Carbapenem-resistant Pseudomonas aeruginosa (CRPA) accounts for approximately 10% to 30% of clinically isolated P aeruginosa strains in the United States.¹ Carbapenemase production represents only a fraction of this resistance, with the Centers for Disease Control and Prevention reporting 1,530 carbapenemase-producing CRPA (CP-CRPA) among more than 68,000 CRPA isolates collected between 2017 and 2022.² Within CP-CRPA in the United States, Verona Integron-encoded Metallo-β-lactamase (VIM) is the predominant carbapenemase detected. While the New Delhi metallo-β-lactamase (NDM) is prevalent in Asia, Europe, and the Middle East, it remains rare in the United States, with few cases of autochthonous transmission. NDM-producing CP-CRPA can be especially challenging to treat, with ~30% treatment failure.³ One such case of NDM-producing CP-CRPA in Southern California is thought to be an endemic strain, given the patient’s lack of international travel and genetic dissimilarity to global strains.⁴

The transmission of carbapenemase genes within and between bacterial populations is of urgent concern, making infection prevention critical to stop the spread of these isolates. A number of outbreaks of CP-CRPA have been reported in health facilities in the United States, including an outbreak of imipenemase metallo-β-lactamase (IMP)-producing strains in an intensive care unit in Idaho,⁵ a cluster of VIM-producing strains in a skilled nursing facility in Chicago,⁶ and a recent report of NDM-1 isolates in Texas, starting with a travel-associated index case.⁷ Certain “high-risk strains” of P aeruginosa, such as ST235, ST111, and ST233, are particularly concerning due to their extensive drug resistance (XDR) and carriage of carbapenemase genes. These sequence types, found globally, can harbor both chromosomal and plasmid carbapenemase genes and are associated with over 60 different β-lactamases.⁸

Here, an outbreak involving local transmission of NDM-producing CP-CRPA in a Southern California hospital is described. Epidemiologic investigations and genomic sequencing identified a sink as the environmental reservoir.

Eight cases of NDM-producing CP-CRPA were identified in the same Southern California hospital: one from April 2023, five from June through August 2023, one from April 2024, and the last one from September 2024 (Table 1). The index case (patient 1) in this cluster was previously described⁴ and reported recent travel to Hawaii. Patient 2 received recent medical care in Iran, and no other patients reported recent travel outside of the area. Two patients (patients 2 and 6) did not receive any antimicrobial treatment due to asymptomatic bacteriuria, two recovered (patients 1 and 7) after treatment with cefiderocol. Three patients (patients 4, 5, and 8) succumbed due to underlying comorbidities.

Notably, patient 3 was admitted for extracorporeal membrane oxygenation support and initially treated with cefiderocol for NDM-producing CRPA identified in a sputum culture. After stabilizing, the patient was transferred to their home institution for management of acute myeloid leukemia. During this time, the patient developed necrotic skin lesions secondary to calciphylaxis, which became superinfected with multidrug-resistant organisms, including a now pan-resistant NDM-producing CRPA. The patient received bacteriophage therapy, compassionate use of cefepime-zidebactam, and sodium thiosulfate for calciphylaxis. The patient has since fully recovered.

Isolates from these NDM-producing CP-CRPA cases were obtained from various sources, including urine, respiratory tract, blood, and peritoneal fluid (Table 1). Consistent with the production of NDM β-lactamase, phenotypic susceptibility testing (Table 2) found that all isolates were resistant to β-lactam antibiotics, including carbapenems. All isolates included in this study were susceptible to cefiderocol, and those tested for colistin susceptibility exhibited intermediate results.

In response to the identification of NDM-producing CP-CRPA in patients 4 and 5, the first epidemiologic investigation was launched. Commonalities between staff, procedures, equipment, and patient care locations were investigated. From this, it was determined that respiratory therapists (RTs) were shared between four patients, and registered nurses between two patients. All patients were either on a ventilator or had a tracheostomy at the time of their positive culture. However, no RT equipment or machines were identified as shared between patients, and single-use nebulizers and airway clearance devices were strictly used before disposal. Additionally, no shared water sources were identified. Extensive environmental sampling of the RT equipment reprocessing area, including sinks, drains, and work surfaces, did not yield any P aeruginosa. As a precaution, the sink in the RT equipment reprocessing area was replaced.

