Microbiological and Molecular Features Associated with Persistent and Relapsing Staphylococcus aureus Prosthetic Joint Infection

Background: Persistent and relapsing prosthetic joint infection (PJI) due to Staphylococcus aureus presents a clinical challenge. This study aimed to provide an extensive description of phenotypic and genomic changes that could be related to persistence or relapse. Methods: Initial and second S. aureus isolates from 6 cases of persistent and relapsing PJI, along with clinical isolates from 8 cases, with favorable outcome were included. All isolates were studied by phenotypic and genotypic approaches. Results: Recurrent S. aureus isolates exhibited a significant increase in adhesive capacity, invasion and persistence compared to resolved isolates. No association was found for the presence or absence of certain genes with the persistence or relapse of PJI. All sequential isolates showed identical sequence type (ST). Resistance gene loss during the infection and a great diversity of variants in different virulence genes between the pair of strains, mainly in genes encoding adhesins such as fnbA, were observed. Conclusions: S. aureus-caused relapse and persistence PJI is associated with bacterial phenotypical and genotypical adaptation. The main paths of adaptation were persistence in the intracellular compartment, and the loss of antibiotic resistance genes and variant acquisition, especially in genes encoding adhesins.


Introduction
Prosthetic joint infections (PJIs) are described as difficult to treat [1]. Although the incidence of PJI is not very high, it is still a devastating complication for patients, often requiring surgical intervention and long-term antibiotics, leading to impaired joint function [2]. Debridement, antibiotics and implant retention (DAIR) may be attempted in selected acute cases to minimize the morbidity associated with prosthesis removal and loss of bone stock [1,3]. Unfortunately, however, a large percentage of patients with an acute episode of Staphylococcus aureus PJI treated with DAIR eventually develop recurrent or persistent infection [4].
S. aureus has been identified as one of the main pathogens responsible for PJI [5]. It is a versatile microorganism with a wide genetic background and virulence factors that play an important role in the pathogenesis of infection and its evolution to persistent or recurrent forms [6,7]. The diversity of virulence factors allows S. aureus to adhere to and invade tissue, form biofilm and evade the immune system [7]. The causes of persistent infection are not well established, but possibilities include a limited antibiotic effect, suboptimal surgical management, host immune response or underlying diseases predisposed to recurrence. However, these could also be related to the intrinsic characteristics of the microorganism [8]. In this respect, previous studies have linked certain intrinsic characteristics of S. aureus to the persistence or chronicity of the infection, which include biofilm formation in bacterial communities [9,10], persistence in osteoblasts [11], increased adhesion capacity [12], accessory gene regulator (agr) dysfunction [13] and the formation of small colony variants (SCVs) [14].
Previous studies have focused on analyzing the clinical characteristics of the Staphylococcal-PJI evolution [4,15], but few have delved into the study of the microbiological characteristics and molecular factors that influence PJI development [16][17][18]. We hypothesized that some clinical isolates of S. aureus exhibit certain phenotypic and genotypic characteristics that eventually favor persistent or relapsing PJIs. We also investigated potential adaptive mechanisms in recurrent S. aureus isolates that allow certain infections to become chronic despite optimal antimicrobial and surgical management.

Setting and Patients
Our group recently conducted a prospective, multicenter, observational study of 88 cases of PJI caused by S. aureus [18]. Most patients were treated with DAIR (debridement < 21 days after symptom onset) and received antibiotic treatment according to the antimicrobial susceptibility profile. Despite this, in six patients, infection persisted or relapsed with the original S. aureus strain. We included the initial (recurrent) isolates of the patients with recurrent staphylococcal infection and a second S. aureus isolate recovered during antimicrobial treatment in three cases (persistent isolates), and after antimicrobial treatment in another three cases (relapsed isolates). A random selection of clinical isolates from eight patients treated with DAIR who had a favorable clinical course with no evidence of recurrent S. aureus infection (resolved isolates) were also included ( Table 1).

