Development of in silico models to predict viscosity and mouse clearance using a comprehensive analytical data set collected on 83 scaffold-consistent monoclonal antibodies.

Biologic drug discovery pipelines are designed to deliver protein therapeutics that have exquisite functional potency and selectivity while also manifesting biophysical characteristics suitable for manufacturing, storage, and convenient administration to patients. The ability to use computational methods to predict biophysical properties from protein sequence, potentially in combination with high throughput assays, could decrease timelines and increase the success rates for therapeutic developability engineering by eliminating lengthy and expensive cycles of recombinant protein production and testing. To support development of high-quality predictive models for antibody developability, we designed a sequence-diverse panel of 83 effector functionless IgG1 antibodies displaying a range of biophysical properties, produced and formulated each protein under standard platform conditions, and collected a comprehensive package of analytical data, including in vitro assays and in vivo mouse pharmacokinetics. We used this robust training data set to build machine learning classifier models that can predict complex protein behavior from these data and features derived from predicted and/or experimental structures. Our models predict with 87% accuracy whether viscosity at 150 mg/mL is above or below a threshold of 15 centipoise (cP) and with 75% accuracy whether the area under the plasma drug concentration-time curve (AUC0-672 h) in normal mouse is above or below a threshold of 3.9 × 106 h x ng/mL.

Toward generalizable prediction of antibody thermostability using machine learning on sequence and structure features.

Over the last three decades, the appeal for monoclonal antibodies (mAbs) as therapeutics has been steadily increasing as evident with FDA's recent landmark approval of the 100th mAb. Unlike mAbs that bind to single targets, multispecific biologics (msAbs) have garnered particular interest owing to the advantage of engaging distinct targets. One important modular component of msAbs is the single-chain variable fragment (scFv). Despite the exquisite specificity and affinity of these scFv modules, their relatively poor thermostability often hampers their development as a potential therapeutic drug. In recent years, engineering antibody sequences to enhance their stability by mutations has gained considerable momentum. As experimental methods for antibody engineering are time-intensive, laborious and expensive, computational methods serve as a fast and inexpensive alternative to conventional routes. In this work, we show two machine learning approaches - one with pre-trained language models (PTLM) capturing functional effects of sequence variation, and second, a supervised convolutional neural network (CNN) trained with Rosetta energetic features - to better classify thermostable scFv variants from sequence. Both of these models are trained over temperature-specific data (TS50 measurements) derived from multiple libraries of scFv sequences. On out-of-distribution (refers to the fact that the out-of-distribution sequnes are blind to the algorithm) sequences, we show that a sufficiently simple CNN model performs better than general pre-trained language models trained on diverse protein sequences (average Spearman correlation coefficient, ρ, of 0.4 as opposed to 0.15). On the other hand, an antibody-specific language model performs comparatively better than the CNN model on the same task (ρ= 0.52). Further, we demonstrate that for an independent mAb with available thermal melting temperatures for 20 experimentally characterized thermostable mutations, these models trained on TS50 data could identify 18 residue positions and 5 identical amino-acid mutations showing remarkable generalizability. Our results suggest that such models can be broadly applicable for improving the biological characteristics of antibodies. Further, transferring such models for alternative physicochemical properties of scFvs can have potential applications in optimizing large-scale production and delivery of mAbs or bsAbs.

Machine learning to design integral membrane channelrhodopsins for efficient eukaryotic expression and plasma membrane localization.

There is growing interest in studying and engineering integral membrane proteins (MPs) that play key roles in sensing and regulating cellular response to diverse external signals. A MP must be expressed, correctly inserted and folded in a lipid bilayer, and trafficked to the proper cellular location in order to function. The sequence and structural determinants of these processes are complex and highly constrained. Here we describe a predictive, machine-learning approach that captures this complexity to facilitate successful MP engineering and design. Machine learning on carefully-chosen training sequences made by structure-guided SCHEMA recombination has enabled us to accurately predict the rare sequences in a diverse library of channelrhodopsins (ChRs) that express and localize to the plasma membrane of mammalian cells. These light-gated channel proteins of microbial origin are of interest for neuroscience applications, where expression and localization to the plasma membrane is a prerequisite for function. We trained Gaussian process (GP) classification and regression models with expression and localization data from 218 ChR chimeras chosen from a 118,098-variant library designed by SCHEMA recombination of three parent ChRs. We use these GP models to identify ChRs that express and localize well and show that our models can elucidate sequence and structure elements important for these processes. We also used the predictive models to convert a naturally occurring ChR incapable of mammalian localization into one that localizes well.

Structure-guided SCHEMA recombination generates diverse chimeric channelrhodopsins.

Integral membrane proteins (MPs) are key engineering targets due to their critical roles in regulating cell function. In engineering MPs, it can be extremely challenging to retain membrane localization capability while changing other desired properties. We have used structure-guided SCHEMA recombination to create a large set of functionally diverse chimeras from three sequence-diverse channelrhodopsins (ChRs). We chose 218 ChR chimeras from two SCHEMA libraries and assayed them for expression and plasma membrane localization in human embryonic kidney cells. The majority of the chimeras express, with 89% of the tested chimeras outperforming the lowest-expressing parent; 12% of the tested chimeras express at even higher levels than any of the parents. A significant fraction (23%) also localize to the membrane better than the lowest-performing parent ChR. Most (93%) of these well-localizing chimeras are also functional light-gated channels. Many chimeras have stronger light-activated inward currents than the three parents, and some have unique off-kinetics and spectral properties relative to the parents. An effective method for generating protein sequence and functional diversity, SCHEMA recombination can be used to gain insights into sequence-function relationships in MPs.

