Biochemistry Focus webinar series: Digital Biology
Hello, and welcome to biochemistry. Focus, webinar, series. On behalf of the biochemical. Society. And portland, press, i am pleased to welcome you to this webinar. Topics. In the series include different research, areas, in the molecular biosciences. As well as practical, sessions to support career, development. Each webinar, will give you the opportunity, to ask questions, via text, and we welcome suggestions, for future speakers. So, welcome to today's webinar again, i am professor, natalia krasnogor, and i hold a chair, in computing, science and synthetic, biology, at newcastle, university. And also royal academy, of engineering, chair in emerging technologies. At newcastle, i, engage in research, that sits at the interface, of computing, science, engineering, biology. And dna, remain on our technology. Today's, webinar. Is called, digital, biology. Advanced computational. Approaches, to biological. Design, and engineering. And we will hear from two researchers. About their work, which has been enabled. By cutting-edge, computational. Methods. First, we will hear from professor, christian, berges chaffetzel. Christian, is a professor, of biochemistry. At the university of bristol. She has studied and worked, both ethioric. And the university, of zurich. And was a team leader, and a, european, research council investigator. At ember. Before coming to bristol, in 2014. Working christian, love ranges from eukaryotic. Translation, initiation. Via regulation, of translocation. Termination, to co-translation. Of membrane, proteins, via the bacteria, called translocon. The lab's research, is funded by bbsrc. And the medical research council. And the group was recently, awarded, a welcome trust investigator. Award. Bristol, is home to a prison, bio a bbsrc. Epsrc. Jointly, funded synthetic biology, research center. An international, team of scientists, labor christian, and herbies, symbiote. Colleague, imri berger. Have recently, discovered. A drag-able, pocket, containing. A linoleic. Acid. In the electron, cryo microscopy. Structure. Of the sarskovit2. Spike, glycoprotein. And this is what we will hear about today. Following. Christian's. Presentation. We will hear from nicole, percy. A research, fellow, at the synthetic, biology, research center in nottingham. Sblc. Nottingham. Is also jointly finally, funded by bbsrc, and the epsrc. And concentrates. On engineering, bacteria. To make industrially, useful products, from, c1, feedstock. Including, the greenhouse, carbon monoxide. Carbon dioxide. And methane. The center researches, both aerobic. And anaerobic. Microbial, systems. But its primary, focus is aerobic, gas fermentation. And the chosen chances for this work is cuprio, vidus necato. Nicole will today present on her work, into the metabolic. Engineering, of c1, gas assimilating. Microbes. Which are being tailored, as industrial, chassis, for the sustainable. Production, of platform chemicals, and biofuels. And on the genome-scale, metabolic, models of aerobic. And anaerobic. Industrially, relevant, chassis. Developed. To practice metabolic, engineering. Effectively. Before i hand over to our speakers, i would like to mention, that. The questions will be answered, at the end of the webinar. But at all time you can send in your questions. By typing, in the box that you have in, front of you. The. At this point, i would like to. Hand over to, a christian. Thank you very much to the organizers. And thank you for the introduction. And for giving me the opportunity. To present, our work. And um i'm going to talk about the fatty acid, binding, pocket, in the structure, of salsa.
Spike Protein, in. So um. You probably all have seen this image a lot of times, it's the sars2, virus, and on top of the, on the surface, we have the spike protein, in. Which is responsible. For docking. Of the. Virus, to the host cells to the ac2, receptor and the other, cells and then, mediates, the membrane, fusion, with the host cell so it is import. Important for infectivity. And this protein. Is the main. Protein. And all the vaccine, development, efforts, are. Against, the spike protein to develop, naturalizing. Anti-protein. Bodies, and to develop. And thereby, interfere, with infection. So what we did at the beginning, of the pandemic, is that we decided, to produce, the spike, protein, in. And. We have basically, replaced, the membrane, part, of the spike, protein, with a primarization. Domain, and produced, the protein. In insect cells. And that is worked by frederick, garzoni, and julian caper, in our group. Now. For cryo, em, we have used the sample, and flash froze, it and here you see a typical, micrograph. We get from cryo, em. And it's not your screen resolution, it is actually a really, noisy image and it's very hard to see the particles. Here and that's a common, feature. In cryo electron, microscopy. So how do we get from these very noisy, images, actually to a high resolution. 3d, reconstruction. Which. Which is called the atomic, resolution. The answer, is. Averaging. So by averaging. Over many, identical. Particles, we can increase. The signal to noise ratio, and actually come. To a resolution. Of three point, um. Uh. Two point eight angstrom which we had achieved here. So we have used about six hundred. 000 particles, compared, each particle, with each particle. Class, them, into different, views, and then averaged. And here are the 2d class averages, now you clearly, see the particle, and all they are high resolution. Features alpha helices. In the particles. And basically, this, process, of comparing, all the particles. With each other. Is very computation. Intensive. But it can be paralyzed. And that makes.
