Staphylococcus aureus bacteremia (SAB) represents a common, life-threatening and emerging bloodstream infection (1, 2) accounting for up to 15% of hospital-acquired infections in the United States (3, 4) and greater frequency worldwide. Among these cases, antibiotic-persistent S. aureus bacteremia is of urgent and growing concern. This condition occurs when the infecting clinical isolate of S. aureus is not cleared from the bloodstream, despite appropriate dosing and pharmacology of anti-staphylococcal therapy to which the isolate is susceptible in vitro per CLSI breakpoints. Such persistent bacteremia cases are prevalent and potentially lethal (5–7), particularly when involving methicillin-resistant S. aureus (MRSA). The etiologies of SAB are diverse, including invasive skin/soft tissue infections, catheter-associated infections, prosthetic joint infections, and endocarditis among others (7–12). In patients with MRSA bacteremia, outcomes vary greatly depending on the source and hematogenous dissemination of infection (13). Antibiotic-persistent MRSA bacteremia (PB) occurs when the infecting isolate is not cleared from the bloodstream despite appropriate treatment with an antibiotic to which it exhibits susceptibility in vitro. This paradox suggests that differential host responses in context of anti-infective therapy contribute to clinical outcomes in the face of SAB due to a particular MRSA isolate (14).

Specific host-pathogen interactions influencing outcomes of MRSA infection have been a topic of increasing research (15, 16). Yet, the intersection of host and S. aureus mechanisms underlying antibiotic persistent vs. resolution of MRSA bacteremia (RB) remain incompletely understood. In recent years, development of high-throughput genomic, transcriptomic and proteomic platforms has enabled identification of disease-associated immune phenotypes (17, 18). In this study, we tested the hypothesis that peripheral blood transcriptional profiling integrated with proteomic and genotypic correlates would offer new insights into immune mechanisms impacting PB vs. RB outcomes in MRSA bacteremia.

There were four explicit goals of the current investigation: 1) identify transcriptional signatures that differentiate PB vs. RB outcomes in the setting of gold-standard vancomycin therapy in the absence of other omics data; 2) discern relationships linking transcriptomic, IL-10 cytokine levels and genotypic signatures to further enhance differential host response signatures in such PB vs. RB outcomes; 3) disclose putative molecular and cellular mechanisms that may impact these differential outcomes; and 4) evaluate the predictive accuracy of signatures identified using a separate cohort of patients with PB or RB outcomes. Patients having PB vs. RB outcomes exhibited significantly differential patterns of gene co-expression. Specifically, up-regulation of T and B cell signaling genes were hallmarks of RB outcomes, particularly when integrated with low IL-10 levels and host DNMT3A A/C genotype associated with resolution. These findings suggest host transcriptional responses in context of genotypic regulation may shape cellular and proteomic host responses necessary for clearing of MRSA infection in the setting of vancomycin treatment. Such findings further substantiate the potential for systems immunology applications to enhance predictive, diagnostic, or prognostic assessment that could guide medical intervention for improved clinical outcomes.

In the present study, we examined the transcriptome, DNMT3A genotype and IL-10 proteome of whole peripheral blood to differentiate immune response pathways associated with RB versus PB outcomes in clinical MRSA bacteremia. Our central findings show upregulation of T and B cell immune response genes early during MRSA infection is associated with decreased incidence of PB in the setting of appropriate vancomycin therapy. Furthermore, transcriptional profiles correlated with lower IL-10 cytokine level and heterozygous DNMT3A A/C genotype, both of which have been associated with reduced risk of PB outcome in human MRSA bacteremia (14, 38).

