROC = Area under the receiver operating characteristic.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/d4c42928c2622072cbf8c5e4.png"},{"id":108410844,"identity":"2f37075d-4fce-4f20-9155-00f70385a43d","added_by":"auto","created_at":"2026-05-04 10:11:05","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":149797,"visible":true,"origin":"","legend":"\u003cp\u003eResults for Model A (UniRef100) and B (UniRef90) after repeated ANOVA (analysis of variance), each reduced by the displayed UniRef trait or clinical factor with the least contribution in the previous cycle of the ANOVA analysis. The dots represent the subsequently computed AUROC (Area under the receiver operating characteristic) values (training cohort = blue; validation cohort = orange) for each set of features while the spreads display the upper and lower 95% confidence intervals. Notably, some UniRef traits co-occurring in the exact same isolates had to be collapsed to one vicarious trait to enable ANOVA analysis. For the sake of visibility, only the vicarious trait is indicated. For Model A, UniRef100_D2J5W0 (position 67) also represents UniRef100_D2J5W8, UniRef100_D2J5Y9, UniRef100_D2J5Z4, UniRef100_O85217 and UniRef100_Q8VVU3). For Model B, UniRef90_A0A0U1MRA0 (position 9) also represents UniRef90_A0A133QB87; UniRef90_Q76LS7 (position 19) also represents UniRef90_Q76LS8; UniRef90_A0A1D0BP50 (position 24) also represents UniRef90_D2J5W8 and UniRef90_D2J5Z4; UniRef90_D2J5Y0 (position 33) also represents UniRef90_E4PYM.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/2eafae9bbba57309b226037e.png"},{"id":108494675,"identity":"749fabfd-6eb0-4994-9f53-084a5b763958","added_by":"auto","created_at":"2026-05-05 10:06:29","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":901899,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/c321ce15-f75e-4e48-955f-fdbd7d90097f.pdf"},{"id":108410833,"identity":"6cdc2840-0d50-442a-aba1-779e2d220dc5","added_by":"auto","created_at":"2026-05-04 10:11:02","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":706212,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigS12Table.docx","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/307e17d2702e82baf7c854cd.docx"},{"id":108410840,"identity":"c8a5f274-70c4-49ac-92a2-8aeb6548671d","added_by":"auto","created_at":"2026-05-04 10:11:03","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":61177,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Table 3: \u003c/strong\u003eFull list of significantly associated traits on in-hospital mortality in \u003cem\u003eS. aureus\u003c/em\u003ebacteremia, including their abundance, type of association, their trait associated in-hospital mortality rate, suggested site of action, rank achieved in the Scoary2 pipeline and clustering with other traits in the absence/presence matrix.\u003c/p\u003e","description":"","filename":"SupplTable3allUniRef90XUniRef100FINAL.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/46580a5b3cdcfc7694f584e0.xlsx"},{"id":108493280,"identity":"dd61ebcb-6c60-4869-ad69-da92e2bbb222","added_by":"auto","created_at":"2026-05-05 09:59:51","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":4085031,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Table 4:\u003c/strong\u003e Output of the Scoary2 analytic pipeline after incorporating the same input used for the training cohort in model A and feeding the maximum likelihood tree of the analyzed cohort for phylogenetic background correction. Significant UniRef100 factors identified in Scoary2 (represented by p≤0.05 in fq*ep) are colored in orange, 77 significant UniRef100 trait of our statistical model are colored in blue.\u003c/p\u003e","description":"","filename":"SupplTable4ResultsScoary2inclfinalphylogenyUniRef100Inputcompletelywithoutourmodel.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/d9beee18e197aaedd80f4be6.xlsx"},{"id":108410838,"identity":"8b0cdf02-801b-4df5-b52c-8e1978c29a23","added_by":"auto","created_at":"2026-05-04 10:11:02","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":856732,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Table 5: \u003c/strong\u003eAbsence-presence matrix of UniRef100 traits across the study isolates. Columns are sorted traits associated with increased in-hospital mortality (left, red headings) and decreased in-hospital mortality (right, green headings). From left to right, traits associated with increased in-hospital mortality are arranged in ascending abundance, and traits associated with decreased in-hospital mortality in descending abundance. Arrows indicate the top 10 traits. The second spreadsheet displays contig localization of traits with start and end nucleotide.\u003c/p\u003e","description":"","filename":"SupplTable5presenceabsencematrixUniRef100inclMLSTMRSAsurvivalsortedfinal.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/dc2a80aa094c825cb148f3cf.xlsx"},{"id":108410841,"identity":"f2769c3f-345c-432a-8903-23e916005b7b","added_by":"auto","created_at":"2026-05-04 10:11:03","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":343452,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Table 6: \u003c/strong\u003eClinical metadata used for the analysis. The second spreadsheet lists the field names with their respective meaning. (NA = information not available)\u003c/p\u003e","description":"","filename":"SupplTable6clinicalmetadata.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/24c389adfed4f4c8a07f0a5c.xlsx"},{"id":108410839,"identity":"06070744-2d31-4ef1-92a4-4c2c561cbf9c","added_by":"auto","created_at":"2026-05-04 10:11:02","extension":"docx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":26521,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-9449451/v1/23446f328c21dd6a16edba6a.docx"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nMaria JGT Vehreschild declares to have the following conflicts of interest:\r\nGrants or contracts from any entity:\r\nMSD, Heel, Roche, Tillotts, Pfizer \r\nPayment made to institution\r\n\r\nConsulting fees\r\nGILEAD, Tillotts, Pfizer, Bioaster, GSK, Ecraid, EUMEDICA, Bactolife, PAION \r\nPayment made to myself\r\n\r\nPayment or honoraria for lectures, presentations, speakers bureaus, manuscript writing or educational events\r\nAkademie für Infektionsmedizin, Astra Zeneca, bioMerieux, Biotest, DGI, European Society of Neurogastroenterology, Falk Foundation, FomF GmbH, GFO Kliniken Bonn, GILEAD, GSK, Helios Kliniken, Hessisches Landessozialgericht, Infektio Forum, Janssen Cilag GmbH, Klinikum Kassel, Klinikverbund St. Antonius \u0026 St. Josef GmbH, Landesärztekammer Hessen, LMU Kliniken, MSD, Pfizer, Streamed up, St. Vincent Hospital, Tillotts, Vivantes \r\nPayment made to myself","formattedTitle":"Multicentric bloodstream infection cohort study reveals new potential Staphylococcus aureus virulence factors influencing in-hospital mortality","fulltext":[{"header":"BACKGROUND","content":"\u003cp\u003e\u003cem\u003eStaphylococcus aureus\u003c/em\u003e is among the most important human pathogens causing a large variety of infections. In 2019 alone, over 1 million deaths were associated with \u003cem\u003eS. aureus\u003c/em\u003e infections globally. Among the clinical manifestations with a deadly outcome due to \u003cem\u003eS. aureus\u003c/em\u003e,\u003cem\u003e\u0026nbsp;\u003c/em\u003ebloodstream infections (BSI) are at the forefront [1]. Clinical risk factors for mortality in BSI have been studied extensively revealing sex, age, high count in comorbidity scores (\u003cem\u003ee.g.,\u003c/em\u003e Charlson Comorbidity Score (CCI)), acquisition (community vs. hospital-acquisition) and focus as prominent determinants of mortality risk in \u003cem\u003eS. aureus\u003c/em\u003e BSI [2,3,4]. Additionally, disease severity at presentation represents an independent variable influencing patient outcome, which has resulted in the establishment of BSI risk scores based on clinical parameters. One well-validated, discriminative score is the Pitt bacteremia score (PBS) which assesses body temperature, mental status, presence of hypotension, mechanical ventilation and cardiac arrest during the initial clinical evaluation [4,5].\u003c/p\u003e\n\u003cp\u003eGiven its clinical relevance and high burden of disease, \u003cem\u003eS. aureus\u003c/em\u003e has been the subject of intensive research for decades. Numerous virulence factors enable \u003cem\u003eS. aureus\u003c/em\u003e to adapt to different niches and metabolic environments. Surprisingly, even in an environment like the human bloodstream, \u003cem\u003eS. aureus\u003c/em\u003e can overcome host immune responses and spread metastatically to various body sites. In the bloodstream, bacteria are confronted with the complement system, phagocytosing and antibody-producing immune cells and their armory of reactive oxygen species, proteases and antimicrobial substances (like defensins), as well as therapeutic antibiotics. Moreover, the battle for nutrients (macro- and especially micronutrients like iron, manganese, zinc) challenges bacterial survival [6]. To ensure survival in harsh environments, \u003cem\u003eS. aureus\u003c/em\u003e has evolved immune-evasion strategies (\u003cem\u003ee.g.\u003c/em\u003e, staphylococcal protein A (spA), staphylococcal superantigen-like proteins (SSLs), chemotaxis-inhibitory protein of \u003cem\u003eS. aureus\u003c/em\u003e (CHIPS), leucocidins, phenol-soluble modulins (PSMs), proteases), the ability to produce toxins (\u003cem\u003ee.g.\u003c/em\u003e, hemolysins, exfoliative toxins ETA-ETE, staphylococcal enterotoxins), adhesion molecules, biofilms, and the ability to induce abscesses (\u003cem\u003ee.g.