In vivo screening for identification of fitness factors

To systematically investigate the genetic basis of Streptococcus sanguinis fitness in IE, we performed a genome-wide screen to identify gene deletions that influence competitiveness of pooled mutants in our in vivo model. The model employs introduction of a catheter into the heart to stimulate the formation of sterile vegetations, followed by inoculation of bacteria through an ear vein (see Materials and Methods) (Fig 1A). The mutants tested were derived from our comprehensive non-essential gene knockout library, comprising 2,011 mutants [30], 27 newly annotated open reading frame (ORF) knockout mutants, and one essential gene mutant, Δf₁f_o [32], for a total of 2,039 mutants (S1 Table). These mutants were tested in 24 pools, with sizes of 85–212 mutants per pool (Fig 1; S2 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Experimental design and screening workflow for identification of IE-related fitness factors in S. sanguinis SK36.

A. Sequential steps for evaluating the fitness of mutants using ORFseq with the animal IE model (see Materials and Methods for detailed procedures). Created in BioRender. Bao, L. (2026) https://BioRender.com/22bvwa5. B. Overview of the stages of mutant screening, including initial screening, re-screening, and confirmation, conducted across a total of 24 animal experiments. Created in BioRender. Bao, L. (2026) https://BioRender.com/eswruus.

https://doi.org/10.1371/journal.ppat.1014156.g001

To ensure the reliability of our screens, we included three known IE fitness-reduced mutants as positive controls—ΔpurB (SSA_0046), ΔssaB (SSA_0260), and ΔssaC (SSA_0261)—and a hypothetical-protein mutant with wild-type-like fitness, ΔSSA_0169 (S3 Table) as the negative control. Each pool was tested in three to five catheterized rabbits (Fig 1A; S3 Table).

Fitness of mutants was assessed by calculating the ratio of mutant abundance in cardiac vegetations [21,24] of catheterized animals approximately 20 hours post-inoculation relative to their abundance in the input pools—a measure we will refer to as the “abundance ratio.” In streptococcal endocarditis, pathology results primarily from growth of the bacteria and the resulting growth of the vegetation [7]; thus, in vivo growth is tantamount to IE-associated fitness. Mutant abundance in input pools (the bacterial inocula for animals) and output vegetations was quantified using ORFseq [31], which identifies mutants via ORF-linked APH(3’)-IIIa (kanamycin resistance gene) tags (see Materials and Methods; S1 Fig).

Pool size optimization and validation

A major challenge in mixed mutant library analysis is the occurrence of severe population bottlenecks during host colonization. In highly diverse mutant libraries, some mutants may be randomly lost because of stochastic sampling effects or low initial abundance rather than genuine fitness defects, potentially leading to false-positive identification of conditionally important genes [29]. To minimize these effects, we experimentally optimized the inoculum pool size to maximize mutant representation and preserve library diversity during IE establishment. Our previous signature-tagged mutagenesis study [19] employed a pool size of 40 mutants using a similar rabbit model. To enhance screening efficiency without inducing bottleneck effects [33], we pilot tested 85, 212, and 159 mutants in Initial-Screen Pool-1, Pool-2, and Pool-3, respectively [33] (S2 and S3 Tables). To reduce variation, we pooled similar cell numbers of each mutant in each experiment (Figs 1A and S1; see Materials and Methods).

For Pool 1, all three fitness-reduced controls (ΔpurB, ΔssaB, and ΔssaC) exhibited reduced recovery, while the WT-like control (ΔSSA_0169) [34] was reproducibly recovered from every animal, confirming the pool’s validity. However, in Pool 2 (212 mutants), the WT control showed significantly reduced abundance, indicating a bottleneck effect [33]. We then examined Pool 3 (159 mutants), where fitness-reduced controls were confirmed, and the WT control was reproducibly recovered. Based on these results, we selected 159 mutants as the optimal pool size and employed this approximate pool size for the 21 remaining pools (S2A-S2B Fig; S2, S3 and S4 Tables). All (100%) of the 66 tests of the fitness-reduced control mutants demonstrated significantly reduced abundance ratios, while 21 of 22 tests (95.45%) showed no significant reduction in abundance ratios for the WT control. The lone exception was “initial screen pool-2,” as discussed above.

Identification, description, and conservation of 146 identified IE fitness determinants

During screening, the distribution of mutant fitness values revealed that although some mutants occasionally exhibited increased abundance in individual animals, these effects were inconsistent and likely reflected stochastic variation. Accordingly, we focused on mutants that consistently showed reduced fitness across replicate experiments, defined by an abundance ratio < 1 and statistically significant attenuation based on combined p-values and the number of independent tests (see Materials and Methods). Using these criteria, we identified 146 high-confidence fitness determinants (with the F₁F₀ ATPase operon treated as a single locus) (S5–S6 Tables). Interestingly, 137 (94%) of the 146 candidates identified in this study are novel IE fitness factors in that they have not been previously associated with IE in any bacterium (S6 Table). These include 17 that are annotated as hypothetical proteins with unknown functions.

