Arm race among closely-related carbapenem-resistant Klebsiella pneumoniae clones | Panda Anku

Few patients carried CRKP on ICU admission, but many acquired CRKP in the ICU

On March 27, 2017, there were 42 patients in the ICU who were screened via rectal swab, of which 8 (19.0%, 8/42) were found to harbour CRKP (Fig. 1). Of the 34 patients with no CRKP on admission, 20 stayed in the ICU for ≥7 days. Follow-up screening swabs were collected from all 20 patients, of which 11 had CRKP recovered from at least one rectal swab (Fig. 1).

Fig. 1: Schematic of patients swabbed in the study and those leading to isolation of CRKP.
figure 1

Patients staying within the ICU for more than 72 h were screened for CRKP using rectal swabs; where possible, swabs were collected within 72 h of their admission. For those patients who were CRKP negative on screening, and continued to stay for at least 7 days, we collected additional swabs to test for nosocomial acquisition. Patients sampled within 72 h of their admission showed a low colonization rate of just 4.5%. However, samples collected after 72 h, or indeed after patients had been within the ICU for some time, showed a much higher prevalence of CRKP (32.7–36.7%), indicating that CRKP is largely acquired within the ICU environment. Blue = patient cohorts, orange = sample collection, purple = CRKP negative, green = CRKP positive.

During the study period, a total of 627 patients were admitted to the ICU and stayed within its care for 3 or more days. Rectal swabs were collected from 596 patients, with 74.8% (n = 446) of these swabs collected within 3 days of their admission, and 25.2% (n = 150) swabbed at some point afterwards. Amongst the 446 patients who were screened on admission, just 4.5% (n = 20) were found to harbour CRKP. For the 426 CRKP negative patients, 180 stayed within the ICU for less than 7 days and follow-up swab were not submitted and 246 stayed for at least 7 days, among which 226 actually submitted a follow-up swab, with CRKP subsequently recovered from 36.7% (n = 83) of samples (Fig. 1).

From the 150 primary samples that were collected after three days of ICU admission, CRKP was recovered from 32.7% (n = 49). The low CRKP carriage rate (4.5%) seen in those who were screened within 3 days of admission suggests that the vast majority (if not all) of the CRKP isolates in these 49 patients were acquired in the ICU. Therefore, a total of 638 patients were swabbed at least once, among which 171 had a CRKP isolate from rectal swabs during the study period (Fig. 1).

Clinical impacts from CRKP are limited, but reflect CRKP carriage populations

Despite the relatively high carriage rates, very few patients developed clinical infections due to CRKP. Non-duplicate CRKP isolates were recovered from clinical samples of 21 patients in the ICU during the study period (Dataset S1 in the Supplementary file). From the 21 patients yielding clinical CRKP samples, 20 also had CRKP-positive rectal swabs; no rectal swabs were collected from the remaining patient. Among the 171 patients with CRKP colonization, 11.7% (20/171) developed infection due to CRKP.

Most CRKP isolates belonged to ST11 and KL64 and carried bla

Among the 171 CRKP isolated, 18 were non-culturable after storage. The remaining 153 CRKP isolates from rectal swabs and 21 from clinical samples (Dataset S1) were subjected to short-read whole genome sequencing. Therefore, a total of 174 CRKP isolates were sequenced.

The 174 CRKP isolates belonged to six STs, ST11, ST37, ST45, ST292, ST661, and ST1640 which is a single allele (rpoB) variant of ST11 (Supplementary Table S1 for the allele profile of the 6 STs). The vast majority (n = 154, 88.5%) of the isolates belong to ST11, with 16 isolates (9.2%) belonging to ST45, and a single isolate for the remaining four STs. In the ST11 isolates, 94.8% (n = 146) belonged to the KL64 capsular type, while the remaining 8 belonged to KL47 (n = 7) or KL39 (n = 1). All of the ST45 isolates belonged to KL62.

