Emergence and Inter- and Intrahost Evolution of Pandrug-Resistant Klebsiella pneumoniae Coharboring tmexCD1-toprJ1, blaNDM-1, and blaKPC-2

Pandrug-resistant (PDR) Klebsiella pneumoniae poses a great challenge to public health, and tigecycline is an essential choice for antimicrobial treatment. In this study, we reported the emergence of PDR K. pneumoniae coharboring tmexCD1-toprJ1, blaNDM-1, and blaKPC-2, which belongs to ST22 and ST3691. ABSTRACT Klebsiella pneumoniae is capable of acquiring various exogenous genetic elements and subsequently conferring high antimicrobial resistance. Recently, a plasmid-mediated RND family multidrug efflux pump gene cluster, tmexCD1-toprJ1, was discovered in K. pneumoniae. In this study, we analyzed tigecycline-resistant K. pneumoniae isolates from patients from surveillance from 2017 to 2021. In addition to phenotype detection, including growth curves, plasmid transferability and stability, hypermucoviscosity, biofilm formation, and serum survival, by whole-genome sequencing, we analyzed the phylogenetic relationships of the isolates harboring tmexCD1-toprJ1 and discovered the composition of plasmids carrying tmexCD1-toprJ1. In total, we discovered that 12 tigecycline-resistant isolates from 5 patients possessed tmexCD1-toprJ1, designated sequence type 22 (ST22) and ST3691. An ST11 isolate harbored a partial tmexD1, and a complete toprJ1 (tmexC1 was lost) was tigecycline sensitive. All the ST22 tigecycline-resistant isolates coharbored tmexCD1-toprJ1, blaNDM-1, and blaKPC-2. tmexCD1-toprJ1 was encoded by a novel IncU plasmid in ST22 and an IncFIB/HI1B plasmid in ST3691, which presented differences in mobility and stability. Interestingly, isolates from the same patients presented heteroresistance to tigecycline, not only among isolates from different specimens but also those from the same sample, which might be attributed to the differential expression of tmexCD1-toprJ1 due to the dynamic genetic heterogeneity caused by relocating tmexCD1-toprJ1 close to the replication origin of plasmid. Here, we reported the emergence of K. pneumoniae isolates coharboring tmexCD1-toprJ1, blaNDM-1, and blaKPC-2. The results highlight the impact of in vivo genetic heterogeneity of tmexCD1-toprJ1-carrying elements on the in vivo variation of tigecycline resistance, which might have notable influences on antimicrobial treatment. IMPORTANCE Pandrug-resistant (PDR) Klebsiella pneumoniae poses a great challenge to public health, and tigecycline is an essential choice for antimicrobial treatment. In this study, we reported the emergence of PDR K. pneumoniae coharboring tmexCD1-toprJ1, blaNDM-1, and blaKPC-2, which belongs to ST22 and ST3691. By whole-genome analysis, we reconstructed the evolutionary map of the ST22 ancestor to become the PDR superbug by acquiring multiple genetic elements encoding tmexCD1-toprJ1 or blaNDM-1. Importantly, the genetic contexts of tmexCD1-toprJ1 among the ST22 isolates are different and present with various mobilities and stabilities. Furthermore, we also discovered the heterogeneity of tigecycline resistance during long-term infection of ST22, which might be attributed to the differential expression of tmexCD1-toprJ1 due to the dynamic genetic heterogeneity caused by relocating tmexCD1-toprJ1 close to the replication origin of plasmid. This study tracks the inter- and intrahost microevolution of the superbug PDR K. pneumoniae and highlights the importance of timely monitoring of the variation of pathogens during antimicrobial treatment.