To define functional relationships among the 146 IE fitness determinants identified in S. sanguinis SK36, these genes were grouped into five functional categories: DNA replication and cell division; transcription, translation and post-translational modification; cell wall synthesis; transport and metabolism; and hypothetical proteins (Fig 2). Genes involved in DNA replication and cell division encoded multiple purine biosynthetic enzymes and factors associated with DNA protection, primosome function, replication initiation, and septation. The transcription, translation, and post-translational modification category encompassed genes involved in transcriptional regulation, RNA degradation, ribosomal proteins, ribosome biogenesis, protein synthesis, proteolysis, and protein secretion. Cell wall synthesis genes included the rml and rgp genes required for rhamnan biosynthesis. Transport-related IE fitness determinants comprised core phosphotransferase system components (HPr and enzyme I), the metal transporter SsaACB, multiple energy-coupling factor (ECF) transporters, and the F₁F_o ATPase. Metabolic genes were enriched for the shikimate pathway, CoA synthesis, pyridoxal phosphate–dependent aminotransferases, serine and tryptophan biosynthesis, and glycolytic enzymes. Seventeen IE fitness determinants were annotated as hypothetical proteins. In addition, two genes—SSA_1509, located within the rgp gene cluster, and SSA_2367, located within an ECF transporter operon—were not directly tested due to their absence from the input library but are likely IE fitness determinants based on genomic context.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Interactions of IE fitness genes.

The panel illustrates pathways in which IE fitness genes interact with other IE fitness genes, organized into five functional groups. Known essential genes within each pathway, as well as mutants unrecovered from the archived library (designated as “not tested), are included. These components are retained within the networks to provide complete biological and metabolic context, allowing experimentally validated fitness factors to be interpreted within fully integrated functional pathways. Pathway names are shown in bold beneath each network. Gene locus identifiers with four digits correspond to SSA_XXXX, whereas those with five digits correspond to J1C87_XXXXX. Numbers inside the circles indicate the number of independent fitness assays performed for each mutant.

https://doi.org/10.1371/journal.ppat.1014156.g002

Several IE fitness genes functionally intersected with pathways containing essential genes, highlighting their integration into core cellular processes. For example, the products of the IE fitness determinants ftsA (SSA_0655) and ezrA (SSA_0879) interact with the essential cell division protein FtsZ [35]; five IE fitness ribosomal proteins interface with an additional 36 essential ribosomal proteins; and the IE fitness genes coaA, coaB, and coaC act together with essential coaD and coaE in coenzyme A biosynthesis. In contrast, other IE fitness pathways—such as rhamnan biosynthesis (rml and rgp genes), purine biosynthesis, the SsaACB metal transporter, and the shikimate pathway—are composed largely of nonessential genes, suggesting requirements for infection that do not apply to in vitro growth.

To assess the evolutionary conservation of these IE fitness determinants, homology searches were performed across 113 genomes from the three genera of Gram-positive cocci that cause the majority of IE cases [36]: streptococci; enterococci; and staphylococci (S6 and S7 Tables). Comparative genomic analysis revealed extensive conservation of IE fitness genes within the genus Streptococcus and substantial conservation across genera. Most IE fitness determinants were present in nearly all Streptococcus strains and were enriched for functions related to DNA replication and repair, ribosome biogenesis and translation, central carbon metabolism, and energy homeostasis, indicating a requirement to maintain core physiological functions in the endocardial niche. Genes involved in amino acid, nucleotide, and cofactor biosynthesis—including the shikimate pathway, purine biosynthesis, folate-mediated one-carbon metabolism, and CoA synthesis—were also conserved. Additionally, genes involved in cell division, envelope integrity, and stress response were among the most conserved IE fitness determinants. While many IE fitness genes were shared across Streptococcus, Enterococcus, and Staphylococcus, a subset—including components of the accessory Sec secretion system, select glycosyltransferases, and several hypothetical or domain-of-unknown-function proteins—was present only within Streptococcus. To determine whether this widespread conservation reflects IE-specific adaptations or general bacterial essentiality, we expanded our comparative genomic analysis to include 73 low-IE-risk control strains (comprising multiple strains of non-endocarditis Streptococcus species [16,36,37] and one strain each of 13 non-streptococcal bacterial species [38–44]. We observed no significant difference in the overall conservation of the IE fitness genes between the 113 IE-associated pathogens and the 73 low-IE-risk strains; while 105 of the 145 fitness genes were present in >50% of IE-associated strains, 98 of these genes met the same conservation threshold among the low-risk pathogens (Fisher's exact test, p = 0.38). Collectively, these data suggest that while these core functions are required for IE, their broad conservation reflects a shared Gram-positive evolutionary background rather than IE-specific adaptation. Thus, the conserved IE fitness architecture comprises general essential functions integrated with species-specific accessory systems.

