Screening and alignment of the mclx genes from public available genomes

The presence and transcriptional integrity of the mclx genes was analysed in a panel of Mtb complex strains and clinical isolates which genome had been completely sequenced. These organisms were selected from the Genome database of the National Center for Biotechnology Information (NCBI), limiting the search by organism—Mycobacterium tuberculosis complex (taxid: 77643)—and including only genomes with the status "Complete" or "Scaffolds or contigs". This search was performed on the 20^th of March 2014 and yielded 187 organisms. To retrieve the mclx gene sequences from these organisms, each genome or set of scaffolds was used as a reference against which the mclx sequences from the reference strain Mtb H37Rv (Rv0386, Rv1358 and Rv2488c) were mapped, using the assembling tools from Geneious R7.1.4 (Biomatters) [14]. Instances when the putative mclx orthologue spanned more than one scaffold/contig lead to the elimination of the respective organism from the analyses, as to avoid sequencing misreads potentially attributed to scaffolds/contigs junctions. One hundred and fifty strains/clinical isolates were retained for further analyses: two of Mycobacterium africanum, 12 of Mycobacterium bovis, eight of Mycobacterium cannetti, and 128 of Mtb (Table 1). The mclx genes were identified and individually aligned against their reference orthologue from Mtb H37Rv to identify nucleotide substitutions, insertions and deletions, using the ClustalW algorithm [15] available in the Geneious software [14] (S1A Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Pairwise identity and indels in the mclx genes in a panel of 150 Mtb complex strains/clinical isolates¹.

https://doi.org/10.1371/journal.pone.0128983.t001

For most Mtb strains/clinical isolates, the information on the lineage could be retrieved from the literature and/or information on the genome. For the few cases in which this information could not be found, each genome or set of scaffolds was used as a reference against which the regions of difference (RD) from the reference strain Mtb H37Rv, described by Gagneux et al. [16] in the definition of the six phylogeographical lineages, were mapped, and the occurrence of the described long sequence polymorphisms was used for the definition of the lineage (marked with an "*" in Table 1). In a few cases the RDs fell in scaffolds/contigs junctions, and therefore the presence of polymorphisms could not be accurately determined, precluding the determination of the lineage (marked with a "?" in Table 1).

Screening and alignment of the mclx genes from clinical isolates of an epidemiologically characterized cohort

To get an insight into the epidemiological/clinical features that may be associated with the occurrence of the mclx pseudogenization, the mclx transcriptional integrity was analysed in a diversified panel of Mtb clinical isolates. This panel is composed of 140 organisms collected and isolated in the Netherlands from 1993 to 2011, which are fully characterized from an epidemiological and clinical point of view (S1 Table) [17]. Demographic and clinical information, provided by the Registration Committee of the Netherlands Tuberculosis Register (NTR) that approved this retrospective study, were linked to the isolates on the basis of gender, date of birth, year of diagnosis and postal code. For 100 of these clinical isolates, whole-genome sequencing was performed ([17] and Nebenzahl-Guimaraes et al., unpublished data) and the single nucleotide polymorphisms and INDELs (insertions or deletions) were called against the reference strain H37Rv using Breseq software (version 0.23) with a minimum depth of 15x [18] (S1B Fig). SNPs with low-quality evidence (i.e. possible mixed read alignment) or within 5 bp of an INDEL were discarded. The presence of INDELs within the mclx coding regions and their potential to disrupt the open reading frame was evaluated. For the remaining 40 clinical isolates, the mclx genes were PCR-amplified in 3 overlapping fragments, each of 1200 to 1400 base pairs (see S2 Table for information on the primers and PCR conditions), and the purified PCR products were sequenced (by GATC Biotech). Each gene fragment was amplified twice and all fragments were sequenced in both directions. The final sequences were mapped against and aligned with their orthologues from the Mtb H37Rv reference strain, using the Geneious R7.1.4 (Biomatters) [14] software (S1C Fig). All alignments were visually inspected and a conservative approach was applied: whenever the sequencing results failed to converge to an obvious consensus, the gene status (functional/pseudogene) was considered to be unknown. For that reason, a few clinical isolates were not considered in the final analyses (final n = 127).

