In Vitro Study of the Synergistic Effect of an Enzyme Cocktail and Antibiotics against Biofilms in a Prosthetic Joint Infection Model

Prosthetic joint infections (PJI) are frequent complications of arthroplasties. Their treatment is made complex by the rapid formation of bacterial biofilms, limiting the effectiveness of antibiotic therapy.

ABSTRACT

Prosthetic joint infections (PJI) are frequent complications of arthroplasties. Their treatment is made complex by the rapid formation of bacterial biofilms, limiting the effectiveness of antibiotic therapy. In this study, we explore the effect of a tri-enzymatic cocktail (TEC) consisting of an endo-1,4-β-d-glucanase, a β-1,6-hexosaminidase, and an RNA/DNA nonspecific endonuclease combined with antibiotics of different classes against biofilms of Staphylococcus aureus, Staphylococcus epidermidis, and Escherichia coli grown on Ti-6Al-4V substrates. Biofilms were grown in Trypticase soy broth (TSB) with 10 g/liter glucose and 20 g/liter NaCl (TGN). Mature biofilms were assigned to a control group or treated with the TEC for 30 min and then either analyzed or reincubated for 24 h in TGN or TGN with antibiotics. The cytotoxicity of the TEC was assayed against MG-63 osteoblasts, primary murine fibroblasts, and J-774 macrophages using the lactate dehydrogenase (LDH) release test. The TEC dispersed 80.3 to 95.2% of the biofilms’ biomass after 30 min. The reincubation of the treated biofilms with antibiotics resulted in a synergistic reduction of the total culturable bacterial count (CFU) compared to that of biofilms treated with antibiotics alone in the three tested species (additional reduction from 2 to more than 3 log₁₀ CFU). No toxicity of the TEC was observed against the tested cell lines after 24 h of incubation. The combination of pretreatment with TEC followed by 24 h of incubation with antibiotics had a synergistic effect against biofilms of S. aureus, S. epidermidis, and E. coli. Further studies should assess the potential of the TEC as an adjuvant therapy in in vivo models of PJI.

INTRODUCTION

Prosthetic joint infections (PJI) are catastrophic complications of joint replacement surgeries. With a prevalence of 2.0% to 2.4% in patients following a total knee or total hip replacement (1), PJIs represent one of the leading causes of failure for arthroplasties (2).

The most prevalent species encountered in PJIs are Staphylococcus aureus, coagulase-negative staphylococci, streptococci, and Gram-negative bacilli (3). These species are known to adhere to the surfaces of implants and to form complex communities embedded in an extracellular matrix called biofilms. In these structures, bacteria become tolerant to antibiotics up to concentrations 1,000 times the MIC of planktonic cells (4–6).

Surgical debridement, implant retention, and antibiotic therapy (DAIR) is the preferred treatment strategy of acute (less than 3 weeks of symptoms) PJI (7), with better functional outcomes (8) and lower costs (9, 10) than alternative strategies. However, the observed failure rate is considerable (11, 12). These failures are partly explained by an insufficient biofilm removal during the debridement surgery, as was shown in vitro by Urish et al. (13) and Poilvache et al. (14).

As a consequence, the need for new adjuvant treatments to facilitate the dispersal of biofilms has been recognized in the literature (15–19). Among mechanical and pharmaceutical approaches, the use of hydrolytic enzymes targeting specific components of the biofilm matrix has been suggested (19). The dispersal potential of single enzymes targeting individual matrix components, such as extracellular DNA (eDNA), exopolysaccharides, or extracellular proteins, has been demonstrated by previous authors in in vitro models (20–23). Moreover, the synergy of the combination of these enzymes with antibiotics (24, 25) or antiseptics (26, 27) against biofilms has been reported in previous studies.

However, these previous studies did not address the variety in matrix composition among bacterial species, as they focused on the effect of single enzymes on selected species. For this reason, a broad-spectrum tri-enzymatic cocktail (hereafter referred to as TEC for tri-enzymatic cocktail), composed of an endo-1,4-β-d-glucanase, a β-1,6-hexosaminidase, and a recombinant RNA/DNA nonspecific endonuclease (see Materials and Methods), was tested, targeting the matrix of Escherichia coli, Enterococcus faecalis, Klebsiella pneumoniae, Pseudomonas aeruginosa, Staphylococcus aureus, and Staphylococcus epidermidis biofilms (patent numbers WO2017/211737 A1 [28] and BE2020/5726 [29]).

In this study, we investigated the dispersal effect of this tri-enzyme cocktail against Staphylococcus aureus, Staphylococcus epidermidis, and Escherichia coli monomicrobial biofilms grown in vitro on titanium alloy surfaces similar to prosthetic implants and examined its synergy in combination with a broad selection of antibiotics belonging to different classes recommended for PJI treatment (7) at concentrations relevant to human use.

RESULTS

Antibiotic susceptibility of planktonic bacteria and of biofilms.