\u003c/em\u003e, fibronectin-binding proteins, polysaccharide intercellular adhesin (PIA), coagulases, fibrinogen-binding proteins, Panton-Valentine-Leucocidin (PVL)), and metabolic singularities (\u003cem\u003ee.g.\u003c/em\u003e, metal capturing strategies, nucleotide and amino acid prototrophy). Moreover, flexible gene regulation of virulence factors under cell stress and antimicrobial pressure is ensured by two component systems such as AgrAC (belonging to the quorum-sensing accessory genome regulator agr) and diverse transcription factors [2,6,7,8,9,10,11].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe relevance of specific virulence factors for human infections has remained largely unclear because many of them are strictly human-specific and cannot be assessed in animal models [12]. Yet, some pathogenicity factors like the superantigen TSST-1 encoded by \u003cem\u003etst\u003c/em\u003e and an undisrupted beta-hemolysin gene (\u003cem\u003ehlb\u003c/em\u003e) have been linked to severe disease progression during systemic \u003cem\u003eS. aureus\u003c/em\u003e infections in humans\u0026nbsp;[2,7,9]. Dysfunction in Agr signaling correlated with an adverse outcome in pneumonia and MRSA (methicillin-resistant \u003cem\u003eStaphylococcus aureus\u003c/em\u003e) infections in a big meta-analysis but not in endocarditis or central-line associated BSI, where a dysfunctional Agr was even beneficial\u0026nbsp;[13]. A certain \u003cem\u003ecapA\u003c/em\u003e gene polymorphism was found to influence mortality in a bacteremia study investigating two common \u003cem\u003eS. aureus\u003c/em\u003e clonal complexes (CC20 and CC30) after taking some clinical factors into account [14]. PVL producing strains promote an increased mortality in necrotizing pneumonia, but data remain inconclusive for \u003cem\u003eS. aureus\u003c/em\u003e BSI [2,15]. Overexpression of \u003cem\u003ehla\u0026nbsp;\u003c/em\u003e(\u0026alpha;-hemolysin) and the presence of the staphylococcal enterotoxin C (Sec)\u0026nbsp;were linked to an adverse outcome in animal models of pneumonia and sepsis [9]. Overall, many of the pathogen factors found to be associated with clinical outcomes were derived from rather small case series, laboratory experiments, often involving only a limited number of strains and animal models, or were not adjusted to relevant host factors known to affect the prognosis.\u003c/p\u003e\n\u003cp\u003eThe BLOOMY study (BLOodstream Infection due to multidrug-resistant and susceptible Organisms Multicenter studY) and the follow-up project BLOOMY-PREDICT investigated the epidemiological, clinical and microbiological aspects of BSI due to \u003cem\u003eS. aureus\u003c/em\u003e and other bacteria in hospitalized adult patients [16]. Here, 643 clinical \u003cem\u003eS. aureus\u003c/em\u003e isolates from monomicrobial index blood cultures were analyzed, and the prognostic value of presence/absence of genetic factors for in-hospital mortality was assessed to potentially serve as biomarkers in \u003cem\u003eS. aureus\u003c/em\u003e BSI.\u003c/p\u003e"},{"header":"RESULTS","content":"\u003cp\u003e\u003cstrong\u003eAnalysis of cohort composition\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWithin the BLOOMY and BLOOMY-PREDICT studies, 970 cases of \u003cem\u003eS. aureus\u003c/em\u003e BSI were documented, of which 928 (95.7%) were monomicrobial BSI. Only isolates from index blood cultures were sequenced and matched with the clinical data set of their infected host. The final cohort comprised 494 isolates from the BLOOMY study and 149 isolates from the PREDICT study phase (total = 643). The initial and final composition of the cohort is shown in Fig. 1. Study site and phase specific distribution of sequenced isolates are shown in Supplementary Fig. 1.\u003c/p\u003e\n\u003cp\u003ePatient demographics, including age, sex, PBS, CCI, discharge status, appropriateness of therapy, acquisition type of BSI (community-acquired vs. hospital-acquired; defined as \u0026le;72h and \u0026gt;72h after hospital admission), and BMI (body mass index), additionally to \u003cem\u003eS. aureus\u003c/em\u003e specific features, are displayed in Supplementary Table 1. To correct for possible genetic variation during the recruitment span, isolates were randomly assigned to training and validation cohorts (75% and 25% of total). Of note, the validation data was handled as true holdout data set and only used to assess the quality of our model in the final ANOVA (analysis of variance). Comparison of the epidemiological features (Supplementary Table 1) between the two cohorts did not show significant differences. The overall in-hospital mortality rate was 24.7% (95% CI 21.4-28.3%). The Odds ratio of female vs. male in-hospital mortality was 1.68 (95% CI 1.16-2.43).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIdentification of clinical confounders for prediction of in-hospital mortality\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo account for confounding clinical variables on in-hospital mortality in \u003cem\u003eS. aureus\u003c/em\u003e BSI, a multiple logistic regression analysis was performed. The following clinical and epidemiological factors as known confounding predictors of in-hospital mortality were preselected: age, sex, BMI, CCI, PBS, BSI focus at discharge and appropriateness and duration of antibiotic therapy. Among those, only sex, CCI and PBS contributed significantly to the final clinical model. Performance of the clinical multiple logistic regression analysis with these three factors resulted in a AUROC (area under the receiver operating characteristic) of 77% for the training and 82% for the validation cohort (Supplementary Table 2).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGenomic features of \u003cem\u003eS. aureus\u0026nbsp;\u003c/em\u003eisolates\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA phylogenetic tree including all \u003cem\u003eS. aureus\u003c/em\u003e isolates with their local origin, MLST type, the associated survival, study cohort and their assignment to the training vs. validation cohort, is shown in Fig. 2. In total, 61 distinct MLST types (most commonly ST45, ST7 and ST22) and 232 unique \u003cem\u003espa\u003c/em\u003e types (predominantly t091, t084 and t002) were identified (Supplementary Fig. 2). ST22 and t002 are known successfully circulating lineages in Europe [17]. Altogether, we postulate a genetically heterogeneous \u003cem\u003eS. aureus\u003c/em\u003e cohort in this study.\u003c/p\u003e\n\u003cp\u003eGenomic factors were identified by annotating the protein-based UniRef100 and UniRef90 trait assigned by the UniProt database. 62.472 unique proteins (99.8% coding, 0.02% non-coding) were annotated with 51.639 unique UniRef100 and 11.221 unique UniRef90 traits. When counting every single annotated genetic feature in all \u003cem\u003eS. aureus\u003c/em\u003e isolates, a total of 1.736.046 traits (1.624.814 coding (cds), 5.859 short coding (sorf) and 105.373 non-coding genes (\u003cem\u003ee.g.,\u003c/em\u003e tRNA, rRNA, ncRNA) were annotated. Among the proteins, 93.4% (n=1.622.156) had a UniRef90 while 89.3% (n=1.549.756) had a UniRef100 annotation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eUniRef traits associated with \u003cem\u003eS. aureus\u003c/em\u003e in-hospital mortality\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe next incorporated annotated coding and non-coding proteins in two distinct statistical models (Fig. 3). First, UniRef100 traits\u0026nbsp;served as a unique protein-based trait for our rigid model (\u003cu\u003eModel A\u003c/u\u003e). UniRef100 includes identical sequences at the amino acid level, and hereby represents a single protein variant. Second, to obtain prognostic traits on a protein group level we utilized the UniRef90\u0026nbsp;traits\u0026nbsp;(\u003cu\u003eModel B\u003c/u\u003e) comprising protein variants with at least 90% amino acid sequence similarity. Depending on the degree of protein conservation the amount of protein variants summarized in the UniRef90 trait differed from 1 \u0026ndash; 45 variants. The final results of the significant traits in both models after correcting the significant UniRef traits for sex, CCI and PBS\u0026nbsp;are shown in Fig. 4. Sorting of the significant traits and clinical factors was achieved by multiple repetitions of the ANOVA analysis each reduced by the traits/factors contributing least to the previous cycle of ANOVA. This sorting of the traits provided insight into the most impactful\u0026nbsp;features in the predictive model.\u003c/p\u003e\n\u003cp\u003eThe top 10 traits for each model, together with their given protein names, associated in-hospital mortality rate and genetic abundance, are listed in Table 1 (A) and (B). The full list of significant traits (77 traits in Model A; 39 traits in Model B) with their suggested mode of action in \u003cem\u003eS. aureus\u003c/em\u003e pathogenicity, the decisive clinical factors (three each in Model A and B), and the AUROC value of each cumulative feature set from the highest to the lowest rank of the training and the validation cohort as a marker for the quality of the prediction model are displayed in Supplementary Table 3. Eleven traits, all associated with increased in-hospital mortality, were concordant in both models highlighting the overlapping consistency of the two annotation approaches (Table 1 (C)). Those eleven traits showed a fairly conserved amino acid sequence reflected by maximum five different protein variants (UniRef100) within the UniRef90 trait. The commonly identified traits included four functionally unknown proteins, one uncharacterized phage protein, the staphylococcal enterotoxins type J and R, two replication proteins and two transcriptional regulators.