Abstract
Objective:
Detection of tuberculosis laboratory cross-contamination using whole-genome sequencing.
Methods:
A total of 22 M. tuberculosis strains with high genotypic homology from one hospital were collected during the drug resistance surveillance. Genome sequencing and epidemiological investigation were conducted to determine the occurrence of cross-contamination.
Results:
The pair wise comparison between the genomes in each cluster indicated that 15 (71.4%) of 21 strains with available genomic data had no SNP differences with at least one other strain within the same cluster. The analysis of the specimen collection time found that, among the 16 strains collected on the same day, 14 (87.5%) of them had no SNP differences with one another strain; meanwhile, among the strains within the same cluster whose SNP distance was 0, 93.3% (14/15) of them had the same collection time, suggesting that these findings were most likely caused by cross contamination.
Conclusion:
A high proportion of M. tuberculosis strains with genotypic homology from the single institute that shared the same process time period was most likely caused by the cross contamination. Whole genome sequencing analysis can help to determine the occurrence of cross contamination.
Keywords: Mycobacterium tuberculosis, Cross-contamination, Whole-genome sequencing
1. Introduction
The World Health Organization (WHO) has reported that about one third of the global population is infected with Mycobacterium tuberculosis. In 2014, there were about 9,600,000 new patients with tuberculosis worldwide, and 1,500,000 patients died of this disease. Currently, tuberculosis and AIDS are responsible for the highest number of annual infectious disease-related deaths [1]. Tuberculosis is predicted to remain one of the most severe diseases in the coming 10 years [2].
Currently, the main method of laboratory diagnosis for tuberculosis is the use of a microbiological test, such as a sputum smear, sputum solid culture, or liquid culture system (e.g., MGIT 960). The accuracy of the test results can be influenced by various factors, including operation proficiency, specimen status, and reagent quality [3]. The results of these tests are used to determine the correct medication and to evaluate the disease transmission and prognosis. Therefore, quality control of tuberculosis diagnosis tests is important. To improve the test quality across laboratories, standard policies have been released; these guidelines propose a series of quality control protocols for the layout, equipment, culture medium, and reagents used in M. tuberculosis laboratories [3,4]. However, there is no universally accepted protocol for the quality control of tuberculosis laboratory test results.
Cross contamination among different specimens is the most frequent problem when conducting a sputum culture test. Cross contamination has been proven to cause false positives, which lead to clinical misdiagnosis and the prescription of incorrect medication [5]; however, few reports in China have addressed this issue. Recently, tuberculosis surveillance uncovered 22 M. tuberculosis strains from a single hospital in Shanghai that all had high genotype homology. To confirm the relationship among these strains and to differentiate which results may have been due to cross contamination, we conducted a molecular epidemiologic investigation.
2. Materials and methods
2.1. Specimen source
The sample collection was based on the Diagnosis for Pulmonary Tuberculosis (WS288–2008), which was released by the Ministry of Health of the People’s Republic of China [6]. A total of 22 strains with high genotype homology from one hospital were collected during the surveillance of drug resistance of tuberculosis.
2.2. Drug sensitivity test
The 1% standard proportioning method recommended by WHO was applied when performing the drug-susceptibility test on solid culture medium. The drug concentrations used for each test were as follows: isoniazid 0.2mg/L, streptomycin 4 mg/L, rifampin 40.0 mg/L, and ethambutol 2 mg/L [7].
2.3. Epidemiological survey
All of the patients in this study signed an informed consent form. Following this, our trained investigators interviewed the patients to collect their epidemiological information, including name, age, address, and disease history. After enquiring about the close contacts of patients, the investigators recorded the locations of activity of the patients. In this study, the people who had direct contact with the patients or those occupying a particular public space together with the patients were considered to have a probable epidemiologic relationship(and vice versa).
2.4. Variable number of tandem repeat (VNTR)genotyping
In this study, we used the 9 + 3 loci VNTR method for the genotyping of M. tuberculosis strains, as this method has been validated to be suitable for typing the prevalent M. tuberculosisstrains in China [8]. PCR amplification combined with gel electrophoresis was used in the detection for VNTR genotyping.
