Staphylococcus Aureus Osteomyelitis as an Inducer of Tolerance to Escherichia Coli Pyelonephritis: an Experimental Study

The high incidence of osteomyelitis in vulnerable populations like those with multiple injuries or elderly undergoing joint arthroplasties generates the question what may be their responses to subsequent infection by high virulent isolates. Rabbits were subject to two operations at three week intervals; sham osteomyelitis and sham pyelonephritis (group S); sham osteomyelitis and Escherichia coli pyelonephritis (group P); and Staphylococcus aureus osteomyelitis and E. coli pyelonephritis (group OP). Survival was recorded; cytokine stimulation of circulating mononuclear cells (PBMCs) and tissue myeloperoxidase (MPO) activity and bacterial growth were monitored. In some experiments, dalbavancin treatment was given before pyelonephritis. Healthy PBMCs were pre-treated with bone homogenate, S. aureus or both. Mortality of groups S, P and OP after induction of pyelonephritis was 0%, 50% and 8.3% respectively. Tumour necrosis factor-alpha (TNFα) production by PBMCs was significantly lower in the OP group at 48 hours. E. coli bacterial load was similar in groups P and OP at death or sacrifice whereas the MPO activity of group OP was decreased. Production of TNFα was further decreased among dalbavancin treated rabbits; in these rabbits tissue MPO was increased. TNFα production decreased when healthy PBMCs pre-treated with bone homogenate, S. aureus (HKSA) or both were stimulated with E. coli (HKEC); production was further decreased in the presence of anti-TLR4 and anti-TLR9. It is concluded that staphylococcal osteomyelitis modulated the innate immune responses of the host leading to protection from death by highly virulent E. coli. Tolerance to TLR ligands is the most likely mechanism of action.

Osteomyelitis is an uncommon infection that mandates prolonged antimicrobial treatment and that often results in patient disability. Osteomyelitis usually runs in chronicity and complicates bone surgery for fractures or joint replacement 1 . As a consequence, the common profiles of patients who develop osteomyelitis are either young people with multiple trauma undergoing bone fracture fixation and who remain post-operatively in an Intensive Care Unit environment 2 or elderly who are operated for hip or knee joint arthroplasty 3 . In these patients the immune system is in constant exposure to the antigen stimuli of Staphylococcus aureus that is the commonest pathogen. Due to their profile these patients are prone to subsequent severe infections like bacteremia, lung infection and urinary tract infection that may put their life in danger 4,5 . Severe infection i.e. sepsis develops as a result of the complex orchestration of host immune responses leading to the release of pro-inflammatory and anti-inflammatory cytokines. Perhaps the most critical step for the activation of the innate immune response is recognition of pathogen-associated molecular patterns (PAMPs) by pattern recognition receptors (PRRs) of blood monocytes and tissue macrophages 6 . In that case, it is questionable how recognition of PAMPs of S. aureus in a patient with chronic osteomyelitis may modulate immune response to a subsequent infection.

Results
All animals survived at good condition the first operation of induction of osteomyelitis by S. aureus whatever allowed them to become subject to a second operation after three weeks. At that time point animals of group S already subject to sham osteomyelitis were subject to sham pyelonephritis; animals of group P already subject to sham osteomyelitis were subject to E. coli pyelonephritis; and animals of group OP already subject to S. aureus osteomyelitis were subject to E. coli pyelonephritis. At the end of the 14-day follow-up from the second operation mortality was 0%, 50% and 8.3% respectively (Fig. 1).
The investigation of the mechanism of the protection of preceding osteomyelitis in the animals of OP group from death compared to the animals of P group comprised: (a) the innate immune response as expressed by the cytokine production of peripheral blood mononuclear cells (PBMCs); and (b) the potential of neutrophils for phagocytosis as expressed by tissue activity of myeloperoxidase (MPO). The main modulation in the production of tumor necrosis factor-alpha (TNFα) by PBMCs following stimulation with lipopolysaccharide (LPS) was found after 24 and 48 hours but not after 72 hours. More precisely, TNFα production by PBMCs of groups S and P steadily increased from baseline (0 hours) remaining, however, lower in group P than in group S. TNFα production by PBMCs of group OP remained stable compared to baseline being significantly lower compared to the other groups ( Fig. 2A). The production of TNFα after stimulation with phytohemagglutin (PHA) was not remarkable in any group (data not shown). A negative correlation was found between the count of S. aureus in cancellous bone of animals of group OP sampled before the induction of pyelonephritis and the production of TNFα by PBMCs isolated at 48 hours post the induction of pyelonephritis (Fig. 2B). Interleukin (IL)-10 in PBMCs supernatants and serum ferritin levels at all time points were below the limit of detection. These findings point towards an attenuation of pro-inflammatory responses by PBMCs as part of the mechanism of protection from death in the OP group mainly expressed after 48 hours. This is compatible with the survival curves of the groups shown in Fig. 1 that start to separate after the first 48 hours.
