Klebsiella pneumoniae is both a natural component of human microbiota, colonizing the gastrointestinal tract and nasopharynx in healthy individuals (Chang et al., 2021), and a dangerous pathogen, responsible for approximately 11–15% of hospital-acquired pneumonia cases and up to 8% of all healthcare-associated infections (Chen et al., 2023; Grosjean et al., 2024; Rahmat Ullah et al., 2024).

Infections caused by K. pneumoniae represent a significant concern in healthcare settings due to the increasing prevalence of antibiotic-resistant strains. In Europe, over one third of K. pneumoniae isolates demonstrate resistance to at least one major antimicrobial drug class, including fluoroquinolones, third-generation cephalosporins, aminoglycosides, or carbapenems (European Centre for Disease Prevention and Control, 2022). These infections are responsible for an estimated 650,000 deaths annually worldwide (Murray et al., 2022). In light of this significant public health concern, the World Health Organization (WHO) has identified K. pneumoniae as a high-priority pathogen in its 2024 Bacterial Priority Pathogens List (World Health Organization, 2024).

One promising alternative for combating infections caused by antibiotic-resistant strains is bacteriophage (phage) therapy and its derivatives, including polysaccharide depolymerases and endolysins (Danis-Wlodarczyk et al., 2021; Durbas and Machnik, 2022). Bacteriophages—viruses that specifically infect and lyse bacteria—have demonstrated promising results in treatment applications (Pirnay et al., 2024). Although clinical data on recombinant phage proteins in humans are still unavailable, their efficacy has been convincingly demonstrated in animal models (Volozhantsev et al., 2020; Gorodnichev et al., 2021a; Hua et al., 2022; Li et al., 2024).

Phage therapy is commonly implemented under Article 37 of the Declaration of Helsinki, providing “compassionate care” when conventional antimicrobial therapies fail. However, standard treatment protocols generally do not recommend antibiotic withdrawal, meaning that phage therapy is often administered concurrently with antibiotics. This raises important questions about the potential interactions between these antimicrobial agents.

The combined use of phages and antibiotics can yield varying outcomes: synergy, additive effect, or antagonism. Synergy occurs when the bactericidal action of antimicrobial agents together is greater than the sum of their individual effects. Additive effects indicate an impact equal to the sum of the individual effects, while antagonism describes an interaction where the action of one agent inhibits the effect of the other (Liu et al., 2020).

In therapeutic contexts, antagonistic interactions between phages and antibiotics, which can reduce overall treatment efficacy, are considered particularly critical—even more so than synergy among antimicrobial agents. Reflecting this, several phage therapy centers have incorporated in vitro tests for detecting combined effects of phages and antibiotics as part of the standard phage selection process (Onallah et al., 2023).

To date, no cases of antagonism between phages and antibiotics targeting K. pneumoniae have been documented. Instead, both in vitro (Bedi et al., 2009; Verma et al., 2010; Qurat-ul-Ain et al., 2021; Han et al., 2022; Xu et al., 2022; Lew et al., 2023) and in vivo (Chhibber et al., 2008; Pacios et al., 2021; Wang et al., 2021) studies have demonstrated high efficacy of these agents when used together. However, differences in bacterial strains, phages, antibiotics, and methodologies often complicate comparisons across studies.

The aim of this work was to compare various in vitro methods for assessing combination effects of antibiotics and phages. Using three strains of K. pneumoniae, we evaluated the individual and combined actions of four antibiotics (gentamicin, levofloxacin, meropenem, and chloramphenicol), three phages (Dlv622, Seu621, and FRZ284), and a depolymerase (Dep622). To quantify combination effects, we applied the most commonly used methods for such studies: the t-test, Bliss independence model, two-way ANOVA, and checkerboard assay.

In recent years, the scientific community has actively explored the potential effects of combined phage-antibiotic therapy. Numerous in vitro studies have demonstrated both synergistic and antagonistic interactions (Schindler, 1964; Tagliaferri et al., 2019; Jiang et al., 2020; Liu et al., 2020; Nicholls et al., 2023; Shein et al., 2024), and these effects have been further validated in in vivo models (Oechslin et al., 2017; Grygorcewicz et al., 2020; Santamaría-Corral et al., 2023). These findings highlight the potential for translating such combined approaches into clinical practice, offering promising avenues for improved therapeutic outcomes.