The MICs and minimal biofilm eradication concentrations (MBECs) are shown in Table 1. The strains were susceptible to all tested antibiotics according to the CLSI breakpoints, except for the ATCC 33591 methicillin-resistant S. aureus (MRSA) strain, which was resistant to doxycycline. For this reason, doxycycline was not used against ATCC 33591 biofilms for the next experiments. The MICs in cation-adjusted Mueller-Hinton broth (MHB-ca) and Trypticase soy broth (TSB) with 10 g/liter glucose and 20 g/liter NaCl (TGN) were within two dilutions of each other. Most of the MBECs exceeded the MICs by a hundred- or thousand-fold factor, to the notable exception of daptomycin, linezolid, and moxifloxacin against the strain ATCC 33591 in MHB-ca. The MBECs could not be established in TGN for the strain ATCC 47076, as we observed inconsistent results across replicates. All MBECs were found to be higher than the concentrations chosen for the coupon assays (Table 2).

TABLE 1.

MICs and minimal biofilm eradication concentrations^a

Antibiotic	Assay	MIC or MBEC value (mg/liter)
		ATCC 33591		ATCC 35984		ATCC 47076b
		MHB-ca	TGN	MHB-ca	TGN	MHB-ca	TGN
Daptomycin	MIC	1	1	1	1
	MBEC	4	>1,024	>1,024	>1,024
Doxycycline	MIC	16	16	0.25	0.25
	MBEC			128	512
Linezolid	MIC	2	1	2	1
	MBEC	8	>1,024	>1,024	>1,024
Moxifloxacin	MIC	0.125	0.125	0.06	0.125
	MBEC	2	>1,024	>1,024	>1,024
Rifampin	MIC	0.004	0.004	0.004	0.004
	MBEC	8	2	128	4
Vancomycin	MIC	1	8	2	8
	MBEC	>1,024	>1,024	>1,024	>1,024
Ceftazidime	MIC					0.25	0.125
	MBEC					>1,024
Ciprofloxacin	MIC					0.016	0.03
	MBEC					128
Colistin	MIC					0.5	0.25
	MBEC					>1,024
Meropenem	MIC					0.03	0.06
	MBEC					>256

^aValues in bold indicate an MIC superior to the CLSI breakpoint values (50).

^bThe MBEC assay results for the ATCC 47076 strain were unreliable and were not included.

TABLE 2.

Antibiotics experimental concentrations and rationales

Antibiotic	Concn (mg/liter)	Rationalea
Daptomycin	30	$\frac{{\text{AUC}}_{0-24\text{\hspace{0.17em}h}}}{24\text{\hspace{0.17em}h}}$
Doxycycline	1.5	$\frac{{\text{AUC}}_{0-24\text{\hspace{0.17em}h}}}{24\text{\hspace{0.17em}h}}$
Linezolid	5	Cminb
Moxifloxacin	1.5	$\frac{{\text{AUC}}_{0-24\text{\hspace{0.17em}h}}}{24\text{\hspace{0.17em}h}}$
Rifampin	2.5	$\frac{{\text{AUC}}_{0-24\text{\hspace{0.17em}h}}}{24\text{\hspace{0.17em}h}}$
Vancomycin	2	Target trough concn (20 mg/liter)
Ceftazidime	3.8	Cmin
Ciprofloxacin	1.625	$\frac{{\text{AUC}}_{0-24\text{\hspace{0.17em}h}}}{24\text{\hspace{0.17em}h}}$
Colistin	2.5	$\frac{{\text{AUC}}_{0-24\text{\hspace{0.17em}h}}}{24\text{\hspace{0.17em}h}}$
Meropenem	0.25	Cmin

^aPharmacokinetic and PK/PD information obtained from the EUCAST rationales for clinical breakpoint documents or from the literature (references in Table S1 in the supplemental material).

^bC_min, minimum concentration of drug in serum.

Kinetics of TEC effect.

The effect of TEC on ATCC 33591 MRSA biofilms is illustrated in Fig. 1 for CFU counts (Fig. 1A) and biomass (Fig. 1B and C). The CFU counts decreased modestly after 15 min of incubation (−0.68 log₁₀ CFU, P = 0.0365 from time zero), and no statistically significant differences were observed between the subsequent incubation times. A sharp decrease in the total biomass was observed after a 15-min incubation (−72.0% from time zero, P = 3.3e−06) and appeared to be maximal after 30 min (−95.82% from time zero, P = 7.0e−07; −19.45% from 15 min, P = 0.014). These effects were corroborated by macroscopic observation of the crystal violet-stained samples (Fig. 1C). Based on these results, an incubation time of 30 min with TEC was adopted for the next experiments.

FIG 1.

Kinetics of biofilm dispersal by TEC against MRSA S. aureus ATCC 33591 biofilms. (A) Log₁₀ CFU counts per coupon. (B) Total biomass assay (expressed as the percentage of time zero control). (C) Pictures of biofilm samples exposed to TEC for the indicated times (C-, negative control [coupon incubated in sterile culture medium]) stained with crystal violet. The values indicate the means of triplicate experiments (± standard error). Statistical analysis, repeated measure ANOVA, followed by paired t tests with Holm correction. *, P < 0.05.

