Maternal Colonization Versus Nosocomial Transmission as the Source of Drug-Resistant Bloodstream Infection in an Indian Neonatal Intensive Care Unit: A Prospective Cohort Study

Abstract Background Drug-resistant gram-negative (GN) pathogens are a common cause of neonatal sepsis in low- and middle-income countries. Identifying GN transmission patterns is vital to inform preventive efforts. Methods We conducted a prospective cohort study, 12 October 2018 to 31 October 2019 to describe the association of maternal and environmental GN colonization with bloodstream infection (BSI) among neonates admitted to a neonatal intensive care unit (NICU) in Western India. We assessed rectal and vaginal colonization in pregnant women presenting for delivery and colonization in neonates and the environment using culture-based methods. We also collected data on BSI for all NICU patients, including neonates born to unenrolled mothers. Organism identification, antibiotic susceptibility testing, and next-generation sequencing (NGS) were performed to compare BSI and related colonization isolates. Results Among 952 enrolled women who delivered, 257 neonates required NICU admission, and 24 (9.3%) developed BSI. Among mothers of neonates with GN BSI (n = 21), 10 (47.7%) had rectal, 5 (23.8%) had vaginal, and 10 (47.7%) had no colonization with resistant GN organisms. No maternal isolates matched the species and resistance pattern of associated neonatal BSI isolates. Thirty GN BSI were observed among neonates born to unenrolled mothers. Among 37 of 51 BSI with available NGS data, 21 (57%) showed a single nucleotide polymorphism distance of ≤5 to another BSI isolate. Conclusions Prospective assessment of maternal GN colonization did not demonstrate linkage to neonatal BSI. Organism-relatedness among neonates with BSI suggests nosocomial spread, highlighting the importance of NICU infection prevention and control practices to reduce GN BSI.

Worldwide, nearly half of early childhood deaths are in neonates, and up to one-half are due to infection [1]. More than 100 000 neonates die annually from infection in India, the highest absolute burden globally [2]. Neonatal sepsis is up to 10-fold more common in India than in high-income countries and has a mortality rate of up to 50% [3]. Gram-negative (GN) pathogens are the predominant cause of neonatal sepsis in India, where antimicrobial resistance (AMR) is highly prevalent [4,5]. Neonatal sepsis caused by multi-drug resistant (MDR) organisms has limited treatment options and higher mortality, especially among neonates admitted to the neonatal intensive care unit (NICU) [5][6][7][8].
It is unclear whether neonatal infections with AMR GN pathogens are the result of early nosocomial transmission or reflect maternal colonization. Linkage between maternal colonization and subsequent neonatal colonization has been reported in high-income settings where AMR GN neonatal sepsis is uncommon [9][10][11][12]. In low-and middle-income country (LMIC) settings, where the prevalence of these infections is higher, some studies have investigated the linkage between maternal and neonatal colonization with AMR GN pathogens [13][14][15]. However, the role of maternal AMR GN colonization in S38 • CID 2023:77 (Suppl 1) • Robinson et al Clinical Infectious Diseases S U P P L E M E N T A R T I C L E neonatal sepsis risk has not been comprehensively assessed in India. We conducted a prospective cohort study to assess the association between maternal and environmental colonization and neonatal colonization and BSI with AMR GN pathogens in Western India. We hypothesized that pregnant women are colonized with AMR GN pathogens, which are transmitted to their neonates with resultant neonatal colonization and infection.

Study Design and Population
We conducted a prospective cohort study from 12 October 2018, until 31 October 2019, to describe the role of maternal and environmental GN colonization in BSI among neonates admitted to the 60-bed NICU at Byramjee Jeejeebhoy Government Medical College (BJGMC) Sassoon General Hospital, a tertiary care facility in Pune, India. Infection control practices at BJGMC during the study period included oversight by an Infection Control Committee, which included microbiologists and infection preventionists, participation in a Comprehensive Unit-based Safety Program (CUSP) to improve hand hygiene, aseptic technique for invasive procedures, and medication and IV fluid preparation and administration [16], and facility-level antimicrobial stewardship without capacity for prospective audit and feedback. Women admitted to Labor & Delivery with risk factors for delivering children developing neonatal sepsis who provided consent were enrolled. Neonates were followed until discharge, transfer, or death. Inclusion criteria for enrollment were expected preterm delivery (<37 weeks gestation), low birth weight (based on estimated fetal weight), prolonged rupture of membranes (>18 hours prior to delivery), meconium-stained amniotic fluid, history of febrile illness within 2 weeks prior to presentation, or otherwise deemed to be high-risk for neonatal sepsis by clinicians. Neonates from enrolled mothers who were admitted to the NICU were included in Cohort A (Supplementary Figure 1).
A separate cohort study prospectively observed all neonates admitted to the BJGMC NICU to characterize the epidemiology of BSI [5]. Neonates admitted to the BJGMC NICU between 12 October 2018, and 31 October 2019 but not enrolled into Cohort A were included in Cohort B to compare BSI strain relatedness among all neonates admitted to the NICU during the study period ( Supplementary Figures 1 and 2). This study was approved by the BJGMC Ethics Committee, the Indian Council of Medical Research, and the Johns Hopkins Medicine Institutional Review Board.