\u003c/p\u003e\n\u003cp\u003eTo validate the relevance of the uncovered traits in a second independent genome-wide approach, we fed the same input used for model A and our phylogenetic tree in the Scoary2 pipeline (for results see Supplementary Table 3 and 4). Scoary2 is a tool for genome-wide association studies reflecting also the phylogenetic background of the strain collection but does not incorporate clinical data. Here, nine of the 77 UniRef100 traits were significant, including all five highest ranked traits from our Model A. Those five traits included an uncharacterized protein and protein variants of the ABC transporter substrate-binding protein FhuD1, the proline\u0026mdash;tRNA ligase ProS (all three prognostically unfavorable), the ATPase component of the ATP binding cassette (ABC) superfamily transporter MntA and a lipoprotein with unknown function (both favorable). Additional commonly identified traits in Model A and Scoary2 were variants of a HTH transcriptional regulator, an unknown amino acid proton symporter, the fibronectin-binding protein A and a staphylococcal tandem lipoprotein. In summary, future studies should focus on those nine traits as promising and independently confirmed candidates influencing \u003cem\u003eS. aureus\u003c/em\u003e virulence in human BSI.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCo-dependency of relevant traits\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo evaluate patterns of co-occurrence in single \u003cem\u003eS. aureus\u003c/em\u003e lineages, traits were displayed in an absence/presence matrix (Supplementary Table 5) sorting the isolates according to the phylogenetic tree as a backbone. Most traits are distributed across several MLST types, indicating fairly lineage-independent virulence proteins. Upon inspection, three clusters became apparent in which various traits frequently occurred in close proximity on the contig, suggesting either a functional or local dependency: cluster I., comprising two uncharacterized lipoproteins, cluster II., comprising a phage protein and integrase, and cluster III., comprising staphylococcal enterotoxins R and J, a MarR transcriptional regulator, a site-specific DNA recombinase, a phage protein, three replication-associated proteins and two uncharacterized proteins (see Supplementary Table 3). Interestingly, all proteins found in cluster III. are located on the pIB485-like plasmid. For each cluster, it has yet to be determined whether the corresponding proteins all actively influence in-hospital mortality, which we consider rather unlikely, or whether they are only surrogates for mortality, at least in part. To detect protein dependencies beyond their close proximity in the genome, UniRef100 traits were further investigated by \u003cem\u003ein silico\u003c/em\u003e functional network analysis using the STRING database. A number of networks became apparent linking the traits by associations in curated databases, experimentally determined functional connections, gene neighborhood and co-occurrence across genomes (Supplementary Figure 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIndexing of the most important traits identified in the analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe most important traits with a known mode of action in our analysis were two protein variants involved in micronutrient transport (FhuD1 (UniRef100_Q93PN3) for iron and MntA (UniRef100_Q99VY2) for manganese and zinc transport) and a variant of the proline\u0026mdash;tRNA ligase ProS (UniRef100_A7X1P3), involved in protein translation. Known virulence factors in \u003cem\u003eS. aureus\u003c/em\u003e pathogenicity identified in our study were protein variants of the fibrinogen-binding protein (UniRef100_P68799) and fibronectin-binding protein A (UniRef100_UPI0002CA386D), both crucial for adhesion processes. Moreover, two staphylococcal enterotoxins (Sej (UniRef100_D2J5X0) and Ser (UniRef100_O85217)) were found among the prognostic traits. Contrarily, proteins that have been closely intertwined with \u003cem\u003eS. aureus\u003c/em\u003e pathogenicity like hemolysins, coagulase, PSMs or superantigens like TSST-1 did not appear as significant risk factors in our models. The gene abundances of \u003cem\u003eeta\u003c/em\u003e, \u003cem\u003eetd\u003c/em\u003e and \u003cem\u003elukF/lukS\u0026nbsp;\u003c/em\u003e(coding for PVL) were below 5%, therefore failing the criterion to be incorporated.\u003c/p\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003ePathogenicity of \u003cem\u003eStaphylococcus aureus\u003c/em\u003e has been studied extensively over the last few decades, addressing an important clinical need. We sought to illuminate the impact of bacterial genetic factors and link them to outcome of BSI in hospitalized patients. Selected variables incorporated in the clinical model were age, sex, BMI, CCI, PBS, ascertained focus at discharge, and appropriateness and length of administered therapeutic substances. Our pragmatic definition for appropriate antibiotic therapy reflects antimicrobial recommendations in international guidelines [18]\u0026nbsp;and was applied to account for in-hospital mortality. However, this definition does not claim to fully consider all recommended therapeutic strategies and interdisciplinary approaches, particularly in cases of deep-seated infections, endocarditis or bone and joint infections. PBS, CCI and sex turned out to be the most impactful clinical variables in predicting in-hospital mortality and are commonly known risk factors for lethal outcome [5,19,20]. However, PBS was used here not as an indicator for in-hospital mortality risk as its actual intended use, but to account for the patient\u0026rsquo;s state at the time of blood culture sampling on day 0. Irrespective of the genetic equipment of the diverse \u003cem\u003eS. aureus\u003c/em\u003e isolates and the elapsed time until BSI was diagnosed, PBS served as a benchmark to level the initial clinical status. Thus, additional potential predictors (particularly specific laboratory values) were not included as confounders. Our data support the importance of gathering clinical scores to estimate in-hospital mortality risk in BSI.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe created a rigid statistical Model A in which a vast number of proteins were annotated based on their 100% identity at the amino acid sequence level. In contrast, Model B was fed with proteins sharing 90% similarity at the amino acid sequence, allowing to associate orthologous or paralogous protein groups. Importantly, we correlated entirely novel factors with in-hospital mortality, but also well-described proteins linked to \u003cem\u003eS. aureus\u003c/em\u003e virulence, highlighting the biological plausibility of our method. Of note, our approach did not consider transcriptional regulation of the underlying genes but focused on the influence of absence or presence of proteins and their amino acid variants. There are further explanations why virulence factors might not have been detected in our study: 1. Low-abundancy of the trait (\u003cem\u003ee.g.,\u003c/em\u003e PVL). 2. Redundancy of the trait (\u003cem\u003ee.g.,\u003c/em\u003e PSMs which are conserved proteins of the core genome). 3. High amino acid variability (\u003cem\u003ee.g.,\u003c/em\u003e \u0026alpha;-hemolysin [21]) potentially reducing the statistical power of the trait. Scoary2 was applied as an orthogonal approach to validate our results and to minimize the phylogenetic background as a surrogate for clonal virulence [22]. The sheer number of significantly associated proteins compared to other genome-wide approaches underlines the importance of correcting not only for the phylogenetic background but also incorporating clinical data to untangle BSI mortality markers, the relevance of which could otherwise be under- or overestimated by those confounders. The 116 relevant proteins can be grouped by their mode of action and contribute to \u003cem\u003eS. aureus\u003c/em\u003e pathogenicity and resilience at different levels: immune modulation, metal homeostasis, gene transcription and translation, recombination and genome editing, oxidative stress reduction, adhesion, abscess formation, metabolic functions, heavy metal resistance and environmental sensing. All these factors can influence bacterial survival chances in the harsh, contested bloodstream niche. The function of other proteins remains obscure.\u0026nbsp;To date, it is unclear if they singularly contribute to high assessment scores like the PBS over time of ongoing BSI. Furthermore, since \u003cem\u003eS. aureus\u003c/em\u003e lineages and their virulence factors are distributed in globally inhomogeneous patterns, the abundance and, thus, impact of the identified virulence factors are not \u003cem\u003eper se\u003c/em\u003e generalizable to other patient cohorts\u0026nbsp;[23]. In the following part, we discuss a selection of the identified proteins to contextualize their importance in \u003cem\u003eS. aureus\u003c/em\u003e BSI.\u003c/p\u003e\n\u003cp\u003eUniRef100_Q93PN3, a variant of the ABC transporter substrate-binding protein FhuD1, is an extracellular surface lipoprotein binding ferric hydroxamate and staphyloferrin, which is a siderophore extracting iron from transferrin [10]. Iron restriction has challenging consequences on both bacteria and the host, as shortage for this valuable co-factor causes deterioration of essential enzymatic functions. Since the host restricts access to iron during acute and chronic infection, in a process known as nutritional immunity, bacteria have evolved efficient iron-capturing strategies to outperform the hosts own capacity [24]. FhuD1 is selectively upregulated under iron restriction to enable uptake of iron-bound siderophores, and also xenosiderophores, a process known as iron piracy [25,26].