Characterization of Hemolytic Activity
Alpha-hemolysin (Hla) activity was analyzed and quantified by a rabbit erythrocyte lysis assay in 96-well plates and by visual inspection and assessment of zones cleared of round bacterial colonies [19]. The activity of the agr operon in gamma hemolysin production was measured and categorized as negative, weak, or strong [20]. Beta hemolysin production was also determined [21].

Biofilm Formation
Biofilm formation was assessed in triplicate by the 0.7% crystal violet method on microtiter plates using 33% glacial acetic acid as the decolorizing solution [18]. Absorbance was measured at 595 nm. The results were interpreted in accordance with Stepanovic [22].

Fibronectin Adhesion Assay
To measure the adhesion of S. aureus to fibronectin, 96-well tissue plates were prepared as described by Seidl et al. [23]. Bound staphylococci were detected by crystal violet staining at 620 nm using an enzyme-linked immunosorbent assay (ELISA) plate reader.

Cytotoxicity Assay on MG63 Osteoblasts
As described previously, the cytotoxicity assay was performed on an MG 63 (CRL-1427) human osteoblast cell line [11] using the Cytotoxicity Detection Kit Plus LDH (Roche Applied Science, Indianapolis, IN, USA). The percentage of cell lysis was calculated relative to the value for the maximum LDH release control (cells lysed with a 1% solution of Triton X-100), following the manufacturer's instructions.

Adhesion, Invasion and Persistence Intracellular Assay
The MG 63 human osteoblast cell line (CRL-1427) (LGC standard, Manassas, VA, USA) was cultured and maintained in Dulbecco's modified eagle medium (DMEM) supplemented with 10% fetal calf serum and 100 µg/mL gentamicin [24]. Cells were infected with stationary phase bacterial suspensions at a multiplicity of infection of 100, equivalent to 10 7 CFU/mL. Inoculum CFU levels were confirmed during each experiment by serial dilution and plating. The first 24-well cell plate was incubated at 37 • C in 5% CO 2 for 2 h (adhesion). In the next two 24-well culture plates, the medium was replaced with DMEM + 10 µg/mL lysostaphin to lyse all extracellular or adherent staphylococci, and the plates were returned to the incubator for a further 3 h (invasion) and 48 h (persistence). At each of these time points, cells were lysed in 1 mL of sterile water for 10 min; serial dilutions were made; and suspensions were plated to calculate the bacterial CFU. Each isolate was tested in triplicate.
Sequence files were deposited at GenBank under BioProject PRJNA774351 and accession numbers JAJHKM000000000-JAJHNN000000000.

Statistical Analysis
All statistical analysis was performed using GraphPad Prism 8.2.1 software (San Diego, CA, USA). Two-way analysis of variance (ANOVA) was used to compare differences between multiple groups, and the unpaired Mann-Whitney test to compare differences between two samples: p values < 0.05 were considered to be significant. The association between the presence or absence of genes and clinical variables (panmicrobial GWAS) was analyzed with Scoary software (version 1.6.16) [38]. The association between the presence or absence of genes and the clinical evolution of persistent and relapsing PJI was studied by adjusting the p value using the Benjamini-Hochberg method (FDR, False Discovery Rate) according to the following criteria: p < 0.10 and FDR < 0.05, p < 0.05 and FDR < 0.10 and, finally, p < 0.05 and FDR < 0.05.