2.5. Genomic DNA extraction
All of the strains were inactivated in a water bath at 85 °C for 30 min in a Biosafety Level 2 Laboratory (with negative pressure), then suspended in DNA extraction liquid and boiled for 10 min in water. After cooling on ice for 2–3 min, cell lysates were centrifuged for 10min (12,000 rpm). The supernatant was used for genotyping assays. Genomic DNA for whole genome sequencing and analysis was extracted by cetyl trimethylammonium bromide (CTAB) method.
2.6. Whole genome sequencing analysis
We constructed genomic libraries with an insert fragment length ~ 300 bp following the standard procedures of paired-end sequencing (Illumina). Whole genome sequencing was performed on an Illumina HiSeq 2500 platform, and the average sequencing depth was about 100- fold. The low-quality reads were trimmed by Sickle software (https://github.com/ucdavis-bioinformatics/sickle). Bowtie 2 software (v2.3.1) was used to map all of the reads to theM. tuberculosis H37Rv genome (NC_000962.2), and remove multiple overlapping regions. SAMtools (v1.4) was used to exclude the nucleotides with sequencing quality and mapping quality below 30. Then, VarScan (v.2.3.9) was used to detect fixed mutations with a depth deeper than 10% that of the average sequencing depth and a frequency higher than 75%. According to the PPE/PE-PGRS gene list provided by NCBI, the fixed mutations located in these genes were filtered before the following analysis.
3. Results
The VNTR (9 + 3) genotyping results of 22 tuberculosis strains isolated from a single hospital formed a total of seven clusters (Table 1). This study collected basic patient information from the patients whose samples were the sources of these strains, and assessed the epidemiology as well as the possible cross contamination within the seven clusters.
Table 1.
9 basic loci and 3 hypervariable loci.
| Loci | PCR primer sequence(5’−3’) | Flanking region Size + repeat Unit size(bp) | Size of amplicons (bp) and repeat unit size(bp) × no.in H37Rv |
|---|---|---|---|
| QUB-11b | CGTAAGGGGGATGCGGGAAATAGG CGAAGTGAATGGTGGCAT |
67 + 69 | 412 69×5+10 |
| QUB-18 | ATCGTCAGCTGCGGAATAGT AATACCGGGGATATCGGTTC |
182 + 78 | 621 78×5+49 |
| MIRU26 | GCGGATAGGTCTACCGTCGAAATC TCCGGGTCATACAGCATGATCA | 243 + 48 | 387 48 × 3 |
| QUB-26 | AACGCTCAGCTGTCGGAT GGCCAGGTCCTTCCCGAT |
129 + 111 | 708 111 × 5 + 24 |
| Mtub21 | AGATCCCAGTTGTCGTCGTC CAACATCGCCTGGTTCTGTA |
92 + 57 | 206 57 × 2 |
| MIRU31 | CGTCGAAGAGAGCCTCATCAATCAT AACCTGCTGACCGATGGCAATATC |
108 + 52 | 264 52 × 3 |
| Mtub04 | GTCCAGGTTGCAAGAGATGG GGCATCCTCAACAACGGTAG |
137 + 51 | 269 51×2+30 |
| VNTR2372 | ACCTCCGTTCCGATAATC CAGCTTTCAGCCTCCACA |
172 + 57 | 298 57×2+12 |
| Miru40 | GGGTTGCTGGATGACAACGTGT GGGTGATCTCGGCGAAATCAGATA |
354 + 54 | 408 54 × 1 |
| VNTR 3820 | TGCGCGGTGAATGAGACG ACCTTCATCCTTGGCGAC |
247 + 57 | 444 57×3+26 |
| VNTR3232 | CCCCAGCCTTACGACTGA GTCGGGCTTGGTGAAGG |
191 + 56 | 407 56×3+48 |
| VNTR4120 | GTTCACCGGAGCCAACC GAGGTGGTTTCGTGGTCG |
310 + 57 | 447 57×2+23 |
3.1. Basic information and epidemiologic survey results
The basic information survey revealed that, except for specimens No. 6, 10, 15, 18, 19, and 22, all of the specimens were tested by sputum smear or sputum culture on the same day as one another (Table 3).
Table 3.