Effect of individual enzymes compared to the TEC.

In a preliminary experiment, the effect of the individual enzymes composing the TEC was tested against biofilms of each of the three species in comparison with the cocktail (Fig. 2). The TEC was found to be at least equivalent to the most effective single enzyme against all three species. While we did not observe any significant biofilm dispersal when treating methicillin-resistant S. epidermidis (MRSE) ATCC 35984 biofilms with the individual enzymes, exposure to the TEC resulted in a 93% reduction of the biomass (P < 0.0001). The RNA/DNA endonuclease was found to be equivalent to the TEC against MRSA ATCC 33591, and the β-1,6-hexosaminidase was equivalent to the TEC against E. coli ATCC 47076.

FIG 2.

Effect of a 30-min treatment with the individual enzymes on the biomass of ATCC 33591, ATCC 35984, and ATCC 47076 biofilms and comparison to the TEC. Hexosaminidase, β-N-acetylhexosaminidase; endonuclease, DNA/RNA endonuclease; endoglucanase, endo-1,4-β-d-glucanase. Data expressed as percentage of the control. The values indicate the means of triplicate experiments (± standard error). Statistical analysis, two-way ANOVA followed by Holm-Sidàk post hoc test. NS, no significant difference; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Effect of TEC and of antibiotics alone or in combination against biofilms.

Biofilms exposed to TEC for 30 min were washed and reincubated in medium supplemented or not with antibiotics at clinically relevant concentrations (Table 2).

The results are illustrated in Fig. 3 (MRSA ATCC 33591), Fig. 4 (MRSE ATCC 35984), and Fig. 5 (E. coli ATCC 47076) for CFU counts (Fig. 3A, Fig. 4A, and Fig. 5A), biomass (Fig. 3B, Fig. 4B, and Fig. 5B), and fluorescence microscopy (Fig. 3C, Fig. 4C, and Fig. 5C). The details of the comparisons are shown in Tables S2 to S7 in the supplemental material.

FIG 3.

Effect of the combination of the TEC with antibiotics on MRSA ATCC 33591 biofilms. (A) CFU counts. (B) Biomass. (C) Fluorescence microscopy. (A and B) Blue, controls; red, samples treated for 30 min with TEC. T0, no reincubation; TGN, 24 h reincubation in TGN; others, 24 h reincubation in TGN with antibiotics at the indicated concentrations (see Table 1). CFU counts expressed in log₁₀ CFU per coupon, and biomass expressed in percentage of T0 controls. The values indicate the means of triplicate experiments (± standard error). Statistical analysis, two-way ANOVA followed by Tukey HSD comparing control samples to samples exposed to the enzymatic cocktail. NS, no significant difference; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. (C) Green, living cells (SYTO 9); red, dead cells (propidium iodide). ×20 magnification. Scale bars represent 50 μm. VAN, vancomycin; RIF, rifampin; LZD, linezolid; MXF, moxifloxacin; DAP, daptomycin.

FIG 4.

Effect of the combination of the TEC with antibiotics on MRSE ATCC biofilms. (A) CFU counts. (B) Biomass. (C) Fluorescence microscopy. (A and B) Blue, controls; red, samples treated for 30 min with TEC. T0, no reincubation; TGN, 24 h reincubation in TGN; others, 24 h reincubation in TGN with antibiotics at the indicated concentrations (see Table 1). CFU counts expressed in log₁₀ CFU per coupon, and biomass expressed in percentage of T0 controls. The values indicate the means of triplicate experiments (± standard error). Statistical analysis, two-way ANOVA followed by Tukey HSD comparing control samples to samples exposed to the enzymatic cocktail. NS, no significant difference; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. (C) Green, living cells (SYTO 9); red, dead cells (propidium iodide). ×20 magnification. Scale bars represent 50 μm. VAN, vancomycin; RIF, rifampin; LZD, linezolid; MXF, moxifloxacin; DAP, daptomycin; DOX, doxycycline.

FIG 5.

Effect of the combination of the TEC with antibiotics on E. coli ATCC 47076 biofilms. (A) CFU counts. (B) Biomass. (C) Fluorescence microscopy. (A and B) Blue, controls; red, samples treated for 30 min with TEC. T0, no reincubation; TGN, 24 h reincubation in TGN; others, 24 h reincubation in TGN with antibiotics at the indicated concentrations (see Table 1). CFU counts expressed in log₁₀ CFU per coupon, biomass expressed in percentage of T0 controls. The values indicate the means of triplicate experiments (± standard error). Statistical analysis, two-way ANOVA followed by Tukey HSD comparing control samples to samples exposed to the enzymatic cocktail. NS, no significant difference; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. (C) Green, living cells (SYTO 9); red, dead cells (propidium iodide). ×20 magnification. Scale bars represent 50 μm. CEF, ceftazidime; CIP, ciprofloxacin; CST, colistin; MER, meropenem.

