New Gene Linked to Increased Skin Infection Virulence

Genetic deletions define the mΦ11 gene(s) responsible for increased skin infection virulence. Prophage mΦ11 promotes USA300-mediated tissue damage by increasing skin abscess size during murine infection (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI177872DS1), as previously reported (5). Often, hypervirulent strains of S. aureus will exhibit differences in growth rates (17), secreted protein production (18), and/or transcriptional profiles (19–21). However, strain USA300 LAC* harboring mΦ11 did not show significantly altered in vitro growth kinetics, hemolysis patterns, biofilm production, exoprotein profiles, or transcript levels of non-mΦ11 genes compared with the parental USA300 LAC*, as determined by RNA-Seq analysis (Supplemental Figure 1, B–F). Therefore, results of the in vitro analyses did not correlate with the increased virulence demonstrated by mΦ11-containing strains during skin infection. These data suggest that an in vivo signal(s) is required for mΦ11-associated virulence (22). Thus, in vivo models of infection are required to identify mechanisms underlying mΦ11-mediated virulence (Figure 1A).

Effect of en bloc deletions on the mΦ11-mediated skin abscess phenotype. (A) Skin infection workflow. Created with BioRender.com. (B) Map of mΦ11 in strain USA300-BKV, adapted with permission from Copin et al. (5), with en bloc deletion locations. Arrows indicate predicted ORFs and the direction of the transcription of genes within the unique mΦ11 modules. Homologous (red) and nonhomologous (blue) ORFs are shown, compared with prototypical Φ11. Black arrow indicates pamA. Black bars beneath the gene map correspond to the gene blocks deleted from the indicated strain. (C) Representative images of skin abscesses 72 hours after subcutaneous infection with the indicated strains. Scale bar (black): 1 cm. (D) Skin abscess infections with en bloc deletion mutants. Skin abscess area at the indicated times after infection with approximately 107 bacterial CFU of LAC* lysogens containing mΦ11 (blue, n = 20, strain BS989), mΦ11Δ32–43 (purple, n = 16–18, strain RU47), mΦ11Δ44–57 (salmon, n = 20, strain RU108), or mΦ11Δ32–64 (cyan, n = 18–20, strain RU42). Data are pooled from 2 independent experiments and represent mean ± SD. Statistical significance was determined by Kruskal-Wallis and Dunn’s tests, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

Annotation of the mosaic portion of mΦ11 failed to identify known virulence factors (5). Consequently, we constructed 3 en bloc deletions within the mosaic region of mΦ11 to identify candidate gene(s) (Figure 1B). The deletions were confirmed using whole genome sequencing (Supplemental Figure 2). As expected, deletion of the entire mosaic region containing genes in the replication and lysogeny modules (Δ32–64) eliminated the mΦ11 skin abscess phenotype in mice (Figure 1, C and D). Although deletion of an upstream fragment (Δ32–43) had no impact on abscess size, deletion of the center gene block (Δ44–57) eliminated the skin abscess phenotype (Figure 1, C and D). These data localized skin abscess candidates to 14 genes (genes 44–57) in mΦ11 for further analysis. Of the 14 genes, 8 gene sequences were unrelated to prototypical Φ11 or other known prophages and therefore were considered promising candidates for further analysis (Supplemental Table 1).

An mΦ11-encoded adenine methyltransferase is responsible for increased skin infection virulence. Examination of the 8 potential virulence genes identified a methyltransferase that was absent in wild-type Φ11 (Supplemental Table 1). The mΦ11-encoded adenine methyltransferase (pamA) shares amino acid sequence homology with DNA adenine methyltransferases (dam) (5), so-called orphan methyltransferases that are not paired with a cognate restriction endonuclease and therefore do not form an obvious restriction-modification system. DNA adenine methyltransferases act independently to regulate gene expression and bacterial replication (23–25). They have also been implicated in prophage-mediated pathogenicity of an outbreak strain of E. coli (26). Thus, pamA represented a promising candidate gene as a virulence regulator during skin infection.

