Phage Therapy Against Multidrug-Resistant Serratia Remains an Investigational Strategy

Multidrug-resistant and carbapenem-resistant Serratia marcescens infections are difficult to treat because the bacterium combines intrinsic and acquired antimicrobial-resistance mechanisms with an ability to persist in biofilms. A 2026 narrative review concludes that bacteriophage therapy is biologically plausible for this opportunistic pathogen and may offer strain-specific killing, including activity against biofilm-associated bacteria, but the evidence remains largely preclinical and does not establish clinical efficacy. The review identifies only one published human therapeutic application specifically directed at S. marcescens, alongside a small number of experimental infection studies, making precision phage therapy an investigational and potentially complementary approach rather than an established replacement for antibiotics.

The clinical problem is especially consequential in intensive care, neonatal units and patients who are immunocompromised, hospitalized for prolonged periods or undergoing invasive procedures. S. marcescens is associated with ventilator-associated pneumonia, catheter-related bloodstream infections, complicated urinary tract infections, postoperative wound infections and empyema. Genomic surveillance across 35 countries has identified 12 phylogenetic lineages, with Sm5, Sm6 and Sm7 showing strong associations with human infection and enrichment for high-risk antimicrobial-resistance genes. The organism’s chromosomal ampC gene contributes to intrinsic resistance to several early-generation beta-lactams, although clinically significant AmpC overexpression is uncommon, reported in fewer than approximately 5% of isolates in the evidence summarized by the review. Mobile elements carrying carbapenemase and extended-spectrum beta-lactamase genes have nevertheless reduced the value of many standard beta-lactams, including carbapenems.

The resulting treatment gap has encouraged renewed interest in phages, viruses that infect and lyse bacteria. Their narrow host specificity can distinguish a clinical isolate from much of the surrounding microbiota, while replication at an infection site may increase local exposure. Some anti-Serratia phages also encode depolymerases that degrade capsular or extracellular polymeric material, potentially exposing bacterial receptors and improving access to cells embedded in biofilms. These properties are attractive in principle, but they are not interchangeable with demonstrated human benefit. A retrospective analysis of 242 infected patients reported 28-day all-cause mortality of 50.0% in those with multidrug-resistant isolates, yet the review found insufficient clinical data from which to calculate response rates, comparative efficacy, optimal dosing or reliable treatment durations for phage therapy.

Phage activity begins with recognition of a specific bacterial receptor. In S. marcescens, lipopolysaccharide and its O-antigen are experimentally established receptors for several phages, including kappa, FSB3, FSB55 and phiIF3. Biochemical work with phage kappa identified LPS as its primary receptor, while mutants defective in the LPS O-side chain became resistant to several LPS-dependent phages. In strain Db11, disruption of an O-antigen transporter impaired both productive infection and adsorption by phiIF3; only two of 25 additional Db11 phages retained the ability to infect that mutant. The flagellum provides a mechanistically different route of entry. Deletion of fliC, which encodes flagellin, made S. marcescens ATCC 274 resistant to the flagellotropic phage chi. These observations explain why a phage active against one isolate may fail against another: strain-specific O-antigens, capsule phenotypes and flagellar expression can change receptor availability.

Resistance to phages can also arise after adsorption. S. marcescens strains carry diverse restriction–modification systems, CRISPR–Cas arrays, cyclic-oligonucleotide-based antiphage signaling systems and abortive-infection-associated defenses, although none is present or functionally equivalent in every isolate. A type II restriction–modification system known as SmaI has been characterized, while genomic surveys have found lineage-dependent differences in the wider restriction–modification repertoire. CRISPR systems have been detected in only a subset of strains, and some spacers resemble sequences from Serratia phages, indicating historical interaction without proving current antiviral activity. An S. marcescens CBASS system uses a sequential “cGASylation” mechanism to activate a cyclic antiviral signal and a cell-death-associated response. Abortive infection can also halt phage propagation at the expense of the infected cell. Because the bacterial defense repertoire varies by strain and phages may evolve countermeasures, the genomic presence of a defense locus alone cannot predict susceptibility to a particular therapeutic phage.

Preclinical development is less mature for S. marcescens than for several other multidrug-resistant Gram-negative pathogens. Most studies concern individual phages or small isolate panels, and relatively few evaluate mature biofilms, representative clinical collections, animal infection models or clinically relevant phage–antibiotic combinations. Candidate viruses are typically enriched from sewage, soil, hospital wastewater or clinical material, plaque-purified and characterized phenotypically. Clearing in a spot assay does not prove productive infection, so apparent activity must be confirmed through plaque formation, efficiency-of-plating measurements, viable-count endpoints and evidence of progeny production. Host-range panels should represent contemporary clinical diversity in resistance profile, lineage, O-antigen, capsule, biofilm behavior, infection source, geography and time; the review notes that no international guideline establishes a universal minimum number of isolates. Complete genome sequencing is required to exclude lysogeny, toxin, virulence, antimicrobial-resistance and other undesirable determinants, alongside functional confirmation of a strictly lytic lifecycle.

