Personalized Phage Therapy in an ICU Patient with Polymicrobial Pulmonary Infections: a case from a single-arm trial

Abstract

Ventilator-associated infections (VAIs) caused by multidrug-resistant (MDR) pathogens poses a significant challenge in intensive care units (ICUs). Phage therapy has emerged as a promising alternative for VAIs, though its efficacy depends on multiple factors. We report a 40-year-old male ICU patient with sequential pulmonary infections of MDR Acinetobacter baumannii, Klebsiella pneumoniae, and Stenotrophomonas maltophilia. Seven rounds of tailored phage therapy—targeting single or dual pathogens—were administered adjunctive to antibiotics. A. baumannii was complete eradication, but K. pneumoniae and S. maltophilia showed cyclical clearance and recurrence due to strain replacement and phage resistance. Phage therapy correlated with reduced bacterial load, fever, procalcitonin, and IL-6 levels, alongside improved pulmonary inflammation, while white blood cell counts and C-reactive protein showed limited response. No severe adverse events occurred. Dynamic shifts in dominant pathogens aligned with clinical progression, with phage therapy providing transient control, extending the window for intervention and reducing high-level antibiotic use. After six months, the patient was weaned from ventilation and discharged. This case underscores the complexity of VAI management and highlights the potential of personalized phage therapy to mitigate MDR infections and minimize antibiotic reliance. This case was enrolled in a single arm trial (ChiCTR2000036801).

This case report presents a patient with sequential pulmonary infections of multidrug resistant (MDR) Acinetobacter baumannii, Klebsiella pneumoniae, and Stenotrophomonas maltophilia, treated with tailored phage therapy adjunctive to antibiotics.

Introduction

Ventilator-associated infections (VAIs) are common ICU-acquired infections which significantly prolong ventilation and hospital stays, increase costs, and reduce recovery rates. Antibiotic use in VAI patients accounts for nearly half of all ICU antibiotic consumption, fueling the emergence of multidrug-resistant (MDR) pathogens^1,2. With the widespread use of carbapenem antibiotics, carbapenem-resistant pathogens such as Acinetobacter baumannii (CRAB) and Klebsiella pneumoniae (CRKP) have become among the most prevalent MDR pathogens in the ICU. Furthermore, pathogens intrinsically resistant to carbapenems, such as Stenotrophomonas maltophilia (SMA), are being detected with increasing frequency in the ICU. These nosocomialorganisms leave clinicians with very limited and often ineffective antibiotic choices, presenting a major challenge in the management of critically ill patients^1,2.

Bacteriophages (phages), viruses of bacteria, offer a promising alternative against MDR infections. Clinical evidence supports their safety and efficacy in pulmonary infections; however, efficacy varies significantly by infection type³. In acute infections, the target bacteria are more homogeneous and less host-adapted, rendering them susceptible to phage clearance, though patients remain prone to reinfection if symptoms do not improve^2–6. In chronic infections, bacteria are more heterogeneous; phage therapy often rapidly reduces the dominant strain and yields symptomatic relief, but rarely achieves complete eradication^7–10.

This study reports the phage therapy applied to a critically ill ICU patient with VAIs. The patient developed successive pulmonary infections caused by three MDR bacterial species, for each of which personalized phage cocktails were selected and administered. Alongside concurrent antibiotic therapy, the phage treatment effectively reduced the bacterial load in the lungs and alleviated clinical symptoms, thereby creating a critical window that facilitated the patient’s recovery. This case represents an instance of adjunctive phage therapy concurrently targeting three distinct MDR organisms in a single patient, highlighting both the potential and challenges of phage therapy in treating severe infections.

Results

Patient history

On March 14, 2023, a 40-year-old male patient with a five-day history of unexplained fever was admitted to our Center. He was diagnosed with viral encephalitis and concurrent central nervous system (CNS) demyelination based on cerebrospinal fluid analysis and cranial CT findings. The patient’s condition deteriorated rapidly, progressing to coma complicated by aspiration pneumonia and respiratory failure, which necessitated transfer to the ICU on March 20. Invasive mechanical ventilation was initiated immediately, followed by tracheostomy two days later. Seven days after ICU admission, the patient’s pneumonia progressed and a CRAB—susceptible only to polymyxin B (PMB), tigecycline (TGC), and trimethoprim/sulfamethoxazole (SMZ/TMP)—was first isolated from respiratory samples (Supplementary Data 1). Despite empirical combination therapy with PMB and cefoperazone/sulbactam (CPZ/SUL), the patient continued to exhibit persistent fever and elevated IL-6 levels (Supplementary Data 2), with a Clinical Pulmonary Infection Score (CPIS) of 7 (Table S1). In combination with the clinical findings, these indicators supported a diagnosis of VAI, prompting a temporary switch to PMB plus TGC based on susceptibility profile (Fig. 3a, Supplementary Data 1).