Uh. High performance, computing. And and computing, in the cloud, possible, and this is what we have done. Now, to kind of. Show you the the process, of um, cryo and single particle, cryo electron microscopy. Again and that you better understand. What we actually, do um by image processing. I'm going to show this movie we have our sample. In an aqueous, buffer we flash freeze the sample, and then bring it to the temperature of liquid nitrogen. The sample, is on a grid we illuminate, it in the cryo electron, microscope. We collect, these very noisy, data, and then we pick individual. Particles. Shown here. We have a starting, 3d, model which doesn't have to be correct but which basically, provides. Initial, ideas, how these different particles, are, related. In view as used, to each other, so by comparing. The different. Images. With the starting, model we can assign, top, side, and front views. And we have many more views, because the particle, is randomly. Oriented. In the eyes. And by collecting, many many data we can fill, all the different, views. And now all the, images, have, attributed. Angles, so we know how they are related, in space, to each other, and this information. Can be used to reconstruct. An own 3d. Structure. From our, actual, data. And, by repeating, this process, and the resulting, 3d structure, is then again, fed in as an initial, model, and do this. Many times we actually can achieve, a very high resolution. Reconstruction. So. This is our crying, m structure, of the spike protein. It's work done by by christine. Turtle. In our groups. And we have in this query, m. Sample we have identified. Two different, conformations. Um, one of them is the open conformation. It's formed by. About 30. Of the particles. And, here in blue you see, the receptor, binding, domain, of the spike protein, pointing upwards. This is the atomic, model where you see the, the, receptor, binding domain which is up. And then we have the majority, of the particles. In a closed conformation, where all the receptor, binding, domains, are pointing, downwards. So none of them is upwards and if you look at this top view, you see, that this particle. Is geometric, and we applied c3, symmetry. And achieved the 2.85. Angstrom. Resolution. Structure. Now when we analyze, this closed conformation. We found, extra. Non-protein. Density. In the receptor, binding domain, it's this tube-shaped. Density. Here, and this has not been reported. In any previous, sars-2. Structure, spike protein, structure. At the point we had, reconstructed. This structure, there were two spike, protein, structures, already, published. And this density, is present, in all three subunits, this is the non-symmetrized. Map and you can see in the three. Rbds. In cyan. Pink, and green, the, extra density. And it's christine's, achievement. To realize. That this. Extra density, is surrounded. By hydrophobic. Amino, acids so it must be a very hydrophobic. Molecule. And with this tube shape density, she speculated. That it is a fatty acid. And she compared. This, density, and structure, with. All the available. Structures, of proteins, with fatty acids bound. And the most similar. Structures, contained, linoleic, acid and here's a fatty. Acid, kindness. B3, and odorant, binding, protein, and both of them have linoleic, acid. Bound and it adopts a very similar, kink shape. We provided, additional. In, evidence, that it's linoleic, acid by using mass spectrometry. And this was done with, collaborators. At the max planck institute, in heidelberg, joachim, schwartz. Oscar, schauffer, performed, the aztov. Mass spectrometry. And he obtained spectra, with a peak at 279. Daltons. This is. Electron, spray, ionization. Mass spec so this is the iron, and. Therefore there is a difference, of one. Between the acid, and the iron which we detect, here, so this fits extremely, well and we have a peak which is compatible, with linoleic, acid in mass back.
What Is special, about, our structure, is that our. Fatty acid binary pocket is b part height. So here you see the column, potential. And one rbd. Provides, a hydrophobic. Pocket, here in white. And the second, rvd. Provides. Positively. Charged, residues. In blue here that interact, with the carboxy, head group of the fatty acid. What's special about linoleic, acid. It's an essential. Free fatty acid. It's also a, poly, so we need to actually take it up with food it's a vitamin. It's as an unsaturated. Fatty, acid. And it's essential for maintaining, membrane fluidity, and surface. Tension in lungs, and alteration. Of the la pathway, lipid composition. Is observed, in acute respiratory. Distress, syndrome, and severe pneumonia. And both of these are key elements, in covid19. Pathology. In the body la is metabolized. To our redonic, acid and prostaglandins. And these are two key molecules, central to inflammation. And immune modulation. Again, key symptoms. Of, covalent, nitric pathology. Importantly. La, levels have been shown to be decreased. In covid19. Patients, here. So la is depleted. In, patients. We have then compared. Our, spike protein, structure with the two previous. Apo, structures, which did not contain. Linoleic, acid. In the overlay, here you see the upper structure in gray and our structure, in cyan. And you can see that there is a gating, helix, moving, away, to make space for the linoleic, acid. And in the skating, helix we have two tyrosines, which need to swing away, and then can interact, with the linoleic, acid in the pot. The second, rbd, contributes, the positive, charge, residue. To interact, with the carboxy, head grade, group, and to do so it moves. Six angstroms, closer, to the first, rbd. And as a consequence. In the rbd, trimer. The confirmation. Is much more compact, and this is shown here so the rbds, are much closer. To each other compared to the upper structure, which where the the, distance, between, the. Uh two rbds, is eleven and stream and more. So we have a compact, structure, here. Now in the previous structure, about seventy, per five percent. Of the, particles. Were in the open conformation. With the one rbd. Up, conformation. And here i've colored, in red, the receptor, binding, motif. And this is the motive, that interacts, with ace2, this is a crystal, structure, shown the receptor, binding motif, interacting, with the ace2, receptor. And, thus this is the infectious, form which is compatible, with. Binding. In contrast, in our structure, 70. Is in the closed conformation. Where the. Receptor, binding, motifs, are. Tucked away at the interface, of the rbds. And the receptor. Cannot, bind and thus this, conformation. Is considered, to be non-infectious. This is again, an. Image. Top view. And you can see the receptor, binding motif, in between. The rbds, it's fully structured. In our cryovm. Density. But it's not compatible, with, acetobinding. So at that point we thought we need to characterize. Further the ace ii binding, of our la bound stars, to spike protein. In a first experiment, we performed. Size exclusion, chromatography. We mixed spike, and ace2, receptor. And we obtained, a peak, which is rooted, at smaller, volume meaning a higher molecular, weight, so there is complex, formation, and we observe, two proteins. The two proteins. In the peak one fraction. Similarly. In, elisa. We see that the, spike protein, can bind to immobilized. Ace2, and compete with soluble, rbd, domain. So it binds ace2. Now this is the important, experiment, where we then decided, to. Compare, a lay bounce bike with upper spike. This has been done by, by our core experiment with immobilized. Ace2. And we removed, from the la bound, spike. The la. With the lipidex, column, to obtain upper spike, and when we remove the linoleic, acid we constantly.
