Sequence-encoded determinants of regional mutational plasticity: comparative analysis of PE_PGRS genes in Mycobacterium tuberculosis and other bacteria

Abstract

The PE_PGRS gene family in Mycobacterium tuberculosis exhibits extensive sequence variability across genotypes, which is consistent with antigenic divergence. Here we investigate how Mtb—despite lacking horizontal gene transfer—balances genomic stability with adaptive plasticity. Comparative analysis of 88 bacterial genomes reveals that PE_PGRS genes exhibit features facilitating mutability, including a significantly elevated CGGC tetramer density (mean 4.97 per 100 nt; range 1.7–7.4) compared with the genome-wide average (1.62 per 100 nt; p = 0.011) and depleted in out-of-frame stop codons, potentially conferring robustness to 1-nt and 2-nt frameshifts. Computational predictions suggest that CGGC motifs may promote secondary DNA structures, potentially destabilizing replication and contributing to replication errors, while the scarcity of out-of-frame stop codons allows continued translation beyond frameshifts, leading to changes in protein sequence and length. This dual organization may contribute to the observed adaptability of M. tuberculosis and could highlight a broader principle by which some pathogens evolve under strong constraints on horizontal gene transfer. We propose that CGGC-rich regions may function as programmed mutational hotspots across a wide range of microorganisms.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-47170-w.

Introduction

Mycobacterium tuberculosis (Mtb) remains one of the most successful human pathogens^1–5. The incidence of tuberculosis (TB) persists despite long-standing national and international programs to control the spread of TB^6–10. Mtb has successfully overcome the selective pressure exerted by the use of an ever-expanding array of anti-TB drugs^10–16. Unlike many other bacteria, Mtb lacks horizontal gene transfer mechanisms^17–19. Mtb were conceptualized as a group of genes moving together through time and space²⁰. To actualize high adaptive fitness, Mtb may maintain mechanisms with promutagenic potential that provide mutational variation at loci important for adaptation and survival of Mtb in equilibrium changing external factors²¹. Understanding how genomes simultaneously maintain stability and permit adaptive variability is a central question in evolutionary biology.

Approximately 10% of the coding capacity of the Mtb genome is associated with the PE (including PE_PGRS) and PPE gene family, present only in the genus Mycobacterium²². Many researchers confirm the considerable genetic variation of PE_PGRS genes basing on the presence of multiple single nucleotides polymorphisms, deletions, and insertions ^23–32. This variability stands in contrast to the relatively conserved nature of the core genome, suggesting the presence of endogenous mechanisms behind such site-specific changes. There is a need for studies that bridge evolutionary, structural, and immunological gaps to unravel the contribution of PE/PPE proteins to the adaptive success of Mtb³³.

Currently, spontaneous mutations are commonly defined as replication errors, which presumes uniform occurrence of mutations throughout the genome. However, this assumption neglects the existence of two opposites – highly conserved and hypervariable regions that enable rapid adaptation, as well does not reflect the long-term genomic stability of species throughout megaevolution.

This study aims to determine whether the unique structural features of the highly variable PE_PGRS genes of Mtb can explain the uneven distribution of spontaneous mutations across the genome.

In this study, we explore the hypothesis that the primary structure of PE_PGRS rich with sterically active tetramers such as CGGC and deficient with out-of-frame stop codons predispose them to mutations. Programmed DNA variability, as exemplified by Mtb, may represent a conserved evolutionary strategy among all DNA-based life forms.

Results

Elevated abundance of secondary-structure-forming CGGC motifs in PE_PGRS genes

The WHO catalog of drug resistance-associated mutations in Mtb includes over 17,000 variants across multiple genes³⁴, linked to resistance to first- and second-line anti-TB drugs. Comprehensive cataloging of mutations in drug resistance-associated genes reveals that not all nucleotides are involved in mutagenesis, and strikingly few synonymous (silent) mutations are observed, despite their functional neutrality under natural selection. The long-term use of diagnostic tools such as Xpert® MTB/RIF assay (Cepheid, USA) and GenoType® MTBDRplus assay (Hain Lifescience, Germany) for the detection of mutations associated with resistance to rifampicin and isoniazid further supports the observation that mutations associated with anti-TB drug resistance are restricted to specific nucleotide positions, and neighboring nucleotides around resistance-associated mutation hotspots are seldom involved in mutagenesis.

The design of probes targeting single nucleotide polimorphism in anti-TB drug resistance loci, using secondary DNA structure prediction tools such as UNAFold (Quikfold)³⁵, reveals that mutations frequently occur at sites prone to hairpin formation. Changes in Na⁺ and Mg²⁺ concentrations, as specified in the Quikfold settings, can alter predicted hairpin structures in silico (Fig. 1, Suppl. Tab. 1). The effectiveness of Quikfold has been validated by its application in predicting hairpin structures and PCR annealing temperatures for diagnostic oligonucleotides. Conformational analysis of Mtb mutation sites associated with drug resistance reveals key sequence motifs—including GCCG, CGGC, GCGC, GGG, GGGG, and CTGC—that play a significant role in ssDNA hairpin formation. DNA secondary structures may redistribute local interaction energies and charges, which can lead to nucleotide misincorporation and replication-associated mutagenesis.

