Page 11234..»

Category : Protein Folding

How DeepMind is unlocking the secrets of dopamine and protein folding with AI – VentureBeat

Demis Hassabis founded DeepMind with the goal of unlocking answers to some of the worlds toughest questions by recreating intelligence itself. His ambition remains just that an ambition but Hassabis and colleagues inched closer to realizing it this week with the publication of papers in Natureaddressing two formidable challenges in biomedicine.

The first paper originated from DeepMinds neuroscience team, and it advances the notion that an AI research development might serve as a framework for understanding how the brain learns. The other paper focuses on DeepMinds work with respect to protein folding work which it detailed in December 2018. Both follow on the heels of DeepMinds work in applying AI to the prediction of acute kidney injury, or AKI, and to challenging game environments such as Go, shogi, chess, dozens of Atari games, and Activision Blizzards StarCraft II.

Its exciting to see how our research in [machine learning] can point to a new understanding of the learning mechanisms at play in the brain, said Hassabis. [Separately, understanding] how proteins fold is a long-standing fundamental scientific question that could one day be key to unlocking new treatments for a whole range of diseases from Alzheimers and Parkinsons to cystic fibrosis and Huntingtons where misfolded proteins are believed to play a role.

In the paper on dopamine, teams hailing from DeepMind and Harvard investigated whether the brain represents possible future rewards not as a single average but as a probability distribution a mathematical function that provides the probabilities of occurrence of different outcomes. They found evidence of distributional reinforcement learning in recordings taken from the ventral tegmental area the midbrain structure that governs the release of dopamine to the limbic and cortical areas in mice. The evidence indicates that reward predictions are represented by multiple future outcomes simultaneously and in parallel.

The idea that AI systems mimic human biology isnt new. A study conducted by researchers at Radboud University in the Netherlands found that recurrent neural networks (RNNs) can predict how the human brain processes sensory information, particularly visual stimuli. But, for the most part, those discoveries have informed machine learning rather than neuroscientific research.

In 2017, DeepMind built an anatomical model of the human brain with an AI algorithm that mimicked the behavior of the prefrontal cortex and a memory network that played the role of the hippocampus, resulting in a system that significantly outperformed most machine learning model architectures. More recently, DeepMind turned its attention to rational machinery, producing synthetic neural networks capable of applying humanlike reasoning skills and logic to problem-solving. And in 2018, DeepMind researchers conducted an experiment suggesting that the prefrontal cortex doesnt rely on synaptic weight changes to learn rule structures, as once thought, but instead uses abstract model-based information directly encoded in dopamine.

Reinforcement learning involves algorithms that learn behaviors using only rewards and punishments as teaching signals. The rewards serve to reinforce whatever behaviors led to their acquisition, more or less.

As the researchers point out, solving a problem requires understanding how current actions result in future rewards. Thats where temporal difference learning (TD) algorithms come in they attempt to predict the immediate reward and their own reward prediction at the next moment in time. When this comes in bearing more information, the algorithms compare the new prediction against what it was expected to be. If the two are different, this temporal difference is used to adjust the old prediction toward the new prediction so that the chain becomes more accurate.

Above: When the future is uncertain, future reward can be represented as a probabilitydistribution. Some possible futures are good (teal), others are bad (red).

Image Credit: DeepMind

Reinforcement learning techniques have been refined over time to bolster the efficiency of training, and one of the recently developed techniques is called distributional reinforcement learning.

The amount of future reward that will result from a particular action is often not a known quantity, but instead involves some randomness. In such situations, a standard TD algorithm learns to predict the future reward that will be received on average, while a distributional reinforcement algorithm predicts the full spectrum of rewards.

Its not unlike how dopamine neurons function in the brains of animals. Some neurons represent reward prediction errors, meaning they fire i.e., send electrical signals upon receiving more or less reward than expected. Its called the reward prediction error theory a reward prediction error is calculated, broadcast to the brain via dopamine signal, and used to drive learning.

Above: Each row of dots corresponds to adopamine cell, and each color corresponds to a different reward size.

Image Credit: DeepMind

Distributional reinforcement learning expands upon the canonical reward prediction error theory of dopamine. It was previously thought that reward predictions were represented only as a single quantity, supporting learning about the mean or average of stochastic (i.e., randomly determined) outcomes, but the work suggests that the brain in fact considers a multiplicity of predictions. In the brain, reinforcement learning is driven by dopamine, said DeepMind research scientist Zeb Kurth-Nelson. What we found in our paper is that each dopamine cell is specially tuned in a way that makes the population of cells exquisitely effective at rewiring those neural networks in a way that hadnt been considered before.

One of the simplest distributional reinforcement algorithms distributional TD assumes that reward-based learning is driven by a reward prediction error that signals the difference between received and anticipated rewards. As opposed to traditional reinforcement learning, however, where the prediction is represented as a single quantity the average over all potential outcomes weighted by their probabilities distributional reinforcement uses several predictions that vary in their degree of optimism about upcoming rewards.

A distributional TD algorithm learns this set of predictions by computing a prediction error describing the difference between consecutive predictions. A collection of predictors within apply different transformations to their respective reward prediction errors, such that some predictors selectively amplify or overweight their reward errors. When the reward prediction error is positive, some predictors learn a more optimistic reward corresponding to a higher part of the distribution, and when the reward prediction is negative, they learn more pessimistic predictions. This results in a diversity of pessimistic or optimistic value estimates that capture the full distribution of rewards.

Above: As a population, dopamine cells encode the shape of the learned reward distribution:We can decode the distribution of rewards from their firing rates. The gray shaded area is the true distribution of rewards encountered in the task.

Image Credit: DeepMind

For the last three decades, our best models of reinforcement learning in AI have focused almost entirely on learning to predict the average future reward. But this doesnt reflect real life, said DeepMind research scientist Will Dabney. [It is in fact possible] to predict the entire distribution of rewarding outcomes moment to moment.

Distributional reinforcement learning is simple in its execution, but its highly effective when used with machine learning systems its able to increase performance by a factor of two or more. Thats perhaps because learning about the distribution of rewards gives the system a more powerful signal for shaping its representation, making it more robust to changes in the environment or a given policy.

The study, then, sought to determine whether the brain uses a form of distributional TD. The team analyzed recordings of dopamine cells in 11 mice that were made while the mice performed a task for which they received stimuli. Five mice were trained on a variable-probability task, while six were trained on a variable-magnitude task. The first group was exposed to one of four randomized odors followed by a squirt of water, an air puff, or nothing. (The first odor signaled a 90% chance of reward, while the second, third, and fourth odors signaled a 50% chance of reward, 10% chance of reward, and 90% chance of reward, respectively.)

Dopamine cells change their firing rate to indicate a prediction error, meaning there should be zero prediction error when a reward is received thats the exact size a cell predicted. With that in mind, the researchers determined the reversal point for each cell the reward size for which a dopamine cell didnt change its firing rate and compared them to see if there were any differences.

They found that some cells predicted large amounts of reward, while others predicted little reward, far beyond the differences that might be expected from variability. They again saw diversity after measuring the degree to which the different cells exhibited amplifications of positive versus negative expectations. And they observed that the same cells that amplified their positive prediction errors had higher reversal point, indicating they were tuned to expect higher reward volumes.

Above: Complex 3D shapes emerge from a string of amino acids.

Image Credit: DeepMind

In a final experiment, the researchers attempted to decode the reward distribution from the firing rates of the dopamine cells. They report success: By performing inference, they managed to reconstruct a distribution that was a match to the actual distribution of rewards in the task in which the mice were engaged.

As the work examines ideas that originated within AI, its tempting to focus on the flow of ideas from AI to neuroscience. However, we think the results are equally important for AI, said DeepMind director of neuroscience research Matt Botvinick. When were able to demonstrate that the brain employs algorithms like those we are using in our AI work, it bolsters our confidence that those algorithms will be useful in the long run that they will scale well to complex real-world problems and interface well with other computational processes. Theres a kind of validation involved: If the brain is doing it, its probably a good idea.

The second of the two papers details DeepMinds work in the area of protein folding, which began over two years ago. As the researchers note, the ability to predict a proteins shape is fundamental to understanding how it performs its function in the body. This has implications beyond health and could help with a number of social challenges, like managing pollutants and breaking down waste.

The recipe for proteins large molecules consisting of amino acids that are the fundamental building block of tissues, muscles, hair, enzymes, antibodies, and other essential parts of living organisms are encoded in DNA. Its these genetic definitions that circumscribe their three-dimensional structure, which in turn determines their capabilities. Antibody proteins are shaped like a Y, for example, enabling them to latch onto viruses and bacteria, while collagen proteins are shaped like cords, which transmit tension between cartilage, bones, skin, and ligaments.

But protein folding, which occurs in milliseconds, is notoriously difficult to determine from a corresponding genetic sequence alone. DNA contains only information about chains of amino acid residues and not those chains final form. In fact, scientists estimate that because of the incalculable number of interactions between the amino acids, it would take longer than 13.8 billion years to figure out all the possible configurations of a typical protein before identifying the right structure (an observation known as Levinthals paradox).

Thats why instead of relying on conventional methods to predict protein structure, such as X-ray crystallography, nuclear magnetic resonance, and cryogenic electron microscopy, the DeepMind team pioneered a machine learning system dubbed AlphaFold. It predicts the distance between every pair of amino acids and the twisting angles between the connecting chemical bonds, which it combines into a score. A separate optimization step refines the score through gradient descent (a mathematical method of improving the structure to better match the predictions), using all distances in aggregate to estimate how close the proposed structure is to the right answer.

The most successful protein folding prediction approaches thus far have leveraged whats known as fragment assembly, where a structure is created through a sampling process that minimizes a statistical potential derived from structures in the Protein Data Bank. (As its name implies, the Protein Data Bank is an open source repository of information about the 3D structures of proteins, nucleic acids, and other complex assemblies.) In fragment assembly, a structure hypothesis is modified repeatedly, typically by changing the shape of a short section while retaining changes that lower the potential, ultimately leading to low potential structures.

With AlphaFold, DeepMinds research team focused on the problem of modeling target shapes from scratch without drawing on solved proteins as templates. Using the aforementioned scoring functions, they searched the protein landscape to find structures that matched their predictions and replaced pieces of the protein structure with new protein fragments. They also trained a generative system to invent new fragments, which they used along with gradient descent optimization to improve the score of the structure.

The models trained on structures extracted from the Protein Data Bank across 31,247 domains, which were split into train and test sets comprising 29,427 and 1,820 proteins, respectively. (The results in the paper reflect a test subset containing 377 domains.) Training was split across eight graphics cards, and it took about five days to complete 600,000 steps.

The fully trained networks predicted the distance of every pair of amino acids from the genetic sequences it took as its input. A sequence with 900 amino acids translated to about 400,000 predictions.

Above: The top figure features the distance matrices for three proteins, where the brightness of each pixel represents the distance between the amino acids in the sequence comprising the protein. The bottom row shows the average of AlphaFolds predicted distancedistributions.