\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003ePatients with iron overload who are administered the chelator desferroxamine B are at risk of \u003cem\u003eS. aureus\u003c/em\u003e infections. Whether this increased risk can be attributed to FhuD1/2 remains controversial\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e[25,27,28].\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eIt is also unclear whether this protein variant of FhuD1\u003cem\u003e\u0026nbsp;\u003c/em\u003econfers advantages in iron scavenging compared to other variants, resulting in better bacterial survival in the bloodstream and thereby increasing human in-hospital mortality.\u003c/p\u003e\n\u003cp\u003eThe presence of UniRef100_Q99VY2, an ATPase variant\u0026nbsp;of the ABC superfamily transporter MntA and component of a Mn\u003csup\u003e2+\u003c/sup\u003e/Zn\u003csup\u003e2+\u003c/sup\u003e transport system, correlated with decreased in-hospital mortality. Zinc is indispensable for enzyme activity, in humans and microorganisms alike, and is thus fiercely competed for in the host [29,30]. Manganese is an essential co-factor for the superoxide dismutase, which eliminates reactive oxygen species and thereby protects the bacterium from oxidative cell stress [31]. The importance of these two metals for bacterial survival is undisputed and the correlation of this protein variant with a beneficial outcome needs further investigation. Differences in the enzymatic activity of this variant\u0026nbsp;might explain the observed association and further studies should focus on comparing the catalytic activity of the various variants.\u003c/p\u003e\n\u003cp\u003eStaphylococcal enterotoxins J (Sej) (UniRef90_Q76LS7) and R (Ser) (UniRef100_O85217/UniRef100_D2J5X0) are known enterotoxins and act as superantigens [7,9]. Both were associated with increased in-hospital mortality and co-detected with eight other traits in our cohort, including a MarR family transcriptional regulator and other uncharacterized protein-encoding genes on the pIB485-like plasmid [32]. The presence of Sed (staphylococcal enterotoxin D), also located on this plasmid, and other enterotoxin-encoding genes was previously associated with the manifestation of endocarditis, although in another cohort, absence of \u003cem\u003esed/j/r\u003c/em\u003e correlated with embolizing endocarditis [33,34]. It is yet unclear whether the identified factors are drivers of in-hospital mortality or simply correlate with other unfavorable factors that are located on the plasmid and were not identified with our approach.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTranscription factors play a crucial role in bacterial adaptation and operate in ubiquitous processes, including biofilm formation, chemotaxis, secretion mechanisms, adhesion and many more. They can either function as inhibitors or activators of transcription and interact with multiple effector proteins (\u003cem\u003ee.g.,\u003c/em\u003e RNA polymerase, DNA, RNA) [35]. We found protein variants of different transcription regulator families often but not exclusively, correlating with elevated death rates. As an example, MarR (Multiple antibiotic resistance regulator), which is represented with two protein variants (UniRef100_D2J5X2/UniRef90_A0A1W5ISY9), comprises a family of transcription factors responding to antibiotics, chemical and oxidative stressors and contributing to virulence in \u003cem\u003eS. aureus\u003c/em\u003e [35-37]. Whether each identified protein variant of those transcription factors has repressive or activating functions and how this is associated with a beneficial or unfavorable BSI outcome remains to be further elucidated.\u003c/p\u003e\n\u003cp\u003eFinally, among the relevant traits were also several phage-originating proteins mostly associated with increased in-hospital mortality. Temperate bacteriophages harbor a substantial amount of virulence determinants and significantly contribute to \u003cem\u003eS. aureus\u003c/em\u003e pathogenicity [38,39]. Moreover, several uncharacterized proteins like domains of unknown function (DUFs) or the highest ranked trait UniRef100_A0A6C0L8I7, also having the highest trait specific in-hospital mortality rate of 60.6%,\u0026nbsp;appeared as relevant factors for outcome prediction. Origin and function of these uncharacterized proteins remain enigmatic requiring further investigation.\u003c/p\u003e\n\u003cp\u003eRecently, a study was published searching for predictive factors discriminating \u003cem\u003eS. aureus\u0026nbsp;\u003c/em\u003estrains based on origin and clinical manifestation (colonization vs. different severity stages of infection). By combining Random Forrest and two genome-wide association methods, annotated genes in coding regions, intergenic regions (IGR) and sRNA from human and animal isolates were associated with the mentioned endpoints. \u003cem\u003emecA\u0026nbsp;\u003c/em\u003eand an adjacent IGR were identified as discriminative factors for (severe) \u003cem\u003eS. aureus\u003c/em\u003e infections versus colonization [40]. We could not confirm this finding, possibly due to different study endpoints and a MRSA rate of 5.1%, which is lower than the 43% in the study of Sassi \u003cem\u003eet al\u003c/em\u003e.. In comparison, our analysis stands out for its large, clearly structured, multi-centric cohort and exclusive focus on \u003cem\u003eS. aureus\u003c/em\u003e blood culture isolates, and for accounting and correcting for a tremendous number of clinical parameters. Nonetheless, a known limitation of our study is the use of short read sequencing, resulting in assembly issues especially affecting genomic low-complexity regions or plasmid assembly. This necessitates confirmation of the traits\u003cem\u003e\u0026nbsp;in vitro\u003c/em\u003e and \u003cem\u003ein vivo\u003c/em\u003e, particularly those that have not yet been characterized.\u003c/p\u003e\n\u003cp\u003eTo the best of our knowledge, this is the first large BSI study on \u003cem\u003eS. aureus\u003c/em\u003e combining a vast, standardized clinical data set with whole genome sequencing analysis to evaluate bacterial traits as predictors of in-hospital mortality in a congruent cohort. These findings provide new evidence that pathogen genomic factors may add clinically relevant prognostic information beyond host characteristics and disease severity, and uncover a set of biologically plausible candidates for translational development. Validation in independent cohorts and functional studies will be important next steps, but the present results offer a valuable framework for improving risk stratification and for guiding the future development of predictive biomarkers and innovative preventive or therapeutic strategies.\u003c/p\u003e"},{"header":"MATERIAL AND METHODS","content":"\u003cp\u003e\u003cstrong\u003eStudy population\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePatient recruitment for the BLOOMY study took place between 2017 and 2018, while recruitment for the BLOOMY-PREDICT cohort was carried out between 2019 and 2020. Participating hospitals consisted of six German university hospitals. Hospitalized patients \u0026ge;18 years of age suffering from bloodstream infections due to ESKAPEE microorganisms (\u003cem\u003eEnterococcus faecium/faecalis\u003c/em\u003e, \u003cem\u003eStaphylococcus aureus\u003c/em\u003e, \u003cem\u003eKlebsiella species\u003c/em\u003e, \u003cem\u003eAcinetobacter baumannii\u003c/em\u003e, \u003cem\u003ePseudomonas aeruginosa\u003c/em\u003e, \u003cem\u003eEnterobacter species\u003c/em\u003e and\u003cem\u003e\u0026nbsp;Escherichia coli\u003c/em\u003e) were eligible for study participation and were asked for consent. Clinical departments with low incidence of bloodstream infections (ophthalmology, psychiatry) were excluded. Clinical record data were reviewed and entered in a database structure (Research Electronic Data Capture, REDCap). A full list of recorded epidemiological and clinical variables, including the PBS and the Charlson comorbidity index (CCI) for each patient, was published previously [16].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo account for antimicrobial therapy as a confounding factor for survival criteria for the appropriateness of therapeutic antimicrobials and duration of therapy were selected in a fairly strict manner reflecting antimicrobial recommendations in international guidelines [18]. In patients with survival of \u0026ge;14 days after BSI onset, appropriate in-hospital therapy was defined as IV administration of one of the following antibiotics for at least 14 days: flucloxacillin, cefazolin, piperacillin/tazobactam (only if not exclusively administered for the duration of treatment due to know elevated mortality risk for piperacillin/tazobactam monotherapy [41]), ampicillin/sulbactam (all four antimicrobials for treatment of MSSA (methicillin-susceptible \u003cem\u003eStaphylococcus aureus)\u0026nbsp;\u003c/em\u003ecases only), vancomycin, daptomycin or linezolid (last three drugs for treatment of both MSSA and MRSA). In patients who died or were transferred before day 14, adequate therapy duration was set to x\u0026minus;1 days, where x is the time from BSI onset to death or transfer. We are aware that this definition does not perfectly reflect the demand for individually tailored therapeutic strategies and interdisciplinary approaches particularly in complicated cases of \u003cem\u003eS. aureus\u003c/em\u003e BSI. It also does not distinguish between empirical or targeted therapy or consider the infection focus or diagnostic procedures like negative follow-up blood cultures. We tried to establish a pragmatic definition to purposefully account for appropriate therapy as a potential confounder in our large cohort and the endpoint \u0026lsquo;in-hospital mortality\u0026rsquo;.