One of the key challenges in studying phage-antibiotic combinations is the difficulty in comparing results across studies due to significant differences in experimental methods, limited characterization of bacterial strains and phages, and a restricted number of tested combinations. These limitations often result in a tendency to describe synergistic effects as unique to specific phage-antibiotic pairs (Bedi et al., 2009; Xu et al., 2022; Zhao et al., 2023). In our study, we aimed to carefully standardize and structure each phase of evaluating the combined effects of phages and antibiotics on K. pneumoniae.

In selecting K. pneumoniae strains, we focused on isolates that were likely to require combination therapy for eradication. We included multidrug-resistant strains with reduced susceptibility to phages, specifically with capsular type KL23 and clinically significant sequence types ST39 and ST1869 (Table 2). The KL23 capsular type is found in approximately 4% of the general population but accounts for 9–17% of carbapenemase-producing strains (Satlin et al., 2017; Sonda et al., 2018). ST39 is associated with several outbreaks of carbapenem-resistant K. pneumoniae in Russia and Greece (Kuzina et al., 2023; Tryfinopoulou et al., 2023). ST1869, although rare in clinical practice, exhibits carbapenem resistance (Li et al., 2019) and differs by only one allele from ST11, a prevalent sequence type in China known for high rates of multidrug resistance and hypervirulence (Gu et al., 2018).

To maximize the diversity of observed effects, we selected phages from different taxonomic families: Dlv622 (Autographiviridae, Slopekvirinae, Drulisvirus), Seu621 (Vequintavirinae, Mydovirus), and FRZ284 (Straboviridae, Tevenvirinae, Jiaodavirus). The lytic phage FRZ284, capable only of lysis from without, was included to demonstrate that even phages unlikely to be chosen for therapy based on spot tests can exhibit synergy with antibiotics. This phage lysed K. pneumoniae strains regardless of capsular type and has a receptor-binding protein of unknown specificity (Gorodnichev et al., 2021b).

The phages Dlv622 and Seu621 possess homologous receptor-binding proteins (query coverage 98%, identity 65.94%), characterized as polysaccharide depolymerases (Gorodnichev et al., 2021a). One of these proteins, Dep622, the receptor-binding protein of phage Dlv622, was used in the current study. Since depolymerase itself does not lyse bacterial cells, investigating patterns in the combined effects of depolymerases and antibiotics offers a novel perspective on the therapeutic application of depolymerases.

We also selected antibiotics from different classes (aminoglycosides, fluoroquinolones, carbapenems, and phenicols) to maximize the diversity of observed effects. These antibiotics are widely used in clinical practice: meropenem is generally considered a first-line treatment for serious K. pneumoniae infections, while gentamicin is frequently used as part of combination therapy for severe infections. The use of chloramphenicol and levofloxacin is typically guided by susceptibility testing results to ensure their effectiveness against resistant strains (CLSI, 2021).

The concentrations of antimicrobial agents used in phage-antibiotic combination studies are pivotal for interpreting results and assessing clinical relevance. Most studies on phage-antibiotic interactions employ sublethal doses of antimicrobials, which may be above or below therapeutic doses depending on the bacterial strain’s susceptibility. In this study, we based antibiotic concentrations on C_max—the maximum serum concentration of an antibiotic reached after a single dose (Table 1). We believe this approach enhances clinical relevance, providing a more clinically relevant evaluation of combination efficacy. While achieving an equivalent concentration for phages would be ideal, this is likely unfeasible due to phage replication at infection sites.

To assess and compare combined antimicrobial effects, we used several established methods: the t-test, Bliss independence model, two-way ANOVA, and checkerboard assay (Liu et al., 2020; Eller et al., 2021; Gordillo Altamirano et al., 2022; Nikolic et al., 2022; Xu et al., 2022). Among these, the t-test, commonly applied in many studies, only compares the combined effect of agents with the effect of the most effective agent alone. This approach does not account for the contribution of the second agent, thus limiting its ability to determine the overall interaction effect. To address this limitation, we refined the criteria for t-test interpretation, defining conditions for synergy, antagonism, and additive effects. In specific cases where the individual effects of both agents were statistically significantly different from the control and the inhibitory effect of their combination was significantly stronger, we introduced the term “tentative synergy.” This term reflects a potential positive interaction, yet the exact contribution of the second agent remains uncertain, precluding a definitive classification as synergy.