Fig. 3. Susceptibility profiles of target pathogens to therapeutic phages and antibiotics.

a-c: Phage and antibiotic susceptibility profiles of representative clinical isolates: (a) A. baumannii, (b) S. maltophilia, (c) K. pneumoniae, for D62/D69, the D62 isolate was used for phage susceptibility testing and the D69 isolate was used for antibiotic susceptibility testing. Color codes indicate susceptibility: green, susceptible; yellow, intermediate; red, resistant. Blue rectangles (labeled with dates) indicate the timing and formulation of phage administrations; light-blue rectangles (labeled with “√” and antibiotic names) represent combination antibiotic regimens used during pathogen-targeted phage therapy. All timelines are relative to Day 0 (D0), the start of phage therapy. (d-f) Plaque formation efficiency (EOP) of representative clinical isolates: (d) A. baumannii, (e) S. maltophilia, (f) K. pneumoniae. The central spots observed within plaques resulted from agar puncture during sample application; the white grid lines visible in the images are background artifacts from the imaging device. CRAB carbapenem-resistant A. baumannii, SMA S. maltophilia, CRKP, carbapenem-resistant K. pneumoniae, WT wild-type, PMB polymyxin B, CPZ/SUL cefoperazone/sulbactam, TGC tigecycline, SMZ/TMP trimethoprim/sulfamethoxazole, MNC minocycline, OMC Omadacycline, AMK amikacin, LEV levofloxacin, CAZ ceftazidime, CAZ/AVI ceftazidime-avibactam, MEM Meropenem, MLST Multilocus-sequence typing, CPS capsular polysaccharide, ΔCPS loss of CPS, CPSmut mutation of CPS, NA not available.

Concurrently, informed consent was obtained for adjunctive phage therapy targeting the predominant strain cultured on March 30.

Phage therapy

In addition to the initially identified CRAB, SMA, and CRKP were subsequently detected in the patient’s respiratory samples, either alternatively or in combination. Based on the dynamic shifts in bacterial pathogens and corresponding antibiotic treatments, the seven courses of phage therapy in this case can be divided into four episodes. To facilitate a clear timeline description, the first administration of phage therapy is designated as Day 0 (April 6). Unless otherwise specified, all subsequent time points in this article are relative to this baseline (Fig. 1).

Fig. 1. Clinical summary from time of initial infection through day 75 of phage treatment.

The x-axis represents the timeline starting from the first detection of CRAB. The first day of phage therapy is designated as Day 0 (D0). The longitudinal panels display oxygen support status (blue), antibiotic administration (gray), and bacterial detection results—color-coded from green to red based on semi-quantitative (semi-q) bacterial burden assessed via streak plate method. Original strains (Ori strain) and phage-resistant strains (Res strain) are differentiated by white and yellow partition lines, respectively. Seven courses of phage therapy were administered via inhalation on the indicated dates; all bacterial data shown were isolated from respiratory samples. Laboratory parameters (Tm, PCT, and IL-6) are indicated by black solid dots. Source data are provided as a Source Data file. Tm body temperature, PCT procalcitonin, IL-6 interleukin-6, CRAB carbapenem-resistant A. baumannii, SMA S. maltophilia, CRKP carbapenem-resistant K. pneumoniae, PMB polymyxin B, CPZ/SUL cefoperazone/sulbactam, TGC tigecycline, SMZ/TMP trimethoprim/sulfamethoxazole, OMC omadacycline, AMK amikacin, CAZ/AVI ceftazidime-avibactam, MEM Meropenem. qd once daily, bid twice daily, tid three times daily, qid four times daily. All antibiotics were administered intravenously.

Episode 1 (day 0 – day 17)

Given that the adjusted antibiotic regimen failed to control the infection, the therapy was switched back from PMB plus TGC to PMB plus CPZ/SUL prior to initiating the first phage therapy course, aiming to broaden coverage against common Gram-negative pathogens in the ICU. A two-phage cocktail targeting CRAB, Cocktail-1 (containing CP-p-AB-022 and CP-p-AB-036; 5 mL per vial with titers of 1 × 10⁹ and 5 × 10⁸ PFU/mL, respectively), was selected and administered via nebulization at 10 mL per dose, twice daily (Course-1). Target CRAB was eliminated from sputum culture starting from the first day after treatment (Fig. 1). However, the patient’s pulmonary inflammatory conditions worsened again after a transient improvement (Fig. 2, C1), with subsequent emergence of CRKP and SMA in respiratory samples. As both bacteria met the inclusion criteria for phage therapy, phage susceptibility testing (PST) was immediately initiated using samples from Day 4. No lytic phage was identified from our existing phagebank for CRKP, prompting de novo screening. For SMA, a single-phage CP-p-PMA-011 (5 mL a vial, 1 × 10⁸ PFU/mL) was selected and administered twice daily via nebulization during the Course-2 treatment. SMA tested negative on days 2, 3, and 5 after Course-2 phage therapy but reappeared on day 7. Meanwhile, CRKP continued to be detected (Fig. 1). Radiographic examination on day 8 after Course-2 (Day 17) showed significant improvement in right lung inflammation, though slight progression was noted in the left lung (Fig. 2, C2).

Fig. 2. Radiologic changes before and after phage therapy.

Serial chest radiographs on indicated days before and after phage therapy (labeled C1 to C5&6, means Course-1 to Course-6). Density, profusion, and confluence of bilateral fluffy opacities are positive correlated with the severity of pneumonia. The first administration day of phage therapy is designated as Day 0 (D0).