Fig. 1.

Predicted secondary DNA structures in the gyrA region encompassing codons 88–94 demonstrate alternative hairpin topologies and interaction energies that may influence the spectrum of quinolone-resistance mutations (Ggc88Tgc, gCg90gTg, Tcg91Ccg, gAc94gGc, Gac94Aac, gAc94gTc, Gac94Cac) (Mfold; Zuker, 2003). (A) – Hairpins in the locus carrying codons 90, 91, and 94 (Energy: ΔG = –3.19, ΔH = –43.10, ΔS = –128.68, Tm = 61.8 °C); (B, C) – Hairpins with alternative secondary structure topologies carrying codon 90 (Energy: ΔG = –2.61, ΔH = –32.00, ΔS = –94.76, Tm = 64.5 °C) and codon 94 (Energy: ΔG = –0.99, ΔH = –22.70, ΔS = –70.00, Tm = 51.1 °C).

In the Mtb genome, only GGG, CGGC, and GCCG among the 28 studied k-mers participating in the formation of DNA conformational structures showed elevated genome-wide densities (~ 1.6 motifs per 100 nt each); for the others, dencity remained below 1 motif per 100 nt. Within PE_PGRS genes, tetramer abundance > 2 per 100 nt was recorded for AACG, GCGC, GCCG, GGG, and CGGC, but frequencies ≥ 4 motifs per 100 nt were limited to GGG (10 PE_PGRS genes) and CGGC (50 PE_PGRS genes, including 30 with ≥ 5 motifs per 100 nt) (Fig. 1).

PE_PGRS genes exhibited significantly higher CGGC tetramer density (per 100 nt) (mean 4.97, range 1.7–7.4; 95% CI: 3.63–6.33) compared to the genome-wide mean (1.62; 95% CI: 0.97–2.27; Mann–Whitney U test: U = 219,210, Z = 12.70, p < 10^–36) (Fig. 2). In contrast to PE_PGRS genes, several genes in the Mtb genome are completely devoid of the CGGC tetramers, indicating a non-random distribution. These include Rv0011c (cell division protein CrgA), Rv0616A (antitoxin VapB29), Rv0722 (50S ribosomal protein L30), Rv1038c, Rv1197, and Rv2347c (ESAT-6–like proteins EsxJ, EsxK, EsxP), Rv3053c (glutaredoxin electron transport protein NrdH), Rv3320c (ribonuclease VapC44 toxin), and a group of genes encoding conserved or hypothetical proteins—Rv0460, Rv2452c, Rv3196A, Rv3678A, Rv2722, Rv0100. Thus, CGGC enrichment in PE_PGRS genes may promote secondary structure formation and create mutational hotspots.

Fig. 2.

(a, b) Abundance of CGGC and pre-stop codons (TAG, TGA, TAA), controlling tolerance to 1nt and 2nt frameshift mutations in PE_PGRS genes. In PE_PGRS genes, Spearman’s correlation coefficients indicate a weak but significant inverse correlation with the frequency of 1-nt pre-stop codons (Spearman’s r = –0.257, n = 61, p = 0.045) and a non-significant very weak correlation with 2-nt pre-stop codons (Spearman’s r = –0.08, n = 61, p = 0.4937). Binomial test showed significant over-representation of CGGC in PE_PGRS genes in the CG-rich model (p = 1.1 × 10⁻³ for ≥ 4 CGGC repeats) and in the Gly-Gly-rich model (p = 0.035 for ≥ 5 repeats).

Although PE_PGRS proteins are highly enriched in glycine residues, which are encoded by GGG, GGA, GGC, and GGT codons, the expected probability of forming a CGG–CGG junction under a random Gly–Gly model is 0.0625, yielding an expected CGGC abundance of ~ 2 motifs per 100 nt. Under a conservative GC-bias and codon-usage null model, CGGC frequency is expected to be ≤ 2 motifs per 100 nt. In contrast, a substantial proportion of PE_PGRS genes (50/62) exhibit 4–7 motifs per 100 nt, significantly exceeding this expectation (exact binomial test, p ≈ 1.1 × 10⁻³ for ≥ 4 motifs). This over-representation indicates that CGGC enrichment in PE_PGRS genes cannot be explained by glycine content or GC bias alone.

To determine whether the absence of CGGC motifs is attributable to gene length, we analyzed 281 small genes (≤ 297 nt, comparable in length to two PE_PGRS genes). CGGC motifs occur even in very short genes: the shortest annotated gene (93 nt) contains one motif. Across genes ≤ 297 nt, the mean CGGC density is 1.46 per 100 nt (range 0.34–5.41). Linear regression showed no relationship between gene length and CGGC abundance (ANOVA: R² = 0.00049, F(1,279) = 0.136, p = 0.712). Thus, the absence of CGGC in some small genes cannot be explained solely by their length.

