Oxacillinase-181 Carbapenemase-Producing Klebsiella pneumoniae in Neonatal Intensive Care Unit, Ghana, 2017–2019

We sequenced 29 carbapenemase-producing Klebsiella pneumoniae isolates from a neonatal intensive care unit in Ghana. Twenty-eight isolates were sequence type 17 with blaOXA-181 and differed by 0–32 single-nucleotide polymorphisms. Improved surveillance and infection control are needed to characterize and curb the spread of multidrug-resistant organisms in sub-Saharan Africa.

axilla, groin and peri-anal region were cultured for gram-negative bacteria on MacConkey agar.
To better understand the role of the environment in the spread of healthcare-associated infections at the NICU, we conducted three environmental screenings of the NICU in September 2017, October 2017, and January 2018. Areas screened included incubator doors, cots, trolley handles, door handles, weighing scales, tables, and desks. Blood cultures were collected by using the BACTEC culture system (Becton Dickinson, Maryland, USA) throughout the study period for all neonates at risk for sepsis or with clinically suspected sepsis. Epidemiologic data were prospectively extracted from clinical notes, including date of sampling for bloodstream infection.
Following the detection of blaOXA-181 carbapenem-producing K. pneumoniae carriage at the NICU, we initiated an outbreak investigation using whole-genome sequencing (WGS) to understand transmission in the NICU. In total, 161 K. pneumoniae isolates were identified from carriage (n = 99), environment (n = 14) and blood cultures (n = 48). All 29 K. pneumoniae isolates with phenotypic carbapenem resistance were included in the sequencing study.

Infection Control Interventions
In response to the outbreak and the discovery of environmental contamination by MDR gram-negative bacteria, three major deep environmental cleaning exercises were conducted besides the routine cleaning at the NICU (October 2017, August 2018, March 2019). These were not part of the initial protocol but deemed necessary by the infection control unit. As part of the original protocol, an alcohol-based hand hygiene intervention using the WHO multimodal hand hygiene strategy (3,5) was instituted at the NICU to improve hand hygiene practice over a 6month period (September 2018-March 2019).

Genome Sequencing and Analyses
We determined genetic relatedness of suspected outbreak isolates by WGS. DNA was extracted and purified with DNeasy Blood and Tissue kit (Qiagen, Hilden, Germany). Isolates were whole-genome sequenced on a MiSeq Instrument (Illumina, San Diego, CA, USA) using paired-end libraries (2× 250bp). Number of reads and total number of bases sequenced per isolate is listed in Appendix Table 1.
We used BacDist to identify single nucleotide polymorphisms (SNPs) in each ST17 isolate relative to Klebsiella pneumoniae strain HS11286 reference genome (RefSeq assembly accession no. GCF_000240185.1) (BacDist: doi: 10.5281/zenodo.3667680) (https://github.com/MigleSur/BacDist). This reference sequence is from a published and fully closed genome of a blaKPC-2 carbapenemase-producing K. pneumoniae (16). BacDist compared SNPs across all isolates to identify those that differed between the isolates, i.e., to identify SNPs that have accumulated since the most recent common ancestor. BacDist filtered mutations to only retain SNPs at positions covered by at least 10 reads in all clones and to exclude mutations where all clones showed more than 80% non-reference reads at the given position. 4 (18)(19)(20)(21)(22), and on initial tests with the following models and priors: substitution models HKY and generalized times reversible clock models strict, relaxed exponential, and relaxed log normal; priors coalescent exponential population and coalescent constant population.
A time-based phylogenetic tree was calculated from a chain of 300 million steps with sampling every 1,000 steps. The first 10% of steps were discarded as burn-in, and effective sample sizes and 95% highest posterior density (HPD) intervals (i.e., an interval in which the modeled parameter resides with 95% probability) were calculated by Tracer version 1.7.1 (23).
We explored the robustness of the BEAST analysis with both respect to use of outgroup (Appendix Table 2) and with respect to different models and priors (Appendix Table 3). First, we  Table 2). Second, we performed the time-based phylogenetic analysis with all 12 possible combinations of the initially tested models and priors (substitution models HKY and GTR; clock models strict, relaxed exponential, and relaxed log normal; priors coalescent exponential population and coalescent constant population). All 12 analyses yielded effective sample sizes (ESS) of all parameters of >2,762 and the time of most recent common ancestor (tMRCA) ranged from year 2017.05 to 2017.09 (Appendix Table 3).
Overall, we obtained similar results across the shown variations with either respect to outgroup (Appendix Table 2) or models and priors (Appendix Table 3), and all variations yielded ESS of all parameters that were adequate according to guidelines for running BEAST (https://beast.community). We have reported results with reference strain HS11286 as outgroup in the main manuscript; nonetheless, we find that all the shown results support the conclusion that the outbreak strains share a recent common ancestor from around early 2017.
Phylogenetic trees were visualized and annotated with metadata using CLC Genomics