Inter-and Intrahost Evolution of PDR K. pneumoniae Microbiology Spectrum Molecular characteristics, features of plasmid types, and resistance genes. Further whole-genome sequencing (WGS) and analysis revealed the molecular characteristics of the isolates that correspond to the observed phenotypes. The 11 PDR isolates among the 12 tigecycline-resistant K. pneumoniae belonged to ST22, and the one MDR isolate was ST3691. All of the isolates harbored tmexCD1-toprJ1. A tigecycline-sensitive ST11 isolate (PEKP3038) harbored truncated tmexD1 and toprJ1, and tmexC1 was lost. We also obtained the genomes of three tigecycline-sensitive ST22 isolates (PEKP4079, PEKP3116, and PEKP3119) and an ST3691 isolate (PEKP4245), which were tmexCD1-toprJ1 negative. All 14 ST22 isolates harbored IncFIB-and IncFII-type plasmid replicons (Fig. 1A). Most of them (13/14) also harbored the IncU-type plasmid replicon, including two isolates without tmexCD1-toprJ1. Compared with the tmexCD1-toprJ1-negative isolates, all 11 tigecyclineresistant ST22 isolates harbored resistance genes, including strAB, tet(A), aac(3)-IId, bla NDM-1 , Inter-and Intrahost Evolution of PDR K. pneumoniae Microbiology Spectrum and catA2 in addition to tmexCD1-toprJ1. Similarly, compared with the tmexCD1-toprJ1negative ST3691 isolate, the tmexCD1-toprJ1-carrying ST3691 isolate PEKP4245 comprised an extra IncH1B plasmid replicon and a number of resistance genes (Fig. 1B). The distribution of resistance genes was in line with the resistance phenotypes (Table S2). Genetic characteristics of tmexCD1-toprJ1-bearing plasmids. To explore the characteristics of tmexCD1-toprJ1-bearing genetic elements, we obtained the complete circularized sequences of chromosomes and plasmids of these isolates. We found that the tmexCD1-toprJ1 gene clusters were carried by two types of plasmids. The plasmids from ST22 isolates were highly similar (Fig. 2). These plasmids were ;230 to 323 kb in size and comprised the IncU replicon, similar to two tmexCD1-toprJ1 negative plasmids (pPEKP3116-217 and pPEKP3119-216, with ;80.44 to 99.89% coverage and ;97.07 to 99.80% identity). We identified two major insertions in tmexCD1-toprJ1-bearing IncU plasmids. The first one was adjacent to an insertion sequence 26 (IS26) element encoding tmexCD1-toprJ1 and six other resistance genes, and the other insertion encoding arr-3 and bla OXA-1 was adjacent to an ISKpn26 element (Fig. 2). The core genetic environment of tmexCD1-toprJ1 comprised Tn501-int-int-hp-hp-tnfxB1-tmexC1-tmexD1-toprJ1 (Fig. 3A). This composition was most similar to the corresponding sequences of the IncFIA plasmid pHNAH8I-1 of ST1 K. pneumoniae and the IncFII(K) plasmid pKA9-4 of ST37 K. pneumoniae isolated from chickens in China ( Fig. S1) (10,13). In addition, this structure was located within the large insertion adjacent to IS26 and comprised other resistance genes, transposase genes, and ISs, suggesting that this novel IncUtype tmexCD1-toprJ1-bearing plasmid might be generated via multiple steps of recombination. In addition, tigecycline-resistant ST22 also harbored multiple IncF plasmids encoding bla KPC-2 , bla NDM-1 , and heavy metal resistance genes (Fig. 4).
