In this study, first we aimed to identify Staphylococcus spp. colonized on the skins healed from a PI. In order to analyze the pathogenicity of human-isolated S. caprae and cross-talk between the S. caprae and S. aureus, next we conducted some experiments such as mouse infection models and in vitro hemolysis assay.

A prospective cohort study was conducted in a long-term care hospital with 500 beds in Ishikawa Prefecture (Japan) from 2018 to 2019. Skin samples were collected from 30 patients who had healed from a PI within 1 month, although one sample was excluded because it was obtained immediately after bathing. After sample collection, whether the RPI (lesion or redness) was present at the sites where the samples were collected was assessed every week. Within 6 weeks, seven from a total of 29 patients suffered from RPI (Shibata et al., 2021). Written informed consent was obtained from the participants or her/his guardians before inclusion in the study. Patients who had a skin disease, hypertrophic scars, keloids, or other PIs were excluded. Further information on the participants was described in a previous report (Shibata et al., 2021).

DNA samples from the 29 patients were obtained by attaching an air-permeable medical tape with urethane glue to a healed PI site or a control site, according to a previous report (Ogai et al., 2021b). The control site was opposite the healed PI site when the healed PI was present on a trochanter or ∼5 cm superior to the healed PI site when present on the sacrum or coccyx (Arisandi et al., 2020). Sequence data of 16S rRNA genes (V3–V4 regions) were obtained in a previous report [DNA Data Bank of Japan (DDBJ) accession no. DRA009678 (Shibata et al., 2021)]. To determine Staphylococcus spp. associated with RPI, this study conducted SLST, a recently developed sequencing method that can type and profile the targeted bacterial family beyond the species level (Ederveen et al., 2019). This study selected a Staphylococcus-specific variable region in a gene encoding 30S ribosomal protein S11, which was chosen by a TaxPhlAn pipeline using 7,247 Staphylococcus genomes, for an SLST target. The regions were amplified by KAPA HiFi HS ReadyMix (Roche, Basel, Swiss) and a pair of SLST PCR primers (Supplementary Table 1), followed by purification by AMPure XP magnetic beads (Beckman Coulter, Brea, CA, United States). After the addition of index sequences using the Nextera XT Index Kit version 2 (Illumina, San Diego, CA, United States) and subsequent purification by AMPure, the equimolar mixtures of the products were applied for MiSeq sequencing with the MiSeq Reagent Kit version 3 (Illumina) and 15% of PhiX Control version 3 (Illumina). Sequence data were deposited to the DDBJ (accession no. DRA013063).

The 300-bp pair-end reads (fastq data) were filtered (QC > 20) and merged by the fastp tool (Chen et al., 2018), followed by conversion into FASTA-format sequence data by the seqtk tool.¹ Sequence data were applied to the TaxPhlAn pipeline (Ederveen et al., 2019). The alleles were built with a Shannon diversity index threshold of 0.6. Unassigned alleles were manually extracted and applied to nucleotide BLAST in the National Center for Biotechnology Information website and regarded as those derived from species with the highest scores.

A 16S rRNA metataxonomic analysis was previously conducted for identical samples, and the abundance of Staphylococcus spp. in the microbiota was determined (Shibata et al., 2021). 16S rRNA and SLST data were assembled by adjusting the abundance of each Staphylococcus spp., which were obtained by SLST, to the abundance of Staphylococcus spp. in the whole microbiota obtained by 16S rRNA metataxonomics (Supplementary Figure 2).

Staphylococcus caprae ATCC 35538 [strain designation: 143.22 (CCM 3573)] (Devriese et al., 1983), derived from goat milk, and Staphylococcus epidermidis ATCC 14990 [strain designation: Fussel (NCTC 11047, R. Hugh 2466)] were obtained from the American Type Culture Collection. S. caprae JMUB145 was isolated from the blood culture of a woman with fever during chemotherapy treatment for uterine cancer at Jichi Medical University Hospital in Japan (Watanabe et al., 2018). S. aureus N315 (agr type II) was kindly provided by Dr. Keiichi Hiramatsu (Kuroda et al., 2001).

After incubation in TE buffer containing lysostaphin and lysozyme at 37°C for 30 min, genomic DNA was extracted by the QIAamp DNA mini kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions.

To analyze the allele types of the agrD gene in the clinical samples, qPCR was conducted using QuantiFast Probe PCR kits (Qiagen), primers, and probes described (Supplementary Table 1) in the AriaMx Real-time PCR System (Agilent Technologies, Santa Clara, CA, United States). Before the analysis, amplification specificity was validated using the indicated copy numbers of ATCC35538 and JMUB145 genomes that possess the agrD_YSTCSYYF and agrD_YRTCNTYF genes, respectively. To prepare standards, genomic DNA was extracted using QIAamp DNA mini kit according to the manufacturer’s instruction. Double-stranded DNA (dsDNA) concentrations were measured using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, Waltham, MA, United States).

