Phage Therapy of Mycobacterium Infections: Compassionate Use of Phages in 20 Patients With Drug-Resistant Mycobacterial Disease

1The therapeutic use of phages for treating drug-resistant bacterial infections has received recent attention, but the types of infections and pathogens deemed suitable; routes, dosage, and frequency of administration; interactions with antibiotics; and pharmacokinetics remain unclear. Unlike small molecule antibiotics, bacteriophages can replicate at the sites of infection, are much larger than standard antimicrobials, and penetration to sites of bacterial replication may be restricted. Immune neutralization may also limit phage activity. Bacteriophages are often highly bacterium-specific, which is advantageous for precise pathogen targeting, but demands personalized phage matching for individual patient isolates. Anecdotal reports support a robust safety profile and clinical improvement has been reported for some but not all cases.

2Because of increasing and widespread antibiotic resistance among Mycobacterium pathogens, alternative therapies are needed. Nontuberculous mycobacterial (NTM) infections—especially those caused by Mycobacterium abscessus—are particularly challenging, as many are refractory to antibiotics and extended drug therapies are poorly tolerated. NTM infections are increasingly common among people with cystic fibrosis (CF), but are also prevalent among non-CF patients, including those with bronchiectasis or Mendelian susceptibility to mycobacterial disease (MSMD). People with CF typically have complex recurrent pulmonary infections, often with mixed flora, but M. abscessus infections are particularly challenging and typically preclude lung transplantation. Phages have been proposed for managing CF, but their utility for Mycobacterium infections, including NTM and tuberculosis, remains unclear.

3There is great variation in phage susceptibilities among M. abscessus clinical isolates. Approximately 40% of M. abscessus isolates have a smooth colony morphotype, and to date no therapeutically useful phages have been identified for these. In contrast, 75%–80% of rough strains are efficiently killed by at least 1 phage, and the low rates of phage resistance in vitro suggest that phage resistance in vivo may not be a limitation. Nonetheless, the repertoire of therapeutically useful phages is small, and mostly limited to phages isolated on Mycobacterium smegmatis; few phages have been isolated directly on any strain of M. abscessus.

4Two case reports have described compassionate use of phages for NTM infections. One was a patient with CF and disseminated M. abscessus infection following bilateral lung transplantation and drug-induced immunosuppression. The second was an immunocompetent patient with non-CF bronchiectasis and a severe M. abscessus pulmonary infection. The same 3-phage cocktail was used to treat both patients and was administered intravenously (IV) twice daily at a dose of 109 plaque-forming units (PFUs) per dose for at least 6 months. Both were also treated with concomitant multidrug antibiotic regimens. The first patient had clinical improvement in lung function, radiographic imaging, and clinical signs and symptoms, although without complete clearance of the infection. The second had initial reduction in M. abscessus colony counts in sputum that was abrogated by emergence of a potent neutralizing antibody response to the phages.

5We report here therapeutic interventions for a pilot cohort of 20 patients with antibiotic-refractory mycobacterial infections. Phage administration was safe, no resistance was observed, and favorable microbiological or clinical outcomes were observed in a majority of cases.

6Since May 2019, we received approximately 200 requests for adjunctive phage treatment for patients with NTM infections that were either refractory to antibiotic therapies or in which extended drug treatments were not tolerated; antibiotic susceptibilities are shown in Supplementary Table 1. Most patients had CF and M. abscessus infections, but some had other underlying diseases complicated by NTM (and in 1 case bacille Calmette-Guerin [BCG]) infections. The following criteria determined eligibility for compassionate use phage therapy: age >5 years, microbiologic documentation of mycobacterial infection based on at least 2 positive cultures of relevant tissue or body fluids; drug susceptibility testing (DST) documenting resistance to multiple antimycobacterial drugs; clinical signs, symptoms and radiographic findings involving at least 1 organ system (eg, NTM lung disease based on American Thoracic Society (ATS), European Respiratory Society (ERS), European Society of Clinical Microbiology and Infectious Diseases (ESCMID), and Infectious Diseases Society of America (IDSA) diagnostic criteria) or evidence of disseminated disease; clinical failure of or intolerance to antimycobacterial treatment; stable underlying conditions with anticipated survival of at least 3 months; and treating physician(s) able and willing to use adjunctive phage therapy. Recent mycobacterial isolates from patients were screened for phages that efficiently infect and kill the isolate, and regulatory approval was provided under an emergency Investigational New Drug application by the United States Food and Drug Administration or comparable processes outside of the United States. Local investigational or ethics review board approvals were also obtained. Twenty patients who met criteria provided informed consent for adjunctive phage therapy (Table 1). See the Supplementary Information for detailed methods.