Significant differences in transcriptional profiles were identified in PB vs. RB patients. Analysis of GE modules revealed two particularly interesting relationships: 1) T cell networks associated with PB vs. RB outcome; and 2) B cell networks associated with DNMT3A genotype and IL-10 cytokine level. Overall, GE modules showed significantly higher co-expression signatures of these T cell and B cell gene networks in RB as compared to PB patients. Specific hub genes within these networks included T cell activation and differentiation ( Figure 1 ), and B cell receptor signaling, activation and proliferation ( Figure 2 ). For example, upregulated T cell genes CD3E, CD6, CD5, and CD27, members of the T cell signaling protein family, were upregulated in RB compared to PB ( Figures 3A , 4 ). The surface proteins CD5 and CD6 modulate T cell activation in response to pathogen associated molecular patterns (PAMPs), including those found in MRSA (39, 40). Interestingly, CD27 differentiates naïve from memory T cell subsets, with greater expression on naïve T cells (41). The finding that CD27 expression is upregulated in RB patients is consistent with the notion that activation of naïve T cells is as important as that of memory T cells in controlling MRSA bacteremia. Thus, T cell receptor-mediated responses appear to play a critical role in the modulation of T cell activation, expansion, and maintenance of long-term memory important in protecting against persistent MRSA bacteremia.

Consistent with this premise, the T cell chemoattractant CXCR3 was significantly upregulated in RB as compared to PB patients. CXCR3 is important for trafficking and recruitment of Th1 and Th17 polarized CD4⁺ T cells in response to infection. This finding is substantiated by the fact that a mixed Th1/Th17 immune response is known to mediate clearance of MRSA infection (42, 43). CXCR3 also recruits cytotoxic CD8⁺ T cells that contribute to clearance of infection, potentially including host cells harboring intracellular S. aureus (44). It should be noted that higher expression of genes does not necessarily imply greater inflammatory response. For example, many T and B regulatory cell genes are involved in modulation of immune response. Among these, IL-10 expression is characteristic of regulatory (Treg) and B10 (regulatory B or Breg) cell subsets. The fact that IL-10 levels were reduced in RB outcomes supports the concept that effector T and B cells are likely prioritized over Treg and Breg for effective clearance of MRSA from the bloodstream. Thus, taken together, the current findings support the hypothesis that appropriate expression and polarization of T and B cell responses are integral to resolution of MRSA bacteremia in the setting of vancomycin therapy.

While Th17 and Th1 T cell polarization appear important to RB outcomes, other T cell polarization pathways likely contribute as well. For instance, the transcription factors GATA3 and TCF7 were also identified as top hub genes in the T cell module ( Figures 3A , 4 ). GATA3 promotes Th2 polarization and humoral immunity which protects against exotoxin-mediated complications of MRSA bacteremia (45–47). On the other hand, Th2 pathways may also modulate hyper-inflammatory immune responses to MRSA that may be detrimental to host clearance of infection (48). TCF-7 is a transcriptional activator with critical roles in the development, differentiation, and durability of CD4⁺ and CD8⁺ T cells (49–53). This relationship with CD8⁺ T cells is of special interest, as persistent MRSA may exploit intracellular invasion to evade immune responses (54). Therefore, it is plausible that TCF-7 upregulation contributes to immune responses that guard against persistence strategies by MRSA. Taken together, these findings suggest that specific pathways of activation and polarization of CD4⁺ and CD8⁺ T cells overall or subsets thereof are important contributors to RB outcomes.

The above results indicate crucial hallmarks of effector T cell transcriptomes in RB outcomes. By comparison, upregulation of B cell networks was identified as contributing to RB vs. PB outcomes in context of DNMT3A genotype and IL-10 levels. Our previous work identified a gain-in-function polymorphism in the human DNMT3A gene as associated with a reduced risk of PB outcomes (14). The heterozygous A/C genotype in the DNMT3A gene was associated with lower IL-10 cytokine level and RB outcomes as compared to the A/A genotype which correlated with PB. Hence, in the current study we sought to identify transcriptional networks corresponding to distinct DNMT3A genotypes using GE module analysis. Expression of GE module ME7 was significantly enriched for genes involved in B cell signaling and upregulated in patients with the DNMT3A A/C genotype correlating with RB (14). The top 20 hub genes in B cell function associated with the DNMT3A A/C genotype included: CD1C, CD19, CD22, CD72, CD79A, CD79B, FCER2, and MS4A1 ( Figures 3B , 5 ). Notably, CD19, CD22, CD72, CD79A, CD79B have been previously implicated in B cell-mediated protection against invasive S. aureus infection (55). Other genes upregulated in this module include members of the B cell signaling protein family, which function in B cell receptor activation and regulation of antigen receptor-mediated signaling ( Figures 3B , 5 ). In contrast, FCRL2, FCRLA and NT5E are modulators of B cell response. Together, the fact that these genes were upregulated in context of the DNMT3A A/C genotype suggests epigenotypic regulation of DNA (methylation) modulates protective B cell responses in MRSA bacteremia.