Biofilm-associated activity is a distinctive target for further testing. The podovirus UFV01, isolated using a clinical S. marcescens strain, carries a tail-fiber-associated depolymerase that degraded extracellular polymeric material and reduced preformed Escherichia coli biofilms without productively infecting the cells. This “bystander” effect demonstrates matrix degradation rather than phage-mediated killing, and it does not establish therapeutic efficacy. A phage such as WSPA, recovered with a multidrug-resistant clinical isolate as enrichment host, showed lytic activity against a panel of contemporary multidrug-resistant strains and no identified virulence or resistance genes, but its tested panel was limited and it had no in vivo or clinical validation. Depolymerase activity, plaque morphology and broad host range are therefore useful screening features, but they cannot substitute for testing under physiological conditions and in infection models.

Direct human evidence remains exceptionally sparse. The review identified a 2025 case report of a 59-year-old patient with persistent postoperative pneumonia and empyema after lung-cancer resection. Pleural fluid remained culture-positive for an extensively drug-resistant S. marcescens strain after 120 days of unsuccessful antibiotic treatment. The strictly lytic phage Spe5P4 was delivered intrapleurally and by nebulization at a reported concentration of 10⁹ PFU/mL, together with meropenem and amikacin, drainage and supportive care. The patient’s dyspnea and chest pain resolved within 72 hours, oxygen saturation normalized, pleural fluid measurements declined and imaging showed lung re-expansion with near-complete resolution of effusion by day 140. No phage-related adverse events were reported, and post-treatment isolates showed reduced fluoroquinolone resistance. However, the uncontrolled single-case design and concurrent antibiotics, drainage and supportive care prevent attribution of improvement to Spe5P4 alone; the report establishes preliminary feasibility in one patient, not efficacy.

Organism-specific preclinical findings are stronger than the clinical record but remain narrow. In a lethal BALB/c mouse model of S. marcescens infection, immediate administration of phage phiSM9-3Y produced 100% survival, while a 60-minute delay reduced protection to approximately 60% even with a higher phage dose. In a Galleria mellonella model infected with the clinical bloodstream isolate UMH9, a single 10⁷-PFU dose of Sm10b_1 increased six-day survival from 0% to 67% relative to untreated larvae. These models demonstrate in vivo antibacterial activity against S. marcescens, but neither reproduces the full immune, anatomical, physiological and comorbidity complexity of human infection. Phage–antibiotic combinations also cannot be assumed to be synergistic. Some antibiotics can inhibit phage replication by suppressing bacterial metabolism, whereas other combinations enhance killing or biofilm disruption. Each phage, isolate, antibiotic, concentration and treatment sequence therefore requires direct testing, with both bacterial killing and viable phage titers measured.

Translation is additionally constrained by variable pharmacokinetics, bacterial resistance, immune neutralization, manufacturing and regulation. Administered PFU does not reliably predict the active concentration at an infection site because distribution, adsorption, in situ amplification, local clearance and bacterial susceptibility all affect exposure. Local delivery may concentrate phages in pleural or pulmonary compartments, whereas systemic administration can expose them to sequestration and neutralizing immune responses; the clinical significance of anti-phage antibodies remains uncertain. Personalized preparations also require consistent identity, titer, purity, sterility, endotoxin control, potency and storage stability, while cocktails require assessment of component balance, stability and interference. The review’s priorities are standardized isolate-specific susceptibility testing, receptor-informed and diverse phage libraries, quantitative cocktail design, pathogen-specific pharmacokinetic and pharmacodynamic studies, serial monitoring of bacteria and phages, and prospective controlled trials with antibiotic-only and combination comparators. Until those steps are completed, the evidence supports biological plausibility and limited proof of concept, but not reproducible efficacy or generalizable clinical feasibility for multidrug-resistant S. marcescens infections.

Subject of Research: Evaluation of bacteriophage therapy for multidrug-resistant Serratia marcescens infections

Article Title: Phage therapy for multidrug-resistant Serratia marcescens infection: mechanisms, preclinical evidence, and translational challenges

Article References: Zhang, S., Tao, H., Zhao, S., & Wei, Y. (2026). Phage therapy for multidrug-resistant Serratia marcescens infection: mechanisms, preclinical evidence, and translational challenges. Frontiers in Microbiology, 17, Article 1959850. https://doi.org/10.3389/fmicb.2026.1959850

Image Credits: AI Generated

DOI: 10.3389/fmicb.2026.1959850

Keywords: bacteriophage therapy, Serratia marcescens, multidrug resistance, carbapenem resistance, biofilms, phage receptors, phage-antibiotic synergy, clinical evidence, antiphage defenses, translational research, Phage, therapy