Episode 2 (day 18 – day 31)

Following 2 courses of phage therapy in Episode 1, CRAB was eradicated, while SMA recurred after initial clearance. Despite continuous use of PMB (for a total of 27 days since ICU admission), CRKP persisted—susceptible to PMB, amikacin (AMK), and ceftazidime-avibactam (CAZ/AVI), and resistant to TGC and CPZ/SUL (Fig. 3c, Supplementary Data 1). Therefore, beginning on Day 18, the antibiotic regimen was adjusted to CAZ/AVI and AMK targeting CRKP, combined with omadacycline (OMC) for SMA (susceptible to minocycline, MNC) (Figs. 1, 3b). On Day 23, a newly isolated phage cocktail targeting CRKP, Cocktail-3 (ΦKP-D4, ΦKP-D5, 5 mL/vial, 3 × 10⁸ and 2 × 10⁹ PFU/mL, respectively), was administered via nebulization twice daily for two consecutive days (Course-3). Although sputum culture turned negative for CRKP one day before this course, the bacterium was detected only once on day 10 within two weeks post Course-3 (Fig. 1). After Course-3 phage therapy, the patient’s pulmonary imaging showed marked bilateral improvement (Fig. 2, C3). Procalcitonin (PCT) and IL-6 levels gradually returned to normal (Fig. 1). During this period, SMA continued to be detected, and CRAB reemerged, prompting preparation for the next course of phage therapy.

Episode 3 (day 32-day 63)

From Day 32, CAZ/AVI (targeting CRKP, which had turned negative) was discontinued, and OMC for SMA was switched to SMZ/TMP. TGC and CPZ/SUL were added to target CRAB (Fig. 1). Meanwhile, PST was performed on CRAB and SMA isolates from Day 28. Results showed that the phagebank susceptibility profile (PSP) of SMA remained consistent with that prior to Course-2. Two CRAB strains were identified: one with the same PSP as before Course-1, and another that had developed resistance to CP-p-AB-022 in Cocktail-1 but remained sensitive to CP-p-AB-036 and newly sensitive to CP-p-AB-011 from the phagebank. Accordingly, a four-phage Cocktail-4 was formulated, comprising CP-p-PMA-011 used in Course-2 for SMA, Cocktail-1 for the original CRAB strain, and CP-p-AB-011 (5 mL/vial, 2 × 10⁹ PFU/mL) for the resistant CRAB strain (Fig. 3). 20 mL of Cocktail-4 was administered twice daily via nebulization in Course-4. CRAB was negative one day before treatment and was detected only on day 3 post Course-4 (resistant to Cocktail-1 but susceptible to CP-p-AB-011), with no further detection throughout the study and follow-up. Although the bacterial load of SMA decreased after phage therapy, it remained detectable. On day 3 after Course-4, CRKP reemerged (Fig. 1). PSP and genome sequencing revealed that both SMA and CRKP represented new strain types: SMA shifted from ST309 to ST31, and CRKP changed from KL19/ST15 to the clinically common KL64/ST11. Phage cocktails targeting these new strains were prepared: Cocktail-5 (vB_SmeS_BUCT700, vB_SmeS_BUCT703, 5 mL/vial, 2 × 10⁸ and 3 × 10⁸ PFU/mL, respectively) for SMA and Cocktail-6 (CP-p-Kpn-005, CP-p-Kpn-065, CP-p-Kpn-107, 5 mL/vial, 2 × 10⁸, 4 × 10⁸ and 2 × 10⁹ PFU/mL, respectively) for CRKP. These were administered on Day 47 (Course-5, 10 mL) and Day 50 (Course-6, 15 mL), respectively, twice daily. On Day 53 (6 days after Course-5), SMA reverted to ST309 and was not detected on days 9 and 11 post Course-5. On Day 53 (3 days after Course-6), CRKP developed resistance mutations against the therapeutic phages and continued to be detected thereafter (Figs. 1,  3c, f).

After Course-4, pulmonary imaging improved significantly but showed progression again before Course-5 and −6, with slight improvement post-treatment (Fig. 2, C4, C5&6). The patient gradually began intermittent weaning from mechanical ventilation and underwent bedside respiratory exercises. During weaning, high-flow nasal oxygen therapy was administered with satisfactory oxygenation.

Episode 4 (day 64-day 75)

Due to severe CNS demyelination, the patient experienced respiratory muscle fatigue during weaning attempts. To reduce oxygen consumption, mechanical ventilation was resumed on Day 64. After phage therapy in the previous episode, CRAB was cleared, SMA switched strain types twice with transient clearance. Antibiotics were adjusted to CAZ/AVI plus meropenem, aiming to better address the persistent CRKP infection. A new cocktail targeting resistant CRKP, Cocktail-7 (CP-p-Kpn-142, CP-p-Kpn-145, 5 mL/vial, 3 × 10⁷ and 3 × 10⁸ PFU/mL, respectively), was administered twice daily via nebulization for two days (Course-7). Although bacterial load decreased post-treatment, it was not eradicated; the strain developed resistance to CP-p-Kpn-142 and intermediate susceptibility to CP-p-Kpn-145 (opaque plaques). Re-sequencing of SMA from Day 69 showed a return to ST31. By this time, the patient’s infection parameters had improved, but limb muscle strength and respiratory function remained poor. The family transferred the patient to another hospital for rehabilitation on Day 75 (June 20) upon securing a bed. Notably, SMA cultures turned negative on the day prior to discharge.