Inter-species and intra-genomic comparisons of the abundance of CGGC tetramers

The distribution of CGGC tetramers was highly variable across bacterial taxa and within the Mtb genome, and was not determined solely by genomic G + C contentc (Figs. 3 and 4). Within genera, closely related species often differed markedly despite identical G + C%: Yersinia pestis vs Y. enterocolitica (1.79 vs 0.53 per 100 nt; genes with ≥ 2.5 CGGC: 659 vs 7; p < 0.01, Mann–Whitney U test), Pseudomonas aeruginosa vs P. fluorescens (1.79 vs 1.13; 659 vs 31), and Klebsiella pneumoniae (1.06–1.07; 37–40) vs Salmonella enterica (0.73–0.75; 3–6) and E. coli (0.55–0.56; 0–4). Neisseria meningitidis showed higher CGGC levels (1.06–1.11; 20–27) than E. coli and Salmonella despite comparable G + C. Among mycobacteria, CGGC density in Mtb (1.62 per 100 nt, number of genes with CGGC ≥ 2.5: 202–243) was higher than in M. leprae (0.99–1.04; 1–10), and lower than values in M. avium, M. intracellulare, M. kansasii (2.05, 1.85, 1.78; 1021, 586, 601). Corynebacterium diphtheriae, a conserved actinobacterial pathogen and the causative agent of diphtheria, exhibits an even lower CGGC content (0.58 per 100 nt) than M. leprae and most mycobacteria or opportunistic actinomycetes. High average CGGC density is observed in Bordetella pertussis and B. parapertussis (1.90 and 2.0; 744 and 953).

Fig. 3.

Inverse correlation between CGGC tetramer abundance and pre-stop codons (TAA, TGA, TAG) controlling tolerance to 1 nt and 2 nt frameshift mutations across 87 bacterial genomes with different total GC content (%). Values represent genome-wide mean frequencies of CGGC tetramer, 1-nt pre-stop codons, and 2-nt pre-stop codons (per 100 nt) calculated across all coding sequences. Spearman’s correlation coefficients indicate a weak but significant inverse correlation between CGGC density and pre-stop codons controlling tolerance to 1 nt (Spearman’s r = –0.207, n = 3749, p = 1.226 × 10⁻³⁷) and 2 nt (Spearman’s r = –0.162, n = 3749, p = 9.837 × 10⁻²⁴) frameshift mutations. This pattern is observed not only in PE_PGRS genes and Mtb genome but also across taxonomically diverse species, including closely related bacteria within the same genus. The observed distribution cannot be explained solely by variation in genomic G + C content.

Fig. 4.

Number of genes in microbial genomes categorized by CGGC tetramer abundance: (A) genes without CGGC; (B) genes with CGGC abundance > 0 to ≤ 1 per 100 nt; (C) genes with CGGC abundance ≥ 2.5 per 100 nt; (D) genes with CGGC abundance > 3.0 to ≤ 4.0 per 100 nt

Importantly, most of the studied microorganisms harbor gene pools completely devoid of CGGC tetramers. In mycobacteria, this fraction ranges from 12 to 104 genes per genome (Mtb: 15–27 genes; M. leprae: up to 104 genes) (Fig. 4).

Analysis of the prevalence of gene pools with varying CGGC content across bacterial species reveals that genes with moderate enrichment (> 2.5– ≤ 3 per 100 nt) are generally more widespread than those with higher enrichment (> 3– ≤ 4 and > 4 per 100 nt). However, the size of the high-CGGC gene pool (> 3 per 100 nt) varies markedly across species: 68–93 genes in Mtb, substantially extended in M. avium subsp. avium (253), M. intracellulare (115), and M. kansasii (216), and remains limited in M. chelonae (8), M. leprae (1), and M. abscessus (17). Other bacteria with a significant pool of highly CGGC-enriched genes (> 3 per 100 nt) include Bordetella pertussis (264 genes) and B. parapertussis (205), Burkholderia mallei (380) and B. cepacia (312), Streptomyces globisporus (227), and Nocardia asteroides (227). Closely related species may exhibit markedly different representation of genes highly enriched in CGGC (> 3 per 100 nt): Yersinia enterocolitica (5 genes) versus Y. pestis (130), Pseudomonas fluorescens (4) versus P. aeruginosa (130), and Actinomyces naeslundii (6) versus A. israelii (93). The apparent heterogeneity of genes by the degree of their CGGC abundance may reflect variation in mutational potential across genes within a genome and between genomes of different bacterial species.

CGGC-induced hairpins as catalysts of replication errors

The CGGC tetramer exhibits structural advantages that may underlie its promutagenic potential. Thermodynamic modeling (mfold³⁵) of DNA hairpins with identical loops but varying stem compositions showed that CGGC forms the most stable hairpin among tested sequences (ΔG = –4.33 kcal/mol, highest Tm), followed by CGGG, CGCG, GGG, GGGG, and non-GC-rich controls (Suppl. Tab. 2).