Another frequently used method is the Bliss independence model, which assumes that two agents act independently through stochastic processes (Eller et al., 2021). Given that antibiotics and bacteriophages target bacteria via distinct biochemical mechanisms, the Bliss model is applicable for analyzing their combined effects. However, a notable drawback of this method is its high sensitivity—it does not account for the absence of an effect. Consequently, if the effects of both agents and their combination do not significantly differ from the control, the Bliss model will classify the interaction as additive, even when a true interaction might be absent.

Another commonly used approach for assessing interactions is two-way ANOVA, which specifically evaluates the interaction term, making it relevant for distinguishing between synergy and antagonism. However, the interaction p-value alone does not provide information on the direction of the effect. This interpretation can be inferred from the slope of interaction plots (Liu et al., 2020), though standard error may limit the accuracy of this approach.

Our results indicate that the t-test and Bliss models identified the highest number of statistically significant effects, likely due to the lack of adjustments for multiple comparisons, allowing for the detection of less pronounced synergy and antagonism effects (Figure 2). In contrast, the more conservative two-way ANOVA, which requires a strong and consistent effect to reach statistical significance, identified considerably fewer effects.

The broad range of results highlights the need to select methods based on specific research objectives. For selecting potential phage-antibiotic combinations for therapeutic application, using the t-test or Bliss (or a combination of these) may be optimal due to their simplicity in calculation and interpretation. Additionally, for therapeutic purposes, identifying potential antagonistic interactions can be more critical than seeking combinations with the highest synergy. In turn, two-way ANOVA could be an effective tool for exploratory research, particularly when the objective is to pinpoint combinations with strong interactions for deeper investigation into bacterial response.

The checkerboard assay, with its evaluation of interactions across a broad concentration range, has unique advantages. While this approach appears more informative, the FICI values calculated for two out of three cases conflicted with the effects observed in t-test, Bliss, and two-way ANOVA analyses. Notably, the checkerboard assay operates at antimicrobial agent concentrations unlikely to be achieved in real therapeutic scenarios. However, checkerboard results, when supplemented with Bliss analysis, can illustrate how the combined effect varies with agent concentrations. For example, when using linezolid near C_max (4–8 μg/mL), the combinations G4-17 + Seu621 + LVX and B536-2-17 + Dlv622 + LVX consistently demonstrate synergy at any phage concentration, while a similar effect for G4-17 + Dlv622 + LVX is observed only at the highest phage concentrations. Such observations may be valuable for understanding the potential variability in phage titers at infection sites (Schooley et al., 2017).

Overall, while these methods cannot be directly compared due to differences in statistical assumptions and sensitivity, their complementary use provides a more robust assessment of phage-antibiotic interactions. The consistency of results across multiple methods strengthens confidence in observed effects, whereas discrepancies highlight potential context-dependent factors such as dose–response relationships. These findings emphasize the importance of selecting the appropriate analytical framework based on the specific goals of a study rather than relying on a single method.

In total, 35 of the 48 phage-antibiotic combinations tested demonstrated statistically significant effects based on at least one method (Figure 4). Two previously described cases (9faiz + FRZ284 + CMP and 9faiz + FRZ284 + GEN) were excluded due to difficulties in accurately interpreting their effects; these combinations were thus classified as statistically insignificant.

Phage and depolymerase combinations with levofloxacin exhibited the highest proportion of synergistic effects. However, only the combinations of levofloxacin with Seu621 phage or Dep622 depolymerase consistently produced effects across all three K. pneumoniae strains tested. Previous studies have shown synergy between fluoroquinolone antibiotics and phages in disrupting K. pneumoniae biofilms, specifically with a ciprofloxacin and T7-like phage combination (Verma et al., 2009); however, this is the first report of synergy involving a Mydovirus phage or depolymerase. Nevertheless, expanding the diversity of bacterial strains tested may impact the consistent synergistic interaction observed between levofloxacin and either Seu621 phage or Dep622 depolymerase. In fact, as shown in the checkerboard assay, certain effects may reverse depending on the antimicrobial concentration (Figure 3).

Gentamicin ranked second in terms of the proportion of synergistic effects observed. Although synergistic effects between aminoglycoside antibiotics and K. pneumoniae phages have been reported in animal models (Wang et al., 2021; Zhao et al., 2023), numerous studies suggest that aminoglycosides may inhibit phages (Schindler, 1964; Bowman, 1967; Jiang et al., 2020; Kever et al., 2022). In our study, all phage and depolymerase combinations with gentamicin showed trends toward synergy or additive effects, with antagonism detected in only one case.