\u003c/p\u003e\n\u003cp\u003eFor the current study, isolates and clinical metadata (Supplementary Table 6) from five study sites of the BLOOMY cohort and two study sites of the BLOOMY-PREDICT cohort were available and included. We only included monomicrobial BSI, defined as no isolation of pathogens other than \u003cem\u003eS. aureus\u003c/em\u003e from a blood culture obtained within 24 hours after the first positive blood culture with \u003cem\u003eS. aureus\u003c/em\u003e (day 0). Common skin contaminants (coagulase-negative staphylococci, \u003cem\u003eCorynebacterium spp.\u003c/em\u003e, \u003cem\u003eBacillus spp.\u003c/em\u003e, and \u003cem\u003eCutibacterium spp.\u003c/em\u003e) detected in one of several blood cultures were not considered as polymicrobial infections.\u003cem\u003e\u0026nbsp;\u003c/em\u003eThe study endpoint of interest in this \u003cem\u003eS. aureus\u003c/em\u003e specific sub-cohort was in-hospital mortality due to BSI. The study was conducted under the ethics approval no. 765/2106BO1 and 584/2019BO1, issued by the local ethics committee of the University Hospital T\u0026uuml;bingen.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCultivation and isolation of \u003cem\u003eStaphylococcus aureus\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIsolation of \u003cem\u003eS. aureus\u003c/em\u003e from blood cultures was performed using commercially available automated blood culture incubator systems following standard protocols at the respective study sites. Blood cultures were streaked on agar growth media and plates were incubated at 37 \u0026deg;C in different atmospheric environments following state of the art standard diagnostic protocols. For identification of microorganisms, rapid diagnostic molecular testing as well as MALDI-TOF (Matrix-assisted laser desorption/ionization-Time of flight) mass spectrometry using the Microflex LT instrument (Bruker Daltonics, Bremen, Germany) were used. Antimicrobial susceptibility testing was performed using commercial semi-automated systems (\u003cem\u003ee.g.,\u003c/em\u003e VITEK 2, bioM\u0026eacute;rieux, N\u0026uuml;rtingen, Germany) and interpreted according to the breakpoints published by EUCAST (European Committee of Antimicrobial Susceptibility Testing; https://www.eucast.org/). Suspected methicillin resistance was confirmed by commercially available assays for molecular \u003cem\u003emecA/C\u003c/em\u003e gene detection.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSequencing of isolates\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIsolates of \u003cem\u003eS. aureus\u003c/em\u003e were frozen at -80 \u0026deg;C until further sequencing analysis. Genomic bacterial DNA was extracted from cultures grown on Columbia agar supplemented with 7% sheep blood (Thermo Fisher Scientific, Schwerte, Germany) using the Qiagen DNeasy\u0026reg; UltraClean\u0026reg; Microbial Kit (Qiagen, Hilden, Germany), following the manufacturer\u0026apos;s instructions with minor modifications. DNA concentrations were quantified using the Qubit\u0026trade; dsDNA BR Assay Kit (Thermo Fisher Scientific, Massachusetts, USA). Whole genome sequencing libraries were prepared with the Illumina DNA Prep Kit (Illumina, San Diego, USA) using a standard protocol, and samples were barcoded using Illumina\u0026reg; DNA/RNA UD Indexes. The barcoded libraries were subsequently quantified using the Qubit\u0026trade; dsDNA BR Assay Kit (Thermo Fisher Scientific). Normalized libraries were pooled equimolarly and sequenced on a NextSeq 500 platform (Illumina, San Diego, USA) using a Mid Output Cartridge v2.5 (2x150 bp), aiming at approximately 100x genome coverage. Whole Genome Sequences are accessible under the accession number ERP191105 (BioProject No. PRJEB110420) at the European Nucleotide Archive (ENA).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBioinformatic analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cu\u003eGenome assembly and annotation\u003c/u\u003e\u003c/p\u003e\n\u003cp\u003eThe sequencing files were demultiplexed with our in-house Nextflow pipeline ncct-mibi/nxf-bcl, which uses bcl2fastq (v2.19.0.316) for demultiplexing, fastp (v0.23.4) for quality check and visualized the results with MultiQC (v1.7). The pipeline was executed with Nextflow (v20.10.0) to allow parallel processing of files. The generated data were analyzed with the Nextflow based Bactopia (v3.0.1) pipeline, which allows a standardized processing of the data in a parallel fashion [42]. After a basic quality check with fastp (v0.23.4) and FastQC (v0.12.1), the raw reads were assembled using unicycler (v0.5.0) [43]. Afterwards, annotation was performed using bakta (v1.9.3) with the database available at that time (zenodo.10522951), which was downloaded on 17.04.2024 [44]. The bakta output included annotation of the protein-based UniRef100 and UniRef90 traits assigned by the UniProt database (https://www.uniprot.org/) on each gene predicted within all samples. UniRef100 and UniRef90 traits were further used as input for the statistical analysis. UniRef100 represents a reference cluster of proteins with a 100% identity at the amino acid sequence level. UniRef90 further groups the UniRef100 clusters with an amino acid sequence identity of 90% allowing to analyze orthologous or paralogous protein families. Finally, an additional taxonomic classification was performed with gtdb-tk (v2.4.0, database release 220, downloaded 04.06.2024) to validate the species and obtain average nucleotide identity (ANI) to the closest reference genome.\u003c/p\u003e\n\u003cp\u003e\u003cu\u003eAdditional classification for MLST (Multilocus Sequence Typing), \u003cem\u003espa\u003c/em\u003e and \u003cem\u003eagr\u003c/em\u003e type\u003c/u\u003e\u003c/p\u003e\n\u003cp\u003eThe resulting assembly was used for automated classifications of the MLST, \u003cem\u003espa\u003c/em\u003e and \u003cem\u003eagr\u003c/em\u003e type. All of them are included in the tools collection (Bactopia Tools) of Bactopia (v3.0.1). The MLST type was determined using mlst (v2.23.0) and the mlst-database (v2.23.0-20240325), the \u003cem\u003espa\u003c/em\u003e type using spatyper (v0.2.1) and the \u003cem\u003eagr\u003c/em\u003e type using agrvate (v1.0.2).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cu\u003ePhylogenetic analysis\u003c/u\u003e\u003c/p\u003e\n\u003cp\u003eCore genome analysis was performed using 645 \u003cem\u003eS. aureus\u003c/em\u003e whole genomes (including 642 isolates of the analyzed cohort) in SPINE (version 0.2) [45], setting LCB (locally collinear blocks) to 1000, permissible SNP variation to 90% and number of included isolates to 95% for core genome identification. The core genome consisted of 2.347.879 bases (2.3 Mb). To check for possible phage insertion in the core genome, the computed core genome was uploaded to PHASTEST.ca (Version \u003cstrong\u003e3.0; upload date\u0026nbsp;\u003c/strong\u003e27.09.2024\u003cstrong\u003e)\u003c/strong\u003e [46] where no prophage sequences could be assigned. For SNP calling and further filtering, GATK (The Genome Analysis Toolkit (GATK) v3.2-2), SAMtools v.0.1.19 and VCFtools - v0.1.11 were used (settings for minimal mapping score: 30; minimal read coverage: 10; maximal read coverage: 3; minimal SNP mapping quality: 30; minimal SNP base quality: 30; SNP percentage 0.8; no indels) [47-49]. For final maximum likelihood (ML) tree construction, we used IQ-TREE multicore version 1.6.3. (settings: UFboot and bootstraps 1000) [50]. Visualization of phylogenetic trees was performed in iTOL v6.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTwo goals were to be achieved: First, we aimed to find a predictive model for in-hospital mortality due to\u0026nbsp;\u003cem\u003eS. aureus\u003c/em\u003e BSI. The prediction ought to be based on a suitable selection of clinical data and genomic information about the \u003cem\u003eS. aureus\u003c/em\u003e bloodstream isolates. The second goal was to identify bacterial traits that were potentially responsible for the outcome. The data structure in the BLOOMY and the PREDICT study were different with regard to coding of some parameters and naming of data fields. Therefore, we first harmonized the data fields and incorporated missing information in single cases in a random fashion. This incorporation of missing information was not applied for BMI (body mass index) due to the high number of missing values. Preparation of the genomic information was straight forward given the results from the bioinformatic preprocessing. Second, we split the data into two cohorts (training and validation cohort) by using a random mechanism. Third, we analyzed the distributions of demographic, clinical and isolate specific parameters (age, sex, infection focus, resistance pattern, etc.). We provide a descriptive overview of the cases and evaluated whether training and validation data were comparable (Supplementary Table 1), using Fisher\u0026rsquo;s exact and Chi Square tests with a significance level of p\u0026le;0.05. Lastly, we built a predictive model using ANOVA and logistic regression, described in detail in the following part.