Compared to GGG/CCC, CGCG, and GCGC motifs, CGGC provides greater stem stability and more pairing options (including the inverse GCCG and submotifs CCG and GCC, with GCC yielding more stable base pairing). This enhanced stability may better counteract loop destabilization and redistribute energy more efficiently during replication, thereby increasing the likelihood of nucleotide misincorporation and mutagenesis.

Notably, CGGC tetramers are present within major resistance-associated sites in katG (Ser315) and gyrA (near Ala90 and Ser91), and are frequently detected near mutation sites in PE_PGRS genes from sequenced Mtb strains (Suppl. Tables 1 and 2). Their presence in these loci indicates that CGGC may shape local sequence architecture, promoting mutations by disrupting replication fidelity via secondary structure formation.

Evolutionary divergence in the abundance of out-of-frame stop codons conferring 1-nt and 2-nt frameshift tolerance across bacteria

A one- or two-nucleotide insertion or deletion within a coding sequence typically induces a frameshift, potentially resulting in a premature stop codon and the production of a truncated, non-functional peptide. A low abundance of premature stop codons (PSC)—TAA, TGA, TAG—allow translation to proceed far downstream of the original frameshift site, conferring a degree of frameshift tolerance. This phenomenon was observed in the PE_PGRS2 gene of the Mtb strain 4860, in which a premature stop codon appears at position 1001 nt, despite the primary deletion occurring at position 591 nt. A low abundance of premature stop codons may lead to the loss of the canonical stop codon and extension of the protein into intergenic DNA regions, as observed for PE_PGRS12 in Mtb strain 5005 and M. bovis BCG Pasteur (Suppl. Fig. 1).

We analyzed the abundance of PSC within PE_PGRS genes of Mtb, as well as in the rest of its genome, and across a range of other bacterial species. Our analysis revealed that PE_PGRS genes with elevated CGGC content tend to exhibit a lower frequency of PSC, terminating translation upon a 1- or 2-nt frameshift mutations.

In PE_PGRS genes, CGGC abundance displayed a weak but significant negative correlation with the frequency of PSCs controlling 1-nt frameshift mutations (Spearman’s ρ = –0.257, n = 61, p = 0.045) and a very weak, non-significant negative correlation with PSCs controlling 2-nt frameshift mutations (Spearman’s ρ = –0.08, n = 61, p = 0.494) (Fig. 2).

Multiple linear regression showed that GC content was the dominant predictor of PSC frequency (partial r =  − 0.557), while CGGC density had a weaker, marginally non-significant effect (partial r =  − 0.233, β =  − 0.052, p = 0.070). Thus, the reduction in AT-rich stop codons is largely a passive consequence of GC pressure rather than CGGC accumulation.

An inverse correlation between CGGC and PSC abundance controlling 1-nt (Spearman’s ρ = –0.207, n = 3751, p = 1.23 × 10⁻³⁷) and 2-nt (Spearman’s ρ = –0.162, n = 3751, p = 9.84 × 10⁻²⁴) frameshift mutations, typical of the M. tuberculosis genome, was also observed across diverse bacterial species (Fig. 3).

Representation of the gene pool lacking PSCs controlling 1-nt and/or 2-nt frameshift tolerance across bacteria

A reduced abundance of PSCs may impair the efficiency of cellular control over the consequences of frameshift mutations and may serve as a genetic mechanism enabling abrupt changes of protein properties which may be important for antigenic variation and protein diversification. Reduced density of PSCs in the genome may promote mutagenesis and act as a promutagenic factor encoded in the DNA, contributing to rapid protein evolution.

Genomic abundance of genes lacking PSCs conferring both 1- and 2-nt frameshift tolerance. Mtb harbors genes completely devoid of PSCs controlling both 1-nt and 2-nt frameshift tolerance (e.g., Rv0064A, Rv1434, Rv1590, Rv2774c, Rv1473A (transcriptional regulator), Rv2056c (30S ribosomal protein S14)). Such gene pools were detected in nearly all studied microbial genomes (Figs. 4 and 5), ranging from 0 to 102 genes per genome (mean 26.3). Notably larger pools were observed in A. israelii, A. naeslundii, B. parapertussis, B. pertussis, B. mallei (both chromosomes), K. pneumoniae subsp. pneumoniae, most mycobacteria (except M. leprae), N. asteroides, Porphyromonas gingivalis, P. aeruginosa, Rhodococcus erythropolis, S. globisporus, Thermus aquaticus, and Y. pestis.

Fig. 5.

Number of genes in microbial genomes categorized by abundance of pre-stop codons, controlling 1 nt and 2nt frameshift mutation tolerance: (A) genes without pre-stop codons, controlling 1nt; (B) genes without pre-stop codons, controlling 2 nt; (C) genes with 1nt > 2nt; (D) genes with 2 nt > 1nt.