7With relatively few potentially therapeutically useful phages for NTM, in 11 cases only a single candidate phage was identified; for others, 2 or more genomically distinct phages were combined into cocktails (Tables 1 and 2). Most patients were initially administered 109 PFU intravenously twice daily; some patients also received the same dose by inhalational nebulization (Table 1). All patients also received antimycobacterial treatment with at least 2 drugs based on prior DST against their own target isolate and on their prior tolerance of available drugs. Initial duration of phage treatment was 6 months, although some patients received shorter or longer courses of treatment directed by clinical and microbiologic responses (Table 1). When available, baseline clinical and laboratory assessments of patients included signs and symptoms of NTM infection, a complete blood count, renal and liver chemistries, sedimentation rate, and C-reactive protein, and, when applicable, radiographic assessment and pulmonary function studies. Where possible, these were monitored weekly during the first month of phage treatment and monthly thereafter; radiographic and pulmonary functions were evaluated at intervals determined by treating clinicians. Microbiological monitoring included acid-fast bacilli (AFB) smear and culture of sputum or other relevant clinical specimens at baseline, at 1 month, and then periodically thereafter based on clinical and microbiologic response. During treatment, Mycobacterium strains from clinical samples were tested for phage resistance (Table 2). Where possible, serum, sputum, and/or bronchoalveolar lavage specimens were collected prior to and after treatment initiation and tested for phage antibodies using neutralization or enzyme-linked immunosorbent assays or both (Table 3). For several patients, both serum and sputum samples were tested by polymerase chain reaction (PCR) for the presence of phage DNA; only sporadic weakly positive signals were observed (Supplementary Figure 1). Antimycobacterial regimens were adjusted by the treating clinicians as needed based on DST and drug tolerability during phage treatment.

8Mycobacterial isolates from individual patients were tested for sensitivity to a panel of approximately 25 phages representing genomic clusters known to infect M. abscessus or Mycobacterium tuberculosis (Figure 1). Of the 200 strains screened between May 2019 and May 2021, 157 (78%) were M. abscessus; 77 and 71, respectively, had rough and smooth colony morphologies; 9 were mixed and both types were purified and cultured. Fifty-five (71%) rough colony strains were infected and killed efficiently by at least 1 phage. If an isolate was efficiently infected and killed by 1 or more phage, and the patient’s clinical status indicated eligibility for compassionate use intervention as previously described, regulatory permissions were obtained and purified phages were dispatched to a local dispensing pharmacy. Of the 20 patients offered treatment, 17 had M. abscessus infections; 14 of these had underlying CF, 1 had bronchiectasis, 1 had scleroderma, and 1 had hypersensitivity pneumonitis. One patient had a Mycobacterium chelonae disseminated skin infection, 1 had CF with pulmonary Mycobacterium avium, and 1 had a disseminated BCG infection (Table 1). Of the 17 M. abscessus strains isolated, 12 were subspecies abscessus and 5 were subspecies massiliense (Table 1).

9All patients were treated based on compassionate use, had disparate underlying conditions, and had complex infections due to diverse mycobacterial species with differing patterns of phage and antimicrobial susceptibility. Therefore, fixed definitions of treatment response were not possible. Generally, a favorable response was defined as mycobacterial smear and culture conversion to negative in at least 1 relevant specimen coupled with clinical and/or radiographic improvement or resolution of signs and symptoms of infection after at least 6–8 weeks of phage treatment. A partial response was defined as either mycobacterial smear or culture conversion or clinical and radiographic improvement. All patients had a previous history of prolonged or relapsing mycobacterial infections often coupled with other drug-resistant bacterial coinfections, underlying organ system consequences of CF or chronic lung disease, and numerous other clinical complications. Patients with CF were generally not receiving or had failed CF transmembrane conductance regulator modulators. Antibiotic therapies were optimized where possible, although most patients had only 1 or 2 at least partially active antibiotics that were used with phages. Underlying conditions and drug toxicities complicated interpretation of phage efficacy in several patients. Within this context, of the 20 patients treated, favorable or partially favorable responses were observed for 11 patients, 5 had inconclusive outcomes, and 4 had no response (Tables 1–3). Groups of patients with differing responses to phage treatment are described below; detailed synopses are in the Supplementary Information.