The mean ± SE log10 cfu/g of S. aureus load in cancellous bone samples of group OP was 3.32 ± 0.52 cfu/g. However, S. aureus did not grow in any other body tissue. The tissue bacterial load of E. coli was similar in groups P and OP (Fig. 3A). However, the tissue MPO activity that is a reflection of the level of infiltration by neutrophils 7 was significantly attenuated in group OP compared to the other groups. As a matter of fact, the MPO activity of the spleen and lung of group P was remarkably increased compared to the MPO activity of the kidney of the same group where the primary infectious site was. This is compatible with a compartmentalization of neutrophils towards the remote organs where E. coli sepsis disseminated in group P; this did not happen with a similar intensity in group OP (Fig. 3B). As shown in Fig. 3C,D, there is negative correlation between E. coli load of the spleen and of the right kidney and of tissue MPO of the animals of the OP group. Although this may be an indicator of more efficient phagocytosis by neutrophils, other explanations may do apply.
The next question was how the production of TNFα and the tissue MPO activity might be modulated in case of rabbits receiving antistaphylococcal treatment. To this end, seven more rabbits were studied to which one single intravenous infusion of dalbavancin was administered at the time of removal of tibial foreign body three weeks from S. aureus challenge. Dalbavancin is a novel long-acting lipopeptide with high activity against S. aureus with resistance to methicillin as it was the studied pathogen. Rabbits were subject to E. coli pyelonephritis seven days after the infusion of dalbavancin (group ODP). In these rabbits, production of TNFα by PBMCs was even lower than the OP group (Fig. 4A). MPO activity in the liver and in the right kidney was greater than the OP group (Fig. 4B). Survival of study groups Rabbits were divided into three groups: S subject to sham osteomyelitis followed three weeks later by sham pyelonephritis; P: subject to sham osteomyelitis followed three weeks later by Escherichia coli pyelonephritis; and OP: subject to Staphylococcus aureus osteomyelitis followed three weeks later by E. coli pyelonephritis. The log-rank tests and the p-values of the indicated comparisons are provided.
In order to better decipher the mechanism of TNFα modulation, the in vivo experiment was simulated in vitro. More precisely, rabbit PBMCs were pre-treated with bone homogenate, heat-killed S. aureus (HKSA) or both and then stimulated with heat-killed E. coli (HKEC). TNFα production was significantly decreased when cells were pre-treated with bone homogenate, HKSA or both (Fig. 5). This production was further decreased in the presence of anti-TLR (Toll-like receptor) 4 and anti-TLR9.

Discussion
The present study showed that preceding S. aureus osteomyelitis offered survival benefit from E. coli pyelonephritis. In the studied model of sepsis, death is mediated through multiple organ dysfunctions as reflected by the intense infiltration of remote organs i.e. liver, spleen and lung by neutrophils where the pathogen was spread. Protection from death of rabbits with osteomyelitis was associated with attenuation of pro-inflammatory responses of E. coli sepsis as reflected by decreased production of TNFα by circulating PBMCs and by decrease of neutrophil inflammation in both the infected kidney and remote organs.
The mechanism behind this protection of the host resembles that of LPS tolerance. LPS tolerance is described as reduced immune responsiveness of immune cells to one LPS challenge subsequent to a previous exposure to E. coli or LPS 8,9 . However, a recent study using mammary epithelial cells from cows showed a similar phenomenon using agonists of TLR2 and TLR6. More precisely, cells were pre-treated with the TLR2/6 agonist Pam2CSK4 and subsequently exposed to LPS; these cells produced lower concentrations of TNFα than not pre-treated cells 10 . This phenomenon could not be reversed with the use of epigenetic inhibitors showing that epigenetic modifications do not mediate tolerance from TLR2/6 agonists. Since structures of S. aureus like muramyl dipeptide and peptidoglycan are mostly recognized via TLR2 interactions, this type of tolerance may be the mechanism underling the decreased TNFα production from the circulating PBMCs of the OP group. Interestingly, in our model, this type of tolerance needed a second in vivo hit to be shown since TNFα produced following ex vivo stimulation of PBMCs before the induction of pyelonephritis was not different between the OP and the P groups. From the clinical perspective, it needs to be outscored that the attenuation of TNFα responses was even more pronounced following antistaphylococcal treatment for one week. Probably in rabbits receiving dalbavancin treatment attenuated responses may be achieved not only through replicating S. aureus but also through dead staphylococcal remnants. This phenomenon of TLR tolerance is fully dependent on the virulence of the pathogen. In a recent study, mice free from specific pathogens (SPF) were compared to mice cohoused with animal from pet stores (CoH). When exposed to low virulence Listeria monocytogenes that interacts with TLR2, CoH showed enhanced immune responses allowing better survival than SPF mice. The opposite took place when these mice were exposed to peritoneal sepsis mediated through cytokine storm 11 . In other terms, tolerance through TLR2 interaction is mostly favored in a setting of chronic well-tolerated infection. This type of infection was osteomyelitis in our setting.