Following identification of the seventh case in April 2024, a point prevalence study was conducted in the unit. While several patients had P aeruginosa identified on rectal culture, none were carbapenemase producers. Additional environmental sampling revealed growth of P aeruginosa in multiple sinks and P-traps, but none were NDM CP-CRPA. After the eighth case, a third investigation was launched, and extensive environmental sampling ultimately detected two NDM-positive environmental P aeruginosa isolates from the drain and P-trap of a sink in an intensive care unit room that previously housed patients 1 and 8.

To determine the relatedness of NDM CP-CRPA isolates from patient and environmental sources, whole-genome sequencing (WGS) using Illumina was performed and the resulting genome assemblies were compared. All isolates were typed as ST235 and harbored the bla_NDM-1 gene. Nucleotide-level comparison of all assemblies revealed that isolate 2 differed by over 1,000 SNPs, indicating it was unrelated to the other isolates (Fig. 1A). After excluding isolate 2, the remaining patient isolates (from patients 1, 3, 4, 5, 6, 7, and 8) and both environmental isolates (sink drain and P-trap) differed by only zero to 14 SNPs, suggesting close relatedness (Fig. 1A). Comparison to other sequenced isolates available at NCBI using Pathogen Detection¹⁷ revealed that environmental isolates, and those from patients 1, 3, 4, 5, 6, 7, and 8, formed a distinct cluster, separate from other isolates (Fig. 1B). In contrast, isolate 2 did not cluster with any of the other isolates (data not shown). Notably, the isolate collected from patient 8 was most similar to the environmental and the index patient isolates, clustering phylogenetically with both (Fig. 1A,B), consistent with the epidemiologic link of the two patients who stayed in the same room with the contaminated sink, albeit 1.5 years apart. These results suggest a high likelihood that the NDM-producing CP-CRPA was introduced by the index patient and subsequently spread to the other six patients over the course of 1.5 years, and a contaminated sink played an important role in its transmission.

WGS revealed comprehensive AMR gene profiles for individual isolates, consistent with the XDR phenotypes observed. All isolates contained several AMR genes in addition to bla_NDM-1 (Fig. 1C). Notably, isolate 2 harbored aminoglycoside resistance genes aph(3′)-IIb and aph(3′)-VI, β-lactamase genes bla_PME-1, bla_PDC-35, and bla_OXA-488, the fosfomycin resistance gene fosA, phenicol resistance genes catB7 and floR2, the sulfonamide resistance gene sul1, and the tetracycline resistance gene tet(G). Additionally, isolate 2 contained an oprD (W339*) mutation associated with β-lactam resistance, as well as gyrA(T83I) and parC(S87W) mutations associated with quinolone resistance. Outbreak isolates shared similar AMR gene profiles, including aph(3′)-IIb and aph(3′)-VIa, bla_PDC-35 and bla_OXA-488, fosA, the fluoroquinolone resistance gene crpP, as well as catB7, floR2, sul1, and tet(G). All outbreak isolates also carried an ambiguous bla_PME gene, with the closest match being bla_PME-1, a truncated form of the β-lactamase gene bla_KBL, and a penicillin-binding protein 3 gene mutation ftsI(R504C) associated with β-lactam resistance. Furthermore, the gyrA(T83I) mutation was observed in all outbreak isolates. However, there were some differences observed between outbreak isolates: environmental isolates and those from patients 1 and 8 contained the aminoglycoside resistance genes aac(6′)-Ib4 and aadA1, the phenicol resistance gene catB3, and the β-lactamase gene bla_OXA-10, while isolates from patients 3, 4, 5, 6, and 7 contained the aminoglycoside resistance gene aadA6. These results demonstrate that isolate 2 had a unique AMR gene profile compared with the other isolates, further supporting its unrelatedness. Moreover, the observed AMR gene differences between environmental isolates and those from patients 1 and 8 compared with the remaining isolates suggest that genomic changes emerged over the course of the outbreak.

