Utilizing whole genome sequencing to delineate relapse and reinfection tuberculosis on the Canadian prairies

Abstract

RATIONALE

Recurrent tuberculosis (TB) accounts for 5% of the Canadian TB burden. Recurrence can occur from relapse or reinfection. Identifying reinfection has implications for developing control policies.

OBJECTIVE

The objectives of this study were to quantify reinfection in recurrent TB using whole genome phylogenetic and single nucleotide polymorphism (SNP) and compare with epidemiological and clinical parameters from the Canadian prairies.

METHODS

DNA sequences of Mycobacterium tuberculosis isolates from recurrent TB cases along with epidemiological and clinical parameters were collected from Alberta and Saskatchewan. Inclusion criteria were two or more culture positive notifications age ≥17 years. SNP and phylogenetic tree scale differences (TIP) were used to determine the relapse and reinfection categories.

RESULTS

Of 7,627 notifications of TB disease, 533 were recurrent. 93 pairs (180 cases) were culture positive from which 26 (50 cases) were available for sequencing. 19 cases with SNP and TIP values ≤25 and ≤.001 were classified relapse. Seven SNP and TIP values >160 and >.001 were classified reinfection. Five of seven reinfections were Indigenous cases from high TB incidence areas. The non-sequenced and sequenced pairs differed only in age.

CONCLUSIONS

Recurrent culture positive TB is uncommon on the Canadian prairies and is more likely to be relapse. Reinfection is more likely in Indigenous persons living in high TB incidence communities. A limitation was the risk of selection bias since only 28% of eligible cases were sequenced. Since we did not know the non-sequenced reinfection risk, we could only conclude that the non-sequenced and sequenced categories were similar.

RÉSUMÉ

JUSTIFICATION:

La tuberculose récurrente représente 5 % du fardeau de la tuberculose au Canada. La récurrence peut résulter d'une rechute ou d’une réinfection. La détection de la réinfection a des implications pour l’élaboration de politiques de lutte contre la tuberculose.

OBJECTIF:

Quantifier les cas de réinfection dans la tuberculose récurrente à l’aide de la phylogénétique du génome entier et du polymorphisme d’un seul nucléotide (PSN), et la comparer aux paramètres épidémiologiques et cliniques des Prairies canadiennes.

MÉTHODES:

Des séquences d’ADN d’isolats de Mycobacterium tuberculosis provenant de cas de tuberculose récurrente, ainsi que des paramètres épidémiologiques et cliniques, ont été recueillis en Alberta et en Saskatchewan. Les critères d’inclusion étaient la notification de deux cultures positives ou plus, et un âge ≥17 ans. Les polymorphismes d’un seul nucléotide et les distances phylogénétiques ont été utilisés pour déterminer les catégories de rechute et de réinfection.

RÉSULTATS:

Sur 7 627 notifications de tuberculose, 533 cas étaient récurrents. 93 paires (180 cas) étaient positives à la culture, dont 26 (50 cas) étaient disponibles pour le séquençage. 19 cas avec des valeurs de polymorphisme d’un seul nucléotide et de ≤25 et ≤.001 ont été classés en rechute. Sept cas avec des valeurs de polymorphisme d’un seul nucléotide et de distance phylogénétique >160 et > 0,001 ont été classés comme une réinfection. Cinq des sept réinfections étaient des cas autochtones provenant de zones à forte incidence de tuberculose. Les paires non séquencées et séquencées ne différaient que par l'âge.

CONCLUSIONS:

La tuberculose récurrente à culture positive est peu fréquente dans les Prairies canadiennes et est plus susceptible d'être une rechute. La réinfection est plus probable chez les autochtones vivant dans des communautés à forte incidence de tuberculose. Le risque de biais de sélection constitue une limite puisque seulement 28 % des cas admissibles ont été séquencés. Comme nous ne connaissions pas le risque de réinfection en l'absence de séquençage, nous ne pouvions que conclure que les catégories non séquencées et séquencées étaient similaires.

