Abstract
The mechanisms that drive the transition from commensality to invasiveness in Staphylococcus aureus are poorly understood. We recently reported that >50% of S. aureus isolates from uninfected diabetic foot ulcers in French patients harbor a prophage, ROSA-like, that is absent from invasive isolates from diabetic foot infections, including osteomyelitis. Here we show that the ROSA-like insertion abolishes the ability of S. aureus to replicate within osteoblasts, the bone-forming cells, greatly reducing damage to infected cells. These results unravel an important mechanism by which particular S. aureus strains are maintained in a commensal state in diabetic foot ulcers.
Staphylococcus aureus is both a frequent colonizer of humans and a versatile pathogen able to elicit invasive infection. Understanding the mechanisms that drive the transition of S. aureus from commensality to invasiveness is important to predict infection risk in colonized patients and to propose targeted decolonization procedures to high-risk patients. Although the molecular epidemiology of carriage and clinical strains can be markedly different [1], no genomic characteristic able to firmly discriminate commensal and invasive isolates has been detected so far [2]. In particular, the presence of toxin-encoding genes or the constitutive overexpression of toxins in certain S. aureus lineages has been unambiguously associated with an increased virulence, based on epidemiological and experimental data [2], yet these so-called hypervirulent S. aureus lineages are also frequently encountered as harmless colonizers, and they are not obligate pathogens [1].
The transition from colonization to invasion is of particular importance in diabetic foot ulcer (DFU), a frequent and chronic complication of diabetes mellitus. S. aureus often colonizes DFU wounds without triggering infection, but it can also invade the underlying tissues, frequently causing diabetic foot osteomyelitis [3]. In previous studies of DFU S. aureus isolates, we demonstrated that uninfecting and invasive strains exhibited markedly different population structures and that strains of clonal complex 5/8 (CC5/8) were overrepresented in uninfected ulcers [4, 5]. The association of S. aureus CC5/8 colonization with favorable evolution of DFUs was later revealed in a multicenter study [6], suggesting a genetic basis for the lower virulence of these strains. Indeed, genome-wide comparisons of invasive and uninfecting strains within CC5/8 led to the identification of a 44-kb prophage insertion that was highly specific to noninfecting isolates [7]. The prophage, designated “ROSA-like” owing to a high sequence similarity with the previously described phage ROSA [8], was inserted in an intergenic region upstream of the isd locus, which encodes a hemoglobin receptor involved in heme-iron uptake [7]. CC5/8 strains with ROSA-like insertion exhibited a decreased virulence in in vivo killing models of nematodes and zebra fish embryos, compared with ROSA-like–negative CC5/8 strains from infected DFUs. Strikingly, the spontaneous expulsion of ROSA-like, selected under iron-depleted growth conditions, restored virulence in the same models [7].
These observations suggested that ROSA-like insertion influences the global virulence of CC5/8 S. aureus. Because osteomyelitis is a frequent and clinically important complication of diabetic foot infection [3], we examined in the present work whether ROSA-like expulsion altered the interactions between S. aureus and osteoblasts, the bone-forming cells. Indeed, osteoblast invasion is increasingly thought to play a key role in the pathogenesis of osteomyelitis, mainly because bacteria that survive and grow in the intracellular compartment are protected from most antimicrobial agents and immune effectors [9] and because the death of infected osteoblasts contributes to bone loss [10]. We used a well-established model of intracellular bacterial challenge of MG-63 osteoblast-like cells to test whether presence or absence of the ROSA-like phage modified intracellular growth and cytotoxicity.
METHODS
Bacterial Strains
S. aureus DFU isolate NSA1385 and its isogenic ROSA-like–negative (Δrosa) variant have been described previously [7]. NSA1385 was isolated from an uninfected DFU that did not evolve to infection during a 1-month follow-up prospective longitudinal study [5]. The Δrosa variant was obtained after spontaneous ROSA-like expulsion under iron-restricted growth conditions in TMS medium [7]. Complete phage loss in the Δrosa variant was confirmed by polymerase chain reaction. Genome comparisons of NSA1385 and the Δrosa variant by entire DNA sequencing confirmed the isogenic nature of the pair of strains and the absence of other modifications apart from phage expulsion [7].
