Near-fatal Legionella pneumonia in a neonate linked to home humidifier by metagenomic next generation sequencing

Neonatal legionellosis is a rare diagnosis but is important to consider in cases of severe pneumonia, as it portends high mortality without effective antimicrobial coverage. Determination of the source of exposure to Legionella has important implications for preventing further infections. Here, we demonstrate use of two advanced molecular assays for rapid non-invasive diagnosis of legionellosis and environmental source investigation.

An 11-day-old full-term female neonate presented in respiratory distress to an emergency department (ED) at an outside hospital. Her mother (P1G1) had received routine prenatal care with negative infectious screening test results. The patient’s birth history was unremarkable with appearance, pulse, grimace, activity, and respiration (AP-GAR) scores of nine at 1 and 5 min. The patient, who was the only child in the family, was exclusively breast-fed and left home only for visits to the pediatrician office. On the day of presentation, the patient’s parents had noted reduced feeding and blue discoloration of her fingers and toes. In the ED, she was tachypneic, hypoxic, hypercapnic, and hypothermic. Her laboratory tests, including blood culture and PCR for SARS-CoV-2, influenza A/B, and respiratory syncytial virus (RSV), were unremarkable except for c-reactive protein (CRP) 17.9 (normal <10 mg/L), procalcitonin 2.1 (normal ≤0.5 ng/mL), and ferritin 577 (normal 13–150 ng/mL). She was intubated and escalated to high frequency oscillatory ventilation for increasing oxygen requirements. Her clinical diagnosis was consistent with acute respiratory distress syndrome (ARDS). She received one dose each of ampicillin, ceftazidime, gentamicin, and acyclovir intravenously and was transferred to our health system the following day. Timeline of clinical course, antimicrobial therapy, and laboratory results are summarized in Figure S1A.

On arrival, given difficulties oxygenating and worsening ARDS, she was cannulated on venoarterial extracorporeal membrane oxygenation (VA-ECMO). Chest radiograph (Figure S1B) showed complete opacification of both lungs. Her antimicrobial regimen was broadened to ampicillin, meropenem, vancomycin, and acyclovir. Gram stain of endotracheal purulent aspirate showed rare faint Gram-negative rods with no growth on routine aerobic, mycobacterial, or fungal culture. Respiratory virus panel and herpes simplex virus (HSV) PCR on plasma were negative. On hospital day 2 azithromycin was started empirically to cover atypical respiratory pathogens. Given the respiratory failure in the setting of neonatal sepsis with a non-diagnostic infectious work-up, pan-bacterial 16S rRNA PCR (Figure S1C) followed by amplicon sequencing was performed on plasma cell-free DNA (cfDNA) on hospital day 4, which detected a Legionella species with 100% identity to Legionella feeleii. Given this result, Legionella culture was ordered on the endotracheal aspirates, which grew numerous (4+) colonies of Legionella identified as L. feeleii with matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF) from cell extracts. Antibiotic therapy was changed on hospital day 5 to single agent azithromycin based on the predictable susceptibility of Legionella to macrolides. The patient improved significantly by day 3 of a planned 10-day course of azithromycin. She was then successfully decannulated from VA-ECMO on hospital day 9 and discharged home on hospital day 23.

Given that the patient was at risk of reinfection from the same source, we attempted to understand and remove the patient’s source of infection. Legionella is transmitted to humans through inhalation of aerosolized bacteria from environmental water sources.¹ Because the patient had rarely left her home, we focused our investigation on the potential water sources in her home. Legionella culture performed on tap water concentrates from the kitchen and bathroom sink, and water concentrate from an air humidifier in the patient’s room were negative for Legionella. However, three Legionella species PCR assays targeting different sequences were positive on the humidifier water but not the kitchen or bathroom sink waters. This finding prompted the removal of the humidifier from the patient’s home prior to hospital discharge. To investigate whether L. feeleii was the species present in the humidifier water and to decipher its ecology, culture-independent metagenomic next generation sequencing (mNGS) was performed. Processed 150-base pair sequence reads (179,223,943 total reads) from the humidifier water were classified based on k-mer mapping to the Kraken2 mNGS database, containing complete, assembled bacterial, fungal, and viral genomes available in the NCBI GenBank. Overall, 49% of reads were classified to the species level, 0.2% of which were assigned to L. feeleii (Figure S1D). Using another metagenomic profiler (Meta-PhlAn 3.0), which identifies organisms based on homology of reads to a database of highly conserved marker genes, L. feeleii is also detected within the sample. Collectively, these mNGS analyses confirmed the presence of L. feeleii within the humidifier water. To determine if the same strain of L. feeleii was present in the humidifier and recovered from the patient, the metagenome assembled genomes (MAGs) from the humidifier water and the patient’s cultured L. feeleii isolate were compared. The MAG assembled from the L. feeleii isolate sequencing sample had 100% completeness and low contamination (0.19%) using the CheckM bioinformatic pipeline. An L. feeleii MAG recovered from the humidifier water had 87.7% completeness and 3.74% contamination. In addition, another 25 genomes were also recovered from the humidifier water, including Psuedoxanthomonas sp., Sphingobacterium sp., and Mehtyloversatilis universalis, illustrating the presence of a rich ecosystem in the humidifier. The average depth of reads per nucleotide across the genome was 45.4 for the humidifier water versus 1,616.3 for the patient’s isolate (Figure S1E). Using inStrain to perform pairwise comparisons of MAGs, the L. feeleii MAGs from the patient’s isolate and the humidifier had 100% average nucleotide identity (3,449,062/3,449,062 nucleotides were identical between the isolate and the MAG from the humidifier sample). However, 17 nucleotides with biallelic polymorphisms were found in patient and the humidifier MAGs at different allele frequencies (Figure S1F). Altogether, these results provide strong evidence that the neonate acquired her L. feeleii infection from the humidifier.