Keywords:

• M. tuberculosis
• genomics
• whole genome sequencing
• epidemiology
• relapse
• reinfection

Introduction

Tuberculosis disease, caused by M. tuberculosis (TB), resulted in 6.4 million cases and 1.6 million deaths globally in 2021.¹ Canada reported 1,796 new and recurrent² cases (definitions in Appendix A1 of the Online Supplementary Material) in 2017,³ the last year for which statistics were available; 1,290 (71.8%) cases originated outside Canada and 95 (5.3%) were recurrent. Furthermore, 148 of 1,459 TB isolates were resistant to one or more first line drugs.⁴ Recurrent disease contributes to ongoing transmission of infection and is more likely to be drug resistant.⁵ Drug resistance further contributes to transmission of infection, increased duration and cost of treatment. Both are significant factors that affect the successful control of TB disease.^4–7

Recurrent disease can be classified as relapse or reinfection (Appendix A1 of the Online Supplementary Material).^2,8–10 However, differentiating relapse from reinfection is a challenge without genotyping or genomic analysis.^8,9 Factors associated with relapse include incomplete treatment, drug resistance, and lineage.^7,8,11 Factors associated with reinfection include resident in area of high incidence TB, (Appendix A1 of the Online Supplementary Material) living in crowded conditions, and contact with infectious TB (Appendix A1 of the Online Supplementary Material).^8,11 An added complexity for identifying reinfection is mixed infection which may be uncharacterized due to limited sampling⁹ and occurs in similar frequency and similar conditions as reinfection^9,11,12 (Appendix A1 of the Online Supplementary Material) The proportion of mixed infections was reported to account for up to 20% of pulmonary TB (PTB) and up to 51% of extrapulmonary TB (EPTB).⁹ Differentiating mixed infection from reinfection is important. It is also important to differentiate relapse from reinfection in developing effective control strategies, especially the resource intensive search for the source of the reinfection.^7,8

Whole-genome sequencing (WGS) has been used to investigate drug resistant mutations, transmission, genomic evolution, social-networks, and distinguishing reinfection from relapse.^11–18 WGS has a higher resolution than genotyping methods for M. tuberculosis, such as restriction fragment length polymorphism (RFLP), mycobacterial interspersed repetitive unit–variable number tandem repeat (MIRU-VNTR) and spacer oligonucleotide genotyping (spoligotyping).^11,12,15 The proportion of reinfection in recurrent cases of TB varied from 5% in HIV negative cases in India,¹⁵ to 9% in Beijing,¹⁹ to 25% in Hunan province China,²⁰ to 27% in Malawi⁸ and 75% in HIV positive cases in India.¹⁵ Reinfection is important since it implies new transmission events, particularly when the source case is drug resistant.^5,6

The Canadian Tuberculosis Standards (CTBS)²¹ state that repeated exposure is rare in low incidence areas such that active TB usually results from a first infection. However, it has been shown that reinfection occurs in high-incidence, high-transmission areas. Our objective was to identify the contribution of relapse and reinfection TB disease in high and low incidence settings in the Canadian prairies and compare WGS analysis with epidemiological and clinical data.

Materials and methods

Study participants and epidemiology

This was a retrospective study of recurrent TB. Inclusion criteria for WGS were two or more notifications of culture positive TB, age 17 years and older, and diagnosed and treated in Alberta and Saskatchewan. Not all isolates were sequenced. Isolates that were not found or could not be regrown were not sequenced. The majority had one recurrence while some had two recurrences: the first notification was episode 1, the second was episode 2 and the third was episode 3. Epidemiological and clinical data – diagnosis date, age, sex, population group, country of origin, domicile area, site of TB, contacts that included travel to endemic countries, community TB incidence, treatment completion, sensitivity tests and episode interval were obtained from the TB program for each province.

Reinfection risk score

Relapse factors comprised incomplete treatment and drug resistance, both for the first episode. Reinfection factors comprised identified contact(s) and living in a high-incidence community at the time of notification (>30/100,000), both for the second episode. A reinfection score was developed for each case based on the risk factors for reinfection: that is, +1 for a high-incidence community, +1 for a contact, −1 for incomplete treatment, −1 for drug resistance and 0 for HIV coinfection, diabetes, silicosis, smoking, air pollution and unavailable data.¹¹ Zero was scored since these comorbidities were associated with both relapse and reinfection.¹¹

Mycobacterial culture and DNA extraction

The details are described in Appendix A2 of the Online Supplementary Material.