Distribution of genes lacking PSCs conferring 1-nt or 2-nt frameshift tolerance. Genes lacking PSCs exclusively for 1-nt frameshift tolerance were considerably more numerous (e.g., 1,148 in B. parapertussis, 1,564 in B. mallei chr. 1, and up to 1,967–1,970 in N. asteroides, S. globisporus, Y. pestis, and P. aeruginosa). The average number of genes without PSCs controlling 2-nt frameshift tolerance across microorganisms was approximately 264.3. Among mycobacteria, the largest gene pools lacking PSCs for 1-nt frameshift tolerance were found in M. avium (1,130 genes), M. intracellulare (848 genes), M. kansasii (564 genes), and M. fortuitum (459 genes). In Mtb, the number of genes lacking PSCs for 2-nt frameshift tolerance ranged from 199 to 237, similar to that in M. abscessus (232 genes).

Frameshift tolerance varies across genes and species

The abundance of PSCs, conferring 1-nt frameshift tolerance, ranged (per 100 nt) from 0.18 in Y. pestis, B. parapertussis, B. cepacia, and S. globisporus to 4.21 in Borrelia recurrentis (mean 1,77); for 2-nt frameshifts, this value ranged from 0.66 in A. israeli to 4.33 in C. botulinum (mean 1,78), depending on the species. In Mtb, the average abundance of PSCs conferring 1-nt and 2-nt frameshift tolerance was 0.53–0.54 (1.09 in the Mtb H37Rv strain), and 0.98 (1.03% in Mtb H37Rv), respectively, while in M. leprae—obligate pathogen and obligate intracellular parasite─these levels were higher: 1.03–1.05 and 1.3–1.34, respectively.

Control over 1-nt and 2-nt frameshift tolerance varies both between genes and across species: some genes and bacterial taxa exhibit a predominance of 1-nt frameshift control, while others show dominance of 2-nt frameshift control. Among the 74 bacterial species analyzed, 33 showed genome-wide dominance of 1-nt frameshift control, whereas 36 were dominated by 2-nt frameshift control (Fig. 5).

Mycobacteria exhibited stronger genomic robustness to 1-nt than to 2-nt frameshifts, as observed in Mtb (2,964–3,079 vs. 510–547 genes), M. fortuitum (3,028 vs. 5,026), M. chelonae (1,646 vs. 3,293), M. abscessus (1,945 vs. 3,711), M. intracellulare (2,994 vs. 4,546), M. avium subsp. avium (2,679 vs. 4,174), M. kansasii (2,918 vs. 4,632), and M. leprae (541 vs. 1,263).

Certain genes (such as PE_PGRS) and certain microbial species (Mtb and non-tuberculosis mycobacteria, Y. pestis, B. pertussis, B. cepacia, S. globisporus, N. asteroids, P. aeruginosa) may have increased adaptive potential due to attenuated control over 1-nt and/or 2-nt frameshifts.

Discussion

Earlier, it was presumed that the proper functioning of bacteria is encoded in their genome at multiple hierarchical levels, each constrained by specific physical forces³⁶. Variations in tetranucleotide frequencies were suggested as an indicator of the structural organization of microbial genomes³⁷.

To explore this further, we conducted a genome-wide analysis of CGGC tetramer abundance across mycobacteria and diverse taxa of bacteria. We hypothesized that these motifs could be associated with genomic regions of high plasticity.

We focused on the highly variable PE_PGRS genes of Mtb. Our analysis revealed that these genes are significantly enriched in CGGC tetramers relative to the genomic average and markedly depleted in PSCs, particularly those associated with robustness to 1-nt frameshifts.

In silico conformational analyses of DNA at mutation sites associated with resistance to anti-TB drugs predicted the formation of short hairpin structures (7–13 nt in length), with stems frequently containing CGGC or related tetramers (e.g., CGCG, GCGC, CCGG).

Based on these observations, we propose a model wherein the overrepresentation of such sterically active motifs (e.g., CGGC) could create a genomic environment that facilitates nucleotide misincorporation: during replication secondary structures formed by sterically active motifs (CGGC or other) —can alter local electrostatic, hydrophobic interactions, potentially compromising replication fidelity and thereby contribute to elevated mutation rates in these regions.

Our findings indicate that the uneven distribution of mutational variability in the Mtb genome, particularly within hypervariable PE_PGRS genes, correlates with their primary sequence composition (elevated content of specific motifs, including sterically active CGGC tetramers). These findings support a model in which the potential for mutational variability in PE_PGRS genes in Mtb as well as in the genomes of other microorganisms, is encoded within the primary DNA sequence. Genes enriched in CGGC tetramers in Mtb (PE_PGRS) may exhibit a significantly higher susceptibility to mutation. However, whether CGGC-rich regions actively promote this variability through secondary DNA structure formation remains to be experimentally validated.

CGGC motifs may facilitate the formation of secondary DNA structures, analogous to telomeric repeats in eukaryotic genomes. The human TTAGGG motif includes the GGG trimer and the out-of-frame stop codon TAG, and contributes to the formation of stable G-quadruplex structures³⁸, thereby influencing replication and termination signals in eukaryotes.

