Asp29Glu marks the re-humanised Mycobacterium tuberculosis complex lineages without affecting the protease melE (Rv2569c)

The lineages of the Mycobacterium tuberculosis complex (MTBC) that returned to a human host after the animal radiation, L6 together with L9 and L10, share a small set of fixed variants. One of them, NC_000962.3:2894321:G:C (c.87C>G, p.Asp29Glu), falls in Rv2569c (melE), a macrophage-infection protease of the transglutaminase family described since 2004 and carrying the strongest in vivo phenotype signal; I tested whether it has any functional effect. Across 142 673 genomes assigned to 705 clades, the variant is carried by 1400 of the 1402 genomes of the L6+L9+L10 node, the two exceptions being low-coverage false negatives (among the lowest call counts and read depths in their sub-clade, both since confirmed as under-called carriers in their raw reads), so 1400/1400 informative members of that snapshot; and by 1 of the 141 271 genomes outside the node, a homoplasy (ordinary-quality genome). A re-examination on a later and larger snapshot finds no genuine exception: it confirms the two as carriers in their reads, and the one apparent non-carrier it adds belongs to another lineage and is misfiled in the node’s container of the source database, so the site is carried by 1490 of the 1490 node members of that snapshot as well. It is the only non-synonymous change of the gene to reach that frequency in the clade, and the rest of the mel1 locus carries none. Residue 29 lies 31.9 Å (Cα-Cα) from the catalytic nucleophile Cys196, placed by two independent routes; the general base and acid are a moderate-confidence hypothesis. The residue is not in the disordered region a sequence predictor had declared, but in a folded, exposed domain (pLDDT 95.7) devoid of any detectable binding pocket, ranking 311th and 309th of 312 residues for cryptic-pocket propensity with the two predictors, the nucleophile and the general base of the catalytic triad (positive control) ranking in the first tenth with the first predictor. Its predicted impact is below the median of 26 matched controls from the same set (S6). The substitution is therefore a near-perfect diagnostic marker for a node that no barcode currently names, and none of the measures applied detects an effect of it. Three parameters of melE are also revised: its catalytic class is transglutaminase-like rather than serine, its active site is numbered here for the first time, and the apparent disagreement about its in vivo role disappears once the host backgrounds are compared, the directed knockouts having been made in the one background in which neither transposon screen detects anything.