Whole genome sequencing

Genome Quebec prepared the libraries using Illumina Truseq Prep Kit and performed on an Ultra II Illumina NovaSeq 6000 S4 sequencer (300 cycles) that generated 2 × 150 bp paired end reads. The defined coverage depth was ≥100x. The required minimum total reads were 1.5 M.

Bioinformatics data analysis

Coverage was calculated with fastq-info v2.0 with H37Rv (accession no: NC 000962.3) as a reference genome²² based on the Lander Waterman equation.²³ For each sample, raw read-pairs were subjected to quality control using Fastqc (v0.11.9).²⁴ The Illumina universal adapters were present in one sample and removed using Trimmomatic (with ILLUMINACLIP option).²⁵ WGS raw reads and trimmed reads were aligned to the reference genome M. tuberculosis H37Rv;²⁶ using the read mapping function (bowtie v 2-2.4.2)²⁷ implemented in Gen2Epi.²⁸ De novo assemblies were generated using SPAdes (v 3.15.1)²⁹ implemented in Gen2Epi²⁸ (with parameters –cov-cutoff auto –careful –kmer 21,33,55,71). The quality of the assembled contigs was checked with QUAST (v 5.0.2)³⁰ using M. tuberculosis H37Rv as a reference26 (Genbank accession number [annotation number]: NC_000962.3) and its respective annotation. The mean genome fraction covered by contigs was 98.3%. Annotation of the de novo assembled contigs was performed using Prokka³¹ (with parameter “-compliant” using a minimum contig length of ≥200 bp for gene prediction and product name assignment). Lineage classification and drug resistance prediction was performed using Mykrobe version 0.10.0.³²

Single nucleotide polymorphism and phylogenetic analysis

Variant calling and pairwise comparison of SNPs between samples were determined by uploading the cleaned WGS reads to the MTBseq pipeline version 1.0.333 using Mycobacterium tuberculosis strains H37Rv (Genbank accession number: NC_000962.3) as reference.²⁶ Variants were filtered automatically based on repetitive regions (including the polymorphic PPE/PE regions³⁴ and regions encoding for phage proteins), genes associated with resistance and SNPs within a window of 12 bp within the same subsets.³³ For variance prediction, MTBseq pipeline used the default parameters: minimum coverage 4 forward reads, 4 reverse reads, and a minimum Phred score ≥20. At the default setting, MTBseq reliably detected variants if they were present in ≥75% allele frequency in the bacterial population.^33,35 The minimum coverage to call a variant was 8x. The consensus fastq file for the filtered variants was used for phylogenetic analysis. The phylogenetic tree was generated using RAxML version 8 using the GTRGAMMA model with 200 bootstrap replications.³⁶ The output tree was visualized using iTOL.³⁷

Threshold determination to identify relapse and reinfection

A numeric table was constructed summing the branch length of each case and episode to calculate the value of the final horizontal position on the phylogenetic tree scale, (ie, tree tip [TIP]), Figure 1. The TIP difference between episodes was determined as a positive value, Online Supplementary Table S1. This value was compared to the SNP results for each pair. Two methods for evolutionary distance were used to delineate similar and unique samples: SNPs based on MTBseq pipeline, and the phylogenetic tree scale based on maximum likelihood RAxML software. They are two methods for identifying relatedness using the same data with different algorithms.^38,39

Figure 1. Phylogenetic tree scale.

Unrooted tree scale that measures genetic distance between 50 genomes comprised of 22 pairs and two triplets. Genomes are listed by CaseNo., episode and lineage. The horizontal axis shows the scale representing the number of single nucleotide polymorphism (SNP)s divided by the sequence length; 0.100 is equivalent to 10%. CaseNo. 1, 9, 11, 14, 20, 24 (all episode 2) and 11 (episode 3) had SNPs >160 whose final tree scale positions were different from Episode 1 and 2. They were all lineage 4. The horizontal final position can also be obtained from Online Supplementary Table S1.