Observe Contin. Consistently. Observed, a higher signal. So more upper spike binds to the surface at same concentration. Compared, to a la bound spike. And consequently. The dissociation. Constant, is smaller, meaning higher. Affinity. For the upper spike compared to the l.a bound spike. This has consequences. For infectivity. We thought, and we'd convinced, our colleague, andrew davidson. To. Perform experiments, with live cells whatever's. And here. You'll see um. Immunofluorescence. Images. The virus is green, and the cells are blue. In the upper row this is the inhibition. Of virus, replication. By rendezvilla. So with increasing, rendezvous. Concentrations. The virus replication. Is blocked. And a much stronger, effect on virus replication. Is seen when we mix, la, and randomly, severe so the two synergize. To block virus, replication. And this can also be followed, by, quantitative. Rt pcr. Where we see, less, viral rna, present, in the sample, when we add randy severe, and linoleic, acid. So la and rendezvous. Synergize. To suppress, us to, virus, replication. The spike protein in structure and sequence, is conserved. Among, the corona, viruses. And here this is a sequence, comparison. Between the seven corona, viruses, that can, infect, humans. The residues. That line the hydrophobic. Pocket, are underlaid, in cyan. Including, the in purple, the two tyrosines, that swing away in the gating, helix. And we have in green the residues. That. Interact, with the carboxyl. Head group. If you compare cells, 2 and sas 1 you see that all these residues. In the hydrophobic. Pocket, are conserved. And also the positive. Residues. So we predict, that size 1, would also have a linolic, acid binding, pocket. In the receptor, binding domain. In males. The sequence, is less well conserved. But the residues, lining the pocket, are still hydrophobic. So there is a hydrophobic. Pocket and there are also positively. Charged residues. In the neighboring, adjacent, rbd, which could interact with the carboxyl, head group. So, possibly. Mass 2 also has a fatty acid. Binding, pocket. Now how can we, carry this forward, i've shown you that.
In The. Upper, structure, 75. Is in the open, conformation. Whereas, in the l.a, bound structure the majority. Is in the closed conformation. So if we now could develop, a molecule, that binds, even better than la. We would have an. Uh potential. Very potential. Um, and good antiviral. Which potentially, could lock it entirely, in the non-infectious. Form. And there and hopefully, irreversibly. Lock it there and thereby, decrease, cells to infectivity. There is a very encouraging. Precedent, for such an approach, from michael rosman's, group. And. This is used to, treat rhinovirus. Infections. Rhinoviruses. Have a receptor. Binding protein, with a fatty acid binding, pocket. And. The group has developed, green compounds. Which bind, into this hydrophobic. Pocket. And, block receptor, binding, proteins. Directly. And thereby they. Completely. Abolish, infectivity. Of rhinoviruses. And this is successfully. Used in the clinic already. Now in summary, i've shown you that, we have discovered, an la binding pocket in the south turkoff 2 spike. This provides, the first direct structure, link between linoleic, acid, covid19. Pathology. And the cells ii virus itself. The pocket, is bipartite. It connects two rbds, within the trimer. It stabilizes. The locked form of the spike protein. And this, locked form is non-infectious. The fatty acid binding pocket appears to be present, in other disease-causing. Coronary, viruses. As well. And we think that the pocket, is, is drivable. And we could develop. Antivirals. Um. Which would, then unlock, the the spy protein, entirely, in the non-infectious. Form. Moreover. Other. Other proteins, along the multi-nodal. La signaling. Access could be targeted. To. Achieve a therapeutic. Intervention. Against, us two infections. Thank you very much for your attention, and with this i would like to thank those who contributed. To the work this was a collaboration. With emre berger. And um christine. Uh. Gupta. Worked in our lab to produce spike, and and the cryovm. Structure, we worked with andrew davidson. And adrian mulholland. Did md. Perform, nd simulations. Which i didn't have the time to show. And i would also like to thank welcome trust, and um and oracle, for helping, to um. To, actually fund this book. Thank you very much for your attention, i'm happy to take, questions. Thank you very much, christian, that was a fascinating. Talk. We will switch over now to. Nicole. Thank you. We can see your slides yes thank you. Great. Thank you. Again thank you to the organizers, for this opportunity. And thank you to nat for the great introduction. Um. So i'm going to talk about some of the modeling, that we're doing at the synthetic, biology, research, center.