We next examined whether IE fitness determinants clustered within shared metabolic pathways or protein complexes. Using genomic organization and functional annotation, we identified seven systems in which multiple genes contributed to IE fitness (Fig 3A): the shikimate pathway (S3A Fig), CoA biosynthesis (S3B Fig), rhamnan synthesis (rmlABCD and rgpABCDF; S3B Fig), the EI and HPr components of the phosphotransferase system (PTS; S3D Fig), energy-coupling factor (ECF) transporters (S3E Fig), serine biosynthesis (S3F Fig), and the SsaACB manganese transporter (S3G Fig) [21,45].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Fitness of selected mutants of S. sanguinis SK36 and S. mutans UA159 in the IE model and the rich medium BHI.

A. Log₂ abundance ratios of selected S. sanguinis mutants in the output pool (IE model) compared to the input pool (inocula) across different animals. Each point represents a value of a mutant tested in one animal. For mutants with zero sequencing read counts, the abundance ratios were set to 0.00001 for plotting log₂ values. Dashed red line indicates an abundance ratio of 1. Abun, abundance. B. Log₂ abundance ratios of selected S. sanguinis mutants in the output pool (BHI) relative to the input pool (inocula) across different BHI replicates. Each point represents a value of a mutant tested in one BHI replicate. Zero sequencing read counts were set to 0.00001 for calculation of abundance ratios. Dashed red line indicates an abundance ratio of 1. C. Pairwise comparison of mean Log₂ abundance ratios of selected S. sanguinis mutants in the IE model and BHI. Each point represents the average Log₂ abundance ratio across multiple experiments. Dashed red lines indicate an abundance ratio of 1. The solid gray diagonal line represents equal fitness in both conditions; mutants plotted above the line exhibit a more pronounced fitness defect in BHI than in the IE model (section I), while those below the line are more deficient in the IE model than in BHI (section II). D. Log₂ abundance ratios of S. mutans mutants in the output pool (IE model) relative to the input pool across different animal replicates. Each point represents a single mutant measured in one animal. Dashed red line indicates an abundance ratio of 1. E. Log₂ abundance ratios of S. mutans mutants in the output pool relative to the input pool across BHI replicates. Each point represents a single mutant measured in one BHI replicate. Dashed red line indicates an abundance ratio of 1. F. CFU per animal in vegetations collected 20 hours after inoculation with ΔcoaC and the wild type (WT) strain of S. mutans in the IE model. CFU values represent total bacterial recovery from infections in which wild-type and coaC mutant strains were co-inoculated into individual animals at a 1:1 ratio in the initial inoculum. Note that the CFU for ΔcoaC was set at the limit of detection (LOD) since no colonies were recovered while assuming one colony formed without dilution. Number of animals used is six. The horizontal bar represents the median of each strain, and the asterisk indicates statistical significance (P < 0.05) by Welch’s t-test after log transformation.

https://doi.org/10.1371/journal.ppat.1014156.g003

In vitro growth assay of fitness factor mutants

To determine whether the observed IE fitness defects reflected general growth impairment or niche-specific requirements, we evaluated in vitro fitness in rich medium (BHI) using pooled ORFseq assays. To simulate the physiological environment of our vertebrate IE model, we utilized conditions representing the oxygen level of the left side of the heart (12% O₂) [24,46] and the elevated temperature observed during active infection (39°C) (S4 Fig). This temperature models the systemic febrile response common to human infective endocarditis, where fevers exceeding 38°C occur in nearly 98% of cases [47]. These analyses included three pools from the original IE fitness screen (final pool_1–3 S2 Table) and a newly generated reconfirmation pool (S8–S10 Tables). Of the 146 IE fitness determinant mutants identified in vivo, 137 were represented across the four pools and evaluated for growth (Fig 3B and 3C). Among these, 65 mutants exhibited reduced competitive fitness in BHI (abundance ratio <0.5) (Fig 3C), including two genes annotated as hypothetical proteins. Notably, 112 mutants were more severely attenuated in IE than in BHI. This subset included all six components of the shikimate pathway, as well as components for CoA and serine synthesis, the two core PTS components (EI and HPr), the SsaACB Mn²⁺ transporter, and at least two of the three core ECF transporters. Conversely, 25 mutants—including most components related to rml/rgp—exhibited a greater growth reduction in BHI than in the IE model (Fig 3C).

Growth phenotypes frequently diverged among genes within the same pathway. For example, coaA mutants exhibited severe growth defects in BHI, whereas downstream CoA biosynthetic mutants (∆coaB and ∆coaC) did not (Fig 3B). Similar dissociations between in vivo fitness and in vitro growth were observed across multiple systems.