Directed Evolution of a Bright Near-Infrared Fluorescent Rhodopsin Using a Synthetic Chromophore.

By engineering a microbial rhodopsin, Archaerhodopsin-3 (Arch), to bind a synthetic chromophore, merocyanine retinal, in place of the natural chromophore all-trans-retinal (ATR), we generated a protein with exceptionally bright and unprecedentedly red-shifted near-infrared (NIR) fluorescence. We show that chromophore substitution generates a fluorescent Arch complex with a 200-nm bathochromic excitation shift relative to ATR-bound wild-type Arch and an emission maximum at 772 nm. Directed evolution of this complex produced variants with pH-sensitive NIR fluorescence and molecular brightness 8.5-fold greater than the brightest ATR-bound Arch variant. The resulting proteins are well suited to bacterial imaging; expression and stability have not been optimized for mammalian cell imaging. By targeting both the protein and its chromophore, we overcome inherent challenges associated with engineering bright NIR fluorescence into Archaerhodopsin. This work demonstrates an efficient strategy for engineering non-natural, tailored properties into microbial opsins, properties relevant for imaging and interrogating biological systems.

Diversity in ABC transporters: type I, II and III importers.

ATP-binding cassette transporters are multi-subunit membrane pumps that transport substrates across membranes. While significant in the transport process, transporter architecture exhibits a range of diversity that we are only beginning to recognize. This divergence may provide insight into the mechanisms of substrate transport and homeostasis. Until recently, ABC importers have been classified into two types, but with the emergence of energy-coupling factor (ECF) transporters there are potentially three types of ABC importers. In this review, we summarize an expansive body of research on the three types of importers with an emphasis on the basics that underlie ABC importers, such as structure, subunit composition and mechanism.

Effects of lipid environment on the conformational changes of an ABC importer.

In order to shuttle substrates across the lipid bilayer, membrane proteins undergo a series of conformation changes that are influenced by protein structure, ligands, and the lipid environment. To test the effect of lipid on conformation change of the ABC transporter MolBC, EPR studies were conducted in lipids and detergents of variable composition. In both a detergent and lipid environment, MolBC underwent the same general conformation changes as detected by site-directed EPR spectroscopy. However, differences in activity and the details of the EPR analysis indicate conformational rigidity that is dependent on the lipid environment. From these observations, we conclude that native-like lipid mixtures provide the transporter with greater activity and conformational flexibility as well as technical advantages such as reconstitution efficiency and protein stability.

Small substrate transport and mechanism of a molybdate ATP binding cassette transporter in a lipid environment.

Embedded in the plasma membrane of all bacteria, ATP binding cassette (ABC) importers facilitate the uptake of several vital nutrients and cofactors. The ABC transporter, MolBC-A, imports molybdate by passing substrate from the binding protein MolA to a membrane-spanning translocation pathway of MolB. To understand the mechanism of transport in the biological membrane as a whole, the effects of the lipid bilayer on transport needed to be addressed. Continuous wave-electron paramagnetic resonance and in vivo molybdate uptake studies were used to test the impact of the lipid environment on the mechanism and function of MolBC-A. Working with the bacterium Haemophilus influenzae, we found that MolBC-A functions as a low affinity molybdate transporter in its native environment. In periods of high extracellular molybdate concentration, H. influenzae makes use of parallel molybdate transport systems (MolBC-A and ModBC-A) to take up a greater amount of molybdate than a strain with ModBC-A alone. In addition, the movement of the translocation pathway in response to nucleotide binding and hydrolysis in a lipid environment is conserved when compared with in-detergent analysis. However, electron paramagnetic resonance spectroscopy indicates that a lipid environment restricts the flexibility of the MolBC translocation pathway. By combining continuous wave-electron paramagnetic resonance spectroscopy and substrate uptake studies, we reveal details of molybdate transport and the logistics of uptake systems that employ multiple transporters for the same substrate, offering insight into the mechanisms of nutrient uptake in bacteria.

EPR spectroscopy of MolB2C2-a reveals mechanism of transport for a bacterial type II molybdate importer.

In bacteria, ATP-binding cassette (ABC) transporters are vital for the uptake of nutrients and cofactors. Based on differences in structure and activity, ABC importers are divided into two types. Type I transporters have been well studied and employ a tightly regulated alternating access mechanism. Less is known about Type II importers, but much of what we do know has been observed in studies of the vitamin B12 importer BtuC2D2. MolB2C2 (formally known as HI1470/71) is also a Type II importer, but its substrate, molybdate, is ∼10-fold smaller than vitamin B12. To understand mechanistic differences among Type II importers, we focused our studies on MolBC, for which alternative conformations may be required to transport its relatively small substrate. To investigate the mechanism of MolBC, we employed disulfide cross-linking and EPR spectroscopy. From these studies, we found that nucleotide binding is coupled to a conformational shift at the periplasmic gate. Unlike the larger conformational changes in BtuCD-F, this shift in MolBC-A is akin to unlocking a swinging door: allowing just enough space for molybdate to slip into the cell. The lower cytoplasmic gate, identified in BtuCD-F as "gate I," remains open throughout the MolBC-A mechanism, and cytoplasmic gate II closes in the presence of nucleotide. Combining our results, we propose a peristaltic mechanism for MolBC-A, which gives new insight in the transport of small substrates by a Type II importer.