Assessment for Neonatal Bloodstream Infections
Blood cultures were obtained at the discretion of treating clinicians and processed per routine clinical care. The BJGMC Microbiology Laboratory performs routine microbiology tests, including organism identification and antimicrobial susceptibility testing (AST), and is accredited by the Indian National Accreditation Board for Testing & Calibration Laboratories. BSI was defined as positive blood culture with a known neonatal pathogen and classified as early onset if collection of positive blood culture specimen was on day of life (DOL) 0-2 and late onset if specimen collection was on DOL 3 or later. Isolates were considered "drug-resistant" if they were non-susceptible to third or fourth generation cephalosporins or piperacillin-tazobactam.

Assessment for Maternal, Neonatal, and Environmental Colonization
Colonization with GN organisms resistant to carbapenems or 3rd generation cephalosporins was assessed in maternal and neonatal samples. Maternal vaginal and rectal swabs were collected at time of enrollment and repeated at delivery if >6 hours after enrollment. Staff collected skin and peri-rectal swabs from neonates admitted to the NICU on DOL 0, 3, 7, and weekly until NICU exit. Environmental colonization was assessed with weekly sampling of unit sinks and the immediate neonatal care environment. Samples were collected using the Eswab collection system (COPAN FLOQSwabs, 1 mL Liquid Amies medium), which were aliquoted and frozen at −80°C (Supplementary Methods). For neonates who had BSI, maternal, neonatal, and environmental samples obtained up to the week of BSI onset were thawed. Colonization sample aliquots were evaluated in duplicate with and without enrichment in tryptic soy broth for 4 hours.
Samples were plated onto MacConkey agar plates with meropenem, ceftriaxone, and ceftazidime discs. Representative colonies growing near antibiotic discs were analyzed using VITEK for organism identification and antimicrobial susceptibility testing. Isolates were stored at −80°C pending next-generation sequencing (NGS).

Next-generation Sequencing and its Interpretation
GN isolates from clinical BSI samples and those recovered from associated maternal colonization and environmental swabs were selected for sequencing. DNA was extracted from bacterial isolates using the QIAamp DNA Mini Kit. DNA extraction quality was assessed using Qubit and QIAXpert. NGS was performed using Illumina HiSeq with a read length of 2 × 150 bp and target sequencing depth of 100×. After quality control, de novo gene assembly was performed (Supplementary Methods). AMR genes were identified using the ResFinder [17], PointFinder [18], Comprehensive Antibiotic Resistance Database (CARD) [19,] and ARG-ANNOT [20] databases. Core genomes among isolates of the same species were assembled to visualize phylogenetic relationships with maximum likelihood trees [21]. Species represented by fewer than 4 isolates were characterized by hierarchical clustering of pairwise single nucleotide polymorphism (SNP) distance. Evaluations of strain relatedness may produce inconsistent results dependent on the subjective choice of reference genome or parameters for core genome construction [22]. Strain relatedness was therefore evaluated by quantifying pairwise SNPs between each pair of isolates of the same species without requirement for reference genome selection or core genome creation [23] and displayed as distance matrices and force-directed network graphs.

RESULTS
The study recruited 1051 women admitted to Labor & Delivery with neonatal sepsis risk factors (Supplementary Figure 1) 1 and  Supplementary Table 1). Antepartum antibiotics were administered to 737 (79%) of women with live births. The most common antibiotics administered were metronidazole, cefotaxime, and ampicillin for both cesarean and vaginal delivery. During the recruitment period for Cohort A, there were 1301 additional neonates admitted to the NICU who were not born to mothers who had been prospectively enrolled into Cohort A (Supplementary Figure 2).