10Favorable clinical and microbiological responses were observed in 5 patients (patients 1, 9, 10, 15, and 16) and partial clinical or microbiological responses were noted in 6 patients (patients 2, 4, 7, 13, 17, and 20) (Table 1); all were treated with IV phages except for patients 13 and 17, who also received nebulized phages after the IV regimen. Eight of these 11 patients had CF and complex pulmonary infections (7 M. abscessus, 1 M. avium), 1 had disseminated skin lesions caused by M. chelonae, 1 had scleroderma with M. abscessus infection, and 1 had disseminated BCG infection (Table 1). In 5 patients (patients 9, 10, 15, 16, and 17), the infections appear to have been largely resolved: 3 with pulmonary M. abscessus infections (patients 9, 10, and 15), 1 with pulmonary M. avium infection (patient 17), and another with M. chelonae skin infection (patient 16); details of patients 15 and 16 were recently reported. Three of these 11 patients (patients 7, 16, and 17) had no change in their antibiotic regimen during phage treatment, but at least 4 had some change to their antibiotic course. Patient 15 cleared M. abscessus cultures and successfully underwent a bilateral lung transplant. Two patients (1 and 13) had substantial clinical improvement, but without clear evidence of culture conversion. One of these (patient 1) was the case reported previously in which posttransplant disseminated M. abscessus infection and clinical signs and symptoms greatly improved, but some skin nodules persisted after >1 year of phage treatment. These were only intermittently culture positive, but the patient subsequently died from CF-related health challenges and organ failure 44 months after the start of phage treatment. In the other (patient 13), there was substantial improvement of symptoms and forced expiratory volume in 1 second by spirometry, but the patient remained culture positive.

11For 4 patients (patients 2, 4, 7, and 20), response to therapy was partial and more difficult to assess largely due to complications from other infections, although there was evidence of improved control of the Mycobacterium infections. One patient (patient 2) had a severe chest infection requiring sternum resection. Phage treatment resulted in AFB smear-negative chest swabs, but the patient died after failing therapy for multiple bacterial and fungal coinfections. For patient 4, phage treatment resulted in conversion to culture negative tracheal aspirates, but systemic adenovirus infection resulted in death. Patient 7 had M. abscessus infection with both rough and smooth colony morphotypes; the rough strain derived from the smooth strain by a mutation in glycopeptidolipid synthesis. The phage identified for the rough strain (Table 1) did not infect the smooth counterpart but was administered on the possibility that removal of the rough variant could be clinically beneficial, even if the smooth strain persisted. The patient remained clinically stable but had persistently positive sputum cultures that grew only smooth strains after phage administration, suggesting that phage treatment had reduced the burden of the rough strain. Finally, patient 20 had MSMD and disseminated BCG infection. Phage treatment resulted in improved clinical signs and symptoms and marked reduction in BCG PCR positivity in weekly blood and urine samples. However, the patient died of other complications.

12Five patients (patients 3, 8, 11, 12, and 18) had inconclusive responses to therapy or had modest short-lived improvements. Patient 8 developed phage neutralizing antibodies and had little clinical improvement with either aerosolized or IV administered phage, likely due to the phage neutralizing immune response. Like patient 7, patient 11 had a mixed infection with both rough and smooth colony morphotypes of M. abscessus, and active phage was identified only for the rough strain (Table 1). While chest radiographs improved, sputum cultures were intermittently positive. Patient 12 showed some reduction in sputum M. abscessus load during the first month of IV administration, but subsequent recrudescence correlated temporally with an increase in antibody-mediated phage neutralization. Patient 18 has not shown substantial clinical improvement, likely also due to serum neutralization of phage; he has been switched to aerosolized phage therapy after 4 months of IV administration and treatment is ongoing. Patient 3 died shortly after the start of phage administration due to multiple organ failure.

13Three patients with CF and 1 with non-CF bronchiectasis (patients 5, 6, 14, and 19), all with pulmonary M. abscessus infections, showed no overall clinical or microbiologic response (Table 1). All had been treated with IV phage, although 1 patient (patient 19) was switched to aerosolized administration. The reasons for lack of response are unclear.