Increased production of IL-10 has been shown in a variety of experimental models and in humans to correlate with worsened outcomes in S. aureus infection (56). Various investigators have also linked DNMT3A change-in-function polymorphisms to impaired T and B cell immune responses via dysregulated cytokine production (57–60).

The current results also suggest novel mechanistic insights underpinning this relationship. For example, the DNMT3A A/C genotype was associated with increased co-expression signatures of T and B cell gene networks. This correlation plausibly aligns to the A/A genotype favoring host susceptibility to PB by IL-10 modulation of protective immune responses ( Figure 6 ). Further supporting this concept, our Random Forest classification model of MRSA clinical outcome using differentially expressed T and B cell hub genes and IL-10 cytokine level had good classification performance (AUC = 0.81). Notably, a subgroup of patients with RB clinical outcome carried the DNMT3A A/C genotype, expressed similar levels of genes in the T cell enriched module, yet had reduced expression of genes in the B cell enriched module. Conversely, a subgroup of DNMT3A A/A positive patients with PB expressed T cell but not B cell module genes. Therefore, the combined DNMT3A genotype, IL-10 level, and T and B cell hub gene signature is a more reliable predictor of RB/PB outcome ( Figure 7 ). The fact that the predictive value of the composite model was affirmed in the validation cohort strengthens the potential to accurately predict RB vs. PB outcomes based on an integration of T and B cell hub gene signatures and IL-10 levels ( Figure 7 ).

Beyond insights into MRSA immune evasion strategies, functions corresponding to identified T and B cell signaling pathway genes associated with PB and RB outcome in relation to IL-10 cytokine level and host DNMT3A genotype may also guide therapeutic development or strategies to address persistence. Several lines of evidence suggest that blocking IL-10 signaling facilitates clearance of viral infection and prevents tumor growth in animal models (61, 62). Likewise, IL-10 blockade increased clearance and abrogated hematogenous dissemination of C. neoformans to the brain implying this strategy has therapeutic potential in treatment of fungal infections (63). Thus, it is tempting to speculate that IL-10 blockade in patients carrying the DNMT3A A/A genotype would promote Th1/Th17-mediated protective immune responses and resolution of MRSA bacteremia. However, IL-10 can also act as a pro-inflammatory cytokine, particularly with respect to CD8⁺ T cell function in certain bacterial infections (64–67). Thus, the specific role and relationship of IL-10 in context of mixed Th17/Th1 responses believed essential for protection against S. aureus bacteremia remain to be explored.

