The role of Klebsiella populations in preterm infants

As of February 2023, the genus Klebsiella (family Enterobacteriaceae) encompassed 13 species of bacteria with validly published names and one species with a non-valid name. Raoultella spp. are also considered part of the genus Klebsiella (Figure 1) [1,2]. Klebsiella pneumoniae, Klebsiella oxytoca and Klebsiella aerogenes (formerly Enterobacter aerogenes [3]) have received most attention from a clinical perspective, with K. pneumoniae responsible for up to 15% of healthcare-associated infections and increasing levels of antimicrobial resistance being reported for all species [4–7]. Klebsiella spp. are found in the environment, and contribute to the commensal gut microbiota of humans and animals. Gut colonization with Klebsiella spp. contributes to extraintestinal infections in the immunocompromised and clinically vulnerable [7]. K. oxytoca and K. aerogenes, and to a lesser extent Raoultella spp., represent emerging pathogens [4–6].

In recent years, the adoption of whole-genome sequencing in taxonomic, clinical and epidemiological studies has led to an increased understanding of the genetic diversity of Klebsiella spp. (Figure 1). K. pneumoniae has a diverse population structure, representing a complex of five species [K. pneumoniae, K. quasipneumoniae (subsp. quasipneumoniae and similipneumoniae), ‘K. quasivariicola', K. africana, K. variicola (subsp. tropica and variicola)] [7]. The K. oxytoca complex comprises several distinct phylogroups (Ko) defined based on differences in bla_OXY sequences [K. michiganensis (Ko1, Ko5), K. oxytoca (Ko2), K. spallanzanii (Ko3), K. pasteurii (Ko4), K. grimontii (Ko6), K. huaxiensis (Ko8) and three unnamed novel species] [4]. The contribution of K. oxytoca to human clinical infections has likely been over-estimated, with K. michiganensis more prevalent based on retrospective analyses of publicly available genome data [4,8].

Preterm describes infants born prematurely [i.e. <37 weeks gestational age (GA)], and they are often of low birth weight (LBW; <2.5 kg). Approximately 11% of global live births are preterm, and account for a third of all neonate deaths [9]. Intestinal colonization of preterm infants occurs in neonatal intensive care units (NICUs), with empiric antibiotics administered to most, but not all, preterm infants in the first days of life to cover possible early-onset infection from birth contributing to the development of a gut microbiota dominated by Enterobacteriaceae, Enterococcus and Staphylococcus spp. [10,11]. Between 1 and 10% of preterm infants harbour Klebsiella spp. in their faecal microbiota [12], but this proportion can be much higher depending on geographical location [10,13]. Colonization with these, and other opportunistic pathogens, along with an unstable microbiome and systemic developmental immaturity (especially with respect to immune and gastrointestinal functions) contribute to nosocomial infections such as early-onset sepsis (EOS; <72 h after birth), late-onset sepsis (LOS; ≥72 h after birth), necrotizing enterocolitis (NEC; infection and inflammation of the small and large intestines that can progress to necrosis, sepsis and death) in this patient group [12]. Infants born at <32 weeks GA and of very low birth weight (VLBW; <1.5 kg) are particularly susceptible to infection. A recent study in China (Sina-Northern Neonatal Network) looked at the incidence of LOS in 6639 VLBW infants admitted to 35 different NICUs over a 3-year period. From the LOS cases recorded, 456/1511 (30%) positive cultures were obtained from blood: among these K. pneumoniae was the bacterium most often associated with LOS (147/456, 32%) [14]. However, it is clear from other retrospective studies that the proportion of preterm infections that K. pneumoniae contributes to (between 9 and 76%; summarizing data for sepsis, which includes EOS and LOS; Supplementary Table S1) depends on geographical location.

In this review, we aim to summarize the current understanding of Klebsiella spp. in relation to preterm infants and other information regarding Klebsiella that may be relevant to these neonates, and to highlight the need for further research to unravel the role(s) of Klebsiella spp. in this patient population.

