An adapted method for Cas9-mediated editing reveals the species-specific role of β-glucoside utilization driving competition between Klebsiella species

ABSTRACT

Cas9-based gene editing tools have revolutionized genetics, enabling the fast and precise manipulation of diverse bacterial species. However, widely applicable genetic tools for non-model gut bacteria are unavailable. Here, we present a two-plasmid Cas9-based system designed for gene deletion and knock-in complementation in three members of the Klebsiella oxytoca species complex (KoSC), which we applied to study the genetic factors underlying the role of these bacteria in competition against Klebsiella pneumoniae. Firstly, the system allowed efficient and precise full-length gene deletion via enhanced lambda Red expression. Furthermore, we tested the efficiency of two independent, functionally validated complementation strategies. Ultimately, the insertion of universal “bookmark” targets during gene deletion subsequently allows the most optimal genetic complementation in K. oxytoca, Klebsiella michiganensis, and Klebsiella grimontii. This approach offers a significant advantage by enabling the use of a single high-efficiency “bookmark” for complementing other loci or strains, eliminating the need for site-specific design. We revealed that the carbohydrate permease CasA is critical in ex vivo assays for K. pneumoniae inhibition by K. oxytoca but is neither sufficient nor required for K. michiganensis and K. grimontii. Thus, the adaptation of state-of-the-art genetic tools to KoSC allows the identification of species-specific functions in microbial competition.

IMPORTANCE

Cas9-based gene editing tools have revolutionized bacterial genetics, yet, their application to non-model gut bacteria is frequently hampered by various limitations. We utilized a two-plasmid Cas9-based system designed for gene deletion in Klebsiella pneumoniae and demonstrate after optimization its utility for gene editing in three members of the Klebsiella oxytoca species complex (KoSC) namely K. oxytoca, Klebsiella michiganensis, and Klebsiella grimontii. We then adapted a recently developed protocol for functional complementation based on universal “bookmark” targets applicable to all tested species. In summary, species-specific adaptation of state-of-the-art genetic tools allows efficient gene deletion and complementation in type strains as well as natural isolates of KoSC members to study microbial interactions.

INTRODUCTION

An increasing number of multidrug-resistant (MDR) pathogens pose a major health threat worldwide. Klebsiella pneumoniae, a common MDR opportunistic pathogen from the family Enterobacteriaceae (1), is estimated to have a prevalence of 32.8% among nosocomial infection isolates worldwide (2). As the discovery of novel antimicrobials has proven to be both increasingly difficult and slow (3), many recent studies have explored alternatives to classic antibiotics, specifically, how the microbiome can be utilized to prevent infections (4–8). It is now widely recognized that the microbiome contributes to colonization resistance (CR) against disease-causing pathogens (9, 10) through various mechanisms. For instance, commensals can protect the host by utilizing growth-limiting carbon sources (5–7), siderophores (11), and oxygen consumption (8) or by producing bioactive molecules such as secondary bile acids (4), short-chain fatty acids (12), and growth-inhibitory toxins (13, 14). Diverse human-isolated commensal bacteria have also been described to enhance CR, either by preventing pathogen outgrowth or facilitating pathogen clearance. Such species include enterobacteria, such as Escherichia coli (7, 8) and members of the Klebsiella oxytoca species complex (KoSC) (5, 6). For instance, K. oxytoca was shown to enhance colonization resistance against K. pneumoniae in ampicillin-treated mice via competition for β-glucosides, which was dependent on the β-glucoside transporter, CasA (6). As in many cases, the mechanisms through which pathogen elimination occurs vary depending on the interacting microbes, and additional functional studies are necessary to gain insights into which properties are critical for optimal exploration as future therapeutics.

Genetic tools in bacteria have been decisive in the mechanistic dissection of bacterial physiology and microbial interactions. For gut commensals, such efforts have vastly focused on model organisms such as E. coli (15) or Bacteroides thetaiotaomicron (16); whereas, for many other commensal bacteria, these tools have been lacking. Notably, the speed and efficiency of genetic engineering have been greatly improved by the development of CRISPR technologies in the past decade (17, 18). For bacterial gene editing, CRISPR-Cas9 systems are frequently used together with phage-derived recombination systems (19), typically with lambda Red (20–22). Similar to traditional recombineering approaches, a DNA template is provided to the recipient cells to repair the chromosomal lesion via homologous recombination facilitated by a recombination system. However, while recombineering relies on the incorporation of an antibiotic resistance marker into the chromosome for selection of edited cells (23), in the case of CRISPR/Cas9-based tools, the double-stranded break (DSB) by Cas9 counter-selects against unedited cells, as DSB likely leads to cell death in bacteria (19). Despite many significant iterative technological advances for model organisms, it has been challenging to adopt them in less-studied species and non-laboratory strains of gut bacteria (24).