One potential explanation for the AMR gene differences observed between outbreak isolates may be explained by gene residence within active MGEs. To enable full resolution of MGEs and infer MGE activities occurred during this outbreak, ONT long-read WGS was performed using select isolates and combined with short reads to generate hybrid assemblies. Isolate 1 from the index case contained a class 1 integron nested within a putative Tn3 family transposon with sul1 and qacEΔ1 in its 3′ conserved region. Within its variable region, this integron had an array of three gene cassettes (Fig. 2): the first cassette harbored aac(6′)-Ib4, the second catB3, and the third bla_OXA-10 and aadA1 in different reading frames. In isolates 4 and 5 collected from later cases, a class 1 integron located at the same genomic position contained only one gene cassette that harbored aadA6. Taken together, these observations suggest that the variability in AMR genes and phenotypic susceptibility observed within this outbreak was due to gene cassette switching in an integron that remained in place.

We also explored the intraspecific transmissibility of the bla_NDM-1-containing MGE. Hybrid genome assemblies of Southern California CP-CRPA isolates revealed that bla_NDM-1 in outbreak isolates was located within a chromosomal array of AMR genes containing sul1, three copies of bla_PME-1, aph(3′)-VIa, floR2, and tet(G), as well as truncated forms of bla_KBL and ble_MBL, which was not detected in short-read-based assemblies. This array was located within the "cargo2" region of a previously described MGE, integrative and conjugative element (ICE) 6660¹⁸ (Fig. 3). This ICE6660 cargo2 region resembled others found in sequenced isolates available at NCBI, including P aeruginosa from South Korea, P juntendi from China, and Aeromonas caviae from Vietnam (Fig. 3), suggesting widespread exchange across species and even genera.

Sequencing revealed close genetic similarity (0-14 SNPs) between all isolates, except isolate 2, which was quickly determined to be unrelated. Given the high prevalence of NDM-1 CP-CRPA in Iran and the Middle East,^19,20 this isolate may be associated with the patient’s extensive health care history in the region prior to visiting the Southern California facility. A prior study demonstrated that the index isolate, isolate 1, was distinct from previously described global isolates, indicating it was likely a unique domestic strain.⁴ The subsequent isolates are also likely descendants of this same strain, given a P aeruginosa mutation rate of between 6 × 10⁻⁴ and 1.6 × 10⁻¹ mutations per genome per generation.²¹ Indeed, phylogenetic analyses using the National Center for Biotechnology Information (NCBI) Pathogen Detection Isolates Browser¹⁷ demonstrated that all the Southern California isolates distinctly cluster together relative to other isolates available at NCBI, indicating a hospital outbreak that was originated from an index patient without international travel history, and transmitted nosocomially including via a contaminated sink. The outbreak progressed slowly, and cases were identified sporadically, spanning 1.5 years. The insidious nature of the transmission, despite exhaustive investigation and numerous negative environmental cultures, made the investigations challenging. The source was not identified until the third intensive culture–based investigation with dozens of environmental samples tested.

MGEs such as plasmids, transposons, and ICEs are major contributors to the spread of AMR genes through horizontal gene transfer.²² Notably, the NDM-1 gene is located on the chromosome and not on a plasmid, which makes this strain inherently resistant to carbapenems. No plasmids were detected in isolates from this outbreak, while multiple AMR genes were found within putative MGEs inserted into the bacterial chromosome. Several AMR genes, including bla_NDM-1, were found in an array within the cargo2 region of an ICE6660 element; 3 copies of bla_PME-1 were present. This region is similar to other ICE6660 cargo2 regions found in clinical P aeruginosa and P juntendi isolates collected in South Korea and China, respectively, as well as an environmental A caviae isolate collected from wastewater in Vietnam, suggesting widespread horizontal gene transfer. Chromosomal insertions suggest AMR genes in the Southern California CP-CRPA strain may be more stable than if they were located on plasmids, which heightens the epidemiologic significance of this XDR pathogen.