Statistical analysis and ethics

Statistical analysis was performed with SPSS v28.0.1.0. Fisher exact test was conducted using a two-sided alpha level 0.05. Ethics approval was granted by University of Saskatchewan REB Bio-782 and by the University of Alberta REB Pro00088456.

Results

Epidemiology and clinical parameters

Alberta and Saskatchewan reported 7,627 new and recurrent cases from 1989 to 2018. A total of 533 were recurrent, of which 93 pairs (180 cases) were culture positive. The remaining 353 cases were not culture confirmed, Figure 2. Twenty-six pairs (50 cases) were sequenced, comprised of 13 foreign-born, 12 Indigenous, and one non-Indigenous cases. Indigenous and non-Indigenous cases were all lineage 4 compared to foreign born cases that were lineage 1 to 4, Figure 1. Sixty-seven pairs (130 cases) that were not sequenced were comprised of one foreign-born, 65 Indigenous, and one non-Indigenous cases (Online Supplementary Table S2). The non-sequenced pairs were younger, mostly Indigenous who lived in remote, high TB rate communities, with longer recurrence intervals. Online Supplementary Table S3 compares the same parameters for Indigenous non-sequenced with sequenced pairs. The not sequenced cases were younger. The remaining nine parameters of interest were not significantly different.

Figure 2. Distribution of reported tuberculosis (TB) cases 1989–2018.

Shows the distribution for all reported TB cases 1989 to 2018 by step to the number of sequenced cases. A total of 67 culture positive recurrent pairs were not sequenced because they could not be found or could not be reconstituted, 26 pairs were sequenced and 67 pairs were not sequenced.

SNP and phylogenetic analysis

The samples where sequence reads percent aligned to H37Rv of two samples was <55% were removed from further analysis. The remaining 50 samples percentage aligned to H37Rv, which was ≥90% (Online Supplementary Table S4). The coverage depth ranged from 34 to 62 M reads. The depth per sample was ≥100x (Online Supplementary Table S4). Total number of reads was 2,225 M. For the 26 pairs (50 genomes) analyzed, SNPs ranged from 0 to 1,116 (Table 1) and the TIP differences ranged from 0.000 to 0.007 (Table 2). Table 3 shows that SNPs ≤ 25 and TIPs ≤ .001 were concordant when identifying relapse and SNPs > 160 and TIPs > .001 were concordant identifying reinfection. Based on the SNP and phylogenetic tree analysis, 19 recurrent episodes were classified relapses and seven were classified reinfections, a 26.9% prevalence. Relapse cases comprised lineages 1, 2, 3 and 4 while reinfection episodes comprised only lineage 4 (Figure 1).

Table 1. Paired SNP matrix for 26 recurrent TB cases.

	Case	1	2	3	4	4	5	6	7	8	9	10	11	11	12	13	14	15	16	17	18	19	20	21	22	23	24
Case	Epis	1	1	1	1	2	1	1	1	1	1	1	1	2	1	1	1	1	1	1	1	1	1	1	1	1	1
1	2	1116
2	2		4
3	2			2
4	2				1
4	3				8	9
5	2						4
6	2							0
7	2								0
8	2									2
9	2										612
10	2											1
11	2												603
11	3												44	607
12	2														0
13	2															5
14	2																161
15	2																	1
16	2																		24
17	2																			0
18	2																				25
19	2																					6
20	2																						570
21	2																							1
22	2																								2
23	2																									2
24	2																										695

Abbreviations: SNP, single nucleotide polymorphism. CaseNo. episodes 2 and 3 are listed on the vertical axis and episodes 1 and 2 are listed on the horizontal axis. SNP is obtained at the intersection of CaseNo. episode 1 on the horizontal axis and episode 2 on the vertical axis. The bold CaseNo. are for SNPs >160. Seven cases were classified reinfection.

Table 2. Paired phylogenetic TIP difference matrix for 26 recurrent TB cases.