Here, we describe the adaptation and enhancement of a CRISPR-Cas9 tool coupled with lambda Red recombination in both human isolates and type strains of three species of the KoSC: K. oxytoca, Klebsiella michiganensis, and Klebsiella grimontii. Characterizing the role of KoSC in the gut has attracted growing interest, as it has been reported to enhance colonization resistance (5, 6) as well as cause nosocomial infections (25, 26), but so far, genetic tools have only been reported for gene knockout in K. oxytoca (6, 23, 25). The optimized protocol allows full-length gene deletion and knock-in complementation without the long-term maintenance of antibiotic-resistance genes, which enables plasmid-free ex vivo and in vivo studies with engineered strains. Notably, careful phenotypical and genotypic characterization of modified strains is necessary, as the method also introduces in some cases unwanted modifications in the targeted locus. Finally, as a proof of concept, we show how competition for β-glucosides as a driving force behind the decolonization of K. pneumoniae is species-specific to K. oxytoca, rather than being a general principle in the KoSC.

RESULTS

Strong induction of homologous recombination is required for high-efficiency gene deletion

Our adaptation and optimization strategy was built on a two-plasmid CRISPR-Cas9-based system originally developed in K. pneumoniae (21), which we recently utilized to target genes in the K. oxytoca strain MK01 (6). Briefly, one plasmid encodes the endonuclease Cas9, as well as the lambda Red genes, a phage-derived recombination system under the transcriptional control of the L-arabinose-inducible P_araB promoter (Fig. 1a). The second plasmid encodes the Cas9 handle of a single-guide RNA (sgRNA) and sacB (27), a negative selection that can be used for plasmid curing (Fig. 1a). We modified this plasmid to encode a GFP dropout sequence upstream of the Cas9 handle to allow rapid selection of correctly ligated plasmids containing the guide of interest. The third component is a linear double-stranded DNA repair template assembled from 500 bp homology arms carrying the desired genetic edit. Gene editing was achieved by the initial transformation of the Cas9-encoding plasmid followed by the L-arabinose induction of the transcription of the lambda Red genes. This step enables the recombination machinery to be present in the cell before Cas9 (pCas-apr) targets and cleaves the chromosome. Next, the sgRNA-encoding plasmid (pSG-spec) is co-transformed with the linear repair template. Successful integration of the repair template will lead to (i) gene deletion and (ii) integration of a small sequence (termed “bookmark”) (28) that can serve as the target for another pSG during subsequent genetic complementation.

Fig 1.

Two-plasmid CRISPR-Cas9 system allows highly efficient deletion of casA. (a) Plasmid maps of pCas-apr and pSG-GFP-spe. (b) Schematic and (c) sequences of spacers designed for casA deletion. (d) Number of colonies under different lambda Red induction conditions. Under uninduced conditions, no repair template was used to show killing efficiency. 0.2% and 2% L-arabinose-induced cells were provided with a repair template. Colony counts were compared to the transformation of pSG-spe expressing a non-targeting gRNA. All transformations were performed in biological duplicates using separately prepared batches of competent cells. Each dot represents CFUs from one biological replicate. (e) Deletion efficiencies of different sgRNAs quantified by colony PCR reactions. Individual dots represent the average of 10 randomly selected colonies screened from targeting samples and three randomly selected colonies screened from non-targeting samples.

To adapt and optimize the protocol, we initially compared five casA-targeting sgRNAs in their ability to facilitate successful full-length gene deletion in K. oxytoca MK01 (Fig. 1b and c). First, the targeting capacity of the sgRNAs was assessed by the transformation of plasmid pSG-spe into K. oxytoca MK01 in the absence of a repair template and without induction of the lambda Red machinery. Under these conditions, CRISPR-Cas9 is toxic to cells owing to its DBS activity. Compared to the non-targeting control, we observed a 10⁴- to 10⁵-fold reduction in CFUs in four out of five sgRNAs and no detectable colonies with sg189 (Fig. 1d), showing that the sgRNAs led to efficient killing in the absence of a repair template with a small number of surviving colonies (Fig. 1e), even though both plasmids were present in the cell at the same time.

