P069

South Africa (SA) has the highest per capita rate of multidrug-resistant (MDR) tuberculosis globally. We performed a retrospective temporal analysis of 2,407 whole-genome sequences (WGS) from clinical Mycobacterium tuberculosis isolates collected nationally across SA with culture-based drug-susceptibility data for at least one drug. We constructed lineage-specific phylogenies and assessed clustering based on node density and pairwise SNP distance in different drug resistance (DR) groups. Additionally, we used Bayesian molecular dating in BEAST 1.10.4 to estimate the ages of 1,392 most recent common susceptible ancestors across 10,347 instances of resistance to fourteen drugs. Of the 2407 isolates, 1,144 (48.9%) were resistant to rifampicin, and 1,028 (42.7%) were MDR; 782 (32.5%) belonged to 102 clusters with average pairwise distance <25 SNPs. The median ancestral age of these clusters was 34.5 years (IQR 21.6-46.6). Cluster membership was significantly associated with DR across all drugs and DR groups (p-values from 5.23e-04 to 3.08e-195). In simulations, the association between clustering and DR was unlikely to be observed because of random oversampling of DR in our dataset (p<0.0001). The number of independent resistance acquisition events was lower than the number of resistant isolates, both within and outside clusters, indicating ongoing transmission of drug resistance. Our findings suggest that the epidemic of MDR in SA is significantly driven by clonal and persistent transmission of highly fit primary resistant isolates, highlighting the need for enhanced surveillance to control the spread of transmissible MDR strains.