Genes encoding the shikimate pathway (aroB, aroD, aroE, aroK, aroA, and aroC) formed a contiguous cluster and were all required for IE fitness, yet none exhibited growth defects in BHI, indicating a niche-specific requirement for aromatic amino acid biosynthesis during IE (Fig 3A and 3B; S8–S10 Tables). In contrast, rhamnan biosynthesis genes (rmlABCD and rgpABCDF) were required for both IE fitness and in vitro growth, consistent with a central role in cell wall integrity.

Core PTS components (EI and HPr) and select EII transporters also contributed to IE fitness, with some mutants displaying growth defects in BHI, suggesting roles in both general metabolism and host adaptation. Similarly, ECF transporters showed substrate-specific contributions to IE fitness, with only a subset required for in vitro growth. All three genes involved in serine biosynthesis (serA, serB, and serC) were required for IE fitness but dispensable for growth in BHI, consistent with sufficient serine availability in rich medium [48] but potentially limited serine availability within cardiac vegetations [49].

Finally, mutants lacking components of the SsaACB Mn²⁺ transporter exhibited reduced fitness in BHI, differing from previous observations [45], likely reflecting the competition assay and the higher temperature and oxygen levels used in this study (Fig 3B; S7–S9 Tables).

Together, these results demonstrate that IE fitness is supported by a combination of pathways required for general bacterial growth and others uniquely required in the host niche, underscoring the importance of metabolic capabilities.

Assessment of the contribution of CoA and shikimate genes to growth and IE fitness in S. mutans UA159

To determine whether IE fitness determinants identified in S. sanguinis SK36 are conserved across IE-associated streptococcal species, we used the IE animal model to examine mutants of selected genes in S. mutans, an abundant oral species that is phylogenetically distant from S. sanguinis and exhibits antagonistic interactions with it in the oral environment [50–53]. We focused on genes involved in CoA synthesis and the shikimate pathway, as deletion of all but one, coaA (SSA_1033), produced no detectable growth defect in BHI in S. sanguinis SK36 (Fig 3B). Using targeted deletion, we successfully generated a coaC mutant and deletion mutants of all eight genes within the shikimate gene cluster, including six key shikimate pathway genes (aroA, aroB, aroC, aroD, aroE, and aroK), despite these genes having been previously reported as essential [54].

The generated mutants were pooled at comparable cell densities, together with the SMU277 deletion strain—which shows normal growth in BHI and was used as a wild-type (WT) reference, analogous to ∆SSA_0169 in S. sanguinis SK36. SMU277 is not a true homolog of SSA_0169, and it was selected for comparative evaluation because prior transposon sequencing (Tn-seq) analysis indicated it does not display a detectable baseline growth defect in rich medium [54]. While ∆SMU277 was recovered in abundance from the in vivo IE model (Fig 3D; S11 Table), the only other tested strain with high recovery was the mutant deleted for SMU782, encoding a YlbF/YmcA family competence regulator. The SMU782 result was consistent with that observed for the S. sanguinis strain deleted for the orthologous gene, SSA_1465. In contrast, SMU781, which encodes prephenate dehydrogenase, displayed reduced fitness in the IE model, differing from the outcome in S. sanguinis for the mutant of the orthologous gene, SSA_1466 (S4 and S5 Tables). Pooled ORFseq assays performed in BHI medium in vitro using the same mutant pool revealed that the control strain, ∆SMU277, the ∆coaC mutant, and the ∆SMU782 mutant all displayed abundant growth in BHI, whereas the remaining six shikimate cluster mutants displayed reduced growth in BHI (Fig 3E; S11 Table).

To independently validate the ORFseq result in S. mutans, we performed a co-inoculation competition assay between the coaC mutant—which showed no growth defect in BHI—and the WT strain in the IE model. While the WT strain demonstrated robust recovery, no colonies of the coaC mutant were recovered, confirming a severe IE-specific fitness defect (Fig 3F), consistent with the pooled ORFseq findings.

The growth defect of shikimate pathway mutants can be restored by increased peptide transport in S. mutans

To more deeply investigate additional IE fitness determinants, we next focused on the shikimate pathway. To determine how disruption of the shikimate pathway impacts IE fitness in S. mutans, we screened for compensatory mutations that could restore growth in shikimate pathway mutants, which exhibited strong growth defects in BHI medium for seven of the eight genes tested. Serial passage experiments were performed using newly generated S. mutans UA159 shikimate gene deletion mutants, with ΔSMU782—which showed no growth defect in BHI—used as a control. For most genes, two independently evolved populations were sequenced, while four populations were sequenced for ΔSMU785. Whole-genome sequencing revealed that the evolved mutant populations of four genes whose deletion resulted in growth defects acquired duplications encompassing a peptide transporter gene cluster (SMU255 to SMU259) (Fig 4A–4D), suggesting that increased peptide uptake can compensate for loss of shikimate pathway function in S. mutans.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Compensatory mutations of IE fitness factor mutants in the S. mutans shikimate gene cluster.