Next, to promote homologous recombination-driven gene deletion of casA, we induced expression of the lambda Red system by adding 0.2% L-arabinose, as reported for K. pneumoniae (21). Surprisingly, we observed an increase in CFU compared to the uninduced conditions for only one of the five sgRNAs, sg185 (approximately 100-fold increase, from 59 to 5,250 CFUs) (Fig. 1d). The deletion efficiency of sg185 was increased from 0% to 100% as evidenced by colony PCR (10/10 tested colonies), but four out of five sgRNAs led to no successful gene deletion of casA (Fig. 1e). As the survival of cells in most cases depends on successful recombination, we hypothesized that stronger induction of lambda Red gene expression may lead to better editing with other sgRNAs. Indeed, increasing the L-arabinose concentration 10-fold increased the CFU count from 9 to >10⁴ for three of the four remaining sgRNAs (Fig. 1d). More importantly, a 65%–95% deletion efficiency of casA was observed for these sgRNAs (Fig. 1e). Collectively, these results show that although sgRNA performance can vary in terms of both surviving CFUs and editing efficiency, sufficient expression of the lambda Red system leads to robust editing with four out of five sgRNAs.

Knock-in complementation results in markerless genotype restoration

To adapt the editing protocol to gene complementation, the same general principles applied within the gene deletion pipeline were initially followed, with specific changes that preclude targeting of the repair template by Cas9. In brief, sgRNAs were designed to target either the upstream adjacent gene casR or the downstream adjacent gene casB, and a repair template containing the deleted gene flanked by homology arms was assembled (Fig. 2a). To avoid targeting and subsequent cleavage of the repair template by Cas9, the target and the protospacer-adjacent motif (PAM) (29) in the repair template were mutated during homology arm amplification using site-directed mutagenesis. After transformation of the repair template and pSG-spe with four different assembled sgRNAs targeting either casR or casB into 2% L-arabinose-induced ΔcasA cells, we observed a 10³- to 10⁵-fold CFU reduction compared to the non-targeting control with varying numbers of surviving colonies (n = 0–596 CFU/sgRNA), demonstrating once again the different targeting efficiencies of sgRNAs (Fig. 2c). In contrast to the gene deletion pipeline, markedly lower recombination efficiency was observed despite induction with 2% arabinose. Specifically, only one out of four sgRNAs showed successful integration of the repair template, with 10% complementation, in contrast to the 65%–95% deletion efficiency (Fig. 1e and 2d). This complementation approach, although at a lower frequency, allowed the generation of a markerless complementation strain, for example, K. oxytoca MK01 ∆casA::casA.

Fig 2.

Two knock-in complementation approaches lead to successful genetic complementation of casA. (a, e) Schematic and (b, f) sequences of spacers designed for casA complementation and bookmark (BM) complementation, respectively, in K. oxytoca MK01. (c, g) Number of colonies after transformation of K. oxytoca MK01 cells for complementation compared to a non-targeting control. All transformations were performed in biological duplicates using separately prepared batches of competent cells, which were all induced with 2% L-arabinose. Each dot represents CFUs from one biological replicate. (d, h) Complementation efficiencies of different sgRNAs. Individual dots represent the average of 10 randomly selected colonies screened from targeting samples and three randomly selected colonies screened from non-targeting samples.

To overcome the low and target-dependent complementation efficiency observed before, we adapted an alternative complementation method (28) that is based on the insertion of a 24 bp “bookmark” sequence as part of the repair template during gene deletion, which can then be utilized as a unique sgRNA target with known efficiency during complementation at any genomic locus (Fig. 2e). First, we designed four bookmark targets with sequences that were not present in K. oxytoca MK01 (Fig. 2f) and tested them in uninduced K. oxytoca wild-type (WT) cells. No CFU reduction was observed compared to the non-targeting control, suggesting that the bookmark sgRNAs did not cause DBS-triggered cell death in WT cells (Fig. 2g). Next, we transformed uninduced ΔcasA::BM cells with pSG-spe and bookmark sgRNAs to validate targeting after bookmark insertion. Compared to the non-targeting controls, each bookmark sgRNA showed a 10³- to 10⁴-fold reduction in CFUs, confirming successful bookmark insertion into the chromosome and targeting by Cas9 (Fig. 2g). Strikingly, following the co-transformation of bookmark-ligated pSG-spe plasmids with a repair template containing casA and homology arms, we achieved 60%–90% PCR-confirmed complementation efficiency with three out of four tested sgRNAs (Fig. 2h). Overall, iterative improvements to the gene editing strategy allowed the highly efficient PCR-confirmed knock-in of a 2.8 kb insert including casA in K. oxytoca MK01.