Test information of 22 strains.
| strain no. | group | cluster. | Smear time | smear results | Culture time | Culture results | SNPs |
|---|---|---|---|---|---|---|---|
| 1 | 1 | 2 | 2013.12.19 | positive | 2013.12.19 | positive | 0 |
| 2 | 2013.12.19 | positive | 2013.12.19 | positive | 0 | ||
| 3 | 2013.12.19 | positive | 2013.12.19 | positive | 0 | ||
| 4 | 2 | 40 | 2013.7.9 | positive | 2013.7.9 | positive | 0 |
| 5 | 2013.7.9 | negative | 2013.7.9 | positive | 0 | ||
| 6 | 2010.9.14 | positive | 2010.9.15 | positive | 71 | ||
| 7 | 3 | 120 | 2013.5.22 | positive | 2013.5.22 | positive | 0 |
| 8 | 2013.5.22 | negative | 2013.5.22 | positive | 0 | ||
| 9 | 2013.5.22 | negative | 2013.5.22 | positive | 0 | ||
| 10 | 4 | 193 | 2014.3.17 | negative | 2014.3.17 | positive | 2 |
| 11 | 2014.6.12 | positive | 2014.6.12 | positive | 0 | ||
| 12 | 2014.6.12 | negative | 2014.6.12 | positive | 0 | ||
| 13 | 5 | 202 | 2011.6.9 | positive | 2011.6.8 | positive | 0 |
| 14 | 2011.6.9 | negative | 2011.6.8 | positive | 0 | ||
| 15 | 2013.10.24 | positive | 2013.10.24 | positive | 4 | ||
| 16 | 6 | 205 | 2014.9.2 | negative | 2014.9.2 | positive | 3 |
| 17 | 2014.9.2 | positive | 2014.9.2 | positive | 4 | ||
| 18 | 2014.9.4 | negative | 2014.9.4 | positive | 3 | ||
| 19 | 7 | 230 | 2010.8.26 | positive | 2010.8.26 | positive | none |
| 20 | 2010.8.16 | positive | 2010.8.13 | positive | 0 | ||
| 21 | 2010.8.16 | negative | 2010.8.13 | positive | 0 | ||
| 22 | 2010.7.26 | negative | 2010.7.26 | positive | 0 |
To investigate the transmission relationship among cases, we performed an epidemiologic investigation on all 22 tuberculosis patients. The collected information included the name, age, address, and disease history of the patient. We also enquired about their close contacts and a range of daily activities. The results revealed that there were no confirmed or possible epidemiologic relationships among the 22 patients (Table 2).
Table 2.
VNTRgenotyping results of the 22 M. tuberculosisstiains.
| strain no. | cluster.100%match | QUB11b | QUB18 | Mtub21 | Miru26 | QUB26 | Mtub04 | Miru31 | Miru40 | VNTR2372 | VNTR3820 | VNTR4120 | VNTR3232 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 3 | 3 | 3 | 7 | 1 | 3 | 3 | 2 | 2 | 5 | 10 |
| 2 | 2 | 3 | 3 | 3 | 3 | 7 | 1 | 3 | 3 | 2 | 2 | 5 | 10 |
| 3 | 2 | 3 | 3 | 3 | 3 | 7 | 1 | 3 | 3 | 2 | 2 | 5 | 10 |
| 4 | 40 | 6 | 10 | 4 | 7 | 6 | 4 | 5 | 3 | 3 | 16 | 20 | 17 |
| 5 | 40 | 6 | 10 | 4 | 7 | 6 | 4 | 5 | 3 | 3 | 16 | NA | 17 |
| 6 | 40 | 6 | 10 | 4 | 7 | 6 | 4 | 5 | 3 | 3 | NA | 9,20 | 17 |
| 7 | 120 | 6 | 8 | 5 | 7 | 8 | 4 | 5 | 3 | 3 | 12 | 7 | 15 |
| 8 | 120 | 6 | 8 | 5 | 7 | 8 | 4 | 5 | 3 | 3 | 12 | 7 | 15 |
| 9 | 120 | 6 | 8 | 5 | 7 | 8 | 4 | 5 | 3 | 3 | 12 | 7 | 15 |
| 10 | 193 | 5 | 7 | 5 | 7 | 8 | 3 | 4 | 3 | 3 | 14 | 10 | 13 |
| 11 | 193 | 5 | 7 | 5 | 7 | 8 | 3 | 4 | 3 | 3 | 14 | 10 | 13 |
| 12 | 193 | 5 | 7 | 5 | 7 | 8 | 3 | 4 | 3 | 3 | 14 | 10 | 13 |
| 13 | 202 | 5 | 8 | 5 | 2 | 8 | 6 | 5 | 3 | 3 | 14 | 11 | 13 |
| 14 | 202 | 5 | 8 | 5 | 2 | 8 | 6 | 5 | 3 | 3 | 14 | 11 | 13 |
