Experimental animals and ethical approval

Eight-week-old female Sprague-Dawley rats were used for this study. All operations were performed under sterile conditions. All animal experiments were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee of our hospital.

Bacterial inoculum preparation

In this study, green fluorescent protein (GFP)-tagged MRSA (ATCC BAA-1556; ATCC, Manassas, Virginia, USA) was used as a representative pathogen for PJI in our rat model. We maintained the MRSA culture on tryptic soy agar (BD Tryptic Soy Agar, BA-256665; BD, Franklin Lakes, New Jersey, USA) supplemented with 5% sheep blood (BD) and 10 µg/ml chloramphenicol (Sigma-Aldrich, St. Louis, Missouri, USA). Several colonies (commonly three to five colonies) were harvested, suspended in 5 ml of tryptic soy broth supplemented with 10 µg/ml chloramphenicol, and incubated overnight on a shaker maintained at 37°C. We then washed the suspended bacteria with phosphate-buffered saline (PBS) three times by successively centrifuging, discarding the supernatant, and resuspending. After the final wash, we adjusted the cells to a concentration of 1 × 10⁸ colony-forming units (CFUs) per millilitre, according to a standard curve produced by measuring optical density (OD).

The minimum inhibitory concentration determinations for the GFP-tagged MRSA strain

Minimum inhibitory concentration (MIC) for GFP-tagged MRSA strain was determined by preparing serial dilutions of the antibiotic in tryptic soy broth (32, 16, 8, 4, 2, 1, 0.8, 0.6, 0.4, 0.2 µg/ml) in triplicate. MRSA was added to the serial dilutions at a concentration of 1 × 10⁴ CFUs/ml. The bacteria were then cultured for 24 hours with shaking at 37°C. Bacterial growth was assessed by the OD at 630 nm. The MIC was determined to be the dose of antibiotic that inhibited bacterial growth by > 90%. The MICs of gentamicin and linezolid were 0.8 µg/ml and 1 µg/ml, respectively.

MRSA intestinal colonization

We developed the protocol for MRSA colonization in rats based on a previous report and our prior pilot work (Figure 1a).¹⁹ Briefly, the rats were deprived of all food during the MRSA intestinal colonization procedure. We first disrupted the normal intestinal microbiota by treating the rats with 200 mg/kg metronidazole (Catalog No. S1907; Selleck Chemicals, Houston, Texas, USA) by oral gavage and 50 mg/kg ampicillin (Catalog No. S4148; Selleck Chemicals) by intramuscular injection every eight hours for 24 hours (four times total). Oral gavage of 1 ml of MRSA solution (1 × 10⁸ CFUs per ml) was performed two hours after each antibiotic treatment for a total four times. To validate successful colonization, we cultured the shed stool (post-colonization day 1, 3, 5, 7, and 9) on tryptic soy agar (BD Diagnostic Systems Europe, Franklin Lakes, New Jersey, USA) supplemented with 5% sheep blood (BD Diagnostic Systems Europe) and 10 µg/ml chloramphenicol (Sigma-Aldrich). The latter was used to select for MRSA colonies.

Fig. 1

a) A schematic protocol of methicillin-resistant Staphylococcus aureus (MRSA) intestinal colonization. Rats received oral gavage of 200 mg/kg metronidazole and intramuscular injection of 50 mg/kg ampicillin every eight hours (solid line). Oral gavage of 1 ml of MRSA solution (1 × 10⁸ CFU per ml) was performed two hours after each antibiotic treatment (dashed line). b) The stools collected from 40 rats were cultured at day 1, 3, 5, 7, and 9 after MRSA intestinal colonization. The intestine tissues were harvested and cultured after final sacrifice.