CRISPR-Cas9-mediated gene deletion and complementation are applicable across the KoSC

Next, we tested the efficacy of the deletion pipeline in additional strains belonging to the KoSC. We included the type strains for all three KoSC species (K. oxytoca DSM5175^T, K. michiganensis DSM25444^T, and K. grimontii DSM105630^T) and one commensal strain for each KoSC species (K. michiganensis LK158 and K. grimontii LK33) that was recently isolated from the stool of healthy human donors (Fig. 3a). For these five strains, sgRNAs sg185 and sg758 with the highest efficiency for casA deletion and subsequent complementation in K. oxytoca MK01, respectively, were evaluated. For sg185, alignment of the targets and flanking PAM sequences to genomic sequences revealed 100% sequence identity between the K. oxytoca and K. michiganensis strains but two mismatches between both K. grimontii strains and sg185 (Fig. 3f). Notably, Cas9 has been reported to retain functionality despite mismatches between the sgRNA guide and genomic target (30), and these mismatches were not anticipated to preclude successful targeting in K. grimontii. A non-targeting guide RNA encoding pSG-spe was included as a control for transformation efficiency. Notably, the transformation efficiency varied for the five strains between 385 and 5 × 10⁶ CFU/transformation (Fig. 3b). Notably, the transformation efficiency appeared to be strain-specific rather than species-specific (Fig. 3b). Colony PCR confirmed successful deletion of casA in four out of five tested species ranging from 10% to 80% in deletion efficiency, demonstrating the robust applicability of our gene deletion method in KoSC in both type strains and human isolates (Fig. 3c).

Fig 3.

Deletion and complementation protocol extends to other KoSC species. (a) Phylogenomic tree representing the evolutionary relations in the KoSC. Symbols indicate the source of the strains. The tree was constructed based on the core genes in the indicated seven genomes (242 genes). (b, d) Number of colonies after transformation of (b) pSG-sp185 for casA deletion and (d) pSG-sp758 for casA complementation compared to a non-targeting control. All transformations were performed in biological duplicates using separately prepared batches of competent cells induced with 2% L-arabinose. Each dot represents CFUs from one biological replicate. (c, e) Editing efficiencies of (c) sp185 and (e) for sp758. Individual dots represent the average of 10 randomly selected colonies screened from targeting samples. (f) Sequence alignment sp185 targeting casA for gene deletion in six KoSC strains. (g) Left: schematic of primers used in casA deletion repair template assembly (1–4) and PCR confirmation of mutant and complemented genotypes (5–6). Right: gel image of PCR amplicons of WT, ΔcasA, and ΔcasA::casA strains in three KoSC members; expected band size for ΔcasA amplicon is ~1,100 and ~3,000 bp for WT and ΔcasA::casA amplicons.

Next, we sought to complement the isolated strains of K. michiganensis and K. grimontii. After selecting sg758 for bookmark complementation because of its robust complementation efficiency in K. oxytoca MK01, we genetically deleted casA while inserting the sg758 target sequence into the chromosomes of K. michiganensis LK158 and K. grimontii LK33 to obtain ΔcasA::sg758 cells. Subsequent transformation of pSG-sg758 into ΔcasA::sg758 cells resulted in a 100-fold CFU reduction compared to the non-targeting control, indicating successful targeting (Fig. 3d). Furthermore, PCR analysis showed that 50% and 10% complementation efficiency were achieved in K. michiganensis LK158 and K. grimontii LK33, respectively (Fig. 3e), demonstrating that the “bookmark” complementation method is successfully transferrable to different strains. Of note, the editing efficiency of the same sgRNA is highly variable across different strains, with those yielding fewer transformants generally leading to poorer editing (Fig. 3c and e).

Genetic complementation frequently introduces mutations with functional consequences casA is part of the casRAB operon, which has been previously identified to be required for the utilization of β-glucosides, such as salicin, arbutin, and cellobiose, in K. oxytoca (31). In K. oxytoca MK01, we identified casA as essential for the utilization of salicin and arbutin, but not cellobiose, and this gene contributes to nutrient competition with multidrug-resistant K. pneumoniae strains ex vivo and in vivo (6). Since casA is present in the KoSC strains tested, we characterized how the WT, ΔcasA, and ΔcasA::casA strains of all three selected species utilize these carbohydrates. In the minimal medium, the utilization of salicin and arbutin was casA-dependent in all three species; whereas, cellobiose utilization was casA-independent (Fig. 4a). For all three species, three different PCR-confirmed ΔcasA::casA complementation clones were functionally tested for their ability to grow in salicin, arbutin, and cellobiose. Notably, most complemented ΔcasA::casA clones matched the growth profiles of their respective WT strains, indicating the full complementation of casA function in the cell. However, we also observed that one out of three clones in K. oxytoca, K. michiganensis, and K. grimontii with PCR-confirmed complementation displayed growth delays or showed no growth on salicin and arbutin (Fig. 4a).