Five-week-old female BALB/c mice (Sankyo Laboratory Service Corporation, Tokyo, Japan) were anesthetized through intraperitoneal administration of a mixture of Dormicum (4 mg/kg; Astellas Pharma, Tokyo, Japan), Vetorphale (5 mg/kg; Meiji Seika Pharma Co., Ltd., Tokyo, Japan), and Domitor (0.3 mg/kg; Nippon Zenyaku Kogyo Co., Ltd., Fukushima, Japan), followed by disinfection of mouse backs by ethanol. A circular full-thickness wound (4 mm in diameter), including the panniculus carnosus muscle, was made using a sterile biopsy punch (Kai Industries Co., Ltd., Gifu, Japan). Seven days after wound formation, the backs of control (no surgery) and healed mice were shaved under anesthesia by the Dormicum/Vetorphale/Domitor mixture. A 10 μL aliquot containing no bacteria [negative control for counting colony-forming units (CFUs) of commensal bacteria], 2.0 × 10⁷ CFUs of S. aureus N315, 3.2 × 10⁷ CFUs of S. caprae JMUB145, or 3.5 × 10⁷ CFUs of S. epidermidis ATCC 14990 [strain designation: Fussel (NCTC 11047, R. Hugh 2466)] in phosphate-buffered saline (PBS) was inoculated directly onto the scabs or control sites by a micropipette. The droplets were confirmed to be dried under anesthesia. On Day 2 post-infection, mice were euthanized with sevoflurane (Maruishi Pharmaceutical Co., Ltd., Japan). The skin (∼1 × 1 cm) was collected by sterile scissors, suspended in sterile PBS, and mixed vigorously by a vortex mixer.

Eight-week-old female BALB/c mice were anesthetized through intraperitoneal administration of the Dormicum/Vetorphale/Domitor mixture. Mouse backs were shaved and disinfected by ethanol. A 100-μl bacterial solution in PBS was injected to the subcutis in two separate experiments in triplicate for each group. The S. aureus strain N315 (1.3 × 10⁸–1.9 × 10⁸ CFU/mouse), S. caprae strain JMUB145 (1.3 × 10⁸–1.9 × 10⁸ CFU/mouse), S. caprae strain ATCC 35538 (1.1 × 10⁸–1.5 × 10⁸ CFU/mouse), S. epidermidis ATCC strain 14490 (1.9 × 10⁸–2.7 × 10⁸ CFU/mouse), and coinjection of S. aureus strain N315 and S. caprae strain JMUB145 were tested. The size and color intensity of lesions were calculated by ImageJ. To quantify the intensity, image files (pictures) were converted to grayscale, the lesions were selected by drawing/selection tools, and the area-integrated intensity was calculated.

Using normal skin with the same size (1 cm apart from the selected area) as the control, the intensity was calculated as (intensity of the lesion)/(intensity of the normal skin).

Staphylococcus aureus N315, S. caprae JMUB145, and S. caprae ATCC 35538 were cultured in tryptic soy broth (TSB; Becton Dickinson, United State) overnight at 37°C. After centrifugation at 6,000 × g at 10 min, spent medium (supernatant) was collected. Cultured bacteria were suspended in fresh TSB or TSB containing spent medium (final 10%) at a density of OD_600nm = 0.001. After overnight culture and centrifugation at 6,000 × g at 10 min, spent media were utilized for analysis.

Human blood was obtained from 22 to 38-year-old healthy volunteers. Human whole blood was collected in EDTA-containing tubes and centrifuged at 1,500 × g at 20 min. After washing with PBS thrice, cells were suspended at 20% (v/v) in PBS containing 0.5% bovine serum albumin (BSA; FujiFilm Wako Pure Chemical Corporation, Osaka, Japan). Rabbit whole blood in Alserver’s solution (Kohjin-Bio, Saitama, Japan) was centrifuged at 1,200 × g at 20 min. Rabbit blood cells were washed with PBS containing 0.1% BSA four times and suspended at 20% (v/v) in PBS containing 0.5% BSA.

The human or rabbit red blood cell (RBC) solution (200 μL) was incubated with 20 μL spent medium or fresh TSB (for blank control) for 6 h (human) or 18 h (rabbit). After centrifugation at 1,500 × g at 20 min, the absorbance of the supernatants was measured at a wavelength of 492 nm.

The spent medium was obtained after overnight culture, centrifuged at 6,000 × g at 10 min, and filtered. S. aureus N315 and S. caprae JMUB145 were suspended in fresh TSB or TSB containing spent medium (final 10%) at a density of OD_600nm = 0.001 and cultured at 37°C with shaking for 10 h.