Sterile surgical procedures and postoperative management

A total of 40 rats were divided into two groups, with one group receiving surgery at eight hours and the other at 72 hours after intestinal MRSA colonization. After the rat was anaesthetized, the surgical site was shaved clean of fur and thoroughly scrubbed with povidone-iodine. A fenestrated sterile drape was positioned over the surgical field, and the patella was exposed by making a longitudinal skin incision over knee. The joint was opened through a suprapatellar approach to expose the intercondylar notch of the distal femur. A 22-gauge needle was first inserted into the femoral canal in a retrograde fashion at the notch, followed by bone reaming with 20- and 18-gauge needles. To simulate a prosthetic joint implant, we used a 20 mm-long Kirschner wire (Kirschner wire (K-wire); 0.088 mm in diameter). We then inserted the sterile K-wire through the reamed tract and left a 1 mm protruding segment in the knee joint (Figure 2). The joint and incision were closed with interrupted 4-0 sutures.

Fig. 2

a) Representative anteroposterior radiograph of a rat femur showing a surgically inserted Kirschner wire within 1 mm of the intra-articular prominence. b) With methicillin-resistant Staphylococcus aureus (MRSA) strain, five out of 20 rats (25.0%) in the eight-hour group developed prosthetic joint infection (PJI) and two out of 20 rats (10.0%) in the 72-hour group developed PJI. With two MRSA isolates, PJI was detected in six and seven out of 20 rats (30.0% and 35.0%), respectively.

To exclude the possibility of intraoperative bacterial contamination, we obtained intra-joint specimens for subsequent culturing with a sterile cotton swab before wound closure.

Acquisition of specimens for bacterial culture

Rats were sacrificed with an overdose of anaesthetic (pentobarbital sodium, 200 mg/kg) ten days after implant insertion in order to obtain specimens for bacterial culturing. Sterile instruments, surgical techniques, and tissue preparations were used throughout the entire specimen harvesting process. The surgical site was first draped in a fashion identical to that done for the simulated implant surgery. Specimens of the knee joint capsule, the femur, and the implanted K-wire were harvested. Tissues were placed in sterile vessels for subsequent culturing. For the bone and joint capsule specimens, the tissues were homogenized with a tissue grinder (Boster Biological Technology, Wuhan, China), and K-wires were sonicated to release bacteria from the biofilm on the wire before culturing.

Bacteria culturing and identification

All specimens were cultured on tryptic soy agar (BD Diagnostic Systems Europe) supplemented with 5% sheep blood (BD Diagnostic Systems Europe) and 10 µg/ml chloramphenicol (Sigma-Aldrich) for 48 hours at 37°C. The presence of PJI was defined as any positive bacterial culture from the bone, joint capsule, or implant. Antibiotic-resistant MRSA colonies, if present, were harvested for bacterial identification by 16S ribosomal DNA and GFP gene expression assessment to rule out contamination. Bacterial identification was conducted with 16S ribosomal DNA sequencing by using the MicroSeq 500 microbial identification system (ThermoFisher, Waltham, Massachusetts, USA). The expression of GFP in MRSA was assessed via real-time polymerase chain reaction (PCR) (Applied Biosystems (ABI); ThermoFisher) to confirm that the MRSA in the gut was the same as that isolated from the joint that caused PJI. Notably, the reliability of PJI diagnosis would be suspected if the sum of three colony counts (bone, joint capsule, and K-wire) was below 20. If present, the outcome from this subject would be abandoned. In this study, no such condition was encountered.

GFP-positive neutrophil isolation by flow cytometry

Neutrophils were first isolated from whole blood with a protocol used in previous studies.^20,21 Briefly, a 5 ml syringe prepared with heparin (ten units) was prepared for blood collection. Rats were then euthanized with pentobarbital sodium (200 mg/kg) and whole blood was collected by cardiac puncture with a 20-gauge needle. We then centrifuged the blood at 200 g for ten minutes at room temperature. The whitish buffy coat layer at the plasma interface was collected after centrifugation. After washing with PBS, the pellet of white blood cells was resuspended in Hank’s balanced salt solution and layered on top of 1079 and 1119 Histopaque density gradient in a centrifuge tube. The solution was centrifuged at 700 g for 20 minutes at room temperature. The neutrophil-rich layer was collected at the interface between the Histopaque 1079 and 1119. Contaminating red blood cells were then removed by briefly lysing with hypotonic lysis buffer, followed by washing with PBS and additional centrifugation. The purity of isolated neutrophils was assessed using a flow cytometer (BD) with allophycocyanin (APC)-conjugated CD11b and PE-Cy7-conjugated CD 45 antibodies (BD) (Supplementary figure a).