Fig 4.

casA contributes to K. pneumoniae reduction in K. oxytoca but not in K. michiganensis or K. grimontii. (a) Heatmap displaying growth on minimal media supplemented with 5 g/L of arbutin or salicin. The color scale shows OD₆₀₀ values relative to the WT growth in each condition after 24 h of incubation. Displayed data represent the means of three technical replicates. (b) Schematic of ex vivo assay using cecum content of germ-free mice in 1:1 dilution with PBS. For longitudinal bioluminescence measurements, KoSC members were co-inoculated with K. pneumoniae lux carrying the luxCDABE operon in the chromosome. Bioluminescence was recorded every hour for 24 h. For CFU quantifications, KoSC members were co-inoculated with MDR human isolate K. pneumoniae MR01. After 24 h of incubation, samples were serially diluted and plated, and CFUs were quantified after overnight incubation. (c) K. pneumoniae MR01 CFUs quantified on LB agar plates selected for K. pneumoniae growth. Bars represent the means of three biological replicates. Each dot represents the means of three technical replicates. The P value indicated represents the Kruskal-Wallis test on the effect of KoSC presence on K. pneumoniae CFUs. The P values for each group represent Dunn’s multiple comparison tests compared to K. pneumoniae control group with **P < 0.01 and *P < 0.05. (d–f) Strength of the bioluminescent signal of K. pneumoniae lux longitudinally in the presence of the three KoSC members or alone as a control. Each dot represents the means of three independent experiments in technical triplicates using different batches of cecum content obtained from different germ-free (GF) animals for each independent experiment. RLU: relative light units.

Next, we performed whole-genome sequencing from the WT, ΔcasA, and ΔcasA::casA strains (n = 15) to identify potential mutations explaining the phenotype. Indeed, strains with growth defects (K. oxytoca ΔcasA::casA clone2, K. michiganensis ΔcasA::casA clone3, K. grimontii ΔcasA::casA clone2) encoded either missense, nonsense, or frameshift mutations in casA or in its proximity (Fig. S1 to S3). Thus, careful sequence analysis of the edited locus, including the upstream and downstream PCR-amplified homology regions, is required to identify strains with correct genotype.

Overall, our method allows both genetic and phenotypic complementation but underscores that sequencing and ideally also functional validation of complemented strains is critical for these types of genetic studies. Furthermore, in vitro growth assays of deletion and complementation strains confirmed the functional homology of casA across KoSCs in β-glucoside utilization.

CasA contributes to competition against K. pneumoniae only in K. oxytoca, not across the KoSC

Finally, to determine whether the presence of casA is sufficient for interspecies competition against K. pneumoniae across the KoSC, as has been reported for K. oxytoca (6), we utilized an ex vivo competition assay using the cecal content of germ-free mice as a media base, as we previously described to reassemble the environment of antibiotic-disturbed gut (6) (Fig. 4b). After 24 h of incubation of KoSC strains with the MDR human isolate K. pneumoniae MD01 (6), we quantified K. pneumoniae CFUs on selective agar plates. As previously reported, we observed a 100-fold, casA-dependent CFU reduction of K. pneumoniae in co-culture with K. oxytoca MK01; whereas, we recorded a moderate 10-fold reduction when co-cultured with K. michiganensis and K. grimontii (Fig. 4c). For time-resolved competition, KoSC strains were co-cultured with a K. pneumoniae strain carrying the luxCDABE operon (32), allowing longitudinal recording of the bioluminescent signal, which was used as a proxy for growth. After 12 h, we recorded a 1,000-fold reduction in K. pneumoniae bioluminescence in co-culture with K. oxytoca WT and ΔcasA::casA strain, which was casA-dependent, as a lower 50-fold reduction was observed when co-cultured with K. oxytoca ΔcasA (Fig. 4d). In contrast, K. michiganensis (Fig. 4e) and K. grimontii (Fig. 4f) showed a less pronounced reduction effect of 10- to 100-fold that was casA-independent (Fig. 4e and f). These results support previous reports on the role of casA in the growth competition of K. pneumoniae with K. oxytoca and suggest that this effect is not a general KoSC feature but rather limited to K. oxytoca, further highlighting the relevance of K. oxytoca in mediating colonization resistance to MDR K. pneumoniae species.