All CRKP isolates were resistant to carbapenems including meropenem (MIC, 4 to 512 mg/L) and imipenem (MIC, 4 to 512 mg/L) and were also resistant to aztreonam, ceftazidime, and piperacillin-tazobactam. Most of the isolates were also resistant to ciprofloxacin (resistance rate, 99.4%; 173/174), tigecycline (83.9% 146/174; MIC, 1 to 16 mg/L), amikacin (79.9%, 139/174), and trimethoprim-sulfamethoxazole (77.0%, 134/174). Only one isolate (carrying blaNDM-1, see below) was resistant to ceftazidime-avibactam (MIC, 512/4 mg/L). Similarly, only one isolate was resistant to colistin (MIC, 4 mg/L), and was found to carry a plasmid-borne colistin resistance gene, mcr-1.1. None of the isolates were resistant to aztreonam-avibactam.

Carbapenemase-encoding genes were found in all but two of the 174 CRKP isolates, with blaKPC-2 being detected in 169 isolates (97.1%), blaKPC-12 in two and blaNDM-1 in one. The two isolates with no known carbapenemase genes exhibited low-level resistance to carbapenems (MICs of imipenem and meropenem, 4 mg/L). Both isolates had a mutation or insertion in their OmpK35 and OmpK36 porins (details are provided in Supplementary Text S1), which have been found to result in reduced permeability of carbapenems [13], and have blaCTX-M genes (blaCTX-M-14 and blaCTX-M-15 in one isolate and blaCTX-M-3 in another) encoding extended-spectrum β-lactamases.

Co-occurrence of seven ST11 clones in the ICU with identification of a dominant clone 1 and a minor clone 2

We further typed ST11 isolates (n = 154) based on high-quality core SNPs. Strain 015093, the first isolate from a single patient who stayed in the ICU, was selected for long-read sequencing, allowing assembly into a complete genome (Supplementary Table S2). This complete genome of 015093 was used as the reference for phylogenetic analysis, with a total of 399 recombination-free variant sites shared by all isolates used to infer a phylogeny (Fig. 2). The 154 isolates could be robustly differentiated into seven distinct clones fully supported with 100% bootstraps, represented by clades (with multiple isolates) or branches (with a single isolate). Among these clones, those with the largest (n = 130) and the second largest (n = 15) populations, here designated as clone 1 and 2, belonged to the ST11 KL64 type (Fig. 2). The mean and the maximum intra-pairwise SNP distance of clone 1 and 2 were 6 and 9, and 18 and 20, SNPs, respectively, while the inter-pairwise SNP distance between these two ranged from 50 to 65 (Dataset S2 in the Supplementary). The remaining strains (n = 9) were located in two clades and three individual branches, here designated as clone 3 to 7. All were separated from each other by at least 35 SNPs on average (Dataset S2). The abovementioned seven clones could be assigned using a cutoff of 20 SNPs. Of the 21 clinically derived isolates, 67% (n = 14) were identified as clone 1, and 14% (n = 3) were identified as clone 2. This is largely reflective of the abundances in carriage (as determined by rectal swab screening), where 75.8% (n = 116) of the 153 rectal swab samples were assigned to clone 1, and 7.8% (n = 12) to clone 2 (Dataset S1).

Fig. 2: Phylogenomic tree of ST11 CRKP isolates (n = 154) collected in this study.
figure 2

The tree was inferred using strain 015093 as the reference. Information on the strains is available in Supplementary Dataset 1 and Dataset 2. The numbers of SNPs are shown in Supplementary Dataset 3. The phylogeny was inferred from core SNP sites under GTR model with site rate variation and a 1,000-bootstrap test. The tree was midpoint-rooted with bootstrap support over 50% shown in gradients. Strains from 7 well-separated clones were labelled with distinct colors which were clarified in the top-left corner.

