A theoretical exploration of protocols for treating prosthetic joint infections with combinations of antibiotics and bacteriophage

Abstract

With the increase in the placement of prosthetic joints and other hardware in the body, associated infections have risen. These infections are complicated to treat due to the underlying bacteria generating matrices that resist clearance by immune system effectors or antibiotics. These matrices, biofilms, have two primary ways of being eradicated: either by physical removal during surgery or by killing the underlying bacteria, usually with antibiotics. The viruses that kill bacteria, bacteriophages, are readily capable of entering biofilms and eradicating the bacteria therein. Therefore, bacteriophages have therapeutic potential as a supplement to antibiotics for the treatment of prosthetic joint infections. In this investigation, we generate a mathematical and computer-simulation model to explore the contributions of the innate immune system with antibiotics, bacteriophage, and the joint action to control biofilm-associated infections. Our results question the proposition that bacteriophages are an effective addition in the treatment of prosthetic joint infections.

Introduction

Over 1,460,000 knee and hip prosthetic joints will be placed in the US annually by 2050^1,2. Prosthetic joint infections (PJIs) with Staphylococcus aureus and Staphylococcus epidermidis in up to 3% of cases are common sequelae to these procedures and require additional surgery and extensive periods of antibiotic treatment to cure^1,2. A major limitation in treating PJIs is the formation of biofilms on the surfaces of the prosthetics, where bacteria are resistant to the action of antibiotics. Another limitation to treating PJIs and other bacterial infections is the emergence of bacteria resistant to the treating antibiotics. There is a critical need to develop effective interventions to prevent and treat PJIs that can destroy the extracellular matrix, make the bacteria residing in biofilms more susceptible to antibiotics at physiologic concentrations in the joint space, and ultimately kill the underlying bacteria. Bacteriophages (phage) have many characteristics that suggest they could meet this need as a therapy that is complementary to antibiotic PJI treatment.

These bacterial viruses can be particular to one species (and indeed one strain in some cases) of bacteria. Upon interacting with their host bacteria, these viruses bind to a receptor, which is generally a protein or polysaccharide displayed on the outside of the bacterial cell; after binding, they inject their genetic material, hijack the bacteria’s metabolic processes to produce more copies of themselves, and then ultimately lyse the host bacterial cell, producing dozens to hundreds of viral particles³. Phages can be beneficial for the treatment of PJIs for several reasons: (1) Some carry enzymes called depolymerases, which degrade the extracellular matrix of biofilms⁴. This degradation allows the phage to penetrate the biofilm, reach the bacterial cells, and enhance antibiotic penetration⁵. (2) Phages have been shown to synergize with other treatments. For example, using phages alongside antibiotics, enzymes, antiseptics, and detergents can enhance the killing of bacteria^4,6–8. (3) By lysing bacteria within the biofilm, phages can help disrupt the biofilm structure, making the remaining bacteria more susceptible to other treatments or clearance by the host’s immune system⁹. (4) Once inside the biofilm, phages can replicate within the bacterial cells and cause cell lysis, releasing new phage particles that can infect neighboring bacteria. However, the effectiveness of phage can vary depending on several factors, like limited penetration (as not all phages deeply penetrate biofilms) or bacteria resistant to the phage. The prevalence of these virtues and limiting factors in clinical settings represents a knowledge gap that impedes the development of phage therapy for PJIs. Several studies support the potential efficacy of phage therapy for bone and joint infections¹⁰. A meta-analysis published in 2024 reported an infection remission rate of 78% after intervention with phage^10–13.

In this study, we employ mathematical and computer-simulation models (as refs. ^14–16) to design and optimize protocols for the joint action of antibiotics and phages in the treatment of PJIs. Our models identify the factors and key parameters contributing to the success and failure of antibiotics, phages, and their combinations in the treatment of PJIs. These parameters can be estimated experimentally, and, most importantly, the predictions of the model can be tested using both in vitro and in vivo experiments. We discuss those experiments and their potential implications in the clinic for treating PJIs.

Results

A mathematical model and its predictions

A diagram of the model we use to simulate the course of infections and their treatment with antibiotics and phage is presented in Fig. 1. The details of the mathematical and computer-simulation models are given in the Supplementary Text, Equations, and Tables.

Fig. 1. Diagram of a mathematical and computer-simulation model of prosthetic joint infections.