Several bacterial species outside the genus Mycobacterium also show relatively high CGGC tetramer abundance combined with low frequencies of PSCs, including B. parapertussis, B. pertussis, B. cepacia, B. mallei, N. asteroides, P. aeruginosa, S. globisporus, Y pestis, A. israelii, and some non-tuberculosis mycobacteria, but not M. leprae. Species within the same genus that differ in their clinical relevance and pathogenicity for humans can exhibit significant differences in CGGC and PSC densities. Pathogenic species such as Y. pestis, P. aeruginosa, and Mtb possess markedly higher CGGC and lower PSC density compared to Y. enterocolitica, P. fluorescens, and M. leprae. Compared to M. leprae, Mtb exhibits higher proportion of genes with CGGC > 3 per 100 nt, which aligns with epidemiological observations of continuing global burden of tuberculosis versus the declining incidence of leprosy—although many other genetic, immunological, and socio-economic factors clearly contribute to these epidemiological differences.

When ranked by decreasing CGGC abundance and increasing PSC frequency, the studied non tuberculosis mycobacteria (NTM) follow the order: M. avium, M. intracellulare, M. kansasii, M. marinum, Mtb, M. fortuitum, M. abscessus, M. chelonae, and M. leprae. Interestingly, this order correlates with ethiological structure of NTM infections, in which M. avium, M. intracellulare dominate over other NTM.

In recent years, NTM have demonstrated remarkable adaptive capabilities. In several regions, reported rates of NTM infections now exceed those of tuberculosis, with increasing cases in apparently immunocompetent children and adults and evidence of person-to-person transmission. The drivers of this rise are only beginning to be elucidated³⁹. Differences in virulence and pathogenesis between Mtb and NTM may arise from variations in virulence determinants, such as cord factor and the type VII secretion system. However, the ability of NTM to persist within the host, evade immune responses, and resist antimicrobial treatment may be partly explained by a substantial gene pool enriched in CGGC tetramers and depleted in premature stop codons. This combination may confer enhanced mutational plasticity, thereby amplifying the adaptive potential of NTM. High CGGC abundance coupled with low premature stop codon density may drive evolution by increasing mutational plasticity in specific gene subsets. These features could underlie the observed differences in adaptive potential and epidemiological success among pathogens, including NTM.

The results obtained in this study advance the concept of DNA as a dual-information carrier: digital information defines protein structure via the genetic code, while analogue information reflects sequence-dependent physicochemical properties of DNA that promote the formation of secondary structures, potentially influencing the fidelity of replication and repair processes. PE_PGRS genes were shown to be evolutionarily optimized to exploit this analogue layer, which allows them to exhibit enhanced variability.

The obtained data challenge the canonical concept of random spontaneity of mutations and instead support a model of “constrained spontaneity” for mutations. While mutations are often described as purely random, our findings indicate that their likelihood is not uniform across the genome. We hypothesize that spontaneous mutations emerge more frequently in specific “turbulence zones” – genomic areas enriched with sterically active motifs like CGGC, which may be prone to conformational instability. An elevated CGGC abundance in specific genes may enhance their evolutionary plasticity; however, even a single CGGC motif (or another sterically active sequence) may be sufficient to trigger mutational events. Thus, what is often considered “spontaneous” may be influenced by the analog information embedded in the DNA sequence.

Secondary structure formation in single-stranded DNA is a flexible process, influenced by many factors, including ion concentration and temperature, and is not a strictly fixed process, which predetermines the spontaneous nature of mutations. Analysis of conformational structures at mutation sites indicates that their formation is a universal process, found in the genomes of bacteria, viruses, and human cells (Suppl. Tab. S1).

Although the replication machinery is generally highly accurate, ensuring the megaevolutionary stability of genomes, certain loci remain structurally predisposed to replication errors and reduced efficiency of repair mechanisms. DNA loci harboring CGGC motifs that form stable hairpins can stall replication forks and reduce the fidelity of both DNA polymerases and repair enzymes, increasing mutation fixation rates. These hairpins, stabilized by strong base-pairing interactions, can resist strand separation and efficient melting and coating by single-stranded DNA-binding proteins (SSBs)^40,41.

As previously shown⁴¹, pathogenic microorganisms can adjust their mutation rates. Mutator strains emerge and persist within bacterial populations at varying prevalence. Our data on the CGGC tetramer densities and PSCs were compared with the frequency of mutator strains in bacterial populations. Microorganisms with higher mutator frequencies, such as P. aeruginosa (19.5–92%) and N. meningitidis (22.2–56.8%), exhibited higher CGGC content (1.79% and 1.11%, respectively), whereas species with low mutator rates, including S. aureus (1.4–14.6%) and Vibrio parahaemolyticus (0–5.13%), showed lower CGGC levels (0.07 and 0.27, respectively) (Suppl. Tab. S3).

Our findings support the idea that the genetic code, through its degeneracy, governs the composition of the primary nucleotide sequence, thereby modulating the physicochemical properties of DNA to influence genomic stability—by minimizing secondary DNA structures—or, conversely, to enhance local mutability via sterically active tetramers.