Neonatal Characteristics
Among 257 neonates admitted to the NICU in Cohort A, median gestational age was 33.9 weeks (IQR 31.9-36.0), and median birth weight was 1700 grams (IQR 1500-2200). Seventy-eight (30.4%) were delivered via cesarean delivery (Table 1)  Clusters of closely related K. pneumoniae BSI isolates were obtained from neonates who were hospitalized within 1 month of another neonate in the same cluster ( Figure 2). All 4 Escherichia coli isolates were of different sequence types (361, 1193, 38, and 73; Supplementary Figure 3A). There were four isolates identified as Burkholderia cenocepacia; all identified as ST 824 (Supplementary Figure 3B) and had SNP distances of ≤4 to the nearest neighbor ( Figure 3). Two of the 3 Acinetobacter baumannii isolates were ST 25 and had a SNP distance to each other of 8; the other was ST 85 (Figure 3 and Supplementary Figure 3C).  Figure 3).

Strain Relatedness Including Maternal and Environmental Samples
Two maternal rectal K. pneumoniae isolates resistant to ceftriaxone or meropenem and underwent NGS were found to have sequence types 551 and 34 and did not match any sequenced neonatal BSI isolates. There were 2 K. pneumoniae ST 14 isolates recovered from sink traps, the most common ST encountered in BSI isolates; their nearest neighbor among ST 14 BSI isolates by pairwise SNP distance were 1 and 15 ( Figure 3).

Identification of AMR Genes and Plasmids
Among BSI isolates, NDM-1 was found in 16 K. pneumoniae isolates, 1 A. baumannii isolate, and 1 E. hormaechi isolates (Figure 2, Supplementary Figure 3). NDM-5 was identified in 1 K. pneumoniae BSI isolate. NDM-1 was found in all sink isolates with available NGS data-2 K. pneumoniae and 1 A. baumannii isolates. No NDM were found in any sequenced maternal samples. OXA-232 was found in 6 K. pneumoniae BSI isolates and 1 K. pneumoniae sink isolate; OXA-420 was found in 1 A. baumannii BSI isolate. The mcr-9 gene was detected in 3 K. pneumoniae BSI isolates and 1 A. baumannii sink isolate. The most detected plasmid groups in Enterobacterales were IncFII and IncFIB (Supplementary Table 3).