Phage combinations with meropenem and chloramphenicol yielded the highest proportion of statistically insignificant results and a higher frequency of antagonistic effects. Beta-lactam antibiotics, particularly meropenem, are frequently explored in studies examining phage synergy. For example, meropenem combined with a K. pneumoniae Webervirus phage showed synergy in biofilm degradation (Xu et al., 2022), while amoxicillin reduced biofilm biomass when combined with an unclassified phage (Bedi et al., 2009). Synergy between K. pneumoniae phages and chloramphenicol has not been previously documented; however, similar effects have been observed with a P. aeruginosa N4-like phage KPP21 (Uchiyama et al., 2018) and Inovirus phages (Hagens et al., 2006).

Notably, combinations involving the FRZ284 phage and the Dep622 depolymerase revealed diverse effects. In plating assays, phage FRZ284 only displayed lysis from without, indicating that while it could attach to bacteria, it could not proliferate within the strain. Despite this, cases of synergy were observed with three out of four antibiotics (gentamicin, levofloxacin, and meropenem), alongside antagonistic effects with all four antibiotics. The Dep622 depolymerase itself did not directly kill bacteria but sensitized them to antimicrobial agents. In this context, synergistic effects with antibiotics were expected, as demonstrated with gentamicin, levofloxacin, and meropenem. However, the antagonistic effect observed with chloramphenicol on strain 9faiz—validated across all testing methods—remains challenging to interpret.

Combinations of phages Dlv622 or FRZ284 with meropenem, as well as FRZ284 with gentamicin, demonstrated all three types of interactions (synergy, additive effect, and antagonism) across the three K. pneumoniae strains. This unpredictability stems not only from methodological inconsistencies but also from the genetic diversity of bacterial strains and the variability of phage lifecycles, which significantly influence outcomes.

The numerous studies have described the mechanisms underlying synergy and antagonism in phage-antibiotic interactions (Liu et al., 2020; Zuo et al., 2021; Bulssico et al., 2023; Luo et al., 2023; Pons et al., 2023; Kunz Coyne et al., 2024; Qin et al., 2024; Yu et al., 2025). However, it is important to recognize that these mechanisms have been described for specific bacteria-phage-antibiotic combinations, and given the vast diversity of bacterial strains and phages, they may not represent the only possible explanations. The variability in effects observed in our study further supports the idea that multiple, context-dependent factors influence phage-antibiotic interactions. Given this complexity, predicting the clinical applicability of specific phage-antibiotic combinations requires careful experimental validation, but their translation into clinical settings remains uncertain without standardized experimental models.

To advance the field, standardizing experimental approaches for assessing combination effects is crucial. The goal of our study is to initiate this discussion by demonstrating how methodological variability can lead to inconsistent interpretations.

Ultimately, the development of effective phage-antibiotic therapies requires a shift from studying isolated mechanisms to a systematic, standardized framework that accounts for the complexity of microbial interactions. Establishing such approaches will improve reproducibility and enable the rational design of phage-based treatments in clinical practice.

The combined application of phages and antibiotics in therapy requires reliable methods to assess their interactions; however, the lack of standardized approaches complicates data interpretation and clinical implementation. In this study, we applied the most commonly used methods (t-test, Bliss method, two-way ANOVA, and checkerboard assay), complemented by the standardization of antibiotic concentrations at C_max. The t-test proved valuable for rapid interaction assessment, especially after expanding the criteria and introducing “tentative synergy,” allowing consideration of complex cases where the contribution of the second agent is less apparent. The Bliss method aligned well with the t-test results and was particularly effective when combined with checkerboard assay data, enabling visualization of the interaction effect based on agent concentration. In contrast, the FICI was less informative and poorly correlated with other methods. Two-way ANOVA identified only the most consistent and substantial effects, highlighting strong interactions and making it useful for comprehensive investigations.

Overall, our approach identified only two combinations with consistent effects across all three K. pneumoniae strains, suggesting that synergy or antagonism is not solely determined by the nature of phage or antibiotic but depends on a complex interplay of factors. These findings emphasize the need for standardized interaction assessment methods to improve comparability of data and enhance the efficacy of combination therapies.