However, the pathogenesis of bone infections is more complex since in parallel to the release of PAMPs, the immune system of the host is exposed to danger-associated molecular patterns (DAMPs) that are cellular constituents released by bone destruction. Examples of DAMPs are histones, adenosine triphosphate, urate and high mobility group box 1. DAMPs mostly interact with TLR4 and TLR9 12,13 . In our in vitro setting, DAMPs are contained in bone homogenates used for the stimulation of PBMCs. The blunt TNFα responses in the presence of anti-TLR4 and anti-TLR9 indicate that the attenuated TNFα production from the rabbits of the OP group is at least partly mediated through blockade of TLR4 and TLR9 on monocytes.
Sepsis following acute pyelonephritis by E. coli may rapidly deteriorate due to the development of sepsis-induced immunosuppression 14 . It may be hypothesized that tolerance developing after staphylococcal osteomyelitis prevented the development of immunosuppression.
The presented study showed for the first time that staphylococcal osteomyelitis modulated the innate immune responses of the host to such an extent that protected from death from a subsequent infection by highly virulent E. coli. The mechanism behind this phenomenon shares the characteristics of tolerance to TLR ligands and it is mediated through attenuated TNFα production and neutrophil activation. This phenomenon might probably be taken into account in light of the safety of several patients receiving outpatient antimicrobial therapy for chronic infections 15 .  www.nature.com/scientificreports www.nature.com/scientificreports/ period. Temperature ranged between 18 °C and 22 °C, relative humidity was between 55% and 65%, and lighting was maintained in a 12 hours cycle. After one week of acclimatization, animals were randomly divided into three different groups, as follows:

Animals and Methods
• Sham surgery (S, n = 8) that was subject to tibia sham injury for 3 weeks followed by sham pyelonephritis • Pyelonephritis group (P, n = 16) that was subject to tibia sham injury for 3 weeks followed by E. coli pyelonephritis • Osteomyelitis and pyelonephritis group (OP, n = 12) that was subject to implant-related staphylococcal osteomyelitis of the tibia for 3 weeks followed by E. coli pyelonephritis.
Three more rabbits were used for blood sampling.  www.nature.com/scientificreports www.nature.com/scientificreports/ One isolate of S. aureus resistant to methicillin (MRSA) previously isolated from a patient with known implant-related osteomyelitis and another E. coli blood isolate from one previous patient with severe pyelonephritis were used for the induction of osteomyelitis and pyelonephritis respectively. Before each experiment, single colonies were suspended in Mueller-Hinton broth (Oxoid Ltd, London, UK) and incubated at 37 °C in a shaking water bath. The resulting inoculum was adjusted to 5 × 10 7 cfu/ml and it was used for animal challenge.
On each day of experimentation at least one animal per group was operated using a randomization chart. Osteomyelitis was surgically induced in the right proximal tibial metaphysis in all rabbits of the OP group according to the guidelines provided by Mader 16 and also described by our group 17,18 . Briefly, animals were sedated by the intramuscular injection of 25 mg/kg ketamine and 5 mg/kg xylazine (Alvet, Athens, Greece). Anaesthesia was maintained by the intramuscular administration of 15 mg/kg xylazine every 30 minutes. Following skin incision below the knee joint, exposing the medial cortex, a hole was made and one 18 G stainless steel needle was inserted to mechanically induce an endosteal trauma, followed by the instillation of 5 × 10 6 colonies of MRSA at a volume of 0.1 ml. One SLS screw (3.2 mm × 10 mm, CliniLab, Athens, Greece) was used to seal the hole and serve as a foreign body. The skin over the insertion point was sutured with 3.0 silk suture and the animals were carried to their cages after recovery. The same procedure was followed for animals of groups S and P where 0.1 ml 0.9% sodium chloride was used for animal challenge instead of the MRSA pathogen. Rabbits then stayed in their cages being under closely monitoring and having access to food and water. Paracetamol suppositories were administered every 12 hours to minimize suffering.