Clinical outcomes

To comprehensively evaluate changes in clinical symptoms, we analyzed serial bedside chest X-rays (Fig. 2) and laboratory parameters (Fig. 4) before and after each course of phage therapy. After Course-1, although target CRAB was eradicated, the patient acquired CRKP and SMA infections, so pulmonary imaging showed no significant change. After Course-2, CRAB and SMA were cleared, but CRKP persisted. Imaging indicated notable improvement in the right lung with progression in the left. Before Course-3, both lungs showed progression, accompanied by rises in PCT and IL-6; after phage therapy, these markers declined to near-normal levels (Fig. 1), and chest X-rays indicated significant bilateral improvement. Pulmonary imaging reached its best state after Course-4, though progression recurred around Course-5, with improvement after phage therapy. Overall, the patient’s pulmonary imaging fluctuated repeatedly; improvement was observed after all courses except Course-1 (Fig. 2).

Fig. 4. Changes of laboratory indicators before and after phage therapy.

The following parameters were measured at four timepoints before (d–7, d–5, d–3, d–1) and four timepoints after (d1, d3, d5, d7) each course of phage therapy: body temperature (Tm), white blood cell count (WBC), neutrophil count (N), C-reactive protein (CRP), procalcitonin (PCT), and interleukin-6 (IL-6). Data are presented as box plots showing the median (centre line), quartiles (box bounds), and range (whiskers from min to max). Statistical significance was assessed using ordinary one-way ANOVA with Tukey’s multiple comparisons test; n = 7 courses. Source data are provided as a Source Data file.

No adverse events were observed throughout phage therapy. The patient’s liver function parameters, including alanine aminotransferase (ALT) and aspartate aminotransferase (AST), showed no significant changes directly attributable to any individual phage therapy course and exhibited an overall downward trend throughout hospitalization. Body temperature fluctuated considerably during the first two episodes and relatively stabilized during the latter two. Blood PCT and IL-6 levels peaked late in Episode 1 and gradually returned to near-normal from Episode 2 onward (Fig. 1). Analysis of laboratory parameters before and after the seven phage therapy courses showed an overall decreasing trend in body temperature, an initial rise followed by a decline in white blood cells (WBC), neutrophils, PCT, and IL-6, and an initial decline followed by a rise in CRP (Fig. 4, Supplementary Data 2). The PaO₂/FiO₂ values, which reflected changes in O₂ requirements were transiently increased after each treatment episode, indicating an improvement in pulmonary function (Supplementary Data 2). Based on the comprehensive evaluation of the above indicators, we calculated the CPIS for the patient. The CPIS exhibited a gradual downward trend throughout the treatment course (Table S1), indicating progressive improvement in the severity of pneumonia.

After the patient was transferred to another hospital, his pulmonary infection did not worsen. CRKP was still detected in sputum culture on Day 82 (June 27). However, the infection gradually improved in July, and the patient was weaned off mechanical ventilation in August, requiring intermittent oxygen support. The tracheostomy tube was successfully removed in early September. As of June 2025, the patient has not experienced severe pulmonary infection recurrence or required mechanical ventilation or high-flow oxygen therapy. The patient is recovering well with satisfactory respiratory and oxygenation function and continues to receive regular rehabilitation.

Etiological outcomes

The patient developed hospital-acquired CRAB on the seventh day after ICU admission, CRKP on the 16^th day, and SMA on the 19^th day. We performed antibiotic susceptibility testing, PST, and genomic sequencing on various strains of these pathogens. Results revealed high pathogen variability typical of ICU settings. The patient’s CRAB was a clinically prevalent KL3 capsular type, sensitive to CP-p-AB-022 and CP-p-AB-036. It was only susceptible to PMB, TGC, SMZ/TMP, and MNC, and intermediate to levofloxacin (LEV) (Supplementary Data 1). CRAB was not detected for 27 days after the first treatment but reappeared on Day 28, yielding one strain matching the pre-treatment isolate and one resistant strain (D28) (Fig. 3a). Sequencing showed a deletion mutation in the gpi gene within the capsular gene loci of the resistant strain (Fig. 5c). Another resistant strain (D40) emerged after Course-4, with a point mutation in gpi causing a His391Asn substitution (Fig. 5c). This strain was resistant to both phages in Cocktail-1, developed resistance to SMZ/TMP, MNC (both previously susceptible), and LEV (previously intermediate) (Fig. 3a, Supplementary Data 1).

Fig. 5. morphology and genomic alterations of A. baumannii and K. pneumoniae isolates during phage therapy.

Transmission electron microscopy (TEM) images showing capsule structures of representative clinical isolates. a A. baumannii; b K. pneumoniae. Scale bar represents 500 nm. c Genetic mutation profiles of A. baumannii and K. pneumoniae isolates, showing key genomic loci affected and the types of genetic variants identified.

Analysis of SMA revealed two strains during treatment: ST309 (D4, D28, D53), sensitive to CP-p-PMA-011 and resistant to MNC; and ST31 (D39, D69), sensitive to vB_SmeS_BUCT700 and vB_SmeS_BUCT703 and susceptible to MNC. The PSPs of both ST types did not change significantly before and after phage therapy (Fig. 3b). SMA turned negative from the day before discharge.

The initially detected strain of CRKP was KL19/ST15 (D4), susceptible to PMB, AMK, and CAZ/AVI, and resistant to TGC. After Course-3 phage therapy, a new strain with the common KL64 capsular type emerged (D42). After Course-6, the bacterium (D53) developed anti-phage resistance and became susceptible to TGC and SMZ-TMP (both of which were previously resistant). After Course-7, newly resistant strains (D72) showed a PSP consistent with the D53 resistant strain, being resistant to CP-p-Kpn-142 and intermediate to CP-p-Kpn-145. Sequencing analysis revealed deletion mutations in the wcaJ gene within the capsular gene loci in both D53 and D72 anti-phage resistant CRKP strains. After three rounds of phage therapy (Course-3, −6, and −7), CRKP was still present in sputum culture on Day 82 (June 27) at the external hospital, with antibiotic susceptibility profile differing significantly from both KL19/ST15 and KL64/ST11 strains—resistant to TGC and tobramycin, and susceptible only to PMB, AMK, ciprofloxacin, and CAZ/AVI—suggesting a distinct strain (Supplementary Data 1).