| 15 | 202 | 5 | 8 | 5 | 2 | 8 | 6 | 5 | 3 | 3 | 14 | 11 | 13 |
| 16 | 205 | 6 | 7 | 5 | 7 | 8 | 4 | 5 | 3 | 3 | 14 | 6 | 15 |
| 17 | 205 | 6 | 7 | 5 | 7 | 8 | 4 | 5 | 3 | 3 | 14 | 6 | 15 |
| 18 | 205 | 6 | 7 | 5 | 7 | 8 | 4 | 5 | 3 | 3 | 14 | 6 | 15 |
| 19 | 230 | 4 | 10 | 4 | 7 | 8 | 3 | 5 | 3 | 3 | 9 | 8 | 8 |
| 20 | 230 | 4 | 10 | 4 | 7 | 8 | 3 | 5 | 3 | 3 | 9 | 8 | 8 |
| 21 | 230 | 4 | 10 | 4 | 7 | 8 | 3 | 5 | 3 | 3 | 9 | 8 | 8 |
| 22 | 230 | 4 | 10 | 4 | 7 | 8 | 3 | 5 | 3 | 3 | 9 | 8 | 8 |
3.2. Whole genome sequencing analysis of the M. tuberculosis strains
Whole genome sequencing data were obtained for 21 of the 22 strains; we were unable to obtain the data for strain No. 19 from cluster No. 230. Pairwise comparisons between the genomes in each cluster indicated that there were 15 strains (71.4%) that had no single nucleotide polymorphisms (SNPs) with at least one other strain within their cluster (Fig. 1). The remaining six strains had 2–71 SNPs compared with other strains in the same cluster, and five of them had only 2–4 SNPs. The differences were not uniformly distributed among the clusters. For example, cluster No. 2 in the first group and cluster No. 120 in the third group had no SNPs with one another, which indicates likely cross contamination. Cluster No. 205 in the sixth group had at least 3–4 SNPs from other strains, which could exclude the possibility of cross contamination in this sample. Other clusters showed varying results.
Fig. 1.
The distribution of genomic SNPs and the time period of the pairwise comparison within each cluster. The size of circle represents the count of the pairs.
An analysis of the specimen collection time indicated that, among the 16 strains collected on the same day, 14 of them had no SNPs with one another (87.5%). Among the strains within the same cluster whoseSNP distance was 0, 93.3% (14/15) of them had the same collection time, suggesting that this result was most likely caused by cross contamination (Fig. 1. Genetic distance of isolate pairs by duration time).
4. Discussion
A measures index based on national criteria is used to guarantee the work quality and improve the test capability in tuberculosis laboratories in each country. For example, some Mycobacterium laboratories in the USA follow the Capability Demonstration Program provided by the College of American Pathologists or participate in one of two self-evaluation projects from the Centers for Disease Control and Prevention, which include AFB smear, nucleic acid amplification, growth detection, identification, and drug sensitivity tests [9,10]. In China, the National Tuberculosis Reference Laboratory of the Chinese Center for Disease Control and Prevention regularly conducts external quality assessments (EQAs) of sputum smears, drug sensitivity test proficiency, and molecular biology assays [11,12]. These series of quality control measures effectively improve the test quality of the laboratories [13,14]. How-ever, there are no clear standards for the definition or detection of cross contamination.
