Large-scale design and refinement of stable proteins using sequence-only models.

Engineered proteins generally must possess a stable structure in order to achieve their designed function. Stable designs, however, are astronomically rare within the space of all possible amino acid sequences. As a consequence, many designs must be tested computationally and experimentally in order to find stable ones, which is expensive in terms of time and resources. Here we use a high-throughput, low-fidelity assay to experimentally evaluate the stability of approximately 200,000 novel proteins. These include a wide range of sequence perturbations, providing a baseline for future work in the field. We build a neural network model that predicts protein stability given only sequences of amino acids, and compare its performance to the assayed values. We also report another network model that is able to generate the amino acid sequences of novel stable proteins given requested secondary sequences. Finally, we show that the predictive model-despite weaknesses including a noisy data set-can be used to substantially increase the stability of both expert-designed and model-generated proteins.

Aquarium: open-source laboratory software for design, execution and data management.

Automation has been shown to improve the replicability and scalability of biomedical and bioindustrial research. Although the work performed in many labs is repetitive and can be standardized, few academic labs can afford the time and money required to automate their workflows with robotics. We propose that human-in-the-loop automation can fill this critical gap. To this end, we present Aquarium, an open-source, web-based software application that integrates experimental design, inventory management, protocol execution and data capture. We provide a high-level view of how researchers can install Aquarium and use it in their own labs. We discuss the impacts of the Aquarium on working practices, use in biofoundries and opportunities it affords for collaboration and education in life science laboratory research and manufacture.

Perturbing the energy landscape for improved packing during computational protein design.

The FastDesign protocol in the molecular modeling program Rosetta iterates between sequence optimization and structure refinement to stabilize de novo designed protein structures and complexes. FastDesign has been used previously to design novel protein folds and assemblies with important applications in research and medicine. To promote sampling of alternative conformations and sequences, FastDesign includes stages where the energy landscape is smoothened by reducing repulsive forces. Here, we discover that this process disfavors larger amino acids in the protein core because the protein compresses in the early stages of refinement. By testing alternative ramping strategies for the repulsive weight, we arrive at a scheme that produces lower energy designs with more native-like sequence composition in the protein core. We further validate the protocol by designing and experimentally characterizing over 4000 proteins and show that the new protocol produces higher stability proteins.

Automated design of thousands of nonrepetitive parts for engineering stable genetic systems.

Engineered genetic systems are prone to failure when their genetic parts contain repetitive sequences. Designing many nonrepetitive genetic parts with desired functionalities remains a difficult challenge with high computational complexity. To overcome this challenge, we developed the Nonrepetitive Parts Calculator to rapidly generate thousands of highly nonrepetitive genetic parts from specified design constraints, including promoters, ribosome-binding sites and terminators. As a demonstration, we designed and experimentally characterized 4,350 nonrepetitive bacterial promoters with transcription rates that varied across a 820,000-fold range, and 1,722 highly nonrepetitive yeast promoters with transcription rates that varied across a 25,000-fold range. We applied machine learning to explain how specific interactions controlled the promoters' transcription rates. We also show that using nonrepetitive genetic parts substantially reduces homologous recombination, resulting in greater genetic stability.

Cell cycle entry triggers a switch between two modes of Cdc42 activation during yeast polarization.

Cell polarization underlies many cellular and organismal functions. The GTPase Cdc42 orchestrates polarization in many contexts. In budding yeast, polarization is associated with a focus of Cdc42•GTP which is thought to self sustain by recruiting a complex containing Cla4, a Cdc42-binding effector, Bem1, a scaffold, and Cdc24, a Cdc42 GEF. Using optogenetics, we probe yeast polarization and find that local recruitment of Cdc24 or Bem1 is sufficient to induce polarization by triggering self-sustaining Cdc42 activity. However, the response to these perturbations depends on the recruited molecule, the cell cycle stage, and existing polarization sites. Before cell cycle entry, recruitment of Cdc24, but not Bem1, induces a metastable pool of Cdc42 that is sustained by positive feedback. Upon Cdk1 activation, recruitment of either Cdc24 or Bem1 creates a stable site of polarization that induces budding and inhibits formation of competing sites. Local perturbations have therefore revealed unexpected features of polarity establishment.

Optical Control of Peroxisomal Trafficking.

The blue-light-responsive LOV2 domain of Avena sativa phototropin1 (AsLOV2) has been used to regulate activity and binding of diverse protein targets with light. Here, we used AsLOV2 to photocage a peroxisomal targeting sequence, allowing light regulation of peroxisomal protein import. We generated a protein tag, LOV-PTS1, that can be appended to proteins of interest to direct their import to the peroxisome with light. This method provides a means to inducibly trigger peroxisomal protein trafficking in specific cells at user-defined times.