Transmission Electron Microscope (TEM) demonstrated that the phage-resistant strains of CRAB (D28, D40) and CRKP (D53) emerging after phage therapy were capsule-deficient, consistent with the disappearance of halo zones in phage plaques (Figs. 5, 3d, f)¹¹. Compared with their wild-type counterparts, these capsule-deficient strains exhibited significantly attenuated virulence in a Galleria mellonella infection model (Fig. S1). For CRAB, survival rates were 73.3% versus 26.7% at 72 hours post-infection (p = 0.0001); for CRKP, survival rates were 33.3% versus 16.7% at 72 hours post-infection (p = 0.0035) (Fig. S1). No consistent pattern of antibiotic susceptibility change was observed among phage-resistant strains in this case (Supplementary Data 1). However, one CRAB isolate (Day 40) showed enhanced resistance to SMZ/TMP, MNC, and LEV, correlating with an adeL mutation (V145G) that may upregulate the AdeFGH efflux pump¹² (Table S2). Among CRKP-KL64 isolates, susceptibility changes varied and did not correlate with capsule loss (e.g., isolates from D52 and D72), suggesting strain heterogeneity rather than a direct consequence of phage resistance.

Discussion

This case represents a typical example of bacterial infection management in an ICU setting and is one of the limited number of cases involving simultaneous phage therapy against multiple VAIs in a single patient^10,13. The patient, critically ill due to underlying conditions, required invasive ventilatory support and sequentially acquired pulmonary infections with CRAB, CRKP, and SMA after ICU admission. Throughout multiple courses of phage therapy, we observed clinical responses, including clearance of target bacteria, strain replacement, and emergence of anti-phage resistance. Before Course-3 and -4 (targeting CRAB), the target bacteria were not detected, making direct assessment of phage efficacy challenging. However, bacteria remained negative after Course-3 for 15 days, when CRKP reappeared persistently; sequencing revealed it was a new KL64/ST11 strain. Given that CRKP persisted with high load despite continuous PMB during Episode 1, phage therapy is inferred to have contributed. This is corroborated by imaging changes around Course-3. Course-4 targeted both CRAB and SMA. CRAB was detected only on day 3 post-treatment; sequencing revealed a phage-resistant mutant with altered PSP, and it was not detected again thereafter, suggesting phage therapy played a key role in its clearance. Although SMA load decreased after Course-4, it remained detectable; analysis of SMA on Day 39 (post-treatment) revealed a new ST31 strain, also indicating a phage-mediated effect.

As in many phage therapy reports, we observed repeated instances of phage resistance, such as the two strain types, each of SMA and CRKP. Additionally, KL3/ST2 CRAB and KL64 CRKP developed resistance mutations post-phage therapy, involving point or deletion mutations in capsular cluster genes (gpi in CRAB, wcaJ in CRKP). These mutations likely cause loss or alteration of the capsule, as supported by disappearance of halo in phage plaques^14,15. Previous studies indicate that gpi deletion in A. baumannii attenuates virulence, promoting rapid clearance in the host^16,17. Similarly, wcaJ deletion in K. pneumoniae facilitates escape from phage killing but increases susceptibility to host innate immunity, reducing overall fitness¹⁸. Mutations in wcaJ lead to a non-mucoid phenotype with enhanced biofilm formation and epithelial adhesion but reduced resistance to macrophage phagocytosis and virulence¹⁹. These mechanisms may partly explain the improvement in pulmonary imaging and eventual clinical recovery after phage therapy.

The loss of the bacterial capsule, which confers phage resistance, may lead to changes in antibiotic susceptibility^17,18. In the present case, a CRAB-resistant strain exhibited marked alterations in its antibiotic susceptibility profile. Comparative genomic analysis of pre- and post-resistance isolates identified mutations in the adeL gene as a potential contributor to these changes¹². However, no such consistent patterns were observed in K. pneumoniae or S. maltophilia, underscoring the complexity of bacterial interactions with phages and antibiotics. Notably, consistent with previous reports, both CRAB and CRKP capsule-deficient mutants displayed attenuated virulence in the G. mellonella infection model, suggesting a fitness cost associated with the acquisition of phage resistance^19,20.

This patient, on prolonged invasive ventilation in the ICU, developed infections with common nosocomial pathogens CRAB, CRKP, and SMA, at times simultaneously. Antibiotic regimens were adjusted four times based on dominant pathogens and antibiotic susceptibility profiles. Concurrently, personalized phage therapy was administered targeting single or multiple bacteria. Unlike the extended regimens reported in some studies, our approach for acute infections employed consistently short courses. These infections typically display high antibiotic resistance, low heterogeneity, and limited host adaptation, making them more susceptible to phage clearance^4,21,22. Consistent with our prior reports^4,21, 1–3 days of therapy is often sufficient to eliminate the target bacterium or select for resistant variants. Extending treatment with the same phages beyond this window is unlikely to provide additional benefit. However, the impact of host immune status on phage therapy remains debated; impaired immunity may slow pathogen clearance yet reduce phage elimination.