After GFP-tagged MRSA is phagocytosed by neutrophils, the neutrophils can be sorted by flow cytometry on the basis of their GFP label fluorescence. Neutrophils were sorted into GFP-positive and GFP-negative subpopulations. A representative image of GFP-positive neutrophils is shown in Figure 3a. In this study, shown in Figure 3b, we used MRSA-carrying neutrophils isolated from other rats injected intravenously with MRSA (MNiv) and MRSA-carrying neutrophils isolated from other rats that had undergone intestinal MRSA colonization (MNin).

Fig. 3

a) A representative image of green fluorescent protein (GFP)-positive neutrophils after flow cytometry isolation. GFP-negative cells obtained from isolation were used as control. b) Schematic definitions of methicillin-resistant Staphylococcus aureus (MRSA)-carrying neutrophils isolated from other rats injected intravenously with MRSA (MNiv) and MRSA-carrying neutrophils isolated from other rats that had undergone intestinal MRSA colonization (MNin).

To determine whether extracellular S. aureus was attached to neutrophils, isolated MNiv and MNin were incubated with S. aureus-specific antibody (Abcam, Cambridge, UK), followed by secondary antibodies conjugated with Cy5 (BosterBio). Cells were then harvested after incubation and analyzed using flow cytometry.

Assessment of intra-neutrophil MRSA viability

From the above isolation experiment, 10,000 MRSA-containing neutrophils (MNin and MNiv) and non-MRSA-containing neutrophils were lysed with 100 µl of autoclaved water, and then inoculated on tryptic soy agar (BD) supplemented with 5% sheep blood (BD) and 10 µg/ml chloramphenicol (Sigma-Aldrich). Bacterial colonies were quantified using ImageJ analysis (National Institute of Health and LOCI, Madison, Wisconsin, USA).

Prevention efficacy of membrane- and non-membrane-permeable antibiotics against PJI

We assessed whether MRSA carried in neutrophils can lead to PJI in the presence or absence of two antibiotics: gentamicin (a non-membrane-permeable antibiotic) and linezolid (a membrane-permeable antibiotic). Previously operated rats (see above, Sterile Surgical Procedures) were injected with MNin, and then two hours later were injected intravenously with either gentamicin (8 mg/kg) or linezolid (10 mg/kg). The strain of MRSA we used is normally susceptible to both gentamicin and linezolid. After ten days, the operated knee joints of the rats were cultured for MRSA, as before.

Statistical analysis

Differences in the proportion of MRSA-positive cultures between the two surgical groups were evaluated statistically using Fisher’s exact test. For comparing colony counts between the different groups, we used the non-parametric Mann-Whitney U test. SPSS v24.0 (SPSS, Armonk, New York, USA) was used for all statistical analyses. Statistical significance was established at p < 0.05.

Establishment of intestinal MRSA colonization

Results from culturing of the shed stool were obtained for days 1, 3, 5, 7, and 9 after intestinal colonization of MRSA. MRSA was detected in all (n = 40) stools on days 1, 3, 5, and 7 after receiving antibiotics, being food-deprived, and undergoing gavage of MRSA. For day 9, the stools of 12 out of 40 rats (30.0%) were MRSA-positive (Figure 1b).Once ten days had passed after implant surgery, the rats were euthanized and samples from their intestines were cultured for the presence of MRSA. The intestines of all 40 rats were cultured after final sacrifice and all were positive for MRSA, suggesting that MRSA colonization of the intestines may be persistent, although we at times could not detect the bacteria with stool cultures (Figure 1b).