	Case	1	2	3	4	4	5	6	7	8	9	10	11	11	12	13	14	15	16	17	18	19	20	21	22	23	24
Case	Epis	1	1	1	1	2	1	1	1	1	1	1	1	2	1	1	1	1	1	1	1	1	1	1	1	1	1
1	2	.007
2	2		.000
3	2			.000
4	2				.000
4	3				.001	.001
5	2						.001
6	2							.000
7	2								.000
8	2									.000
9	2										.004
10	2											.000
11	2												.003
11	3												.001	.002
12	2														.000
13	2															.001
14	2																.004
15	2																	.000
16	2																		.000
17	2																			.000
18	2																				.000
19	2																					.001
20	2																						.006
21	2																							.000
22	2																								.000
23	2																									.001
24	2																										.003

Abbreviations: TIP, phylogenetic tree scale differences; TB, tuberculosis. CaseNo. episodes 2 and 3 are listed on the vertical axis and episodes 1 and 2 are listed on the horizontal axis. TIP difference is obtained at the intersection of CaseNo. episode 1 on the horizontal axis and episode 2 on the vertical axis. The bold CaseNo. are for TIPs > .001. Seven cases were classified reinfection.

Table 3. Determining relapse and reinfection combining SNPs and TIPs.

Clinical reinfection risk score

The reinfection risk score, SNP and TIP results are listed in Table 4. The best fit threshold for reinfection was defined as a risk score >1 and for relapse ≤1. A sensitivity analysis comparing the reinfection score to the SNP standard showed that the reinfection score specificity was 0.842 and the negative predictive value was 0.882. Six of seven reinfections occurred in Indigenous persons, of which five were resident in high TB incidence communities. Reinfection cases were associated with Indigenous population, lineage 4, contact with TB, living in high TB incidence community and reinfection risk score >1 (Table 5). Multivariate logistic regression analysis was not significant.

Table 4. Reinfection risk score.

Table 5. Epidemiological and clinical comparison of relapse and reinfection cases.

Parameter	Relapse	Reinfection	p value	Test
Number	19	7
Age Mean (SD)	53.2 (19.9)	38.6 (14.8)	.107	Mann-Whitney
Sex
Female	9	4	1.000	Fisher’s Exact
Male	10	3
Pop Group
Foreign-born	12	1	<.05	Fisher’s Exact
Indigenous	6	6
Non-Indigenous	1	0
Diagnosis Year
1989–2003	5	1	1.000	Fisher’s Exact
2004–2018	14	6
Comm TB Rate
High	3	5	<.05	Fisher’s Exact
Low	16	2
Lineage
1–3	8	0	<.05	Fisher’s Exact
4	11	7
Contacts
Yes	2	4	<.05	Fisher’s Exact
No	16	3
Unknown	1	0
Episode Interval
Mean (SD)	5.1 (5.1)	9.12 (8.2)	.094	Mann-Whitney
Site
EPTB	3	0	.540	Fisher’s Exact
PTB	16	7
Reinfect Score
≤1	15	2	<.05	Fisher’s Exact
>1	4	5

Abbreviations: SD, standard deviation; TB, tuberculosis; EPTB, extrapulmonary TB; PTB, pulmonary TB; SNP, single nucleotide polymorphism; TIP, phylogenetic tree scale differences. The table compares epidemiological and clinical parameters by SNP and TIP delineation using the Fisher’s exact and Mann-Whitney tests. SNPs ≤ 25 and TIPs ≤ .001 were considered relapse and SNPs > 160 and TIP > .001 were considered reinfection. Relapse and reinfection numbers were obtained from episode 2 and 3 for 26 sequenced pairs. Reinfection parameters that were significant for reinfection were Indigenous persons, living in high TB incidence community, lineage 4, history of TB contact, and clinical reinfection risk score >1.

Phenotypic resistance and genotypic mutations

Phenotypic drug resistance was reported in six, and drug sensitive in 44 of 50 isolates (Online Supplementary Table S5). Three episode 2 isolates, two rifampicin and one isoniazid acquired drug resistance. Genotypic resistant mutations were observed in four and zero mutations in 46 of 50 genomes. Case 8 showed phenotypic INH resistance and Case 19 phenotypic streptomycin resistance. Both were genetically not predicted to be resistant to these agents.