Overall, the combination of phage and antibiotic therapy was associated with improvements in bacteriological, radiographic, and inflammatory parameters. However, bacterial recurrence, anti-phage resistance, and emergence of new strains were also observed. The complex and fluctuating pathogen dynamics may stem from both the ease of transmission of resistant bacteria in ICUs^1,2 and the strong selective pressure exerted by phage therapy^3,4. Phage therapy may benefit such patients in three ways: first, by rapidly reducing dominant bacterial load, mitigating infection severity, and buying time for critical care^3,23; second, by enabling use of less toxic antibiotics (e.g., PMB was discontinued after Episode 2); third, by selecting for resistant mutants with attenuated virulence due to fitness costs^20,21,24. Crucially, studies indicate that even without full bacterial clearance, phage therapy can ameliorate pulmonary symptoms by selecting attenuated mutants^21,25.

Several limitations should be acknowledged. This single-patient report limits generalizability, and concurrent antibiotic adjustments may confound assessment of phage-specific effects. Unavailable data—including neutralizing antibody titers and phage quantification in respiratory samples—preclude detailed mechanistic analysis. Furthermore, while aerosol nebulization proved clinically effective with rapid microbiological response in this case, the dosage and frequency used here were based primarily on prior clinical experience and published reports⁴. Systematic preclinical and clinical investigations are warranted to optimize these parameters.

In summary, this case illustrates the potential of personalized phage therapy for managing ventilator-associated pneumonia in ICU patients. Phage therapy was associated with a reduction in the target bacterial load and an improvement in clinical symptoms, without significant adverse effects, thereby contributing to the patient’s eventual recovery. It also highlights the complexity of ICU infections among severely ill patients. As a specific therapeutic with immunogenic and lytic properties, phages cannot persist long-term in the human body. Therefore, establishing a broadly covered phagebank against prevalent bacteria and a rapid personalized phage therapy process is essential to effectively address the rapidly changing bacterial landscape in ICU patients. This case serves as a typical example of simultaneous phage therapy, in conjunction with antibiotics, against A. baumannii, S. maltophilia, and K. pneumoniae. It demonstrates both the promise and challenges of phage therapy in controlling drug-resistant bacterial infections in critically ill patients.

Methods ethical statement

This case was enrolled in a single-arm trial, which was approved by the institutional ethics committee of the Shanghai Public Health Clinical Center (2020-S15701) and registered with the Chinese Clinical Trial Registry (ChiCTR2000036801). The patient’s guardians provided informed consent for phage therapy and for the use of samples and data in research and publication.

Clinical isolates and antibiotic susceptibility testing (AST)

Clinical isolates were obtained from the patient’s sputum or bronchoalveolar lavage fluid samples and identified by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (Bruker Daltonik, Bremen, Germany) at the Shanghai Public Health Clinical Center. Strains were designated by species and isolation date, with Day 0 (D0) defined as the initiation date of phage therapy. Antibiotic susceptibility of clinical isolates was determined using the Vitek 2 automated system (bioMérieux, Rhône, France) or using the broth microdilution method.

Phage susceptibility testing (PST, phage-typing)

The susceptibilities of the target bacteria to each phage in the corresponding phagebank from CreatiPhage were measured by spot assays. All potentially distinct bacterial strains from a single sample, distinguished by colony size, color, or viscosity, were analyzed. The lytic efficacy was assessed by the clarity of plaques, scored as transparent (susceptible), opaque (intermediate), or no plaque (resistant)⁴. The susceptibility profiles of different phages in the phagebank against target bacteria can be used to type the bacteria, which we call the phagebank susceptibility profile (PSP). Since there was no lytic phage in the existing phagebank targeting CRKP-D4, this strain was used as host bacterium to isolate new lytic phages ΦKP-D4, ΦKP-D5 from environmental samples, via culture-dependent methods²¹. Briefly, 5 mL of 2 × LB liquid medium was mixed with an equal volume of sewage sample and 200 μL of overnight bacterial culture. The mixture was incubated with shaking at 200 rpm overnight at 37 °C. Subsequently, the culture was centrifuged at 6000 × g for 5 min at 4 °C, and the supernatant was filtered using a sterile 0.22 µm pore-size filter (Millex-GP Filter Unit; Millipore). Phages were detected and purified using the double-layer agar plate assay. The phages vB_SmeS_BUCT700 and vB_SmeS_BUCT703 (targeting SMA-D39) were provided by Prof. Yigang Tong at Beijing University of Chemical Technology.