Image Credit: DeepMind

AlphaFold participated in the December 2018 Critical Assessment of protein Structure Prediction competition (CASP13), a competition that has been held every every two years since 1994 and offers groups an opportunity to test and validate their protein folding methods. Predictions are assessed on protein structures that have been solved experimentally but whose structures have not been published, demonstrating whether methods generalize to new proteins.

AlphaFold won the 2018 CASP13 by predicting the most accurate structure for 24 out of 43 proteins. DeepMind contributed five submissions chosen from eight structures produced by three different variations of the system, all of which used potentials based on the AI model distance predictions, and some of which tapped structures generated by the gradient descent system. DeepMind reports that AlphaFold performed particularly well in the free modeling category, creating models where no similar template exists. In point of fact, it achieved a summed z-score a measure of how well systems perform against the average of 52.8 in this category, ahead of 36.6 for the next-best model.

The 3D structure of a protein is probably the single most useful piece of information scientists can obtain to help understand what the protein does and how it works in cells, wrote head of the UCL bioinformatics group David Jones, who advised the DeepMind team on parts of the project. Experimental techniques to determine protein structures are time-consuming and expensive, so theres a huge demand for better computer algorithms to calculate the structures of proteins directly from the gene sequences which encode them, and DeepMinds work on applying AI to this long-standing problem in molecular biology is a definite advance. One eventual goal will be to determine accurate structures for every human protein, which could ultimately lead to new discoveries in molecular medicine.

Read the original post:
How DeepMind is unlocking the secrets of dopamine and protein folding with AI - VentureBeat

Recommendation and review posted by Alexandra Lee Anderson

The Importance of Understanding TargetProtein Interactions in Drug Discovery – Technology Networks

Youre unwell, you see a doctor, they prescribe you a medicine and you take it. But how exactly is that drug having an effect? What is its mechanism of action? Drugs exhibit their effects through specific protein-target interactions.

But in some cases, there may not be a treatment available. In approximately 30% of cases, drugs fail during clinical development, and toxicity which can be caused by off-target binding is often to blame.

Andrew Lynn, Chief Executive Officer at Fluidic Analytics discusses why understanding proteintarget interactions is so important, the common challenges researchers face when attempting to determine these interactions, and touches on the relationship between the drug "attrition rate" crisis and the off-target effects of drugs.

Laura Lansdowne (LL): Could you discuss the importance of understanding proteintarget interactions in drug discovery, and the implications of not knowing your target?Andrew Lynn (AL): Understanding proteintarget interactions is crucial we are talking about the difference between finding a lifesaving drug/therapy and wasting hundreds of millions of dollars developing a drug with the wrong mechanism of action.A recent paper from Jason Sheltzers group showed that ten anticancer drugs undergoing clinical trials had a completely different mechanism of action from the one originally attributed to them. Briefly, when the protein targeted by each of the drugs was removed from cancer cells, the group expected the drugs to stop working. But what they found was that the drugs continued to work as normal and thus had to be working through off-target binding.This is crucial because it means potentially there are many more drugs out there that are working through off-target binding; it also means that many other drug candidates that have previously been disregarded may have unrecognized promise. This problem is about to become even more acute as research expands into conditions with difficult targets like Alzheimer's disease.The way in which we discover the exact mechanism of action between proteins and potential drug candidates needs better technologies for characterizing on-target and off-target interactions We cannot discover new information relying solely on technologies that have fallen short for decades.LL: What challenges do drug discovery researchers face when trying to identify targetprotein interactions?AL: Drug discovery and development is a lengthy, complex and costly process with a high degree of uncertainty whether a drug will succeed. The two biggest challenges are: First, not understanding the pathophysiology of many disorders, such as neurodegenerative disorders, which makes target identification challenging. Second, the lack of validated diagnostic and therapeutic biomarkers to objectively detect and measure biological states.At the heart of both challenges is the ability to characterize protein-drug target interactions. Unfortunately, the methods currently employed by researchers to do this research are outdated.

An example of this can be seen when scientists try to characterize interactions involving intrinsically disordered proteins (IDPs) such as the ones associated with Parkinsons disease. Current characterization methods modify proteins by fixing them to a surface or putting them in artificial environments. So, its no surprise that many drugs are great at targeting proteins with these modifications but poor at targeting these same proteins as they exist in vivo in solution and not tethered to an artificial surface.

This is why were building new tools and methods for researchers to more accurately characterize binding events in solution: to better understand how drugs interact with their protein targets in their native environment.

LL: What is microfluidic diffusional sizing and how can this be used to measure the binding affinity of proteinprotein interactions?AL: Microfluidic diffusional sizing (MDS) characterizes proteins and their interactions in solution based on the size (or more specifically hydrodynamic radius) of proteins and protein complexes as they diffuse within a microfluidic laminar flow. Characterizing in solution avoids artefacts from surfaces or matrices; gathering information about size to give crucial insights into stoichiometry, on- and off-target binding, oligomerization and folding.

MDS can be used to measure binding affinity by tracking changes in the size of a protein as it binds at different concentrations. The size of the complex can also give a strong indication of whether the protein is forming a protein-target complex at the expected size (on-target binding) or something with a completely different or unexpected size (off-target binding). A major additional advantage of MDS is that, because of the absence of surfaces or matrices, it can be used to characterize binding involving difficult targets such as intrinsically disordered proteins and membrane proteins.

LL: Could you discuss the relationship between the drug "attrition rate" crisis and the off-target effects of drugs?AL: Compound failure rates due to toxicity before human testing is very high. A recent review from a top-20 pharma company cited toxicity as the reason why, between 2005-2010, 82% of drugs were rejected at the preclinical stage and 35% in phase 2a. Overall, concerns surrounding toxicity account for as much as 30% of drug attrition occurring during the clinical stage of development.For many potential drugs, toxicity is due to off-target binding. By employing new methods to characterize drug candidates binding to protein targets in native conditions, we can identify off-target binding more effectively. This could help save billions of dollars in development costs and reduce the attrition rate we are currently facing.

LL: There has currently been very limited success in the development of effective therapies for Alzheimers disease (AD). Could you touch on some of the successes and highlight the molecules of interest in AD as well as the challenges related to their study.AL: One recent success is the anti-amyloid drug, aducanumab. After Biogen re-examined the data from the clinical trials, they found that exposure to high doses of Aducanumab reduced clinical decline in patients exhibiting early stages of Alzheimers disease.If approved, aducanumab would become the first therapy to slow the cognitive decline that accompanies Alzheimer's disease. This a massive step forward and a much-needed source of hope for patients and their families.But aducanumab doesnt cure Alzheimers disease. A major challenge impeding the development of further AD drugs is the ability to understand the mechanism of action via which candidate drugs interact with targets. Amyloid- is known to be a particularly difficult-to-characterize peptide, and even aducanumab doesnt have a well-understood mechanism of action. Any breakthroughs in being able to characterize how it or other Alzheimers disease drugs interact with difficult targets would be a major breakthrough in drug development.However, the majority of Alzheimers patients do not carry the dominantly inherited genetic mutation for the disease, and we dont know why amyloid proteins aggregate within their brains.

It follows that there wont be a single cause but rather many causes. Thus, the common consensus is that there wont be a single miracle drug that cures Alzheimers disease for everyone.

Andrew Lynn was speaking with Laura Elizabeth Lansdowne, Senior Science Writer, Technology Networks.

See more here:
The Importance of Understanding TargetProtein Interactions in Drug Discovery - Technology Networks

Recommendation and review posted by Alexandra Lee Anderson

Gocycle to partner with nutrition brand Fuel10k to promote benefits of e-bikes – Bike Biz

Gocycle is set to partner with protein breakfast brand Fuel10k to increase awareness of how e-bikes can help more people to lead an active lifestyle.

Fuel10k will give away five fast-folding Gocycle GX electric bikes as part of its biggest-ever on-pack promotion between January and April.

The GX will feature on three million of the brands high-protein breakfast drinks and porridge pots in outlets nationwide.

Richard Thorpe, Gocycle designer and founder, said: We are really excited about the opportunity to spread the message of the enormous health benefits of e-bikes to millions of people across the UK. E-bikes are the perfect travel solution for people who want to lead a more active and sustainable lifestyle and above all they are fun!

This partnership is all about fuelling more people to lead a more active lifestyle in the long-term. E-bikes are a great way to get back out onto two wheels. Having the electrical assistance on tap removes many of the daunting elements of cycling and encourages more people to cycle more of the time which can only be a good thing.

Individuals can enter the competition by purchasing a Fuel10k breakfast drink or porridge pot that features a Gocycle on the packaging. They will be presented with a unique code which they can enter on Fuel10ks competition site to be in with a chance of winning a fast-folding Gocycle GX and other prizes such as sports T-shirts, water bottles or discount codes.

Scott Chassels, Fuel10k managing director, added: We are an increasingly time-poor society and everyone seems to be busier than ever, but that shouldnt be at the detriment of our health. Fuel10k exists to give people a better for you, protein-based, breakfast on-the-go, which helps them to maximise the precious little time they have in the morning and fuel their active day ahead.

We are really excited by this partnership as e-bikes can really enhance the lifestyles of busy people by helping them to have a healthier, more sustainable and speedier commute.

The fast-folding Gocycle GX is available to order now online and through select resellers throughout US, Canada, UK, and EU.

Read more:
Gocycle to partner with nutrition brand Fuel10k to promote benefits of e-bikes - Bike Biz

Recommendation and review posted by Alexandra Lee Anderson

Structure of Drosophila melanogaster ARC1 reveals a repurposed molecule with characteristics of retroviral Gag – Science Advances


Activity-regulated cytoskeleton-associated protein (ARC) is an immediate early gene product induced in response to high levels of synaptic activity and is directed to neuronal synapses through signaling sequences in its 3 untranslated region (1). Mammalian ARC (mam-ARC) is essential for neuronal plasticity and is involved in memory (2) acting as a regulator of AMPA receptors (AMPARs) (3, 4). ARC has also been implicated in neurological disorders, including Alzheimers disease (5), fragile X syndrome (6), and schizophrenia (7, 8). In Drosophila melanogaster, two homologs of mam-ARC are expressed: dARC1 and dARC2 (9). dARC1 is present at neuromuscular junctions and, along with its mRNA, has been implicated in regulating the behavioral starvation response but is not involved in synaptic plasticity (10). Therefore, comparing the structural and functional properties of mam-ARC and dARC1 might lead to a better understanding of cognition and memory consolidation.