Genotypic and Phenotypic Comparisons of Isolates from Resolved and Recurrent PJIs
Isolates leading to recurrent PJI showed increased levels of resistance to all antimicrobials tested (Table 1). One isolate (FailPer118) was methicillin-resistant and harbored SCCmec IVc. Three recurrent isolates (FailRel107, FailRel409 and FailPer118) showed resistance to fluoroquinolones, while all isolates in the resolved group were susceptible. FailRel409 isolates were resistant to ciprofloxacin but susceptible to levofloxacin and harbored the norA gene, a mutation in the grlA gene but not the gyrA gene. FailRel107 isolates were resistant to clindamycin but susceptible to erythromycin. Overall, the genotypic and phenotypic antimicrobial resistance correlation was high, with the exceptions of resistance of fosfomycin and cotrimoxazole antibiotics ( Table 2).
a In relapsing or persistent cases, it has only been included the first strain for this analysis.
We analyzed the intracellular activity of the clinical isolates and found that recurrent S. aureus isolates exhibited a significant increase in capacity for adhesion (12.0%, (IQR: 7.8-13.4%) vs. 2.1%, (IQR: 0.9-3.6%), p = 0.01), invasion (4.3%, (IQR: 1.9-7.8%) vs. 0.3%, (IQR:0.03-0.5%), p = 0.001) and persistence (0.2%, (IQR:0.1-0.9%)) vs. 0.01% (IQR: 0.004-0.002%), p = 0.01) compared to those of the resolved PJIs ( Figure 1). Hemolytic activity was similar in both groups of isolates (Table 1). On sheep blood agar, all isolates displayed a hemolysis pattern characteristic of Hla-induced beta-hemolysis although by rabbit erythrocyte lysis assay, three resolved isolates and one recurrent isolate did not hemolyze the erythrocytes (Supplementary Figure S1A). β-hemolysin with a phage insertion harboring immune evasion genes was detected in all strains except FailRel107, and δ-hemolysin activity was observed in five recurrent isolates (83.3%) and six isolates (75%) recovered from cured patients. Similarly, under conditions of stasis, all S. aureus isolates formed biofilm, but it was weak in most and no significant differences between the two groups of isolates was observed (Table 1 Figure 2 shows the phylogenetic tree of all isolates causing persistent, relapsing, and resolved PJIs together with the main resistance and virulence genes. The isolates were grouped according to CC and MLST similarity, forming two clearly differentiated clusters. Supplementary Table 2 shows the virulence genes of all strains causing PJI according to clinical evolution. All isolates harbored genes encoding microbial surface components recognizing adhesive matrix molecules (MSCRAMMs) and genes involved in biofilm formation. The genes fnbpA and fnbpB were identified in 100 and 85.7% of isolates, respectively, while cna was present in only 64.3%. In microbial pan-GWAS, no association was found between the presence or absence of genes and persistent or/relapsing PJI versus resolved PJI. Similarly, no significant variants were identified between strains causing persistent or relapsing PJI and those leading to a favorable outcome.   Figure 2 shows the phylogenetic tree of all isolates causing persistent, relapsing, and resolved PJIs together with the main resistance and virulence genes. The isolates were grouped according to CC and MLST similarity, forming two clearly differentiated clusters. Supplementary Table S2 shows the virulence genes of all strains causing PJI according to clinical evolution. All isolates harbored genes encoding microbial surface components recognizing adhesive matrix molecules (MSCRAMMs) and genes involved in biofilm formation. The genes fnbpA and fnbpB were identified in 100 and 85.7% of isolates, respectively, while cna was present in only 64.3%. In microbial pan-GWAS, no association was found between the presence or absence of genes and persistent or/relapsing PJI versus resolved PJI. Similarly, no significant variants were identified between strains causing persistent or relapsing PJI and those leading to a favorable outcome.

Phenotypic and Genotypic Changes during Persistent and Relapsing S. aureus PJIs
Sequential isolates from six recurrent PJIs were available for phenotypic and genotypic studies. In the relapsing cases, the second isolate was obtained 20, 55 and 217 days after the end of antimicrobial treatment. All sequential isolates showed an identical sequence type (ST) (Supplementary Table S1). All pairs of isolates showed identical antimicrobial resistance patterns except for persistent case 401, which harbored the aminoglycoside resistance gene aac(6)-aph(2 ), implying that FailPer401.1 was resistant to gentamicin, while FailPer401.2 was susceptible. Furthermore, unlike FailPer402.2, FailPer402.1 exhibited the H31N mutation in the dfrB gene, which confers resistance to trimethoprim although both strains were susceptible to cotrimoxazole. Finally, unlike FailRel104.2, FailRel104.1 harbored the blaZ gene, but we were unable to obtain a definitive resistance phenotype for FailRel104.2 because of the SCV phenotype ( Figure 2). ognizing adhesive matrix molecules (MSCRAMMs) and genes involved in biofilm formation. The genes fnbpA and fnbpB were identified in 100 and 85.7% of isolates, respectively, while cna was present in only 64.3%. In microbial pan-GWAS, no association was found between the presence or absence of genes and persistent or/relapsing PJI versus resolved PJI. Similarly, no significant variants were identified between strains causing persistent or relapsing PJI and those leading to a favorable outcome.