Statistical analyses

To envisage the relationship between the clinical and epidemiological features of the TB infection and the status of the mclx genes, a separate logistic regression-based analysis was carried out for mclx2 and mclx3. Firstly, simple binary logistic regressions were performed to identify the significant predictors of each mclx genes status. Afterwards, two multiple binary logistic regressions were performed considering only those considered to be significant predictor variables (p<0.2 in the univariate analysis). Particularly in the case of mclx3, a sequential binary regression model was performed considering three sets of variables (patient-related, microorganism-related and disease-related). All statistical analyses were performed using the IBM SPSS Statistics, version 22 (IBM).

Phylogenetic tree construction

The FASTA files of publicly available NCBI strains were downloaded and pair aligned with the H37Rv reference sequence (NC_000962.gbk) using MAUVE v2.4.0. Multiple sequence alignments (MSA) using 62 robust SNP markers that have been shown to construct high resolution and reproducible phylogenies [19] were then made for 100 of the clinical isolates and 100 of the publicly available NCBI strains. The MSAs were subsequently used to generate a parsimony-based tree using the DNA parsimony algorithm version 3.69 from the Phylip package.

A number of mclx genes among the M. tuberculosis complex species are likely pseudogenized

The mclx genes were previously identified as a family of genes nearly exclusive to the Mycobacterium genus and particularly abundant among the members of Mtb complex [4]. As a distinctive feature, these genes encode proteins with a particular domain architecture, including a CHD (cyclase homology domain) and a LuxR HTH domain. In order to acquire a better understanding of the distribution of this particular group of genes among the Mtb complex, an alignment-based screening was performed against a diversified panel of organisms with their genome fully-sequenced, including elements from the species M. africanum, M. bovis, M. cannetti and Mtb. Interestingly, this strategy revealed that even though mclx genes have in general an overall high degree of sequence conservation, usually with more than 99% nucleotides identical to their orthologues from the reference strain Mtb H37Rv, in a number of them a few INDELs were present, which caused a disruption in the open reading frame leading to the accumulation of stop codons and to the truncation of the respective sequence (Table 1). To avoid misinterpretations, and keeping in mind that gene size can vary to a certain extent without necessarily implying loss of function, the following criterion was established: any given mclx gene was considered to be a pseudogene whenever its sequence was truncated in more than 25% of the size of the corresponding Mtb H37Rv reference orthologue. While this criterion would require further functional validation, we consider it appropriate for this initial screening, as it likely minimizes the chance of false positives (i.e., genes that could be considered pseudogenes given a small size variation but that in fact retain their full functionality).

Overall, our analysis revealed that the pseudogenization is a rather common phenomenon among the mclx genes: 51.3% of the organisms in the analysed panel have at least one of their mclx genes truncated (Table 1 and Fig 1). As shown in Fig 1, the occurrence of the pseudogenes is not evenly distributed among the three different genes: the number of species with a mclx2 pseudogene is much higher than that of species that suffered pseudogenization in the mclx1 and/or the mclx3. In fact, whereas 7.3% and 8.7% of the organisms in the analysed panel have a pseudogenized mclx1 and mclx3, respectively, 39.3% do so for the mcx2. Accordingly, the percentage of organisms with more than 1 mclx pseudogenized is relatively low (3.3%). Notwithstanding, it should be highlighted that in three (out of the four) organisms with two pseudogenes, the pseudogenization events occur in the mclx1 and mclx3 (Table 1 and Fig 1), the two genes with lower pseudogenization occurrence, suggesting that the simultaneous loss of mclx2 and any other mclx may be somehow harmful to the microorganisms.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Relative frequency of the organisms harbouring pseudogenes in each of the mclx genes.

The percentages are relative to the total number of strains/organisms listed in Table 1, and are stratified according to the three main species: Mtb (Mycobacterium tuberculosis), Mbv (Mycobacterium bovis) and Maf (Mycobacterium africanum).