The ARC gene is thought to be derived from the gag gene of a Ty3/Gypsy retrotransposon (11) that, subsequent to genomic insertion, has been repurposed to perform an advantageous function to the host (12). This connection between ARC and retrotransposons was made when sequence alignments revealed that the ARC proteins shared sequence similarity with the Gag protein of retroviruses or retrotransposons (11). These data also suggested that ARC is evolutionarily related to the Ty3/Gypsy family of retrotransposons. Further evidence came from crystal structures of two -helical domains from Rattus norvegicus ARC (rARC) (13), which revealed that rARC N- and C-terminal capsid (CA) domains were structurally homologous to the N- and C-terminal CA domains of both Orthoretrovirinae (13) and Spumaretrovirinae (14). Further phylogenetic analysis revealed that, despite mam-ARC and dARC1 seemingly providing related functions in the host, dARC1 and the tetrapod ARCs most likely arose from separate lineages of Ty3/Gypsy, because dARC1 clustered with insect Ty3/Gypsy retrotransposons and tetrapod ARCs clustered with fish Ty3/Gypsy retrotransposons (12).

The relevance of ARCs retrotransposon origin to its function in synaptic plasticity was not immediately obvious until the recent observation that mam-ARC and dARC1 can self-assemble into particles and package RNA for potential transfer between cells (9, 12), similarly to retrotransposons and retroviruses (15, 16). In D. melanogaster, it is proposed that dARC1 expressed at neuromuscular junction presynaptic boutons assembles into particles that encapsidate dARC1 mRNA. Loaded particles might then be packaged and released as extracellular vesicles for intercellular transfer to the postsynapse, where mRNA release and translation can take place (9, 12). Similarly, mam-ARC can also encapsidate ARC mRNA into particles, allowing transfer from donor to recipient neurons, where ARC mRNA can be translated (12).

Because both dARC1 and mam-ARC are able to form CA-like particles (9, 12), it seems likely that they share a degree of structural similarity. To date, crystal structures of the individual domains from rARC have been determined (13), along with the solution nuclear magnetic resonance (NMR) structure of the rARC CA (17). Here, we report two crystal structures of the entire CA region of dARC1 at 1.7 and 2.3 and consider these structures in comparison to those of rARC and retroviral CA. dARC1 comprises two -helical domains with a fold related to that observed in the CA-NTD and CA-CTD of orthoretroviral and spumaretroviral CA. However, we observe significant divergence in the NTD of dARC1, where an extended hydrophobic strand that packs against 1 and 3 of the core fold replaces the N-terminal hairpin and helix 1 found in orthoretroviral CAs. In the rARC structure, this hydrophobic strand is replaced by peptides from the binding partners Ca2+/calmodulin-dependent protein kinase 2A (CamK2A) and transmembrane AMPAR regulatory protein 2 (TARP2) and may represent a functional adaptation for the recruitment of partner proteins. We also show that dARC1 uses the same CTD-CTD interface required for the assembly of retroviral CA into mature particles and propose that this obligate dimer represents a building block for dARC1 particle assembly. Further examination of the relationship between dARC1, mam-ARC, and Gag from Ty retrotransposon families reveals that, although dARC1 and mam-ARC are functional orthologs, the structural divergence in dARC1 and mam-ARC CA domains is consistent with the notion of Ty3/Gypsy Gag exaptation on two separate occasions. We suggest that they may have undergone different adaptations after appropriation into the tetrapod and insect genomes.

We determined the crystal structure of the CA domain region of dARC1, residues S39 to N205 (dARC1 CA), using single-wavelength anomalous diffraction (SAD) and crystals of Se-Met substituted protein. The structure was determined in both an orthorhombic and a hexagonal crystal form. The orthorhombic crystals diffracted to higher resolution, allowing the structure to be refined to a final resolution of 1.7 with an R factor of 18.1% and a free R factor of 21.3%. Details of data collection, phasing, and refinement are presented in table S1. The asymmetric unit (ASU) contains two chains, each containing an -helical N-terminal (CA-NTD) and C-terminal domain (CA-CTD) (Fig. 1A). The chains are arranged in a dimer with a distinct U-shape reminiscent of a glacial trough (Fig. 1A, right). The CA-CTDs form the base of the trough and pack together to form a homodimer interface, and the CA-NTDs form the sides of the trough and are separated by ~45 . Inspection of each domain reveals that the CA-NTD is made up from an extended N-terminal strand and a four-helix core (1 to 4), and the CA-CTD comprises a further five -helix bundle (5 to 9) (Fig. 1B, i and ii). The tertiary folds of each domain are particularly similar and can be superimposed with a root mean square deviation (RMSD) of 2.2 over 49 C atoms (Fig. 1C). Moreover, it can be seen that the dARC1 N-terminal strand is topologically equivalent to 5 in the CTD, while NTD 1 to 4 are equivalent to CTD 6 to 9. This strong similarity of dARC1 CA domains provides further evidence for the notion that tandem domains of CA arose as the result of a gene duplication event (14). The hexagonal crystal form was independently solved and refined to a resolution of 2.3 and reveals an almost identical dimeric ASU that aligns with an RMSD of only 0.247 over 133 C pairs (fig. S1, A to C). Both structures appear especially stable around the CTD-mediated dimeric interface and, when aligned through their CTDs, show only small differences in the positioning of NTDs with respect to the CTDs (fig. S1D).

(A) Cartoon representation of the dARC1 CA dimer. The N-terminal extended strand and helices are numbered sequentially from the N terminus to the C terminus. Monomer A is colored cyan, and monomer B is colored wheat. The right-hand panel is a view at 90 relative to the left-hand panel. (B) Close-up cartoon representations of dARC1 CA-NTD (left) and dARC CA-CTD (right) showing the helical topology of each domain. (C) Three-dimensional (3D) C structural alignment of dARC1 CA-NTD (blue cartoon) with dARC1 CA-CTD (red cartoon), with secondary structure elements labeled.

The dARC1 CA-CTD monomer consists of a five-helix core comprising 5 (residues A125 to Q134), 6 (residues I143 to Q156), 7 (residues E164 to L171), 8 (residues I177 to H182), and 9 (residues F191 to N204). The dimer interface is located between CA-CTDs, where the outer surfaces of 5 and 7 pack against 5 and 7 of the opposing monomer (Fig. 2A). The homodimer interface encompasses 768 2 of the buried surface and is defined by numerous intermolecular interactions. The interface is largely hydrophobic with contributions from side-chain packing of the Y126, Y129, M130, F133, L170, F172, and L174 hydrophobic and aromatic residues that are exposed on 5 and 7 and form a continuous apolar network with Y129 and F133 at its center (Fig. 2A, left). This is apparent in the analysis of the dARC1 CA surface hydrophobicity profile, which reveals a distinct apolar patch that locates to the center of the CA-CTD homodimer interface (fig. S2A). In addition, at the periphery of the interface, there is also a salt bridge between R161 on the 6-7 connecting loop with D169 at the C terminus of 7, providing further stabilization (Fig. 2A, right). The number and hydrophobic nature of interactions within the homodimer interface suggest that the dimer constitutes a relatively stable or obligate structure.

(A) Cartoon representation of dARC1 CA-CTD dimer. Helices are numbered sequentially from the N terminus to the C terminus. Monomer A is colored cyan, and monomer B is colored wheat. The right-hand panel is a view at 180 relative to the left-hand panel. Insets: Close-up views of molecular details of interactions at the dARC1 dimer interface. Residues that make interactions are shown in stick representation colored by atom type. Salt-bridge interactions between R161 and D169 are shown as dashed lines. (B) SEC-MALLS analysis of dARC1 CA. The sample loading concentrations were 400 M (8 mg/ml) (red), 200 M (4 mg/ml) (orange), 100 M (2 mg/ml) (yellow), 50 M (1 mg/ml) (green), and 25 M (0.5 mg/ml) (blue). The differential refractive index is plotted against column retention time, and the molar mass, determined at 1-s intervals throughout the elution of each peak, is plotted as points. The dARC1 CA monomer and dimer molecular mass are indicated with the gray dashed lines. (C) C(S) distributions derived from sedimentation velocity data recorded from dARC1 CA at 25 M (blue), 50 M (green), and 100 M (red). The curves represent the distribution of the sedimentation coefficients that best fit the sedimentation data (/0 = 1.41). (D) Multispeed sedimentation equilibrium profile determined from interference data collected on dARC1 CA at 70 M. Data were recorded at the three speeds indicated. The solid lines represent the global best fit to the data using a single-species model (Mw = 38.9 1 kDa). The lower panel shows the residuals to the fit.

Given the unexpected nature of the dimer observed in the crystal structure, the solution molecular mass, conformation, and self-association properties of dARC1 CA were examined using a variety of solution hydrodynamic methods. Initial assessment by size exclusion chromatographycoupled multi-angle laser light scattering (SEC-MALLS) was performed with protein concentrations ranging from 25 to 400 M that yielded an invariant solution molecular weight of 40.0 kDa for dARC1 CA (Fig. 2B). By comparison, the dARC1 CA sequence-derived molecular weight is 19.6 kDa. Given this value, together with the lack of a concentration dependency of the molecular weight, it is apparent that dARC1 CA also forms strong dimers in solution. To confirm and better analyze dARC1 CA oligomerization, we measured the hydrodynamic properties using sedimentation velocity (SV-AUC) and sedimentation equilibrium (SE-AUC) analytical ultracentrifugation. A summary of the experimental parameters, molecular weights derived from these data, and statistics relating to the quality of fits are shown in table S2. Analysis of the sedimentation velocity data for dARC1 CA using both discrete component and the C(S) continuous sedimentation coefficient distribution function (Fig. 2C) revealed a predominant single species with S20,w of 2.92 0.03 S and no significant concentration dependency of the sedimentation coefficient over the range measured (25 to 90 M). These data show that dARC1 CA comprises a single stable 2.92 S species with a molecular weight derived from either the C(S) function or discrete component analysis (S20,w/D20,w) of 38 kDa (table S2), consistent with a dARC1 CA dimer. The frictional ratio (f/fo) obtained from the analysis of the sedimentation coefficients is 1.41 (table S2), suggesting that the solution dimer has an elongated conformation and is consistent with the U-shaped conformation observed in the crystal structures. Moreover, analysis of the crystal structure using HYDROpro (18) gives calculated S20,w and D20,w values in close agreement with that observed in solution (table S2), supporting the idea that the dimer observed in the crystal structures is wholly representative of the solution conformation. To further ascertain the affinity of dARC1 CA self-association, multispeed SE-AUC studies at varying protein concentration were carried out and typical equilibrium distributions for dARC1 CA are presented in Fig. 2D. Analysis of individual gradient profiles showed no concentration dependency of the molecular weight, and so, all the data were fitted globally with a single ideal molecular species model, producing a weight-averaged molecular weight of 38.9 kDa (table S2). The lack of any concentration dependency precludes any analysis of homodimer affinity but confirms that dARC1 CA forms a stable dimeric structure that has the expected properties of the dimer we observe in the crystal structure.

