Staphylococcus aureus USA500 was used as the wild-type strain throughout the study. All strains used in this study are described in Table 1. The plasmid, pmx16, was constructed for continuous expression of exogenous genes. Briefly, primers pmx16-f and pmx16-r (Supplementary Table S1) were mixed, cooled from 72°C to 4°C, and ligated to one segment. The segment with multiple restriction sites was digested with KpnI and SacI (New England Biolabs, United States) and inserted into pCM29 (Pang et al., 2010) to replace the sGFP-encoding gene.

All strains were grown to the stationary phase, then diluted to 1:1,000 in 5 ml of TSB medium, and incubated at 37°C and 220 rpm. Among them, USA500∆hemA, USA500∆hemB was diluted to 7:1000 in 5 ml of TSB medium. Colony-forming units (CFU) were counted for the USA500, USA500∆hemA, USA500∆hemB, USA500∆hemA::pRB473-hemAX, and USA500∆hemB::pRB473-hemB strains at 0, 3, 6, 10, 16, 24, 36, and 48 h by plating 10 μl aliquots of 10-fold serial dilutions on TSA in triplicate. For the hemin complementation experiment, all strains were grown to the stationary phase, then diluted to 1:1,000 in 5 ml of TSB medium, and incubated at 37°C and 220 rpm. Hemin was added at a concentration of 5 μg/ml from the beginning of growth. All experiments were repeated three times.

All strains were diluted to 1:100 in the stationary phase in a fresh TSB. Because of their slight growth deficiency, the USA500∆hemA, USA500∆hemB, USA500∆hemA::pRB473-hemAX, and USA500∆hemB::pRB473-hemB mutants were cultured for 28 h, while USA500 was cultured for 24 h so that all of them reached an equivalent stationary phase. The final bacterial concentration was adjusted to the same level (approximately 1 × 10⁹ CFU/ml) to minimize the quorum sensing influence on the persister assay. After centrifugation, 1 × 10⁹ CFU of the stationary phase bacteria were resuspended in 1 ml of TSB and incubated under different stress conditions, including heat (57°C, 3 h), acid (pH 3.0, 48 h), oxidative stress (50 mM H₂O₂; 3 h), and ciprofloxacin (400 μg/ml, 7 days). Aliquots were withdrawn from each treatment at different time points, and live bacteria were counted as described above. For the hemin complementation experiment, hemin was added at a concentration of 5 μg/ml from the beginning of growth before all kinds of treatment. The bacterial load detection limit was 10¹ CFU. All the experiments were replicated at least three times. Because USA500∆hemA and its derivatives were very sensitive to all stresses, to compare their differences, the heat and oxidative stress conditions were changed to 55°C and 10 mM H₂O₂, respectively.

The previously constructed transposon mutant library was further screened for mutants vulnerable to levofloxacin (12.5 μg/ml) under the same conditions (Li et al., 2009; Wang et al., 2015). The transposon mutants were identified using two rounds of inverse PCR, as described previously (Wang et al., 2015).

To confirm their decreased survival phenotype, hemA insertion mutants were treated with levofloxacin (12.5 μg/ml) in TSB medium for 3–6 days at 37°C. The replica was transferred to Tryptone Soy Agar (TSA) plates and compared with the wild-type strain.

An allelic replacement method was used to knock out hemA and hemB, as described previously (Bae and Schneewind, 2006; Wang et al., 2015; Xu et al., 2017). For hemA, the upstream and downstream allelic sequences were amplified with the following primers: hemA-1-KpnI, hemAX-1Rev, hemAX-2Rev, and hemAX-2-EagI (Supplementary Table S1). Subsequently, the two PCR fragments were ligated using overlapping PCR. The allelic fragment was cut by KpnI and EagI and inserted into the plasmid PKOR2 to construct the plasmid PKOR2-hemA. For hemB, the arc and alsSD operons were knocked out, and gene-1-KpnI/NcoI-1, gene-1 Rev./gene-2 Rev., and gene-2-EagI/EcoRI-2 were used as primers (Supplementary Table S1).

After the candidate genes were knocked out, a shuttle plasmid, pRB473, was used to construct complementation strains. For hemAX and hemB complementation, the genes were amplified with the primers listed in Supplementary Table S1 and ligated to the plasmid, pRB473. For hemB complementation, hemB-Pro pstI-1F and Hem-promoter-1R were used as primers to amplify the promoter, which was then ligated to the hemB part. To complement certain genes in S. aureus USA500∆hemA, the plasmid, pmx16, was used. For overexpression of epiF and the hutIU, USA300HOU1605–09, and asp23 operons, gene-F and gene-R were used as primers for amplification. The primers used are listed in Supplementary Table S1. The recombinant plasmids were transformed into the target strain by electroporation.

