There is evidence that some members of the Mycobacterium tuberculosis PE_PGRS gene subfamily, including PE_PGRS33, may have a specific function in M. tuberculosis persistence. The impact of naturally-occurring PE_PGRS33 genetic variations on the virulence and transmissibility of clinical M. tuberculosis isolates is not known. We used PCR and DNA sequencing to identify genetic variations in the PE_PGRS33 gene in comparison with the sequenced laboratory strain, H37Rv, among 649 isolates from a population-based sample. The PE_PGRS33 alleles were placed into two groups, based on the effect of the sequence variations on the PE_PGRS33 protein, and their associations with clinical and epidemiological characteristics were assessed using multivariate logistic regression to control for potential confounding of host-related factors. Of the 639 isolates for which sequence data were obtained, 139 (21.8%) had PE_PGRS33 alleles that would result in a significant change to the PE_PGRS33 protein due to large insertions/deletions or frameshift mutations. These isolates were significantly associated with clustering based on genotype and absence of cavitations in the lungs, compared to isolates having PE_PGRS33 alleles that would result in no or minimal change to the PE_PGRS33 protein. The association of significant changes to PE_PGRS33 with clinical and epidemiological characteristics suggests that PE_PGRS33 may have an important role in M. tuberculosis persistence.