A-D. Gene duplications arising within the peptide ABC transporter gene cluster from four independently evolved IE fitness factor mutants in the shikimate gene cluster (SMU778, SMU779, SMU781 and SMU784) of the populations SMU778#1 (A), SMU779#1(B), SMU781#1 (C) and SMU784#3 (D). Whole-genome sequencing was used to confirm genotypes of the target deletion mutants and gene duplications after alignment to the reference genome of S. mutans UA159 (NC_003450). The red boxes indicate the enlarged peptide ABC transporter gene cluster. Vertical red arrow, site of original gene deletion; the height of the blue segments indicates the number of sequence reads mapped to the reference sequence at the coordinates shown on the X axis. A doubling of sequence reads indicates a duplication of the affected region.

https://doi.org/10.1371/journal.ppat.1014156.g004

Competitiveness of S. sanguinis shikimate pathway cluster mutants in BHI and rabbit serum

To further validate the results in the ORFseq in BHI (Fig 3B) and to evaluate mutant behavior in rabbit serum, competition experiments were performed for eight S. sanguinis mutants within the shikimate gene cluster. These included six shikimate pathway genes identified as IE fitness factors. The WT-like strain ΔSSA_0169 was used as a control, and all assays were performed under physiological oxygen levels representative of the left side of the heart (12% O₂) and at 39°C. For competition assays, each mutant strain, as well as the control ΔSSA_0169 strain carrying a kanamycin resistance cassette, was mixed at approximately equal cell numbers with the reference strain JFP36, which harbors an erythromycin resistance cassette inserted at the SSA_0169 locus, and CFUs were enumerated after 24 hours of co-culture. In BHI medium and rabbit serum, mutants deleted for each of the seven genes from SSA_1463 through SSA_1469 showed no detectable fitness defect under either condition (S5A and S5B Fig). In contrast, ∆SSA_1470 exhibited reduced fitness in rabbit serum but not in BHI medium (S5A and S5B Fig).

The in vitro growth defect of S. sanguinis CoA mutants can be restored by increased fatty acid synthesis

To investigate the mechanism by which genes contributed to IE fitness, we first focused on CoA biosynthesis in S. sanguinis. Three CoA synthesis genes—coaA (SSA_1033), coaB (SSA_1201), and coaC (SSA_1202)—were examined using targeted deletion mutants to assess their fitness in BHI medium and rabbit serum; the WT-like strain ΔSSA_0169 was used as a control, and all experiments followed the same design used for shikimate pathway mutants above. Consistent with pooled ORFseq results (Fig 3B), the ΔcoaA mutant exhibited reduced growth in BHI and ΔcoaB and ΔcoaC showed no growth defect (Fig 5A). In addition, in rabbit serum, ΔcoaA, ΔcoaB and ΔcoaC displayed reduced fitness (Fig 5B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Growth phenotypes and compensatory mutations of CoA-biosynthesis IE fitness factor mutants in S. sanguinis.

A–B. Competitive growth of selected mutants relative to the JFP36 strain in BHI medium (A) and rabbit serum (B). Bars represent CFU measured 24 hours after coculture of mutants with the JFP36 wild-type strain. The limit of detection (LOD) indicates the minimum detectable CFU, assuming one colony formed without dilution since no colonies were recovered. The asterisk indicates statistical significance (P < 0.05) by Welch’s t-test after log transformation. C. Colony counts (CFU/mL) across three evolved ΔcoaA mutants (SSA_1033#12, #13, #14_P6), library stock ΔcoaA mutant (SSA_1033_Lib) and WT (SSA_0169). Bars represent the mean colony counts ± standard deviation from three biological replicates. Black circles indicate individual biological replicates. Statistical significance was assessed using one-way ANOVA followed by Tukey's honestly significant difference post hoc test after log transformation. Different letters denote statistically significant differences among groups (p < 0.05); groups sharing the same letter are not significantly different. D. Suppressor mutations arising in the FabT transcriptional repressor from five independently evolved ΔcoaA mutants (#12, #13, #14, #17 and #20). Red boxes indicate the enlarged fatty acid synthesis (FASII) gene cluster. The rRNA operons are indicated by red boxes in the panels. The height of the blue segments indicates the number of sequence reads mapped to the reference sequence at the coordinates shown on the X axis. The vertical arrow in red indicates the deleted coaA gene. The vertical gold bars indicate the positions of the mutations in the fabT gene. E–I. Gene duplications arising within the FASII gene cluster regulated by FabT from three independently evolved ΔcoaA mutant populations #13 (E), #23 (F), #26 (G), #28 (H) and #34 (I). Note that the duplicated regions are flanked by directly repeated rRNA operons (indicated by red boxes in panel E). Vertical red arrow, site of original gene deletion; the height of the blue segments indicates the number of sequence reads mapped to the reference sequence at the coordinates shown on the X axis. A doubling of sequence reads indicates a duplication of the affected region. The red box indicates the region of the fatty acid synthesis (FASII) gene cluster.

