WGS technology is revolutionizing IPAC. Here, routine surveillance of MDR pathogens across the US MHS was critical in uncovering a decades-long P. aeruginosa epidemic cluster. With traditional approaches alone (Giani et al., 2018), comparable outbreaks may avoid detection due to the sporadic nature of patient infections, scattering of cases throughout the hospital, and changing antibiotic susceptibility profiles. Furthermore, because of budget constraints, surveillance programs focusing on ‘high-risk’ isolates carrying select resistance genes (e.g. ESBLs, carbapenemases) (World Health Organization, 2017; Hong et al., 2015; Del Barrio-Tofiño et al., 2020) would also fail to detect this ST-621 outbreak clone. Indeed, while the ST-621 lineage was the focus of several studies in the late 2000s, these examined a blaIMP-13 carrying epidemic clone which spread throughout Europe and has since been sporadically detected in South America and Southeast Asia (Mereuţă et al., 2007; Fournier et al., 2012; Santella et al., 2010; Teo et al., 2021). To our knowledge, this IMP-carrying strain has not been reported in the USA to date, and the outbreak in Facility A is the result of a distinct MDR clone lacking the carbapenemase.
Similar to the US MHS, healthcare networks worldwide that already benefit from prospective WGS surveillance programs are reporting large numbers of unrecognized outbreaks (Parcell et al., 2018). However, two resources rarely available in other settings were key to this investigation: a genome repository of isolates from the preceding decade and the ability to conduct prospective environmental sampling. The former was integral to date the presumed origin of the outbreak to the late 1990s, soon after the hospital opened, and to trace the evolution and separate spread of two subclones throughout floors and wards. The latter revealed that, more than 20 years later, reservoirs of both subclones persist in sink drains from patient rooms throughout the facility. This supports previous evidence implicating these sites as major reservoirs for P. aeruginosa in the clinic (Kerr and Snelling, 2009; Kotay et al., 2017). Decontamination of drains can be extremely challenging due to the limited penetration of disinfectants, the lack of access (e.g. to perform scrubbing), recolonization (e.g. disposal of contaminated patient specimens), or retrograde growth from p-traps (Kerr and Snelling, 2009; Smith and Hunter, 2008; Kotay et al., 2017). Furthermore, sinks with shallow basins and gooseneck faucets directing water straight into the drain, similar to those in Facility A (Figure 4A), have been linked to increased backsplash onto nearby surfaces and medical equipment (Hota et al., 2009).
Besides environmental contamination, reservoirs within patients likely contributed to the spread and longevity of the outbreak. In particular, long-term infections, documented for a significant fraction of patients, provided a recurring source of the epidemic clone. It is noteworthy that 5 of the 6 patients with long-term (>1 year) infections (and 9 out of 11 if reduced to 6 months) were carriers of subclone SC2. Compared to SC1, all but one SC2 isolates carried a truncated WbpX glycosyltransferase, an essential component of the common polysaccharide antigen (CPA) biosynthesis (Lam et al., 2011). While this mutation could result in increased immune evasion, the rough LPS phenotype recurrently observed in vivo (in particular in CF patients) usually results from a functional CPA but the lack of O-specific antigen (Jurado-MartÃn et al., 2021; Hu et al., 2017). In further evidence for a progression toward a host-adapted lifestyle, a convergence of mutations in genes involved in alginate production, quorum-sensing deficiency, loss of motility, and decreased protease secretion, all phenotypes associated with chronic infections (Marvig et al., 2015), was observed throughout the 10 years of sampling.
Considering its role in long-term infections and the distinctive association with patients and sinks in the Tower Building of Facility A, the origin and emergence of subclone SC2 is of particular importance. Although obfuscated by the limited sampling, a plausible scenario (supported by BEAST2 inferences) would be that a patient with an initial infection acquired during a stay in the Main Building before the SC1-2 split (ca. 2004) proceeded to shed the epidemic strain in the newly opened (2012) Tower Building during successive visits. Importantly, the Main and Tower Buildings of Facility A have distinct plumbing systems, providing the necessary ecological separation (minus inter-hospital patient transfers) for the further spread and divergence of the two sub-clones. Patient 27 is just such an example, but the lack of early sampling (2000–2010) and incomplete collection thereafter precludes a definitive answer. Indeed, other scenarios cannot be discounted, with a recent study showing sinks in a newly built ICU were already contaminated with an outbreak clone of P. aeruginosa before the arrival of the first patients (Sukhum et al., 2022).
One of the most notable features of this outbreak was the evolution of antibiotic resistance over time, with the emergence of resistance to first (cephalosporin), second (carbapenems), and third (colistin) line treatments. The potent R504C substitution in PBP3 was selected in each subclone, and in addition to facilitating cephalosporin resistance, has also been linked to ceftazidime/avibactam resistance (López-Causapé et al., 2017). While no patient prescription data is available from Facility A, it can be speculated that the high prevalence of cephalosporin resistance in early outbreak isolates led to increased carbapenem use, which in turn selected for the many independently evolved OprD mutants, one of the most common mechanisms for carbapenem resistance in P. aeruginosa (Jurado-MartÃn et al., 2021). Finally, albeit limited to a single patient, the emergence of colistin resistance, through a well-characterized mechanism, is a reminder that the threat of extensively drug-resistant P. aeruginosa is only a prescription and a few mutations away (Chambers and Sauer, 2013).
Some limitations were noted for this study and complicated the genomic inferences. First, although the criteria remained the same, the partial and inconsistent sampling of clinical P. aeruginosa isolates prevented from capturing the exact magnitude of the outbreak through the years, and sampling bias cannot be ruled out. Second, detailed epidemiological data for the patients (date of admission(s), bed location, reason for sampling, medications prescribed, etc.) were unavailable. As a result, possible patient overlaps were only inferred from the location and culture date of the isolates, and many were likely not detected. Third, fixed thresholds (i.e. genetic relatedness, time between isolates collection, and shared hospital location) were applied to predict a possible origin of infection. Acknowledging the other limitations, these were designed to predict the most likely cases of direct patient-to-patient transmission, when an overlap was identified, or environment-to-patient transmission, when a patient overlap was ruled out. Because of conservative criteria, the origins of most cases remained undetermined, and an alternative to fixed threshold, using TransPhylo (Didelot et al., 2021), is being explored in a follow-up study. This approach uses a probabilistic model to predict whom infected whom in outbreak scenarios and reconstruct transmission trees (Didelot et al., 2021). Fitting our dataset, TransPhylo was recently extended to allow the use of multiple genomes per host (Carson et al., 2024) and to incorporate epidemiological data into the analysis (Carson et al., 2025).
The data generated during this study has resulted in various ongoing interventions (e.g. closing sinks, replacing tubing, using foaming detergents Jones et al., 2020) at Facility A. Sampling and sequencing, in near real-time, of all P. aeruginosa clinical isolates (and not just MDR) also commenced in 2021. Notably, as of May 2025, no new ST-621 patient cases have been reported in over 15 months, and the unbiased contemporary data using all P. aeruginosa from patients at this facility confirms the spread has slowed, with just four cases identified in 2022–2024. Similar to Facility A, the roll-out of routine WGS surveillance in the clinic has the potential to improve patient care and prevent some of the estimated 136 million hospital-associated drug-resistant infections per year globally (Balasubramanian et al., 2023).