Morphing of MhsT models obtained from SPEACH_AF illustrating its conformational changes.
The Idea
The lab of Hassane Mchaourab, director of the Center for Applied Artificial Intelligence in Protein Dynamics and professor of molecular physiology and biophysics, has developed a methodological blueprint that couples experimental double electron–electron resonance spectroscopy with an AI approach to help describe the conformational landscapes of a diverse spectrum of cell membrane transporters.
The research, led by recent Ph.D. graduate and first author Alexandra C. Schwartz, was published in PNAS and focuses on a bacterial homolog of neurotransmitter:sodium symporters. These proteins regulate neurotransmitter levels in synapses by pumping them back into cells.
Schwartz used DEER spectroscopy to probe the structure and energetics of a Bacillus halodurans protein, MhsT, as a model. The Mchaourab lab, which has specialized in DEER for over a decade, engineers “spin” labels at two locations on a protein at a time, typically on transmembrane helices. This approach allows the researchers to measure how the distance between the two spin labels changes between conditions and infer their position and movement with one another.
“DEER is a low-throughput method where each piece of data is akin to a puzzle piece,” Schwartz said. “You need to put the pieces together to understand how the protein elements come together to form the overall structure and the global conformational changes that underlie its function.”
This process presented a challenge when she started her Ph.D. in 2018, but technological innovations sometimes come at opportune moments: The development of the AI tool AlphaFold (now in its third iteration, AlphaFold3) by Vanderbilt alum and 2024 Nobel laureate in chemistry John Jumper, BS’07, gave Schwartz a way to surpass what has traditionally been a huge technical challenge for DEER spectroscopists.
To obtain structures that align with the movements observed in DEER, Schwartz and colleagues integrated an AlphaFold2-based approach called SPEACH_AF into their workflow. SPEACH_AF, which was developed by Richard A. Stein, research associate professor of molecular physiology and biophysics and Mchaourab lab member, generates alternate protein conformations by computationally modifying the protein sequence alignment before feeding it into AlphaFold2, which then predicts the protein’s 3D structure. This process helped establish a cohesive conformational landscape that explained MhsT’s physical transformations during ligand transport.
Why it matters
The neurotransmitter transporters for serotonin, dopamine, and norepinephrine are implicated in a variety of neuropsychiatric disorders and diseases, including Parkinson’s disease, attention deficit and hyperactivity disorder, depression, anxiety, and autism spectrum disorder. NSSs are also molecular targets for psychoactive drugs—including multiple classes of antidepressants—and addictive substances—including cocaine and amphetamines. Using bacterial NSS homologs allows researchers like Schwartz to gain insights into the underlying mechanisms of the human proteins, and using DEER and AlphaFold2 enables a broader perspective of the protein’s structural dynamics, a process that is not as straightforward when using other structural biology techniques.
One key element of the paper that the authors were not expecting at the outset was MhsT’s divergent behavior when tested in different membrane environments. Traditionally, detergent micelles (a monolayer) have been used to evaluate membrane protein dynamics. However, their less rigid structure relative to the more biologically relevant lipid nanodisc environment (a bilayer) seems to influence the protein’s behavior. Schwartz’s data suggest that nanodiscs play a substantial role in the structure and mechanism of NSSs more than previously demonstrated.
“My original goal with the nanodisc experiments was just to validate the micelle data, but instead I found a unique effect that suggested that MhsT’s energetics were distinct to each membrane environment,” Schwartz said. “Not only that, but the lipid nanodiscs actually enable MhsT to undergo coordinated shifts not observed in micelles.”
What’s next
The Mchaourab lab, together with the Center for Applied AI in Protein Dynamics, is now exploring the integration of AlphaFold and other AI approaches into the design, analysis, and interpretation of experimental data. These efforts aim to generate increasingly accurate structural models from experimental data, formulate early hypotheses, and streamline the currently low-throughput DEER data collection process.
“You can capture the structural architecture with cryo-electron microscopy or X-ray crystallography, but the conformations they provide are usually snapshots of a single state,” Schwartz said. “With this new method, we can obtain structures that complement the structural and energetics information gathered from the DEER data.”
As Schwartz transitions to the workforce, the Mchaourab lab will continue its work with DEER and AI and is planning to apply their integrated approach to a wide range of membrane proteins with different functions, albeit with a major focus on the NSS family, particularly mammalian NSSs.
The Mchaourab lab is currently working on the human serotonin transporter, hSERT, and will examine its transport mechanism to identify similarities and differences with other members of the transporter family and hopefully provide insights into the evolutionary conservation of NSSs.
“Understanding how disease-linked mutations affect transporter function could also inform the development of targeted therapeutics that compensate for specific changes in their mechanisms,” Schwartz said.
Go deeper
The paper “Alternating access of a bacterial homolog of neurotransmitter: sodium symporters determined from AlphaFold2 ensembles and DEER spectroscopy” was published in PNAS in September 2024.
Funding
This research used funds from the National Institute of Neurological Disorders and Stroke and the National Institute of General Medical Sciences.