A 30-min exposure of the coupons to TEC significantly reduced the CFU counts and biomass of all strains. The CFU counts were reduced by 2.04 log₁₀ in strain ATCC 33591, 2.54 log₁₀ in strain ATCC 35984, and 1.24 log₁₀ in strain ATCC 47076. Biomass was reduced by 95.2% in strain ATCC 33591, 88.6% in strain ATCC 35984, and 80.3% in strain ATCC 47076.

A 24-h reincubation of control samples was not associated with significant changes in CFU counts. Their biomass remained stable in strains ATCC 35984 and ATCC 47076 but increased by 24.4% in strain ATCC 33591.

Similarly, the CFU counts of the samples treated with the enzyme cocktail and then reincubated for 24 h in TGN were similar to those of nontreated samples for all three strains. The biomass was however significantly lower than that for control group samples for ATCC 33591 (−19.1%) and ATCC 35984 (−34.7%) biofilms.

The effect of a 24-h exposure to antibiotics on control biofilms was variable versus CFU counts and generally weak or nonexistent versus the biomass when compared to initial controls (statistical analyses shown in Tables S2 to S7 in the supplemental material).

Against staphylococci, vancomycin (20 mg/liter) caused a 1.84 log₁₀ CFU increase in the CFU counts of MRSE ATCC 35984 biofilms but no changes MRSA ATCC 33591. Conversely, the biomass of MRSA samples increased by 22.3% but did not change significantly in MRSE samples. Rifampin (2.5 mg/liter) caused a significant reduction of the CFU counts in both strains (ATCC 33591, −1.59 log₁₀ CFU; ATCC 35984, −3.39 log₁₀ CFU) but did not affect the biomass of either strain. Linezolid (5 mg/liter) had a limited effect on the CFU counts of strain ATCC 33591 (−1.50 log₁₀) and did not affect the CFU counts of strain ATCC 35984 nor the biomass of both strains. Moxifloxacin (1.5 mg/liter) impacted both strains differently despite similar MICs. In ATCC 33591 samples, the CFU counts did not change, and the biomass increased by 29.7%, while in ATCC 35984, a 2.52 log₁₀ CFU decrease was observed and the biomass remained stable. Daptomycin (30 mg/liter) had a strong effect on the CFU counts of both strains (ATCC 33591, −4.99 log₁₀; ATCC 35984, −3.50 log₁₀) as well as a moderate effect on the biomass of strain 33591 (−20.2%), while the biomass of strain ATCC 35984 remained unchanged. Doxycycline (1.5 mg/liter) was used against MRSE ATCC 35984 only as MRSA ATCC 33591 was resistant. No significant effect was observed for either CFU counts or biomass.

Against E. coli ATCC 47076 biofilms, ceftazidime (3.8 mg/liter) and ciprofloxacin (1.625 mg/liter) had a significant effect on CFU counts (ceftazidime, −2.06 log₁₀; ciprofloxacin, −3.21 log₁₀) and biomass (ceftazidime, −61.7%; ciprofloxacin, −48.33%) while colistin (2.5 mg/liter) and meropenem (0.25 mg/liter) did not.

Turning now to the results for biofilms of the three strains successively exposed to TEC and antibiotics at set concentrations, we observed significant reductions of CFU counts and biomass compared to those of samples treated only with antibiotics, except for the CFU counts of MRSA ATCC 33591 samples treated with daptomycin (which were already markedly decreased by the antibiotic alone) and the biomass of E. coli ATCC 47076 treated with ceftazidime (for which the difference between samples preexposed or not to the TEC cocktail did not reach significance). The difference in CFU counts between the samples preexposed to the enzymes and the control group samples was always higher than 2 log₁₀ and exceeded 3 log₁₀ in MRSA ATCC 33591 samples treated with vancomycin (−4.22 log₁₀, P < 0.0001) or moxifloxacin (−3.87 log₁₀, P < 0.0001), in MRSE ATCC 35984 samples treated with vancomycin (−6.84 log₁₀, P < 0.0001), and in all E. coli ATCC 470476 samples exposed to any of the 4 tested antibiotics (ceftazidime, −5.23 log₁₀, P < 0.0001); ciprofloxacin, −3.98 log₁₀, P < 0.0001; colistin, −5.57 log₁₀, P < 0.0001; meropenem, −6.11 log₁₀, P < 0.0001).

The difference in biomass between samples treated with antibiotics alone or with the combination of TEC and antibiotics exceeded 80% in most conditions, except for strain ATCC 33591 samples treated with daptomycin and strain ATCC 47076 samples treated with ciprofloxacin and colistin.