The bacteria (yellow rods), bacteriophage (blue phage), resource (purple R particles), immune cells (green), and the antibiotic (striped pill) can be in two sites: the body, where they are planktonic (blue compartment), or in a biofilm on the prosthetic joint (pink compartment). A limiting resource and immune cells enter the environment at a continuous rate. This is the same rate at which unconsumed resources, bacteria, phage, immune cells, and antibiotics exit the system. Antibiotics and bacteriophages can be dosed independently.

In Figs. 2 and 3, we consider the changes in the densities of bacteria, phage, the immune cells, the concentration of the limiting resource, and the antibiotic concentration in the planktonic and biofilm states. These changes are given by Supplementary equations 1–13. The definitions of the variables and values of the parameters in the following simulations are those presented in Supplementary Tables 1 and 2, unless otherwise stated. All treatment simulations are conducted at a multiplicity of infection (MOI) of 0.1 to mirror experiments conducted in refs. ^15,17,18.

Fig. 2. Changes in the densities of bacteria and phage in an infection with two habitats, planktonic and biofilm with an innate immune system.

A An infection without treatment. B Treatment with 5 µg/mL of an antibiotic. C Treatment with 10⁶ PFU/mL of phage for an initial multiplicity of infection (MOI) of 0.1. D Treatment with an antibiotic and phage.

Fig. 3. Changes in the densities of bacteria and phage in an infection with two habitats, planktonic and biofilm, with an innate immune system and antibiotic-resistant subpopulations.

A An infection without treatment. B Treatment with 5 µg/mL of an antibiotic. C Treatment with 10⁶ PFU/mL of phage for an initial multiplicity of infection (MOI) of 0.1. D Treatment with an antibiotic and phage. E Treatment with a phage and then an antibiotic 8 h later.

In Figs. 2 and 3, we follow these simulated dynamics with a biofilm site without and with the emergence of antibiotic resistance, respectively. These simulations are initiated with 1/100-th of the densities of the populations of the limiting resource at equilibrium in the absence of treatment. These equilibria are obtained when one sensitive cell is grown up in the absence of antibiotics, phage, or the innate immune system (the experimental analog for this would be a 1/100 dilution of an overnight bacterial culture). In Figs. 2A and 3A, we consider the control of an infection by the innate immune system in the absence of treatment. In Figs. 2B and 3B, we consider the addition of fixed doses of a bactericidal antibiotic. In Figs. 2C and 3C, we consider the addition of fixed doses of bacteriophage, which are able to enter the biofilm. In Figs. 2D and 3D, we consider the joint action of an antibiotic and phage in controlling the infection. In Fig. 3E, we consider the joint action of an antibiotic and phage when the phage is given 8 h after the antibiotic.

In Fig. 2, the immune system alone cannot control the infection, and, within short order, the system becomes resource-limited (Fig. 2A). The addition of an antibiotic can suppress the planktonic cells. However, it can only suppress the biofilm-dwelling bacteria by less than one log (Fig. 2B). The phage, on the other hand, is able to control the infection initially; however, “boom-and-bust” type cycles emerge after the initial control of the planktonic cells and thus the biofilm-dwelling bacteria are maintained around the initial infection density¹⁹ (Fig. 2C). Surprisingly, the combination of the antibiotic and phage yields a dynamic nearly identical to that of the antibiotic alone—implying that the bacteriophage is not affecting this dynamic (Fig. 2D). Taken together, these simulations predict that the concomitant use of antibiotics and phage will result in single agent dominance, particularly the action of the antibiotic being favored.

In Fig. 3, like in the previous figure, the immune system is unable to control the infection. However, antibiotic-resistant subpopulations do not arise and, moreover, are lost (Fig. 3A). Treatment with just an antibiotic is not able to control the infection; the antibiotic-resistant subpopulations are able to increase in density but not in relative frequency; and the planktonic antibiotic-sensitive bacteria are maintained as a minor population due to transition from the biofilm-associated antibiotic-sensitive cells (Fig. 3B). The addition of the phage gives quantitively and qualitatively the same results as in Fig. 2C, as the antibiotic-resistant cells are rapidly lost (Fig. 3C). Interestingly, compared to 2D, the antibiotic in Fig. 3D is not able to govern the dynamics of the infection as well, but the antibiotic-resistant subpopulations are still not able to increase in frequency but are maintained. Taken together, administration of a bacteriophage without antibiotic is capable of eliminating any antibiotic-resistant subpopulations, which could make subsequent antibiotic treatment more effective. The co-administration of these antimicrobial agents is able to control the infection such that it does not rise to the level of an untreated infection, and in some cases, the density of the planktonic bacteria is suppressed (but not the biofilm bacteria) by up to three logs from the initial infection density, but still, the infection is not cleared. When the antibiotic administration is delayed such that the phage is dosed first and then the antibiotic begins to be dosed 8 h later, the cycles observed in Fig. 3D are lower in magnitude, and the antibiotic-resistant populations are lost. This delayed treatment can decrease the planktonic bacteria density by nearly six logs (Fig. 3E).