To determine whether pamA is necessary for the enhanced skin abscess phenotype, we engineered an in-frame, unmarked pamA deletion in a USA300 LAC* mΦ11 lysogen (mΦ11ΔpamA). Sanger and whole genome sequencing confirmed the deletion and absence, respectively, of adventitious secondary mutations in mΦ11ΔpamA (Supplemental Figure 3). Infection of mice with mΦ11ΔpamA resulted in a nearly identical average skin abscess size compared with that of the control Φ11 lysogen (Figure 2A), suggesting that pamA was necessary for increased virulence. Complementation, by integration of constitutively expressed pamA into the staphylococcal chromosome in a single copy using the S. aureus pathogenicity island (SapI) att site (27), verified that pamA is responsible for the skin abscess phenotype (Figure 2C). We did not observe a difference in bacterial burden at 72 hours after infection (Figure 2, B and D) or dermonecrosis area (Supplemental Figure 5A), as previously reported for comparisons between mΦ11 and Φ11 lysogens (5). Collectively, these data demonstrated that mΦ11-encoded pamA was required for increased skin abscess size in mice.

mΦ11 phage adenine methyltransferase increases skin abscess size without affecting tissue bacterial burden. (A) Effect of pamA on the mΦ11 skin abscess phenotype. Abscess area at the indicated times after infection with approximately 1.5 × 107 bacterial CFU of LAC* containing Φ11 (orange, n = 50 abscesses, strain BS990), mΦ11 (blue, n = 48–50 abscesses, strain BS989), or mΦ11ΔpamA (green, n = 50 abscesses, strain RU39). Data are pooled from 4 independent experiments and represent mean ± SD. Statistical significance determined by Kruskal-Wallis and Dunn’s tests, ***P ≤ 0.001, ****P ≤ 0.0001. (B) pamA skin abscess bacterial burden. Skin abscesses from A with LAC* containing Φ11 (orange, n = 30 abscesses, strain BS990), mΦ11 (blue, n = 30 abscesses, strain BS989), or mΦ11ΔpamA (green, n = 30 abscesses, strain RU39) were harvested at 72 hours and abscess CFU enumerated. Data represent mean ± SD. Statistical significance determined by Kruskal-Wallis and Dunn’s tests. (C) Effect of pamA complementation on abscess size. Abscess area at the indicated times after infection with approximately 1 × 107 bacterial CFU of LAC* containing mΦ11:EV (blue, n = 36 abscesses, strain RU138), mΦ11ΔpamA:EV (green, n = 36 abscesses, strain RU128), or mΦ11ΔpamA:pamA (red, n = 34–36 abscesses, strain RU131). EV, empty vector. Data are pooled from 4 independent experiments and represent mean ± SD. Statistical significance determined by Kruskal-Wallis and Dunn’s tests, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. (D) Bacterial burden in abscesses. Skin abscesses from C of LAC* containing mΦ11:EV (blue, n = 22 abscesses, strain RU138), mΦ11ΔpamA:EV (green, n = 22 abscesses, strain RU128), and mΦ11ΔpamA:pamA (red, n = 20 abscesses, strain RU131) were harvested at 72 hours and CFU enumerated. CFU/abscess is shown due to missing abscess weights during one of the replicate experiments. With the available weight-adjusted data, we found no significant differences between strains (not shown). Data represent mean ± SD. Statistical significance determined by Kruskal-Wallis and Dunn’s tests.

pamA increases CA-MRSA skin abscess size irrespective of other mΦ11 genes and in non-USA300 backgrounds. Next, we hypothesized that pamA expression would increase CA-MRSA virulence independent of other mΦ11 genes. Indeed, wild-type USA300 LAC*-expressing pamA (LAC*:pamA) produced larger abscesses than an empty vector control strain (LAC*:EV), as shown in Figure 3A. A maximal increase in abscess area of 91% was observed at 48 hours after infection. Even during an extended 10 days of postinfection monitoring, despite an overall decrease in abscess size, LAC*:pamA produced significantly larger skin abscesses than those of LAC*:EV (Supplemental Figure 4A). Therefore, pamA was sufficient to increase CA-MRSA skin virulence. As with mΦ11 lysogens, LAC*:pamA did not affect bacterial CFU recovered from skin abscesses (Figure 3B and Supplemental Figure 4B) or lesion dermonecrosis area (Supplemental Figure 5B), indicating that pamA increases abscess size through a mechanism independent of bacterial burden or toxin production, respectively. Considering these findings, we posited that pamA increases abscess size by increasing tissue inflammation. To test this, we compared skin abscess histology and murine cytokine production of LAC*:pamA and LAC*:EV at 72 hours after infection. Skin abscess inflammatory burden (Figure 3, C and D) and proinflammatory cytokine/chemokine production (Figure 3E) were increased in LAC*:pamA skin abscesses compared with control LAC*:EV. Together, these data demonstrated that insertion of pamA into a USA300 LAC* background without the surrounding mΦ11 genes was sufficient to increase local tissue inflammation and thereby skin abscess size.