Intestinal MRSA colonization causes PJI

To determine whether intestinal MRSA carriage can cause PJI, a total of 40 rats were divided into two groups, with one group receiving surgery at eight hours and the other at 72 hours after intestinal MRSA colonization. Rats in both groups were sacrificed ten days after implant insertion. We cultured the operated joints of rats that underwent previous intestinal MRSA colonization for determining the presence of PJI. As a control for inadvertent surgical site MRSA contamination, we cultured tissue samples prior to wound closure. The intraoperative cultures from all 40 rats were negative, indicating the absence of intraoperative contamination. By contrast, cultures of the operated joints (culturing bone, joint capsule, and implant) revealed that five out of 20 rats (25.0%) in the eight-hour group developed PJI and two out of 20 rats (10.0%) in the 72-hour group developed PJI (Figure 2b). We then repeated the eight-hour group experiments with two MRSA isolates from PJI patients. The intraoperative cultures from all 40 rats were negative and PJI was detected in six and seven out of 20 rats (30.0% and 35.0%) respectively for the two isolates (Figure 2b). These results suggest that colonized MRSA originating from the gut can cause infections at distal surgical sites, in our experiment, at the operated knee joint with an intra-articular implant.

Wound-site MRSA exposure is not the cause of PJI after intestinal MRSA colonization

To exclude the possibility that postoperative wound site contamination with MRSA contributed to PJI in our paradigm, we assessed operated joints of 20 rats lacking prior experimental intestinal MRSA colonization. In this control, MRSA solution (1 ml/day of 1 × 10⁸ CFUs/ml MRSA) was applied daily on the rats around the skin incision for ten days. Post-mortem cultures of the operated joints revealed that none of these rats developed PJI after ten days of wound site MRSA exposure.

Extracellular MRSA bacteraemia is not the cause of PJI after intestinal MRSA colonization

We then sought to exclude the possibility that PJI caused by intestine-derived MRSA actually occurs via extracellular bacteraemia. To address this possibility, we injected 20 rats with MRSA (1 × 10⁸ CFUs) intravenously on three consecutive days after knee joint surgery. None of the 20 rats developed PJI, indicating that extracellular bacteraemia does not cause PJI in this experimental paradigm.

MRSA-carrying neutrophils isolated after intestinal MRSA colonization harbours viable MRSA

To determine whether intestinal MRSA reaches knee joints via a Trojan Horse mechanism, we analyzed the neutrophils of rats that had undergone intestinal MRSA colonization. Whole blood samples were collected from rats after 1, 3, 5, and 7 days of GFP-MRSA colonization, and neutrophils were isolated. Flow cytometry revealed that 15.3%, 7.5%, 5.6%, and 5.1% of the circulating neutrophils were GFP-MRSA positive at day 1, 3, 5, and 7, respectively (Figure 4a).

Fig. 4

a) Bar graph showing the mean percentage of methicillin-resistant Staphylococcus aureus (MRSA)-harbouring neutrophils in rats 1, 3, 5, and 7 days after intestinal colonization with MRSA and eight hours after intravenous injection of MRSA. b) Bar graph showing the mean number of colony-forming units (CFUs) of neutrophil samples 1, 3, 5, and 7 days after intestinal colonization with MRSA and eight hours after intravenous injection of MRSA. Error bars: SD.

Next, we cultured cellular lysates derived from 10,000 of these MNin and non-MRSA-containing neutrophils. We also cultured samples of blood plasma from the rats to exclude the possibility of extracellular bacteraemia following intestinal MRSA colonization. As expected, no MRSA-positive colonies formed on either the cultures of plasma or lysates from non-MRSA-containing neutrophils. By contrast, MRSA-positive colonies formed on the cultures of lysates from MNin. The colony counts per thousand MNin were stable throughout the seven-day study period (Figure 4b).

Intravenous injection of MNin causes PJI

To further confirm that PJI develops from MRSA carried within neutrophils, we intravenously injected naïve rats with MNin (1 × 10⁶ cells). The naïve rats received injections for three consecutive days after knee surgery. Samples from the rats’ joints were cultured ten days after surgery as before, and we observed that 11 out of 20 rats (55.0%) developed PJI. These results suggest that intestinal MRSA carried by neutrophils could cause PJI in our rat model.