Attempts to mildly disrupt the central apolar network by introduction of an F133A mutation had no effect on dimerization when assessed by SEC-MALLS (fig. S2B). More aggressive mutations F133A + Y129A and F133A + R161A resulted in complete loss of protein solubility and an inability to purify the constructs, further suggesting that, in dARC1 CA, homodimerization is a requirement for protein folding/structural integrity and likely forms a key building block of dARC1 particle assembly. Analysis of the electrostatic surface potential of the dimeric structure reveals a differential distribution of charge, where the surface of the glacial trough has a net negative charge that spreads across both domains of each dARC1, and the underside where the C-terminus projects has a more positively charged character (fig. S2C), suggesting that, upon assembly, dARC1 particles would have a negatively charged exterior and a more positively charged interior where nucleic acid is contained.

Given that mam-ARC and dARC1 share functional similarities, we assessed the relationship between rARC and dARC1 by comparing the dARC1 structure with the individual domains from rARC. Overall, the alignments are excellent, reflecting the evolutionary relationship, but there are significant differences between dARC1 and rARC in both their NTDs and CTDs.

There are two crystal structures of the rARC NTD in complex with peptide ligands [Protein Data Bank (PDB): 4X3H and 4X3I] (13) and a recent solution NMR structure [6GSE; (17)] of the entire rARC CA domain that resolves the NTD in an apo form. Superficially, the dARC1 CA-NTD aligns well with all available structures of the rARC CA-NTD, with DALI Z scores of 8 to 10 and RMSDs between 1.5 and 1.9 (Fig. 3A).

(A) Left: 3D structural alignment of dARC1 CA-NTD (teal cartoon) and apo-rARC CA-NTD (PDB: 6GSE; lilac cartoon). Secondary structure elements are labeled. Circled are the ordered N-terminal strand of dARC1 and the disordered N-terminal strand of apo-rARC. Right: 3D structural alignment of dARC1 CA-NTD and the peptide-complex structures of rARC CA-NTDs (PDB: 4X3H and 4X3I). The protein backbones are shown in cartoon representation, colored according to the legend. Secondary structure elements are labeled. The arrow indicates the different positioning of the extended N-terminal strand between the dARC1 and rARC structures. (B, i to iv) Individual views of the structures presented in (A): (i) apo-dARC1, (ii) apo-rARC, (iii) rARC-TARP2, and (iv) rARC-CaMK2B. Residues that constitute the hydrophobic NTD cleft are shown in stick format, colored by atom type. In each structure, the side chains of the aromatic residues buried in the interface (F45 and F52, dARC1 CA-NTD; Y229*, rARC CA-NTDTARP2; F313*, rARC CA-NTDCaMK2B) are colored purple, yellow, and orange, respectively. The conserved main-chain hydrogen bonding interactions between the backbone amide and carbonyl of F52 with the carbonyl of L89 and the amide of Y91 (dARC1), of Y229 with the carbonyl of H245 and the amide of N247 (rARC CA-NTDTARP2), and of F313 (rARC CA-NTDCaMK2B) with the carbonyl of H245 and the amide of N247 are shown as dashed lines.

Examination of the dARC1 CA-NTD reveals an N-terminal extended strand (NT-strand), residues G43 to R56, with a short configuration that packs against the core of the NTD. The NT-strand makes many interactions with the apolar and aromatic side chains that extend from 1, 2, and 4, burying 803 2 of surface in the interface [Fig. 3, A and B (i), and fig. S3A], and the same configuration is observed in all four instances of the NTDs that we see in our two crystal structures (fig. S3B). The NT-strand residues are highly conserved in dARC genes across Drosophilidae but not with the mam-ARCs (fig. S3C). In particular, two highly conserved aromatic residues, F45 and F52, are entirely buried, surrounded by the conserved side chains of F64, L89, I115, and F119, and act to anchor the NT-strand into the hydrophobic 1-to-4 cleft of the CA-NTD. In addition, there is a main-chain interaction between the backbone amide and carbonyl of F52 with the carbonyl of L89 and the amide of Y91 that further stabilizes the conformation of the NT-strand [Fig. 3B (i) and fig. S3A].

In apo-rARC CA-NTD (6GSE), the helical core aligns very well with the corresponding region of dARC1 (RMSD = 1.45 ). However, here, the rARC NT-strand residues D210 to E216 have a disordered conformation (Fig. 3, A and B, ii), and the 1-to-4 hydrophobic cleft, which in dARC1 contains the native NT-strand, is unoccupied in rARC, suggesting that there is a functional divergence for the NT-strand between the dARC1 and mam-ARC families. This notion is supported by the inspection of the rARC CA-NTDTARP2 and CA NTDCaMK2B complexes (4X3H and 4X3I), where the 1-to-4 cleft of rARC is now occupied by the bound TARP2- or CaMK2B-derived peptides (Fig. 3B, iii and iv), and the bound peptides adopt the same extended configuration as the native NT-strand in the dARC1 structure (fig. S3D) and bury a comparable amount of surface, 772 and 641 , respectively. Moreover, both bound peptides contain an aromatic residue equivalent to dARC1 F52, Y229 in TARP2, and F313 in CaMK2B that packs into the core of rARC CA-NTD and makes an identical main-chain interaction with the backbone carbonyl of H245 and the amide of N247 as that observed between the backbone amide and carbonyl of F52 with the carbonyl of L89 and the amide of Y91 in dARC1 (fig. S3D). In these peptide-complex structures, the rARC NT-strand, D210 to E216, that is disordered in the apo structure now adopts a parallel configuration to pack against the bound peptides (Fig. 3B, iii and iv), and it is possible that the propensity to form this stabilizing configuration has been selected for. This notion is supported by the inspection of the dARC and mam-Arc multiple sequence alignment (fig. S3C) that reveals a conserved TQIF motif in Amniota that retains -branched residues, favored in structure, at the T and I position. This motif is not present in amphibians or in Latimeria chalumnae Gypsy2, the closest known relative to the transposon from which tetrapod ARC was exapted, suggesting that this feature, and possibly peptide binding ability, arose within Amniota.

The structures of dARC1 CA-CTD and rARC CA-CTD (PDB: 4X3X) also superimpose well (RMSD = 2.7 ). However, the CTD of the apo-rARC CA NMR structure more closely matched the structure of dARC1 CA-CTD (RMSD = 2.2 ), with all five helices overlaying (Fig. 4A). However, in contrast to our solution studies of dARC1 (Fig. 2, A to C, and fig. S2A), the rARC CA domain was monomeric in solution, even at the high concentrations under which NMR was performed (17).

(A) 3D structural alignment of dARC1 CA-CTD and rARC CA-CTD from apo-rARC (PDB: 6GSE). The structures are shown in cartoon, with equivalent helices labeled and shown as cylinders. dARC1 is colored cyan, and rARC is colored light blue. (B and C) Details of the CA-CTD homodimer interfaces. Cartoon representations of the protein backbone of dARC1 CA-CTD (B) and rARC CA-CTD (C) are shown, colored as in (A). The view is of one monomer looking into the dimer interface. Residues that make interactions in dARC1 CA and their equivalents in rARC are shown in stick representation, color-coded by residue type (purple, hydrophobic/aromatic; green, polar; red, acidic; blue, basic). (D and E) Hydrophobic surface representations of (B) and (C), respectively. Circled in (D) is a distinct hydrophobic patch on the surface of dARC1 CA-CTD, which is absent in rARC. (F) Multiple sequence alignment of ARC, dARC1, and dARC2 CA-CTDs and parent retrotransposon sequences. Group 1 contains tetrapod ARC (tARC) sequences and the closely related Latimeria chalumnae (L. ch) Gypsy2 transposon. Top: Secondary structure of rARC; numbers according to the rARC (R. norvegicus) sequence. Group 2 contains dARC1, dARC2, and closely related Linepithema humile (L. h) Gypsy11 retrotransposon. Bottom: Secondary structure of dARC1; numbers according to the dARC1 (D. melanogaster) sequence. Red box and white text represent invariant residues shared between groups. Red text represents residues conserved within a group. Asterisks mark the residues at the dARC1 CTD dimer interface and their equivalents in tARCs, as shown in (B) and (C).

In dARC1, a large proportion of the CTD dimer interface results from the packing of hydrophobic side chains projecting from helices 5 and 7 (Fig. 2A). However, upon comparison of the external 5/7 surfaces of dARC1 and rARC (Fig. 4, B and C), it is apparent that the exposed Y126, Y129, M130, F133, L170, F172, and L174 side chains that are responsible for the hydrophobic character of the dARC1 dimer interface are not conserved in rARC and are replaced by E282, Q285, R286, D289, Y324, V326, and T328 in rARC. Therefore, the hydrophobic patch present on the surface of dARC1 is not evident in the same surface on rARC (Fig. 4, D and E). In addition, R161 and D169, which make a salt bridge interaction in the dARC1 interface, are also not conserved, being replaced by D315 and Q323 in rARC (Fig. 4, B and C). These sequence differences are also apparent throughout the entire dARC and mam-ARC families. Hence, there is strong sequence conservation of residues that constitute the core fold of the CA-CTD across both dARC and mam-ARCs, but the hydrophobic CA-CTD dimer interface residues are only present in the dARC lineage (Fig. 4F). Together, these data reveal that, while tertiary structure topology of dARC1 and rARC CA-CTDs is conserved, there are substantial differences in the character of the surface that is presented around 5 to 7; in dARC1, the hydrophobic nature of this surface drives the formation of a strong CTD dimer, whereas in rARC, the more polar nature of this surface may explain why the protein is monomeric in solution. Given these differences, although there is strong evidence for the assembly of both dARC1 and mam-ARC into CA-like particles (9, 12), it seems likely that if dARC1 and mam-ARC use the 5/7 interface in a particle assembly pathway, the interface may be substantially weaker for mam-ARC.

(A) Pairwise DALI 3D C structural alignment of dARC1 CA-NTD with HIV CA-CTD (left), RSV CA-CTD (middle), and HIV-NTD (right). In each panel, the cartoon of the dARC1 CA-NTD backbone is shown in blue, and the backbone of the aligned structures is shown in gray. (B) Pairwise 3D C structural alignment of dARC1 CA-CTD with HIV CA-CTD (left) and RSV CA-CTD (right). In each panel, the cartoon of the dARC1 CA-CTD backbone is shown in red, and the backbone of the aligned structures is shown in gray. (C) Pairwise 3D C structural alignment of dARC1 CA-NTD with prototypic foamy virus (PFV) CA-NTD (left) and dARC1 CA-CTD with PFV CA-CTD (right). (D) DALI Z scores, RMSD, number of aligned residues, and sequence identities for 3D C alignments.