Phenotypic and Genotypic Changes during Persistent and Relapsing S. aureus PJIs
Sequential isolates from six recurrent PJIs were available for phenotypic and genotypic studies. In the relapsing cases, the second isolate was obtained 20, 55 and 217 days after the end of antimicrobial treatment. All sequential isolates showed an identical sequence type (ST) (Supplementary Table S1). All pairs of isolates showed identical antimi- Analyses of cell adhesion, invasion, and persistence in the sequential isolates revealed no significant differences between initial and second isolates obtained from the same patient (Supplementary Figure S2). However, we observed certain changes in five pairs of isolates ( Figure 3). In four pairs (FailRel104, FailRel107, FailPer401 and FailPer402), the initial isolate showed a greater capacity for adhesion and invasion than the second one (p < 0.05), while in just one relapse case (FailRel409), the second isolate showed a greater capacity for adhesion and internalization (adhesion: 1.1 ± 0.8% vs. 6.3 ± 2.8%; p < 0.0001 and invasion 1.5 ± 0.2% vs. 5.2 ± 2.7%; p = 0.001).
Antibiotics 2022, 11, x FOR PEER REVIEW crobial resistance patterns except for persistent case 401, which harbored side resistance gene aac(6´)-aph(2´´), implying that FailPer401.1 was resista while FailPer401.2 was susceptible. Furthermore, unlike FailPer402.2, Fa ited the H31N mutation in the dfrB gene, which confers resistance to t hough both strains were susceptible to cotrimoxazole. Finally, unlike F Rel104.1 harbored the blaZ gene, but we were unable to obtain a definitiv notype for FailRel104.2 because of the SCV phenotype ( Figure 2).
Analyses of cell adhesion, invasion, and persistence in the seque vealed no significant differences between initial and second isolates ob same patient (Supplementary Figure S2). However, we observed certain pairs of isolates ( Figure 3). In four pairs (FailRel104, FailRel107, FailPer40 the initial isolate showed a greater capacity for adhesion and invasion tha (p < 0.05), while in just one relapse case (FailRel409), the second isolate s capacity for adhesion and internalization (adhesion: 1.1 ± 0.8% vs. 6.3 ± and invasion 1.5 ± 0.2% vs. 5.2 ± 2.7%; p = 0.001). Figure 3. Comparison of the adhesion, invasion and persistence capacities betw ond isolates of persistent-and relapsing-PJI cases recovered from the same patien means ± standard deviation of experiments conducted in triplicate. The statist performed by using two-way ANOVA (* p < 0.05; ** p < 0.01; **** p < 0.0001). ns: n p ≥ 0.05).
In the same way, all sequential pairs of isolates showed the same and hemolytic pattern, except for the FailRel104.2 relapse isolate, which hla activity and conversion to the SCV phenotype. Surprisingly, we did Adhesion Invasion Persistence Figure 3. Comparison of the adhesion, invasion and persistence capacities between initial and second isolates of persistent-and relapsing-PJI cases recovered from the same patient. Results represent means ± standard deviation of experiments conducted in triplicate. The statistical analyses were performed by using two-way ANOVA (* p < 0.05; ** p < 0.01; **** p < 0.0001). ns: non-significant (i.e., p ≥ 0.05).
In the same way, all sequential pairs of isolates showed the same agr functionality and hemolytic pattern, except for the FailRel104.2 relapse isolate, which showed a loss of hla activity and conversion to the SCV phenotype. Surprisingly, we did not observe an increased capacity for biofilm formation in the second isolates (Supplementary Figure S3), nor did we observe any difference in cytotoxic effect or fibronectin-binding capacity ( Supplementary  Figures S4 and S5).
With respect to the presence or absence of virulence genes, we found no differences between the initial and second isolates but observed a wide diversity of variants in different virulence genes between strain pairs in relapsing and persistent PJIs (Tables 3, S1 and S2). In the FailRel104 and FailRel409 cases, the same non-synonymous SNP was found in the fnbA gene in the initial and relapsing strain: Q819H (reference MSSA476 (BX571857.1)). Nevertheless, no differential variants were found between all first and second isolates in the persistent and relapsing cases.