Inter-and intrahost genetic heterogeneity and microevolution of the ST22 population. To explore the evolutionary history of the ST22 isolates from our hospital, we further analyzed the phylogenetic relationships and genetic differences. The tmexCD1-toprJ1-harboring isolates clustered together with three tmexCD1-toprJ1-negative isolates (Fig. S1D). One of the three tmexCD1-toprJ1-negative isolates, PEKP4079, showed only one nucleotide difference compared with the tmexCD1-toprJ1-harboring isolate PEKP4087. Between two other tmexCD1-toprJ1-negative isolates, PEKP3116 and PEKP3119, only two single nucleotide polymorphisms (SNPs) were identified, which could be regarded as the same clone. Meanwhile, we found 549 shared SNPs within the tmexCD1-toprJ1-harboring cluster compared with PEKP3116 and PEKP3119 (Table S1). In contrast, only 1 to 11 SNPs were found among the tmexCD1-toprJ1-harboring cluster. We also identified 2 lineages among the tmexCD1-toprJ1-harboring isolates, differing by 11 common SNPs to each other ( Fig. 1A and C). Lineage A included the isolates from patient 3 (PEKP4026) and patient 2 (PEKP4104, PEKP4087, and PEKP4069), and lineage B included the isolates from patient 4 (81, PEKP4007, PEKP5001, PEKP5006, PEKP4009, and PEKP3087) and patient 5 (D6).
We then reconstructed the evolutionary map by using chromosomal SNPs first and then used the variations in plasmids and clinical information to illustrate the details (Fig. 5). During inter-and intrahost evolution, plasmids acquire various AMR and/or heavy metal resistance genes (Fig. 4). In the ancestor ST22 isolate, whose genetic backbone was similar to PEKP3116 and PEKP3119, an insertion encoding tmexCD1-toprJ1 might occur on the IncU plasmid, as well as the other insertion encoding bla NDM-1 on the IncFII plasmid, to become the most recent ancestor (MRCA) of the two ST22 lineages harboring tmexCD1-oprJ1. The two insertion events might occur together or gradually. This MRCA isolate might accumulate three SNPs to obtain a 10-kb ColRNAI-type plasmid and a 217-kb IncFIB/IncFII plasmid together or gradually to evolve lineage A, meanwhile accumulating 11 SNPs in parallel to The matched regions between two sequences are displayed by light blue blocks, and the identities are marked. The arrows represent the genes related to resistance and transfer (red, AMR; green, integrase recombinase and transposase; purple, transfer associated; dark blue, plasmid replication; gray, other functions).
Inter-and Intrahost Evolution of PDR K. pneumoniae Microbiology Spectrum evolve lineage B. An isolate of lineage A, PEKP4079, was transmitted from patient 2 to another patient, and the IncU plasmid seemed to be lost during interhost transmission. Both lineages also exhibited intrahost population heterogeneity. Three lineage A isolates, PEKP4104, PEKP4087, and PEKP4069, were isolated from patient 2 for three consecutive months. There were 2 to 5 chromosomal SNPs among these phylogenetically close isolates, and the IncU plasmid bearing tmexCD1-toprJ1 varied from ;230 to 307 kb in size, with large insertions, deletions, and rearrangements. PEKP4104 also lost the IncFIB/IncFII plasmid. Six lineage B isolates were isolated from patient 4 for three consecutive months. There were 1 to 20 chromosomal SNPs, and the tmexCD1-toprJ1-bearing IncU plasmid varied from ;286 to 323 kb in size. In addition, the long-fragment variations mainly occurred in two sites: one was related to the IS26 element near the 39 end of the sequence, and the other was related to the inverted repeats (IRs) comprising IS26, qnrS1, and ISKpn19. There was only one copy of the sequence located near the IncU replicon in the ancestor plasmid, and its reverse complement was inserted downstream of tmexCD1-toprJ1 in most of the tmexCD1-toprJ1-bearing IncU plasmids, except pPEKP4104-307 and pPEKP4069-230 of lineage A. The fragments between the two copies of IRs were ;182 kb in size and inverted in pPEKP4007-323, pPEKP5006-322, and pPEKP4009-286 of isolates from patient 4 and pD6-300 of isolate D6 from patient 5.