The clinical study (Kanazawa University no. 44) and animal experiments (Kanazawa University no. AP-214236) were approved by the Medical Ethics Committee of Kanazawa University and conducted in accordance with the Declaration of Helsinki and the Microorganism Safety Management Regulations of Kanazawa University.

We focused on skin microbiome associated with RPI. In a previous report, 16S rRNA metataxonomics revealed that the abundance of Staphylococcus spp. was significantly higher in RPI-healed sites than non-RPI-healed sites (p = 0.002, Mann–Whitney U test) (Shibata et al., 2021) and that the abundance was also significantly in the healed sites (including RPI and non-RPI) than the control sites (p < 0.01, Welch’s t-test), indicating composition of staphylococci is involved in RPI. In this study, Staphylococcus spp. were identified using a recently developed sequencing method named SLST. A Staphylococcus-specific region in a gene encoding 30S ribosomal protein S11 was amplified by PCR and sequenced by the MiSeq system. After merging pair-end reads, removing low-quality reads, and converting read data to multi-FASTA format, data were applied to the TaxPhlAn pipeline (Ederveen et al., 2019). The pipeline gave numbers of sequences derived from each Staphylococcus. The abundance of each Staphylococcus was adjusted to that of Staphylococcus spp. obtained by 16S rRNA sequencing in the previous report (Figure 1 and Supplementary Figures 2, 3). Staphylococcus spp. were mainly dominated by S. caprae and S. aureus, whereas the abundance of S. epidermidis was low. The abundance of S. caprae was significantly higher in healed sites than control sites (p < 0.01, Welch’s t-test; Figure 2A). The abundance of S. caprae was also significantly higher RPI-healed sites than non-RPI-healed sites (p = 0.01, Kruskal–Wallis test and Bonferroni correction; Figure 2B). S. aureus was high in healed sites (p = 0.07, Welch’s t-test; Figure 2C). The multiple comparison procedure (Kruskal–Wallis test) did not detect significance among the four groups (Figure 2D). There was no significant difference in the abundance s of Staphylococcus haemolyticus or S. epidermidis (Supplementary Figure 4). As shown in our previous report, there was no significant difference in the abundance of Corynebacterium spp. between RPI-healed sites and non-RPI-healed sites (Shibata et al., 2021) but RPI-control sites and RPI-healed sites (p = 0.02, Kruskal–Wallis test and Bonferroni correction) (Supplementary Figure 4).

This study observed the Staphylococcus composition of the seven healed skins that suffered recurrence 1–6 weeks later. The dominance of S. caprae was found in four of the seven patients, whereas that of S. aureus was found in the other two samples (RS #6 and #7 in Figure 3). In one patient’s skin (RS #5) that suffered recurrence in 4 weeks, both S. caprae and S. aureus colonized. Patients who had the S. aureus-dominant microbiome suffered from RPI within 1 week, whereas those with the S. caprae-dominant one did after 2–6 weeks. There were no significant characteristics in the patient information.

Staphylococcus caprae, CoNS, usually colonize goat skin and milk and occasionally cause goat mastitis (Devriese et al., 1983). Recent reports showed that S. caprae also reside as human commensal bacteria and occasionally cause nosocomial infections (Seng et al., 2014). Two S. caprae strains ATCC 35538 and JMUB145, isolated from goat and human, respectively, showed 98.5% average nucleotide identity (Yoon et al., 2017), indicating that their characteristics were slightly different. Some of staphylococci such as S. aureus is typed by an accessory-gene-regulator (agr) gene sequence (Dufour et al., 2002). Paharik et al. (2017) reported that S. caprae from goat milk possesses an agr gene encoding autoinducing peptide YSTCSYYF. Watanabe et al. (2018) reported that clinical isolate S. caprae possesses another allele type of the agrD gene encoding YRTCNTYF (Figure 4A), although mature structures of the peptides have not been confirmed yet. To determine which type of S. caprae was found in the healed ulcers, qPCR was conducted to count the copy numbers of agrD_YSTCSYYF and agrD_YRTCNTYF genes (Supplementary Figure 6). Most agrD genes in the samples were agrD_YRTCNTYF in healed sites (Figure 4B), indicating that genotypes of S. caprae strains in the clinical samples are similar to S. caprae JMUB145 rather than ATCC 35538. Thereafter, this study mainly utilized the human isolate S. caprae JMUB145 for analysis.

