全基因组测序技术筛选儿童肺炎链球菌致病因子

Screening of pathogenic factors of Streptococcus pneumoniae in children based on whole-genome sequencing technology

目的全基因组测序技术筛选不同类型肺炎链球菌(感染与定植菌株)的分子特征,揭示其致病相关标志物。方法2015年12月-2021年1月在柳州市妇幼保健院收集儿童的临床常见标本(血液、脑脊液、胸膜液、痰液、鼻窦吸入和耳分泌物等)为感染组。收集2020年10月-12月在佛山市采用整群抽样方法抽取7所幼儿园,开展健康幼儿园儿童的鼻咽拭子采样为定植组。感染菌株为医院感染儿童的临床感染肺炎链球菌,定植菌株为社区健康儿童的鼻咽定植肺炎链球菌。采用全基因测序技术全面检测肺炎链球菌的耐药基因和毒力基因。采用单因素分析(χ^(2)检验)和机器学习方法(随机森林)筛选出儿童肺炎链球菌的致病相关标志物。结果感染菌株pce、pitA、pitB、sipA、rrgA、rrgB、rrgC、srtG1、srtG2、srtC1、srtC2、srtC3、nanA、nanB、cps4A、cps4B、cps4D和zmpC毒力基因携带率高于定植菌株(P<0.05);cbipG、pfbA、lytA和lytB基因携带率低于定植菌株(P<0.05)。感染菌株ermC、mefA和msrD耐药基因携带率高于定植菌株(P<0.05);cat(pC194)基因的携带率低于定植菌株(P<0.05)。随机森林模型筛选出6个致病分子特征(毒力基因pce、rrgA、lytA、lytB和zmpC,耐药基因msrD),交叉验证准确率为70.11%,曲线下面积(AUC)为0.73。结论运用随机森林有效筛选肺炎链球菌致病相关标志物,为追溯高致病性传染源和开展精准的靶向干预提供遗传学证据。

OBJECTIVE To screen the molecular characteristics of Streptococcus pneumoniae strains(the infection and colonization strains)by the whole-genome sequencing technology so as to reveal the pathogenesis-related markers.METHODS The common clinical specimens(blood,cerebrospinal fluid,pleural fluid,sputum,paranasal sinus inhalation and otic secretions)were collected from the children who were treated in Liuzhou Maternity and Children Healthcare Hospital from Dec 2015 to Jan 2021 and were assigned as the infection group.7 kinder-gartens of Foshan were randomly selected by cluster sampling method from Oct 2020 to Dec 2020,the nasopha-ryngeal swab specimens were collected from the healthy children of the kindergartens and were assigned as the col-onization group.The strains of infection were the S.pneumoniae strains causing the clinical infection in the chil-dren,and the strains of colonization were the nasopharyngeal colonized S.pneumoniae strains in the healthy chil-dren from communities.The drug resistance genes and virulence genes in the S.pneumoniae strains were fully detected by the whole-genome sequencing technology.The pathogenesis-related markers of the S.pneumoniae strains were screened out by means of univariate analysis(chi-square test)and machine learning method(random forest).RESULTS The carrying rates of virulence genes pce,pitA,pitB,sip A,rrgA,rrgB,rrgC,srtG1,srt G2,srtC1,srtC2,srtC3,nanA,nanB,cps4A,cps4B,cps4D and zmpC of the infection strains were higher than those of the colonization strains(P<0.05).The carrying rates of cbipG,pfbA,lytA and lytB genes of the infec-tion strains were lower than those of the colonization strains(P<0.05).The carrying rates of drug resistance gene ermC,mef A and msrD of the infection strains were higher than those of the colonization strains(P<0.05);the carrying rate of cat(pC194)gene of the infection strains was lower than that of the colonization strains(P<0.05).Totally 6 pathogenic molecular characteristics(virulence genes pce,rrgA,lytA,lytB and zmpC,drug resistance gene msrD)were screened out by random forest model,the accurate rate of cross validation was 70.11%,and the area under curve(AUC)was 0.73.CONCLUSION The random forest model may effectively screen out the patho-genesis-related markers of the S.pneumoniae strains and provide genetic evidence for tracing high-pathogenicity in-fectious sources and carrying out precise targeted interventions.