Phage preparation, quality control, and administration

Amplification and purification of therapeutic phages were performed by culture-dependent methods⁴. Briefly, 350 μL of overnight bacterial culture was added in 35 mL LB liquid medium and the mixture was incubated with shaking at 200 rpm 1 h at 37 °C. Then, phage was added at MOI of 0.01 and shaking at 200 rpm until the mixture became clear. Next, the mixture was centrifuged at 6000 × g for 5 min at 4 °C to remove debris and the supernatant filtered using a sterile 0.22 µm pore-size filter (Millex-GP Filter Unit; Millipore). The concentrate was dialyzed against 0.9% sodium chloride physiological solution (Shandong Qidu Pharmaceutical) 3 times for a minimum of 3 h each. All phage preparations, with a concentration of 10⁸ PFU/mL (on amplification host), were sterilized, aliquoted and packaged in accordance with GMP standards at a facility partnered with CreatiPhage. The quality specifications for the phage formulations used in this study were adapted from the group standard (T/SHPPA 028-2024, Quality requirements of phage preparations used for clinical therapy [in Chinese]). As the preparations were intended for localized nebulized delivery, they were processed as non-sterile preparations with a six-month shelf life at 4 °C, and endotoxin removal and testing were not required⁶. For microbiological quality control, 10% of each production batch was randomly sampled; 500 μL from each vial was plated onto blood agar and incubated at 37 °C for 48 h. The absence of bacterial growth indicated acceptable microbiological quality. Prior to phage therapy, the titers and plaque clarity of ready-to-use phage vials were determined against target bacteria using the efficiency of plating (EOP) assay⁴. Briefly, 200 μL of mid-log phase bacterial cultures from each strain were mixed with 4 mL 50 °C LB top agar and pour onto LB agar plates. Phage lysate was serially diluted 10-fold to 10⁻⁶ in LB. A 2 μL volume of each phage dilution was spotted onto the top agar, followed by an 8 h incubation at 37 °C to calculate the titer. Only phage lots with titers exceeding 1 × 10⁷ pfu/mL were selected for therapeutic use. Selected phages were combined into a cocktail and administered via respiratory nebulization using a vibrating-mesh nebulizer (Aerogen Solo) connected to the humidifier of a ventilator (Evita V300, Dräger). Each treatment course consisted of two doses of lytic phages administered within one day, either on a single day or on two consecutive days.

Primary and secondary indicators

The primary indicators were changes in target bacteria load and bacterial anti-phage resistance measured by semi-quantitative streak-plate method (or colony counting method for bronchoalveolar lavage fluid samples) and phage susceptibility spot test, respectively⁶. The streak-plate assay was performed by streaking specimens onto agar plates using a sterile inoculating loop, employing four sequential quadrant streaks of decreasing intensity. Bacterial growth was scored semi-quantitatively as follows: negative, no colonies observed; 1 + , growth confined to the first quadrant; 2 + , growth extending into the second quadrant; 3 + , growth extending into the third quadrant; 4 + , growth present in all four quadrants.

The secondary indicators include changes in inflammatory indicators, including chest radiographies, body temperature, white blood cells, neutrophils, procalcitonin, hypersensitive C-reactive protein, IL-6, as well as organ function indicators (e.g., alanine aminotransferase and aspartate aminotransferase) before and after phage therapy (Supplementary Data 2). As this study involved localized phage therapy with short treatment courses, testing for phage-neutralizing antibodies was not required. All secondary indicators, along with the quantitative and semi-quantitative bacterial assessments included in the primary indicators, were conducted in the hospital’s microbiology laboratory in accordance with standard protocols.

Genome sequencing and analyzing

Genomic DNA was extracted from isolates of A. baumannii, S. maltophilia, and K. pneumoniae using the M5 Bacteria Genomic DNA Kit (Mei5bio) according to the manufacturer’s instructions. High-quality DNA library of A. baumannii and K. pneumoniae was processed using the TELL-Seq Microbial Library Prep Kit (UST) and sequenced by the Illumina NovaSeq 6000 platform. The raw sequencing data was processed by the TELL-Read pipeline to produce linked read data in FASTQ format and the quality-controlled data were assembled using Tell-Link to generate scaffold sequences (https://universalsequencing.com/en-cn/pages/software#). The genome sequencing of S. maltophilia was performed on the Illumina NovaSeq 6000 platform with Hieff NGS®MaxUp II DNA Library Prep Kit (YEASEN) for library preparation. After sequencing, raw reads were filtered via Fastp (v0.23.0) by removing adaptors and low-quality reads, and then clean reads were obtained. Genome assembly was done using SPAdes (v3.5.0), and the Gapfiller (v1.11) was used for filling gaps. In addition, Pilon (v3.5.0) was used to improve draft genome accuracy by correcting base errors and filling gaps.

Capsular typing of CRAB and CRKP was performed using Kaptive (v2.0.6), and genetic variations were analyzed with SnapGene (v4.2.4). Multilocus sequence typing (MLST) was conducted for CRAB, CRKP, and SMA in accordance with the schemes provided on the respective MLST websites: http://pubmlst.org/abaumannii/, https://bigsdb.pasteur.fr/klebsiella/, and https://pubmlst.org/organisms/stenotrophomonas-maltophilia.

Genomic DNA of newly isolated phages KP-D4 and KP-D5 was extracted using a viral genome extraction kit (QIAamp MinElute Viral Spin Kit, Qiagen) according to the manufacturer’s instructions. Sequencing was performed on the Illumina HiSeq 2500 platform, and genome assembly was conducted using SPAdes (v3.15.5). Gene prediction and functional annotation of potential ORFs were performed using Pharokka. Predicted virulence genes were identified by comparison with the VFDB (https://www.mgc.ac.cn/VFs/main.htm), and antibiotic resistance genes were analyzed using the CARD.