This study indicated that S. caprae and S. aureus colonized on the healed PI site. To examine the ability of colonization on the skin after wound healing, a full-thickness wound on the back of mice was made, and 7 days were allowed to pass for the wound to heal (Figure 5A). Using this skin scab model, PBS containing S. caprae, S. aureus, and S. epidermidis were spotted (inoculated) onto the control (i.e., uninjured) and scab sites. The colonized bacteria were collected and counted on Day 2 postinoculation. S. epidermidis (strain ATCC 14990) colonized on the control skin at a significantly higher ratio than skin scab (p < 0.05, one-way analysis of variance) than S. aureus (agr type II strain N315) and S. caprae (strain JMUB145), although the ratio was not altered when this bacterium was inoculated on skin scab (Figure 5B). In contrast, S. aureus N315 and S. caprae JMUB145 colonized on skin scab at a significantly higher ratio than control skin (p < 0.05, Student’s t-test), suggesting that both S. aureus and S. caprae prefer colonization on healed skin rather than on the normal skin.

The mouse model revealed that both S. caprae and S. aureus could colonize on healed skin (Figure 5). However, it remained unknown whether the dominance of these bacteria is directly associated with the deterioration of skin integrity, such as RPIs. In the colonization experiments (Figure 5), there was no significant inflammation or ulcer recurrence at the bacteria-inoculated healed skin, suggesting that another model is required to be developed to resemble RPI. Although several reports have shown how to mimic PI in animal models, it is still very difficult to mimic the RPI situation with animal models. Thus, this study analyzed the pathogenicity of S. caprae and S. aureus by a subcutaneous injection model. Subcutaneous injection of S. aureus strain N315 resulted in lesion formation with the largest size and highest color intensity on Day 4 (Figure 6 and Supplementary Figure 7). In contrast, injection of S. caprae JMUB145 alone or S. caprae ATCC 35538 alone did not result in lesion formation. When S. aureus N315 and S. caprae JMUB145 were co-injected, the resultant lesion sizes were significantly smaller than those caused by S. aureus alone (Figure 6).

α-Hemolysin (Hla) plays a key virulence factor in the pathogenicity of S. aureus (Bayer et al., 1997). Regarding the suppression of Hla production, Paharik et al. (2017) reported that an autoinducer peptide composed of eight amino acids (YSTCSYYF) with a C-terminal five-membered thiolactone ring produced by S. caprae from goat milk could suppress Hla production and lesion formation of S. aureus on mouse skin. Their result was consistent with this study. Watanabe et al. (2018) reported the complete genome sequences of three S. caprae strains isolated from patients with nosocomial infections in Japan. Inconsistent with the report by Paharik et al. (2017) the three S. caprae strains isolated in Japan, including strain JMUB145, possessed agrD genes that encode different eight amino acids (YRTCNTYF; Figure 4A). To examine whether the two S. caprae agrD peptides (YSTCSYYF and YRTCNTYF) exhibit different characteristics to Hla production by S. aureus, this study tested the effects of AgrD peptides from S. caprae strains ATCC 35538 (agrD_YSTCSYYF) and JMUB145 (agrD_YRTCNTYF) on the hemolysis of rabbit RBCs by S. aureus. Spent media obtained from overnight TSB culture was utilized for hemolytic assay. Bacteria were also cultured in the presence of 10% spent medium. Hemolytic activity was tested using spent medium. As shown in Figure 7A, S. aureus N315 exhibited hemolytic activity on rabbit RBCs. The hemolytic activity was significantly enhanced when S. aureus N315 was cultured in the presence of 10% S. aureus N315 spent medium (P = 0.003, Student’s t-test). In contrast, the hemolytic activity using spent medium was partially suppressed by culture in the presence of 10% S. caprae JMUB145 (p < 0.001) but not ATCC 35538 (p = 0.898), indicating that ability of suppressing Hla production by S. aureus is different between the two S. caprae strains. In the previous report by Paharik et al. (2017) half maximal effective concentration (EC₅₀) was shown, indicating that concentration might be required to examine the inhibitory effects of the ATCC 35538 supernatants.

Hla exhibited hemolytic activity on rabbit RBCs but not human RBCs (Klainer et al., 1972). Consistently, the spent medium of S. aureus N315 did not exhibit hemolytic activity on human RBCs (Figure 7B). In contrast, S. caprae strains ATCC 35538 and JMUB145 possess several genes encoding β-hemolysin, δ-hemolysin, and putative hemolysins (Watanabe et al., 2018). The two S. caprae strains exhibited hemolytic activity on human RBCs (Figure 7B). Interestingly, the activity of S. caprae strain JMUB145 was significantly higher than the S. caprae strain ATCC 35538 (p < 0.001, Student’s t-test). Moreover, hemolysis was dramatically suppressed by culture in the presence of 10% S. aureus N315 spent medium (p < 0.001), showing that hemolysin production by S. caprae is suppressed by S. aureus. The spent medium did not affect the growth speed of S. aureus or S. caprae (Supplementary Figure 8).