Cell Invasion Assay
Intracellular bacterial growth and cytotoxicity during osteoblast infection were quantified as described previously [11], using a well-established ex vivo model whose readouts represent good correlates of clinical virulence in S. aureus osteomyelitis. Briefly, washed and sonicated bacterial suspensions were cocultured for 2 hours with MG-63 osteoblast-like cells (LGC Standards, Teddington, United Kingdom) at a multiplicity of infection of 100:1. After this invasion step, extracellular bacteria were eliminated by 1-hour incubation in 200 mg/L gentamicin-containing medium, followed by incubation in 40 mg/L gentamicin until the end of the experiment. At indicated time points (3, 24, or 48 hours after infection), viable intracellular bacteria were released by osmotic lysis of infected cells and quantified by plating suitable lysate dilutions on agar. Cytotoxicity was assessed 24 and 48 hours after infection by quantifying lactate dehydrogenase (LDH) activity in cell culture supernatant on a Dimension Vista automated analyzer (Siemens Healthcare, Tarrytown, NY). At least 3 independent experiments were performed in triplicate wells, and CFU counts were averaged from 2 series of plate countings per well. Intracellular CFUs were expressed as percentages relative to the initial inoculum of the coculture experiment.
Statistical Analysis
Intracellular CFUs and LDH release were log transformed before inclusion in statistical analyses. Two-group comparisons were performed using a 2-tailed Welch t test. Multiple regression models were used to detect independent associations after controlling for potential confounders. In particular, possible intracellular growth of strains NSA1385 and Δrosa was assessed by performing linear regression of intracellular CFUs on time, including experiment identifier as covariate to control for interexperiment variation. Nested regression models were compared using the omnibus F test. Regression validity was controlled by QQ plot inspection. The significance threshold was set at 0.05 for all tests. All computations were performed using R software, version 3.0.1 Good Sport (R Foundation for Statistical Computing, Vienna, Austria).
RESULTS
By comparing the number of viable intracellular bacteria within osteoblasts infected by S. aureus NSA1385 and its Δrosa variant after 3, 24, and 48 hours of incubation, we observed that ROSA-like expulsion correlated with an enhanced ability to grow in the intracellular compartment (Figure 1A). The proportion of the infecting inoculum that survived within osteoblasts was higher at all time points in the Δrosa strain as compared to NSA1385. The difference increased with time, with intracellular Δrosa CFUs being 1.6-, 2.6-, and 5.6-fold higher after 3, 24, and 48 hours, respectively. Interestingly, this pattern was related to an inability of NSA1385 but not the Δrosa variant to grow within osteoblasts, as confirmed by the fact that intracellular Δrosa CFUs increased over time (P < .001, by the F test for simple regression on time) but not NSA1385 CFUs (P > .05). Finally, in a multiple regression model of relative intracellular CFUs that controlled for incubation duration and interexperiment variation, the inclusion of ROSA-like positivity (NSA1385 vs Δrosa) dramatically improved model fit (P < .0001, by the analysis of variance F test), confirming the difference of intracellular fitness associated with ROSA-like insertion.
Loss of ROSA-like phage restores Staphylococcus aureus intracellular growth and cytotoxicity toward osteoblasts in an ex vivo model of intracellular infection. MG-63 osteoblast-like cells were challenged with S. aureus NSA1385, an avirulent strain isolated from uninfected diabetic foot ulcer (open marks), and its variant lacking the phage ROSA-like (Δrosa; gray marks). All axes use log scales. Black horizontal bars denote geometric means with 95% confidence intervals. A, Viable intracellular S. aureus counts relative to the infecting inoculum. B, Lactate dehydrogenase (LDH) released by membrane leakage in infected osteoblasts, used as a proxy measure of host cell damage. C and D, Correlation between intracellular inoculum and LDH release after 24 and 48 hours of infection, respectively. Gray lines are linear regression slopes with confidence bands. Values in upper left corners denote hours after infection. *P < .05 and ***P < .001, by the Welch t test after log transformation (A and B) or Pearson and Spearman tests of correlation estimating linear and possibly nonlinear correlation, respectively (C and D). Abbreviation: NS, nonsignificant.
The intracellular replication of the Δrosa variant was accompanied by an enhanced cytotoxic effect toward osteoblasts (Figure 1B). Indeed, LDH release in culture supernatant, practically measuring cytoplasmic membrane damage, was 1.2-fold higher after 24-hour incubation in Δrosa-infected as compared to NSA1385-infected osteoblasts, and this difference was markedly increased (by 1.6-fold) after 48 hours.
We then examined in more depth the relationships between intracellular growth and cytotoxicity to determine whether the latter was independently influenced by ROSA-like insertion (Figure 1C and 1D). In a pooled analysis of NSA1385- and Δrosa-infected cells, intracellular CFUs were positively associated with LDH release both at 24 hours (Spearman's rho = 0.83, P < .0001) and 48 hours after infection (rho = 0.52, P = .03). Moreover, in a multiple regression model of LDH release, controlled as before for incubation duration and interexperiment variation, most of the explanatory power was brought in by the inclusion of intracellular CFUs (P < .0001), while the additional inclusion of ROSA-like positivity failed to significantly improve model fit (P = .06). These results thus indicated that the observed cytotoxicity in our model was mostly a consequence of intracellular bacterial replication and that ROSA-like insertion had little, if any, independent influence on the cytotoxicity of the infective isolate.