DISCUSSION

Although massive improvements have been made in the development of CRISPR-based and other gene editing technologies, the implementation of such tools is often limited to laboratory strains or model organisms, which can vastly differ from less characterized species or even wild isolates of model organisms in terms of transformability and cultivability (33). In this study, we present an optimized Cas9-based pipeline for full-length markerless gene deletion combined with a gene knock-in protocol for complementation, enabling downstream functional analyses that do not require the presence of selection markers for plasmid maintenance, which can limit experimental design.

We optimized our pipeline for K. oxytoca MK01, which was previously used to delete casA (6). We observed high transformation efficiencies in K. oxytoca MK01, making it an ideal candidate for benchmarking our Cas9-based genetic toolbox. Transformation of K. oxytoca with sgRNA-targeted Cas9 without induction of the lambda Red system revealed that a small fraction of cells were able to survive or escape the lethal effects of the Cas9-generated double-stranded DNA break, as previously observed in E. coli (34, 35) and P. putida (19). These escapers may occur as a result of several possible genetic events interfering with effective gene targeting, such as deletion of sgRNA (19), mutations in Cas9 (34), mutation of the target site, or homologous recombination-mediated repair in case of weakly targeting sgRNAs (35). In our study, the number of escapers varied depending on the target sgRNA sequence, although we generally observed strong CFU reductions with all tested sgRNAs.

Comparison of several sgRNAs targeting casA (or adjacent genes for casA complementation) revealed vast differences in editing efficiencies ranging from 0% to 100%. Although casA mutants were obtained for 4/5 sgRNAs, the differences in editing efficiencies highlight the need for testing different sgRNAs in the gene of interest to successfully obtain deletion mutants as well as complementation strains. Notably, in most publications reporting similar CRISPR-based tools, differences between sgRNA performances have rarely been reported (19, 21, 36). Furthermore, we observed that sgRNA, which gave rise to more escapers, also allowed better editing when the lambda Red genes were induced, resulting in higher editing efficiency. This was recently demonstrated in a study that obtained an increased editing efficiency by attenuating sgRNA (30). Although designing multiple sgRNAs can circumvent this heterogeneity for gene deletion, in the case of other edits aimed at a specific genomic location, attenuation could potentially be useful for sgRNAs yielding no colonies, such as sg189 (Fig. 1d). Another key factor contributing to successful gene editing was the 10-fold increase in L-arabinose concentration for the induction of lambda Red gene expression compared to recently published CRISPR/Cas9 tools in K. pneumoniae (21, 37), resulting in an increase in editing efficiencies from 0% to 75% in three previously unsuccessful sgRNAs. This approach is particularly beneficial for hypermucoid strains, which have low transformation efficiency and are common in the KoSC and Klebsiella genera in general (38, 39). The increase in the expression of the recombination machinery would circumvent the low transformability and would allow the isolation of the desired genetic edit, requiring fewer transformants.

Developing a protocol for genetic complementation is essential to definitively verify if the phenotype observed following gene deletion is associated with the gene-of-interest or is a by-product of the gene editing process itself. Since in the current study we specifically focused on non-model gut bacteria, re-integration of the deleted segment is easiest to its native locus. Although other gene editing tools might use a distant non-coding segment of the genome for complementation, this requires a detailed genomic analysis that is often unavailable in recently isolated bacterial strains. The establishment of robust and high-efficiency editing combined with the implementation of the recently reported “bookmark” complementation method (28) enabled the development of successful in situ complementation, which to our knowledge is reported for the first time for KoSC. Although complementation follows the same principle as deletion, it also poses the challenge of exclusively targeting the chromosome, but not the repair template by Cas9. To circumvent this issue, we first mutated PAM in the repair template, which can be both laborious and inefficient. Adapting a recently reported protocol developed in Clostridium autoethanogenum (28) allowed us to simplify the design process and increase complementation efficiency with unique sgRNAs inserted into the chromosome during gene deletion. This led to a similar editing efficiency for complementation as for deletion. Of note, we found that some complementation clones did not restore β-glucoside utilization despite the integration of the repair template. Sequencing revealed that these instances were caused by either nonsense or frameshift mutations in casA or its proximity (see Fig. S1 to S3). Therefore, we advise careful sequence confirmation of the edited genotype, especially in the case of genes that are difficult to phenotypically characterize or if prior phenotypic information is not available. This may be conducted via whole-genome sequencing or targeted resequencing of the gene locus and its adjacent regions, depending on the size of the gene and the availability of resources.