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe workflow of the predictive model is displayed in Figure 3. Regarding the clinical data, we applied multiple logistic regression analysis with likelihood tests and three out of eight clinical factors with known influence on mortality in\u0026nbsp;\u003cem\u003eS. aureus\u003c/em\u003e BSI were significant (p\u0026le;0.05) (Step 1): PBS, CCI and sex. For the genomic content of isolates, we used the annotated protein-based UniRef100 traits (resulting in Model A) and UniRef90 traits (resulting in Model B) as input (Step 2 and 3). Coding and non-coding genomic regions were included. We applied univariate logistic regression analysis on protein absence/presence for the outcome survival vs. death during hospitalization to extract p-values (Step 4). Traits were incorporated in the multiple logistic regression analysis when their p-value was less or equal to 5% (Step 5). Of note, these p-values were used as condensed information for sieving purposes and not for hypothesis testing. Next, highly and rarely abundant traits (\u0026lt;5% and \u0026ge;95%) were removed (Step 6). Then, we ran a multiple logistic regression analysis in sets of preselected clinical data (PBS, CCI, sex) and each remaining trait to find out about the surplus of that specific trait to the clinical data (Step 7). This resulted in another list of p-values used in a sieving step (Step 8). Only data from the training cohort were used from Step 1-8. Only statistically significant traits (p\u0026le;0.05) were considered eligible for inclusion in the final analysis of variance (ANOVA). The validation cohort was not used for cross validation purposes but was handled as a true holdout data set and used to assess the quality of our model in the ANOVA analysis. By performing repeated cycles of ANOVA, we reduced the number of UniRef traits and clinical factors (traits and factors further on summarized as \u0026lsquo;features\u0026rsquo;) step by step (Step 9 and 10). At each step, the traits/clinical factor contributing least to the model was removed and were sorted accordingly to reflect their impact in the predictive model. The resulting cascade of ANOVA, including the initial set of features down to the smallest, was each submitted to an AUROC computation (Step 11). We calculated the AUROC value for each feature set from the ANOVA analysis on the training as well as on the validation data. Notably, some UniRef traits co-occurring in the exact same isolates had to be collapsed to one vicarious trait to enable ANOVA analysis leading to a supposedly reduced number of traits in Figure 4 compared to Supplementary Figure 2. Hidden traits were de-collapsed after ANOVA.\u003c/p\u003e\n\u003cp\u003eAll codes used in the data analysis were organized in a sequence of distinct R scripts. Each script processes the data from the preceding step and passes its results to the next. Parameter and data paths were stored in a separate data file so that flexibility and transparency are guaranteed. The code is available on GitHub in repository [http://github.com/UliSchopp/BloomyStaphAu]. We provide detailed description of the code including information about the R packages in the repository.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo compare our method with a second independent approach we used Scoary2\u0026nbsp;(version scoary-2:0.0.15), a software for genome-wide association studies with, however, limitations to incorporate clinical data [22]. The calculations were based on all annotated UniRef100 traits of the training cohort isolates to proceed with the same input. Moreover, the phylogenetic tree prepared in a preceding step was incorporated to account for the phylogenetic background in our cohort. The values shown in output column with title fq*ep calculated and used by the Scoary2 pipeline as criteria for significance (significance level p\u0026le;0.05) were utilized to identify relevant traits.\u003c/p\u003e\n\u003cp\u003eVisualization of figures and analyses were performed in R, GraphPad Prism (10.0.0), GIMP (3.0) and Microsoft Excel (version 2408).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAbsence-presence matrix of significant UniRef100 trait\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo determine the absence and presence of significant UniRef100 traits in all samples a custom python3 script was used. The script was created with the help of ChatGPT and requires the python packages argparse (v1.1), pandas (v2.2.2) and pathlib. The script was validated for functionality by the bioinformatician (JM). The input of the script is a gff3-like file with the annotations of all samples and a list of UniRef100 traits. The script extracts the samples, which contain the UniRef100 trait and creates an absence-presence matrix with UniRef100 traits as columns and samples as rows. For each sample-UniRef100 trait pair the location of the UniRef100 trait on the assembly is tracked as well.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProtein network prediction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo detect protein dependencies beyond their close localization in the genome, traits were further investigated by \u003cem\u003ein silico\u003c/em\u003e functional network analysis using the STRING database (accessed on February, 2026, [51]). Within this database, protein-protein dependencies can be predicted from curated databases, experimentally determined functional connections, gene neighborhood and co-occurrence across genomes. Amino acid sequences of all significant UniRef100 traits were loaded into the database. Sequence identity to other homologues in the database was \u0026gt;90%, and names were assigned based on the UniProt name. Networks with at least three homologues and with minimum required confidence score set to 0.2 (between medium to low confidence) were retrieved and analyzed.\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding:\u0026nbsp;\u003c/strong\u003eThe BLOOMY (TTU 08.810) and BLOOMY-PREDICT study (TTU 08.821) were funded by the German Center for Infection Research (DZIF) and supported by the DZIF Clinical Research Unit (TTU 08.701), TK is funded by the German Center for Infection Research (DZIF, TTU06.716 and TTU08.716).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors contributions:\u003c/strong\u003e Conceptualization of study (SP, MW, and KS), development of study design (SP, KS, US, and JM), sequencing of study isolates (BB and SP), data verification, bioinformatic and statistical analysis (JM, US, TK, and KS), review of study progress (KS and SP), development of the study concept, design and protocol of the BLOOMY/BLOOMY-PREDICT study (ET, WVK, SE, SP, HS, MJGTV, JR, TC, and CG), data collection, coordination and interpretation of the BLOOMY/BLOOMY-PREDICT study (ET, WVK, SE, BPG, SG, SR, MJGTV, HS, JR, TC, CG, LMB, EK, CI, NK, KS, KX and PGH), administration and validation of the clinical data base (BPG), writing of the first draft (KS, SP, LLP, and AP), contribution to the manuscript and approval of the final version of the article (all authors), submission of the manuscript (KS and SP).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCONFLICTS OF INTEREST\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cu\u003eMaria JGT Vehreschild\u0026nbsp;\u003c/u\u003e\u003c/p\u003e\n\u003cp\u003eGrants or contracts from any entity:\u003c/p\u003e\n\u003cp\u003eMSD, Heel, Roche, Tillotts, Pfizer\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePayment made to institution\u003c/p\u003e\n\u003cp\u003eConsulting fees\u003c/p\u003e\n\u003cp\u003eGILEAD, Tillotts, Pfizer, Bioaster, GSK, Ecraid, EUMEDICA, Bactolife, PAION\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePayment made to myself\u003c/p\u003e\n\u003cp\u003ePayment or honoraria for lectures, presentations, speakers bureaus, manuscript writing or educational events\u003c/p\u003e\n\u003cp\u003eAkademie f\u0026uuml;r Infektionsmedizin, Astra Zeneca, bioMerieux, Biotest, DGI, European Society of Neurogastroenterology, Falk Foundation, FomF GmbH, GFO Kliniken Bonn, GILEAD, GSK, Helios Kliniken, Hessisches Landessozialgericht, Infektio Forum, Janssen Cilag GmbH, Klinikum Kassel, Klinikverbund St. Antonius \u0026amp; St. Josef GmbH, Landes\u0026auml;rztekammer Hessen, LMU Kliniken, MSD, Pfizer, Streamed up, St. Vincent Hospital, Tillotts, Vivantes\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePayment made to myself\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eGBD 2019 Antimicrobial Resistance Collaborators. Global mortality associated with 33 bacterial pathogens in 2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet. 2022 Dec 17;400(10369):2221-2248. doi: 10.1016/S0140-6736(22)02185-7. Epub 2022 Nov 21. PMID: 36423648; PMCID: PMC9763654.\u003c/li\u003e\n\u003cli\u003evan Hal SJ, Jensen SO, Vaska VL, Espedido BA, Paterson DL, Gosbell IB. Predictors of mortality in Staphylococcus aureus Bacteremia. Clin Microbiol Rev. 2012 Apr;25(2):362-86. doi: 10.1128/CMR.05022-11. PMID: 22491776; PMCID: PMC3346297.\u003c/li\u003e\n\u003cli\u003eWestgeest AC, Lambregts MMC, Ruffin F, Korn RE, Webster ME, Kair JL, Parsons JB, Maskarinec SA, Kaplan S, Dekkers OM, de Boer MGJ, Fowler VG Jr, Thaden JT. Female Sex and Mortality in Patients with Staphylococcus aureus Bacteremia: A Systematic Review and Meta-analysis. JAMA Netw Open. 2024 Feb 5;7(2):e240473. doi: 10.1001/jamanetworkopen.2024.0473. PMID: 38411961; PMCID: PMC10900971.\u003c/li\u003e\n\u003cli\u003eVaron B, Palacios-Baena ZR, de Kraker MEA, Rodr\u0026iacute;guez-Ba\u0026ntilde;o J, Leibovici L, Paul M; UNIFORM research group. Universal Risk Factors for Mortality in Bloodstream Infections (UNIFORM): a systematic review and Delphi survey. Clin Microbiol Infect. 2024 Apr;30(4):453-461. doi: 10.1016/j.cmi.2023.12.030. Epub 2024 Jan 3. PMID: 38182050.