The topology of the -helical two-domain fold of dARC1 is highly reminiscent of retroviral CA structures. Interrogation of the PDB database with dARC1 CA using the DALI alignment/search engine (19) produced an overwhelming number of matches to Gag proteins (87%, Z score 5.0) and identified rARC, together with many orthoretroviral and spumaretroviral CA-NTD and CA-CTD structures. Alignments with CA-NTDs and CA-CTDs from HIV CA, Rous sarcoma virus (RSV) CA, and prototypic foamy virus CA (PFV) are presented in Fig. 5. The best structural alignments to dARC1-NTD were with retroviral Gag CA-CTD structures rather than with Gag CA-NTD structures (Fig. 5, A and D), indicating that the dARC1 CA-NTD is more closely related to the orthoretroviral CA-CTD than to the orthoretroviral CA-NTD. Alignments with dARC1-CTD also had the best structural alignment with orthoretroviral Gag CA-CTD structures (Fig. 5, B and D), perhaps not unexpected given the observation of close resemblance of dARC1 CA-NTD to dARC1 CA-CTD (Fig. 1B, iii). Alignments with PFV CA-NTD and CA-CTD were also found (Fig. 5, C and D); although not as significant as with the orthoretroviral CA, these data support previous observations of a relationship of spumaretroviral Gag with mam-ARC (14).

These data provide evidence for a structural conservation between orthoretroviral CA and ARC proteins, and the weaker alignments observed with orthoretroviral CA-NTDs suggest that orthoretroviral CA-NTDs have undergone much more structural divergence than has occurred in the Ty3 family or ARC proteins. Moreover, these data further support the previously proposed idea that a duplication of a CA-CTD progenitor first gave rise to double domain ancestors and that subsequent divergence of domains resulted in spumaretroviral, orthoretroviral, and Metaviridae-derived proteins, such as ARC, that are found presently (14, 20).

Given the existence of the dARC1 CA dimer and the distant relationship with orthoretroviral CA, we next looked to see whether the dimer interface was conserved between dARC1 and the CTD dimers of HIV-1 CA and RSV CA that are known to be essential for CA assembly in orthoretroviruses. For these comparisons, the interhexamer CA CTD-CTD dimers observed in HIV-1 and RSV CA-hexamer crystal structures (21, 22) were used, as these most closely relate to those observed in cryo-electron microscopy (cEM) studies of whole CA assemblies (22, 23). Cartoon representations of the dARC1, HIV-1, and RSV CA-CTD dimers are shown in Fig. 6 (A to C). In each, the domain arrangement that presents the dimer interface is the same, and this is also seen in the CA-CTD dimer of native Ty3 particles visualized by cEM (24), but with some repositioning of the CA-NTDs (fig. S4). The structures have been aligned to find the best C alignment over the entire dimer (HIV, RMSD = 2.8 over 117 C; RSV, RMSD = 3.1 over 101 C) (Fig. 6, D and E), and it is apparent that each interface is made up from interactions between residues on CTD helices 5 and 7 of dARC1, which correspond to 7 and 8 in the orthoretroviral CA-CTD structures. Notably, in the orthoretroviruses, 7 is reduced to a single turn, and the monomers are rotated with respect to each other. Therefore, in dARC1, residues on 5 and 7 contribute equally to the interface, while in the orthoretroviruses, 8 contributes more to the interface than does 7. This combination of the larger contribution of 5 in dARC1, together with the rotation and displacement of CA-CTDs seen in the orthoretroviruses, has the effect of reducing the surface area that is buried at the interface from 768 2 in dARC1 to 452 2 in HIV-1. Notably, the homodimer affinity for orthoretroviral CA-CTD dimers is much weaker than the dARC1 dimer. Equilibrium dissociation constants ranging between 10 and 20 M have been reported for HIV-1 (25, 26), and CA-CTD dimerization is undetectable for other genera (2729). Nevertheless, given the domain organization and the similarity in character of the orthoretroviral and dARC1 CA-CTD dimers, we suggest that this interface is a key building block of CA assembly, retained in dARC1 and conserved from Ty3/Gypsy transposable elements to orthoretroviridae.

(A to C) Cartoon representations of CA-CTD dimers. (A) dARC1 is colored cyan and wheat. (B) HIV-1 is colored magenta and pale green (PDB: 2XFX). (C) RSV is colored gray and red (PDB: 3G21). The orthoretroviral structures are aligned with respect to the dARC1 dimer. CTD helices 5 to 9 are labeled in the dARC1 structure, and the equivalent 7 to 10 are labeled in the orthoretroviral structures. The buried surface area (2) and free energy of interaction (iG) of each interface, calculated in PDBePISA, are displayed below each structure. (D and E) Structural alignment of dARC1 CA with HIV-1 CA and RSV CA dimers, respectively. Protein backbones are colored as in (A) to (C).

Our crystal structures demonstrate that the central region of dARC1 contains two largely -helical domains that, despite the lack of sequence conservation, have the same predominantly -helical folds observed in the structures of CA domains from the ortho- and spumaretroviruses. A more detailed inspection of dARC1 CA-NTD and CA-CTD reveals that they comprise four- and five-helix bundles, respectively, with a topology that aligns well with the arrangement of secondary structure elements observed in orthoretroviral CA NTDs and CTDs (Fig. 5). However, it is apparent that both the ARC CA-NTD and CA-CTD are much more closely related to the orthoretroviral CA-CTDs than they are to orthoretroviral CA-NTDs (Fig. 5), consistent with our previous notion that an ancient domain duplication was a key event during retrotransposon evolution (14). Notably, orthoretroviral CA-NTDs contain an extra N-terminal hairpin and an additional two helices compared to the ARCs and the CA domains of Ty3/Gypsy transposons (fig. S4) (24). This suggests that unique aspects of the retroviral life cycle might be driving specific changes in the structure of the retroviral CA-NTD. One such pressure might be associated with the process of maturation that follows retrovirus budding from the cell. Maturation involves proteolytic cleavage of immature viral cores, followed by CA reassembly to yield mature virions and although it is proposed that dARC1 and mam-ARC transport mRNA between cells, it is thought likely that particles are packaged into extracellular vesicles for cell-to-cell transfer (9, 12). Similarly, maturation events do not occur in Ty3 elements, which also do not bud from the cell and have Gag that assembles directly into mature forms (24). The absence of maturation also characterizes spumaviruses, and it was observed previously that the CA NTDequivalent region of PFV Gag showed greater similarity to rARC than to orthoretroviral CA (14).

Our three-dimensional (3D) superimpositions have demonstrated that there is a large degree of structural conservation between the dARC1 and mam-ARC CA structures. However, despite this strong similarity, two regions of distinct differences between the dARC1 and rARC structure are apparent. The first region concerns the ARC CA-NTD and the interaction with potential binding partners; the second region concerns the putative dimerization domain of the CTD.

Functionally important interactions between mam-ARC and a variety of neuronal proteins, including the TARP2 and CaMK2B proteins, as well as the NMDA (N-methyl-d-aspartate) receptor, have been defined (13, 17). However, no such interactions have been reported for dARC1. In the rARC structures with bound TARP2 or CaMK2B peptides, the disordered N-terminal region of rARC seen in the apo structure now forms a short parallel sheet, with the bound peptide stabilizing the peptide binding within a hydrophobic cleft on rARC. It is apparent that the conformation of these rARC-bound peptides strongly resembles that of the NT-strand of dARC1 NTD (Fig. 3). Therefore, given the sequence differences in the NT-strand region between the dARC and mam-ARCs (fig. S3C), one notion is that mam-ARC has evolved an N-terminal strand that no longer binds into the CA-NTD hydrophobic cleft but has gained the ability to promote the binding of synaptic protein ligands, perhaps acting as a sensor of synaptic stimuli. This sensing property might then contribute control to a functional role for ARC based on assembly and mRNA trafficking.

There are also significant differences between dARC1 and rARC CA-CTD, illustrated in Fig. 4. Overall, our crystal structure of dARC1 and the NMR structure of full-length rARC (17) are very similar, with good overlay in all five helices. However, inspection of the dARC1 surface reveals a substantial hydrophobic patch that is absent in rARC (Fig. 4, D and E). This hydrophobic patch is shared with the orthoretroviruses (25, 30) and seems to be associated with the formation of stable dARC1 dimers, whereas rARC is monomeric. Whether this translates to differences in the stability of assembled particles in vivo remains to be determined; however, it is possible that differences in the physiological roles of dARC1 and mam-ARC may mean that mam-ARC has evolved to require a weaker interface that facilitates disassembly. Alternatively, it is possible that mam-ARC may require a conformational change to facilitate dimerization or uses a completely different assembly mechanism that uses other surfaces of the molecule.

The observation that residues at the dARC1 CA-CTD interface are not conserved between the insect and mam-ARC lineages suggests the possibility that, although mam-ARC particles have been observed in vitro and in cells, their mode of assembly may not use an obligate CA-CTD dimer as a building block. This type of observation has been made with orthoretroviruses that assemble through a combination of NTD-NTD, NTD-CTD, and CTD-CTD interactions to form the viral CA shell, where the relative contribution that different types of CA interaction make to the overall formation of the viral core varies depending on the retroviral genera. For instance, in lentiviruses, it is apparent that CA assembly requires a strong intrinsic CTD-CTD dimeric interaction (25, 30). However, more generally, CA shell formation requires three types of interaction: intrahexamer NTD-NTD self-association (3033), intrahexamer NTD-CTD interactions between adjacent CA monomers (30, 34, 35), and interhexamer CTD-CTD interactions (25, 30). Therefore, it is entirely possible that, in dARC1 and mam-ARC particles, the relative contributions of each type of interface may also differ.

Mam-ARC and dARC1 appear to have different biological properties. However, it remains to be determined whether these differences result from the capture of two different Ty3/Gypsy elements or they reflect evolutionary adaptations. Perhaps the best studied example of the appropriation of retroelement encoded genes by mammalian hosts is the case of syncytin, a fusagenic protein essential for proper placenta formation (36). It is evident that syncytin capture appears to have occurred on multiple independent occasions, involving envelope proteins from different retroviruses (37, 38), resulting in placentae with subtly different morphologies (39). Determining whether this is also the case with the ARC genes, as well as their close relatives in the mammalian genome (11), will require further characterization of existing retrotransposon elements using structural methods not reliant on the comparative similarities in related nucleic acid sequences that have disappeared with the passage of time.