Specifically, Today i'm going to talk about the genome scale metabolic, modeling work that we've been doing, to try and help the experimental. Work that we're doing. So nat already gave a really nice introduction, to the sbrc. But just a quick. Recap. So, we're. Engineering, bacteria. To provide, alternative. And sustainable. Routes, to producing, platform, chemicals. That are otherwise. Produced, via, fossil fuels. And also, importantly. These bacteria. Can grow, on c1, feedstocks. Such as carbon dioxide. And. Carbon monoxide. That are produced, in, large, quantities, in industry, such as steel mills. That contribute. To. Uh greenhouse, gases, and global, warming. So the idea here is that we'll, use these bacteria. As microbial. Factories. For converting. These, waste. Gases, into, more valuable, added. Chemicals. So how does the genome scale metabolic, modeling work, that we're doing in the sbrc. Fit into the overall, work that the sbrc, are doing. So we start off from selecting, a wild-type, strain. So as we've already mentioned, the first criteria. Is that, the bacteria. Have to grow on c1 compounds. And we also want them to be able to grow to high cell densities. Since the more cells we have the more product we have. We then have to do some data collection. Uh initially, it's very important that we have an accurate genome annotation. And then we want to start. To collect. Uh physiological. Characterization. Growth curves. Um nutrient demand, etc. And this then allows us to construct, a, genome scale model. And so this starts from. The genome annotation, we can build a genome scale model. And then validate, it using our experimental. Data. Once we're happy with our initial model we can start to simulate, the behavior. Of these bacteria. So we can assess. Different, pathways, for producing, our, product, see how, which pathways, have optimal, yield or, minimum, energy requirements, for example. We can also, look to use these models for identifying. Genes, to either delete. Or insert, into the bacteria. For ensuring, that we've got our product being made. And we can also use these models to just have a general. Um. Better understanding. At the system, level. Of the bacteria's. Cellular, metabolism. So once we've made some predictions, we can then. Test these in the lab, the sbrc, experimentalists. Have constructed. A lot of genetic, tools, that allow us to. Insert, or delete, genes. Into our bacteria. And then after this process. Ideally, we'll end up with a chassis, that isn't, is efficient, at producing, our, product. And ideally, also, has high growth rates. More often than not however you can end up with a sub-optimal, strain. And this is just to highlight that we go through this process. Several, times. It's an iterative, cycle where we keep informing, each step, as we go around. So hopefully that gives you an idea of how the genome scale models. Modeling work fits in with the overall work that's being done in the sbrc. What, actually is a genome scale model. So i mentioned, that it starts from the annotated. Genome, so this provides, us with all the information, of the genes. That are in the bacteria. The gene functions, that we can then ex, allow us to extract the entire, set of biochemical. Reactions. This. Results. In a large. Intricate. Metabolic, network, of metabolites. And reactions. And the genes that are associated, to them reactions. To get to this point it's quite a fast process, so long as you have an accurate, genome annotation. There's lots of automated. Tools, available. That allow you to just extract, this information. Extract, the information, from.
Database, Such as kegs such as biocyc. However, at this point. You'll, it's very unlikely, that the model is going to be functioning, and, producing, results, that are comparable, to those from the lab. So then we do some model refinement. This is where it's, a lot more work involved. So we need to make sure the reactions, are mass balanced. Check the directions. There's going to be some missing annotation, in our genome, so we have to do some gap filling. And then we add a biomass, reaction. To. Represent, cellular, growth, of, the bacteria. And we've, carried out this process. For three different, bacteria. That we're interested, in at the sbrc. So, coopervedas, necata, that's our core. Bacteria, that we're working with in the sprc. And we've got clostridium, autoethanogenium. Both of which were constructed, in collaboration. With oxford, brooks university. We also have a model for, eu bacterial mimosa. Again this was constructed, in collaboration. With, the korean, advanced. Institutes. Of science and technology. And also the university, of toulouse. So here in the table this is just to highlight, the number of reactions, metabolites. And genes. That are in these. Models just to show you the kind of size. Of the models that we're working, with. Once we have our construction. We can then start to simulate, the behavior, of our bacteria. Using the models. We do this, using, an approach, called flux balance analysis. This is a constraint, based approach. So before we have any constraints. In the network. We have an unlimited, amount of flux, or. Flow of metabolites. Coming into the system, or being produced. Out of the system. And so we have an unconstrained. Solution, space as is highlighted, in this three-dimensional. Figure, here. By adding constraints. So a fundamental, constraint, for flux balance analysis, is the steady state constraint. But also adding constraints, such as substrate, availability. So here highlighted, in the diagram on the left, constraining, how much of your substrate, is coming in, and what this does it con it reduces. Your. Solution, space your feasible, solution space, to a convex, cone. So now any feasible, solution, flux distribution, through this metabolic, network. Has to lie somewhere, within this convex, code. Now, ideally, we'd want to explore the entire, solution, space. But due to the size of these, models. It's very computationally. Expensive. So what flux balance analysis, does it. Maximizes. Or minimizes. An objective, function. To identify. One optimal. Flux distribution. And very often. In this area what is used as the objective, function, is that biomass, reaction, that i previously, mentioned. So this is under the assumption, that the bacteria, will want to, maximize, their cellular, growth. So hopefully that gave you an idea, of what genome skill models are and. The flux balance analysis, method. Now i'm going to. Provide, a, few examples. Of how. We're using, these techniques. To try and gain more information. About. The bacteria. That we're interested, in at the sprc. So this first example, is actually, some work that my colleague, carried out rupert, norman. He was he carried out two simulations. In the first one on the left, he's, mimicking, carbon limitation. And on the right he's mimicking. Non-carbon, limitation. And comparing, what the uh. Product, uh formation. Is in the two approaches. So clostridium. Autocytogenum. Is the bacteria, rupert was using. And this can grow on carbon monoxide. So what he did in the left, example. To maximize. Uh to do to simulate, carbon limitation. He maximized, that growth rate as previously, mentioned. And increased, the carbon monoxide. Um along the x-axis. And as you can see the main product, being. Formed, is acetate. And then co2. However, then when he simulated, non-carbon, limiting, conditions. This is where he's now fixed his growth rate, his biomass, reaction. And still. Increased, co, so we've got excess. Carbon monoxide, in the system. And now what we see is a, a switch between the products being formed so a switch between, acetate. And ethanol, being produced. He then took this further by, did the same simulation. But continued, to increase the carbon monoxide. Uptake, rate. And, now he then found, found hydrogen, being produced, as the main product.
