Antimicrobial resistance studies often focus on individual protein variants, neglecting broader protein family dynamics. Dihydrofolate reductase (DHFR), a key antibiotic target, has been extensively studied using deep mutational scanning, yet resistance mechanisms across this diverse protein family remain poorly understood. Here, we developed a synthetic metagenomics approach using DropSynth, a scalable gene synthesis platform, to construct a phylogenetically diverse library of 1536 DHFR homologs. These sequences, primarily derived from host-associated metagenomes, represent 759 bacterial species, including many clinically relevant pathogens. A multiplexed in vivo assay tested their ability to restore metabolic function and confer trimethoprim resistance in an ∆ strain. Half of the synthetic homologs rescued the phenotype without supplementation, and mutant variants with up to five amino acid substitutions increased rescue rates to 90%, highlighting DHFR's evolutionary resilience. Broad mutational scanning of DHFR homologs and 100,000 mutants revealed key insights into fitness and resistance, offering the most comprehensive analysis of complementation and inhibitor tolerance to date.