pamA increases skin abscess size and inflammation in the absence of mΦ11. (A) Effect of pamA on skin abscess size. Abscess area at the indicated times after infection with approximately 1 × 107 bacterial CFU of LAC* with empty vector (EV) (maroon, n = 50 abscesses, strain RU129) or constitutively expressed pamA (purple, n = 50 abscesses, strain RU121) integrated into the chromosome in single copy. Data are pooled from 4 independent experiments and represent mean ± SD. Statistical significance was determined by Mann-Whitney test, ****P ≤ 0.0001. (B) Effect of pamA on CFU recovered from skin abscesses. Skin abscesses (n = 25 abscesses per strain) from 2 independent infections in A were harvested at 72 hours and CFU enumerated. Data represent mean ± SD. Statistical significance determined by Mann-Whitney test. (C) Effect of pamA on skin inflammation. Biopsies of skin abscess (n = 15 per strain, pooled from 2 independent experiments) 72 hours after infection with LAC* containing pamA (strain RU121) or EV control (strain RU129) were H&E stained and inflammatory burden was graded by a blinded dermatopathologist. Statistical significance determined by χ2 test (P = 0.0014). (D) Representative images of skin abscess biopsies from C. One image from each strain is presented according to dermatopathologist architecture classification as nodular (above) or diffuse (below). (E) Effect of pamA on local proinflammatory and vascular proliferation cytokines. Skin abscess biopsies from 3 independent experiments of LAC* with pamA (n = 24 abscesses, strain RU121) or EV control (n = 22 abscesses, strain RU129) were homogenized, and levels of the indicated cytokines were measured. Data represent mean ± SD. Statistical significance was determined by Mann-Whitney test, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

Although pamA was discovered in USA300, a strain within clonal complex (CC) 8, we wanted to test whether the pamA virulence effects are relevant in other clinical MRSA strain backgrounds. While mΦ11 failed to lysogenize non-CC8 strains, suggesting that mΦ11 is lineage specific to USA300/CC8, we continued our investigation by inserting the pamA expression vector into the chromosome of a CC1 MRSA strain from our clinical collection (BVED028). This strain was selected because it represents the same lineage as the well-characterized CA-MRSA prototype MW2, but unlike MW2, its resistance profile was compatible with our vector system. Even in the CC1 background, pamA increased abscess size to a similar degree as our findings in a CC8 (LAC*) background without affecting CFU recovery (Supplemental Figure 6). These data demonstrate that the virulence effects of pamA could extend to MRSA strains from different genetic backgrounds.

pamA-associated skin abscess virulence depends on methyltransferase activity. To determine whether pamA increases skin abscess virulence through methylase activity, we identified the conserved Dam active site (NPPY) in PamA (Figure 4A) and individually introduced several point mutations in residues previously reported to inactivate methyltransferase activity (28). To confirm that the PamA point mutants were inactive, we digested genomic DNA with the restriction endonuclease DpnI, an enzyme that digests at the methylated target of Dam (GATC) (29). As expected, pamA-containing strains, but not those with point mutations in pamA, were susceptible to DpnI digestion (Figure 4B).

The pamA-mediated skin abscess phenotype depends on the methylase activity of PamA. (A) Predicted structure of mΦ11 PamA. Amino acid backbone represented in green, with N-terminus (M1), C-terminus (Q141), and putative active site (N64, P65, P66, Y67) highlighted. Generated by AlphaFold, visualized using PyMol Molecular Graphics System, version 2.5.2 (Schrödinger, LLC). (B) Effect of PamA point mutants on methylase activity. Genomic DNA was isolated from LAC* strains containing the indicated pamA alleles and digested with DpnI (DpnI+) or PBS control (DpnI–), then visualized on a 1% agarose gel. The analysis confirms that PamA methylates at the predicted GATC site and that PamA point mutants lack methylation activity. EV, empty vector. (C) Skin abscess size. Abscess area of LAC* with EV (maroon, n = 20 abscesses, strain RU129), pamA (purple, n = 20 abscesses, strain RU121), and pamAP65T (cyan, n = 20 abscesses, strain RU162) at the indicated time points after skin infection with approximately 1 × 107 CFU of bacteria per abscess. Data are pooled from 2 independent experiments and represent mean ± SD. Statistical significance was determined by Kruskal-Wallis and Dunn’s tests, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. (D) Bacterial burden in abscesses. Skin abscesses from infections in panel C (n = 9–11 abscesses per strain) were harvested at 72 hours and CFU enumerated. Data represent mean ± SD. Statistical significance was determined by Kruskal-Wallis and Dunn’s tests.