https://doi.org/10.1371/journal.ppat.1014156.g005

To explore how altered CoA synthesis might impact IE survival, we screened for compensatory mutations that could restore growth in the S. sanguinis ΔcoaA mutant, which displayed a strong growth defect in BHI medium. Newly generated ΔcoaA mutant populations were subjected to serial passage for six cycles, resulting in multiple independently evolved populations. Comparative growth analysis showed that the evolved ΔcoaA populations formed larger colonies and exhibited markedly improved growth in BHI medium relative to the unevolved library strain (Figs 5C and S6). Whole-genome sequencing revealed that five populations (#12, #13, #14, #17, and #20) acquired mutations in fabT (Fig 5D), a negative transcriptional regulator of the type II fatty acid synthesis (FASII) pathway [55], including one truncation mutation in population #20. In addition, five populations (#13, #23, #26, #28, and #34) exhibited duplications of the FASII gene cluster, with population #13 harboring both a fabT mutation and FASII duplication (Fig 5D–5I). Sequencing of the original ΔcoaA mutant from the ORFseq library confirmed the absence of secondary mutations. Together, these results suggest that increased FASII expression can compensate for impaired CoA synthesis, linking CoA availability to fatty acid biosynthesis.

Ethics statement

All animal experiments were conducted in accordance with institutional and national guidelines for the care and use of laboratory animals. Experimental procedures, including the rabbit infective endocarditis model and blood collection, were approved by the Institutional Animal Care and Use Committee (IACUC) at Virginia Commonwealth University (Protocol No. AM10030; Principal Investigator: Todd Kitten). This protocol has been continuously maintained since 2003 and was most recently approved on May 5, 2025. All procedures adhered to the Guide for the Care and Use of Laboratory Animals and the American Veterinary Medical Association (AVMA) Guidelines for the Euthanasia of Animals. Surgical cardiac catheterization and bacterial inoculations were performed in ABSL-2–certified facilities at Sanger Hall of Virginia Commonwealth University, which provide specialized housing for the New Zealand White rabbit model. Postoperative monitoring and care were conducted in dedicated vivarium facilities with 24-hour access for study personnel under the oversight of the Division of Animal Resources (DAR).

Strains

The 2,048 non-essential mutant library used in this study was sourced from a previous study [30]. Additionally, 27 knockout mutants of newly annotated open reading frames were generated using the same method as in the previous study [30], while Δf₁f_o mutants were generated in our recent study [32] (S1 Table). S. mutans UA159 was obtained from Ann Progulske-Fox (University of Florida).

S. mutans target deletion construction

Streptococcus mutans UA159 knockout strains were generated using an overlap extension PCR strategy as previously described [32]. Briefly, deletion constructs were assembled by fusing approximately 1 kb of the 5′ and 3′ flanking regions of each target gene to a kanamycin resistance cassette. Primers used for construct assembly are listed in S11 Table. For genetic transformation, 50–500 ng of the purified overlapping PCR product was mixed with 200 ng of competence-stimulating peptide (CSP: SGSLSTFFRLFNRSFTQALGK) and 300 µL of competent S. mutans cells. Transformation mixtures were incubated under anaerobic conditions at 37°C for 24 hours. Aliquots of the transformation reactions were subsequently plated onto BHI agar supplemented with appropriate selective antibiotics (kanamycin at 500 µg/mL). Following plate drying, transformants were sealed using an agar overlay technique by adding 1 mL of cooled BHI agar containing the corresponding antibiotic. Plates were incubated at 37°C under anaerobic conditions for 4–5 days to select for stable transformants, which were subsequently verified by PCR.

Mutant pooling

All mutants were individually preserved in 20% glycerol stocks and stored at -80°C until use. Selected mutants were inoculated from the -80°C glycerol stock into Eppendorf tubes containing 300 μL of BHI medium and grown overnight under microaerobic conditions (6% O₂, 7.2% CO₂, 7.2% H₂, and 79.6% N₂) using an Anoxomat (Advanced Instruments, Norwood, MA) jar at 37°C. Following overnight growth, the 300 μL cultures were transferred into 1.2 mL of fresh BHI medium and cultured for an additional 3 hours under the same microaerobic conditions. A 100 μL sample was collected to measure cell density by optical density at 600 nm (OD₆₀₀). Equal OD₆₀₀ values of different mutants were pooled to create input pools (Fig 1).