https://doi.org/10.1371/journal.pone.0128983.g001

Epidemiological assessment of the mclx pseudogenization

In order to evaluate whether there was a relationship between the inactivation of these genes in certain bacterial isolates and the epidemiological and clinical characteristics of the disease caused by those isolates, the presence and functionality of the mclx genes was analysed in a panel of Mtb strains isolated in the Netherlands from 1993 to 2011 and fully characterized regarding their epidemiological and clinical features (S1 and S2 Tables). Even though all organisms in this panel were isolated in the Netherlands, only a minority of them (22.0%) were from native Dutch individuals. In fact, the patients' birth region is quite diversified: whereas 29.9% of the patients were born in Europe, 18.1% were born in Africa, 18.1% in the Eastern Mediterranean area, 13.4% in the South East Asia, 10.2% in the Western Pacific and 9.4% in the Americas. Accordingly, and given the strong phylogeographical nature of the Mtb lineages, this panel of isolates holds representatives of the EAm lineage (51.2%), Indo-Oceanic (IO) lineage (24.4%), East-African Indian (EAI) lineage (14.2%), and East-Asian (EAs) lineage (7.9%). From the classical risk factors commonly associated with active TB, the one that stands out in this panel is the origin from an endemic region. As for the other frequently-mentioned risk factors, their occurrence is rather low: only 3.1% of the strains were known to be isolated from homeless patients, 8.7% from known drug and/or alcohol users, and 14.2% from patients with co-morbidities (10.2% were HIV-positive, 3.1% had diabetes mellitus, 0.8% reported a malignancy and none were diagnosed with renal insufficiency or had been through organ transplantation). Regarding the TB localization, 64.6% of the isolates were retrieved from patients diagnosed with pulmonary TB, 18.9% from patients with extra-pulmonary TB, and 15.7% from patients reported to have both pulmonary and extra-pulmonary TB. Microbial transmissibility was defined following the work of Nebenzahl-Guimaraes et al. [17], and 61.4% of the clinical isolates analysed were considered "transmissible". In what concerns the transcriptional integrity of the mclx genes, no pseudogenes were found among mclx1 following the criteria described in the material and methods, whereas 18.9% and 24.4% of the mclx2 and mclx3, respectively, had suffered pseudogenization (S1 Table).

To identify the significant predictors of each mclx2 and mclx3 gene status, simple binary logistic regressions were performed (Table 2). Interestingly, the results were dissimilar for both genes, i.e., the independent variables for which the different categories presented odd ratios (ORs) for pseudogenization statistically different from the reference were not the same for mclx2 and mclx3. The only variable that had a statistically significant association with the pseudogenization for both genes was “transmissibility” (Table 2). However, the effect of this variable had a different direction in each of the genes, i.e., non-transmissible isolates were around four times more likely to carry a pseudogenized copy of mclx2, but around two times more likely to carry a non-pseudogenized copy of mclx3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Univariate ORs for mclx2 and mclx3 pseudogenization (p values <0.2 are highlighted).

https://doi.org/10.1371/journal.pone.0128983.t002

The other variable revealing an association to mclx2 was ethnicity—being a native Dutch represented a 4.5-fold increased risk of carrying a pseudogene (Table 2). The calculation of the ORs for the different Mtb lineages did not yield significant results, as most of the isolates with the mclx2 pseudogenized belong to the EAm lineage (with a single exception—S1 Table), and for a number of lineages the number of pseudogenizations is null. However, isolates belonging to the EAI lineage did have a decreased risk for pseudogenization when compared to isolates belonging to the EAm lineage. Among the completely sequenced and publicly available Mtb complex genomes, there was also only one strain outside the EAm lineage that had a pseudogenized mclx2 (Table 1).

Concerning the mclx3 four other factors besides transmissibility were shown to have a significant association with its pseudogenization: gender, birth region, house setting and localization of the TB infection (Table 2). Being a female represented a 2.2-increased risk for having a pseudogenized form of mclx3, whereas living in an urban area represented a decrease of this risk to 0.345. On the other hand, isolates from strictly extra-pulmonary infections had an increased OR (more than five-fold higher) for mclx3 pseudogenization when compared to strictly pulmonary strains. Finally, birth region was strongly associated with the mclx3 gene status, with no pseudogenes identified in isolates from patients born in Africa, a decreased OR (0.886 and 0.664) for pseudogenization in patients born in the Americas and Eastern Mediterranean, and an increased OR (3.100 and 53.143) for those born in South East Asia and Western Pacific, when compared to Europe (Table 2).