\u003c/li\u003e\n\u003cli\u003eRoth JA, Tschudin-Sutter S, Dangel M, Frei R, Battegay M, Widmer AF. Value of the Pitt Bacteraemia Score to predict short-term mortality in Staphylococcus aureus bloodstream infection: a validation study. Swiss Med Wkly. 2017 Aug 10;147:w14482. doi: 10.4414/smw.2017.14482. PMID: 28804865.\u003c/li\u003e\n\u003cli\u003eL\u0026ecirc;-Bury P, Echenique-Rivera H, Pizarro-Cerd\u0026aacute; J, Dussurget O. Determinants of bacterial survival and proliferation in blood. FEMS Microbiol Rev. 2024 May 8;48(3):fuae013. doi: 10.1093/femsre/fuae013. PMID: 38734892; PMCID: PMC11163986.\u003c/li\u003e\n\u003cli\u003eCheung GYC, Bae JS, Otto M. Pathogenicity and virulence of \u003cem\u003eStaphylococcus aureus\u003c/em\u003e. Virulence. 2021 Dec;12(1):547-569. doi: 10.1080/21505594.2021.1878688. PMID: 33522395; PMCID: PMC7872022.\u003c/li\u003e\n\u003cli\u003eJenul C, Horswill AR. Regulation of \u003cem\u003eStaphylococcus aureus\u003c/em\u003e Virulence. Microbiol Spectr. 2019 Apr 5;7(2):10.1128/microbiolspec.gpp3-0031-2018. doi: 10.1128/microbiolspec.GPP3-0031-2018. PMID: 30953424; PMCID: PMC6452892.\u003c/li\u003e\n\u003cli\u003eAhmad-Mansour N, Loubet P, Pouget C, Dunyach-Remy C, Sotto A, Lavigne JP, Molle V. \u003cem\u003eStaphylococcus aureus\u003c/em\u003e Toxins: An Update on Their Pathogenic Properties and Potential Treatments. Toxins (Basel). 2021 Sep 23;13(10):677. doi: 10.3390/toxins13100677. PMID: 34678970; PMCID: PMC8540901.\u003c/li\u003e\n\u003cli\u003eMlynarczyk-Bonikowska B, Rudnicka L. The Pathogenicity Mechanisms of \u003cem\u003eStaphylococcus aureus\u003c/em\u003e. Int J Mol Sci. 2025 Dec 6;26(24):11803. doi: 10.3390/ijms262411803. PMID: 41465232; PMCID: PMC12733317.\u003c/li\u003e\n\u003cli\u003eAravind L, Anantharaman V, Balaji S, Babu MM, Iyer LM. The many faces of the helix-turn-helix domain: transcription regulation and beyond. FEMS Microbiol Rev. 2005 Apr;29(2):231-62. doi: 10.1016/j.femsre.2004.12.008. PMID: 15808743.\u003c/li\u003e\n\u003cli\u003eSpaan AN, Surewaard BG, Nijland R, van Strijp JA. Neutrophils versus Staphylococcus aureus: a biological tug of war. Annu Rev Microbiol. 2013;67:629-50. doi: 10.1146/annurev-micro-092412-155746. Epub 2013 Jul 3. PMID: 23834243.\u003c/li\u003e\n\u003cli\u003eLee SO, Lee S, Lee JE, Song KH, Kang CK, Wi YM, San-Juan R, L\u0026oacute;pez-Cort\u0026eacute;s LE, Lacoma A, Prat C, Jang HC, Kim ES, Kim HB, Lee SH. Dysfunctional accessory gene regulator (agr) as a prognostic factor in invasive Staphylococcus aureus infection: a systematic review and meta-analysis. Sci Rep. 2020 Nov 26;10(1):20697. doi: 10.1038/s41598-020-77729-0. PMID: 33244173; PMCID: PMC7691521.\u003c/li\u003e\n\u003cli\u003eRecker M, Laabei M, Toleman MS, Reuter S, Saunderson RB, Blane B, T\u0026ouml;r\u0026ouml;k ME, Ouadi K, Stevens E, Yokoyama M, Steventon J, Thompson L, Milne G, Bayliss S, Bacon L, Peacock SJ, Massey RC. Clonal differences in Staphylococcus aureus bacteraemia-associated mortality. Nat Microbiol. 2017 Oct;2(10):1381-1388. doi: 10.1038/s41564-017-0001-x. Epub 2017 Aug 7. PMID: 28785103.\u003c/li\u003e\n\u003cli\u003eCampbell AJ, Mowlaboccus S, Coombs GW, Daley DA, Al Yazidi LS, Phuong LK, Leung C, Best EJ, Webb RH, Voss L, Athan E, Britton PN, Bryant PA, Butters CT, Carapetis JR, Ching NS, Francis J, Hung TY, Nourse C, Ojaimi S, Tai A, Vasilunas N, McMullan B, Bowen AC, Blyth CC; Australian and New Zealand Pediatric Infectious Diseases (ANZPID) clinical research network (CRN) of the Australasian Society of Infectious Diseases (ASID) and the Australian Group on Antimicrobial Resistance (AGAR). Whole genome sequencing and molecular epidemiology of paediatric Staphylococcus aureus bacteraemia. J Glob Antimicrob Resist. 2022 Jun;29:197-206. doi: 10.1016/j.jgar.2022.03.012. Epub 2022 Mar 25. PMID: 35342022.\u003c/li\u003e\n\u003cli\u003eTacconelli E, G\u0026ouml;pel S, Gladstone BP, Eisenbeis S, H\u0026ouml;lzl F, Buhl M, G\u0026oacute;rska A, Cattaneo C, Mischnik A, Rieg S, Rohde AM, Kohlmorgen B, Falgenhauer J, Trauth J, K\u0026auml;ding N, Kramme E, Biehl LM, Walker SV, Peter S, Gastmeier P, Chakraborty T, Vehreschild MJ, Seifert H, Rupp J, Kern WV; DZIF BLOOMY study group. Development and validation of BLOOMY prediction scores for 14-day and 6-month mortality in hospitalised adults with bloodstream infections: a multicentre, prospective, cohort study. Lancet Infect Dis. 2022 May;22(5):731-741. doi: 10.1016/S1473-3099(21)00587-9. Epub 2022 Jan 19. Erratum in: Lancet Infect Dis. 2022 Apr;22(4):e101. doi: 10.1016/S1473-3099(22)00160-8. PMID: 35065060.\u003c/li\u003e\n\u003cli\u003eAsadollahi P, Farahani NN, Mirzaii M, Khoramrooz SS, van Belkum A, Asadollahi K, Dadashi M, Darban-Sarokhalil D. Distribution of the Most Prevalent \u003cem\u003eSpa\u003c/em\u003e Types among Clinical Isolates of Methicillin-Resistant and -Susceptible \u003cem\u003eStaphylococcus aureus\u003c/em\u003e around the World: A Review. Front Microbiol. 2018 Feb 12;9:163. doi: 10.3389/fmicb.2018.00163. PMID: 29487578; PMCID: PMC5816571.\u003c/li\u003e\n\u003cli\u003eDelgado V, Ajmone Marsan N, de Waha S, Bonaros N, Brida M, Burri H, Caselli S, Doenst T, Ederhy S, Erba PA, Foldager D, Fosb\u0026oslash;l EL, Kovac J, Mestres CA, Miller OI, Miro JM, Pazdernik M, Pizzi MN, Quintana E, Rasmussen TB, Ristić AD, Rod\u0026eacute;s-Cabau J, Sionis A, Z\u0026uuml;hlke LJ, Borger MA; ESC Scientific Document Group. 2023 ESC Guidelines for the management of endocarditis. Eur Heart J. 2023 Oct 14;44(39):3948-4042. doi: 10.1093/eurheartj/ehad193. Erratum in: Eur Heart J. 2023 Dec 1;44(45):4780. doi: 10.1093/eurheartj/ehad625. Erratum in: Eur Heart J. 2024 Jan 1;45(1):56. doi: 10.1093/eurheartj/ehad776. Erratum in: Eur Heart J. 2025 Mar 13;46(11):1082. doi: 10.1093/eurheartj/ehae877. PMID: 37622656.\u003c/li\u003e\n\u003cli\u003eGBD 2013 Mortality and Causes of Death Collaborators. Global, regional, and national age-sex specific all-cause and cause-specific mortality for 240 causes of death, 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet. 2015 Jan 10;385(9963):117-71. doi: 10.1016/S0140-6736(14)61682-2. Epub 2014 Dec 18. PMID: 25530442; PMCID: PMC4340604.\u003c/li\u003e\n\u003cli\u003eD\u0026apos;Hoore W, Sicotte C, Tilquin C. Risk adjustment in outcome assessment: the Charlson comorbidity index. Methods Inf Med. 1993 Nov;32(5):382-7. PMID: 8295545.\u003c/li\u003e\n\u003cli\u003eXiao M, Zhao R, Zhang Q, Fan X, O\u0026apos;Sullivan MV, Li DF, Wang XY, Wu HL, Kong F, Xu YC. Genotypic Diversity of Staphylococcus aureus \u0026alpha;-Hemolysin Gene (hla) and Its Association with Clonal Background: Implications for Vaccine Development. PLoS One. 2016 Feb 11;11(2):e0149112. doi: 10.1371/journal.pone.0149112. PMID: 26866483; PMCID: PMC4750931.\u003c/li\u003e\n\u003cli\u003eRoder T, Pimentel G, Fuchsmann P, Stern MT, von Ah U, Verg\u0026egrave;res G, Peischl S, Brynildsrud O, Bruggmann R, B\u0026auml;r C. Scoary2: rapid association of phenotypic multi-omics data with microbial pan-genomes. Genome Biol. 2024 Apr 11;25(1):93. doi: 10.1186/s13059-024-03233-7. PMID: 38605417; PMCID: PMC11007987.\u003c/li\u003e\n\u003cli\u003eJolley KA, Bray JE, Maiden MCJ. Open-access bacterial population genomics: BIGSdb software, the PubMLST.org website and their applications. Wellcome Open Res. 2018 Sep 24;3:124. doi: 10.12688/wellcomeopenres.14826.1. PMID: 30345391; PMCID: PMC6192448.\u003c/li\u003e\n\u003cli\u003eSebulsky MT, Speziali CD, Shilton BH, Edgell DR, Heinrichs DE. FhuD1, a ferric hydroxamate-binding lipoprotein in Staphylococcus aureus: a case of gene duplication and lateral transfer. J Biol Chem. 2004 Dec 17;279(51):53152-9. doi: 10.1074/jbc.M409793200. Epub 2004 Oct 8. PMID: 15475351.\u003c/li\u003e\n\u003cli\u003eEndicott NP, Lee E, Wencewicz TA. Structural Basis for Xenosiderophore Utilization by the Human Pathogen Staphylococcus aureus. ACS Infect Dis. 2017 Jul 14;3(7):542-553. doi: 10.1021/acsinfecdis.7b00036. Epub 2017 May 30. PMID: 28505405.\u003c/li\u003e\n\u003cli\u003eSebulsky MT, Heinrichs DE. Identification and characterization of fhuD1 and fhuD2, two genes involved in iron-hydroxamate uptake in Staphylococcus aureus. J Bacteriol. 2001 Sep;183(17):4994-5000. doi: 10.1128/JB.183.17.4994-5000.2001. PMID: 11489851; PMCID: PMC95374.\u003c/li\u003e\n\u003cli\u003eMishra RP, Mariotti P, Fiaschi L, Nosari S, Maccari S, Liberatori S, Fontana MR, Pezzicoli A, De Falco MG, Falugi F, Altindis E, Serruto D, Grandi G, Bagnoli F. Staphylococcus aureus FhuD2 is involved in the early phase of staphylococcal dissemination and generates protective immunity in mice. J Infect Dis. 2012 Oct 1;206(7):1041-9. doi: 10.1093/infdis/jis463. Epub 2012 Jul 24. PMID: 22829645.\u003c/li\u003e\n\u003cli\u003eArifin AJ, Hannauer M, Welch I, Heinrichs DE. Deferoxamine mesylate enhances virulence of community-associated methicillin resistant Staphylococcus aureus. Microbes Infect. 2014 Nov;16(11):967-72. doi: 10.1016/j.micinf.2014.09.003. Epub 2014 Sep 22. PMID: 25251026.\u003c/li\u003e\n\u003cli\u003eXia P, Lian S, Wu Y, Yan L, Quan G, Zhu G. Zinc is an important inter-kingdom signal between the host and microbe. Vet Res. 2021 Mar 4;52(1):39. doi: 10.1186/s13567-021-00913-1. PMID: 33663613; PMCID: PMC7931793.\u003c/li\u003e\n\u003cli\u003eGrim KP, San Francisco B, Radin JN, Brazel EB, Kelliher JL, P\u0026aacute;rraga Sol\u0026oacute;rzano PK, Kim PC, McDevitt CA, Kehl-Fie TE. The Metallophore Staphylopine Enables \u003cem\u003eStaphylococcus aureus\u003c/em\u003e To Compete with the Host for Zinc and Overcome Nutritional Immunity. mBio. 2017 Oct 31;8(5):e01281-17. doi: 10.1128/mBio.01281-17. PMID: 29089427; PMCID: PMC5666155.