One important advantage of the bookmark-based complementation approach is that it can be readily used in other bacterial strains or genomic loci without prior target-gene-specific design. After selecting the optimal deletion and complementation conditions for K. oxytoca MK01, we successfully applied our approach to four out of five other types and human isolate strains belonging to two additional species in the KoSC, namely K. michiganensis and K. grimontii.

Finally, the generation of ΔcasA and ΔcasA::casA strains in a human-derived strain of each KoSC representative species revealed that casA-dependent reduction of K. pneumoniae can only be observed in the case of K. oxytoca. This demonstrates that the presence of CasA is not sufficient but rather contributes in a K. oxytoca-specific manner against colonization by K. pneumoniae. Although we observed a moderate, 10- to 100-fold reduction of K. pneumoniae in the presence of K. michiganensis and K. grimontii, these interactions can most probably be attributed to genes other than casA. Hence, more extensive characterization of nutrient competition between members of the KoSC and other Enterobacteriaceae, including K. pneumoniae, is required to identify species- versus strain-specific effects.

In summary, we presented a robust and easily applicable two-plasmid CRISPR-Cas9 system for KoSC, enabling full-length gene deletion and subsequent in situ complementation. Following gene editing, both plasmids could be easily cured, creating markerless strains for downstream analyses. Given the increasing interest in characterizing competitive interactions between gut bacteria and Klebsiella species in particular (5, 6, 13, 40, 41), we anticipate that our tool will contribute to mechanistic studies of microbiome communities.

MATERIALS AND METHODS

Plasmid construction

Plasmids pCas and pSG were obtained from the Zhang lab. The sequences were previously published (21), and the plasmids are available in Addgene’s depositor collection (Accession numbers 117231 and 117234, respectively). The single-guide RNA-expressing plasmids were assembled using standard cloning procedures. Briefly, spacers were ordered as oligos with 5´ overhangs corresponding to the BbsI digestion site in the plasmid pSG-GFP-spe. The forward and reverse spacer oligos (10 µM) were heated to 95°C in a thermocycler, then cooled down gradually at −0.1 °C/s, and finally diluted 1:100. One microliter of annealed oligos was ligated into BbsI-digested pSG-spe following the manufacturer’s instructions, transformed into competent E. coli DH5α cells, and plated on lysogeny broth (LB) agar plates supplemented with 75 µg/mL spectinomycin. Following overnight incubation at 37°C, successfully annealed plasmid-bearing colonies appeared white, in contrast to GFP-expressing green colonies.

Gene targeting in KoSC members

KoSC members were cultured aerobically overnight in LB medium, then back-diluted 50-fold and subsequently grown until cultures reached an OD₆₀₀ of 0.6–0.8. Cells were then made electrocompetent by three rounds of washing with 10% ice-cold glycerol solution and centrifuging at 7,200 × g. Competent cells (50 µL) were transformed with 50 ng pCas-apr, and after 60 min of incubation, the cells were recovered in 950 µL liquid LB. The cultures were then centrifuged for 3 min at 8,000 × g, and the bacterial pellets were resuspended in 50 µL liquid LB and plated on LB agar plates supplemented with 60 µg/mL apramycin.

In the second round of transformation, transformants were cultured in liquid LB supplemented with 60 µg/mL apramycin overnight, then back-diluted 50-fold and grown at 30°C until an OD₆₀₀ of 0.15–0.2, then induced with 10% (vol/vol) 20 g/L L-arabinose in liquid LB supplemented with 60 µg/mL apramycin. Induction continued for 2 h either at room temperature or at 30°C, depending on the strain’s growth rate—until cultures reached an OD₆₀₀ of ca. 0.8. The cells were made electrocompetent as described above.

For genes targeting 50 µL of L-arabinose, competent cells were co-transformed with 200 ng of sgRNA plasmids and ~400 ng of linear repair template assembled with 500 bp upstream and downstream homology arms via overlap extension PCR. Notably, for bookmark spacer insertion, spacer sequences were included at the 3´ end of primers used for homology arm amplification during oligo synthesis. The cells were recovered after 60 min and plated in serial dilutions on LB agar plates supplemented with 60 µg/mL apramycin and 300 µg/mL spectinomycin. The limit of detection was determined based on the plated volume. Following overnight incubation, colony PCRs were performed on 10 and 3 randomly selected colonies with targeting and non-targeting sgRNA expressing pSGs, respectively. Successful repair template integration was verified on a 1% Tris-acetate-EDTA (TAE) agarose gel run at 130 V for 15 min, and then editing efficiencies were quantified.