To detect confounding and/or mediation factors in the relation between the different variables with a significant association, multivariate binary logistic regression analyses were performed for the pseudogenization of mclx2 and mclx3 (Tables 3 and 4). These analyses included, for each case, all variables that presented a p below 0.200 in the univariate binary logistic regression (Table 2). These variables were organized into three different blocks (patient-related, microorganism-related and disease-related), which were sequentially added to each multivariate model.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Multivariate logistic regression model for mclx2 pseudogenization (p values < 0.05 are highlighted).

https://doi.org/10.1371/journal.pone.0128983.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Multivariate logistic regression models for mclx3 pseudogenization (p values < 0.05 are highlighted).

https://doi.org/10.1371/journal.pone.0128983.t004

For mclx2, a single model was built including the microorganism-related variable transmissibility, and the patient-related variables age and ethnicity (Table 3). Transmissibility was no longer significant upon correcting for age and ethnicity. In fact, in the multivariate model only ethnicity and age present significant associations: microorganisms isolated from native Dutch have an increased tendency to be carriers of mclx2 pseudogenes, as do microorganisms isolated from younger people. However, it should be noticed that the p value for the microorganism lineage is close to the cut-off (0.200), and one of its categories actually has a p value of 0.036. Adding this variable to the multivariate model, both age and ethnicity lose its significance (S3 Table).

For mclx3, three different models were built: the first one using only the disease-related variable TB localization, the second one including the microorganism-related variable transmissibility, and the third one incorporating the patient-related variables gender, house setting and birth region (Table 4). As for transmissibility, gender and house setting, their associations to mclx3 status were no longer significant after correcting for the other variables in the model. However, birth region remained as a significant variable, with patients born in the Western Pacific having an increased (77-fold) probability of carrying a pseudogenized form of this gene when compared to patients from Europe (Table 4). Since no pseudogenes were found among strains isolated from African patients, it was not possible to perform the mathematical computation of the OR and confidence interval for this category (and for the-Log likelihood of the model no final solution could be found). However, excluding African individuals from the sample had a negligible effect on the calculation of the parameters for the other categories/variables and on the overall significance of this model (data not shown). Finally, disease localization remains as a significant variable even after correcting for the microorganism- and patient-related variables. Clinical isolates from strictly extra-pulmonary infections have a nearly 10-fold increased probability of carrying a pseudogenized form of mclx3 when compared to isolates from strictly pulmonary forms of the disease (Model 3, Table 4). This increased propensity for mclx3 pseudogenized forms is maintained for isolates from disseminated (pulmonary and extra-pulmonary) infections, although in a non-significant way (Model 3, Table 4).

To avoid phylogenetic redundancy, i.e., to ensure that the observed results were due to actual relations observed in the sample and not to the relative abundance of certain genotypes, the analyses were repeated using a single representative for each VNTR and RFLP type (the excluded isolates are annotated in the S1 Table). The results were similar to the previous ones in both the univariate (S4 Table) and multivariate (S5 and S6 Tables) analyses, supporting the initial deductions.

To gain some phylogenetic insight into the distribution of these pseudogenization events, a phylogenetic tree encompassing a number of genomes analysed in this article was constructed and the position of putative pseudogene-causing INDELS annotated (Fig 2). The frameshift-causing INDEL events in the different mclx genes are not unique in each organism, but are often found repeated across different strains/clinical isolates. On the other hand, although most pseudogenes in a given gene are concentrated in the same part of the tree, a few others appear scattered throughout the organisms, preventing a clear phylogenetic signal. Particularly, although most strains with pseudogenization events among mclx genes belong to Mtb, two M. africanum representatives have pseudogenized forms of the mclx1 and there is one M. bovis BCG with a pseudogenized mclx3 (Table 1, Figs 1 and 2). This somewhat dispersed distribution of pseudogenization events, together with the fact that different INDELs occur in different strains/clinical isolates but result in the pseudogenization of the same gene, suggest that the pseudogenization of each mclx may have occurred more than once in their phylogenetic history. This is consistent with a scenario where strong selective pressures are at the basis of these inactivation events, leading to the same overall result—the pseudogenization of a given mclx, although sometimes following different pathways (different INDELs), as opposed to a scenario of random evolution, where the pseudogenization of a given gene would have likely occurred once and dispersed throughout the lineage.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. A parsimony-based phylogenetic tree depicting the distribution of pseudogenization events across the 3 mclx genes in both epidemiologically characterized clinical isolates (n = 100; denoted in black) and publicly available NCBI strains (n = 100; denoted in red).

LSP-defined MTBC lineages/sublineages are colour-coded and indicated in the outer arc (Purple – Mycobacterium africanum; Brown—Mycobacterium bovis; blue—Euro-American lineage; red—East-Asian lineage; orange—Indo-Oceanic; green—East-African Indian lineage).

https://doi.org/10.1371/journal.pone.0128983.g002