Discussion

The main finding was the 19 cases classified as relapse and 7 cases classified as reinfection. The reinfection cases were likely to be Indigenous and resident in remote high TB incidence communities. The significant reinfection risk factors for this culture positive recurrent TB cohort were Indigenous ethnicity, lineage 4, contact with TB, living in high TB incidence community and reinfection risk score >1. Although the reinfection burden was low, false positive reinfections have program implications resulting from new infection follow-up policy.

Our data delineated cases of relapse and reinfection. We used two methods, SNP and phylogenetic tree TIP differences, to determine relapse and reinfection. There were discordant results between SNP and TIP thresholds. By removing the discordant thresholds, relapse and reinfection were identified to the exclusion of all other cases. Because there were no SNPs between 25 and 161, a SNP threshold could not be defined in this population since the intermediate values were indeterminate.⁴⁰ More studies that include intermediate SNP values between 25 and 161 need to be done in this population to define the threshold.

In a review of published reports that used SNPs to identify TB genomes, the lower SNP thresholds below which cases were considered relapse, varied from <2 to <50 and an upper threshold from >12 to >1,300, above which cases were considered reinfection.⁴⁰ Both categories were within the range of previous reports.⁴⁰ The epidemiological context in which the SNP thresholds were applied to delineate different genomes, differed from study to study. They were different populations with different objectives, with variable HIV infection prevalence, different definitions of recurrent TB (episodes intervals between 17 wk to greater than one year, ie, they included recurrent intervals consistent with treatment failure).^{8,12,20,41,42} The results were not comparable between studies and showed a wide range of SNPs that could reasonably be applied.⁴⁰ Strict SNP thresholds to infer transmission may not be universally valid.⁴³ SNP thresholds are also thought to be influenced by mutation rates that ranged from 0 to 17 SNPs/genome/year with means of 0.3 to 5.37.^19,41,44,45 One review concluded that SNPs varied by lineage⁴⁶ and could not be reliably calculated for intervals less than 15 years.⁴⁶ These differences were most likely the result of different assumptions and methods.⁴⁶ In another review, the molecular clock data did not show a steady mutation rate/year and was thought to be dependent on patient characteristics (age, country of origin, disease site, drug resistance, treatment outcome), treatment regimen, treatment compliance and strain type or lineage.⁴⁴

SNPs may also arise from mixed TB infection, the prevalence of which was reported to be as high as 20% in PTB cases^9,47 and 51% in EPTB cases.⁹ Mixed infections occurred in the same settings in which reinfections occurred, mostly high-incidence communities.¹² Moreover, they were demonstrated in postmortem TB cases dated as early as eighteenth century Hungary, during the peak TB wave in Europe.⁴⁷ In our study, the clinical assessment for Case 1 was mixed infection at the time of the second episode. Episode 1 showed INH and streptomycin resistance while episode 2 showed fully sensitive organisms, with no history of contact or travel to an endemic area. The left upper lobe was destroyed during the initial episode. TB organisms have a propensity to recur in damaged tissue,¹¹ with a four-fold increased risk following residual lung cavitation,⁴⁸ suggesting that persistence of the organism is enhanced. Mixed infection connotes infection with more than one strain. Recurrence of TB with a strain that was initially present after completed treatment is the definition of relapse.² Genomic assessment of Case 11 also raised the possibility of mixed infection. Episodes 1 and 2 and 2 and 3 differed by 603 and 607 SNPs, respectively. Episode 1 and 3 differed by 44 SNPs and a TIP difference of 0.001, that is, they were similar genomes as defined for this cohort. Case 11 was an Indigenous resident in a high TB incidence community, a setting in which mixed infections have been reported.¹² It was possible that the episode 2 genome was part of the initial infection. Since mixed infection analysis was not included in the protocol, both cases could not be verified as mixed infection.