For complementation and downstream functional analyses, the edited strains were cured with pSG-spe and both plasmids. Briefly, colonies were inoculated under non-selective conditions and then plated on LB agar plates supplemented with 5% sucrose, and were incubated at 30°C or 37°C depending on pCas-apr curing. Following overnight incubation, the colonies were streaked on selective and non-selective agar plates to confirm successful plasmid curing.

Bookmark complementation spacer selection

sgRNAs for bookmark complementation must fulfill two important criteria: (i) they must not show targeting activity in the WT genome, which would lead to off-target effects during complementation, and (ii) they are suitable for successful gene editing. To eliminate the possibility of off-target activity, we utilized spacer sequences that showed sequence homology to casA but contained two to seven mismatches in all tested KoSC strains along the 20 nt spacer sequence. Furthermore, we compared the targeting activity of all bookmark spacers in uninduced WT KoSC cells to that of a non-targeting spacer.

Phylogenetic tree construction

The phylogenomic tree in Fig. 3a representing the evolutionary relationships in the Klebsiella oxytoca species complex was constructed

In vitro growth assay

To assess the sugar utilization of WT, ΔcasA, and ΔcasA::casA strains of KoSC members, strains were cultured overnight in liquid LB, and then 50 µL of the culture was streaked on R2A agar plates. After 16 h, the cells were adjusted to an OD₆₀₀ of 0.2 in phosphate-buffered saline (PBS). Cultures were inoculated at 5% vol/vol in MM9 medium with a single carbon source of arbutin, salicin, or cellobiose at a concentration of 5 g/L. OD₆₀₀ was recorded every hour. Growth assay data were visualized using GraphPad Prism software (version 8.2.1).

Whole-genome sequencing

Genome libraries were prepared using the Illumina DNA PCR-Free Library Kit and the IDT for Illumina DNA/RNA UD Indexes for previously isolated DNA. The libraries were prepared following the manufacturer’s instructions. Quantification of library concentrations was performed using Qubit ssDNA Assay Kit followed by further quantification with KAPA Library Quantification Kit for Illumina. Sequencing was performed on the NovaSeq S4 PE150 platform targeting a depth of 1 million reads per sample. Sequence analysis, nucleotide alignment, and translation of the casRAB operon were performed with Geneious Prime (2019.2.1), and amino acid alignments were illustrated using Jalview 2.11.3.2.

Mice

Germ-free animals were used for intestinal content harvesting in ex vivo competition assays. WT C57BL/6NTac mice were raised in a germ-free breeding facility at the Helmholtz Centre for Infection Research and were maintained for breeding purposes only. In parallel to the renewal of breeding animals, old breeding pairs were sacrificed, and cecum content was isolated and then stored at −20°C until further use.

Ex vivo competition assay

The isolated cecum contents were thawed and diluted 1:1 with PBS. KoSC and K. pneumoniae lux strains were cultured in liquid LB overnight, and then the OD was adjusted to OD₆₀₀ of 1 and OD₆₀₀ of 0.2, respectively. Twenty microliters of KoSC members and 10 µL of K. pneumoniae lux was co-inoculated into 250 µL of PBS-diluted germ-free cecum contents in a 96-well format. The bioluminescence was recorded every hour using a microplate spectrophotometer. After 24 h, the assays were plated in serial dilutions on LB agar plates supplemented with 50 µg/mL kanamycin or 50 µg/mL chloramphenicol for K. pneumoniae quantification against K. oxytoca/K. grimontii and K. michiganensis, respectively, and on MacConkey agar plates for KoSC quantification of K. oxytoca and K. grimontii, which are visually distinguishable from K. pneumoniae unlike K. michiganensis.

ETHICS APPROVAL

All animal experiments were performed in accordance with the guidelines of the Helmholtz Center for Infection Research (Braunschweig, Germany), the National Animal Protection Law (Tierschutzgesetz, TierSchG), and Animal Experiment Regulations (Tierschutz Versuchstierverordnung, TierSchVersV), and with the recommendations of the Federation of European Laboratory Animal Science Association (FELASA).

DATA AVAILABILITY

The authors confirm that data supporting the findings of this study are available within the article.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/jb.00317-23.

Supplemental figures. jb.00317-23-s0001.pdf.

Figures S1 to S3.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.