For in vivo studies, we used the S. aureus strain containing pamAP65T, since this substitution exhibited the most substantial decrease in Dam methylation activity (28). Consistent with the hypothesis that the methylation activity of PamA contributes to the increased abscess size, LAC*:pamAP65T produced abscesses that were 32%–47% smaller than LAC*:pamA abscesses and similar in size to the LAC*:EV control (Figure 4C). The P65T amino acid change eliminated the pamA-mediated large-size skin abscess size phenotype without affecting tissue bacterial burden in the underlying tissues (Figure 4D). We conclude that the DNA methylation activity of PamA increases abscess virulence.

Identification of genes involved in pamA-mediated virulence. We proceeded to investigate whether pamA epigenetically regulates bacterial gene(s) that result in the hyper-abscess phenotype. Notably, pamA is constitutively expressed in LAC*:pamA, allowing us to bypass the in vivo induction needed to produce mΦ11-related phenotypes. The need for constitutive expression to explore in vitro pamA methylation effects was exemplified by DpnI digest of LAC*:pamA and parental mΦ11 strains, in which the mΦ11 strains showed minimal pamA methylation changes (Supplemental Figure 7). This likely contributes to the absence of in vitro phenotypes in the parental mΦ11 strain (Supplemental Figure 1, B–F) and informed us that LAC*:pamA is required to investigate the effects of pamA methylation on global transcription. Thus, we performed RNA-Seq with LAC*:pamA and LAC*:EV strains during exponential growth in nutrient-restrictive (RPMI) medium chosen to resemble nutrient availability under infectious conditions in human plasma (30). LAC*:pamA induced widespread transcriptional changes in CA-MRSA compared with the LAC*:EV control, with 483 genes differentially expressed (232 overexpressed, 250 underexpressed, adjusted P < 0.05) (Figure 5A). The most significantly upregulated gene in LAC*:pamA compared with LAC*:EV encodes fibronectin-binding protein A (fnbA; FnBPA) (Figure 5A). Quantitative reverse transcriptase PCR (qRT-PCR) confirmed a 15-fold increase in fnbA transcription in the LAC*:pamA strain compared with LAC*:EV (Figure 5B). FnBPA is an S. aureus cell wall–anchored protein that binds adhesive matrix molecules, increasing S. aureus invasion into nonprofessional phagocytic cells (31–33). FnBPA also induces platelet aggregation (34), promotes biofilm formation (35–37), and has been implicated as a virulence factor in endocarditis (38), sepsis (39), implant infections (40), and skin and soft tissue infections (41). Collectively, these observations suggest that fnbA plays a role in pamA-mediated virulence.

pamA induces widespread transcriptional changes, including a large increase in the expression of fibronectin-binding protein A (fnbA; FnBPA). (A) Whole genome transcriptome. Volcano plot of RNA-Seq data comparing LAC* strains containing pamA (n = 3 biological replicates, strain RU121) or empty vector (EV) control (n = 2 biological replicates, strain RU129) after 5 hours of growth in RPMI media. Data points to the right of 0 (green arrow) represent upregulated genes in LAC*:pamA and data points to the left of 0 (red arrow) represent downregulated genes in LAC*:pamA; pamA and fnbA are highlighted. Blue data points represent genes that achieved statistical significance (P ≤ 0.05); pink data points indicate genes that did not. (B) Effect of pamA on fnbA expression. Quantitative real-time PCR of fnbA expression in LAC* strains containing pamA or EV control. Strains were grown and prepared in the same manner as A. Data represent mean SD of 3 biological replicates.

pamA enhances biofilm production in vitro and in vivo by increasing FnBPA. The upregulation of fnbA expression observed in pamA-containing strains, coupled with its association with biofilm-related infections (42), suggest that pamA increases the formation of biofilms. Indeed, we found that LAC*:pamA nearly doubled biofilm production compared with LAC*:EV in an in vitro biofilm assay (Figure 6A). This phenotype reverted to LAC*:EV when pamA contained an inactivating point mutation (Figure 6A). Scanning electron microscopy of in vitro biofilms revealed a more robust bacterial aggregation and architecture, while SYTOX staining demonstrated increased extracellular DNA in the LAC*:pamA biofilms compared with LAC*:EV (Supplemental Figure 8). Thus, the methylase activity of pamA increased biofilm production.

