In spite of variations in the sequences of tRNAs, the genetic code (anticodon trinucleotides) is conserved in evolution. However, non-anticodon nucleotides which are species specific are known to prevent a given tRNA from functioning in all organisms. Conversely, species-specific tRNA contact residues in synthetases should also prevent cross-species acylation in a predictable way. To address this question, we investigated the relatively small tyrosine tRNA synthetase where contacts of Escherichia coli tRNA(Tyr) with the alpha2 dimeric protein have been localized by others to four specific sequence clusters on the three-dimensional structure of the Bacillus stearothermophilus enzyme. We used specific functional tests with a previously not-sequenced and not-characterized Mycobacterium tuberculosis enzyme and showed that it demonstrates species-specific aminoacylation in vivo and in vitro. The specificity observed fits exactly with the presence of the clusters characteristic of those established as important for recognition of E. coli tRNA. Conversely, we noted that a recent analysis of the tyrosine enzyme from the eukaryote pathogen Pneumocystis carinii showed just the opposite species specificity of tRNA recognition. According to our alignments, the sequences of the clusters diverge substantially from those seen with the M. tuberculosis, B. stearothermophilus and other enzymes. Thus, the presence or absence of species-specific residues in tRNA synthetases correlates in both directions with cross-species aminoacylation phenotypes, without reference to the associated tRNA sequences. We suggest that this kind of analysis can identify those synthetase-tRNA covariations which are needed to preserve the genetic code. These co-variations might be exploited to develop novel antibiotics against pathogens such as M. tuberculosis and P. carinii.