Clone 1 was introduced into the hospital after clone 2

It is possible that the dominant clone 1 may have emerged in our hospital prior to the initiation of this study. Whole genome sequencing of CRKP isolates recovered from clinical samples has been performed in West China hospital since 2014. To investigate when clone 1 and clone 2 emerged in our hospital, we interrogated our genome collection and included all ST11 CRKP isolates from 2014 to March 26, 2017 and created an additional SNP-based phylogenetic analysis as described above. There were 25 ST11 KL64 CRKP clinical isolates including 6 from the ICU between December 11, 2015 and March 26, 2017 (Dataset S3 in the Supplementary). Among these 25 strains, 16 belonged to clone 1, while 3 belonged to clone 2 (Fig. 3). Strain 020130, which was recovered from a non-ICU patient, was the closest strain to clone 1. Although there were only 9 to 18 core-genome SNPs between 020130 and clone 1 genomes, strain 020130 was not considered to be part of clone 1, as it formed a separate distinct branch in the phylogenomic tree (Fig. 3). Strain 020120 that was recovered on January 5, 2017 from a patient in the Emergency Department in this hospital was the earliest isolate of clone 1. The earliest isolate of clone 2 was strain 095080 recovered on June 18, 2016 from another patient, also in the Emergency Department. Both strain 020120 and strain 095080 were most likely introduced into the hospital by these patients as they were recovered from samples collected on the day of admission to the Emergency Department. The first isolate (strain 040011) of clone 1 in the ICU was recovered from a urine sample collected on March 16, 2017, while the first isolate (strain 015247) of clone 2 emerged in the ICU on May 9, 2019. Both 040011 and 015247 were likely introduced to the ICU as they were recovered from samples collected within 48 h of admission to ICU. Therefore, clone 1 was introduced into our hospital later than clone 2 but emerged in the ICU earlier than clone 2.

Fig. 3: The time-calibrated phylogenetic tree of 171 strains of ST11 KL64 CRKP isolates in this study and collected in the hospital before the study.
figure 3

The dating was performed based on the maximum-likelihood tree inferred from core SNP sites under GTR model with site rate variation and a 1,000-bootstrap test. Strains collected in and before this study were highlighted in red and blue, respectively. Branches with bootstrap support over 50% were coloured in gradient and the average nucleotide substitution rate was estimated to be 4.85 per genome per year. MRCA refers to the most recent common ancestor. The 95% confidence intervals of estimated dates are as below: 1. October 1991 to May 2012; 2. November 2004 to February 2014; 3. May 2010 to November 2014; 4. May 2013 and December 2015; 5. September 2015 and July 2016; 6. January 2011 and April 2016; and 7. April 2015 and October 2016 with numbers 1 to 7 in circle. The estimated emergence of clone 1 and clone 2 corresponds to 7 and 5 marked with an asterisk, respectively.

To estimate the date of the emergence of ST11 KL64 clones, we also inferred a time-calibrated phylogenetic tree (Fig. 3) based on the 171 strains with the effective sample size of all estimates over 400 to ensure the duration of MCMC chain was long enough. Using a relaxed clock model, the mean substitution rate was estimated 4.85 (3.31 to 6.68) substitutions per genome per year. The estimated emergence of clone 1 and clone 2 was dated April 2016 (95% CI, September 2015 to July 2016) and May 2016 (95% CI, April 2015 to October 2016), respectively (Fig. 3), suggesting that the two clones emerged almost at the same time. Despite this, clone 1 went on to colonize 8.6-fold more patients than clone 2, while circulating within the ICU over the same time-period.

Clone 1 has decreased ability to form biofilms, shorter environmental survival, and attenuated virulence in vitro

Clone 1 has 13 unique SNPs including 11 in protein-coding sequences (CDS) and 2 in a spacer region (Supplementary Table S3), a truncated ulaB (Supplementary Fig. S1), and a 10.5-kb deletion which results in loss off type 1 fimbriae encoding fim genes (Supplementary Fig. S2, Table S4, and Table S5), when compared to clone 2 (Supplementary Text S2 for details). Only six SNPs were non-synonymous mutations. One of the missense mutations occurred in rcsC, which encodes a sensor kinase as part of the Rcs phosphorelay. The Rcs phosphorelay is conserved throughout the Enterobacteriaceae as an important signaling pathway and has also been found to actively participate in biofilm formation [14] and survival outside of the host [15]. Mutations of rcsC have been associated with decreased biofilm formation [16] and influence bacterial survival in environment [17]. In addition, there was a three-nucleotide insertion in igaA, leading to an aa insertion at position 662. igaA has been found to negatively regulate the Rcs phosphorelay [18,19,20].