The observed differences in detected AMR genes between genetically related isolates suggested some AMR genes were lost and acquired within this cluster. Integrons are genetic elements containing a variable region flanked by attI and attC recombination sites allowing for the capture and exchange of different gene cassettes, and can be carried by transposons integrated into the genome or plasmids.²³ The isolate from the index case within this outbreak had a class 1 integron containing three AMR gene cassettes harboring aac (6′)-Ib4, catB3, bla_OXA-10, and aadA1, while isolates from later cases had only one AMR gene cassette with aadA6. As these integrons were inserted at the same location on the bacterial chromosome, these findings suggest that AMR gene cassettes were swapped during the course of this outbreak, providing an example of AMR gene dynamics that can occur in related isolates. Therefore, phenotypic susceptibility results should be interpreted with caution for outbreak investigations.

Nosocomial spread of Pseudomonas via sinks has been demonstrated in numerous studies, as the bacteria are well-adapted to thrive in moist environments and can form a biofilm extending from the P-trap to the drain.^24-27 Based on our investigations and findings, several interventions were implemented, including: (1) replacing the sink P-trap in the affected room; (2) weekly application of an EPA-registered disinfectant against Pseudomonas biofilms (Virasept, Ecolab, Inc) in all sinks; (3) staff education; (4) installation of splash guards in the medication preparation areas; (5) reminders to staff to keep patient care supplies out of sink splash zones; (6) adjusting faucets so they do not pour directly on the drain; (7) regular audits of environmental services cleaning; and (8) use of the CDC Water Infection Control Risk Assessment (WICRA) for Healthcare Settings to evaluate and prioritize water risk interventions. Follow-up environmental cultures obtained six weeks later, did not detect NDM CP-CRPA. No additional cases were identified as of the submission of this paper (February 24, 2025).

While the contaminated sink was strongly implicated in the transmission of NDM-producing CP-CRPA to patients 1 and 8, not all cases in this outbreak could be directly epi-linked to that environmental reservoir. For instance, patients 4 and 5 were identified on the same date (July 7, 2023). Although they were not housed in the same room or hospitalized concurrently, our review of care patterns revealed overlapping involvement of respiratory therapy and nursing staff, as well as shared use of procedural rooms for similar interventions. These observations raise the possibility of indirect cross-transmission via health care personnel or contaminated procedural environments. While causality cannot be confirmed retrospectively, such overlaps represent plausible transmission routes that merit consideration in outbreak investigations. This highlights the complexity of health care-associated pathogen transmission and underscores the importance of evaluating both environmental and operational factors in prolonged outbreak responses. Contaminated sink drains have been clearly linked to numerous outbreaks of infections due to gram-negative bacilli, including Pseudomonas, and have received increasing attention as an environmental reservoir.²⁸ Colonized drains can result in intermittent dispersal of organisms to other environmental surfaces such as countertops, to health care worker or patient hands or clothing, or medical equipment.^29-31 In outbreaks similar to this one, prolonged durations have been noted lasting many months or years. Longer patient length of stay may be a risk factor for patient acquisition.^25,32 Once established, eradication of sink drain colonization is challenging due to persistence of organisms in biofilms and inadequate contact time achieved by disinfectants when poured into a sink drain. In this outbreak, the lengthy time in-between cases and initially negative environmental cultures highlight the challenges with outbreak investigation of potentially contaminated sinks. Intensive water management interventions, such as application of EPA-registered disinfectant with biofilm claims, may be required to control transmission.^25,31,32 Sink drain and wastewater plumbing should be included in health care water infection control risk assessments and investigated as a potential source of outbreaks due to gram-negative bacilli.^25,32

NDM-producing CRPA poses an extremely high risk to patients, especially those with reduced immune function and extended hospitalization. Use of WGS contributed to the rapid detection, characterization, and subsequent infection control response, which is critical to curbing the spread of these organisms in health care settings. We demonstrate that a unique strain of NDM CP-CRPA was transmitted among seven patients in a single hospital over the course of 1.5 years, with two cases linked to a contaminated sink. Implementation of real-time pathogen surveillance enabled effective response to an ongoing outbreak that involved environmental sampling, microbiological testing, WGS, facility management, and updated infection prevention policies.