The cell cultures were then pelleted by centrifugation at 800 × g for 10 minutes at room temperature, and the supernatant was discarded. The resulting 3 mL pellet was thoroughly mixed with 1 mL of 80% glycerol, aliquoted, and stored at -80°C until further use. For preparing inoculum, the 1 mL glycerol stock of a mixed mutant pool was washed twice in 10 mL PBS and adjusted to OD₆₀₀ of 0.8, approximating 10⁸ CFU per mL.

IE animal model

To assess fitness, an endocarditis model was employed as described previously [24] with modifications. In brief, New Zealand White rabbits were sedated, anesthetized, and provided an extended-release analgesic prior to the procedure. Endocardial damage was induced by inserting a PE-90 catheter into the right carotid artery until it met or passed a short distance through the aortic valve, with placement monitored via ultrasound imaging. The catheter was sealed and the incision site was then sutured closed. After a two-day recovery period, sedated rabbits were inoculated with 0.5 ml of a pooled mutant strain suspension prepared as described above via a peripheral ear vein. Approximately 20 hours post-inoculation, the rabbits were sedated and euthanized through intravenous administration of Euthasol. Cardiac vegetations were collected and homogenized in PBS.

DNA isolation

Genomic DNA (gDNA) was isolated from the homogenized vegetation samples and from the inocula. Briefly, cells were pelleted by centrifugation at 9,000 × g for 10 minutes at room temperature, then resuspended in 200 μL of resuspension buffer (20 mM EDTA, 200 mM Tris-HCl, 2% Triton X-100). For lysis, 200 μL of AL lysis buffer (Qiagen, 19075) was added, and the mixture was incubated for 1 hour at room temperature. DNA was precipitated by adding 1 mL of 100% ethanol containing 100 mM sodium acetate. Following washing and drying, the DNA was resuspended in 150 μL of water and prepared for ORFseq library construction or whole-genome sequencing.

• ORFseq library preparation

ORFseq library preparation followed a modification of a previously described protocol [31]. Genomic DNA (gDNA) was fragmented to approximately 500 bp using a Covaris S2 Ultrasonicator under the following settings: Duty cycle - 5%; Intensity - 5.0; Bursts per second - 200; Power - 23 W; Mode - Frequency sweeping; Treatment time - 1/2 = 60 sec/40 sec. PolyC tails were then added to the 3′ ends of the fragmented DNA using terminal deoxynucleotidyl transferase (Promega, USA) at 37°C for one hour, followed by enzyme inactivation at 75°C for 20 minutes. PolyC-tailed DNA fragments were purified with AMPure XP beads (Beckman, USA) and used as templates in PCR amplification with Platinum Taq DNA Polymerase (Invitrogen, 10966026).The first round of PCR was performed with primers olj376 and K10_Truseq (S12 Table) under the following conditions: an initial denaturation at 94°C for 2 minutes, followed by 25 cycles of 94°C for 30 seconds, 60°C for 30 seconds, and 68°C for 30 seconds, with a final extension at 68°C for 5 minutes, and a hold at 4°C. PCR products were then purified using AMPure XP beads. A second PCR round was performed using PE1npKan as the universal 3′ primer (S12 Table) and distinct Truseq_HT primers (S12 Table) to index the samples. The PCR conditions were identical to the first round. The final PCR products were purified again with AMPure XP beads and submitted to the VCU DNA Core Facility for NGS sequencing.

ORFseq library sequencing and quantification of mutants

Sequencing was conducted on the Illumina platform with 100 cycles of single-end sequencing. The leading sequence (TTTTAGTACCTGGAGGGAATAATG), corresponding to the 3’-end of the APH(3’)-IIIa (kanamycin resistance) gene, was trimmed from the reads using Cutadapt [70]. For the ORFseq for S. sanguinis, the trimmed reads were then aligned to the reference S. sanguinis SK36 genome (CP071435.1) using Bowtie2 [71], and read counts were calculated with featureCounts [72]. For ORFseq in S. mutans, the trimmed reads were aligned to the reference S. mutans UA159 genome (NC_004350.2) using Bowtie2, and read counts were calculated based on base coverage within the corresponding genes. Mutant abundance was calculated by averaging two to three technical replicates from input or output samples following normalization by total read counts. The abundance ratio for each biological replicate was calculated by dividing output abundance by input abundance. The overall abundance ratio was determined by averaging abundance ratios across biological replicates.

Screening workflow

The fitness factor screenings were conducted in three stages (Stages I–III). In Stage I, seventeen initial mutant pools (“initial screen pool-1” to “initial screen pool-17”) were tested. Mutants showing significant fitness differences in Stage I were retested in Stage II using three re-screen pools (“re-screen pool-1” to “re-screen pool-3”) (Fig 1B). In Stage III, candidates with abundance ratios significantly different from 1 were further validated and compared directly using four final pools (“final pool-1” to “final pool-4”), enabling relative fitness comparisons among candidates. This stage involved testing in four final pools (“final pool-1” to “final pool-4”) in which the relative fitness of each candidate could be compared with the others. In total, the screening process involved 24 mixed mutant pools, yielding 3,435 tests for 2,039 unique ORF deletion mutants. Of these, 1,210 mutants were tested once with the majority identified as fitness normal, 829 mutants (40.6%) were tested in at least two experiments, while 37 mutants remained untested.