Under anaerobic conditions, strain 020120 (the representative strain of clone 1, referred as C1_020120 thereafter to increase clarity) and strain 020130 (the closest strain of clone 1) exhibited no obvious biofilm formation (absorption at OD590 nm, mean ± standard deviations (SD), −0.01 ± 0.06 and 0.01 ± 0.02, respectively), while strain 020115 (the representative strain of clone 2, referred as C2_0201115 thereafter) formed biofilms (0.33 ± 0.06) (Fig. 4A). Under aerobic conditions, strain C1_020120 exhibited significantly less biofilm formation (0.06 ± 0.05) than strain 020130 (0.17 ± 0.07, P < 0.001) and strain C2_020115 (0.53 ± 0.10, P < 0.001) (Fig. 4A). In in vitro survival assays mimicking the ICU environment, E. coli ATCC 25922 survived only one day, while strain C1_020120 and C2_020115 survived 7 and 11 days (Supplementary Fig. S3), respectively.

Fig. 4: Biofilm formation, virulence assays, qRT-PCR, and ethanolamine utilization results.
figure 4

A Biofilm formation of C1_020120, C2_020115, and 020130. Absorption values of strains C1_020120, C2_020115, and 020130 (the closest strain of clone 1) at both OD 590 nm are shown. B Survival of G. mellonella after infection by bacterial strains. The effect of 1 × 103, 1 × 104, 1 × 105, and 1 × 106 CFU of each strain on survival of G. mellonella at 72 h after infection is shown. KP767, a hypervirulent K. pneumoniae, was used as a positive control, while E. coli DH5α was used as a negative control. C qRT-PCR for eutB (encoding the ethanolamine ammonia-lyase heavy chain) and eutR (encoding an HTH-type transcriptional regulator) genes for strain C1_020120 and C2_020115. Values are the mean ± standard deviation (SD). *p < 0.05. D utilization of ethanolamine (EA) as a sole carbon (adding NH4Cl) or nitrogen (adding glucose) source. C1_020120 and C2_020115 were able to use ethanolamine as a nitrogen source but not a carbon source.

The 50% lethal dose (LD50) at 72 h of strain C1_020120 against G. mellonella larvae was 1.4 × 105 CFU, one log higher than the 1.2 × 104 CFU of C2_020115, and higher than the 9.5 × 103 CFU of the hypervirulent K. pneumoniae strain KP767 (Fig. 4B). Therefore, strain C1_020120 displays decreased virulence compared to the clone 2 strain.

Clone 1 significantly outcompetes clone 2 in a murine intestinal colonization model

To confirm that mice had no CRKP intestinal carriage, a control group was orally administered meropenem for 3 days. No colonies were observed after overnight culture of fecal suspensions on SCAI medium containing 2 mg/L meropenem and 32 mg/L linezolid. Both tetracycline resistance in C2_020115 and amikacin resistance in C1_020120 were not transferrable and could therefore be used as the markers to differentiate the two strains.

In mice fed with either C1_020120 or C2_020115, the levels of colonization by both strains increased until day 2, and then steadily decreased for up to 28 days, with no significant level of difference between the two clones (Fig. 5A). A similar trend of colonization occurred in mice co-fed with both C1_020120 and C2_020115 (Fig. 5B). However, C1_020120 rapidly reach the highest density and then remained for up to 12 days. In contrast, levels of C2_020115 were almost undetectable by day three (Fig. 5C). This clearly shows that clone 1 has a significant fitness advantage over clone 2 when in direct competition in the intestines of mice.

