Antimicrobial resistance is a global health threat and a growing barrier to the treatment of Mycobacterium tuberculosis ( M. tb ), the deadliest human pathogen. For over a hundred years, antibiotics were restricted to small molecules. However, advances in computational biophysics have enabled highly accurate design of proteins, unlocking possibilities for protein-based therapeutics. We seek to engineer a protein-based binder capable of inhibiting multiple protein targets for treating drug-resistant tuberculosis. By targeting the mmpL family of transporters, which are involved in key fitness and resistance mechanisms, we intend to reduce the capacity of M. tb to survive and escape antibiotic effects. Employing machine learning-based methods for protein design, we computationally generated de novo proteins predicted to bind to the essential mycolic acid transporter mmpL3. We then sought to elucidate whether they inhibit mmpL3 in the model organism Mycobacterium smegmatis ( M. smeg ), a species closely related to M. tb . By inducing expression of the candidate de novo protein antibiotics in M. smeg and examining the impact on bacterial growth, the effectiveness of the proteins at blocking mmpL3 function can be assessed. If these efforts to design an mmpL3 inhibitor are successful, we will increase the number of mmpL transporters targeted by our designs, beginning with mmpL5, a critical mediator of resistance to last-resort antibiotics. Ultimately, we seek to engineer a binder that can inhibit numerous mmpL family members, expanding the physiochemical diversity that can be recognized by a single protein.
Johnson et al. (Sun,) studied this question.