dARC1 residues S39 to N205 were determined to represent the CA domain according to multiple sequence alignment and secondary structural analysis performed in ClustalX (40) and Psipred (41). An Escherichia coli codon-optimized complementary DNA (cDNA) for D. melanogaster dARC1 (UniProt, Q7K1U0) was synthesized (GeneArt), and the relevant sequence was polymerase chain reactionamplified and subcloned into a pET22b plasmid (Novagen). The resulting construct comprised residues 39 to 205 of dARC1, with an N-terminal Met and a C-terminal PLEHHHHHH His-tag extension. Proteins were expressed in E. coli strain BL21 (DE3) grown in LB broth by induction of log-phase cultures with 1 mM isopropyl--d-thiogalactopyranoside (IPTG) and incubated overnight at 20C. Cells were pelleted and resuspended in 50 mM tris-HCl, 150 mM NaCl, 10 mM imidazole, 5 mM MgCl2, and 1 mM dithiothreitol (pH 8.0), supplemented with lysozyme (1 mg/ml; Sigma-Aldrich), deoxyribonuclease (DNase) I (10 g/ml; Sigma-Aldrich), and one Protease Inhibitor cocktail tablet (EDTA-free, Pierce) per 40 ml of buffer. Cells were lysed using an EmulsiFlex-C5 homogenizer (Avestin), and dARC1 CA was captured from clarified lysate using immobilized metal ion affinity on a 5-ml Ni2+-NTA superflow column (Qiagen). Bound dARC1 CA was eluted in nonreducing buffer (50 mM tris-HCl, 150 mM NaCl, and 300 mM imidazole), and carboxypeptidase A (CPA; Sigma-Aldrich, C9268) was added at a ratio of ~100 mg of dARC1 per mg of CPA. The resulting mixture was incubated overnight at 4C to allow digestion of the C-terminal His-tag. The CPA was inactivated by the addition of TCEP-HCl [tris (2-carboxyethyl) phosphine hydrochloride] to 2 mM. dARC1 CA was further purified by size exclusion chromatography using a Superdex 75 (26/60) (GE Healthcare) column, equilibrated in 20 mM tris-HCl, 150 mM NaCl, and 1 mM TCEP (pH 8.0). Purified protein eluted in a single peak. Selenomethionine derivative protein was produced using an identical procedure, but with Methionine auxotroph E. coli B834 (DE3) cells, grown in selenomethionine medium (Molecular Dimensions, Newmarket, United Kingdom), used to express the protein. Electrospray-ionization mass spectrometry was used to confirm the identity of dARC1 and, where applicable, selenomethionine incorporation. It also confirmed that the N-terminal Met had been processed and that the His-tag had been completely digested, leaving the motif PLE at the C terminus. Protein was concentrated by centrifugal ultrafiltration (Vivaspin; molecular weight cutoff, 10 kDa), then snap-frozen, and stored at 80C. Protein concentrations were determined by ultraviolet-visible absorbance spectroscopy using an extinction coefficient at 280 nm derived from the tyrosine and tryptophan content.

dARC1 CA was crystallized using sitting drop vapor diffusion at 18C using Swissci MRC two-drop trays (Molecular Dimensions), with drops set using a Mosquito LCP robot with a humidity chamber (TTP Labtech). Native protein was initially concentrated to 20 mg/ml. Typically, drops were 200 to 300 nl, made by mixing protein:mother liquor in a 3:1 or 1:1 ratio, with a 75-l reservoir. Initial crystal hits were obtained using the Structure Screen 1&2 (Molecular Dimensions) under a condition containing 4.3 M NaCl and 0.1 M Hepes (pH 7.5). Two crystal forms could be observed in these conditions: thin rods, which had a primitive orthorhombic (oP) lattice, and hexagonal disks or trapezoidal prisms, which had a primitive hexagonal (hP) lattice. Datasets were collected for these native crystals, but they could not be solved by molecular replacement methods. SeMet dARC1 CA was crystallized under conditions that optimized protein concentration, NaCl concentration, and pH. The best crystals grew in 300- to 400-nl drops set with protein at 12.5 to 16 mg/ml, with mother liquor NaCl ranging between 2.8 and 3.3 M. Rods were ~400 m 30 m 30 m, and hexagons/trapezoids were ~130 m across and up to 30 m thick. Crystals were harvested using MiTeGen lithographic loops. The best cryoprotection was achieved using sodium malonate mixed into mother liquor to a concentration of 1.6 M. This was added directly to the drop, or crystals were bathed in this solution before flash freezing in liquid nitrogen.

Data were collected at the tunable SLS beamline PXIII. For the orthorhombic crystal form, a peak dataset was collected to 2.06 (see table S1). Data were processed by the SLS GoPy pipeline in P212121 using XDS (42) and showed significant anomalous signal to 2.82 . The resultant dataset was solved using SAD methods with Phenix (43), and despite a relatively low Figure of Merit (FOM), the experimental map was readily interpretable and it was possible to almost completely autobuild an initial structure with BUCCANEER (44). A higher-resolution (1.55 ) dataset was collected at a non-anomalous, low-energy remote wavelength (table S1). This dataset was processed using the Xia2 (45) pipeline, DIALS (46) for indexing and integration, and AIMLESS (47) for scaling and merging. This dataset was initially used for refinement to 1.7 and manual model building in COOT (48). It was evident that the data were anisotropic and that they might benefit from anisotropic correction. Diffraction images were reprocessed using the autoPROC pipeline (49), XDS, POINTLESS (50), AIMLESS, and STARANISO ( This dataset was used for further refinement of the model, and there was an improvement in map quality, and in agreement between model and data. For the hexagonal crystal form, a highly redundant peak dataset was collected to 2.14 . This was processed using the Xia2 pipeline, DIALS for indexing and integration, and AIMLESS for scaling and merging, showing significant anomalous signal to 2.59 , in P6122. This dataset was solved using SAD methods in Phenix. Again, the experimental map was readily interpretable, and it was possible to almost completely autobuild an initial structure with BUCCANEER. Refinement and model building were carried out in Phenix and COOT, respectively. Anomalous signal was very strong in this dataset, and so, Friedel pairs were treated separately during refinement. MolProbity (51) and PDB_REDO (52) were used to monitor and assess model geometry. Details of data collection, phasing, and structure refinement statistics are presented in table S1.

SEC-MALLS was used to determine the molar mass of dARC CA. Samples ranging from 25 to 400 M were applied in a volume of 100 l to a Superdex INCREASE 200 10/300 GL column equilibrated in 20 mM tris-HCl, 150 mM NaCl, 0.5 mM TCEP, and 3 mM NaN3 (pH 8.0) at a flow rate of 1.0 ml/min. The scattered light intensity and the protein concentration of the column eluate were recorded using a DAWN HELEOS laser photometer and an OPTILAB-rEX differential refractometer, respectively. The weight-averaged molecular mass of material contained in chromatographic peaks was determined from the combined data from both detectors using the ASTRA software version 6.0.3 (Wyatt Technology Corp., Santa Barbara, CA).

Sedimentation velocity experiments were performed in a Beckman Optima Xl-I analytical ultracentrifuge using conventional aluminum double-sector centerpieces and sapphire windows. Solvent density and the protein partial specific volumes were determined as described (53). Before centrifugation, dARC1 CA samples were prepared by exhaustive dialysis against the buffer blank solution, 20 mM tris-HCl (pH 8), 150 mM NaCl, and 0.5 mM TCEP (tris buffer). Samples (420 l) and buffer blanks (426 l) were loaded into the cells, and centrifugation was performed at 50,000 rpm and 293 K in an An50-Ti rotor. Interference data were acquired at time intervals of 180 s at varying sample concentrations (25, 50, and 100 M). Data recorded from moving boundaries were analyzed in terms of the size distribution functions C(S) using the program Sedfit (54).

Sedimentation equilibrium experiments were performed in a Beckman Optima XL-I analytical ultracentrifuge using aluminum double-sector centerpieces in an An-50 Ti rotor. Before centrifugation, samples were dialyzed exhaustively against the buffer blank (tris buffer). Samples (150 l) and buffer blanks (160 l) were loaded into the cells, and after centrifugation for 30 hours, interference data were collected at 2 hourly intervals until no further change in the profiles was observed. The rotor speed was then increased, and the procedure was repeated. Data were collected on samples of different concentrations of dARC1 CA (25, 50, and 70 M) at three speeds, and the program SEDPHAT (55) was used to determine weight-averaged molecular masses by nonlinear fitting of individual multispeed equilibrium profiles to a single-species ideal solution model. Inspection of these data revealed that the molecular mass of dARC1 CA showed no significant concentration dependency, and so, global fitting incorporating the data from multiple speeds and multiple sample concentrations was applied to extract a final weight-averaged molecular mass.

Amino acid alignments were produced with MAFFT v7.271 (57), within tcoffee v11.00.8cbe486 (58), weighting alignments using three-state secondary-structure predictions produced with RaptorX Property v1.02 (59). Alignment images were produced with ESPript (60).

Acknowledgments: We thank the Swiss Light Source for beamtime and the staff of beamline PXIII. Funding: This work was supported by the Francis Crick Institute, which receives its core funding from the Cancer Research UK (FC001162 and FC001178), the UK Medical Research Council (FC001162 and FC001178), and the Wellcome Trust (FC001162 and FC001178), and by the Wellcome Trust (108014/Z/15/Z and 108012/Z/15/Z). Author contributions: M.A.C., S.C.L., and I.A.T. performed experiments. M.A.C., S.C.L., G.R.Y., J.P.S., and I.A.T. contributed to experimental design, data analysis, and manuscript writing. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors. The coordinates and structure factors for dARC1 CA (S39 to N205) have been deposited in the PDB under accession numbers 6S7X and 6S7Y.

See the rest here:

Structure of Drosophila melanogaster ARC1 reveals a repurposed molecule with characteristics of retroviral Gag - Science Advances

Recommendation and review posted by Alexandra Lee Anderson

The best WIRED long reads of 2019 –

Its time to relax and do some reading. While youre waiting to tuck into some turkey, put your feet up and settle into some of the best long reads weve published this year.

Here is a selection of our best long form journalism from 2019. Not enough? You can find even more of our in-depth journalism in our long form archive. And well be back in 2020 with more stories about how science and technology are changing our world, for better or worse.

Samples in United Neuroscience's laboratory

Alzheimers robs patients of their memories, as plaque builds in the brain, fibres get tangled and connections are lost between nerve cells. When Mei Mei Hu realised her mother, Chang Yi Wang, had created a completely unique vaccine to prevent the disease, she urged her to set up a new company. At United Neuroscience, the mother and daughter duo have combined their scientific and consulting knowledge to find a potential cure that has long eluded researchers. And Alzheimers isnt the only disease in their sights cancer and HIV are next on the list.

Read the full story

Economist Mariana Mazzucato has demonstrated that the real driver of innovation isn't lone geniuses but state investment. Her work to break down tired myths about innovation is now informing governments around the world. Shes currently working with the UK government, EU and UN to apply her moonshot approach to the world's biggest challenges.

Read the full story

In 2010, SoftBank Group CEO Masayoshi Son unveiled his 300-year vision for the future. The company's $100 billion investment arm, the Vision Fund is the biggest tech fund in history

Ryan Pfluger / August

SoftBank is taking over tech one company at a time, with Masayoshi Son as its leader. In 2017 he compared the company to the gentry of the Industrial Revolution the powerful, monied few who funded huge technological and societal changes. Softbank owns stakes in Uber, WeWork and Sprint, among others, and while it may not be a household name like Google or Microsoft, Son has been striving for decades to make it the biggest company in the world.

Read the full story

A Jakarta resident wades past a flooded mosque near the waterfront

Christoffer Rudquist

Jakarta, one of the worlds fastest growing megacities, has a problem: its sinking. Taking clean water from the underground reservoirs that prop up the city means it is slowly collapsing into the mud. The number of people in the city and the timescale needed to solve the problem means that authorities are scrambling to save the Indonesian capital. But the desperate efforts could come too late.