The two-way analysis of variance (ANOVA) (Tables S8 to S13 in the supplemental material) showed significant interaction parameters in all cases when considering the exposure to the enzyme cocktail and the reincubation of the samples as factors. This suggests that the tri-enzyme cocktail and the antibiotics have a synergistic effect on both CFU counts and biomass of all three strains.

The analysis of biofilms using fluorescence microscopy after LIVE/DEAD staining (Fig. 3C, Fig. 4C, and Fig. 5C) illustrates the observed CFU counts and biomass assays. The treatment of samples with the TEC dispersed an important proportion of the bacterial cells, which were enough to regrow a biofilm after 24 h of reincubation in TGN. This recovery appeared to be inhibited by the reincubation with antibiotics. The aspect of control samples exposed during 24 h to vancomycin, ciprofloxacin, and meropenem appeared unchanged from the samples incubated in TGN, keeping a majority of live cells (green), while the samples incubated with daptomycin exhibited a high proportion of dead cells (red).

Cytotoxicity tests.

The cytotoxicity of the tri-enzyme cocktail for different types of eukaryotic cells was then evaluated by measuring the release of the cytosolic enzyme lactate dehydrogenase (LDH) in the culture medium after 1-h and 24-h exposures to TEC at two concentrations (concentration used against biofilms and a 1:1 dilution). At 1 h, no significant toxicity was detected (values of <5% in all cases). At 24 h, the release of the cytosolic enzyme lactate dehydrogenase (LDH) reached a mean percentage (95% confidence interval) of 3.0% (−1.1–7.0) for primary murine fibroblasts, −1.6% (−4.7–1.4) for MG-63 human osteosarcoma cells, and −3.2% (−17.0–10.7) for J774 murine macrophages.

DISCUSSION

This study found that a tri-enzymatic cocktail containing two glycoside hydrolases (an endo-1,4-β-d-glucanase and a β-1,6-hexosaminidase) and a nonspecific RNA/DNA endonuclease had a significant dispersal effect on mature biofilms of S. aureus, S. epidermidis, and E. coli reference strains grown on titanium alloy and that the successive treatment of these biofilms with this enzyme cocktail and antibiotics at concentrations relevant to the clinical setting was synergistic, resulting in significantly lower residual CFU counts than those with either individual treatment.

While the potential of glycoside hydrolases and DNase I as biofilm dispersal agents has been previously described in the literature (19, 22), the originality of our work resides in the association of enzymes targeting biofilm matrix components found in biofilms of varied species (30), which is hypothesized to help sidestep the limited spectrum of individual enzymes caused by the significant variations in matrix contents observed among bacterial species and strains (31, 32). Indeed, the results of individual enzyme assays performed in this study underline the added value of combining enzymes as was done in the TEC. The β-1,6-hexosaminidase used in this study is a dispersin B homolog targeting the poly-β-1,6-N-acetylglucosamine (PNAG) component of the biofilm matrix (33, 34). PNAG is an important constituent of the matrix of multiple species’ biofilms and is required for the formation of biofilm by S. epidermidis (32, 35, 36) and S. aureus strains expressing the icaABCD operon (32, 37) and of E. coli strains expressing the pgaABCD locus (31, 38). Dispersin B was shown to disperse the biofilms of PNAG-producing strains of multiple species, including those included in our study (20, 39, 40). Cellulase is as endo-1,4-β-d-glucanase targeting cellulose, which is an important constituent of the biofilm matrix of Enterobacter spp. (41), including E. coli (42). It has been shown to disperse E. coli (27) and Pseudomonas aeruginosa (21) biofilms in vitro. The recombinant nonspecific endonuclease targets both DNA and RNA phosphodiester bonds (43). DNA is a common component of the biofilm matrix of most bacterial species, including S. aureus, S. epidermidis, and E. coli (44). While the biofilm dispersal properties of this enzyme had not been previously investigated, it has been favorably compared to the widely studied DNase I (22) for the hydrolysis of DNA in sputum (45).

The dispersal of biofilms by enzymes similar to the ones constituting TEC was shown in previous publications to reduce the tolerance of bacterial biofilms to antibiotics (21, 25, 46) and antiseptics (27). However, limitations related to the use of single enzymes have been reported. Izano et al. (39) observed that DNase I did not disperse biofilms of S. epidermidis contrarily to S. aureus and that dispersin B had an inverse effect by dispersing S. epidermidis biofilms and leaving S. aureus biofilms intact. Accordingly, they found that the dispersal of these biofilms by the appropriate enzyme (DNase I for S. aureus of dispersin B for S. epidermidis) improved the effect of the detergent cetylpyridinium chloride on CFU counts (39).

Our results illustrate the broad spectrum of the tri-enzymatic cocktail, which stems from the combination of enzymes with different activity spectrums (30). Furthermore, these results support the hypothesis of a synergy of enzymatic dispersal and a broad selection of antibiotics at concentrations clinically relevant to the treatment of PJIs against biofilms. As could be expected from the observed MBEC values, the antibiotic concentrations used for the coupon assays had a minimal effect on control biofilms, but their combination with an enzymatic pretreatment resulted in significant reductions of culturable cell numbers as well as total biomass in nearly all combinations of bacterial species and antibiotics.