14Phage administration by either IV or aerosolized routes was well-tolerated with no serious adverse reactions related to the phage in any patient. Phage preparations were highly purified, certified to be sterile, and had undetectable endotoxin levels, an advantage of using a lipopolysaccharide-free bacterial host (M. smegmatis) for phage growth. Eleven patients were treated with just a single phage (Table 1; 8 with Muddy, 3 with BPsΔ33HTH_HRMGD03), and in the 6 patients from whom rough colony M. abscessus isolates were recovered after the start of phage treatment, all remained fully phage sensitive. Indeed, changes in phage sensitivity were only observed in 1 phage in a cocktail of 3 (Table 2, patients 1 and 12).

15Sera from 15 patients were tested for immune reactions, and robust neutralization of at least 1 phage was observed in 8 of the patients following IV treatment (Table 3). Although neutralization correlated temporally with loss of clinical response in patient 12, there was no correlation between neutralization and outcomes in 8 remaining patients. Patients 8, 18, and 19 also developed neutralizing antibodies, plausibly contributing to reduction in therapeutic response, but strong neutralization to at least 1 of the phages in patients 15 and 16 did not prevent favorable or partial responses (Tables 1 and 3). Patient 1 had no phage neutralization until 2 years after starting therapy, when only 1 of the 3 phages had decreased activity in serum. Curiously, in some patients (patients 12, 15, and 16), pre-phage sera recognized the phages, although these were not neutralizing. Presumably, this reflects prior exposure to related but distinct phages in the environment.

16This series of 20 patients treated with phages on a compassionate use basis provides support for further evaluation of phages for treatment of mycobacterial infections. Phage administration was well-tolerated, and phage resistance was not observed even when using a single phage. Favorable responses were observed in more than half of the patients, including complete resolution of some infections, and successful lung transplantation in 1 patient. However, some patients saw little clinical benefit, and the basis for this variability in response is unclear. Although phage treatment of mycobacterial infections shows promise, this cohort illustrates some key limitations and lessons.

17First, the repertoire of therapeutically useful phages is small, and expansion requires further phage isolation, developing phages induced from lysogenic strains or using synthetic phages. However, the lack of phage resistance (Table 2) supports use of a single phage, and where >1 phage is available, cycling their administration to circumvent neutralization. Second, optimal routes of phage administration and adequacy of tissue penetration are unclear. IV administration may be preferable for treatment of disseminated infections and appears effective for at least some lung infections, particularly when there is structural lung damage due to fibrosis, severe bronchiectasis, or mucoid plugging that compromises delivery by nebulization alone. We note that for patient 9, in whom the pulmonary M. abscessus infection was fully resolved, the phages were also deposited bronchoscopically, which may have contributed to more effective phage delivery to the infection. Nebulization may also avoid systemic neutralization. The immune status of the patient is also important; immunocompromised patients may tolerate extended phage administration without antibody-mediated neutralization. However, little is known about intracellular penetration or uptake of phages, particularly by macrophages, where most replicating mycobacteria are found. Third, dosage and regimens warrant optimization. As the treatments are well-tolerated, higher doses could be contemplated, using longer interdose intervals. Further exploration of pharmacodynamics and tissue penetration of phage is critical.

18The lack of therapeutically useful phages for smooth M. abscessus strains, the unpredictable specificity for rough strains, and the limited phage repertoire represent current impediments to broad implementation of phage treatments. However, these limitations are not insurmountable, and these case studies suggest that phage treatments may be valuable tools for clinical control of NTM infections. Successful outcomes for M. chelonae, M. avium, and BCGosis, as well as M. abscessus, suggest a large spectrum of target Mycobacterium diseases.

19Although compassionate use case studies such as these lack the rigor and consistency of treatment and patient monitoring possible in carefully controlled blinded clinical trials, they provide a wealth of insights for designing such trials. Variations in antibiotic regimens, surgical interventions, and management of coinfections can all potentially influence patient status, and direct linkage of phage treatments with outcomes in individual patients is perilous. Nonetheless, this series of case studies strengthens the likelihood of direct benefits from phage treatments of Mycobacterium infections and the potential for infection control when none other is effective.