What do we know or understand about Klebsiella spp. and their interactions with preterm infants? The answer is very little. This is largely because there has, until recently, been very little interest in or focus on these bacteria in this patient group. Curation of a PubMed and Web Of Science search for original research articles published in the last 5–10 years using the terms ‘Klebsiella' and ‘preterm infants' reveals many available scientific articles are retrospective/observational studies (Supplementary Table S1) [15–50]. Most of these studies do not focus on Klebsiella spp. specifically, instead they include information on a range of bacteria, predominantly ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, Enterobacter spp.), that are known to be associated with nosocomial infections, antimicrobial resistance and/or preterm infants and different neonatal health outcomes. The overriding message is that Klebsiella spp., along with a range of other opportunistic pathogens, contribute to disease in preterm populations globally. No single bacterium has been linked to preterm-associated infections. However, faecal microbiotas dominated by Klebsiella, or Klebsiella and Enterococcus spp. have been described as two of six preterm gut community types, with the latter type more commonly associated with a NEC diagnosis [10].

While retrospective studies are relatively inexpensive to run and allow consideration of multiple outcomes, with outcome data already available, they are not without disadvantages. They do not provide a true understanding of the relationship between preterm infants and different bacterial groups, including Klebsiella spp. It is often not possible to link time of exposure of a preterm infant to a specific bacterium to subsequent infection, and high numbers of patients are required to detect rare associations. Differences in definitions of disease state and inclusion criteria also influence study outcomes [10]. Significant biases may be introduced to retrospective studies based on culture medium and/or cultivation environment selected: many of the documented studies have relied on clinical microbiological data generated using a broad range of microbiological cultivation techniques known to pick up easily cultivated microbes or predominant clinical isolates. To investigate the association/role of Klebsiella spp. with preterm infant health outcomes, study design should include specific focus on Klebsiella, whether that be using appropriate cultivation methods [Klebsiella-specific agar(s) and/or culture conditions] or molecular tools and analyses focused on Klebsiella-specific genetic targets. In addition, such research should involve as wide a range of preterm infants as possible and a range of sample types, not simply those preterm infants with specific clinical outcomes/conditions which necessitate sampling (e.g. blood, sputum, stool) for clinical microbiology aimed at informing subsequent therapeutic regimes and identifying/monitoring outbreaks.

The younger their GA, the more underdeveloped are the organs and immune system of preterm infants. However, as the immune system of full-term infants is also immature at birth, the immaturity of preterm infants’ other systems (including intestinal motility and secretions, digestion, absorption, mucosal surfaces, barrier function and circulatory regulation) also contributes to their increased susceptibility to neonatal infections [51].

Increased exposure to maternal cytokines (e.g. TNF-α, IL-1, IL-6 and IL-8) can have immunomodulatory effects on preterm infants [9,52]. It has been proposed that these may contribute to the protection of VLBW infants against infection and the acceleration of lung maturation [53]. Although the preterm immune system is underdeveloped, the gut of these neonates is highly immunoreactive and has an exaggerated pro-inflammatory response in NEC [51,54]. The lipopolysaccharide receptor TLR-4 contributes to normal development of the small intestine, but in preterm infants its expression is increased compared with full term infants; its in vivo activity can be inhibited by amniotic fluid and breast milk, and reduced by sodium butyrate [54]. TLR-4 of preterm infants is activated in utero and by Gram-negative members of the preterm gut microbiota (Enterobacteriaceae, including Klebsiella spp.): its activation leads to recruitment of pro-inflammatory T helper 17 cells and release of pro-inflammatory cytokines (IL-17, IL-22), eventually leading to erythrocyte death, mucosal injury and translocation of bacteria to the microvasculature underlying the intestinal epithelium [54]. TLR-4 activation also mediates loss of enteric glia, resulting in impaired intestinal motility and hyperinflammation [54]. The cytokine IL-8, which mediates migration and activation of neutrophils to sites of inflammation, is produced by intestinal epithelial cells, and can contribute to necrosis in the preterm gut [51].

The differential exposure of preterm and full-term infants to antibiotic therapy, microbiological and environmental components, immune status and hospital stay influence neonatal microbiome development in these infant groups [55]. Several recent studies have investigated the gut microbiota of preterm infants, in some cases focusing on specific neonatal clinical outcomes (Supplementary Table S1) and in other cases considering diet [synbiotics, prebiotics, probiotics or different milks (breast milk, fortified breast milk, formula milks)] or therapeutic regimes targeting improved clinical outcomes [56–61].

Ho et al. [62] used 16S rRNA gene amplicon sequencing to characterize the faecal microbiota of VLBW infants (n = 45; GA 28 ± 2 weeks; birth weight 1126 ± 208 g) in a NICU (South Florida and Tampa General Hospital, U.S.A.) during the first month of life, sampling each infant at ≤2 weeks, 3 weeks and 4 weeks postnatally. Proteobacteria (46% of total microbiota, mostly Gammaproteobacteria) and Firmicutes (41%, mostly Bacilli) predominated initial samples (≤2 weeks). Actinobacteria and Bacteroidetes were minor components of the gut microbiota of this study cohort. Relative abundance of Gammaproteobacteria increased throughout the study: from 42.5% (≤2 weeks) to 69.7% (3 weeks) to 75.5% (1 month). However, there was wide interindividual variation in carriage of Gammaproteobacteria in neonates initially (0–90%), with two clusters defined based on Gammaproteobacteria abundance (cluster 1, 2.1 ± 5.9%, n = 20 infants; cluster 2, 79.2 ± 21.6, n = 24 infants). Cluster 1 infants were more likely to have been delivered by Caesarean section and had lower birth weight than cluster 2 infants, and seemingly had delayed establishment of Gammaproteobacteria (in terms of abundance) compared with cluster 2. Klebsiella was the most abundant genus observed in the gut microbiota of cluster 2 infants, with a single amplicon sequence variant dominating. A subsequent publication from the same authors examined the relationship between faecal Gammaproteobacteria and faecal calprotectin (a biomarker for mucosal inflammation) in this study cohort [63]. While they again discussed two subgroups of the cohort (clusters 1 and 2), their distinction of clusters was different — based on Klebsiella [63] rather than Gammaproteobacteria abundance [62] (with three infants moving from cluster 2 [62] to cluster 1 [63]). Faecal calprotectin levels were significantly correlated with Klebsiella abundance (r = 0.207, P < 0.05), with cluster 1 having lower faecal calprotectin than cluster 2 (148 vs 226 µg/g stool at ≤2 weeks postnatally, P < 0.05). Correlation does not equate with causation or mechanistic understanding of associations. Therefore, the question arises as to whether mucosal inflammation influences Klebsiella colonization, or vice versa.

Two recent studies characterized the faecal microbiota of preterm infants during the first 4–6 weeks of life to determine early microbiota development [13,64]. Heida et al. [64] demonstrated an initial abundance of Staphylococcus spp., which transitioned to an Enterobacteriaceae-dominated microbiota during the first month postnatally in a weight-associated manner. Delivery mode was shown to influence the initial gut microbiota, with Escherichia and Bacteroides spp. more common in the faeces of vaginally delivered infants. Klebsiella spp. were a common/normal member of the developing neonatal microbiota [64]. Rao et al. [13] collected faecal samples from 178 preterm infants fortnightly during the first 6 weeks of life (at 1, 14, 28 and 42 days), together with near daily sampling from 13 of the cohort. Similar to other studies [10,64], initial dominance of the faecal microbiota with staphylococci was seen, with subsequent transition to a Klebsiella-, Enterococcus- or Escherichia-dominated microbiota with age. By using a scalable multi-kingdom absolute abundance quantification method, ecological modelling and in vitro/in vivo validations, Rao et al. [13] were able to demonstrate the predictability of assemblage of the preterm microbiota, driven by simple microbe–microbe interactions. Establishment of Klebsiella spp. in the gut was facilitated by Staphylococcus spp., with klebsiella suppressing growth of staphylococci. Klebsiella spp. were inhibited by enterococci. The factors produced by staphylococci and enterococci that influence the growth of Klebsiella spp. warrant further attention. Rao et al. [13] also demonstrated that reliance on relative abundance data in microbiota profiling studies can skew findings, potentially masking microbial dynamics (especially Klebsiella and Escherichia ‘blooms’ commonly associated with pre-disease states in preterm infants). In addition, they highlighted that correlation analysis in no way predicted the dynamics or ecological processes underlying development of the preterm gut microbiota. These findings have implications for the wider microbiome research community.

Seki et al. [65] examined multiple data points of extremely preterm neonates (<28 weeks GA) related to the gut microbiota–immune–brain axis. This observational study included 53 infants (n = 15 diagnosed with severe brain injury; n = 38 had no/mild brain injury), with magnetic resonance imaging scans (brain development), peripheral blood samples (immune markers) and regular stool samples (gut microbiota; day 3, day 7 and fortnightly samples week 2 to week 12 after birth) analyzed. Klebsiella spp. contributed to the gut microbiota of extremely preterm neonates, together with nine other genera described as prevalent (found at ≥10⁵ cells in at least 1/5 faecal samples). Klebsiella spp. were less abundant in early stool samples (days 3 and 7) of neonates with severe brain injury, but more abundant (1.7 times higher) at 4 and 6 weeks of age compared with neonates with no/mild brain injury. Elevated Klebsiella abundance was associated with white-matter injury, as were a number of immune-related components in peripheral blood (C-reactive protein, specific T-cell populations, and several cytokines/chemokines) [65]. Mechanistic studies are required to unravel the relevance of these associations with preterm infant health.

Olm et al. [66] investigated the gut microbiota associated with NEC, using shotgun metagenomic analysis of biobanked faecal samples of NEC patients and matched controls. UniFrac analysis of microbiota data did not show distinct clustering of NEC and control neonates. However, K. pneumoniae was enriched in samples from infants subsequently diagnosed with NEC, detected in 52% of pre-NEC samples compared with 23% of control samples. When bacterial replication rates (iRep values) were considered rather than relative abundances from metagenomic data, significantly higher replication rates (total microbiota) were seen in pre-NEC samples compared with controls. Machine learning-informed analyses of metagenomic data identified four aspects of the preterm gut microbiome that differed between pre-NEC and control samples: iRep value (total microbiota), Klebsiella spp., secondary metabolite clusters and fimbriae that could elicit a TLR4-mediated pro-inflammatory response [66]. Similar to Rao et al. [13], these authors highlighted reliance of relative abundance data (without consideration of measures such as iRep) could lead to misleading interpretation of study findings.

Diet-related differences in the faecal microbiota of preterm infants have been observed. Pärnänen et al. [67] found several ESKAPE organisms (including K. pneumoniae), as well as K. oxytoca and Clostridioides difficile, enriched in the faecal microbiota of formula-fed infants compared with breast-fed or fortified human-milk-fed infants. A study focused on feeding intolerance and the gut microbiota of preterm infants suggested that the relative diversity of the gut microbiota significantly decreased in association with a diagnosis of feeding intolerance [68]. A Klebsiella-dominant faecal microbiota was observed for the feeding-intolerance group when this intolerance was diagnosed (stool collected within 24 h of diagnosis), although this may partially reflect the reduced diversity of the microbiota (i.e. loss of species richness rather than an increased abundance of Klebsiella spp.) [68]. However, it should be noted that Klebsiella spp. are lactose-fermenting bacteria — increased abundance of these bacteria in feeding-intolerant preterm infants may be due to microbial catabolism of lactose included in some enteral feeds.

We investigated Klebsiella populations associated with the faecal microbiota in a U.K. preterm cohort (n = 109, <37 weeks GA) [12]. Microbiota profiling (amplicon-based) demonstrated 38.5% of infants harboured Enterobacteriaceae in their first available stool sample after birth. Cultivation work recovered Enterobacteriaceae from 42.2% of the same faecal samples. Multiple species of Enterobacteriaceae were harboured by some infants, while others appeared to have a single predominant species (based on colony morphology and biochemical data from distinctive colony types) [12]. Most infants harbouring lactose-fermenting Enterobacteriaceae (i.e. Klebsiella, Escherichia and Enterobacter spp.) were healthy preterm infants (n = 23), three had a NEC diagnosis during their stay in NICU, there were eight cases of suspected sepsis, one infant had an eye infection and one had an operation (gastroschisis). No common Klebsiella strains were found among the infants in the cohort, with virulence- and antimicrobial-associated differences observed among genome sequences of isolates from each infant, even when they shared sequence, capsule and/or O-antigen types. Eight K. pneumoniae, three K. grimontii, two K. michiganensis and one K. quasipneumoniae strains were isolated. Faecal Klebsiella isolates (from both healthy and sick infants) were able to reside, persist and potentially replicate in macrophages, suggesting they could all evade the host immune system and had the potential to cause opportunistic infections [12]. Preterm infants receive a range of iron supplements (blood transfusions, parenteral feeding, oral) while in NICUs during the first weeks of life. All Klebsiella strains recovered from healthy and sick infants produced siderophores (iron scavengers) in vitro, demonstrating no difference in the colonization or virulence potential of these bacteria [12]. Consequently, much research is required to determine the interactions between neonate and Klebsiella spp. that lead to disease in some preterm infants.

Tilimycin (TM) and tilivalline (TV) are cytotoxic pyrrolobenzodiazepine metabolites (Figure 1). TM, a DNA-damaging agent, is encoded by a biosynthetic gene cluster (BGC) in the genomes of some strains of species belonging to the K. oxytoca complex (specifically K. oxytoca, K. michiganensis, K. grimontii and K. pasteurii); TM spontaneously reacts with indole to form TV, which targets tubulin and disrupts the spindle apparatus of eukaryotic cells [8,69]. These metabolites are causative agents of antibiotic-associated haemorrhagic colitis (AAHC; diffuse mucosal oedema, haemorrhagic erosions, bloody diarrhoea), with disease caused by the overgrowth of cytotoxin-producing strains secondary to the use of antibiotics [70,71]. A recent study in mice has demonstrated the DNA-alkylating metabolite TM causes the accumulation of mutations in cycling intestinal stem cells within weeks of a single K. oxytoca overgrowth, driving somatic changes that could hypothetically contribute to disease susceptibility in some preterm infants who are subject to transient ‘blooms’ of TM-producing bacteria in their gut [72]. We have identified strains of K. grimontii in the faecal microbiota of preterm infants that encode the BGC, along with metagenome-assembled genomes of K. michiganensis and K. oxytoca recovered from faecal samples of preterm infants [8,12]. Around the same time as our study [8], using a m-hydroxybenzoate agar that selects for K. oxytoca-related bacteria over other Klebsiella spp., Paveglio et al. [73] recovered strains from preterm infants (<32 weeks GA) with NEC that could produce both TM and TV (confirmed by mass spectrometry analysis and apoptosis assays). They identified the strains as cytotoxin-producing K. oxytoca based on 16S rRNA, pehX (a marker for K. oxytoca sensu lato) and npsAB (non-ribosomal peptide synthetase genes A and B essential for the synthesis of TM) gene sequence analyses, but they were in fact strains of K. grimontii and K. pasteurii based on reanalysis of their multilocus sequence data [8]. They then used frozen stool samples from NEC and non-NEC matched controls to isolate K. oxytoca from the gut microbiota. Cytotoxin-producing K. oxytoca was isolated from 6/10 and 4/5 of the NEC and non-NEC infants, respectively, with 4/10 and 1/5 harbouring cytotoxin-negative strains of the bacterium. Interestingly, the relative abundance of K. oxytoca from 16S rRNA gene-based amplicon profiling from these infants’ faecal samples differentiated NEC infants harbouring nspAB-positive K. oxytoca (high relative abundance of K. oxytoca) from all other infants (NEC patients with toxin-negative K. oxytoca and control infants, low abundance) [73]. Furthermore, the presentation of NEC symptoms was different for patients in accordance with the timing of high abundance of K. oxytoca relative to NEC onset (high K. oxytoca abundance ‘prior to’ versus ‘near or shortly after’ onset). This may reflect differences in antibiotic treatment regimens (course duration and proximity to NEC onset) used for these cases given the known association of AAHC with antibiotic administration.

As mentioned above, preterm infants are given empiric antibiotics in the first days to weeks of life, and — similar to AAHC — blood in the stool and intestinal necrosis are hallmarks of NEC. The finding of cytotoxin-producing strains of the K. oxytoca complex in the faeces of preterm infants has led to the suggestion that these bacteria could contribute to NEC in a proportion of preterm infants [8,73]. Healthy infants (weeks 0–8 of life) in an Austrian cohort had a carriage rate of K. oxytoca sensu lato of 49–73%, with PCR-based assays leading to higher rates of detection than cultivation work; ∼50% of recovered isolates were cytotoxic, though it was noted that not all npsAB-positive strains produced cytotoxin as assessed by MTT and chemical analyses [74]. A recent study examined a published metagenomic dataset [75] derived from 829 faecal samples from 571 full-term infants born in the U.K. for the presence of K. oxytoca complex bacteria [72]. At days 7 and 21 of life, 76/504 (15%) and 74/325 (23%) samples, respectively, harboured toxigenic Klebsiella spp. Of the complex-positive samples, 46/76 encoded the minimal til locus (BGC) at day 7 (in order of species prevalence: K. grimontii, n = 34/37; K. michiganensis, n = 4/25; K. oxytoca, n = 6/11; K. pasteurii, n = 2/3), with prevalence of each BGC-positive species increasing between days 7 and 21 of life (from 9.1 to 12.6%, n = 41/74). Across all BGC-positive samples, the relative abundance of BGC-encoding bacteria was high (range 0.92–94.1%, median 12.6%) [72]. No robust data are available for the gut carriage of K. oxytoca sensu lato (cytotoxic or otherwise) by preterm infants. Therefore, it is clear the association of cytotoxin-producing Klebsiella spp. with diseases affecting preterm infants is complex and warrants further study. Adoption of the real-time PCR assay for npsAB genes [76], with a sensitivity of 15 cfu/ml of sample, for the detection of potential cytotoxin-producing bacteria in the faeces of preterm infants would allow us to better define the relationship between these pathogens and the presentation of NEC in this cohort. In addition, it is clear that refined molecular-based identification [whole genome (15), shotgun metagenomic [12] or rpoB gene [77] sequencing] should be adopted to accurately identify members of the K. oxytoca complex in the gut microbiota of preterm infants and to correctly identify clinically relevant isolates.

TM can inhibit the growth of Lactobacillus, Bacteroides, Fusobacterium, Proteus and Bifidobacterium spp., members of the commensal gut microbiota [69,78]. In the U.K., probiotic interventions in neonates most frequently involve feeding Bifidobacterium and Lactobacillus-containing multi-strain preparations [57]. Whether TM inhibits and/or influences intestinal colonization with probiotic bacteria remains to be studied. It has been hypothesized that TM confers cytotoxin-producing strains with a competitive advantage over other gut bacteria when in the presence of an appropriate carbon source (e.g. glucose). Support for this suggestion comes from a study done with in vitro systems inoculated with human faeces that showed TM (1–170 µM) exerted broad-spectrum activity against a range of Gram-positive and Gram-negative gut bacteria [79]. In mice TM caused reductions in the species richness and evenness of the murine gut microbiota, driving compositional changes [79]. In addition, and of great concern with respect to the global and preterm burden of antimicrobial resistance in ESKAPE pathogens, TM directly contributed to de novo mutations in the genomes of Escherichia coli and K. pneumoniae strains that led to their resistance to rifampicin and nalidixic acid; a strain of Pseudomonas aeruginosa also acquired a resistance determinant associated with rifampicin upon exposure to TM [79]. No antibacterial effect has been demonstrated for TV to date, either in pure or mixed culture or the murine microbiota [79].

• Klebsiella spp. are often part of the normal gut microbiota of preterm infants, and sometimes contribute to infectious diseases affecting this patient population.

• The current literature is dominated by retrospective studies. Shotgun metagenomic and cultivation-based studies of the preterm faecal microbiota have shown Klebsiella spp. contributing to infections in preterm infants are phenotypically and genotypically diverse.

• There is an urgent need for mechanistic studies focusing on host–Klebsiella interactions to determine why only some preterm infants develop infections caused by these bacteria. In addition, there is a need for Klebsiella–microbiota studies to elucidate the mechanisms contributing to establishment and maintenance of Klebsiella populations in the preterm infant gut microbiota.

The authors declare that there are no competing interests associated with the manuscript.

Open access for this article was enabled by the participation of Nottingham Trent University in an all-inclusive Read & Publish agreement with Portland Press and the Biochemical Society under a transformative agreement with JISC.

Conceptualization, formal analysis, visualization, writing (original draft, reviewing and editing) — A.L.M., L.H.

AAHC

antibiotic-associated haemorrhagic colitis

BGC

biosynthetic gene cluster

EOS

early-onset sepsis

GA

gestational age

LBW

low birth weight

LOS

late-onset sepsis

NEC

necrotizing enterocolitis

NICU

neonatal intensive care unit

TM

tilimycin

TV

tilivalline

VLBW

very low birth weight