pamA methylase increases biofilm production through fnbA (FnBPA). (A) Effect of pamA methylase on biofilm production. In vitro biofilm quantified by OD after static growth for 24 hours by LAC* strains with the indicated pamA alleles integrated into the chromosome. EV, empty vector. Data represent mean ± SD and are pooled from 2 independent experiments. Statistical significance was determined by ANOVA with Tukey’s test, ****P ≤ 0.0001. (B) Effect of pamA on abscess biofilm. Representative images of skin abscess tissue stained with DAPI (blue) and 5-methylcytosine (5-mC, green) 72 hours after infection with approximately 1 × 107 CFU of LAC* containing pamA (strain RU121) or EV (strain RU129). Scale bar: 200 μm. (C) Biofilm area of LAC* containing pamA (n = 12 abscesses, strain RU121) or EV (n = 11 abscesses, strain RU129) quantified as the difference between DAPI (total extracellular DNA) and 5-mC (eukaryotic host extracellular DNA) (48). Red data points correspond to representative images in B. Data are pooled from 2 independent experiments; individual results are shown in Supplemental Figure 9A. Statistical significance was determined by Mann-Whitney test, **P ≤ 0.01. The difference remained significant (P = 0.007) after removal of the pamA strain outlier (Supplemental Figure 9B). (D) Cell wall proteins. Cell wall–associated proteins from biofilms of LAC* strains containing pamA or EV (3 biological replicates each) were separated by SDS-PAGE. Gel image represents 2 independent experiments. Yellow star = band of interest. (E) Identification of FnBPA bands. Western blot of cell wall–associated protein bands from D using polyclonal anti-FnBPA. (F) Biofilm production. In vitro biofilms from LAC* strains containing the indicated genetic changes quantified by OD. Data represent mean ± SD of 6 biological replicates per strain, pooled from 2 independent experiments. (G) FnBPA production. Western blot of cell wall–associated proteins during in vitro biofilm production by the indicated strains.

S. aureus biofilm has traditionally been associated with device-related infections (43, 44), endocarditis (45), and osteomyelitis (46). However, we and others have found that biofilms also form during S. aureus deep tissue abscess infections (47, 48). Additionally, skin abscess size correlates with in vitro biofilm formation with S. aureus (49). To determine whether pamA-mediated biofilms form in vivo, we quantified biofilm production in skin abscess tissue of LAC*:pamA and LAC*:EV control strains by immunofluorescent staining of extracellular bacterial DNA (48), an abundant component of S. aureus biofilms (50). LAC*:pamA strains produced 6-fold more biofilm compared with the LAC*:EV control (Figure 6, B and C, and Supplemental Figure 9). Thus, pamA stimulated biofilm production in skin abscesses, supporting the idea that pamA-mediated biofilm production is important for pathogenesis of the skin abscess phenotype.

To test whether FnBPA production was increased in pamA-associated biofilms, we compared levels of FnBPA in biofilms from LAC*:pamA and LAC*:EV control strains. Bacterial cell wall–associated proteins from in vitro biofilms, as determined by SDS-PAGE, are shown in Figure 6D. A distinct, high–molecular weight protein band was observed to be more abundant in LAC*:pamA compared with control strain LAC*:EV; there was otherwise considerable similarity in the distributions of the corresponding bands obtained from the 2 strains. Consistent with our transcriptional data, the high–molecular weight protein band was identified as FnBPA by mass spectrometry (Supplemental Figure 10) and confirmed by Western blot (Figure 6E).

To determine whether FnBPA was responsible for increased biofilm production, we compared biofilm formation in an fnbA-inactivated mutant of LAC*:pamA (LAC*:pamA plus fnbA:bursa) and a control strain carrying an empty vector (LAC*:EV plus fnbA:bursa). The results showed that LAC*:pamA plus fnbA:bursa phenocopied the biofilm production of the LAC*:EV strain (Figure 6F); therefore, the fnbA inactivation reversed the biofilm-enhancing effect of pamA. Western blot of cell wall–associated proteins from biofilm-associated bacteria confirmed that LAC*:pamA increased FnBPA production (Figure 6G). Thus, pamA increases biofilm production in USA300 LAC* by increasing production of FnBPA.

pamA increases skin infection virulence through fnbA. To investigate whether fnbA was responsible for pamA-associated skin abscess virulence, we compared LAC*:pamA plus fnbA:bursa mutant and control LAC*:pamA strains. Strain LAC*:pamA plus fnbA:bursa produced 56%–62% smaller abscesses than strain LAC*:pamA (Figure 7A). At the same time, inactivation of fnbA did not affect abscess size in the LAC*:EV control, indicating that the upregulation of fnbA by pamA is required for the observed phenotype. We found no difference in abscess tissue bacterial CFU related to the presence of pamA or fnbA (Figure 7B), supporting the idea that fnbA is necessary for the increased inflammatory response seen in LAC*:pamA.