The cytotoxicity of TEC to different cell lines relevant to the joint environment (macrophages, fibroblasts, osteoblasts) was found to be negligible after a 24-h exposure. An explanation to this observation could be that the targets of these enzymes are either not produced by mammalian cells (cellulose) or irrelevant to their proliferation (eDNA and PNAG).

These results should however be interpreted with caution, as our study presents some limitations. We limited ourselves to the study of biofilms of reference strains. While this facilitates the replication of our results by other teams in comparison to clinical strains, the loss of diversity in gene expression could mask the effect of the variability in matrix composition observed in some species (47). The antibiotic concentrations considered in our study did not take into account the serum protein binding. Among the drugs used in this study, doxycycline and daptomycin show a high serum protein binding, which could lead to an overestimation of their effects in our experimental setting. However, the albumin concentration in synovial fluid is three to four times inferior to the serum concentration (48). This difference in albumin concentrations could lead to a higher free fraction of drugs in the synovial fluid than in serum. For this reason, the total serum concentration appeared to be a reasonable estimate that limited the risk of underestimating the effects of antibiotics. The in vitro nature of our experiments limits extrapolation to the clinical setting, as it does not represent the diversity of the situations encountered in a clinical setting nor does the use of set concentrations of antibiotics reproduce the complex pharmacokinetics observed in the bone and joint environment (49).

However, these observations support our hypothesis that a treatment approach combining local administration, such as an intraarticular injection, of a broad-spectrum enzymatic cocktail and antibiotics could prove to be beneficial in a PJI setting where the variety of pathological bacterial species encountered is important (3). Further studies in more refined models, such as animal PJI models, should therefore be undertaken to confirm our in vitro results and determine an adequate strategy around this approach in complement to DAIR.

MATERIALS AND METHODS

Bacterial strains, antibiotics, and enzymes.

(i) Strains. Methicillin-resistant Staphylococcus aureus ATCC 33591, methicillin-resistant Staphylococcus epidermidis ATCC 35984, and Escherichia coli ATCC 47076 were used as representatives of the bacterial species usually encountered in PJI.

(ii) Antibiotics. The following antibiotics were used as laboratory standards: ciprofloxacin HCl (potency, 93.9%) and moxifloxacin HCl (potency, 90.9%) from Bayer AG (Leverkusen, Germany); colistin (potency, 79.6%) and doxycycline (potency, 90.2%) from Sigma-Aldrich (Saint-Louis, MO, USA); and linezolid (potency, 99.2%) from RibX Pharmaceuticals (Morristown, NJ, USA). The following antibiotics were used as powders for injections in humans: ceftazidime (potency, 72.5%; Panpharma France, Luitré, France), daptomycin (potency, 100%; MSD, Puteaux, France), meropenem (potency, 92%; Hospira Benelux BVBA, Brussels, Belgium), rifampin (potency, 100%; Aventis, Schiltigheim, France), and vancomycin (potency, 97.5%; Mylan, Hoeilaart, Belgium). The antibiotic concentrations were set on the basis of the pharmacokinetic/pharmacodynamic (PK/PD) characteristics of the antibiotics studied and their pharmacokinetic parameters in humans described in the EUCAST rationales for clinical breakpoints or in the literature (Table 2).

(iii) Tri-enzymatic cocktail. The tri-enzymatic cocktail (TEC; OrthenzyQure) used in this study was designed and provided by OneLife SA (Louvain-la-Neuve, Belgium). It consists of a sterile Tris-buffered (pH = 7) aqueous solution containing 400 U/ml of aspecific DNA/RNA endonuclease from Serratia marcescens (c-LEcta GmbH, Leipzig, Germany), 50 U/ml of endo-1,4-β-d-glucanase from Aspergillus niger (Sigma), and 0.06 U/ml of β-N-acetylhexosaminidase from Actinobacillus pleuropneumoniae (GIGA, Liège, Belgium) (composition described in patent publication numbers WO2017/211737 [28] and BE2020/5726 [29]). The TEC was stored at 4°C and was heated at 37°C for 30 min before use. The TEC is available upon request at OneLife SA for research purposes (under R&D reference OL-RD1000/OrthenzyQure).

In order to assay the biofilm dispersal effect of the individual enzymes, OneLife provided solutions of the individual enzymes at concentrations identical to those used in the TEC.

Susceptibility testing.

MICs were determined by microdilution following the CLSI protocol (50) in MHB-ca and TGN (TSB + 2% NaCl + 1% glucose), i.e., the medium used for growing biofilms. MICs of daptomycin were tested in MHB-ca and TGN containing 50 mg/liter of Ca²⁺.