However. It's known that hydrogen. Has a limit, on its, production, due to thermodynamic. Constraints. And so when rupert, constrained, this in the model, he then repeated, the simulation. And found. Butane dial being produced, and this was a really nice result because this has not been found. Using any other, uh genome scale model of sea auto, and butandal. Is an interesting, product. To try and produce. And if anyone's, interested. In, looking at these results. In more detail. Rupert's actually already published this work in engineering, biology. Just last year. So now moving on to the work on cooper vedas next. Uh this slide is just to give you an idea of the kind of data that we are collect. Collecting. Uh for. Uh, using with our models. This is just for fructose, but we're gathering, we've collected a lot of this data now for, autotrophic. Conditions, as well so growth under co2. And hydrogen, which coopervedas. Can grow on. So, for example, we've been collecting. 13c. Metabolic, flux, analysis. So this is carbon, tracing, experiments. This is limited to just small networks, however we've used it, this is for this, picture that we see here is just central carbon metabolism. But we've been using it to validate, our genome scale model. And as you can see in the left hand uh sorry the right hand plot here, we have a reasonable correlation. Between the two. Two approaches. We're also, very interested, at the sprc. In gene essentiality. Of these bacteria. We want to know which, genes we can knock out it's feasible to knock out, um. In terms of increasing, our product yield and it can also be helpful in actually identifying. Growth coupling. So we've actually. Predicted, gene essentiality. Using our using our genome scale model. And we've compared, this to an experimental. Technique, for identifying. Gene essentiality. Which is called tradis. And so this this table here. Shows the number of true positives. False positives, etc. But the overall, accuracy, is around 80. Which is reasonable, for a bacteria, that's not, um, too well studied. So then we did a similar thing to what rupert did with, sea auto, with cupra vedas. So cooper venus, is. Um. Is is a really interesting, bacteria. Due to the fact that, it can produce, as a storage, compound. A polymer, called polyhydroxy. Uh hollowed. Polyhydroxybutyric. Acid or php. And it produces, this as a storage, compound. Which is actually also. A biodegradable. Plastic. And it produces, it when it has excess, carbon. And a limit, in some other, nutri, nutrients, such as nitrogen. So again we try to, simulate, this using our genome scale model, very similar to what rupert did, we're fixing, the growth, rate. And we're forcing, fructose, so forcing, excess, carbon. Into our. Model into our system. And then looking what products. We're producing. Due to that excess carbon. And what we have here is a lot of pyruvate, is being made, and then initially, we have some acetate, being made. However, from our experiments, from the espresso, experiments, we know that acetate, isn't produced very high quantities. And we're missing that php. Being produced. So what we then tried, is integrating. Omics, data, or particularly gene expression, data, to see if we could get the model to. Produce, more. Accurate, phenotypes. And what we're doing here is where. In the simulation, in the flux balance analysis, simulation. Where penalizing. Reactions. For carrying, flocks. If the gene, if their corresponding. Gene. Has a low expression. In the in the gene expression, data. And as you can see here, we now have, we still have, pyruvate, being made, but we also now have our php, being made, so this was a nice example. To show, where. Using omix, data can help improve our model for, predicting. More accurate phenotypes. And this is just preliminary, work but we're hoping to publish this really soon. And then as a final, example, this is actually an example. Where we've, increased, ethylene, production, in e coli. It's not an example of a c1, feeding bacteria. However it's a nice example, where we've gone through that cycle, that we showed that i showed at the beginning. To increase the product. So in this in this example, what we were doing here, we were inserting. An enzyme, the ethylene, forming enzyme that's found in found native in other bacteria. Inserting, this into e coli.