Differential expression of tmexCD1-toprJ1 contributed to tigecycline heteroresistance. In line with the intrahost population heterogeneity, we also observed varied levels of tigecycline resistance among tmexCD1-oprJ1-harboring ST22 K. pneumoniae isolated from the same patients ( Fig. 6; Table S2). The MICs of tigecycline varied from 6 to 16 mg/L against the isolates from patient 4 and 4 to 24 mg/L against the isolates from patient 2. The highest MICs were observed during treatment or pretreatment with Inter-and Intrahost Evolution of PDR K. pneumoniae Microbiology Spectrum tigecycline in patient 4 (16 mg/L against PEKP4007 and PEKP5006) and patient 2 (24 mg/ L against PEKP4087). Interestingly, the MICs decreased against the isolates when tigecycline treatment was withdrawn for both patients and increased again in patient 4, who reused tigecycline (Fig. 6). By quantitative PCR (qPCR) detection of the tmexCD1-toprJ1 genes, the copy numbers were approximately one per cell in all the isolates, indicating that no copy number variation occurred on the tmexCD1-toprJ1-bearing plasmid (DNA level). By reverse transcription-qPCR (RT-qPCR) detection of tmexCD1-toprJ1 mRNA, we observed differential expression of these genes in the intrahost population ( Fig. S3 and S4). In line with the variation in the MICs of tigecycline, the expression of tmexCD1-toprJ1 exhibited the same trends. The growth of isolates was not significantly different within the two intrahost populations, indicating that there might not be significant fitness cost due to the presence and expression of tmexCD1-toprJ1. PEKP4087 exhibited the highest MIC (24 mg/L) of tigecycline and presented no significant differences in the growth curve compared with other isolates (PEKP4069, PEKP4104) isolated from patient 2. Moreover, isolate PEKP4007 from patient 4 presented higher MICs (16 mg/L) of tigecycline and a more prominent growth rate than the other isolates (Fig. S5). In addition, we observed an interesting phenomenon that the large fragment containing tmexCD1-toprJ1 in the IncU plasmid might invert naturally. Among the assembled plasmid sequences, this region might be randomly inverted (Fig. 2). We designed a PCR experiment to detect the inversion in PEKP5001 and PEKP5006 and found that the inversion and original status both existed in the two isolates, although the assembled sequences were different (Fig. S6). We speculated that this region might be unstable and related to the differential expression of tmexCD1-toprJ1 and different levels of tigecycline resistance.

DISCUSSION
Here, from long-term surveillance within our hospital, we comprehensively demonstrated the clinical, epidemiological, and genomic characteristics of pandrug-resistant K. pneumoniae coharboring tmexCD1-toprJ1, bla KPC-2 , and bla NDM-1 . Importantly, the ST22 K. pneumoniae isolates caused infections in patients with poor prognosis. The full tmexCD1-toprJ1 gene cluster was discovered in two types of plasmids harbored by K. pneumoniae strains of two different STs. Genetic recombination and acquisition occurred during interand intrahost evolution. Importantly, it is the expression of tmexCD1-toprJ1 that plays a key role in the heteroresistance of tigecycline. The tmexCD1-toprJ1 and its subtype were mainly discovered among Enterobacteriaceae, including K. quasipneumoniae, Klebsiella variicola, Klebsiella michiganensis, and Proteus mirabilis (14,15,19). This efflux pump confers resistance to multiple drugs, including tigecycline, quinolones, cephalosporins, and aminoglycosides, in K. pneumoniae strains of animal origin and was then widely identified in K. pneumoniae strains derived from animals (13,17,18). To date, tmexCD1-toprJ1-harboring K. pneumoniae strains have also been discovered within various STs in hospitals, especially the endemic clones ST15 and ST11 (10,14,21,22). Additionally, these genes were also detected in bacteria from various environmental samples, including urban drainage and food samples from market and slaughterhouse sewage (12,16,20). The spread of the tmexCD1-toprJ1-like gene cluster should be of great concern. Two tmexCD1-toprJ1-negative strains were isolated 3 months earlier than tmexCD1-toprJ1positive strains in our hospital and belonged to a phylogenetically earlier lineage. In addition, it is reported that tmexCD1-toprJ1 newly emerged into K. pneumoniae, but no tmexCD1-toprJ1-positive ST22 K. pneumoniae strains were reported previously. Therefore, we speculate that the tmexCD1-toprJ1-positive ST22 strains in our hospital evolved from the tmexCD1-toprJ1-negative strain by acquisition of the tmexCD1-toprJ1-harboring plasmid.