As Legionella species are typically intracellular organisms that grow in amoeba in their natural environmental niches,² we attempted to identify an amoeba host that L. feeleii may have used for intracellular replication in the humidifier water. Unlike prokaryotic genomes, eukaryotic genomes can be challenging to assemble from mNGS reads because of their large size and low abundance. Thus, sequencing reads from the humidifier were searched for eukaryotic mitochondrial marker genes; this is ideal for the purpose of this investigation because each eukaryotic cell has up to thousands of copies of mitochondrial genome, which lowers the limit of detection. A hidden markov model (HMM) was used to recognize the genes encoding mitochondrial COX1 protein family among assembled contigs. When we used BLAST to align the sequences from the HMM results, we discovered a putative COX1 gene that shared 100% (543/543) amino acid sequence identity with the COX1 protein from Vermamoeba vermiformis, an amoeba species that has previously been reported to serve as host to Legionella species³ (Figure S1G).

Neonatal legionellosis remains a rare infection with a high mortality. It is often nosocomial and linked to specific environmental water sources including humidifiers, water-birthing pools, and others.⁴ The inability of Legionella to grow on routine culture medium delays microbiological diagnosis and contributes to morbidity and mortality reported for neonatal legionellosis. The near-fatal case reported here illustrates the utility of two complementary culture-free nucleic acid tests used to non-invasively diagnose neonatal legionellosis and to culture-independently decipher its environmental source in order to prevent reinfection. First, targeted 16S rRNA PCR amplicon sequencing on plasma cfDNA enabled non-invasive diagnosis and tailored treatment of L. feeleii infection. Targeted sequencing has been in clinical use for some time for culture-free diagnosis of bacterial, fungal, and parasitic infections through detection of microbial genomic DNA in tissue and sterile body fluids.⁵ However, application of targeted sequencing for detection of pathogen-derived cfDNA in plasma, also known as “liquid biopsy,” is relatively new⁶ and is motivated by clinical mNGS performed on plasma and optimization of cfDNA pre-analytical and analytical factors.⁷ Although not used in this case, mNGS of plasma cfDNA has emerged in clinical microbiology for unbiased and non-invasive diagnosis of infections caused by different categories of pathogens.⁸ mNGS is currently offered only by few reference laboratories and its role in diagnosis of infectious diseases needs clarification.⁹ Furthermore, head-to-head comparisons are needed to better define advantages and limitations of mNGS versus targeted sequencing for detection of pathogen cfDNA in plasma.

Exceeding its clinical application, mNGS is commonly used to investigate microbial ecology in biological and environmental samples.¹⁰ This report describes a unique application of mNGS to decipher the source of L. feeleii in the patient’s home. Metagenomic sequencing of the humidifier water sample not only showed the presence of L. feeleii, which was not possible with microbiological culture, it also linked the patient’s infection to the humidifier by confirming the presence of an L. feeleii strain identical to the patient’s isolate. This finding identified and thus triggered elimination of the source of L. feeleii from patient’s home. In addition, mNGS allowed us to identify the diverse community of organisms, including a potential amoebic host, that maintained L. feeleii in the humidifier.

In conclusion, rapid and non-invasive testing of a neonate’s plasma with targeted PCR amplicon sequencing identified a rare Legionella species as the cause of respiratory decline, and mNGS identified the humidifier as the likely source of the pathogen.