Discussion

Infections of prosthetic joints with Staphylococcus spp. and other bacteria are common sequelae to surgeries that place hardware into the body^20,21. These infections are particularly difficult to treat due to the emergence of biofilms on the hardware, which decreases the efficacy of both the innate immune system and antibiotics²². Bacteriophage (phage), capable of penetrating the biofilm⁶, would seem to be the perfect addition to antibiotic therapy to eliminate the bacteria that the antibiotics themselves cannot kill. The results of the simulations we present in this study support the complementary nature of phage in treating PJIs in only limited circumstances, particularly when the phage is applied before the antibiotic. While generally this may not be feasible in the clinic, in the case of PJIs it is as these patients are generally not septic and these infections are usually subacute/chronic rather than acute and do not necessitate emergency treatment.

With a biofilm, both antimicrobial agents are able to suppress the infection but fail to clear it. Most notably, the antibiotic, and not the phage, seems to be the determining factor in the dynamics of treatment despite the inability to kill cells in the biofilm. The primary reason for the limited effectiveness of phage can be attributed to the predator-prey-type dynamics that emerge¹⁹. As the density of the bacteria is brought down by the phage, the rate of phage killing tends to zero as phage and bacteria interact in a density-dependent manner. Consequently, the phage begins to be washed out, and its density declines; thus, the bacteria can increase in density again. On the other hand, antibiotics still kill at low bacterial densities; therefore, the antibiotic does not cause predator-prey-type cycles to emerge.

That being said, there is still likely clinical utility to phages. One is that these viruses can kill and control antibiotic-resistant bacteria and, as demonstrated above, could have utility if used before antibiotics to prevent the ascent of antibiotic-resistant subpopulations. Second, while the phage is not capable of clearing the infection, it does control the density of the infecting bacteria, potentially precluding the exhaustion of the innate immune system. Our conclusions are based solely on mathematical and computer-simulation models. It could well be that different results are obtained when in vitro or in vivo experiments are performed. The utility of theory at large is in generating testable hypotheses and predictions that can be supported or rejected by experiments.

The models employed here do not take into account four things: (1) the interactions that occur within a biofilm are in a structured environment rather than the mass-action environment simulated here²³; (2) bacterial populations that are resistant to the phage could emerge; (3) populations which are resistant to both the phage and the antibiotic could also emerge; and (4) the role of the adaptive immune system. We do not believe these complexities require further modeling. First, models that consider structured habitats are much more complicated, and their predictions are more difficult to test. Second, there are many recently described phages for which resistance is extremely rare or cannot be generated^15,24. We believe the next step in the development of phages for treating PJIs is experimentation, particularly with different phages and varying antibiotics. Fundamentally, the role of modeling is to generate hypotheses which can be tested by experimentation. Here, we have generated several hypotheses which are amenable to testing (and indeed rejection) by the proposed experiments.

The starting conditions we use in our model can readily be recreated in vitro^25,26. For the in vitro experimentation, many of the common biofilm models would successfully recreate the conditions simulated here. These experiments could also be performed in a range of model systems, such as mice, as ultimately, the utility of this work comes in the clinical application of phages and antibiotics to humans.

Methods

Numerical solutions (simulations)

For our numerical analysis of the coupled, ordered differential equations presented (Equations 1–13), we used Berkeley Madonna²⁷ with the parameters presented in Supplementary Table 2. Copies of the Berkeley Madonna programs used for these simulations are available at www.eclf.net.

Author contributions

Conceptualization: B.R.L., T.G.G., B.A.B., M.H.W. Methodology: B.R.L., T.G.G., B.A.B. Investigation: B.R.L., T.G.G., B.A.B. Visualization: B.R.L., T.G.G., B.A.B. Funding acquisition: B.R.L. Project administration: B.R.L., T.G.G., B.A.B., M.H.W. Supervision: B.R.L., T.G.G., B.A.B., M.H.W. Writing—initial draft: B.R.L., T.G.G., B.A.B., M.H.W. Writing—review & editing: B.R.L., T.G.G., B.A.B., M.H.W.

Data availability

The Berkeley Madonna programs used for these simulations are available at EcLF.net. All data are presented in this article or its supplementary materials.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41522-025-00908-2.