Of note, transposon 5393 (Tn5393) and IS26 conferred tmexCD1-toprJ1 acquisition. We also discovered the two types of genetic contexts of tmexCD1-toprJ1. In ST22 K. pneumoniae, the genetic element was flanked with Tn5393 at the 59 end, whereas IS26 was flanked in ST3691 and ST11. Most previous studies confirmed that the plasmid harboring tmexCD1-toprJ1 could be successfully transferred into E. coli (10,14,22). Importantly, it was confirmed that IS26, Tn5393, and ICEKP were conferred with tmexCD1-toprJ1 acquisition and mobilization (9,10,16,18). A previous study also demonstrated that the tmexCD1-toprJ1 plasmid was unsuccessfully transferred into E. coli but successfully transferred into the hygromycin-resistant ST11 K. pneumoniae HS11286YZ6 (12). Conjunctive elements might play important roles in the rapid transmission of tmexCD1-toprJ1 (10,19,22). In this study, we also discovered that the tmexCD1-toprJ1 cluster was inserted into a nonconjugative IncU plasmid of ST22 K. pneumoniae via Tn5393, and ST3691 K. pneumoniae might have acquired tmexCD1-toprJ1 bearing the IncFIB/IncHI1B plasmid via plasmid conjugation.
Most of the tmexCD1-toprJ1-like positive strains identified to date were recently discovered in China (9, 10), in addition to several reports of strains from Vietnam and Kenya (11,16,20). A previous study estimated that tmexCD1-toprJ1 is rare in clinical K. pneumoniae strains in China (,0.1%) (9), and the positivity rate of animal-derived K. pneumoniae was 3.37% (12). Surprisingly, infection caused by ST3691 and ST11 was defined as communityacquired infection in our study. Furthermore, the plasmid harboring tmexCD1-toprJ1 was stable after the 30th passage and tended to acquire heavy metal resistance genes during in vivo evolution, indicating that its persistent existence within the hospital environment might impel transmission and evolution, driving the emergence of the rapidly transferable "superbug." Therefore, long-term genomic surveillance is essential, especially in areas with heavy antibiotic consumption.
Since the first report of tmexCD1-toprJ1, various MDR strains harboring tmexCD1-toprJ1 and other AMR genes have been recovered (15). Importantly, the convergence and cotransmission risk of both resistance genes within endemic clones emerged (21). Sun et al. reported the coexistence of tmexCD1-toprJ1 and mcr-8 on the same plasmid in Inter-and Intrahost Evolution of PDR K. pneumoniae Microbiology Spectrum animal-derived K. pneumoniae strains (10). Moreover, it has been reported that CRKP strains coharbored the carbapenemase gene and tmexCD1-toprJ1 in the same plasmid (20,22). A previous study concluded that selective pressure imposed by the heavy use of older tetracyclines in China could have contributed to the emergence of this efflux pump (9). In this study, we found that ST22 K. pneumoniae became a PDR strain by acquiring and coharboring plasmids carrying bla KPC-2 and bla NDM-1 in the hospital and causing fatal infections.