\u003c/li\u003e\n\u003cli\u003eKehl-Fie TE, Zhang Y, Moore JL, Farrand AJ, Hood MI, Rathi S, Chazin WJ, Caprioli RM, Skaar EP. MntABC and MntH contribute to systemic Staphylococcus aureus infection by competing with calprotectin for nutrient manganese. Infect Immun. 2013 Sep;81(9):3395-405. doi: 10.1128/IAI.00420-13. Epub 2013 Jul 1. PMID: 23817615; PMCID: PMC3754211.\u003c/li\u003e\n\u003cli\u003eShearer JE, Wireman J, Hostetler J, Forberger H, Borman J, Gill J, Sanchez S, Mankin A, Lamarre J, Lindsay JA, Bayles K, Nicholson A, O\u0026apos;Brien F, Jensen SO, Firth N, Skurray RA, Summers AO. Major families of multiresistant plasmids from geographically and epidemiologically diverse staphylococci. G3 (Bethesda). 2011 Dec;1(7):581-91. doi: 10.1534/g3.111.000760. Epub 2011 Dec 1. PMID: 22384369; PMCID: PMC3276174.\u003c/li\u003e\n\u003cli\u003eRasigade JP, Lecl\u0026egrave;re A, Alla F, Tessier A, Bes M, Lechiche C, Vernet-Garnier V, Laou\u0026eacute;nan C, Vandenesch F, Leport C; AEPEI Study Group. \u003cem\u003eStaphylococcus aureus\u003c/em\u003e CC30 Lineage and Absence of \u003cem\u003esed\u003c/em\u003e,\u003cem\u003ej\u003c/em\u003e,\u003cem\u003er\u003c/em\u003e-Harboring Plasmid Predict Embolism in Infective Endocarditis. Front Cell Infect Microbiol. 2018 Jun 8;8:187. doi: 10.3389/fcimb.2018.00187. PMID: 29938201; PMCID: PMC6003251.\u003c/li\u003e\n\u003cli\u003eNienaber JJ, Sharma Kuinkel BK, Clarke-Pearson M, Lamlertthon S, Park L, Rude TH, Barriere S, Woods CW, Chu VH, Mar\u0026iacute;n M, Bukovski S, Garcia P, Corey GR, Korman T, Doco-Lecompte T, Murdoch DR, Reller LB, Fowler VG Jr; International Collaboration on Endocarditis-Microbiology Investigators. Methicillin-susceptible Staphylococcus aureus endocarditis isolates are associated with clonal complex 30 genotype and a distinct repertoire of enterotoxins and adhesins. J Infect Dis. 2011 Sep 1;204(5):704-13. doi: 10.1093/infdis/jir389. Erratum in: J Infect Dis. 2013 Feb 1;207(3):546. PMID: 21844296; PMCID: PMC3156104.\u003c/li\u003e\n\u003cli\u003eAl-Tohamy A, Grove A. Targeting bacterial transcription factors for infection control: opportunities and challenges. Transcription. 2025 Feb;16(1):141-168. doi: 10.1080/21541264.2023.2293523. Epub 2023 Dec 21. PMID: 38126125; PMCID: PMC11970743.\u003c/li\u003e\n\u003cli\u003eChen PR, Nishida S, Poor CB, Cheng A, Bae T, Kuechenmeister L, Dunman PM, Missiakas D, He C. A new oxidative sensing and regulation pathway mediated by the MgrA homologue SarZ in Staphylococcus aureus. Mol Microbiol. 2009 Jan;71(1):198-211. doi: 10.1111/j.1365-2958.2008.06518.x. Epub 2008 Nov 5. PMID: 19007410; PMCID: PMC2698432.\u003c/li\u003e\n\u003cli\u003eDeochand DK, Grove A. MarR family transcription factors: dynamic variations on a common scaffold. Crit Rev Biochem Mol Biol. 2017 Dec;52(6):595-613. doi: 10.1080/10409238.2017.1344612. Epub 2017 Jul 3. PMID: 28670937.\u003c/li\u003e\n\u003cli\u003eNovick RP, Ram G. Staphylococcal pathogenicity islands-movers and shakers in the genomic firmament. Curr Opin Microbiol. 2017 Aug;38:197-204. doi: 10.1016/j.mib.2017.08.001. Epub 2017 Nov 1. PMID: 29100762; PMCID: PMC5884141.\u003c/li\u003e\n\u003cli\u003eIngmer H, Gerlach D, Wolz C. Temperate Phages of \u003cem\u003eStaphylococcus aureus\u003c/em\u003e. Microbiol Spectr. 2019 Sep;7(5):10.1128/microbiolspec.gpp3-0058-2018. doi: 10.1128/microbiolspec.GPP3-0058-2018. PMID: 31562736; PMCID: PMC10921950.\u003c/li\u003e\n\u003cli\u003eSassi M, Bronsard J, Pascreau G, Emily M, Donnio PY, Revest M, Felden B, Wirth T, Augagneur Y. Forecasting Staphylococcus aureus Infections Using Genome-Wide Association Studies, Machine Learning, and Transcriptomic Approaches. mSystems. 2022 Aug 30;7(4):e0037822. doi: 10.1128/msystems.00378-22. Epub 2022 Jul 5. PMID: 35862809; PMCID: PMC9426533.\u003c/li\u003e\n\u003cli\u003eBeganovic M, Cusumano JA, Lopes V, LaPlante KL, Caffrey AR. Comparative Effectiveness of Exclusive Exposure to Nafcillin or Oxacillin, Cefazolin, Piperacillin/Tazobactam, and Fluoroquinolones Among a National Cohort of Veterans With Methicillin-Susceptible \u003cem\u003eStaphylococcus aureus\u003c/em\u003e Bloodstream Infection. Open Forum Infect Dis. 2019 Jun 6;6(7):ofz270. doi: 10.1093/ofid/ofz270. PMID: 31281864; PMCID: PMC6602898.\u003c/li\u003e\n\u003cli\u003ePetit RA 3rd, Read TD. Bactopia: a Flexible Pipeline for Complete Analysis of Bacterial Genomes. mSystems. 2020 Aug 4;5(4):e00190-20. doi: 10.1128/mSystems.00190-20. PMID: 32753501; PMCID: PMC7406220.\u003c/li\u003e\n\u003cli\u003eWick RR, Judd LM, Gorrie CL, Holt KE. Unicycler: Resolving bacterial genome assemblies from short and long sequencing reads. PLoS Comput Biol. 2017 Jun 8;13(6):e1005595. doi: 10.1371/journal.pcbi.1005595. PMID: 28594827; PMCID: PMC5481147.\u003c/li\u003e\n\u003cli\u003eSchwengers O, Jelonek L, Dieckmann MA, Beyvers S, Blom J, Goesmann A. Bakta: rapid and standardized annotation of bacterial genomes via alignment-free sequence identification. Microb Genom. 2021 Nov;7(11):000685. doi: 10.1099/mgen.0.000685. PMID: 34739369; PMCID: PMC8743544.\u003c/li\u003e\n\u003cli\u003eOzer EA, Allen JP, Hauser AR. Characterization of the core and accessory genomes of Pseudomonas aeruginosa using bioinformatic tools Spine and AGEnt. BMC Genomics. 2014 Aug 29;15(1):737. doi: 10.1186/1471-2164-15-737. PMID: 25168460; PMCID: PMC4155085.\u003c/li\u003e\n\u003cli\u003eWishart DS, Han S, Saha S, Oler E, Peters H, Grant JR, Stothard P, Gautam V. PHASTEST: faster than PHASTER, better than PHAST. Nucleic Acids Res. 2023 Jul 5;51(W1):W443-W450. doi: 10.1093/nar/gkad382. PMID: 37194694; PMCID: PMC10320120.\u003c/li\u003e\n\u003cli\u003eDanecek P, Auton A, Abecasis G, Albers CA, Banks E, DePristo MA, Handsaker RE, Lunter G, Marth GT, Sherry ST, McVean G, Durbin R; 1000 Genomes Project Analysis Group. The variant call format and VCFtools. Bioinformatics. 2011 Aug 1;27(15):2156-8. doi: 10.1093/bioinformatics/btr330. Epub 2011 Jun 7. PMID: 21653522; PMCID: PMC3137218.\u003c/li\u003e\n\u003cli\u003eLi H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G, Durbin R; 1000 Genome Project Data Processing Subgroup. The Sequence Alignment/Map format and SAMtools. Bioinformatics. 2009 Aug 15;25(16):2078-9. doi: 10.1093/bioinformatics/btp352. Epub 2009 Jun 8. PMID: 19505943; PMCID: PMC2723002.\u003c/li\u003e\n\u003cli\u003eMcKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, Garimella K, Altshuler D, Gabriel S, Daly M, DePristo MA. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 2010 Sep;20(9):1297-303. doi: 10.1101/gr.107524.110. Epub 2010 Jul 19. PMID: 20644199; PMCID: PMC2928508.\u003c/li\u003e\n\u003cli\u003eNguyen LT, Schmidt HA, von Haeseler A, Minh BQ. IQ-TREE: a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol Biol Evol. 2015 Jan;32(1):268-74. doi: 10.1093/molbev/msu300. Epub 2014 Nov 3. PMID: 25371430; PMCID: PMC4271533.\u003c/li\u003e\n\u003cli\u003eSzklarczyk D, Kirsch R, Koutrouli M, Nastou K, Mehryary F, Hachilif R, Gable AL, Fang T, Doncheva NT, Pyysalo S, Bork P, Jensen LJ, von Mering C. The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res. 2023 Jan 6;51(D1):D638-D646. doi: 10.1093/nar/gkac1000. PMID: 36370105; PMCID: PMC9825434.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003eTable 1 is available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Staphylococcus aureus, bloodstream infection, virulence, prognostic markers, risk factors, in-hospital mortality","lastPublishedDoi":"10.21203/rs.3.rs-9449451/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9449451/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Mortality in Staphylococcus aureus bloodstream infections (BSI) is high. While clinical scores and host risk factors have been evaluated in large clinical cohorts, the relevance of the plethora of virulence factors produced by S. aureus for BSI mortality has remained elusive. By combining a comprehensive, standardized clinical data set with whole genome sequencing analysis of 643 S. aureus isolates from a multicentric BSI study and applying logistic regression analyses, we identified proteins associated with both increased and decreased in-hospital mortality. Adjustment with two clinical severity scores and sex revealed a diverse set of 116 staphylococcal proteins involved in immune modulation, metal homeostasis, adhesion, transcription, and translation (among other functions) as prognostic markers for in-hospital mortality. But also several uncharacterized proteins were associated with in-hospital mortality. Nine predictive proteins were confirmed with an orthogonal statistical approach. Most of the identified proteins have not previously been linked to infection outcomes and represent promising candidates for predictive biomarkers.","manuscriptTitle":"Multicentric bloodstream infection cohort study reveals new potential Staphylococcus aureus virulence factors influencing in-hospital mortality","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-04 10:10:56","doi":"10.21203/rs.3.rs-9449451/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"12be22af-0b39-4a72-ab16-58d4ce668448","owner":[],"postedDate":"May 4th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":67317450,"name":"Health sciences/Diseases/Infectious diseases/Bacterial infection"},{"id":67317451,"name":"Health sciences/Pathogenesis/Infection"},{"id":67317452,"name":"Health sciences/Risk factors"},{"id":67317453,"name":"Health sciences/Medical research/Clinical trial design/Clinical trials"}],"tags":[],"updatedAt":"2026-05-04T10:10:57+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-04 10:10:56","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9449451","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9449451","identity":"rs-9449451","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}