Read the full story

16-year-old Greta Thunberg has mobilised millions of young people to demand action on the world's climate crisis


Greta Thunberg has become the face of the climate crisis protest movement, travelling across the world to urge those in power to act decisively before its too late. Despite her 3.7 million Twitter followers and nine million Instagram followers, the 16-year-old doesnt see herself as a celebrity. I just hope that this movement will continue and we do something about the climate because that is the only thing that matters.

Read the full story

Angela Saini's new book, Superior, exposes the re-emergence of dangerous race science based on genetics

Sebastian Nevols

In the world of genetics, race has long been a factor that scientists have tried to pin down. Some have tried to say that certain races are less intelligent, or more adept at certain tasks. And when a study is published appearing to corroborate such claims, racists eat it up. We keep looking into race, but find very little. In this edited extract from her book, Superior, Angela Saini examines the dangerous belief that with enough data, science could take race a set of categories invented by the powerful to control the weak and somehow make it real.

Read the full story

Makenna Kelly, 13. Her YouTube channel, Life With MaK, has nearly 1.4 million subscribers


Makenna Kelly, a 13-year old YouTube star, gets sent money to eat cookies, drink milk and tap on objects for money. $50 buys a ten minute video, while $30 gets you five minutes. Its all in the name of ASMR the euphoric feeling people get from certain audio stimuli. But videos like these are controversial. Is it right for children as young as five to make videos that give adults brain orgasms?

Read the full story

When the first Fast Radio Bursts, or FRBs were detected by a physics student, it seemed like an incredibly rare phenomenon. Now astronomers agree that one probably happens every second. Thanks to Yuri Milner, a US-based Israeli-Russian billionaire, and his obsession with finding extraterrestrial life, one of the most complex and far-reaching scans has received much-needed funding.

Read the full story

Amos Chapple / Radio Free Europe / Radio Liberty

As global temperatures rise, Siberia is melting. Amongst the thawing tundra, hunters are searching for tusks. China banned the import and sales of elephant ivory in 2017, but finding long-dead mammoths provides a loophole. While it may seem like a safe option, encouraging ivory sales is fraught with risk.

Read the full story

DeepMind co-founder and CEO Demis Hassabis

Jason Madara

DeepMind's algorithms have conquered games. Now they're taking on something much harder: science. In September we profiled the Google-owned artificial intelligence firm as it sets its smarts on protein folding, which biologists consider to be the building blocks of life. As it continues to pursue its stated mission to solve intelligence, we go inside the secretive London firm to explore exactly what its up to.

Read the full story

See the original post here:

The best WIRED long reads of 2019 -

Recommendation and review posted by Alexandra Lee Anderson

Our best recipes from 2019 | Food and cooking –

Beef Wellington for Christmas Dinner, Thursday, Dec. 5, 2019. Photo by Hillary Levin,

Yield: 8 to 10 servings

3 pints (1 pounds) white button mushrooms

2 shallots, peeled and roughly chopped

4 garlic cloves

8 sprigs fresh thyme, leaves only, divided

Salt and pepper

1 (3-pound) center-cut beef tenderloin, trimmed

Olive oil

12 thin slices prosciutto

2 tablespoons Dijon or English mustard

Flour, for rolling out puff pastry

1 pound puff pastry, homemade (see recipe) or store-bought, thawed if frozen.

2 large eggs, lightly beaten

teaspoon coarse sea salt

Note: The duxelles and the homemade puff pastry (if using) can be made a day or two ahead of time.

1. For the duxelles: Add mushrooms, shallots, garlic and the leaves of 2 of the sprigs of thyme to a food processor and pulse until finely chopped. Place a large saut pan over medium heat, add the shallot-and-mushroom mixture, and saut until most of the liquid it releases has evaporated. Season with salt and pepper, and set aside to cool. May be refrigerated for up to 3 days.

2. For the beef: Tie the tenderloin in 4 places so it holds its cylindrical shape while cooking. Drizzle with olive oil, then season with salt and pepper and sear all over, including the ends, in a hot, heavy-bottomed skillet lightly coated with olive oil.

3. Meanwhile, set out your prosciutto on a sheet of plastic wrap at least a foot and a half in length. Shingle the prosciutto so it forms a rectangle that is big enough to encompass the entire filet of beef. Using a rubber spatula, cover prosciutto evenly with a thin layer of duxelles, and season with salt and pepper. Sprinkle with leaves from the remaining 6 sprigs of thyme.

4. When the beef is seared, remove from heat, cut off twine and smear lightly all over with mustard. Allow to cool slightly, then roll up in the duxelles-covered prosciutto, using the plastic wrap to tie it up tightly. Tuck in the ends of the prosciutto as you roll to completely encompass the beef. Twist ends of plastic to seal it completely and hold it in a log shape. Refrigerate 30 minutes to ensure it maintains its shape.

5. Preheat oven to 425 degrees.

6. On a lightly floured surface, roll out the puff pastry to form a rectangle large enough to completely encompass the beef (this is vital if necessary, overlap 2 sheets and press them together). Remove plastic from beef and set meat in middle of the pastry. Fold the longer sides over the meat, brushing the edges with beaten egg to seal. Brush ends with beaten egg to seal, and fold over to completely seal the beef. Trim ends if necessary. Top with coarse sea salt. Place seam-side down on a baking sheet.

7. Brush the top of the pastry with egg, then make a few slits in the top of the pastry, using the tip of a paring knife, to allow steam to escape while cooking. Bake 35 to 45 minutes until pastry is golden brown and beef registers 125 to 130 degrees on a meat thermometer for medium rare, 135 to 140 degrees for medium, 140 to 145 degrees for medium well or 150 to 155 for well done.

8. Allow to rest before cutting into thick slices.

Per serving (based on 8): 762 calories; 41g fat; 11g saturated fat; 194mg cholesterol; 64g protein; 33g carbohydrate; 3g sugar; 2g fiber; 1,779mg sodium; 68mg calcium

Adapted from a recipe by Tyler Florence, via Food Network


Yield: 12 servings

2 cups all-purpose flour, preferably chilled

teaspoon fine sea salt

20 tablespoons (2 sticks) unsalted butter, chilled and diced

cup ice-cold water

Note:This is best prepared in a cool kitchen, on a cool work surface, using light and assertive gestures to prevent overheating the dough. Dont attempt it when the oven is on.

1.In a medium bowl, sift together the flour and salt. Using a pastry blender or two knives, cut the butter into the flour, stopping when the mixture looks crumbly but fairly even, with the average piece of butter about the size of a large pea.

2.Turn out onto a clean and cool work surface and form a well in the center. Pour in the water and work it into the flour and butter mixture with a bench scraper or a wooden spoon. Knead lightly, just enough so that the dough comes together in a ball, and shape into a rough square. There should be little pieces of butter visible in the dough. If you have time, refrigerate 30 minutes.

3.Flour your work surface lightly. Using a lightly floured rolling pin, roll out the dough in one direction into a rectangle about 20 inches long. Add more flour as needed to prevent sticking. Brush to remove excess flour and fold the dough in three, like a letter, so the top and bottom overlap, dusting again after the first fold.

4.Give the dough a quarter of a turn, and repeat the rolling and folding steps. Repeat until youve rolled and folded a total of four times. You should get a neat rectangle or square pad of dough. If you find the dough becomes sticky at any point, refrigerate for 30 minutes to cool again.

5.Put the dough on a plate, cover and refrigerate for at least 1 hour or overnight before using. If the dough seems too stiff when you take it out of the fridge, let it come to room temperature for 15 to 20 minutes before using.

Per serving:246 calories; 19g fat; 12g saturated fat; 51mg cholesterol; 2g protein; 16g carbohydrate; no sugar; 1g fiber; 100mg sodium; 9mg calcium

Adapted from Tasting Paris, by Clotilde Dusoulier

View original post here:

Our best recipes from 2019 | Food and cooking -

Recommendation and review posted by Alexandra Lee Anderson

The 10 most compelling product innovations of 2019 – Fast Company

As the 10 most important product innovations of 2019 showfrom plant-based burgers to alternate reality headsetsthe world still has plenty of room for innovation in meatspace. The brightest products of the past year arent just enticing or convenient for users. These products are often using design to question the ritual of consumption as we know it.

[Photo: Motorola]The original Motorola Razr (2004) changed the world of technology forever, turning clunky phones into sleek, fashion-forward objects of desire. The followup, 15 years later, features a folding OLED display. Its a mechanical marvel, and the first flexible screen device that makes any real sense at all, because it turns our too-large smartphones into pocketable devices. The Razr offers an early taste of the bendable, ergonomic electronics to come. [Link]

[Photo: Logitech]Video game controllers are now so advanced that the military uses them to control tanks and drones. But they require fine motor control that can leave people with disabilities behind. The Xbox Adaptive Controller launched last year, with two giant buttons and lots of extra input options to allow users to connect extra buttons as needed. Then, this year, Logitech decided to actually make those extra buttons. Its $99 kit includes mix-and-match hardware thats built less for profitability than the needs of diverse users. [Link]

[Photo: Air Co.]My dad always said there was no problem too great that you couldnt drink yourself out of it. Okay, that isnt true. But carbon-capturing vodka comes pretty close. A company called Air Co. uses recaptured carbon in the place of yeast to produce vodka. Each bottle scrubs the air as effectively as eight trees breathing for a day. And as an added bonus, Air Co.s production footprint needs just 500 to 1,000 square feet compared with the acres of land required by traditional distilling. Then take a sip while wearing this compelling, carbon-negative raincoat, and you wont have to worry about spilling on yourself. [Link]

[Photo: Motorola]First responders are going into some of the most dangerous places on earth, and in these places, your typical iPhone wont do because it relies on an infrastructure of fallible antennas to work. Instead, first responders still rely on long-range walkie-talkies. Anew walkie-talkie from Motorola Solutions, the APX Next, can be used both hands-free and without an operator on the other end of the line, thanks to a novel voice assistant that helps you access private information without direct internet access. Siri may be an overrated way to find sushi. But the APX Next can literally help save lives; as a firefighter or police officer uses two hands to free someone from a pile of rubble, she can use the APX Next to simultaneously call for help. [Link]

[Photo: Analogue]The Nintendo Switch is the best portable gaming system ever madethanks to a perfect size, a massive library of games, the option to seamlessly dock it to a TV, and controllers that put smartphones to shame. And yet, 2019 brought us two compelling handheld video game consoles (both expected to be released in 2020). Each proves that the independent spirit of hardware design is alive and well.