A previous study reported that the K. pneumoniae strains presented with various MDR/ virulence phenotypes during intrahost evolution, which might confer with different host response. During in vivo evolution, the pathogen population presented dynamic changes and intrahost heterogeneity under antibiotic and/or immune selective pressure (23). In this study, we found that the expression, but not the copy number, of tmexCD1-toprJ1 contributed to the varied MICs of tigecycline when the population was under tigecycline pressure. We discovered genomic and phenotypical heterogeneity, whether within the same sample or among consecutive samples from the same host. Both the PEKP4007 and PEKP5006 strains isolated from the same sample (11 SNPs) showed increased MICs of tigecycline during in vivo evolution but presented a significant growth rate. PEKP4009 and PEKP3087 presented similar growth rates but displayed different MICs of tigecycline. A previous study reported that gene mutations were responsible for tigecycline or colistin resistance during treatment with tigecycline and polymyxin (24). However, we identified no mutation in tmexCD1-toprJ1. A previous study also demonstrated that gene expression changes triggered by ineffectual antibiotic treatment cause pathogens to transition between states of low and high virulence (25). An inversion in the IncU plasmid that caused tmexCD1-toprJ1 to be relocated closer to the plasmid replicon might be responsible for the higher tmexCD1-toprJ1 expression and higher level of tigecycline resistance. It has been confirmed that the genes that are closer to the origin site (ori proximal gene) are replicated first and are more highly expressed (26,27). In E. coli, higher expression of oriCproximal genes has been discovered on the chromosome, which is caused by the replication-induced increase of copy numbers (up to eight copies) of the oriC proximal region (28,29). However, we did not find any evidence in previous studies about this phenomenon in the expression of genes in plasmids, which should be validated in the future.
In conclusion, by genomic surveillance, we discovered novel clinical tmexCD1-toprJ1harboring K. pneumoniae subtypes and plasmids. Although the occurrence was rare, ST22 K. pneumoniae coharboring tmexCD1-toprJ1, bla NDM-1 , and bla KPC-2 emerged in our hospital and caused fatal infection. Furthermore, in ST22 K. pneumoniae, the plasmid harboring tmexCD1-toprJ1 displayed intrahost heterogeneity, and the inversion caused relocation of tmexCD1-toprJ1 close to the replication origin of plasmid, which might be associated with the high expression of tmexCD1-toprJ1. These results call for further investigation of the prevalence of tmexCD1-toprJ1 harboring K. pneumoniae, as well as intensive study of the mechanism of heteroresistance to improve antimicrobial treatment in the clinic.

MATERIALS AND METHODS
Patients and strains. From 2017 to 2021, we conducted a cohort study at the Peking University Third Hospital to reveal the dynamic genomic epidemiology of Klebsiella spp. Clinical information of the infected patients was obtained from electronic medical records in the hospital. We calculated the Charlson comorbidity index (CCI) of the patients to evaluate the severity of illness and defined death or withholding life-sustainable therapy within 28 days as poor prognosis.
All the Klebsiella species isolates were stored in a 280°C freezer. The Klebsiella species isolates were initially confirmed by mass spectrometry and then by the Vitek 2 compact system. Finally, the Klebsiella species isolates were further identified by whole-genome sequencing (WGS) and analyses using Kleborate software (30). Isolates that harbored the tmexCD1-toprJ1-like gene cluster were analyzed together in the study. To clarify the unique genetic context of tmexCD1-toprJ1, we chose the same STs as the matched isolates, regardless of whether they were isolated from the same host.
Plasmid transferability and stability. To identify the transferability of plasmids carrying tmexCD1-toprJ1, we selected E. coli J53 as the recipient to conduct the conjugation experiment (12,20). Briefly, the donor and recipient were mixed at a ratio of 1:1 and then cocultured in LB broth overnight. Subsequently, the mixture was spotted on MacConkey agar containing 1 mg/L tigecycline and 100 mg/L sodium azide. After overnight culture, the transconjugants were finally screened. To detect the plasmid stability, the tmexCD1-toprJ1-positive isolates were subcultured serially to the 30th passage (34). The descendants were also evaluated by AST and qPCR to determine the phenotype and genotype.