Transmission electron microscopy of bacterial capsule

A. baumannii KL3_WT (D7), KL3_ΔCPS (D28), KL3_CPSmut (D40), K. pneumoniae KL64_WT (D41), and KL64_ΔCPS (D53) were cultured to the logarithmic growth phase. A 2 mL aliquot of each culture was collected by centrifugation at 6000 × g for 3 min, and the pellet was resuspended in 1 mL of 20 mM HEPES buffer (Coolaber, Beijing). Subsequently, 80 μL of ferritin (Sigma-Aldrich, Darmstadt, Hessen) was added to the suspension, gently mixed, and incubated at room temperature for 30 min. Following centrifugation at 6000 × g for 3 min, the supernatant was removed, and the pellet was washed three times with 20 mM HEPES buffer. Samples were then fixed with glutaraldehyde to a final concentration of 2.5% and examined using transmission electron microscopy (TEM).

Virulence assay in Galleria mellonella

The Galleria mellonella larval infection model was employed to evaluate the impact of capsule loss on bacterial virulence. A. baumannii KL3_WT (D7), KL3_ΔCPS (D28), and K. pneumoniae KL64_WT (D41), KL64_ΔCPS (D53) were cultured to the logarithmic phase. Cultures were centrifuged at 6000 × g for 2 min, the supernatant was discarded, and pellets were washed twice with sterile physiological saline. The bacterial suspensions were then adjusted to a concentration of 2 × 10⁸ CFU/mL. Prior to infection, larvae were incubated at 37 °C in the dark for 30 min, and only active larvae of uniform size were selected. Each experimental group consisted of three replicate plates, each containing 10 larvae. A 10 μL aliquot of bacterial suspension was injected into the right limb of the last abdominal segment using a microsyringe. Infected larvae were incubated at 37 °C, and survival was recorded every 12 h. Differences in survival curves were analyzed using the Log-rank (Mantel-Cox) test.

Clinical pulmonary infection score

The Clinical Pulmonary Infection Score (CPIS) was employed to assess pneumonia severity in the patient. The score comprises six components: temperature, WBC count, tracheal secretions, the PaO₂/FiO₂ ratio, chest radiographic findings, and pathogenic bacteria detected in tracheal aspirate culture. Each component is assigned a score ranging from 0 to 2. A cumulative CPIS score ≥6 is considered indicative of a high probability of VAP²⁶. This assessment was mainly performed by the clinical team.

Statistics

Body temperature (Tm), white blood cell count (WBC), neutrophil count (N), C-reactive protein (CRP), procalcitonin (PCT), and interleukin-6 (IL-6) were measured at four time points before (day −7, day −5, day −3, day −1) and four time points after (day 1, day 3, day 5, day 7) each course of phage therapy. Data were presented as box plots showing the median, quartiles, and range by using GraphPad Prism (v9.5.0). Statistical significance of Galleria mellonella survival was determined using ordinary one-way ANOVA followed by Tukey’s multiple comparisons test.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization and methodology, J.B., T.Z., M.C., N.W., H.W.; project administration, Y.X.S. (Yongxin Shi), J.L. (Jie Li), Q.Y.; supervision, M.C., N.W., H.W.; investigation, Y.X.S., J.L., Q.Y., J.H.L. (Jianhui Li), L.L., JF.L. (Jianfang Liu), J.D., Y.M.S. (Yumeng Shang), X.Z., L.Y., C.C., Z.Z., H.Z.; data curation, Y.X.S., J.L., J.B., M.C.; visualization, Y.X.S., M. C., N.W.; writing—original draft, Y.X.S., J.L., H.W.; writing—review and editing, T.Z., M.C., N.W.; funding acquisition, J.L., J.B., T.Z., N.W.; resources, J.B., T.Z., N.W., H.W.

Peer review

Peer review information

Nature Communications thanks Jean-Paul Pirnay, Jon Koff, and Sabrina Green for their contribution to the peer review of this work. A peer review file is available.

Funding

This study was supported by grants from the National Key Research and Development Program of China (2021YFA0911200 to N.W.), the National Natural Science Foundation of China (82070772 to T.Z., 32470145 to N.W., 32100113 to J.C.), In-hospital Research Project of Shanghai Public Health Clinical Center (KY-GW-2025-11 to J.L.).

Data availability

The supplementary data and the source data generated in this study are provided in the Supplementary Information/Source Data. The complete genome sequences have been deposited in the National Center for Biotechnology Information (NCBI) under accession numbers SRR35253276, SRR35253275, SRR35253274, SRR35253273 for A. baumannii and SRR35249311, SRR35249310, SRR35249309, SRR35249308 for K. pneumoniae, and SRR35333929, SRR35296177, SRR35296176, SRR35333928 for S. maltophilia, respectively (Supplementary Data 1). The genome sequences of phages have been deposited in the NCBI under accession numbers CP-P-AB022 (PX056322), CP-P-AB011 (PX872918), CP-p-PMA-011 (PX872916), Vb_SmeS-BUCT700 (OM735686.1), Vb_SmeS-BUCT703 (OM735688.1), KP-D4 (PX872917), KP-D5 (PX880621), CP-p-KP005 (PX872919), CP-p-KP065 (PX872919), CP-p-Kpn-107 (PX872921), CP-p-Kpn-142 (PX872915), and CP-p-Kpn-145 (PX872922). Source data are provided with this paper.

Competing interests

Q.Y., J.F.L., X.Z., C.C., Z.Z., H.Z., and N.W. are employees of CreatiPhage Biotechnology. Q.Y. and J.F.L. participated in data analysis, and X.Z., C.C., Z.Z., and H.Z. were responsible for phage preparation. Prior to joining CreatiPhage, N.W. was affiliated with the Shanghai Institute of Phage and jointly supervised the implementation and analysis of this case. The other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75735-w.