Analogue Pocket is a $199 Game Boy reboot, which runs vintage console cartridges but in an industrial design that meshes stark minimalism with a cutting-edge display. Oh, its also an instrument for electronic music. What? [Link]

The Playdate is another enticing bit of gaming hardware, but its more experimental. A surprising crank on the side offers a zany way to play games. And its being released with software partners who are designing new, bite-sized titles for the Playdate and the Playdate alone. Playdate teased a model in which you could buy new seasons of games in packs, and in doing so, Playdate is combining a closed hardware/software ecosystem in a way that only giants such as Apple and Nintendo have ever managed to pull off. [Link]

[Photos: Impossible Foods, Burger King]If 2019 was the year of anything, it was the year of fake meat. Beyond Meat and Impossible both made their mainstream mark. The Impossible Whopperwas such a hit, it gave Burger King its best quarter in four years, cementing nearly a decade of investment in the biology, flavor, and mouthfeel behind a fully engineered burger.Even if faux meats dont outright replace real meat, a little savings in the flexitarian market goes a long way: A pound of beef costs 1,800 gallons of water on top of all sorts of other environmental hazards, which is why experts would like to see beef consumption drop by 50% to save the planet. The Impossible Whopper might not be the best burger youve ever had, but then again, neither is anything else you get at Burger King. [Link]

[Photo: Microsoft]If theres a more complicated industrial design story in 2019 than how Microsoft designed the Hololens 2 augmented reality headset, I havent read it. Its an AR headset that goes on as easily as a baseball cap, making it easy and effortless to hop into the digital world. The combination of materials and hard and soft parts in this design is staggering. And its full of tiny decisions of ergonomics, which work in harmony with technology that requires picometer-level precision (if some parts of the headset come out of the tiniest threshold of alignment, it would literally make you want to vomit). [Link]

[Photo: courtesy Korvaa]The headband is made from lactic acid produced by yeast. The ear padding is a bubbling protein produced by fungus. The leather is mycelium, or the core of a mushroom. And the mesh on your ears is biosynthetic spider silk. Dubbed Korvaa, this is the worlds first microbe-grown pair of headphones. And they are beautiful in their own way. As we reckon with our environmental footprint, projects such as Korvaa are a reminder that there really is another way than simply producing more plastic. [Link]

[Photo: Adidas]The Adidas Loop is a shoethat can be ground down at the end of its life and used to help make new Loop shoes. Whether its the textiles made from plastic, or the business modelwhich may require Adidas to incentivize buybacks of old shoes to make new onesLoop teases an increasingly complicated future for consumer goods (and consumption) in which companies and customers alike are forced to deal with the long-term impacts of products. None of this would matter if Loop shoes were terrible, of course. But they are also a tantalizing garment in their own right, with a shimmery woven plastic thats both beautiful and comfortable. [Link]

Read more:

The 10 most compelling product innovations of 2019 - Fast Company

Recommendation and review posted by Alexandra Lee Anderson

Wow your New Year’s Eve guests with a puff pastry appetizer –

GOLDEN VALLEY, Minn. Chef Lindsay Guentzel stopped by KARE 11 Saturday to share a simple and delicious appetizer idea for New Year's Eve celebrations. Her recipe for Holiday Brie En Croute uses puff pastry with an egg wash that helps the pastry bake perfectly, as the protein and fats in the egg give you that perfectly golden brown finish.

Holiday Brie En Croute

1 sheet frozen puff pastry, thawed out8 oz. brie cheese, sliced cup dried cranberries cup walnuts, choppedHoney, drizzledEgg wash

Preheat oven to 400.

Line baking sheet with parchment paper. Lay out rectangular puff pastry.

Using palms, gently spread out dough.

Starting at one end of pastry shell, place brie in a line down the center running the long way (think long like a hot dog bun, not short like a hamburger bun). The slices will overlap.

Spread cranberries and walnuts over brie evenly and drizzle with honey.

Starting at one end, slowly fold the sides of the pastry shell up over the brie by pinching the corners of the dough between your fingers, lifting up and twisting over (the twists add texture and dimension to the top of the pastry). Move a few inches down the pastry shell and repeat folding movements, gently shaping the dough as you go along.

Using a pastry brush, gently brush egg wash over the pastry shell.

Bake for 20 minutes until golden brown.

Serve on platter warm with knife and serving spatula.

Follow this link:

Wow your New Year's Eve guests with a puff pastry appetizer -

Recommendation and review posted by Alexandra Lee Anderson

The Art of Origami is Now A Key Tool That Helps Doctors Save Lives – Nature World News

Dec 23, 2019 05:03 AM EST

Origami's new role in the field of science and technology has definitely taken a turn for the better in the recent decade. Better known as origami engineering, the practice is used to reduce structures or maximize space and function.

Origami engineering has made great strides in the medical field in particular. The same principles used in origami, when applied to medical devices, allows implants to be folded to minuscule sizes and then unfolded to its actual size. The reverse is also applicable, where like toothpaste tubes, can be fully de-compressed.

Folding techniques could transform flat objects with wrinkles to increase resilience, shock-absorbance, strength, or rigidity. Origami provides a unique insight into how single pieces could sustainably be packaged without cutting, welding, or riveting, allowing for cheaper manufacturing costs and easier assembly.

The utility of origami engineering has captured the attention of people such as Rebecca Taylor, assistant professor at Carnegie Mellon University's Department of Mechanical Engineering. Taylor specializes in microfabrication and biomechanics, a study that has helped her fabricate microscale sensors to reliably assess cardiomyocytes derived from stem cells. A natural inclination to similar practice, Dr. Taylor has developed an origami-based DNA synthetic cardiac contractile protein, which allowed her to observe merging mechanics in multiprotein, acto-myosinc contractile systems.

As a professor, Taylor expands on the utilization of DNA origami in medicine. This technique (also referred to by Dr. Taylor as "bottom-up manufacturing"), allows improvement in nanomanufacturing and nanomechanics of multiprotein systems, paving the way for heart stents that could unfold in a very precise location.

The problem, however, is on how to deploy these structures in a 100% fault-free way. To illustrate this, a common problem that impedes the creation of pop-up tents that could self-assemble at the press of the button is when the folds of the tent get stuck during the folding process on occasion.

Understandably, this raises some concern among those who are keen to use self-folding nanomachines in medicine.

So this is where origami comes in.

According to University of Chicago scientists, the limits of self-folding structures could be intrinsic in that so-called "sticking points" seem to be unavoidable.

Previously thought possible to engineer around, the researchers observed the capacity of foldable structures by creating mathematical models. During the experiment, the team had designed structures capable of self-folding, such as paper origami and nanobots, and creating creases in them beforehand. The result was that when more pre-creases were added to the folds, the more branches in the next folding process could form and the more likely the self-folding mechanism is to get stuck.

Origami engineering is a relatively new innovation. Its application is vast and can be of use to not only technology but to medicine as well. The development of the field itself, then, needs to pick up at a faster pace in order to cater to the intelligent design of foldable structures and materials. But while there are creases in the field that needs to be smoothed out, the greater promise of origami engineering has brought about several research papers in its wake.

RELATED ARTICLE: Swallowed a Battery? Ingestible Origami Robot Made from Pig Gut Can Remove It,Stop Stomach Bleeding

2018 All rights reserved. Do not reproduce without permission.

See more here:

The Art of Origami is Now A Key Tool That Helps Doctors Save Lives - Nature World News

Recommendation and review posted by Alexandra Lee Anderson

Nanopores can identify the amino acids in proteins, the first step to sequencing – University of Illinois News

CHAMPAIGN, Ill. While DNA sequencing is a useful tool for determining whats going on in a cell or a persons body, it only tells part of the story. Protein sequencing could soon give researchers a wider window into a cells workings. A new study demonstrates that nanopores can be used to identify all 20 amino acids in proteins, a major step toward protein sequencing.

Researchers at the University of Illinois at Urbana-Champaign, Cergy-Pontoise University in France and the University of Freiburg in Germany published the findings in the journal Nature Biotechnology.

Graduate student Kumar Sarthak and physics professor Aleksei Aksimentiev were part of a research team that demonstrated that nanopores could sequence proteins, giving reserachers and clinicians insight into activity within a cell.

Photo by L. Brian Stauffer

Edit embedded media in the Files Tab and re-insert as needed.

DNA codes for many things that can happen; it tells us what is potentially possible. The actual product that comes out the proteins that do the work in the cell you cant tell from the DNA alone, said Illinois physics professor Aleksei Aksimentiev, a co-leader of the study. Many modifications happen along the way during the process of making protein from DNA. The proteins are spliced, chemically modified, folded, and more.

A DNA molecule is itself a template designed for replication, so making copies for sequencing is relatively easy. For proteins, there is no such natural machinery by which to make copies or to read them. Adding to the difficulty, 20 amino acids make up proteins, as compared with the four bases in DNA, and numerous small modifications can be made to each amino acid during protein production and folding.

Many amino acids are very similar, Aksimentiev said. For example, if you look at leucine and isoleucine, they have the same atoms, the same molecular weight, and the only difference is that the atoms are connected in a slightly different order.

Nanopores, small protein channels embedded in a membrane, are a popular tool for DNA sequencing. Previously, scientists thought that the differences in amino acids were too small to register with nanopore technology. The new study shows otherwise.

The researchers used a membrane channel naturally made by bacteria, called aerolysin, as their nanopore. In both computer modeling and experimental work, they chopped up proteins and used a chemical carrier to drive the amino acids into the nanopore. The carrier molecule also kept the amino acids inside the pore long enough for it to register a measurable difference in the electrical signature of each amino acid even leucine and isoleucine, the near-identical twins.

This work builds confidence and reassures the nanopore community that protein sequencing is indeed possible, said Abdelghani Oukhaled, a professor of biophysics at Cergy-Pontoise whose team carried out much of the experimental work.

The researchers found they could further differentiate modified forms of amino acids by using a more sensitive measurement apparatus or by treating the protein with a chemical to improve differentiation. The measurements are precise enough to potentially identify hundreds of modifications, Aksimentiev said, and even more may be recognized by tweaking the pore.

This is a proof-of-concept study showing that we can identify the different amino acids, he said. The current method for protein characterization is mass spectrometry, but that does not determine the sequence; it compares a sample to whats already in the database. Its ability to characterize new variations or mutations is limited. With nanopores, we finally could look at those modifications which have not yet been studied.

The aerolysin nanopore could be integrated into standard nanopore setups, Aksimentiev said, making it accessible to other scientists. The researchers are now exploring approaches to read the amino acids in sequential order as they are cut from the protein. They also are considering other applications for the system.

One potential application would be to combine this with immunoassays to fish out proteins of interest and then sequence them. Sequencing them will tell us whether theyre modified or not, and that could lead to a clinical diagnostic tool, Aksimentiev said.

This work shows that theres really no limit to how precisely we can characterize biological molecules, he said. Very likely, one day we will be able to tell the molecular makeup of the cell what we are made of, down to the level of individual atoms.

The National Institutes of Health and the National Science Foundation supported this work. Computer modeling was done on the Blue Waters supercomputer at the National Center for Supercomputing Applications at the U. of I.

Read the original post:

Nanopores can identify the amino acids in proteins, the first step to sequencing - University of Illinois News

Recommendation and review posted by Alexandra Lee Anderson

Page 11234..»