Genomic DNA and RNA extraction and qPCR. Whole-genome DNA was extracted using Tiangen magnetic universal genomic DNA kit (catalog no. DP705). To determine the copy number of the tmexCD1-toprJ1 gene cluster, qPCR was conducted using NEB Luna universal qPCR master mix (catalog no. M3003S). The primers are listed in Table S4 in the supplemental material. To determine the expression level of tmexCD1-toprJ1 in isolates continually isolated from the same patients, whole-genome RNA was extracted using Tiangen RNAprep pure cell/bacteria kit (catalog no. DP430), and RT-qPCR was also conducted using Qiagen QuantiTect SYBR green RT-PCR kit (catalog no. 204243). The PCR amplification consisted of the initial denaturation for 3 min at 95°C, followed by 35 cycles of denaturation for 30 s at 95°C, annealing for 30 s at 54°C, and extension for 30 s at 72°C, and then the final extension for 5 min at 72°C. We used the 16S rRNA gene as an internal control in PCR analysis.
WGS and public data collection. DNA previously extracted was further used for next-generation sequencing, and the libraries were prepared using Nextera technology as described in our previous study (2). Paired-end reads of 150 bp were generated by an Illumina NovaSeq 6000, and all the enrolled isolates were further sequenced by the Nanopore minION platform. Additionally, all of the ST22 and ST3691 genomes that were available in the GenBank database (28 genomes, accessed on 18 October 2021) were included in downstream analysis, resulting in 40 ST22 and 4 ST3691 genome sequences (Table S1). We also collected 18 previously published tmexCD1-toprJ1-positive plasmids (accessed on 18 October 2021) to understand their genetic contexts.
Bioinformatic analysis. Complete genomes were assembled by combining highly accurate short reads by Illumina sequencing and long reads by Nanopore sequencing via de novo assembly with a hybrid strategy according to published methods using Unicycler v0.4.4 (35). Primary gene prediction and annotation were performed by Prokka software as previously described (36). STs and serotypes were further detected by Kleborate software (30). Resistance genes, virulence genes, IS sequences, and plasmid replicon types were determined based on the comparison with ResFinder (37), Virulence Factor Database (38), IsFinder (39), and plasmidFinder (40) databases by BLAST.
To perform phylogenetic analysis, the complete chromosomal sequence of isolate PEKP3116 was used as the reference for analysis. The short reads of each isolate by Illumina sequencing were mapped to the reference by Bowtie 2 v2.2.8 first (41). The single nucleotide polymorphisms (SNPs) of each isolate compared with the reference were identified by using SAMtools v1.9, and the polymorphic sites among all isolates were combined together according to the reference genome using the previously published intrahost single-nucleotide variant (iSNV) calling pipeline (https://github.com/generality/iSNVcalling). We retained the high-quality SNPs (.5 reads of mapping quality .20) and removed the recombination sites detected by Gubbins v2.4.1 (42). Finally, the concatenated sequences of filtered polymorphic sites conserved in all genomes (core genome SNPs [cgSNPs]) were identified to construct the phylogenetic tree using the maximum-likelihood method by IQ-TREE v2.1.2 (43).
Ethical approval statement. All methods were performed in accordance with the Declaration of Helsinki. The protocol for this study was approved by the Peking University Third Hospital Medical Science Research Ethics Committee (M2021545). Due to the retrospective nature of the study, the need for approval was waived by the Peking University Third Hospital Medical Science Research Ethics Committee, and all the patient data enrolled in this study were anonymized.
Data availability. The genome sequences in this study were deposited into the NCBI database under BioProject accession no. PRJNA793277. The data sets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 2.5 MB. SUPPLEMENTAL FILE 2, XLSX file, 0.05 MB.

ACKNOWLEDGMENTS
This work was supported by National Natural Science Foundation of China (82200012) and Beijing Key Clinical Specialty Program (010071).
The research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. M.L. and N.S. designed the study, supervised the whole project, and revised the manuscript. C.L., P.D., and P.Y. performed the bioinformatic analysis and data collection, performed the experiments, and contributed to writing and revising the manuscript. J.Z. performed the experiments, collected the clinical data, and revised the manuscript. J.Y. participated in the experimental performance. All authors read and approved the final manuscript.