The reinfection risk score was based on a history of TB contact, resident in a high TB incidence community, treatment completion and drug sensitivity. Compared to the SNP standard, the reinfection risk score ≤1 had a specificity of 0.84. It meant that the reinfection risk score correctly classified recurrent TB as relapse 84% of the time. The negative predictive value was 0.88. It meant that the reinfection risk score <1 correctly identified a relapse case 88% of the time. Due to the small number of cases and possible selection bias, these results may not be representative of all recurrent cases and cannot be recommended as a proxy for sequencing.

An incidental finding was the discordance between phenotypic resistance and genotypic analysis for INH and streptomycin. Discordance with INH and other drugs has been previously reported.^49–52 Other reports noted an association between drug resistance and lineage 2.⁵³,⁵⁴ Our study noted this association as well. Notably, we observed three cases of acquired drug resistance. This suggested that drug adherence played a role in phenotypic drug resistance.⁵⁵

There were limitations in this study. The main limitation was that 67 of 93 culture positive pairs were not sequenced. This represented a risk of selection bias. Since only two non-Indigenous pairs were not sequenced, the risk arose from the 65 Indigenous pairs that were not sequenced (Online Supplementary Table S4). In the ten parameters of interest that are listed in Online Supplementary Table S4, nine were not significantly different from the sequenced pairs. The difference was the non-sequenced cases were younger. Table 5 shows that age was not associated with reinfection. The reinfection risk of the 65 not sequenced pairs was not available.

Another limitation was mixed infection that we did not identify given that it was not part of the protocol. Since mixed infections occur in the same settings as reinfections,¹² it was possible that some cases that were classified as reinfection were mixed infections. Assessment of two cases indicate that this might have been possible. This emphasizes the challenge in attempting to infer disease transmission. Unidentified mixed infection might have overestimated reinfection. A third limitation was the omission of clinical data such as sputum quality or delay in processing⁹ and the size of the mycobacterial population (extent of disease).⁴⁰,⁴⁸

In summary, recurrent culture positive TB is uncommon on the Canadian prairies and more likely to be relapse than reinfection. Reinfection is more likely to occur in Indigenous persons living in remote high TB incidence communities. WGS might have overestimated the reinfection rate since the protocol did not include mixed infection. A limitation was the risk of selection bias since only 28% of eligible cases were sequenced.

Author contributions

R. Singh, J.R. Dillon and M.K. Sharma were responsible for the study design, analysis, interpretation, drafting, and critical review of the manuscript. F. Jamieson, R. Long and M. Richard-Greenblatt were responsible for acquisition data, analysis, interpretation, and critical review of the manuscript. I. Khan and G.J. Tyrrell were responsible for acquisition data, interpretation and critical review of the manuscript. R. McDonald, J. Minion and S. Shokoples were responsible for acquisition data, data analysis and critical review of the manuscript. E. Rea was responsible for data analysis, interpretation and critical review of the manuscript. V. Hoeppner was responsible for concept, study design, acquisition data, data analysis, interpretation, drafting and critical review of the manuscript. W. Wobeser was responsible for concept, study design, data acquisition, data analysis, interpretation and critical review manuscript.

All authors approved the final draft and agreed to accountability for all aspects.

Acknowledgments

The authors thank the following for their contribution to grant application: Donna Goodridge PhD, Dept. Med, University of Saskatchewan; Ida Lemaigre, Indigenous Community Representative; Kathleen McMullin MEd, Dept. Comm Health and Epi, University of Saskatchewan; Michael Patterson MD, Chief Public Health Officer Nunavut; Lisa Puchalski Ritchie MD PhD, Dept. Clin Epidemiol and Health Care Research, University of Toronto; Keith Travers MSc, Dept. Health and Soc Services, Govt. Nunavut. The authors also wish to thank Jennifer Guthrie PhD, Dept. Microbiol and Immunol, University Western Ontario, for critical review of the manuscript and Josh Lawson PhD, Canadian Centre for Health and Safety in Agriculture, University of Saskatchewan for critical review of the data and interpretation. We are pleased to acknowledge the support of this manuscript by the director of VIDO-InterVac as journal series no. 1043.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Canadian Institutes of Health Research under Grant PJT − 153220.