Multidrug-resistant (MDR) Pseudomonas aeruginosa is an increasing threat world-wide. To anticipate and treat MDR bacterial infections, an in-depth understanding of resistance acquisition in clinical settings is paramount. In this case, we used serial genomic analysis to study in-host evolution of a recurrent P. aeruginosa bloodstream infection in a severely immunocompromised patient. We found that infections were caused by the same strain of MDR P. aeruginosa which acquired de-novo resistance to increasingly broad spectrum, guideline-directed, anti-pseudomonal antibiotics through serial acquisition of a previously undescribed ampD deletion, and a porin mutation, respectively. This study demonstrates both the potential for in-host evolution of bacterial pathogens in immunocompromised hosts undergoing antimicrobial therapy, and highlights the utility of clinical WGS in understanding MDR organism epidemiology and in-host evolution.

Multidrug resistant (MDR) Pseudomonas aeruginosa, as defined by resistance to three or more of the major anti-pseudomonal antibiotic classes, is a global health threat of increasing importance (Magiorakos et al., 2012; CDC, 2019). P. aeruginosa can cause a multitude of infectious syndromes, including pneumonia, bloodstream infections, and wound infections. P. aeruginosa is also an infamous cause of nosocomial infections, because it can transmit effectively within the hospital environment, by contaminating surfaces such as sinks, ultrasound gel, endoscopes, or beds; it is secondary only to Acinetobacter baumanii (also designed a top threat by the CDC) in its ability to contaminate healthcare workers’ gowns and gloves (Morgan et al., 2012). Complementing this ability to transmit in-hospital, P. aeruginosa has a remarkable propensity to acquire resistance through multiple mechanisms, including alteration of porins, efflux pumps, acquisition of beta-lactamases such as ampC or extended-spectrum beta-lactamases (ESBL), carbapenemases, and aminoglycoside nucleotidyl-transferases. Recognizing this growing threat, the Infectious Disease Society recently released an updated dedicated guideline document recommending standardized treatment of treatment of MDR P. aeruginosa and other highly resistant gram-negative bacterial organisms (Tamma et al, 2023).

Key to combating MDR P. aeruginosa is a detailed understanding of its epidemiology and mechanisms of resistance. Patients may acquire MDR P. aeruginosa infection endogenously by in-host evolution of P. aeruginosa under antibiotic pressure, or exogenous infection with MDR P. aeruginosa from another patient, healthcare-workers, or the healthcare environment (Johnson et al., 2009). A study of the acquisition of imipenem-resistant P. aeruginosa in intensive care units between 2001-2006 noted that approximately 50% of imipenem-resistant P. aeruginosa cases were linked to in-patient evolution and 50% were directly acquired from another patient or the hospital environment (Morgan et al., 2012).

The use of whole-genome sequencing (WGS) as part of hospital epidemiology and outbreak investigations holds considerable promise. Our hospital routinely performs WGS to evaluate suspected healthcare transmission of microbes (Crawford et al., 2020), track newly emerging pathogens (Woodworth et al., 2019), and analyze cases of unusual antimicrobial resistance patterns (Hao et al., 2020a). Here, we performed WGS of serially collected P. aeruginosa isolates to study progressive acquisition of resistance despite guideline-directed antimicrobial treatment in an immunocompromised host. We found that in-host evolution rather than exogenous acquisition of an antimicrobial resistance gene was responsible, and involved a novel deletion in ampD and ampE, and a mutation in the porin gene OprD.

A 55-year-old woman with a history of B-cell lymphoma, allogeneic stem cell transplant complicated by graft versus host disease, and renal failure on intermittent hemodialysis, was admitted for sepsis secondary to pan-susceptible P. aeruginosa bacteremia ( Figure 1 ). Direct time to positivity (DTTP) blood cultures drawn through her tunneled dialysis catheter (TDC) and peripherally inserted central catheter (PICC) did not implicate the lines as the source of infection. Both lines were removed on day five, after which her cultures cleared. A transthoracic echocardiogram demonstrated no evidence of endocarditis, and a transesophageal echocardiogram (TEE) demonstrated mitral valve thickening without vegetation. New PICC and TDC lines were placed after 48 hours, and she was discharged on two weeks of cefepime for treatment of bacteremia ( Figure 1 ).

One day after completing her cefepime regimen, she developed fevers and re-presented to the hospital, where blood cultures drawn from her PICC, TDC, and peripheral blood again grew P. aeruginosa, but now with new intermediate resistance to cefepime, as well as resistance to aztreonam, piperacillin-tazobactam and ceftazidime. Even though DTTP blood cultures did not clearly implicate the TDC and PICC as sources of infection, both lines were again removed and later replaced, and she was prescribed a two-week course of meropenem starting from the date of culture clearance. Following clinical improvement, she was discharged home.

Two days after completion of her course of meropenem, the patient developed fevers, chills, and right upper quadrant pain. She re-presented to the hospital for a third time and was found to be growing P. aeruginosa from peripheral blood cultures, and the isolate was now additionally resistant to meropenem. She was started on ceftolozane-tazobactam. A repeat TEE was performed and now revealed a small mobile echodensity on the anterior leaflet of the mitral valve, thought to represent a possible vegetation. An abdominal CT scan demonstrated hepatic micro abscesses that were not amenable to drainage.