Fig. 5: Murine intestinal colonization by C1_020120, C2_020115, and two strains in combination.
figure 5

A bacterial densities and the colonization days of fecal samples of mice fed with C1_020120 or C2_020115 alone. B bacterial densities and the colonization days of fecal samples of mice co-fed with C1_020120 and C2_020115 in combination. C the proportion of C1_020120 and C2_020115 in fecal samples of mice co-fed with C1_020120 and C2_020115 in combination.

Ethanolamine and 1,2-propanediol metabolism pathways are significantly upregulated in clone 1 during competition with clone 2

As C1_020120 and C2_020115 have highly similar genome sequences, we were unable to differentiate them in direct transcriptomic sequencing mice fecal samples co-colonizing both strains. As an alternative approach, we had to plate the mice fecal samples co-colonizing both strains in selective media to obtain the two strains separately, which were then subjected to transcriptomic sequencing. Although this approach would introduce confounding factors, we compared the separated in vitro growth of C1_020120 or C2_020115 on distinct selective media after passage together in murine gut to that of the same strain after passage alone, which also underwent in vitro growth on the same selective media to minimize confounding effects. Differential gene expression analysis of transcriptomic sequencing data revealed 74 significantly (>3 log2fold) upregulated genes in separated in vitro growth of C1_020120 on distinct selective media after passage together with C2_020115 in murine gut than that after passage alone (without C2_020115) (Dataset S4 in the Supplementary). In contrast, there were no significantly upregulated genes in C2_020115 in separated in vitro growth on distinct selective media after passage together with C2_020120 in murine gut. Among the 74 significantly upregulated genes, 50 have a functional annotation. Most of these (39/50) are involved in metabolic processes including propanediol utilization, ethanolamine utilization, and cobalamin biosynthesis. Interestingly, cobalamin (vitamin B12) is an essential part of the activation process for both ethanolamine and propanediol catabolism [21]. The ethanolamine utilization operon (the eut gene cluster), 1,2-propanediol operon (the pdu gene cluster), and the cobalamin biosynthesis operon (the cbi gene cluster) were significantly (>3 log2fold) upregulated in C1_020120 (Fig. 6, and Dataset S5 in the Supplementary). In contrast, none of ethanolamine and 1,2-propanediol metabolism pathways nor the cobalamin biosynthesis pathway had a significant change in levels of expression in C2_020115.

Fig. 6: Differential expression level of C1_020120 or C2_020115 in ethanolamine, 1,2-propanediol, and cobalamin pathways.
figure 6

Differential expression level of C1_020120 or C2_020115 between when co-fed together and when fed alone in pathways for ethanolamine utilization (the eut gene cluster), 1,2-propanediol metabolism (the pdu gene cluster), and cobalamin biosynthesis (the cbi/cob gene cluster) are demonstrated in colours. The exact change in log2fold of these genes are shown in Dataset S5, and their function is listed in Dataset S4.

To confirm this finding, we examined the expression of eutB and eutR for C1_020120 and C2_020115 after passage in murine gut by qRT-PCR. We found that the expression of eutB and eutR of C1_020120 was significantly increased (1.30 ± 0.25 [P = 0.037] and 2.01 ± 0.88 [P = 0.046] folds, respectively) in separated in vitro growth on distinct selective media after passage together with C2_020115 in murine gut than that after passage alone. In contrast, the expression of eutB and eutR of C2_020115 in separated in vitro growth on distinct selective media after passage together with C1_020120 was only 0.46 ± 0.05 and 0.71 ± 0.23 folds, respectively, of that after passage alone (Fig. 4C). On the other hand, we also compared the expression of eutB and eutR between C1_020120 and C2_020115 in separated culture in LB broth and after separate passage in murine gut. We did not find significant differences in the expression of eutB and eutR between C1_020120 and C2_020115 in broth culture (P = 0.372 and P = 0.368) nor after passage alone in murine gut (P = 0.492 and P = 0.481) (Supplementary Fig. S4).