And Then we use the genome scale model of e coli. To, simulate, gene knockouts. Using, two approaches. One was flux the standard flux balance analysis. And another was an alternative. To flux balance analysis which is called, minimization. Of metabolic, adjustment, or moment for short. And using these two approaches, we had two candidates. The first candidate. Was knocking out the gene, suck a. This. Is the reaction, here where we've got alpha ketoglutarate. Going to sucks and alkaway, which is part of the tca, cycle. Um. And what we found using the modelling, is that what's happening here. Is that the, um, alpha, the blocking, the block through this flux here, means that you get higher, alpha-ketoglutarate. Availability. And so then, redirects, flux through this efe, enzyme. This was tested in the lab, however, and we found a mixture, of some strains produced in ethylene, and some not producing, ethylene so it wasn't 100. Guaranteed. We then tested. We then looked at the second candidate, that we had that was actually identified. Via, the fba. Approach. And also the moment approach. And involves, knocking out either, of pro a or probe b. These are, genes corresponding. To proline, biosynthesis. Pathway. And results, in. Uh blocking, any flux towards proline. And the only way of restoring, this flux is then again by this efe, enzyme. This was carried out in the lab by samantha, bryan, and her phd, students. And they were able to. Uh. Find, a. Triple the amount of ethylene, yield, using, this knockout, so we've got two plots here. Where we've used two different plasmids, the second plasmid, has a higher copy number, but both result, in around, three times as much ethylene. We then did, a random, mutagenesis. And enzyme evolution, to see if we could then further, increase. The yield of ethylene. So the top plot is the random mutagenesis. And the bottom plot is enzyme, evolution. And again this resulted. In around, double, the amount of ethylene, being produced. What we then did is we. Selected, the strains, that had the highest amount of ethylene. And we resequenced. Them to identify. The snips. That could potentially, be causing this increase, in ethylene. We also collected, metabolomics. Data. And we also then, went back to the genome scale model to carry out what's called a flux response, analysis. So this is where we, increase, ethylene, in the model. From zero to the maximum. Amount of ethylene. Ethylene. And we look to see how the rest of, the reactions, in the model. Are responding. So this. Plot here this figure here, just shows, the central carbon metabolism. Where the blue reactions, are those that, decrease, to increase ethylene, and the red reactions, are those that increase. With increased ethylene, and we're particularly, interested. In looking at how. The the, reactions, that corresponded. To the genes. That had a snipping, how they responded. And we also compared. To the metabolomics. Um analysis. And the general conclusion, from this analysis, was that. The snips were potentially. Causing an increase. In the alpha-ketoglutarate. Substrate, availability. So this is a substrate, to the efe, enzyme. And we speculated. That, the snips were potentially. Disrupting. The ammonia. Simulation. And regulation. Pathways. That allowed for this increase, of alpha-ketoglutarate. That is otherwise. Quite tightly regulated. So hopefully that's given you, um. Some examples, of the kind of modeling that we're doing at the sbrc. To help the experimental. Work that we're doing. But just to summarize. We have, a number of different genome scale models constructed. Within the sbrc, in collaboration. With. Different universities. In particular. With the, collaboration, with oxford brooks university. We also. Are working, on methanol. Genome scale models. And also. Acetobacterium. Woody eye genome scale model. We've also collected. A lot of data for refining, these models. Making them more condition, specific. In particular, we're very interested, in there, and we've constructed, we've carried out a lot of tradis, experiments, for our bacteria. Which as i mentioned, is useful, for. Knowing which genes we can and can't knock out or potentially, helping. Predict. Growth coupling strategies. We've also now collected, a lot of omics, data. In particular, for autotrophic, conditions, with probe we've got proteomics. Transcriptomics. And metabolomics. That we're now integrating, into our genome scale model, making a more condition, specific, model. And then finally. Finally we're also, very interested. In developing. Modelling, tools, for helping. To analyze, these genome scale models. We have a tool published. Already, called json, modules. You can find this available, on the github, link provided, there.
This Is basically, a tool for, managing, genome scale models. As i showed at the beginning, of the talk that cycle. Where we're constantly. Feeding, new experimental. Data into the model. Making, changes. To, to adjust, to the results. And, doing further, tests to validate, the model, and we want to make sure we don't break, any other, validated. Tests in doing so so this is a nice tool for managing. The the changes, that you're making in these models. Are also very interested. In developing, tools for integrating, thermodynamics. So, ensuring. Ensuring. That any. Flux distributions. We predict, using our models, are thermodynamically. Feasible. And then since we have a lot of omics data that we've collected, now we're also very interested, in developing. Novel ways, of integrating. These tech these, this data. Into, our genome scale models. So thank you very much, for listening. There's a lot of people to thanking the sbrc. And also, our collaborators, dr brooks, car student, the kaist institute. The supervisors. And experimentalists. And. Funders. Um so i'll just leave that slide there for a minute and also you can find my, email, address, there. Thank you very much for listening. Okay, wonderful, these were, a couple of wonderful, talks thank you very much a, christian, and nicole. We have quite a, lot of questions, so. Without further ado i will start. Reading them. So, the. First question. Is. For christian. And come from joshua. Bitton. With supplementation. Of linoleic. Acid, prior, to infection, with cov19. A with mitigating, the severity. Of the symptoms. Um. Yeah i mean obviously. We take up a lot of linoleic, acid already with our food, so the. Lenoric, acid is a vitamin. And there is lots of linoleic, acid in seeds nuts and fish and so usually with a normal diet you have. Enough, um. But it seems like once you are infected with the virus, you quickly, deplete. The linoleic, acid. Pool and especially in severe, disease, these. Patients. This can even, be measured, in this era, of. Of, of the patients. And so, our idea, is not really that prior, to infection, you should use la. Uh but. Once you have been exposed. To somebody. Who, like basically, when the tracking. The tracing, app shows you that you have been in. Exposed, to covet, 19. That then you would maybe. Uh, take, linoleic. Acid. And it is probably, not global, with food, where you will have to absorb, it and. Fatty acids are not very soluble, so. Intravenous. Or with the food it's not very efficient but it would probably be better if one could up. Could apply, it with a nasal spray. Or. Or maybe a mouthwash. Basically, to to the infected, region, should be directly. Treated with linoleic, acid which makes it potentially, easier. Having, said that that's a proper clinical, trial so we need to do the toxicology. We need to get proper dose response, curves. And we and, cell cultures, are not enough so we need to go into animals, to test the idea, that linoleic, acid is a drug and find out how good it is as a drug. Okay. Say thank you for that, a a, related, question is the following. By a adrian. Banzer. Linoleic, acid decreases. The kd, of spike, to ace 2, by a factor of 2.. Is this slide reducing, affinity, sufficient, to explain. The in vivo effect of suscovit. To replication. Or does linoleic, acid have additional. Effects. There are, additional. Effects, of. Linoleic. Acid so. A we don't, really. So this has been. Done, in vitro, and we don't really know, in vivo how much linoleic, as it is there so it could be that the concentration. In cell culture, and afterwards, in vivo. Is a completely, different compared to the one which we use. With um. Epithelial. Cells that's my first one the second point i mean linoleic, acid is metabolized. In the body, and it is incorporated. In membranes, as i said, and one thing the virus does when it enters the cell is actually. Activate, a phospholipase. And this phospholipase. Then helps, the virus, to remodel, the biomembranes. And make these compartments. In which it replicates. And so it needs these four fatty acids. Um. The phospholipase. The cytosolic, phospholipase. It activates, doesn't only release, linoleic, acid from the membranes, and damage, the membrane. But linoleic, acid, also then, at a certain concentration. Starts blocking, the phospholipase.