In an effort to determine whether her multiple episodes of P. aeruginosa bacteremia with progressively increased drug resistance were due to independent infection events or due to evolution of resistance within a single strain, Illumina whole-genome sequencing (WGS) was performed on DNA extracted from blood culture isolates obtained on days 0, 20 and 44, according to an established protocol (Crawford et al., 2020) ( Figure 1 ). Phylogenetic analysis demonstrated that all isolates represented the same strain, differing by a maximum of 3 single nucleotide polymorphisms (SNPs) across a core shared genome of 6,368,008 nucleotides ( Figure 2A ). This confirmed that resistance had progressively evolved, and ruled-out de novo acquisition of independent infections. Prediction of multilocus sequence type (MLST) from WGS data demonstrated that the isolate was most closely related to sequence type 167, and thus differed from several well-known high-risk international strains (Jolley et al., 2018).

Interrogation of the WGS data was then carried out to identify a molecular explanation for the observed resistance. Intriguingly, a 3,794 base pair deletion involving ampD and ampE was identified in isolates from days 20 and 44 ( Figure 2B ) that coincided with acquisition of new beta lactam resistance and cefepime clinical treatment failure. AmpD represses the expression of the ampC beta lactamase gene, and in vitro deletions in ampD have been associated with augmented expression of ampC and increased beta lactam resistance (Zamorano et al., 2010). Only one prior case of P. aeruginosa resistance attributed to an ampD mutation has been reported, and this involved an ampD missense mutation implicated in the rapid acquisition of resistance to multiple beta-lactam antibiotics (Barbosa et al., 2020).

In addition to the novel deletion, the blood culture isolate from day 44 was also found to contain a C748T nonsense mutation in OprD, paralleling the acquisition of carbapenem resistance during the patient’s secondary treatment regimen with meropenem. OprD encodes a 443-amino acid porin that forms a monomeric 18-stranded β barrel with 9 loops; it interacts with carbapenems and permits their entrance to the cell (Li et al., 2012; Jumper et al., 2021; Varadi et al., 2022). The mutation we identified, C748T, would be predicted to generate a truncated 249-amino acid protein that would most likely be non-functional. Prior work has associated other OprD mutations with carbapenem resistance (Marvig et al., 2013; Wheatley et al., 2021). Increased resistance to ceftazidime also characterized the day 44 isolate, which could potentially be due to OprD loss of function, which has been associated with increased resistance to ceftazidime in a large study of clinical isolates (Castanheira et al., 2014).

The patient was ultimately treated for endocarditis with four weeks of ceftolozane-tazobactam plus ciprofloxacin, according to clinical guidelines (Baddour et al., 2015). Following completion of therapy, the patient has not experienced any further recurrent episodes of bacteremia due to MDR P. aeruginosa.

MDR P. aeruginosa infections can arise in the hospital from direct human to human transmission, environmental acquisition, or from endogenous evolution in the setting of antibiotic selective pressure. While the latter appears to be the least common, it may account for as many as 19% of hospital-onset carbapenem-resistant P. aeruginosa infections (Johnson et al., 2009). The progressive acquisition of increasing resistance in the setting of appropriate anti-pseudomonal antibiotics, however, has not been widely reported. Reflecting this, guidelines recommend a non-carbapenem anti-pseudomonal agent as a first-line approach for susceptible isolates (Tamma et al., 2023).

In this case, initial imaging did not demonstrate discrete foci of infection amenable to source control, or an intravascular infection. It is possible that clinically occult endocarditis or an abdominal infection could have been present at the time of the patient’s initial hospitalization, but was too subtle to be confidently detected by echocardiography or CT scan. It is perhaps more likely that the patient’s underlying immunocompromised state impaired effective clearance of the pathogen, facilitating de-novo evolution of resistance. This phenomenon has been recently well described in the setting of SARS-CoV-2 respiratory infections in severely immunocompromised individuals (Scherer et al., 2022), but is not well described in the setting of bacteremia.

WGS is increasingly becoming a tool of choice for understanding the epidemiology and drug resistance mechanisms of MDR P. aeruginosa (Sundermann et al., 2021). While WGS is an established tool for outbreak investigation and pathogen surveillance (Stanton et al., 2022), its use to inform care for individual patient cases has been more limited. Several case studies have used WGS to identify the mechanisms underlying P. aeruginosa drug resistance in single patients (Alamri et al., 2020; Goldenberger et al., 2020; Hao et al., 2021). Few studies, however, have longitudinally assessed in host evolution using WGS, and those to date have focused exclusively on pulmonary infections (Datar et al., 2021; Wheatley et al., 2021; Rojas et al., 2022). An important recent study described the progressive acquisition, and loss, of carbapenem resistance during a prolonged P. aeruginosa pulmonary infection associated with the acquisition of mutations in oprD and other genes (Wheatley et al., 2021).

This study extends prior in vivo longitudinal genomic analyses of this pathogen to the setting of bacterial bloodstream infections. Furthermore, it highlights the potential clinical utility of WGS for understanding progressive and unexplained drug resistance during treatment, and for disambiguating hospital-acquired infections from cases of in-host evolution. Clinical deployment of WGS is not yet widely available, but may soon offer a sufficiently rapid turn-around time to directly impact clinical care.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/, PRJNA888413.

The studies involving humans were approved by University of California San Francisco Institutional Review Board. The studies were conducted in accordance with the local legislation and institutional requirements. The human samples used in this study were acquired from a by- product of routine care or industry. Written informed consent for participation was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and institutional requirements.

NS, SH, and EG: Equal contributions. NS lead writing, clinical data review and integration. SH lead genomic and computational analyses. EG lead microbiology and whole genome sequencing. AD, HMe, NN, and HMa provided computational analysis and pipeline support. BS and JE advised clinical data review and microbiological interpretation. JLD, SM, and CL oversaw all aspects of study and writing. All authors contributed to the article and approved the submitted version.