MBECs were assayed by growing biofilms in Calgary biofilm devices (CBD; Innovotech, Edmonton, Canada) in MHB-ca and TGN. Bacterial suspensions were prepared in phosphate-buffered saline (PBS), starting from overnight cultures on Trypticase soy agar (TSA) from frozen stocks. The suspensions were adjusted to a turbidity of 0.5 McFarland and diluted 1:10 in medium. A total of 150 μl was added per well of the CBDs. The CBDs were incubated for 24 h at 37°C under a continuous rotation of 50 rpm to allow biofilm formation. Antibiotic dilutions in the two media—with 50 mg/liter of Ca²⁺ when assaying daptomycin—were prepared in new 96-well plates at concentrations ranging from 1,024 mg/liter to 0.5 mg/liter (256 mg/liter to 0.125 mg/liter for meropenem). The lids of the CBDs were rinsed twice in PBS before being transferred to the plates containing the antibiotics. The plates were then incubated for 24 h at 37°C with a continuous rotation of 50 rpm. At the end of the antibiotic challenge, the lids were rinsed twice in PBS before being transferred to new plates containing plain medium. The plates were then sonicated for 10 min (Branson 5510 ultrasonic bath; Emerson Electric, Saint-Louis, MO, USA). New sterile lids were placed, and the plates were incubated for 24 h at 37°C without agitation. At the end of the incubation period, the plates were examined for bacterial growth, and optical density measures at 650 nm (OD₆₅₀) were performed using a SpectraMax M3 spectrophotometer (Molecular Devices, San Jose, CA, USA).

Biofilm culture.

Bacterial suspensions were prepared in PBS, starting from overnight cultures on TSA from frozen stocks. The suspensions were adjusted to an OD₆₂₀ of 0.5 (∼8.5 log₁₀ CFU/ml) using a CECIL 2021 spectrophotometer (Cecil Instruments Ltd, Milton, United-Kingdom). The suspension was then diluted 1:100 in TGN. The biofilms were grown on titanium alloy (Ti-6Al-4V) coupons (BioSurface, Bozeman, MN, USA) in 12-well plates using 3 ml of the final suspension per coupon. The plates were incubated for 24 h (ATCC 33591 and ATCC 35984) or 48 h (ATCC 47076) at 37°C, with a constant orbital shaking of 50 rpm. In the case of strain ATCC 47076 biofilms, the medium was changed after 24 h. These culture times were selected, as they allowed the biofilms to reach steady state over the study period and were determined in previous growth kinetic experiments for S. aureus (14). The culture duration period for S. epidermidis was derived from these results. Growth kinetic experiments were performed in this study for E. coli biofilms and showed that a steady state was reached after 48 h of culture. Samples were analyzed for CFU counts and biomass every 24 h for the growth kinetic experiments.

Biofilm treatments.

(i) Individual enzyme assays. The biofilm samples were rinsed twice with sterile PBS. They were then transferred to 24-well plates filled with either 1 ml of TEC or 1 ml of one of the individual enzymes at 37°C. The plates were then incubated for 30 min at 37°C. At the end of the incubation period, the biofilms were rinsed twice in sterile PBS and assayed for total biomass (see below).

(ii) Combination of enzymatic cocktail and antibiotics. The biofilm samples were separated into a control group and an enzyme cocktail group. The coupons of both groups were rinsed twice in sterile PBS before being submitted to the treatments. The tri-enzyme cocktail group samples were transferred to a 24-well plate, in which the wells were previously filled with 1 ml of enzyme cocktail at 37°C. The plates were then incubated at 37°C with no agitation for either 15, 30, and 60 min (when studying the kinetics of activity) or 30 min (other experiments). After incubation, the samples were rinsed twice in sterile PBS before either immediate analysis (see below for analysis technique descriptions) or reincubation. Control group samples were directly rinsed in sterile PBS. All samples were then reincubated for 24 h at 37°C under a constant orbital shaking of 50 rpm in TGN or TGN containing antibiotics at set concentrations (TGN + 50 mg/liter of Ca²⁺ for samples exposed to daptomycin) (see Table 2) in 12-well plates containing 3 ml of medium per well.

Biofilms analysis.

The coupons were rinsed twice in sterile PBS and then analyzed using the following techniques.

(i) CFU counts. Coupons were placed in 15-ml conical tubes (Greiner Bio-One International GmbH, Kremsmünster, Austria) filled with 2 ml of sterile PBS. In order to detach the biofilms and suspend the bacteria, the tubes were then vortexed for 30 s at maximum intensity (Vortex-Genie 2; Scientific Industries, Inc., Bohemia, NY, USA), sonicated for 5 min (Branson 5510 Ultrasonic bath, Emerson Electric, Saint-Louis, MO, USA), and vortexed again for 30 s. The suspension was then sampled and serially diluted before plating on TSA. The TSA plates were then incubated for 24 to 48 h at 37°C until the colonies reached an adequate size. CFU counts were then obtained using an automated method (image acquisition using Gel Doc XR+ and image processing using Quantity One (Bio-Rad, Hercules, CA, USA).