So This could be an additional, effect, of linoleic, acid and there's a third potential, mechanism, where it which is too complicated, to explain, but basically, it blocks the aggress, of the virus when it comes out, so it's it's much more complicated, than just blocking, down the infectivity. I think we just. Need to do more experiments. Um in order to fully understand. Where actually linoleic, acid. Is important. And then this is such a versatile, molecule, that it's really difficult to grab all the different functions. Because as i said it's metabolized. To other hedonic, acid, and prostaglandins. So there is a lot of downstream. Biological. Activity, of that molecule, as well. Okay thank you and now we'll ask one more question, for christian, then we'll move the question for nicole that, people are also hearing. To hear her response. So. Following, on on that. Previous response. The question, from lindsay. Mcdermott. Is whether you have tested. For binding, of other, polyunsaturated. Fatty acids to the suscovit. Spike protein. Yes we get a lot of these questions. And there's a lot of modeling, in the field to kind of see what else fits. Into that hydrophobic. Pocket. So we have actually. Tried, to. Bind, also other molecules. Um. To the the receptor, binding domain but it seems to be not that simple. So. For me. This. Means, that whatever, we have modeled. May fit nicely, into the pocket but there is a high energy of actually, opening, up that model. That pocket and getting the molecule. In. And. Maybe, our essay doesn't, really monitor, that nicely so the only thing we could bind up to now to the rbe, is knowledge asset. Um. But, having said that we are not entirely, sure that the pocket, can. Like even in vivo we are not sure whether arachidonic. Acid or other. Unsaturated. Fatty acids couldn't bind as well it's just that linoleic, acid is what we found. Because it was in our media. Also we have, done the modeling, and other molecules, fit in there, as well but it could well be that these molecules, need to be present, when the spike protein, is produced, and folds. So in the cell and not once it's out um, and and fully folded, so we there is a, lots of additional, work to do to understand, how we can make an antiviral. And what is exactly, the specificity. Of that pocket. Okay, thank you thank you very much if we have time we'll come for more for more questions that we have in the list. And, nicole i have a question. For you from. Becnus. No rousey. He's asking whether the. Certain, c metabolic. Flux analysis. Was done, with labeled, fructose. Was gene essentiality. Prediction, done by having, lower growth. And where your experiment, performed. Performance, batch fermentation. In bioreactor. So essentially, three questions in one. Okay so sorry so the first question was about the metabolic, flux the 13c, metabolic, flux analysis.
If It was. Labeled with fructose. Yeah labeled with fructose. In batch culture. Okay is that um. Yes, that's that's part of the question the other, is, is asking whether gene essentiality. Prediction, was done, by having lower growth, or. Or some other bioinformatic. Methodology. So in the genome scale model or in the tradis, data. I suppose in the gene scale model. Uh the genome scale model it was just. I would. Uh, fix the growth, rate, and then just look to see if by knocking that gene out you can get any flux. On or on any growth rate so, usually, it would be that you can't get flux on any, any growth at all. Under fructose. I don't know sorry. Question. And the. And the, last part of the question was whether the experiments, were performed. In batch fermentation. By your react. So. Initial. Initial, experiments, have been done in batch but the. Analysis, the results that we've got now have been done in keemstar. Okay thank you. Another, question for you nicole, from. Ian hunter, actually there are several, questions in one. How define. Is your composition. For biomass. For each bacterial, species. And he, goes on to say. That presumably. The content related. To each amino acid per biomass, dna, rna, base, fatty acid lipid etc. So. It's different, for different for the different bacteria. So, the cooper venus. We've actually used the biomass, that's already been published. In a paper by, a. Part. It's a korean group. I can send if they're interested i can send them the paper, but i just use the biomass, composition, given in there. For the. Clostridium, auto ethanol genome, sbrc. Constructed. Experimentally. Um. Obtained, these values, for the biomass, composition. And for eu bacterium limosum. The, um, the kais group the, korean advanced, institute, of science and technology, they did that their end, they they calculated. All the experimentally. Calculated, the. Biomass, composition. Okay. Great thank you. Another question for you nicole, from. Kevin thomas. When you apply your, expression, data, how do you use to, uh how do you use. Do you use this to limit, fluxes. Were you tuning, reactions, on and off or just limiting, total flux, depending, on expression. So, there are there's different ways that you can, that you can, integrate, this, expression, data so there's one technique that we've been exploring, at the minute. That i can't give too much details, on because we've not it's a novel approach, and, we've not uh, published it yet and it's also with collaboration, with oxford brooke so i'm not sure how much details i can give, but another. One of the approaches, that we've also tried. Is a pro is an approach which i'm not sure he's familiar, with, but it's called um, gimme, it's a method in already published, i think by, um paulson's, group he's a, big gene and scale modeling, guy. In the us. And basically, what it does it doesn't, um. It doesn't weigh all of the reactions. According to the gene expression. But it actually just. It basically, turns off you give a threshold. Based on you you set that threshold, so it might be that, you look at the. Uh i don't know, the quantile. For example the tenth. Uh, or the 25. The 25. Of the data. And anything, anything below that, you cut it off, or whatever you've decided, that should be a threshold. But then what it does because you can't. Have a then solution, is probably some reactions, that may only have. Very small. Only require, a gene that has very small gene expression. What it then does it allows them to carry flocks, but they'll have a penalty. On the objective, function to be on or in the flux balance analysis, to be on. Okay, thank you very much for, for that question. I have another, a question for for you nicole.