After three weeks, the needle was removed and a small cancellous bone sample was obtained from the entry point for quantitative culture. Then animals were subject to obstructive E. coli pyelonephritis using the procedure already performed by our group 14,[19][20][21] . Briefly, after midline abdominal incision and removal of the intestines, the right ureter was recognized and ligated just below the renal pelvis. One 0.3 ml suspension containing 1 × 10 7 cfu/ml of the E. coli isolate was instilled into the renal pelvis, proximally to the ligation. After the end of the operation and for the first 48 hours rabbits were intramuscularly administered 1 mg/kg of meloxicam (Alvet, Athens, Greece) once daily. Analgesia with paracetamol suppositories 250 mg/kg was given to reduce suffering. Previous experiments of our group have shown that pain was well-managed through the use of paracetamol. This also allowed avoiding opiod sedation that modulates immune responses 22 .
Before the induction of pyelonephritis and at 24, 48 and 72 post pyelonephritis, six ml of heparinized blood was collected after venipuncture of the dorsal auricular artery. Survival was recorded every 12 hours for 14 days. Survivors were sacrificed by the intravenous administration of sodium thiopental (Alvet). After death or sacrifice, specimens of liver, spleen, lower lobe of the right lung and right kidney were aseptically collected into separate sterile containers with 0.9% N/S. Some experiments were repeated with rabbits subject to osteomyelitis as described above (ODP group, n = 7). After three weeks, the needle was removed and rabbits were given one single intravenous infusion of 20 mg/kg of dalbavancin (Angelini, Athens, Greece), as done in previous experiments 23 . One week after the dalbavancin infusion, rabbits of this group were subject to pyelonephritis, as described above. tissue handling. Tissue segments were weighed and mechanically homogenized using a grinder (Hilden, Germany). One aliquot of 0.1 ml was serially diluted 1:10 into Mueller-Hinton broth (Oxoid); 0.1 ml of each dilution was plated onto MacConkey agar (Oxoid). After incubation for 24 hours at 37 °C, the number of viable colonies was counted and expressed as log10 of colony forming units per gram tissue (cfu/g). The lower detection limit was 10 cfu/g. Removed needles were sonicated in a water bath for 3 min (BANDELIN BactoSonic, Berlin, Germany) and suspensions were quantitatively cultured onto Champan's medium (Oxoid). Cancellous bone aliquots were also weighed and homogenized as described above and quantitatively cultured onto Chapman plates.
Activity of MPO was calculated as previously described 14 . Results were adjusted for tissue sample protein content on Bradford assay (Sigma-Aldrich) and they were expressed as MPO units/mg protein/g. Laboratory investigation. Peripheral blood mononuclear cells (PBMCs) were isolated after gradient centrifugation of heparinized whole blood over Ficoll-Hypaque (Biochrom AG, Berlin, Germany), washed and stimulated for cytokine production with purified 10 ng/ml LPS (lipopolysaccharide) of E. coli O55:B5 or 5 μg/ml PHA (InvivoGen, San Siego, USA), as previously described 14 . Concentrations of TNFα and IL-10 were measured in duplicate by an enzyme immunosorbent assay (R&D Inc, Minneapolis, USA). The lower detection limit was 156.5 pg/ml for TNFα; and 31.1 pg/ml for IL-10. Ferritin was measured by an enzyme immunoassay (R&D Inc) in all serum samples as a marker of acute pro-inflammation 24 ; the lower detection limit was 75 ng/ml.
In separate experiments, cancellous bone was sampled from the tibia of three healthy rabbits as described above, weighted, homogenized and dissolved into 0.9% N/S. Before bone sampling, PMBCs were isolated from these rabbits and distributed at a density of 5 × 10 6 cells/ml at the growth conditions mentioned above in duplicate. For these experiments, PBMCs were incubated in the absence or presence of 5 × 10 5 cfu/ml of HKSA and/ or 1 mg/ml of fresh bone homogenate. After 24 hours of incubation at 37 °C at 5% CO 2 , the medium was renewed and cells were left unstimulated or were stimulated with 5 × 10 5 cfu/ml of HKEC. Experiments were repeated with 1 μg/ml of anti-TLR4 (LPS-EK, InvivoGen) or anti-TLR9 (ODN D-SL01, InvivoGen). After 24 hours of incubation, plates were centrifuged and TNFα was measured in supernatants by an enzyme immunosorbent assay. power of the study and statistical analysis. The primary study endpoint was the comparative 14-day mortality between groups P and OP. Based on previous experiments 14, [19][20][21] , the 14-day mortality of group P is anticipated to be 60%. With the assumption that this would have been decreased to 10% in the OP group, 14 rabbits per group would be needed to demonstrate this difference with 80% power at the 10% level of significance.
Survival was compared between groups by the log-rank test. Quantitative variables were expressed as mean ± SE. Comparisons between groups were done by ANOVA. Comparisons between serial time points within the same group were performed by the Wilcoxon ranked-sum test. Correlations between variables were