To date, there have been only a few reports of cross contamination occurrence in tuberculosis laboratories. Previous studies indicate that the incidence of cross contamination in the test cases was about 1%–10% [15–18]. Generally, if different clinical strains isolated at the same time from the same laboratory testing location are detected as having the same genotype, cross contamination is suspected, although additional evidence is needed to prove its occurrence. For example, in 1995, strains from three tuberculosis patients in Alabama, USA, from non-overlapping locations were identified by a single laboratory as having the same genotype, suggesting cross contamination may have occurred in that laboratory [16]. A similar event was reported for a tuberculosis laboratory in the northwest of Iran [17]. For laboratories in countries and regions with a low tuberculosis incidence, multiple positive culture cases appearing within a short period of time may also be caused by cross contamination; this possibility could be confirmed using a genotyping method [18].
Here, we used whole genome sequencing to analyze the cross contamination occurrence in tuberculosis strains cultured from specimens collected at the same medical institution. The results reveal that 66.7% (14/21) of the strains from one hospital with reported genotype homology had cross contamination; specifically, some of the specimens that shared the same processing time had no difference in genotype, which strongly indicates that cross contamination has occurred. Thus, if the genotypes of strains isolated from samples that were processed together in the same batch appear to be the same, the possibility of cross contamination occurrence during operation should be considered, and corresponding intervention measures should be taken.
In the second, fourth, and fifth groups, one specimen from each group had a SNP from two other specimens in the same group. Notably, these samples were not processed at the same time as the others in the group, which suggests that using whole genome sequencing to detect cross contamination in laboratories is reliable. Data from the sixth group showed that although three specimens had the same or close processing times, they had 3–4 SNPs based on whole genome sequencing. These results suggested that the similarities among these strains were unlikely caused by cross contamination but instead might have been caused by a common infection source, and the small SNP differences among these isolates may have been generated during transmission or be due to host diversity within the strain population. In the seventh group, specimen No. 19 did not undergo whole genome sequencing, but the other three strains in this group were found to lack any SNPs, even though the operation time for specimen No. 22 was different from that of the other two specimens. However, specimens No. 20–22 were assessed in a drug sensitivity test in the same batch on the same day, so the cross contamination in these samples might have occurred during that period.
The laboratory processes performed on each sputum specimen include vortexing, digesting by a NaCl—NaOH solution, neutralizing by phosphate-buffered saline (PBS), centrifuging, and discarding the supernatant. If these procedures are not performed properly, cross contamination can easily occur. For example, a sputum smear and culture are commonly performed in the same biosafety cabinet, and no sterilization is typically conducted between these two operations. Additionally, when a sputum specimen is processed by the digesting solution, the dropper dispensing this reagent may come in contact with multiple samples. Furthermore, the aerosol generated when adding liquid using a dropper, centrifuging, or concentrating could easily cause cross contamination. Thus, having standard operating procedures that prevent cross contamination is of great importance in clinical laboratories. For instance, to limit the chances of cross contamination, specimens that are negative based on the results of a sputum smear are processed first, before those with positive results.
Traditional methods are unable to identify cross contamination among tuberculosis specimens, but the application of molecular techniques has made the detection of cross contamination possible. Here, the VNTR genotyping technique was initially used to screen the clustered tuberculosis strains, but no epidemiologic relationship was identified among them. Whole genome sequencing was subsequently performed on 21 of the strains. Among the 16 strains that shared the same operation time, 14 (87.5%) of them had no SNPs, which suggested the occurrence of cross contamination. The molecular epidemiologic study of tuberculosis often excludes the possibility of transmission among genotypically-clustered strains that were processed at the same time, identifying it as likely cross contamination [9]. Currently, the most common genotyping method for cross contamination detection in tuberculosis samples is VNTR. However, because only part of the genetic information in the genome is utilized by the VNTR method, it cannot be used to judge the overall genetic relationship between different strains. Whole genome sequencing greatly improves the reliability of differentiation among strains. Therefore, for emergency public health events, such as an epidemic outbreak that requires etiologic testing within a short period of time, whole genome sequencing could be applied to the identification of highly homologous strains, in addition to being used for surveillance of potential cross contamination.
This is a retrospective study, so it is unlikely to trace back all possible sources of cross contamination during sample processing. The use of a prospective experimental design and appropriate controls is needed to clarify this issue. Furthermore, there is still limited report to evaluate whether the slight genome differences among strains from specimens processed at the same time are due to cross contamination or pathogen heterogeneity; this distinction needs further exploration.