MRSA-carrying neutrophils isolated after intravenous injection of MRSA harbour fewer viable MRSA

As mentioned above, direct intravenous injection of MRSA leads to no PJI development. We next explored the potential mechanism. After intravenous injection of MRSA (1 × 10⁸ CFUs), GFP-tagged MRSA was detected by flow cytometry in 55.2% of circulating neutrophils eight hours later (Figure 4a). This result indicated that circulating neutrophils would also harbour MRSA after intravenous injection of MRSA.

We then assessed whether these MNiv could also cause PJI in naïve rats. To test this idea, we repeated the above experiment but this time injected naïve rats with MNiv instead of MNin. None of the rats developed PJI.

To understand the different infectivity of MNin and MNiv, we next assessed the viability of MRSA in these rats by culturing cell lysates derived from MNiv. The number of resulting MRSA-positive colonies was less than 1% of MNin (Figure 4b). The mean colony number was 2.2 (SD 2.5), and the mean colony number in each of the four MNin groups (days 1, 3, 5, and 7) was 250.6 (SD 113.6), 234.7 (SD 98.3), 245.7 (SD 95.4), and 267.3 (SD 123.5), respectively. We recollected MRSA derived from MNin and MNiv. We then established intestinal colonization with either one of these two MRSA isolates (MNin-derived or MNiv-derived). After surgery, PJI was detected in four out of 20 rats for both groups. Direct intravenous injection of either one of these two MRSA isolates would not cause PJI in our rat model. These results together suggested that injection of MNiv did not lead to PJI mainly because MNiv carried fewer viable MRSA.

Prevention efficacy of membrane-permeable and non-membrane-permeable antibiotics

To further confirm that the gut-to-implant metastasis of MRSA takes place via a Trojan Horse mechanism, we assessed the ability of MNin injection to cause PJI in the presence or absence of two antibiotics: gentamicin (a non-membrane-permeable antibiotic) and linezolid (a membrane-permeable antibiotic). Membrane-permeable antibiotics are able to kill intracellular bacteria in contrast to non-membrane-permeable antibiotics. If the Trojan Horse mechanism operates in our paradigm, we would expect a lower incidence of PJI among linezolid-treated rats.

First, rats were injected with MNin, and then two hours later, injected intravenously with either gentamicin (8 mg/kg) or linezolid (10 mg/kg). The strain of MRSA we used is susceptible to both gentamicin and linezolid. Of the rats that received gentamicin, ten out of 20 (50.0%) developed PJI. By contrast, of the rats that received linezolid, only one out of 20 (5.0%) developed PJI. Furthermore, we treated MNin with control, gentamicin (16 μg/ml), or linezolid (20 μg/ml) for four hours before intravenous injection. After intravenous injection of these pretreated MNin, no rat (zero out of 20) developed PJI in the linezolid group in contrast to seven out of 20 in the control group and six out of 20 in the gentamicin group.

These findings show that the membrane-permeable antibiotic linezolid is more effective in preventing PJI in this model than the non-membrane-permeable antibiotic gentamicin. The higher incidence of PJI among gentamicin-treated rats is consistent with the Trojan Horse hypothesis in that MRSA sequestered in neutrophils was shielded from this non-membrane-permeable antibiotic and thus able to eventually reach and infect the knee joints. The incidence of PJI among control and gentamicin- and linezolid-treated rats is summarized in Table I.

All bacteria grown on culture in this study were identified as the identical strain of inoculated MRSA without colonization of any other bacteria, indicating that there was no unexpected microbial contamination. No extracellular MRSA was detected from MNiv and MNin (Supplementary figure b). Different treatments and their PJI incidence are summarized in Table II. Key findings of this study are summarized in Figure 5.

Fig. 5

Summary of key findings. a) Preoperative intestinal methicillin-resistant Staphylococcus aureus (MRSA) colonization and implant insertion after eight or 72 hours. b) Implant insertion and postoperative intravenous injection of MRSA. c) Implant insertion and postoperative intravenous injection of MRSA-carrying neutrophils isolated from other rats that had undergone intestinal MRSA colonization (MNin). d) Implant insertion and postoperative intravenous injection of MRSA-carrying neutrophils isolated from other rats injected intravenously with MRSA (MNiv). PJI, prosthetic joint infection.