DISCUSSION
Prospective evaluation of maternal colonization prior to delivery and observation for subsequent neonatal BSI did not find evidence of maternal colonization as the source for AMR GN organisms in a cohort of Indian neonates. Most neonatal BSI isolates, however, showed a high degree of strain relatedness to other neonatal BSI occurring during a similar timeframe, suggesting that nosocomial transmission fuels AMR in this setting. A high degree of strain relatedness among neonatal BSI isolates was also found in secondary analysis of a large multinational cohort of neonatal BSI in LMICs [24]. These findings suggest that nosocomial rather than vertical transmission may be the primary driver of drug-resistant GN BSI in this setting. Among 21 neonates with GN BSI, no isolates matched the species and resistance profile of bacteria isolated in maternal samples, suggesting that maternal colonization is unlikely to be the primary source of AMR in neonatal GN BSI in our study. Prior efforts have not consistently identified mother-child transmission of drug-resistant GN colonization in LMICs. A meta-analysis reported a MDR GN transmission rate of 27% from colonized mothers to neonates [25], but only 2 of 8 studies were conducted in LMICs and none in India. A study of very low birth weight (VLBW) inborn neonates in India showed that by the first week of life 68% were colonized with extended-spectrum-beta-lactamase (ESBL)-producing GN bacteria and 5% with carbapenem-resistant organisms (CRO), but no association between maternal and neonatal colonization was identified [15]. Studies in Sri Lanka have shown concordant Enterobacteriaceae colonization in <10% motherneonate pairs and that most neonatal isolates are not clonally related to maternal isolates [13,14]. The Burden of Antibiotic Resistance in Neonates from Developing Societies (BARNARDS) study recently reported eight clonal carbapenem-resistant isolate pairs in maternal and neonate gut microbiota samples, suggesting that mother to neonate transmission of colonizing resistant organisms does occur infrequently, although the study did not link colonization directly to BSI [26].
Lack of evidence of mother-to-child transmission of GN organisms causing BSI suggests that neonates must acquire these organisms during hospitalization. We found that among 37 GN BSI cases with available sequencing data, more than half showed fewer than 5 SNPs to another BSI case, suggesting likely association with at least 1 other case of GN BSI during the study period. Genetic similarity between isolates from neonates with GN BSI supports nosocomial transmission as the primary driver of GN BSI. Nearly all K. pneumoniae BSI isolates showed genetic similarity to isolates obtained from other neonates with K. pneumoniae BSI hospitalized within the same month. K. pneumoniae is an organism of global concern and has been linked to outbreaks in LMIC NICUs [27][28][29]. The National Institute for Health Research Global Health Research Unit (GHRU) on Genomic Surveillance of AMR has prioritized sequencing of K. pneumoniae isolates in consortium countries, including in India. The K. pneumoniae ST identified in this study were among the most common ST found at the GHRU surveillance site in another part of India [30].
Although other species of GN causing BSI had fewer isolates to compare in our cohort, clusters of closely related B. cenocepacia and A. baumannii were identified, which in aggregate with K. pneumoniae isolates constituted the majority of observed and analyzed BSI. Although worldwide surveys of GN bacterial strain types do show regional and national clustering of strain types, the SNP distance of clusters of GN BSI in this study show closer relatedness than expected by regional clustering [24,31]. Close genetic similarity between isolates presented here demonstrates that nosocomial transmission is common in this setting. The BARNARDS study also reported related clusters of neonatal gut colonization [26] and BSI, suggesting that nosocomial transmission is common in other LMIC facilities. Study of the maternal and neonate gut microbiota in the BARNARDS study revealed clusters of clonal carbapenem-resistant isolates among neonatal samples and some maternal-neonate pairs [24].
Given our cohort's predominance of late-onset BSI and clustering of neonatal BSI isolates, AMR GN pathogen transmission likely occurs primarily within the NICU. Prevention of NICU healthcare-associated infection (HAI) relies on strong infection prevention and control (IPC) policies and practices [32]. In the NICU, special considerations for IPC exist, due to a particularly vulnerable population, as well as specialized equipment and practices unique to neonates [33]. In LMIC settings, additional challenges include unreliable water supply, reuse of single-use equipment, insufficient personal protective equipment and capacity for transmission-based isolation, high patient-to-staff ratio, and overcrowding of healthcare facilities [33,34]. Strategies recommended by the World Health Organization to reduce risk of nosocomial transmission of AMR GN pathogens include implementation of multimodal IPC strategies, including hand hygiene, AMR surveillance, use of contact precautions and patient isolation, and environmental cleaning, although capacity is resource-dependent [35][36][37][38].
Efforts to reduce neonatal sepsis mortality in LMICs have been constrained by the fundamental knowledge gap of whether maternal colonization or nosocomial spread is responsible for the high burden of AMR in GN BSI. The concern that maternal colonization serves as the reservoir for AMR in subsequent neonatal GN BSI is driving investment in preventative efforts including investigation of maternal vaccination to reduce colonization with AMR organisms such as K. pneumoniae [39]. The findings from this study suggest that interventions to reduce maternal AMR colonization may have limited impact in settings where maternal rectal or vaginal colonization is not the chief driver of these infections.
Performance at a single site may limit our study's generalizability. Further study is needed in larger populations to determine the relationship between maternal colonization and neonatal sepsis, including in non-hospital settings. The frequency of neonatal sampling could have missed mother-to-neonate colonization transmission. The small number of neonatal GN BSI cases were insufficient to address other factors such as antibiotic use which may impact the risk of a neonate developing an AMR infection. Limited bacterial recovery in environmental samples other than sink traps fails to reveal the environment as a transmission reservoir. The study includes relatively few BSI cases compared to adult studies, but unique risk factors and infection control challenges for BSI prevention in neonates justifies focused study of this vulnerable population. Study strengths include prospective enrollment of mothers prior to delivery and evaluation of strain relatedness among multiple species and sequence types. Most reports of strain relatedness occur in the context of outbreak investigation and do not characterize strain relatedness in the broader context of BSI.
To reduce the global burden of neonatal sepsis and associated mortality due to AMR GN pathogens, elements of healthcare delivery responsible for nosocomial transmission must be identified to guide prevention. Our study did not show evidence of mother-to-child transmission of drug-resistant GN organisms causing BSI but did show that nosocomial transmission is a significant factor in transmission of these organisms, including K. pneumoniae, a major cause of neonatal sepsis in India and other LMIC settings. IPC initiatives must be prioritized to reduce HAI among hospitalized neonates in LMIC settings.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.