The strains of USA500, namely USA500∆hemA, USA500∆hemB, USA500∆hemA::pRB473-hemAX, and USA500∆hemB::pRB473-hemB, were cultured until all of them reached an equivalent stationary phase. They were then treated with levofloxacin (25 mg/ml, 48 h). Total RNA was isolated from these five strains using the RNeasy mini kit (Qiagen, United States) according to the manufacturer’s protocol. For library construction, rRNA was depleted from the total RNA extracted, followed by RNA fragmentation. After first- and second-strand cDNA synthesis, end repair, adding poly(A) tails, adapter ligation, PCR, and product purification, the library was successfully constructed. For quality analysis, SOAP (Li et al., 2008) was used to screen the RNA-seq data, and HISAT (Kim et al., 2015) was used to align the clean reads to the USA300 TCH1516 genome sequence. Rockhopper (Tjaden, 2015) was used to reconstruct the transcripts and classify them into the intergenic region, sense, partially overlapping, and antisense transcripts. Finally, Bowtie2 (Langmead and Salzberg, 2012) was used to align reads to known and novel mRNAs, both of which were regarded as references. The expression levels were calculated using the RSEM software.

Three replicate total RNA samples were separately extracted from strains USA500, USA500∆hemA, USA500∆hemB, USA500∆hemA::pRB473-hemAX, and USA500∆hemB::pRB473-hemB using the RNeasy mini kit (Qiagen) and reverse transcribed using the PrimeScript RT reagent kit (TaKaRa, Japan). Real-time PCR was performed using SYBR Premix Ex Taq II (TaKaRa, Japan) with the primers listed in Supplementary Table S2 to compare expression levels of 38 selected genes. All the kits were used according to the manufacturer’s instructions. The real-time PCR conditions were as follows: 95°C for 30 s, followed by 40 cycles of 95°C for 5 s, 60°C for 30 s, and 95°C for 15 s and then 60°C for 1 min, 95°C for 15 s, and cooling at 4°C. The gene expression levels were calculated using the 2^−∆∆Ct method (Livak and Schmittgen, 2001) with pta and hu (Valihrach and Demnerova, 2012) as the housekeeping genes.

Statistical analysis was performed using the one-way ANOVA with Dunnett’s post hoc test in multiple strains in the GraphPad Prism 8 software (GraphPad Software, San Diego, CA, United States). For in vitro experiments, data from three independent experiments were pooled. p < 0.05 was considered statistically significant. The PossionDis algorithm was used to screen for differentially expressed genes using the RNA-seq data.

In addition to the 260 clones that were found to have decreased survival after 6 days of levofloxacin exposure in the previous study (Wang et al., 2015), we identified seven clones with deficiency in persister formation that had a transposon insertion between nucleotides 854 and 855 of hemA (Figure 1A; Supplementary Figure S1). And mutant in hemA resulting in SCV has been reported before (Curtis et al., 2016; Hubbard et al., 2021).

To examine whether the heme biosynthesis pathway is associated with persister formation, we knocked out hemA and the downstream hemB gene by allelic gene replacement from S. aureus USA500 wildtype (Figure 1B). When either of these two genes was knocked out, the strains displayed a typical SCV phenotype (Figures 1C,D; Jonsson et al., 2003; Garcia et al., 2012; Proctor et al., 2014; Hubbard et al., 2021). After complementing the knockout genes separately, the phenotype was both reverted and the reversion was complete (Figures 1E,F). After culturing on TSA plates for 24 h, the colonies of the mutants were only 0.1 mm in diameter compared with those of the wild-type strain, which had a diameter of 2 mm and were lighter yellow in color. In addition, the growth rates of both ∆hemA and ∆hemB strains were lower (Figure 2A). After 24 h, the CFU counts for the ∆hemA and ∆hemB strains were both 0.68 log lower than that for USA500. After culturing for 48 h, the stationary-phase bacterial concentrations of USA500∆hemA and USA500∆hemB were approximately seven times lower than that of the wild type.

When the heme genes were complemented or hemin was added to the culture medium at a concentration of 5 μg/ml, the growth deficiency and the SCV phenotype in the mutant strains were reversed and the same phenomenon had been reported before (Balwit et al., 1994; Batko et al., 2021), which demonstrated that the growth deficiency was due to the lack of heme biosynthesis (Figure 2B). hemA and hemB encode the first and third enzyme in the biosynthesis process of heme synthesis (Figure 2C), and our results showed knockout either of the two genes impair the production of heme and therefore cause growth deficiency.

Persister formation by these SCV strains was tested under stress conditions, including heat, acidity, and oxidative stress. Under heat stress, the persister levels were lower in the ∆hemA and ∆hemB mutants than in the wild-type strain, by 3.0 log (p < 0.0001) and 3.8 log (p < 0.0001) after 1 h, respectively (Figure 3A). The numbers of persister cells in the ∆hemA and ∆hemB mutants were reduced to 0 after 2 and 3 h, respectively. When the corresponding genes were complemented in the ∆hemA and ∆hemB mutants, the persister levels were restored to the equivalent level in the wild-type strain (p > 0.05; Figure 3A).