Selection criteria for fitness factor candidates

Fitness factor candidates were identified based on p-values, abundance ratios, and number of tests performed per mutant, as follows.1) Single-Experiment Test Mutants: Mutants tested once were considered fitness-reduced if the abundance ratio was significantly different from 1 and the ratio was < 0.025. This value was chosen because it was half the average ratio for the three fitness-reduced control mutants (ΔSSA_0046, ΔSSA_0260 and ΔSSA_0261). 2) Two-Experiment Test Mutants: Mutants tested twice were classified as fitness-reduced if both results were significant and the overall abundance ratio was < 0.2. This was the rounded highest value for the fitness-reduced controls. 3) Three or More-Experiment Test mutants: Mutants tested three or more times were considered fitness-reduced if the number of significant tests exceeded non-significant ones, all significant tests showed reduced abundance, and the overall ratio was < 0.2.

In vitro competition assay and growth assay

For competition assays, JFP36 (Erm^r; ∆SSA_0169::pSerm [34]) and selected Km-resistant mutants were cultured separately from -80°C glycerol stocks in 2 mL of BHI medium without antibiotics in 4-mL tubes. The cultures were incubated overnight (~16 hours) at 37°C under microaerobic conditions (6% O₂). The next day, cultures of these strains were diluted 1:1,000,000 in BHI or serum. Each of the Km-resistant mutants was then mixed with JFP36 at a 1:1 ratio. Colony-forming units (CFUs) of the 1:1 mixtures were determined at time 0 by plating on BHI-agar containing either 10 μg/mL Erm or 500 μg/mL Km. The mixtures were then incubated for 24 hours at 39°C under 12% O². After 24 hours of growth, CFUs were measured again by plating on BHI-agar containing 10 μg/mL Erm or 500 μg/mL Km.

For the growth assay in Fig 5C, the wild-type (WT; ΔSSA_0196) strain was cultured overnight in 1 mL of BHI broth supplemented with kanamycin, whereas the ΔSSA_1033_lib library stock and three evolved mutants (ΔSSA_1033#12P6, ΔSSA_1033#13P6, and ΔSSA#1033_14P6) were cultured for 48 h in 1 mL of the same medium. To prepare the inocula, each culture was diluted 10,000-fold in triplicate into fresh BHI broth supplemented with kanamycin and incubated anaerobically at 37°C for 20 h. Final bacterial densities were determined by serial dilution and colony-forming unit (CFU) enumeration after plating on BHI agar supplemented with kanamycin and incubating the plates anaerobically at 37°C for 48 h.

Identification of compensatory mutations

Antibiotic-resistant colonies selected on agar plates were inoculated into 1 mL of BHI broth containing the appropriate antibiotics and incubated anaerobically at 37°C for 2–3 days. Cultures were grown until the OD₆₀₀ reached 0.1–0.5; this initial culture was designated P0. For the first passage (P1), 300 µL of the P0 culture was transferred into 3 mL of BHI containing antibiotics and grown to saturation. The P1 culture was mixed thoroughly by pipetting five times with a P1000 pipette, and 50 µL was then transferred into 1 mL of fresh BHI with antibiotics to initiate P2. Cultures were grown to an OD₆₀₀ of 0.1–0.5 before the next transfer. This sequential passaging procedure was repeated through P6. Passages P2–P5 were each grown in 1 mL volumes, whereas the final passage (P6) was grown in 3 mL. Aliquots of P1 and P6 cultures were preserved at −80°C in BHI supplemented with 20% glycerol (final concentration). DNA extraction and whole-genome sequencing were performed using cells from P6. For variant calling, whole-genome sequencing was carried out by SeqCenter (https://www.seqcenter.com/) using the shotgun method with 2 × 150 paired-end sequencing.

Fastq files were aligned to an updated SK36 reference genome sequence (CP071435.1) or S. mutans UA159 genome (NC_004350.2) using Geneious Prime software (https://www.geneious.com/) after trimming via the BBDuk method. Sequences with an average coverage of ≥100 were used for subsequent analysis. Variations in the genome were exported from Geneious Prime. To identify the mutated segment, the frequencies (percentages) of all mutations belonging to a certain segment were determined using the procedure recommended by the makers of Geneious Prime.

Statistics

Statistical analysis. Statistical analyses were performed using Microsoft Excel. Unless otherwise indicated, comparisons of competitiveness between two strains were conducted using two-tailed Welch’s t-tests. Statistical details of specific tests used, are provided in the corresponding figure legends. A P value of <0.05 was considered statistically significant.