Acknowledgement
CGY receives funding from the Robert E. Leet and Clara Guthrie Patterson Trust Fellowship Award and from the MIDAS Center for Communicable Disease Dynamics (grant U54 GM088558).
References
- [1].WHO. Global tuberculosis report. World Health Organization; 2015. [Google Scholar]
- [2].Murray CJ, Salomon JA. Modeling the impact of global tuberculosis control strategies. Proc Natl Acad S c i U S A 1998;95(23):13881–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].STOP TB. Mycobacteriology laboratory manual. A publication of the Global Laboratory Initiative a Working Group of the Stop TB Partnership; 2014.
- [4].Tuberculosis laboratory biosafety manual. World Health Organization. [PubMed] [Google Scholar]
- [5].Fitzpatrick L, Braden C, Cronin W, et al. Investigation of Laboratory cross-contamination of Mycobacterium tuberculosis cultures. Clin Infect Dis 2004;38(6):e52–4. [DOI] [PubMed] [Google Scholar]
- [6].People’s Medical Publishing House. Diagnostic criteria for tuberculosis. Ministry of health of the People’s Republic of China; 2008. [Google Scholar]
- [7].Chinese Anti-tuberculosis Association The laboratory science procedure of diagnostic bacteriology in tuberculosis. 2006. p. 46–51. [Google Scholar]
- [8].Yang C, Shen X, Peng Y, et al. Transmission of Mycobacterium tuberculosis in China: a population-based molecular epidemiologic study. Clin Infect Dis 2013. 10.1093/cid/civ255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Drobniewski FA, Hoffner S, Rusch-Gerdes S, et al. Recommended standards for modern tuberculosis laboratory services in Europe. Eur Respir J 2006;28(5):903–9. [DOI] [PubMed] [Google Scholar]
- [10].Drobniewski FA, Caws M, Gibson A, et al. Modern laboratory diagnosis of tuberculosis. Lancet Infect Dis 2003;3(3):141–7. [DOI] [PubMed] [Google Scholar]
- [11].Guanglu Jiang, Mei Shang, Yuanyuan Song, et al. Analysis of first round external quality assessment for MTB drug sensitivity test within disease control laboratories in China. Chin. J. Antituberculosis 2010;32:806–10. [Google Scholar]
- [12].Van Deun A, Wright A, Zignol M, et al. Drug susceptibility testing proficiency in the network of supranational tuberculosis reference laboratories. Int J Tuberc Lung Dis 2011;15:116–24. [PubMed] [Google Scholar]
- [13].Sumin Wang, Yuhong Liu, Guanglu Jiang, et al. Evaluation of quality control system for direct smear microscopy in National Tuberculosis Program in China. Chin. J. Antituberculosis 2000;22:209–10. [Google Scholar]
- [14].Peng Y, Yang C, Li X, Luo T, Li F, Gao Q. Multiple samples improve the sensitivity for detection of mixed Mycobacterium infections. Tuberculosis 2013;93(5):548–50. [DOI] [PubMed] [Google Scholar]
- [15].Dunlap NE, Harris RH, Benjamin Jr. WH, Harden JW, Hafner D. Laboratory contamination of Mycobacterium tuberculosis cultures. Am J Respir Crit Care Med 1995;152(5 Pt 1):1702–4. [DOI] [PubMed] [Google Scholar]
- [16].Asgharzadeh M, Kafil HS, Ghorghanlu S, Pourostadi M. Laboratory cross-contamination of Mycobacterium tuberculosis in northwest of Iran. Egypt J Chest Dis Tuberc 2015;64:665–9. [Google Scholar]
- [17].Fitzpatrick L, Braden C, Cronin W, et al. Investigation of laboratory cross-contamination of mycobacterium tuberculosis cultures. Clin Infect Dis 2004;38(6):E52–4. [DOI] [PubMed] [Google Scholar]
- [18].de CRM, Soini H, Roscanni GC, Jaques M, Villares MC, Musser JM. Extensive cross- contamination of specimens with Mycobacterium tuberculosis in a reference laboratory. J Clin Microbiol Apr. 1999:916–9. [DOI] [PMC free article] [PubMed] [Google Scholar]