The influence of DNA k-mers (e.g., CGGC) on mutability, as revealed by our study, has implications for genome engineering. Since modulating nucleotide composition via editing can inadvertently alter local DNA stability and mutation rates, the analog architecture of DNA must be accounted for in genome editing design. Identifying motifs that predispose DNA to form secondary structures will enable genome engineering with controlled stability.

PE_PGRS genes of Mtb display approximately threefold higher mutational variability than non-PE/PPE genes, driven by non-synonymous substitutions and frameshift-inducing indels, and occurring largely under near-neutral selection (dN/dS ≈ 0.87). Several mechanisms have been implicated in elevated mutagenesis in Mtb, including mutations in gyrA affecting DNA gyrase activity^46,47 and lineage-specific alterations in DNA repair and recombination (3R) systems⁴⁵, both of which are expected to increase mutation rates genome-wide but do not explain the locus-specific increase in mutational burden observed in PE_PGRS genes. In contrast, the presence of CGGC repeats defines a distinct sequence architecture that promotes elevated mutational plasticity specifically within PE_PGRS loci. Recent studies of RNA G-quadruplexes (rG4s) in mycobacteria have shown that PE_PGRS genes exhibit an rG4 density more than tenfold higher than that observed across the full set of Mtb transcripts. For rG4s, as for CGGC motifs, no correlation was detected between glycine content and rG4 density in PE/PPE transcripts⁴⁹.

Our findings provide new insights into the role of PSCs. While the prevailing hypothesis posits that gene sequences are enriched with PSCs to promptly terminate translation upon frameshifts and minimize energy and resource expenditure⁴³, our data suggest a more complex context. The control of robustness against 1-nt and 2-nt frameshift mutations appears to be less stringent in certain genomic regions. Across all analyzed genomes, we identified gene subsets lacking PSCs associated with 1-nt or 2-nt frameshift mutation control, as well as subsets entirely devoid of PSCs. Furthermore, in the studied microorganisms PSC density varied markedly among different gene pools, indicating diverse genomic strategies for managing frameshift errors. The absence of PSCs decouples frameshift mutations from early translational termination and translation to continue beyond frameshift mutations. This increases the coding potential of the genome by allowing production of altered polypeptides in which functionally critical regions may retain activity, thereby reducing the selective cost of frameshift events⁵⁰. Markedly reduced PSC density may represent a genetic mechanism for altering PE_PGRS protein composition and extending protein length through incorporation of intergenic sequences when canonical stop codon is displaced from frame, as observed for PE_PGRS12 in BCG vaccine strains and the Mtb strain 5005 (Suppl. Fig. 1), that may potentially contribute to antigenic drift and immune evasion. This reduction in PSC density may also constrain complete replacement of AT with GC under GC bias pressure, as unrestricted CG enrichment (including CGGC motifs) leads to automatic depletion of AT-rich codons, including premature stop codons (TAA/TAG/TGA). Low PSC density in certain gene subsets may reduce frameshift robustness and permit synthesis of proteins with modified composition and length.

Mutational plasticity in Mtb and other bacterial species may not arise solely from external selective pressures or random replication errors, but could be intrinsically encoded in the genome. In Mtb, we identified a substantial gene set with elevated mutational plasticity, characterized by high CGGC density and reduced PSC abundance. This genomic architecture positions Mtb as a pathogen with high adaptive potential, where primary DNA sequences in certain genes exhibit pro-mutagenic properties. A similar gene pool enriched in CGGC tetramers and depleted in PSCs was also found in other successful pathogens, including P. aeruginosa, Y. pestis, A. israelii, B. pertussis, B. parapertussis, N. asteroides, and B. cepacia.

Collectively, our findings support the dialectical hypothesis of evolution driven by a dynamic balance between adaptive mutagenesis and genomic stability³⁸: while CG-rich regions maintain DNA stability and replication fidelity, conformationally active CGGC tetramers in ssDNA serve as localized mutation hotspots. These structures disrupt replication fidelity through secondary conformational changes, altering electrostatic and hydrophobic interactions and increasing nucleotide misincorporation.

Limitations

This study focuses on the potential link between CGGC tetramers, DNA mutability and elevated variability of PE_PGRS genes in Mtb, and it has several limitations: it is restricted to a single CG-rich tetramer (CGGC) and a single gene family (PE_PGRS) within Mtb. Predicted secondary structures formed by CGGC motifs were not experimentally validated in this work. However, prior studies from our group demonstrated that conformational structures significantly affect the melting temperature (Tm) of dual diagnostic PCR probes designed to detect polymorphisms in anti-TB drug resistance genes (gyrA, rpoB, rrs, katG) and genotypic markers (Rv2994, pstA1, Rv1956, mce3B, Rv0557, mce3F, rpoC, pncA). The proposed association between CGGC enrichment and DNA mutability relies on comparative genomic analyses rather than direct experiments. Despite these constraints, the large-scale dataset supports a relationship between CGGC density and regional variation in PE_PGRS genes.