Benchmarking of optical dimerizer systems.

Optical dimerizers are a powerful new class of optogenetic tools that allow light-inducible control of protein-protein interactions. Such tools have been useful for regulating cellular pathways and processes with high spatiotemporal resolution in live cells, and a growing number of dimerizer systems are available. As these systems have been characterized by different groups using different methods, it has been difficult for users to compare their properties. Here, we set about to systematically benchmark the properties of four optical dimerizer systems, CRY2/CIB1, TULIPs, phyB/PIF3, and phyB/PIF6. Using a yeast transcriptional assay, we find significant differences in light sensitivity and fold-activation levels between the red light regulated systems but similar responses between the CRY2/CIB and TULIP systems. Further comparison of the ability of the CRY2/CIB1 and TULIP systems to regulate a yeast MAPK signaling pathway also showed similar responses, with slightly less background activity in the dark observed with CRY2/CIB. In the process of developing this work, we also generated an improved blue-light-regulated transcriptional system using CRY2/CIB in yeast. In addition, we demonstrate successful application of the CRY2/CIB dimerizers using a membrane-tethered CRY2, which may allow for better local control of protein interactions. Taken together, this work allows for a better understanding of the capacities of these different dimerization systems and demonstrates new uses of these dimerizers to control signaling and transcription in yeast.

TULIPs: tunable, light-controlled interacting protein tags for cell biology.

Naturally photoswitchable proteins offer a means of directly manipulating the formation of protein complexes that drive a diversity of cellular processes. We developed tunable light-inducible dimerization tags (TULIPs) based on a synthetic interaction between the LOV2 domain of Avena sativa phototropin 1 (AsLOV2) and an engineered PDZ domain (ePDZ). TULIPs can recruit proteins to diverse structures in living yeast and mammalian cells, either globally or with precise spatial control using a steerable laser. The equilibrium binding and kinetic parameters of the interaction are tunable by mutation, making TULIPs readily adaptable to signaling pathways with varying sensitivities and response times. We demonstrate the utility of TULIPs by conferring light sensitivity to functionally distinct components of the yeast mating pathway and by directing the site of cell polarization.

Actin behavior in bulk cytoplasm is cell cycle regulated in early vertebrate embryos.

The mechanical properties of cells change as they proceed through the cell cycle, primarily owing to regulation of actin and myosin II. Most models for cell mechanics focus on actomyosin in the cortex and ignore possible roles in bulk cytoplasm. We explored cell cycle regulation of bulk cytoplasmic actomyosin in Xenopus egg extracts, which is almost undiluted cytoplasm from unfertilized eggs. We observed dramatic gelation-contraction of actomyosin in mitotic (M phase) extract where Cdk1 activity is high, but not in interphase (I-phase) extract. In spread droplets, M-phase extract exhibited regular, periodic pulses of gelation-contraction a few minutes apart that continued for many minutes. Comparing actin nucleation, disassembly and myosin II activity between M-phase and I-phase extracts, we conclude that regulation of nucleation is likely to be the most important for cell cycle regulation. We then imaged F-actin in early zebrafish blastomeres using a GFP-Utrophin probe. Polymerization in bulk cytoplasm around vesicles increased dramatically during mitosis, consistent with enhanced nucleation. We conclude that F-actin polymerization in bulk cytoplasm is cell cycle regulated in early vertebrate embryos and discuss possible biological functions of this regulation.

Rationally improving LOV domain-based photoswitches.

Genetically encoded protein photosensors are promising tools for engineering optical control of cellular behavior; we are only beginning to understand how to couple these light detectors to effectors of choice. Here we report a method that increases the dynamic range of an artificial photoswitch based on the LOV2 domain of Avena sativa phototropin 1 (AsLOV2). This approach can potentially be used to improve many AsLOV2-based photoswitches.

Light-activated DNA binding in a designed allosteric protein.

An understanding of how allostery, the conformational coupling of distant functional sites, arises in highly evolvable systems is of considerable interest in areas ranging from cell biology to protein design and signaling networks. We reasoned that the rigidity and defined geometry of an alpha-helical domain linker would make it effective as a conduit for allosteric signals. To test this idea, we rationally designed 12 fusions between the naturally photoactive LOV2 domain from Avena sativa phototropin 1 and the Escherichia coli trp repressor. When illuminated, one of the fusions selectively binds operator DNA and protects it from nuclease digestion. The ready success of our rational design strategy suggests that the helical "allosteric lever arm" is a general scheme for coupling the function of two proteins.