Discussion
Our study provides an extensive description of the adaptation of S. aureus strains during the clinical evolution of persistent and relapsing PJIs and a comprehensive analysis of possible phenotypic and genomics variations of all strains isolated in the same PJI. The availability of sequential isolates offers the opportunity to gain insight into adaptation during S. aureus infection. Other authors were limited by a small number of selected cases and so provided only a limited understanding of the general pattern of S. aureus evolution during a PJI [17].
A large panel of phenotypic experiments was used to investigate possible pathogenic traits associated with the S. aureus isolates that cause recurrent PJI. We found no evidence to support changes in bacterial virulence due to agr dysfunction, increased SCV and biofilm formation, adhesion to fibronectin protein or decreased alpha-hemolysin and LDH production. However, we identified one phenotypic feature associated with recurrent but not resolved PJI isolates: an increased capacity for adhesion, invasion and persistence in osteoblasts, which allows S. aureus to hide in a protected zone of infection and then re-emerge over time. These results could be important from a clinical point of view since they could indicate that there are isolates with greater potential to cause recurrence, either because infection persists despite treatment or because infection recurs after having been considered cured. After a thorough genetic analysis, we discovered a wide variety of clonal groups and virulence factors in S. aureus causing both PJI groups but no genetic molecular signature related to recurrence. The only potential factor that appeared to be associated with PJI recurrence was the higher frequency of quinolone resistance in the recurrent group [39].
Next, we investigated possible phenotypic and genotypic traits of sequential isolates recovered from the same patient at different time points of infection. In one case, the loss of antibiotic resistance genes was observed and antibiotic susceptibility increased. This result contrasted with a recent study that showed increased antibiotic resistance in persistent S. aureus bacteremia [40]. Interestingly, the FailRel107 strain showed an infrequent pattern of antibiotic susceptibility, with phenotypic susceptibility to erythromycin and phenotypic resistance to clindamycin. This could be caused by the lnu gene, which appears to be related to animal clonal lineage CC398 [41]. In addition, this strain did not present the phage carrying the immune system evasion genes, which allows the expression of βhemolysin [21]. Therefore, it showed turbid halos of hemolysis on blood agar. Phenotypic analysis did not reveal any trait associated with the adaptation of S. aureus to persistence or recurrence of infection. In this context, we did not observe any differences between sequential isolates concerning cytotoxicity, biofilm formation or intracellular persistence. This contrasted with a previous study in which recurrent isolates were less cytotoxic, formed more biofilm and persisted longer in the intracellular compartment. However, that study included a small number of strains, which may be problematic for drawing general conclusions [17].
With respect to the role of MSCRAMMs in the pathogenesis of S. aureus in implantassociated infections [42], in all cases except one (FailPer402), significant variants in adhesinencoding genes were produced between the initial and second strains although we did not find the same variants in all or most of the strains causing persistent or relapsing cases. Despite the fact that one study on S. aureus PJI did not find any association between polymorphisms in fibronectin-binding protein (FnBP) and PJI [43], other studies suggested that antigenic variation in FnBPA could contribute to S. aureus evasion of host immune response [44,45]. In connection with this, FailRel104 and FailRel409 showed variants in different MSCRAMMS (bbp, sdrD, clfA, fnbA and fnbB) and we found the same nonsynonymous SNP in the fnbA gene between the initial and second strain in both isolates (Q819N). In the intracellular assays, however, the two isolates behaved differently: the initial FailRel104 and the second FailRel409 showed greater adhesion and invasion than the corresponding paired isolate in the sequence.
A non-synonymous SNP was found in the fmt gene of the FailRel104 strain, which has been associated with decreased production of extracellular virulence factors and decreased pathogenesis, which may contribute to increased bacterial survival [46] and persistence. Similarly, a deletion in DNA-3-methyladenine glycosylase, which catalyzes the removal of chemical DNA base lesions, was also observed in FailPer118 cases [47]. This variant could play a role in bacterial adaptation to stressful environments. It is likely that there are reservoirs of latent microorganisms in recurrent infections that are not eliminated by antibiotic treatment or by the host immune system. Under certain conditions, these microorganisms could reactivate and cause recurrence of infection. However, despite the diversity of variants found, we did not observe any in the agr operon, the main global regulator linked to the ability to downregulate pro-inflammatory virulence factors and increase the expression of factors promoting persistence.
The involvement of S. aureus SCVs in chronic prosthetic joint infections has been reported [48]. In the course of chronic osteomyelitis, after host cell invasion, S. aureus strains have been shown to develop adaptive mechanisms that decrease cytotoxicity and increase the percentage of SCV formation that persist in osteoblasts [16]. In relapsing case 104 (the initial strain was wild phenotype; the second one an SCV), mutations in exotoxins were observed. It has been reported that the SCV phenotype in S. aureus is favored by disrupting the electron transport chain with the consequent reduction in ATP production, which inhibits agr growth and functionality. Under these conditions, there is a reduction in hemolysin production and an increase in FnBP expression, providing SCV mutants with a high capacity for invasion into new host cells [44].
Our study did have some limitations. First, when analyzing the differences between strains, many variants were found in intergenic regions, hypothetical proteins and synonymous SNPs, so we do not know their possible involvement in these infections. Second, many variants were identified between the strains of cases 104 and 409, probably because the reference strain used for analysis (MSSA476) did not belong to the same CC. Third, we did not carry out a transcriptomic analysis of the genome, which may provide more accurate information about the molecular factors of S. aureus involved in the persistence and recurrence of PJI.
In our study, the phenotypic assays showed that the recurrent isolates had greater capacity for adhesion, invasion, and persistence in osteoblasts than those that caused resolved PJI. Conversely, the behavior between pairs of PJI isolates was very heterogeneous, and we were therefore unable to identify any phenotypic pattern associated with persistence or recurrence.