Methods

Whole-genome sequencing

All experimental procedures involving live Mtb cultures were performed in a certified Biosafety Level 3 (BSL-3) laboratory in accordance with institutional and national biosafety guidelines.

High-throughput sequencing of M. tuberculosis genomes was performed using MiSeq (Illumina) and MinION (Oxford Nanopore) platforms. Mycobacterial DNA was isolated using the QIAGEN Blood & Cell Culture DNA Maxi Kit (Cat. No./ID: 13,362) or the QIAamp DNA Mini Kit (Cat. No./ID: 51,304). Genome assemblies were generated with SPAdes v3.14, A5-miseq v20160825, Flye v2.7b, and Canu v1.8. The obtained complete genome sequences were deposited in GenBank: M. tuberculosis 11,502 (CP070338.1, BioSample SAMN17832565), M. tuberculosis 5005 (CP053092.1, SAMN14150054), M. tuberculosis 4860 (CP049108.1, SAMN14598146), M. tuberculosis 31d (CP110674.1, SAMN31644110), M. tuberculosis 4299 (CP125620.1, SAMN32350377), and M. tuberculosis BLR 9248 2019 (CP115447.1).

Bioinformatics research

Thermodynamic modeling of DNA secondary structures was carried out using the Quikfold web-based tool (UNAFold Web Server; Zuker, 2003). To assess the contribution of specific stem motifs to hairpin stability, a set of oligonucleotides was in silico generated with identical loop sequences and variable stem regions. The tested combinations included representative CG-rich configurations observed in Mtb genes, as well as alternative CG arrangements used for comparison. Melting temperatures (Tm) and Gibbs free energies (ΔG) were computed in silico for the following sequences (stem regions in brackets): AA[CGGC]ACCA[GCCG]AA; AA[CGGC]ACCA[GCC]AA; AA[CGGG]ACCA [CCCG] AA; AA[GGGC]ACCA[GCCC]AA; AA[CGGC]ACCA[CCG]AA; CAA[GGG]ACCA [CCC]AAG; AA[GGG]ACCA[CCC]AA; AA[GGG]ACGCGTACA[CCC]AA; AA[GGGG] ACCA[CCCC]AA; CAA[CGG]ACCA[CCG]AAG; A[CTTC]ACCA[GAAG]A (control); A[GTTG]ACCA[CAAC]A (control).

The abundance of nucleotide motifs involved in the formation of secondary conformational structures (GCCGGTGTTG, CGGCGGCAA, AAAA, TTTT, GTCGGC, GCCGAC, TTCT, TAA, CAGA, CCCC, TAG, GTCGG, TCTG, AACC, GGGG, TTGG, GGTT, CCAA, AACG, CGTT, TCGT, GCCGA, ACGA, AGCG, CGCT, GTCG, GCGC, TGA, GGG, CGGC, and GCCG) was assessed across all coding sequences of the Mtb genome, including PE_PGRS genes. Genomic analysis of CGGC motifs, and triplets TGA, TAG, and TAA (expressed as percentages or counts per 100 bp) was performed across coding genes of Mtb and compared with bacterial genomes from diverse taxa (88 genomes, 74 species). All genome assemblies processed in this study are available in GenBank (NCBI; Suppl. Tab. 4). Analyses were conducted using java based tool, available on GitHub (https://github.com/ashakirin/microbilogy-2; A. Shakirin, 2022, VMWare Tanzu Labs, Germany).

Statistical analysis

The CGGC tetramer and PSC densities (per 100 nt) were calculated as the mean for each annotated coding sequence. Statistical analyses, including Shapiro–Wilk normality test, Mann–Whitney U test, Spearman’s rank correlation, linear regression with ANOVA, and binomial tests, were conducted using the online platform StatsKingdom.

Author contributions

Veranika Slizen (VS) and Henadz Hurevich (HH) designed the study. V.S. performed the sequencing of Mtb isolates, carried out all comparative genomic analyses, wrote the original draft. H.H. provided scientific supervision, performed the statistical analyses, and critically revised the manuscript. The use of AI was limited to language editing and formatting support.

Funding

The genome sequencing part of this study was supported by the State Program of Scientific Research of the Republic of Belarus ‘Fundamental and Applied Sciences for Medicine’ (project No. 20190333, 2016–2020). No external funding supported the comparative genomics analysis or manuscript preparation.

Data availability

Processed datasets regarding the distribution of CGGC tetramers and TGA, TAG, and TAA out-of-frame stop codons in Mtb and 81 additional microorganisms are publicly accessible via GitHub (https://github.com/Vero-Slizen/Data-for-Scientific-Reports.git). Additional data supporting the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

This study used only de-identified archived Mtb isolates from an institutional culture collection. The original sputum samples were collected as part of routine diagnostic procedures in accordance with the Clinical Protocol of the Ministry of Health of the Republic of Belarus. Given the retrospective use of archival bacterial isolates with no patient identifiers and no direct patient involvement, no additional ethics approval or informed consent was required for this study.