Clone 1 has a fitness advantage over clone 2 in the presence of ethanolamine as a sole nitrogen source

Previous studies have found that ethanolamine can be used as nitrogen and carbon sources by some bacteria [22, 23]. We found that both C1_020120 and C2_020115 were able to use ethanolamine as a nitrogen source by measuring OD600 nm of individual overnight bacterial cultures (mean ± SD, 0.68 ± 0.03 and 0.66 ± 0.04, respectively), but not a carbon source (0.05 ± 0.01 and 0.05 ± 0.00, respectively) (Fig. 4D). When both strains were co-cultured in the presence of ethanolamine as the sole nitrogen source, C1_020120 had a clear fitness advantage (w, 1.19 ± 0.15) over C2_020115.

A chromosomal inversion occurs within a prophage upstream of the eut operon in C1_020120

There are no SNPs or indels in clone 1 which would account for a such a significant phenotypic adaptation in comparison to clone 2. We, therefore, re-analyzed the combined long-read and short-read complete genome assemblies of C1_020120 and C2_020115 to look for structural rearrangements. We discovered that the 10.5-kb deletion resulting in loss off type1 fimbriae encoding fim genes (Fig. S2) in clone 1 occurred in the region of the genome containing the pdu-cbi and eut operons (Fig. 7). Furthermore, C1_020120 and C2_020115 had a rearrangement (reversion) of a large 160-kb region due to the recombination between the flanking two 226-bp group II introns, switching confirmation (Fig. 7). In strain C1_020120, The two group II introns and their reverse transcriptase-encoding genes ltrA belong to two different intact prophages, 61.3-kb and 53.9-kb in size, respectively. There were no other structural or genome architecture differences between the two clones, and no other SNPs or indels which would affect the expression of the eut operon. Mapping reads of all our ST11 isolates against the clone 1 reference genome confirmed that all clone 1 isolates have this structural rearrangement, which is notably absent from clone 2. This 160-kb region was also found to be fully conserved across all publicly available and fully completed K. pneumoniae genomes, which consisted of 205 STs (n = 1,012, accessed on 2021-10-28, Dataset S6). In the vast majority of genomes (n = 997, 98.5%), this region occurs in the same orientation as clone 1, with only 15 genomes matching the orientation found in clone 2.

Fig. 7: The reversion of a region between the pducbi module and the eut operon.
figure 7

Compared to that of C2_020115, a large region between the module of pdu (for utilizing 1,2-propanediol) and cbi (for cobalamin biosynthesis) of C1_020120 was reversed. This reversed region is flanked by two identical 226-bp group II introns and their reverse transcriptase (RT)-encoding genes ltrA. In strain C1_020120, the two group II introns and ltrA genes belong to two different prophages, 61.3-kb prophage_1 (in light green) and 53.9-kb prophage_2 (in cyan), respectively. The two prophages are intact and have the attachment sites, attL and attR, at both sides. Homologous recombination between the two group II introns could reverse the intervening region including parts of the two prophages. The 10.5-kb region absent from clone 1 (Fig. S2) is also located between the pducbi module and the eut operon and is indicated as ‘deletion region’ here. Other genes shown are fuc (encoding metabolism for L-fucose), gud (encoding glucarate dehydratase), hmu (encoding an ABC transporter complex involved in hemin import), PTS (encoding aphosphoenolpyruvate-dependent sugar phosphotransferase system), hyp-hyc (involving in hydrogenase biosynthesis), gut-srl (involving in metabolism of the hexitol D-glucitol), yga (encoding inner membrane and ribosome binding proteins with putative rhodanese activity), nor (for the expression of anaerobic nitric oxide [NO] reductase), znu (involving in the high-affinity zinc uptake transport), pro (involving in glycine betaine and proline betaine uptake), and nrd (encoding a ribonucleotide reductase system). This figure is not scaled.

Leave a Comment