In The in silicon, analysis. You. You do knock ins and knock down, analysis. But this is a, combinatorial. Problem very rapidly, can get. Quite quite hard, so how do you deal with this, combinatorial. Explosion. Especially, with the noxian. Both for the simulation. And then for a certainty. Ascertaining. Which, potential, solutions, to carry on experiment, experimenting. With. So you mean because, if you were to go above like, two knockouts, i think is what they're meaning so in terms of that, so the, the example that i showed was actually only exploring, double knockouts, that's quite straightforward. To just loop through your model and do that, however, in order to then do you know if you want to explore. Five. Five genes, being knocked out, what we can then do or what we've explored, techniques. Of using. Is optimization. Approach, uh, optimization. Approaches. That use genetic, algorithms. Such as um. A tool called. Opt flux. And this uses a genetic, algorithm, to try and it'll look for a, gene to knock out it'll search, for. If one gene. May provide you a small amount of flux it'll then build and build up on that until you get. A. Strain, that's producing. Um. More flux based on how many, genes you've knocked out so it's it's not an exhaustive. Search of the entire solution, space, it may get because of that it's a genetic, algorithm. You may get stuck in a local, optima, it's not necessarily, going to be the, the highest. Producing. Strain. Um. However, it's it's a way of, getting around that, problem there's also techniques, such as, opnoc. Um, is another tool. In the in the area. Um this just which i think is a mixed industry linear programming. Tool that allows you to then, uh try and solve this issue of finding. More than just two. Knockouts. Okay, wonderful, thank you very much. I have a, few more questions, for. Coming up for a. Christian. So, this is from. A. Maitre. Chief kumar. And says, that. Would you expect. The drugs the target, that target the hydrophobic. Pockets. In the rbd, to be sufficient, to block infection. Or would they have to be used alongside. An antiviral, like rem de severe. As a synergistic. Effect. That's of course, excellent, questions we don't, that we don't have an antiviral. There so one encouraging. Uh, fact. Also it for future pandemics, with corona, viruses, that they said the pocket is conserved. So. It could be really useful to target the pocket but then there is always, the danger that there, is a mutant, which makes the pocket, resistant, to close the pocket, a single. Negatively. Charged, residue, would potentially. Be enough at the entrance of the pocket to completely.
Interfere, With entrance. And so, therefore. I would think that antivirals. Should be combination. Therapy, like in hiv. Where you where a single mutation, cannot, make the virus resistant. Against the therapy. So i yes i would think that we should, kind of envision. An idea therapy, to consist of more than one molecule. To avoid the coronaviruses. To become, resistant. Okay thank you and related to that question. Another one from jack. Star. Is that, you showed experiments. Uh combining, the red. Remember. Severe, with linoleic, acid. Have you done any experiments. Uh using, other, combinations. Drugs with for example dexter metasome. No we didn't, but. What is quite. Interesting. Is that it seems that dexamethasone. And other. Glucocorticoid. Steroids, seem to fit really nicely into that pocket. Um no we didn't do the experiment. But it's something, to. Be done. Okay, and i have a here the last question. Uh, for the evening it seems. Uh how selective, is the binding, of the fatty acid analogues, drugs to the targets. Yeah. So the fatty. Acid, analog, drugs. Yeah unfortunately, there are none. Um there's, a. A very quick question i don't know a quick answer to a, question. Okay. Okay with this we will, uh we'll finish, this is the, the last question. I would like first of all to, uh, thank the speaker, for two wonderful. Presentations. And and and really nice. Session of questions and answers. And of course the audience. And i will invite you, all to continue the discussions, online via the twitter, handles of the biochemical. Society. And portland, press. Just to. Mention that it's not a coincidence. That. We are focusing. Uh, this. Webinar, series, on the area of synthetic biology, this one, now and the one next week. Uh. Because. We were supposed to be having the synthetic, biology. Uh uk, 2020. Conference, meeting, in nottingham, unfortunately. Uh, this didn't take place due to the pandemics. But we are all looking forward to a brighter, future so you can now register, your interest for sbuk. In november, 2021. In the biochemical. Society. Website. Um. There is a there is going to be the follow-up to this. To this webinar. Next. Thursday, 26, at 3. With a topic of developments. In industrial, biotechnology. And again you can register. To that, website. To that webinar. In the society, website. Finally. I would like to, to, to invite you to, join the biochemical. Society. Community, it has great resources. For. All career, stages. A grants. A number of discounts, for conference, in some meetings, so it's a good, opportunity, to check the website, and and see whether, it was. Joining up, with this i will finish here thank you all again, and have a great weekend, bye.