Under inorganic acid stress, the persister levels in the ∆hemA and ∆hemB mutants were 4.2 and 3.4 log lower, respectively, than those in the wild-type strain after 6 h exposure. Consistently, both ∆hemA and ∆hemB mutants were killed after 24 h exposure (Figure 3B). Similarly, under oxidative stress, the persister levels in the ∆hemA and ∆hemB mutants were 2.4 log (p < 0.0001) and 1.2 log (p < 0.0001) lower than those in the wild-type strain after 1 h, respectively, and both mutants were eliminated after 2 h (Figure 3C). The overexpressing strains, ∆hemA::pRB473-hemAX and ∆hemB::pRB473-hemB, showed that their persister levels were restored under both oxidative and inorganic acid stress conditions. Not only complementing the heme biosynthesis genes could restore persistence of the mutants under stress conditions but adding hemin (5 μg/ml) to the medium could also rescue persister levels in both ∆hemA and ∆hemB strains to almost those in the wild-type strain under heat, inorganic acid stress, and oxidative conditions (Figures 3D–F).

When the cells were treated with ciprofloxacin at the concentration of 400 mg/ml (Figure 3G), after 5 days, persister level of ∆hemA and ∆hemB was 1.59-log(p < 0.001) and 1.40-log(p < 0.01) lower than the wild-type strain and ∆hemA::pRB473-hemAX and ∆hemB::pRB473-hemB were 0.74-log (p < 0.01) and 0.37-log lower (p > 0.05). After 7 days, persister levels of ∆hemA and ∆hemB were 1.83-log (p < 0.001) and 1.18-log (p < 0.01) lower than the wild-type strain and ∆hemA::pRB473-hemAX and ∆hemB::pRB473-hemB were 1.11-log (p < 0.01) and 0.47-log (p > 0.05) lower. These results showed that persister level of ∆hemA and ∆hemB mutants was reduced under ciprofloxacin stress and complementing the respective genes may not always restore the phenotype.

The data indicate that heme was involved in persister formation in S. aureus, and the lack of heme biosynthesis could result in severe deficiency in persister formation under various stress conditions.

RNA-seq analysis was conducted in the ∆hemA and ∆hemB strains to investigate the downstream genes and pathways after the heme biosynthesis pathway was disrupted, which could provide a clue for the deficiency of persister formation. A total of 64 genes were downregulated, and 23 genes were upregulated in both ∆hemA and ∆hemB mutants (Supplementary Tables S3 and S4). The downregulated genes were mainly associated with amino acid metabolism (hutI, hutU, hutG, and hutH), nucleotide metabolism (purE, purK, purL, purF, purM, purH, and purD), carbohydrate metabolism (acsA1), transporters (mnhA1, mnhB1, mnhC1, mnhD1, mnhE1, glpF, glnQ, and glpT), lipid metabolism (lip), carotenoid biosynthesis (crtM), and environmental information processing (epiF, epiE, epiG, hld, and asp23). Some of these differentially expressed genes were clustered in the same operons (Figure 4).

The upregulated genes were associated with energy metabolism (pgk, tpi, pgm, and fdaB), butanoate metabolism (alsS and alsD), arginine biosynthesis (arcD3, arcB3, and arcA3), transporters (USA300HOU_0762 and USA300HOU_0897), and a multidrug efflux pump (USA300HOU_2333 and USA300HOU_2334).

The RNA-seq results were further confirmed by real-time PCR using the wild type as a control (Supplementary Table S5). Among 34 genes that were selected, transcriptional changes were confirmed in 21 genes, indicating that a large part of the RNA-seq results was reliable.

Since some operons were entirely upregulated or downregulated, they were more likely to be consequent downstream genes or complementary genes of heme deficiency. We therefore constructed strains overexpressing four downregulated genes or operons in USA500∆hemA, including the hutIU (involved in Histidine metabolism), USA300HOU1605–09, and asp23 (encoding alkaline shock protein 23 and other protein) operons and epiF (involved in Quorum sensing), and constructed two double-knockout strains, including ∆arc (encoding arginine/ornithine antiporter) and ∆alsSD (involved in Butanoate metabolism) in USA500∆hemA, to further investigate the mechanism of heme effects on persister formation.

Under acid stress, the overexpression of the asp23 operon restored persister formation in the ∆hemA mutant, with 0.56 and 0.93 log higher survival than that in the ∆hemA control after 2 h and 3 h treatment, respectively (p < 0.05). However, the double ∆hemA∆arc mutants had 1.16log fewer CFU than that in the USA500∆hemA mutant after 2 h treatment (p < 0.05; Figure 5A), suggesting that knocking out the arc operons further inhibited persister formation under acid stress. Under oxidative stress, the overexpression of the hutIU operon and the knockout arc operon further decreased persister formation. The USA500 ∆hemA::pmx16-hutIU strain had 3.27 log fewer CFU than those in the ∆hemA control after 2 h treatment (p < 0.05). The double USA500∆hemA∆arc mutant had 5.35 log fewer CFU than those in the control after 3 h treatment (p < 0.05; Figure 5B). No significant changes were observed under heat stress. (p > 0.05; Figure 5C).

The RNA-sequencing data was uploaded to the SRA database under the Bioproject ID of PRJNA750238.