DISCUSSION
By comparing the outcomes of the intracellular infection of MG-63 osteoblast-like cells by isogenic variants of a DFU-colonizing S. aureus strain, we demonstrated that the insertion of phage ROSA-like, previously associated with favorable DFU evolution in an epidemiological study [7], impaired the ability of S. aureus to replicate in the intracellular compartment and favored the survival of infected osteoblasts. Inasmuch as osteoblast invasion and killing and the intracellular replication of S. aureus are important aspects of osteomyelitis pathophysiology [9], our results identify a key role of phage ROSA-like in the avirulence of DFU-colonizing strains with ROSA-like insertion.
Interestingly, a positive correlation of intracellular growth and cytotoxicity was observed in our experiments, while these readouts were negatively correlated in our previous work on the virulence toward osteoblasts of healthcare-associated methicillin-resistant S. aureus (MRSA) and hypervirulent community-acquired MRSA [11]. In hypervirulent strains, a negative association of toxin-mediated cytotoxicity with intracellular survival was understandably related to the fact that more-aggressive strains, by damaging their host cell, are quickly released into the extracellular compartment. Contrary to this situation, the cytotoxic impact of ROSA-like expulsion in strain NSA1385 should rather be considered as a direct consequence of intracellular S. aureus replication, because the association between phage loss and LDH release vanished after controlling for intracellular growth. This, in turn, suggests that the avirulence of ROSA-like–positive strains is mainly a consequence of their inability to replicate after cell invasion. Although we cannot exclude that ROSA-like insertion influences toxin-mediated cytotoxicity (eg, by interfering with virulence regulation systems such as the accessory gene regulator agr), such an influence was not supported by our data.
The mechanisms by which ROSA-like insertion restrains intracellular growth are currently unknown. The previously described phage ROSA, which has closest DNA sequence similarity with the ROSA-like phage in BLAST analysis [7], has unknown function [8], and the coding portions of the ROSA-like phage sequence do not include analogs of known staphylococcal virulence factors. Nevertheless, several arguments led us to speculate that a possible interference of phage ROSA-like with iron uptake modulation might be a promising candidate mechanism. These arguments, from the most causal to the most consequential with respect to iron uptake, include (1) the proximity of the intergenic insertion site of ROSA-like to the iron surface determinant locus isd, whose expression might be altered; (2) our previous observation that spontaneous ROSA-like expulsion restored the ability of NSA1385 to overcome iron deprivation [7]; and (3) the importance of iron uptake for survival in iron-restricted conditions [12], such as those that S. aureus must face after host cell invasion. Further research in this direction could include a comprehensive examination of the impact of ROSA-like insertion on the expression of regulators and effectors of iron uptake.
A question that arises naturally from our observations is whether ROSA-like expulsion occurs in vivo in DFUs colonized with ROSA-like-positive S. aureus, possibly favoring the transition to DFU invasion. Indeed, in vivo phage dynamics in S. aureus have been previously linked with virulence modulation in a mouse model study of phage PhiSa3 [13], whose expulsion restores the expression of the functional form of the cytolytic sphingomyelinase β-toxin. However, the frequency of ROSA-like expulsion in vitro (<1/108) is orders of magnitude lower than that of the PhiSa3 β-toxin–converting phage (approximately 1/102) [7, 14], suggesting that within-patient loss of ROSA-like is a very unlikely event and, thus, that avirulent ROSA-like–positive strains are most likely to remain so. Finally, whether and how the phage ROSA-like is positively selected in CC5/8 DFU–colonizing strains remains an open question.
To conclude, our data indicate that a colonization-associated phage restrains S. aureus transition to an invasive state by abolishing intracellular growth after host cell invasion, a feature that might explain why strains harboring this phage were not isolated from diabetic foot osteomyelitis. This observation further supports our previous proposition that patients with DFUs colonized with ROSA-like–positive strains, which represented >50% of patients with uninfected DFUs in a French cohort [7], would probably not benefit from S. aureus eradication. If confirmed in further studies, this proposition could contribute to spare antistaphylococcal treatments in an appreciable proportion of patients with DFUs, in whom sustainable antimicrobial use is of utmost importance given their high risk for recurrent infections.
Notes
Financial support. This work was supported by the Groupement Inter-Régional à la Recherche Clinique et à l′Innovation (grant GIRCI-D50829 to J. P. R.).
Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