(ii) Biomass assay. Coupons were dried overnight at 60°C. They were then stained with 1 ml of 1% crystal violet solution for 10 min (Sigma-Aldrich Corp.). The coupons were then rinsed twice using deionized water to remove the excess dye. Afterward, they were placed for 1 h in 24-well plates filled with 1 ml of 66% acetic acid solution (Merck KGaA, Darmstadt, Germany) to resolubilize the dye. The coupons were removed from the wells, and the absorbance of the solution was read at 570 nm using a SpectraMax M3 spectrophotometer (Molecular Devices, San Jose, CA, USA).

(iii) Fluorescence microscopy. Coupons were rinsed twice in sterile deionized water to avoid interference from phosphate ions. They were then stained using the Filmtracer LIVE/DEAD biofilm viability kit (Thermo Fisher Scientific Inc., Waltham, MA, USA) according to the manufacturer’s instructions. Coupons were then rinsed again twice in sterile deionized water before imaging using an Axio Imager Z1 microscope fitted with an ApoTome 1 attachment (Zeiss, Oberkochen, Germany) at a ×20 magnification. The images were postprocessed using FIJI (51, 52), using the maximum intensity projection technique, and by adjusting the brightness of the channels.

Cytotoxicity assays.

(i) Cell culture. J774 murine macrophages (Sandoz Forschung Laboratories, Vienna, Austria) (53), MG-63 human osteoblasts (ATCC CRL-1427), and primary murine fibroblasts (54) were used for the cytotoxicity assays. The cells were grown in RPMI 1640 or Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum (all products from Thermo Fisher Scientific Inc., Waltham, MA, USA) at 37°C in a 5% CO₂ atmosphere. The cells were used for the assays between passages 2 and 5.

The cells were cultured initially in 250-cm³ cell culture flasks before being transferred to 96-well plates at a concentration of 10⁴ cells per well. The plates were incubated for 24 h at 37°C with 5% CO₂ to allow cellular adhesion to the wells before testing.

(ii) Assay.Cytotoxicity was assessed by measuring the release of the cytoplasmic enzyme lactate dehydrogenase (LDH) in the culture medium, using the cytotoxicity detection kit (Sigma-Aldrich). The wells were rinsed with sterile PBS before adding 100 μl of medium with 1% decomplemented fetal bovine serum. A total of 100 μl of the following test substances were then added to the wells (final concentrations): TEC (concentration used against biofilms and half of this concentration), Triton X-100 solution (1%, positive control), medium with 1% decomplemented serum (negative control). Three empty wells per experiment were filled with 200 μl of medium with 1% decomplemented serum to measure the background signal.

The plates were then incubated for 1 h or 24 h at 37°C in a 5% CO₂ atmosphere. A total of 100 μl of the supernatant of each well was aspirated and transferred to a new 96-well plate and added by 100 μl of the extemporaneously prepared LDH assay solution as per the manufacturer’s instructions. The plates were then incubated for 30 min at room temperature in the dark. The light absorbance was then measured at 490 nm and 650 nm using a SpectraMax M3 spectrophotometer (Molecular Devices, San Jose, CA, USA).

Data transformation and statistical analysis.

CFU counts were normalized using a base 10 logarithmic transformation before analysis. Absorbance values of the biomass assays were transformed into percentages of the T0 (no reincubation) controls after subtracting the average value of negative controls. The absorbance values of the LDH assays were transformed into percentages using the following equation:

$$\text{Cytotoxicity\hspace{0.17em}}\left(\text{\%}\right)=\frac{\text{Abs.\hspace{0.17em}exp}{\text{.\hspace{0.17em}value}}_{420\text{nm}-650\text{nm}}\hspace{0.17em}-\hspace{0.17em}\text{Abs.\hspace{0.17em}control\hspace{0.17em}}{(-)}_{420\text{nm}-650\text{nm}}}{\text{Abs.\hspace{0.17em}control\hspace{0.17em}}{(+)}_{420\text{nm}-650\text{nm}}\hspace{0.17em}-\hspace{0.17em}\text{Abs.\hspace{0.17em}control\hspace{0.17em}}{(-)}_{420\text{nm}-650\text{nm}}}\times 100$$

The statistical analysis was then performed using R (version 3.6.1) and the package car (version 3.0.6) or in GraphPad 8.4.3 (GraphPad Software, San Diego, CA). The TEC kinetics experiment data were analyzed using repeated measures ANOVA followed by paired t tests with Holm correction. The data relating to the combined effect of TEC and antibiotics were analyzed using two-way ANOVA after assessing the normality of residues with QQ plots and the equality of variances using the Levene test. The Tukey honestly significant difference (HSD) test was used to perform pairwise comparisons. The synergy of treatments was defined as a statistically significant interaction parameter of the two-way ANOVAs (55). The threshold for alpha errors is 0.05. The data reported in the text are the means with 95% confidence interval.

Data availability.

The data for the experiments reported in this publication are available upon request to the corresponding author.