Conclusions
Our results showed that S. aureus is capable of modifying its genotype during the clinical course of PJI by adapting to recur or persist, especially by the loss of antibiotic resistance genes and the acquisition of variants, mainly in genes encoding adhesins. However, there was no convergence at the genetic level, indicating that adaptation pathways may be episode specific. This, coupled with other key aspects of the pathogenic complexity of PJI, such as the impact of host immunity on the evolution of S. aureus, could lead to the progression of infection to a cure or chronicity.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10.339 0/antibiotics11081119/s1, Figure S1: Cytotoxicity activity, alpha-haemolysin, biofilm formation and fibronectin binding capacity by S. aureus strains included in our study, Figure S2: Comparisons in the adhesion, invasion and persistence capacities between initial and second S. aureus isolates. Figure S3: Ability to form biofilm of initial-and second S. aureus isolates retrieved from the same patient at the initial time of diagnosis and at the time of persistence or relapsed. Figure S4: Cytotoxic effect after infection between the different pairs of S. aureus grouped into relapse-or persistent-in human osteoblasts. Figure S5: Abilities of S. aureus strains to bind to fibronectin. Table S1: Molecular epidemiology of S. aureus strains causing PJI and reference strains used for each strain and each bioinformatic analysis. Table S2: Virulence genes of all the strains causing the 14 prosthetic joint infections included in this study. Funding: This work was supported by the Plan Nacional I + D + I (PI18/01623 and PI19/00571), Instituto de Salud Carlos III, Ministerio de Economía y Competitividad. MA.M was supported by a "Río Hortega" research contract (CM20/00146, Instituto de Salud Carlos III).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available in supplementary material.