It is important to consider the limitations of this study. The current investigation was designed to detect patterns of host immune response signals, networks and pathways that differ in PB vs. RB outcomes in MRSA bacteremia. While not designed to determine immunologic mechanisms, T and B cell signatures reported in this study were obtained and validated using three patient cohorts from a single center. Further validation with an independent cohort is an ensuing goal. Despite this limitation, using this carefully matched case-controlled study cohort, we not only identified associated transcriptional signatures of protective T and B cell gene networks, but also confirmed previous reports linking the DNM3TA A/C genotype and lower IL-10 cytokine level infection with PB ( Figure 6 ). Future investigation to determine if these signatures exist in MRSA bacteremia patients from other geographic populations will be informative. The present studies derive from patients treated with vancomycin, the gold-standard therapy for MRSA bacteremia. One notable aspect is that not every patient with PB underwent full source control, which may contribute factors influencing the clinical outcome of PB. We are lacking specific clinical information for Minimum inhibitory concentration (MIC) values, time to therapeutic vancomycin levels, and potential missed doses in this study. However, all patients with suspected bacterial sepsis received broad spectrum antibiotic therapy that typically included vancomycin for MRSA due to high prevalence of MRSA at our institution. For this reason, there were likely no delays in vancomycin therapy. Although we are unable to confirm whether the groups were similar in terms of time to achievement of therapeutic vancomycin levels, we believe this is unlikely because vancomycin was administered based on therapeutic drug monitoring for all patients and all patients in each group were already on vancomycin when blood was drawn for transcriptomic analysis. The duration of therapy varied based on the extent of the infection, but generally ranged from two to six weeks. Further studies are also warranted to mechanistically explore whether the transcriptomic signatures associated with RB versus PB in patients treated with vancomycin holds true for other anti-infective therapies used in MRSA bacteremia. Lastly, DNMT3A encodes a DNA methyltransferase that plays a crucial role in epigenetic modifications which influencing gene expression patterns. The relationship between DNMT3A, IL-10 and adaptive immune responses involves intricate regulatory mechanisms within the immune system. Thus, altered DNMT3A activity may influence the expression of T and B cell genes involved in immune regulation, potentially impacting the production of multiple cytokines in response to MRSA (57, 68).

In summary, current findings underscore the importance of lymphocyte-mediated immunity for resolution of MRSA infection. The protective response was associated with genes corresponding to T and B cell functions in context of established IL-10 responses and genotypic relationships in host defense against S. aureus. The fact that patients with RB predominantly carried the DMN3TA A/C genotype is consistent with our prior findings that DNA methylation and epigenetic status governs integrated regulation of T and B cell function, and IL-10 expression shaping PB vs. RB outcomes. Finally, a major strength of this study was validation of key findings from training cohorts using an independent patient cohort analyzed in a masked manner. The implications for future studies include validating the predictive value of omics classification of PB or RB risks in larger cohorts of patients with MRSA bacteremia from diverse populations across clinical endotypes, exploring T and B cell signaling networks eliciting protective immunity to MRSA, and understanding antibiotic-specific relationships. S. aureus has a remarkable ability to evade host innate and adaptive immune defenses (69, 70), thus these insights may shed new light on the dynamic interplay between immune response and organism persistence strategies. The rapid onset of lymphocyte gene expression profiles linked to dysregulated host immunity in PB clinical outcomes makes this gene signature a prime target in the search for the regulatory origins of T cell and B cell dysfunction in MRSA infection ( Figure 6 ). Translation of this knowledge holds promise for guiding development of novel diagnostic methods, innovative anti-infectives, vaccines and immunotherapeutic to address the challenge of persistence in S. aureus and perhaps other infections as well.

The data presented in the this study are deposited in the Sequence Read Archive, NCBI project repository, accession number PRJNA914756. The cytokine and clinical data that support the findings of this study are available on request from the corresponding author, [EFR].

The studies involving humans were approved by Duke University Medical Centre (DUMC). The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation in this study was provided by the participants’ legal guardians/next of kin.

RP: Formal analysis, Investigation, Methodology, Software, Writing – original draft, Writing – review & editing. HP: Data curation, Formal analysis, Methodology, Software, Writing – review & editing. RA: Writing – review & editing, Formal analysis. MR: Writing – review & editing, Conceptualization, Methodology, Supervision. DWG: Funding acquisition, Writing – review & editing, Formal analysis. FR: Data curation, Writing – review & editing, Funding acquisition, Methodology. LC: Data curation, Writing – review & editing. VGF: Funding acquisition, Resources, Writing – review & editing, Conceptualization, Data curation. MRY: Conceptualization, Funding acquisition, Project administration, Writing – review & editing, Resources, Supervision, Validation. EFR: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing.