Carbohydrate competition by Enterobacteriaceae enhances colonization resistance to carbapenem-resistant hypervirulent K. pneumoniae

Abstract

Background

Carbapenem-resistant hypervirulent K. pneumoniae (CR-HvKP) is a growing public health threat due to its virulence and limited treatment options. While prevalent in hospitals, its presence in livestock, particularly pigs, is poorly understood. The gut microbiome provides colonization resistance, but how it restricts CR-HvKP remains unclear.

Results

To further elucidate the colonization resistance mechanisms of the gut microbiota against CR-HvKP, we analyzed stool samples from piglets (L), nursery (N), fattening (F), and sows (PS) using microbiome modeling (Micolo) and competition assays. ST290 K. pneumoniae isolated from PS inhibited CR-HvKP via carbohydrate competition, with a pronounced effect observed for sucrose. Niche-specific supplementation with methyl pyruvate was found to partially alleviate this ecological inhibitory effect.

Conclusions

Carbohydrate-based interventions could be explored as potential therapeutic or prophylactic strategies to combat CR-HvKP colonization, thereby potentially improving animal and public health outcomes.

Download video file ^{(46.6MB, mp4)}

Video Abstract

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s40168-025-02245-0.

Introduction

Carbapenem-resistant Klebsiella pneumoniae (CRKP) has emerged as a leading cause of nosocomial infections, contributing to high morbidity and mortality rates and posing a significant public health threat [1, 2]. The acquisition of a hypervirulent plasmid by a CRKP strain, or the transfer of a carbapenemase-encoding plasmid into a hypervirulent strain, leads to the emergence of carbapenem-resistant hypervirulent K. pneumoniae (CR-HvKP), posing an even greater threat to both human and animal health [3]. The CR-HvKP isolates were initially identified in the Asia–Pacific region; however, they have since been reported with increasing frequency worldwide, including in Europe, South America, Africa, and North America [4–7]. CR-HvKP has been reported in healthcare settings, where it causes higher mortality rates, indicating its potential to spread rapidly in clinical environments and trigger fatal outbreaks [8, 9]. This heightens the apprehension that CR-HvKP may emerge as the next multidrug-resistant "superbug" pathogen, presenting a significant threat to immunocompromised patients.

For CR-HvKP, colonization of the gut is typically the first step toward subsequent invasive infections [10, 11]. The normal resident bacteria of the gut can protect the host against colonization by CR-HvKP, a phenomenon known as "colonization resistance" [12]. Decolonizing CR-HvKP from the gut of hospitalized patients could prevent infections [13]; however, antibiotic-mediated decolonization has been associated with the development of resistance and the selection of other resistant Gram-negative organisms [14]. Recent advances in decolonization strategies have increasingly focused on modulating gut microbiota. For instance, fecal microbiota transplantation (FMT) has been successfully employed to eliminate carbapenem-resistant Enterobacteriaceae (CRE) colonization in infected individuals [15, 16]. Furthermore, the application of engineered microbial consortia has demonstrated efficacy in controlling ecological niches, thereby restoring colonization resistance [17, 18]. Additionally, treatment with specific Klebsiella species has shown promise in combating CRE infections through gluconate competition [19, 20]. Multiple therapeutic agents based on processed fecal microbiota are currently under investigation for the treatment of recurrent infections [21, 22]. Notably, certain defined bacterial consortia, including VOWST, have already received approval from the US FDA [22]. Collectively, these findings highlight the potential of microbiome-based interventions as promising strategies for eradicating CR-HvKP in the intestinal environment.

Current epidemiological evidence indicates that CR-HvKP demonstrates widespread environmental presence, with confirmed detection in multiple reservoirs, including hospitalized patients, retail vegetables, and companion animals such as domestic dogs. However, surveillance data reveal significantly lower detection frequencies in porcine populations [1, 23], suggesting potential species-specific barriers to colonization or ecological establishment. Diverse isolated commensal bacteria play a significant role in promoting pathogen clearance. Given the significant influence of microorganisms on host physiology and the immune system [24], the gut microbiota may play a critical role in driving decolonization capacity in pigs. However, the specific characteristics of the pig gut microbiota at different ages and their effects on decolonization against CR-HvKP remain poorly understood. Elucidating how the microbiome inhibits CR-HvKP colonization is fundamental for developing effective preventative strategies against infection. Recent studies have proposed several mechanisms by which resident bacteria might protect against Enterobacteriaceae pathogens, including the production of inhibitory molecules, competition for nutrients, and induction of the immune response [17, 20, 25–29]. Unraveling these mechanisms could lead to novel microbiome-based approaches to restrict CR-HvKP colonization and subsequent infections. However, the mechanisms by which pig-derived resident bacteria prevent CR-HvKP colonization are still unclear.

To address this, we evaluated the gut microbiomes of suckling piglets (L), nursery pigs (N), fattening pigs (F), and pregnant sows (PS) to identify protective microbes and understand the strategies these microbes employ to confer colonization resistance against CRKP/CR-HvKP. Our findings demonstrate that ST290 K. pneumoniae derived from PS confers robust colonization resistance against multiple CR-HvKP strains, both in vitro and in vivo. We further showed that this effect is mediated, at least in part, through competitive exclusion for carbohydrates. These results highlight the importance of interactions involving resident bacteria in developing microbiome-based therapies aimed at preventing infections caused by CR-HvKP.

Results

Gut microbiota from pregnant sows (PS) confers potent colonization resistance against carbapenem-resistant hypervirulent K. pneumoniae (CR-HvKP)

Previous studies have employed co-culture assays involving intestinal microbial communities and pathogens to investigate their competitive interactions [18, 20, 30]. By quantifying pathogen levels, this approach facilitates the assessment of the gut microbiota's capacity to suppress pathogen growth. This method is distinguished by its strong controllability, high reproducibility, cost-effectiveness, and reduced dependence on animal experiments. Notably, Éva d. H. Almási et al. utilized luminescence labeling of pathogens, which significantly improved the sensitivity of pathogen quantification and enabled real-time monitoring of pathogen dynamics [31]. For this, in vitro models capable of preserving the functional and compositional profiles of in vivo gut microbiomes would be highly valuable. Mipro broth, which supports the growth of a wide range of microorganisms, including aerobic, anaerobic, and facultative anaerobic bacteria, has demonstrated enhanced performance in maintaining the viability, diversity, and compositional and functional profiles of the inoculum microbiome [32]. This makes it particularly suitable for cultivating complex microbial communities, such as those found in the gut. Building on previous research, we first employed Mipro broth as the microbial incubation medium to develop a streamlined colonization evaluation model, named the Micolo model (Fig. 1a). This model enables the assessment of the colonization ability of targeted strains both in vitro and in vivo through bioluminescent measurements. In brief, we constructed an IS26-based transposon-luxCDABE system (p26-RP4-Lux) to tag the luxCDABE gene cluster in the CR-HvKP strain, leveraging the transposition characteristics of IS26. Through conjugation experiments, the luxCDABE gene cluster was successfully tagged onto the hypervirulent plasmid (pVH1-2-VIR) in CR-HvKP VH1-2, forming VH1-2-Lux (Fig S1, Appendix 2).

Fig. 1.

Gut microbiota from pregnant sows (PS) confers potent colonization resistance against carbapenem-resistant hypervirulent K. pneumoniae (CR-HvKP). a Schematic workflow of bioluminescence-based ecological invasion assay. The luxCDABE-tagged CR-HvKP VH1-2 (VH1-2-Lux) was co-incubated with intestinal microbiota derived from four porcine developmental stages: suckling piglets (L), nursery pigs (N), fattening pigs (F), and PS. Pathogen proliferation was quantified through both bioluminescence intensity and colony-forming unit (CFU) enumeration. b Comparative luminescence signals of CR-HvKP pre- and post-exposure to gut microbiota from different swine groups (L: n = 59; N: n = 41; F: n = 91; PS: n = 91). c Corresponding CFU counts of CR-HvKP before and after microbial challenge across experimental groups. d Validation of microbiota functional activity through thermal inactivation (boiling for 20 min). e Differential bioluminescence signals following co-culture with viable versus heat-inactivated PS microbiota (n = 3). f VH1-2-Lux CFU quantification confirming functional requirement of live microbiota (n = 3). g Proposed actor of PS microbiota-mediated colonization resistance. ns = not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

To explore the decolonization potential of intestinal microflora from pigs, fresh fecal samples from suckling piglets (L), nursery pigs (N), fattening pigs (F), and pregnant sows (PS) were subsequently inoculated with VH1-2-Lux and incubated under anaerobic conditions. Compared to control group, a reduction in the bacterial load of VH1-2-Lux was observed across all four treatment groups. Notably, a significant decrease in bioluminescence and CFU counts was observed in the PS samples, indicating robust colonization resistance against VH1-2-Lux (Fig. 1b-c). Fecal samples contain not only live bacteria but also metabolites produced during fermentation. To identify the key component mediating decolonization, nine fecal samples were subjected to heat inactivation, and the decolonization assay was repeated (Fig. 1d). Bioluminescence of VH1-2-Lux was significantly reduced after co-incubation with normal PS samples but not with heat-inactivated samples (Fig. 1e), which was corroborated by colony counts (Fig. 1f). These data suggest that the colonization resistance is dependent on live bacteria from PS (PS flora, Fig. 1g).

Gut microbiota of PS exhibits significant enrichment of Clostridiaceae and Enterobacteriaceae

To identify microbiome changes associated with colonization resistance against CR-HvKP and pinpoint commensal bacteria critical for this resistance in PS, we analyzed the microbiota of L, N, F, and PS using 16S rRNA gene sequencing. The OTU Venn analysis revealed 1,682, 1,565, 1,633, and 3,680 unique OTUs in the L, N, F, and PS groups, respectively (Fig. 2a). Alpha diversity, assessed using the Faith and Simpson indices, showed no significant differences in microbial diversity among all groups (Fig. 2b-d). Beta diversity analysis using Bray–Curtis dissimilarity, visualized through principal component analysis, indicated significant differences in microbiota composition between the groups (Fig. 2f).

Fig. 2.

Gut microbiota of PS exhibits significant enrichment of Clostridiaceae and Enterobacteriaceae. a Observed OTU counts, b Faith's phylogenetic diversity and c Simpson index across experimental groups. d Principal coordinates analysis (PCoA) of microbial communities based on Bray–Curtis dissimilarity. e Linear discriminant analysis Effect Size (LEfSe) identifying differentially abundant bacterial families. Relative abundance of (f) Lactobacillaceae, (g) Streptococcaceae, (h) Clostridiaceae, and (i) Enterobacteriaceae. n = 10 per group; ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001*

To identify bacterial taxa associated with colonization resistance against CR-HvKP, we conducted LEfSe analysis at the family level (Fig. 2g). The PS group exhibited significant alterations in gut microbiota composition characterized by a marked depletion of Lactobacillaceae coupled with concurrent enrichment of Clostridiaceae and Enterobacteriaceae, as compared to control groups (p < 0.05). No statistically significant difference in Clostridiaceae abundance was observed between PS and F groups (Fig. 2h–k). This suggests that an increase in Clostridiaceae and Enterobacteriaceae abundance may be associated with colonization resistance against CR-HvKP colonization.

Enterobacteriaceae species are essential for colonization resistance against multiple CR-HvKP strains in vitro

To investigate microbiota-mediated colonization resistance against CR-HvKP in the PS, we first performed controlled co-culture assays in Mipro broth using C. butyricum S1 (a probiotic strain) and C. sordellii S2 (an opportunistic pathogen) with CR-HvKP VH1-2-Lux. CFU enumeration revealed enhancement of CR-HvKP colonization in the presence of C. butyricum S1 (2.77-fold increase vs. monoculture, p < 0.05) and C. sordellii S2 (2.74-fold increase, p < 0.05) (Fig S2). These results suggest that Clostridiaceae members may not mediate CR-HvKP suppression in PS, and point instead to members of Enterobacteriaceae as potential antagonists against CR-HvKP colonization. Then, we isolated 13 strains of Enterobacteriaceae from PS. These isolates, along with E. coli Nissle 1917, were Subjected to testing in both Mipro and LB broth. In Mipro broth, both bioluminescence and CFUs exhibited a significant reduction, exceeding a tenfold decrease, following co-incubation with the 14 Enterobacteriaceae strains (p < 0.01, Fig. 3a). In LB broth, E. coli Nissle 1917 exhibited a consistent protective effect comparable to that observed in Mipro broth. Aside from E. coli Nissle 1917, the other eight strains (GX2-13, GX2-22, GX2-24, GX2-28, GX3-1, GX3-10, GX3-20, and GX3-23) significantly reduced the VH1-2-Lux bacterial load by more than tenfold (p < 0.01, Fig. 3b).

Fig. 3.

Enterobacteriaceae species mediate colonization resistance against diverse CR-HvKP and CRKP strains in vitro. a Real-time bioluminescence kinetics and CFU quantification of CR-HvKP strain VH1-2-Lux co-cultured with PS-derived Enterobacteriaceae isolates in Mipro medium. b Comparative CFU counts of VH1-2-Lux after co-incubation with Enterobacteriaceae in nutrient-rich LB medium. c Antimicrobial efficacy against six clinical CR-HvKP/CRKP strains, with E. coli Nissle 1917 (1917) serving as a probiotic control. d Enterobacteriaceae species protect against both hypervirulent and conventional CRKP lineages. *p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001

Furthermore, we examined whether the Enterobacteriaceae strains isolated from PS could confer protection against other CR-HvKP or CRKP strains. For this purpose, VH1-2-Lux was substituted with three CR-HvKP strains (DH1, DH28, DH36) and 3 CRKP strains (DH6, DH19, DH20). Each strain was individually co-incubated with five Enterobacteriaceae strains isolated from PS (GX1-10, GX2-22, GX2-28, GX3-10, and GX3-23), in addition to E. coli Nissle 1917. GX1-10 significantly decreased the bacterial load of three CRKP strains and two CR-HvKP strains, whereas GX3-23 notably reduced the bacterial load of three CR-HvKP strains and two CRKP strains. GX3-10 demonstrated a significant reduction in the bacterial load of CRKP DH19 and two CR-HvKP strains (DH28 and DH36). Conversely, GX2-28 only significantly decreased the bacterial load of CRKP DH6. Similar to E. coli Nissle 1917, GX2-22 significantly reduced the bacterial load of all tested CRKP and CR-HvKP strains (p < 0.01, Fig. 3c). The findings suggest that GX2-22 may exert a broad decolonization effect not only against CR-HvKP but also against various CRKP strains.

Therapeutic administration of GX2-22 isolated from PS enhances gut clearance of CR-HvKP

Given the demonstrated colonization resistance of isolated Enterobacteriaceae strain GX2-22 against CR-HvKP in vitro, we further investigated whether this inhibitory effect could be replicated in an in vivo model system. Previous studies have shown that antibiotic treatment in mice induces microbiome changes, facilitating intestinal colonization by opportunistic pathogens [33]. In this study, we established streptomycin-treated mouse model in which mice received the streptomycin over four days (Fig. 4a). Streptomycin treatment significantly depleted major Enterobacteriaceae (p < 0.05, Fig. 4b), which allowed high levels of CR-HvKP colonization (Fig. 4c). VH1-2-Lux counts were below the detection Limit after 1 day and increased approximately to the 10⁵ CFU/g at day 2 after infection (Fig. 4d). In vivo live imaging confirmed that the bioluminescence signal in mice was significantly higher in the streptomycin-treated group than in the control group (p < 0.05, Fig. 4e-f). These results imply that streptomycin-sensitive Enterobacteriaceae might contribute to limiting CR-HvKP colonization.

Fig. 4.

Therapeutic administration of GX2-22 isolated from PS enhances gut clearance of CR-HvKP. a Streptomycin pre-treatment protocol for murine colonization assays: C57BL/6 mice received 5 g/L streptomycin in drinking water for 4 days, followed by oral gavage with 10⁷ CFU/mL CR-HvKP (VH1-2-Lux, ST23). Fecal samples were collected serially to monitor pathogen burden. b Comparative CR-HvKP colonization levels in streptomycin-treated vs untreated mice at day 4 post-inoculation (n = 4). c Representative bioluminescence imaging of bacterial culture plates from fecal samples. d Temporal dynamics of CR-HvKP fecal shedding from day 1 to 4 post-treatment. e In vivo bioluminescence signals localized to cecal regions on day 4. f Quantification of cecal radiance. g Mice pre-treated with streptomycin received CR-HvKP VH1-2-Lux co-administered with sterile water (Con), probiotic E. coli Nissle 1917 (1917), or K. pneumoniae GX2-22 (GX2-22). h Fecal abundances of VH1-2-Lux in control and intervention groups (n = 8; *p < 0.05)

To evaluate the protective role of Enterobacteriaceae against CR-HvKP colonization, we utilized a streptomycin-treated murine model. Following streptomycin administration, mice were inoculated with a mixture of VH1-2-Lux and either GX2-22 or E. coli Nissle 1917 (EcN 1917) (Fig. 4g). Microbiota analysis revealed distinct community structures across groups (Control: 111 OTUs; 1917: 108 OTUs; GX2-22: 112 OTUs) (Fig.S3a), with GX2-22 treatment significantly reducing alpha diversity (Simpson indices, p < 0.05) (Fig.S3b). Beta diversity analysis (weighted/unweighted UniFrac PCoA) confirmed compositional shifts (Fig.S3c-d), including decreased Clostridiaceae but increased Lactobacillaceae and Tannerellaceae (p < 0.05) (Fig.S3e-g). LEfSe identified genus-level differences: controls were enriched in Coriobacteriaceae_UCG_002, Clostridium, Monoglobus, Terrisporobacter and achnospiraceae_UCG_006, whereas GX2-22-treated mice showed higher abundances of Ligilactobacillus, Parabacteroides, Lachnoclostridium and Romboutsia (LDA score > 2.0, Fig S3h). As expected, 1917 reduced VH1-2-Lux gut colonization (p < 0.05 vs control).Notably, GX2-22 achieved comparable Suppression as that of 1917 (Fig. 4h), demonstrating its efficacy in accelerating CR-HvKP clearance. These results indicate that therapeutic intervention with GX2-22 isolated from PS accelerates the clearance of CR-HvKP from the gut.

Protective Enterobacteriaceae strains suppress CR-HvKP through carbohydrate competition

The gut microbiota confers protection against Enterobacteriaceae pathogens through multiple mechanisms, including the production of antimicrobial compounds (e.g., bacteriocins), nutrient competition, and immune system modulation [17, 20, 25–29]. To elucidate the protective mechanisms of Enterobacteriaceae strains GX2-22 and 1917 against VH1-2-Lux, we first employed an agar well diffusion assay to evaluate the antibacterial activity of their cell-free supernatants (CFSs). As demonstrated in Fig. 5a, colistin (positive control, 128 μg/mL) exhibited significant inhibitory activity against VH1-2-Lux (zone of inhibition: 15.02 mm), whereas CFSs from both GX2-22 and 1917 showed no detectable antibacterial effects (zone of inhibition: 0 mm). Furthermore, growth curve analysis revealed that supplementation with CFSs from either strain failed to suppress VH1-2-Lux proliferation (Fig. 5b and S4). These collective data demonstrate that the competitive exclusion of VH1-2-Lux by protective strains is not mediated by secreted inhibitory molecules (e.g., bacteriocins). Instead, we hypothesize direct niche competition mechanisms.

Fig. 5.

Protective Enterobacteriaceae strains suppress CR-HvKP through carbohydrate competition. a Quantitative evaluation of anti-CR-HvKP activity in cell-free supernatants (CFSs) using VH1-2-Lux strain, determined by agar diffusion assay with inhibition zone diameter quantification. b Time-course growth inhibition kinetics of VH1-2-Lux exposed to GX2-22-derived CFSs. c Phylogenetic classification of Enterobacteriaceae isolates based on whole-genome phylogeny. d Comparative genomic analysis of carbohydrate-active enzyme (CAZyme)-encoding genes across all isolates. e Venn diagram depicting shared genomic features between GX2-22 and VH1-2-Lux. f In vitro co-culture competition dynamics under various carbon conditions: monosaccharides, disaccharides, oligosaccharides, glucosides, and sugar alcohols. g Sucrose utilization capabilities of GX2-22, VH1-2-Lux, and their dual-species co-culture. h Carbon source competition assay using methyl pyruvate as VH1-2-Lux-specific carbon substrate. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. n = 3

We next performed whole-genome sequencing on four strains exhibiting strong protective effects (GX1-10, GX3-10, GX3-23, and GX2-22). Genomic analysis revealed that GX1-10, GX3-10, and GX3-23 belong to the ST617, ST34, and ST10 lineages of E. coli, respectively, while GX2-22 was classified as an ST290 K. pneumoniae (Fig. 5c). The carbapenemase-encoding genes were not detected in these strains, all of which were susceptible to carbapenems (Table S3). We then analyzed the carbohydrate-active enzymes (CAZymes) present in the genomes of these strains. CAZymes, responsible for the breakdown of carbohydrates, are classified into several classes based on their catalytic functions, including glycoside hydrolases (GHs), carbohydrate esterases (CEs), polysaccharide lyases (PLs), glycosyltransferases (GTs), and auxiliary activities (AA) [34]. Genomic analysis revealed that the six protective strains contained between 57—62 CAZyme-encoding genes, with 39 of these genes being common across all strains (Fig. 5d and Fig S5a). In comparison to VH1-2-Lux, 1917, GX1-10, GX3-10, GX3-23, and GX2-22 exhibited 15, 18, 15, 15, and 8 unique CAZyme-encoding genes, respectively (Fig. 5e and Fig.S5b). To pinpoint the most valuable CAZyme-encoding genes, a comparative genomic analysis was conducted. Among the 166 identified CAZyme-encoding genes, 12 are associated with carbohydrate binding, six with the hydrolysis of carbohydrate esters, 44 with the hydrolysis and/or rearrangement of glycosidic bonds, 25 with glycosidic bond formation, 75 with auxiliary activities, and four with non-hydrolytic cleavage of glycosidic bonds. Significantly, genes encoding glycoside hydrolase 73 (GH73) and glycosyl transferase 26 (GT26) were found in all protective strains but were absent in VH1-2-Lux (Fig S5c).

GX2-22 exhibited superior growth compared to VH1-2-Lux when monosaccharides (glucose, arabinose), disaccharides (sucrose, D-trehalose), oligosaccharides (chitooligosaccharides [COS], mannooligosaccharides [MOS], raffinose), and sugar alcohols (D-arabitol) served as the exclusive carbon source (Fig S6). When VH1-2-Lux was co-cultured with either 1917 or GX2-22 in media containing a single carbon source, its CFUs were consistently reduced compared to monoculture conditions, indicating that both GX2-22 and 1917 retain decolonization activity against VH1-2-Lux across a range of carbon sources. Notably, co-cultivation with GX2-22 led to a reduction in VH1-2-Lux CFUs by about 2 to 9 orders of magnitude (Fig. 5f). Following validation, quantitative analysis demonstrated distinct Sucrose utilization patterns among cultivation systems. After 24-h incubation, VH1-2-Lux monocultures retained 1.93 ± 0.06 mg mL⁻¹ residual Sucrose, which was 298% higher than that observed in GX2-22 monocultures (0.48 ± 0.07 mg mL⁻¹, p < 0.0001). The co-culture system exhibited a 45.6% reduction in residual sucrose (1.05 ± 0.11 mg mL⁻¹, p < 0.01 vs. VH1-2-Lux, Fig. 5g). Kinetic modeling revealed that GX2-22 significantly enhances sucrose utilization efficiency compared to VH1-2-Lux monocultures.

We then performed metabolic phenotyping of strains using Biolog Phenotype MicroArray™ System to assess carbohydrate utilization efficiency. Quantification of carbon source metabolism revealed that VH1-2-Lux exhibited significantly enhanced metabolic activity towards Glycyl-L-proline, methyl pyruvate, and γ-aminobutyric acid compared to GX2-22 and 1917 (Table S4). Competitive co-culture assays were established between GX22-2 and VH1-2-Lux under 5 mg/mL Sucrose-only or Sucrose plus 5 mg/mL methyl pyruvate. Consistent with carbon competition, VH1-2-Lux load decreased 832.4-fold in sucrose-only medium versus monoculture controls (p < 0.01). Remarkably, supplementation with methyl pyruvate significantly enhanced VH1-2-Lux biomass production by 5.12-fold (p < 0.001, Fig. 5h). These findings collectively support that GX2-22 mediates colonization resistance through preferential carbon source competition, while niche-specific nutritional supplementation can partially counteract such ecological pressure.

Discussion

Colonization of the intestinal tract by CR-HvKP predisposes individuals to infections that are challenging to treat clinically [3, 35]. The gut microbiota, which harbors trillions of bacteria, plays a crucial role in reducing the risk of such infections by preventing CR-HvKP colonization [28, 36]. Although CRKP/CR-HvKP strains are increasingly endemic in some regions and classified as critical organisms [37], they are relatively less prevalent in pigs. To initially characterize the microbiota-determined differences in colonization resistance among pigs, we developed the Micolo colonization evaluation model, which enabled the identification of key species that conferred effective resistance to pathogens. By labeling the CR-HvKP strain VH1-2 with IS-luxCDABE, we facilitated real-time observation of bacterial physiology, distribution, migration, and infection within living organisms [38]. This transposon-based tagging method facilitates transfer and integration between Gram-negative bacterial species and is therefore also applicable for labeling and tracking other pathogenic Enterobacteriaceae strains [39, 40].

Similar to humans, the age of pigs shapes their gut microbiota [41]. Triggered by the observation of a high relative abundance of Enterobacteriaceae in a particularly protected sample, specifically in the PS population, we decided to investigate the effect of commensal Enterobacteriaceae on CR-HvKP colonization. Through screening a collection of pig gut symbionts, we discovered that live Enterobacteriaceae bacteria enriched in the intestines of PS provided effective resistance to CR-HvKP. Our findings support the view that commensal Enterobacteriaceae protect against pathogen colonization, not only for Salmonella Enteritidis but also for CR-HvKP [38]. To our knowledge, this is the first study demonstrating that commensal Enterobacteriaceae from PS can resist CR-HvKP colonization. Notably, not all Enterobacteriaceae isolates from PS exhibited strong protective effects, suggesting strain-specific differences within this bacterial family. Strikingly, K. pneumoniae GX2-22, isolated from the PS sample, strongly reduced CR-HvKP colonization in vitro. The animal experiments further demonstrated that removing Enterobacteriaceae via streptomycin treatment promotes CR-HvKP colonization. Conversely, supplementation with K. pneumoniae GX2-22 or E. coli Nissle 1917 significantly reduced CR-HvKP VH1-2-Lux colonization in the mouse model, indicating that Enterobacteriaceae species are essential for exhibiting protective effects. However, it is worth noting that the colonization level of VH1-2-Lux was low at day post-infection 1, and any loss of VH1-2-Lux during infection may be ascribed to a bottleneck effect [42]. In murine models, GX2-22 intervention significantly reduced the bacterial load of VH1-2-Lux while concurrently enriching genera including Ligilactobacillus, Parabacteroides, Lachnoclostridium, and Romboutsia compared to controls. Notably, recent studies by Yin et al. demonstrated that organic acids derived from Ligilactobacillus metabolites exhibit antagonistic activity against K. pneumoniae [43]. Collectively, our findings demonstrate that GX2-22 may enhance the relative abundance of Ligilactobacillus and related genera, thereby synergistically contributing to the suppression of VH1-2-Lux colonization.

Compared to the carbon-limited Mipro medium, K. pneumoniae GX2-22 exhibited superior colonization resistance in carbon-rich LB broth [32]. This phenotype suggests that its competitive dominance may result from increased metabolic versatility in nutrient utilization. Notably, GX2-22 exhibited carbohydrates (such as sucrose, COS, raffinose, or PNPG) competition-mediated suppression of VH1-2-Lux growth. Supplementation with methyl pyruvate partially rescued VH1-2-Lux biomass production. Collectively, these findings indicate that K. pneumoniae GX2-22 plays a role in colonization resistance by competing with VH1-2-Lux for essential carbon sources. Although supplementation with methyl pyruvate did not restore the bacterial load of VH1-2-Lux to the level observed in monoculture, this may be attributed to GX2-22 competing not only for sucrose but also for the metabolic products derived from methyl pyruvate that are utilized by VH1-2-Lux.

Comparative genomic analysis has identified multiple genes in protective strains linked to GH73 and GT26. Notably, GH73, a member of the third glycoside hydrolase family, exhibits lysozyme activity, which compromises cellular membrane integrity [44, 45]. Lysozyme is increasingly regarded as a potential alternative to antibiotics and is currently garnering significant attention within the animal husbandry industry [44]. Therefore, it is plausible that GH73 in the protective strains contributes to the elimination of CRKP and CR-HvKP. Furthermore, GHs are a class of enzymes primarily involved in the catalysis of carbohydrate metabolism [46]. These enzymes enhance the ability of organisms to digest dietary fibers, thereby promoting bacterial proliferation [47, 48]. Previous studies have demonstrated that GHs can act on mucin glycan core structures, thereby facilitating the prolonged colonization of B. bifidum in the intestinal environment [49, 50]. Consequently, GH73 may enhance the ability to utilize fermentation substrates. GTs facilitate the transfer of sugar moieties between molecules, leading to the formation of glycosidic bonds, which promote bacterial growth [51]. However, attempts to knock out GH73 and GT26 genes in the protective strains were unsuccessful, likely due to their essential roles in cell division [52, 53], which prevented further function validation.

K. pneumoniae GX2-22 is classified as sequence type 290 (ST290) and lacks carbapenem resistance genes. Nonetheless, it possesses the potential to acquire carbapenem resistance plasmids from other strains via horizontal gene transfer, thereby potentially evolving into CRKP or even CR-HvKP [54]. While the observation that K. pneumoniae GX2-22 can reduce colonization by multiple CR-HvKP strains is promising, it is premature to consider its application as a probiotic. This caution is due to the inherent risk of the strain acquiring resistance genes. The CRISPR/Cas system presents a promising approach by conferring immunity against exogenous nucleic acids originating from plasmids [40, 55]. Consequently, a strategically designed probiotic intervention utilizing K. pneumoniae GX2-22 equipped with the CRISPR/Cas system could be developed to prevent or treat colonization by CR-HvKP in the host [48, 56]. Additionally, the incorporation of genes conferring an advantage in carbon source utilization, such as those encoding GH73 or GT26, into probiotic strains and ensuring their robust expression could enhance their growth performance. This enhancement may enable probiotics to more effectively inhibit colonization by CR-HvKP in the gut. Furthermore, targeted dietary interventions utilizing carbohydrates such as sucrose, COS, raffinose, or PNPG may provide a therapeutic approach by fostering the proliferation of commensal Enterobacteriaceae or modulating gut homeostasis, thereby contributing to the reduction of CR-HvKP colonization.

This study, however, presents certain limitations that merit consideration. Firstly, while we have assessed the antagonistic effects of porcine-derived GX2-22 and carbon sources on CR-HvKP colonization in murine models using the concept of proof, the lack of validation in porcine models constitutes a significant limitation. Given the considerable interspecies differences in microbiota complexity, pharmacokinetic properties, and immune system characteristics between murine and porcine systems [57], direct validation in swine models is crucial to confirm the translational potential of these findings. Considering the porcine origin of GX2-22 and the physiological similarities between porcine and human gastrointestinal systems, future research should prioritize comprehensive evaluation in porcine models to establish species-specific efficacy and safety profiles.

Conclusion

This study demonstrated that the gut microbiota of PS exhibits potent antagonistic activity against CR-HvKP growth and colonization. We pinpointed enrichment of Enterobacteriaceae, particularly a K. pneumoniae ST290 strain, as the primary mediator of this effect. Crucially, this ST290 strain substantially suppressed CR-HvKP proliferation through competitive utilization of carbon sources, notably sucrose. Consequently, these findings highlight carbohydrate-targeted interventions as a viable therapeutic/prophylactic strategy to combat CR-HvKP colonization and improve treatment efficacy.

Methods

Bacterial strains and plasmids

The bacterial strains and plasmids used in this study are listed in Table S1. Among these, the CR-HvKP isolate VH1-2 (ST23) was sourced from vegetables, while the remaining isolates were obtained from patients with hospital-acquired infections. In addition to VH1-2, the isolates DH1 (ST11), DH36 (ST11), and DH28 (ST23) were categorized as CR-HvKP according to our previous study [8]. The other 3 ST11 isolates, namely DH6, DH19, and DH20, were identified as CRKP. All strains were cultured aerobically in Lysogeny Broth (LB) at 37 °C with shaking at 180 rpm. When necessary, the following antibiotics were added: chloramphenicol (CHL) at 25 mg/L, meropenem (MEM) at 1 mg/L or 4 mg/L, ampicillin (AMP) at 100 mg/L, fosfomycin (FOS) at 32 mg/L, and sodium tellurite (ST) at 25 mg/L. Diaminopimelic acid (DAP) was added at 50 mg/L for the growth of E. coli WM3064.

Genetic engineering of CR-HvKP strains

We constructed the suicide plasmid p26-RP4-Lux, which includes the luxCDABE cassette flanked by IS26, the RP4 oriT fragment from pCVD442 [58], and the R6K replication origin from pSV033 [59]. The tellurite resistance gene tpm, encoding thiopurine S-methyltransferase, was used as the selection marker for sodium tellurite resistance [60]. The detailed plasmid construction process is described in Appendix 1 of the supplemental material. To engineer CR-HvKP VH1-2, we co-cultured it with E. coli WM3064/p26-RP4-Lux in LB broth Supplemented with DAP at 37 °C for 4 h. The cultures (10 μL each) were then mixed and incubated on prewarmed LB agar at 37 °C for 12 h. The resulting bacterial Library was collected in 900 μL of PBS and plated on ST 25 mg/L LB to select for luxCDABE-labeled mutants. Mutants were verified using bioluminescent imaging (IVIS Lumina Imaging System with Living Image software, version 4.2), PCR, and Sanger sequencing.

Pig microbiome samples

Stool samples were collected from a pig farm in Guangxi province, China. Each sample (10—20 g) was placed in a sterile sampling bag and maintained under anaerobic conditions. Samples were collected in the morning and processed within 12 h. To prepare a 10% (w/v) fecal slurry, 0.25 g of each sample was resuspended in 2.5 mL of PBS (pH 7.4) containing 0.1% (w/v) L-cysteine. The slurry was homogenized by vortexing for 10 min, allowed to settle for 10 min, and then filtered through a double layer of sterile gauze to remove fecal debris. The resulting slurry was autoclaved at 100 °C for 20 min to sterilize it. In addition, a 2 g fecal sample from each subject was stored at − 80 °C for 16S rRNA amplicon sequencing of fecal bacteria.

Pig stool colonization assay

VH1-2-Lux cultures were washed twice with PBS and resuspended to approximately 10⁸ CFU/mL. 100 μL of the CR-HvKP VH1-2-Lux culture was added to 900 μL of filtered fecal slurry in 2 mL EP tubes. Each mixture (200 μL) was dispensed in triplicate into 96-well black plates and incubated anaerobically at 37 °C for 24 h. Colony counts were determined at 0, 24 h using LB plates containing MEM (4 mg/L) and FOS (32 mg/L). Luminescence values were also measured at 0 and 24 h.

16S rRNA gene amplification and sequencing

Microbial DNA was extracted from fecal samples using the QIAamp PowerFecal DNA Kit (QIAGEN) per the manufacturer’s instructions. DNA concentration and Purity were measured with a NanoDrop 2000 spectrophotometer (Thermo Scientific, USA). The V4 region of the bacterial 16S rRNA gene was amplified using primers 515 F (5’-GTGCCAGCMGCCGCGGTAA-3’) and 806R (5’-GGACTACHVGGGTWTCTAAT-3’) in a GeneAmp 9700 PCR system (ABI, USA) under the following conditions: 95 °C for 3 min; 27 cycles of 95 °C for 30 s, 55 °C for 30 s, 72 °C for 45 s; and a final extension at 72 °C for 10 min [61]. Each 20 μL PCR reaction contained 4 μL 5 × FastPfu Buffer, 2 μL dNTPs (2.5 mM), 0.8 μL of each primer (5 μM), 0.4 μL FastPfu Polymerase, and 10 ng template DNA. PCR products were Purified from a 2% agarose gel (AxyPrep Kit, Axygen, USA) and quantified (QuantiFluor™-ST, Promega, USA). Equimolar amplicons were pooled and sequenced (2 × 300 bp) on an Illumina MiSeq platform following standard protocols. Raw reads were processed using QIIME2 with custom scripts (https://docs.qiime2.org/2019.1/). The sequencing services were provided by Wekemo Tech Group Co., Ltd., located in Shenzhen, China.

Enterobacteriaceae strains competition assay

100 µL of fecal slurry from PS was spread onto eosin methylene blue agar (EMBA) plates using a sterile cotton swab. Thirty colonies with different morphologies were Sub-cultured to obtain Pure cultures. Each competition assay was set up by mixing 100 µL of the CR-HvKP culture (10⁷ CFU/mL) with 100 µL of the Enterobacteriaceae culture (10⁷ CFU/mL) in 2 mL EP tubes containing 800 µL of Mipro broth or LB broth. The mixtures (200 µL) were then dispensed in triplicate into 96-well black plates and incubated anaerobically at 37 °C for 24 h. Colony counts were determined at 0, 24 h using LB plates containing MEM (4 mg/L) and FOS (32 mg/L). OD₆₀₀ and luminescence values were monitored at 0 and 12 h.

Animals

Healthy six- to eight-week-old female C57BL/6 J mice were purchased from Guangdong Scarstar Biotechnology (Guangzhou, Guangdong, China) and housed under specific pathogen-free conditions. All animal experiments were conducted at the Animal Center of South China Agricultural University (Guangzhou, Guangdong, China) in accordance with the guidelines of the Animal Experimentation Ethics Committee at South China Agricultural University (Animal Ethics Number: 2024c005).

In vivo competition assay

Mice were administered streptomycin (5 g/L) in their drinking water for four consecutive days before being colonized with CR-HvKP strains. Fresh fecal samples were collected on day 4, and the weight was recorded. Fecal samples were diluted in 1 mL of PBS and homogenized by vortexing. Serial dilutions of the homogenized samples were plated on EMBA to assess the effectiveness of streptomycin in eliminating Enterobacteriaceae. The CR-HvKP VH1-2-Lux strain was cultured in LB broth and resuspended in 4 mL of PBS to approximately 10⁸ CFU/mL. Mice were orally inoculated with 200 μL of VH1-2-Lux. Four days post-infection, feces were collected and plated on EMBA containing 1 mg/L MEM to enumerate bacteria. For the in vivo competition assay, VH1-2-Lux, E. coli Nissle 1917, and GX2-22 were each resuspended in 4 mL of PBS to 10⁸ CFU/mL. VH1-2-Lux was mixed with an equal volume of PBS, E. coli Nissle 1917, or GX2-22, respectively, and 200 μL of each mixture was administered orally to the mice. Fresh fecal samples were collected at 2 days post-infection (dpi 2) and plated on EMBA containing 1 mg/L MEM to quantify VH1-2-Lux.

Assessment of in vitro antimicrobial activity of cell-free supernatants

The cell-free supernatants (CFSs) of GX2-22 and E. coli 1917 were prepared as previously described [62]. Briefly, bacterial cultures were centrifuged at 3,000 rpm for 10 min at 4 °C, and the Supernatants were filtered through a 0.2 μm membrane. The antimicrobial activity of the CFSs was assessed using the well diffusion method [63]. Twenty Brain Heart Infusion (BHI) agar plates were inoculated with VH1-2-Lux bacterial cultures, and wells (6 mm in diameter) were Punched into the agar. Each well was loaded with 20 μL of CFSs. Following a 24-h incubation at 37 °C, inhibition zones were measured. All experiments were performed in triplicate.

The antimicrobial activity of CFSs against VH1-2-Lux was evaluated using growth inhibition assay in sterile 96-well plates (Corning® Incorporated Life Sciences, NY, USA). Bacterial suspensions of VH1-2-Lux were standardized to 10⁸ CFU/mL in Mipro broth. Each well contained 160 μL of Mipro broth, 20 μL of CFS (with 20 μL Mipro broth substituted as the negative control), and 20 μL VH1-2-Lux (10⁶ CFU/ml). Following homogenization, plates were incubated at 37 °C for 24 h with continuous optical density (OD₆₀₀) monitoring. Growth inhibition was quantified by comparing OD₆₀₀ reduction in treatment groups versus controls. All experiments included three biological replicates with technical duplicates.

Whole genome sequencing and analysis

Genomic DNA was extracted from bacterial samples using the Bacterial Genomic Extraction Kit (Tiangen, Beijing) following the manufacturer’s instructions. Whole genome sequencing (WGS) was performed by Tianjin Biochip Corporation using the Illumina MiSeq system (Illumina). Paired-end Illumina reads were assembled using SPAdes v3.6.2 software. Carbohydrate-active enzyme (CAZyme)-encoding genes were annotated using dbCAN (v2) with default parameters²⁵. Antibiotic resistance genes (ARGs) were identified using ResFinder 3.1 (https://cge.cbs.dtu.dk/services/ResFinder/). Genome annotation was carried out through the Prokaryotic Genome Annotation Pipeline (PGAP) server.

In vitro assays

GX2-22, Nissle 1917, and CR-HvKP VH1-2-Lux strains were cultured and adjusted to approximately 10⁸ CFU/mL in PBS. Subsequently, 10 μL of GX2-22 (PBS or Nissle 1917) and VH1-2-Lux were transferred to 980 μL of M9 basal medium containing different carbon sources at a final concentration of 5 mg/mL. A total of 200 μL of the cell Suspension was loaded into 96-well plates and incubated at 37 °C for 24 h. After incubation, serial dilutions were prepared from each well, and the quantification of VH1-2-Lux was performed by plating samples on LB agar plates Supplemented with 1 mg/L MEM.

Sucrose quantification

A 10 mL M9 basal medium was prepared containing a homogeneous suspension of bacterial culture (VH1-2-Lux, GX2-22 or VH1-2-Lux + GX2-22) at a final concentration of 1 × 10⁶ CFU/mL and 5 mg/mL Sucrose. Cultures were gradient-incubated at 37 °C with shaking (180 rpm) for 0, 6, 12, or 24 h, respectively. Post-incubation, samples were centrifuged at 4 °C and 5000 rpm for 5 min, and supernatants were carefully harvested for downstream analysis. Sucrose levels were determined using the acid hydrolysis-Lane-Eynon method (GB 5009.8–2023) [64]. Sample hydrolysis involved acid-catalyzed sucrose-to-reducing sugar conversion, followed by quantification via alkaline tartrate copper reagent. All assays were conducted in three independent biological replicates to ensure reproducibility.

Biolog assays

The carbon sources assimilation ability was assayed using the Biolog Phenotype MicroArray system (Biolog, Hayward, CA, USA), according to the manufacturers’ instructions [65, 66]. Briefly, a pure culture of the strain (VH1-2-Lux, 1917 or GX2-22) was incubated at 37 °C, and then suspended in a special inoculating fluid at the predetermined cell density (90–98% transmittance). Then, 100 μL of the cell suspension was inoculated into each well of the GEN III MicroPlate™. The microplate was incubated at 37 °C for 24 h, during which time kinetic information was recorded and quantified using Biolog’s GEN III OmniLog II ComboPlus kinetic software (Biolog, United States) followed by data analysis.

Statistical analysis

Statistical analyses were conducted using GraphPad Prism 8 software. Results are presented as the mean ± standard deviation from at least three independent experiments. Significant differences between groups were determined using unpaired t-tests or one-way analysis of variance (ANOVA), with a p-value of less than 0.05 considered statistically significant.

Authors’ contributions

JS contributed to the conception and design drafting of the work. GL, LJ, JL, AG, LJX, SYZ, YHL, LW, YZH, RYS, XC and JHL contributed to the acquisition of data. GL, LJ, AG, DHZ and JTY contributed to the analysis of data. XPL and YHL supervised the project. GL wrote the manuscript. LJ, HR, XLL and LC revised manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by the Foundation for Innovative Research Groups of the National Natural Science Foundation of China (32121004), National Natural Science Foundation of China (32402943,32202859), Guangdong Provincial Natural Science Foundation (2025A1515012412), Local Innovative and Research Teams Project of Guangdong Pearl River Talents Program (2019BT02N054), Guangdong Major Project of Basic and Applied Basic Research (2020B0301030007), the 111 Center (D20008), Specific University Discipline Construction Project (2023B10564003), and the school level scientific research project of Jiangsu Agri-animal Husbandry Vocational College (NSF2025CB07).

Data availability

The WGS data has been deposited in GenBank under BioProject accession number PRJNA1135928. The raw sequence data of 16S rRNA gene sequencing generated and analyzed during the current study are available at NCBI under Bioproject PRJNA1135795 and PRJNA1275953.

Declarations

Ethics approval and consent to participate

Animal experimentation was approved by the by the ethics Committee of the laboratory animal center of South China Agricultural University (Guangzhou, China), approval number:2024c005.

Competing interests

The authors declare no competing interests.