L-Homophenylalanine (L-Hph) is a useful chiral building block for synthesis of several drugs, including angiotensin-converting enzyme inhibitors and the novel proteasome inhibitor carfilzomib. While the chemoenzymatic route of synthesis is fully developed, we investigated microbial production of L-Hph to explore the possibility of a more efficient and sustainable approach to L-Hph production. We hypothesized that L-Hph is synthesized from L-Phe via a mechanism homologous to 3-methyl-2-oxobutanoic acid conversion to 4-methyl-2-oxopentanoic acid during leucine biosynthesis. Based on bioinformatics analysis, we found three putative homophenylalanine biosynthesis genes, hphA (Npun_F2464), hphB (Npun_F2457), and hphCD (Npun_F2458), in the cyanobacterium Nostoc punctiforme PCC73102, located around the gene cluster responsible for anabaenopeptin biosynthesis. We constructed Escherichia coli strains harboring hphABCD-expressing plasmids and achieved the fermentative production of L-Hph from L-Phe. To our knowledge, this is the first identification of the genes responsible for homophenylalanine synthesis in any organism. Furthermore, to improve the low conversion efficiency of the initial strain, we optimized the expression of hphA, hphB, and hphCD, which increased the yield to ∼630 mg/liter. The L-Hph biosynthesis and L-Leu biosynthesis genes from E. coli were also compared. This analysis revealed that HphB has comparatively relaxed substrate specificity and can perform the function of LeuB, but HphA and HphCD show tight substrate specificity and cannot complement the LeuA and LeuC/LeuD functions, and vice versa. Finally, the range of substrate tolerance of the L-Hph-producing strain was examined, which showed that m-fluorophenylalanine, o-fluorophenylalanine, and L-tyrosine were accepted as substrates and that the corresponding homoamino acids were generated.