 OK, so how we would like to organise the discussion is that we want to have some specific questions from Nico first, and then specific questions for Luis, and then we'll go on with the discussion. So specific questions relating to the technical aspects of the papers, etc. So first of all, questions for Nico. I have a question about, I apologize, I have a mathematical discussion about my question, it's very nice. And the technique you do is to make a modification of the situation of your vehicle. It's just by putting an enzyme, or the cause, there's some kind, and with some density. And you expect that statistically it will evolve in this direction of this one. But in a legal sense, probably there is something which makes the enzyme much more local in its action, and it's because it interacts with other actors. So do you try to develop this kind of interaction between local acts? I see your point and it's not at all naive by any means. So the targeting of the glycosyl transferase is actually the key point in getting this pathway efficient. And so several of these glycosyl transferases in the Golgi do interact with each other, and it remains very poorly characterised how they do that, but we know which parts of the sequences are involved in doing that. And so when we fuse human glycosyl transferase catalytic domains to the targeting signals too, the signals contain the information where the protein will end up in the cell. This is in part of the protein sequence, and the structural basis of that is not understood, but we know that it's in that sequence that the information is somewhere. And we use that entire sequence from the yeast glycosyl transferases to fuse it to the human ones, to try and do exactly what you said. And it's a critical point. If you don't take it into consideration where you make that fusion, you get like a bit conversion, maybe 10% from one substrate to the product, but you'll never get 95%. So we took that into account in the design. That's one of the things that took most time actually to optimise that. Because you're so proud to discuss all of mathematics, in fact. But at this level, when you try to manipulate several actants, perhaps you need some control of mathematical control to explain the data. So I can comment on this, isn't it? I've been thinking about this as well. How did we manage to get all of this done without any mathematical modelling? And so I think one of the reasons is that this is not a very... I'm going to have to be careful with what I'm saying, but it's not a very carefully controlled pathway in the sense that this happens in the goji apparatus when the protein is on its way out of the cell. So whatever happens to these carbohydrates to a very large degree of approximation now will not be toxic to the cells very much, because the cell is in the process of throwing it out anyway. This is taking abstracting a lot of biology as to why that may not be the case. But you can consider it as a secondary metabolised pathway. It's not very stringently regulated. There's no feedback or very few feedback inhibitions in that system. Which is probably why just pumping up the levels of enzymes that works, you know, you get all the products, but there's very few products that actually do feedback inhibition in previous steps, which is very different from primary metabolism. So that's my take on this. May I say something directly on that? I think that a clinical biology is not in the attention so much of mathematical modelling. As far as I can see there would be a need for that, but it would be a great help if the biologists could identify topics where they may be useful for. And I think if you need a model, for instance, to do good experiment inside, what you cannot, because you have to have an understanding of underlying process. So I can comment on that. So what I was just saying is about the steps of the pathway that happened in the goji. All these goji enzymes need sugar nucleotides, and the sugar nucleotide derived directly from primary metabolism, that happens in the cytosol of the cells, and there you do have these typical feedback control mechanisms that you have in other aspects of metabolism as well. And so there certainly, if we would have to optimise, if you had to build in new sugar nucleotide metabolism, I think that would be a very fruitful area of modelling. But we should take some time outside of this current. I think it's not so very well aware in the community that there is a really open field for it. Okay, yeah. Other questions from Nico? A brief question. Do I could elit for us done in yeast or in an animal? It was done in mammalian cells in yeast and in plants. Can you introduce this into any mammalian cells? Well, we've done this in hecton and tree cells, human cells. We've also done it in CHO cells by now. And I think there's basically no reason why it couldn't be extrapolated, except perhaps in fully differentiated cells. Cells that have to make up a particular tissue. I think there it would be more difficult because that's where carbohydrates play an important role in cell-cell contacts. I just want to say it was a great talk, very much enjoyed. Just playing devil's advocate here. Could you not just introduce, let's say, a clickable amino acid, unnatural amino acids, and then attach whatever carbohydrate you would like in which position? You could. And some people are doing things like that. But the point is a little bit, one has to generate the native linkage for most purposes. Because some of these applications are going to be for therapies that have to be injected lifelong, so any suspicion of immunogenicity would completely kill your project. I think that's the problem. Diagnostic, too, is really no problem. For therapeutics, I would be a bit more concerned. It's a general question, but I wonder if you could comment. The diversity of glycans is phenomenal, and it seems to be overlooked a lot in molecular and mechanistic understanding and biology in general. There's a lot of energy required to build all these chains for themselves. Do you want to comment on the trade-offs between that in general terms? I think it relates to probably the question as to why, in evolution, you get this huge diversity at a substantial energetic cost. I think carbohydrates, so when I show this to my students, if you look at an electron micrograph, for example, of an erythrocyte, red blood cell, the top 15 nanometer or so is all carbohydrate. For anything that approaches a cell in general, the first thing that is encountered is a layer of carbohydrates, which you call the glycocalyx of the cells. That's why so many viruses and pathogens in general recognize the carbohydrates first. I think that's also where the diversity comes from. You're at the battle front between pathogens and mammalian cells. It's well known in influenza, for example, that influenza virus in birds will recognize a different linkage of cialic acids than the one in mammalian bacterial cells. Actually, the WHO is now using that to monitor new viral isolates for sequence variants that threaten to become infectious for humans. I think that's really where it comes from. I think there would be a chance to use mathematical modeling in this case, because I think recognition of cells very much goes on this phase. But it's not done. In fact, it was Ed Dell who told me we should look at that. It could be in some problems, which we have with our immune system, that layers of cysts dies out, destroyed. I'm too naive, but it could happen. I think you're very close to the truth. Yeah. So you mentioned in yeast that it has these gigantic high-lanos clusters. And yet with this OCH1 mutant, you obliterate all of that. So what's the phenotypic consequence of that? It seems like you were just saying that there's a tremendous energy investment in there, and it seems like cells should do better. So different yeast species differ as to the extent to which they are dependent on that particular layer of the cell for their stability. So if you take sacramascia cerevisiae, which most of us use in the paper's yeast, that has a very thick layer of nanocortiles. So there the carbohydrate is really gigantic, up to 200 mannose residues. Now, one of the reasons for us to choose pichia pastores is because by nature it only has like 14 or 15 mannose residues. So if you kill that, the cell is going to be suffering much less than a sacramascia cell in which you kill the same pound away. And so in pichia, when we do that, we reduce it from 14 or 15 mannose residues to 8, which is not that much of a difference. And then the yeast cell doesn't suffer at all. At least you have to be a little bit careful, you have to keep a few percent of the activity for it to be happy. But it doesn't suffer so much as sacramascias would do. So it's a question, I mean, you have these pathways, and we think they're conservative, and the Thungal Kingdom may have as much diversity as we would have in any kingdom of life. So you can choose a bit. Okay, two quick questions there, because I want to get to this at any time. No, I was just wondering, you removed the glycorization pathway from eukaryotic system, but what would it take to get a similar production system in bacteria? Right. Yeah, I don't know. Okay, so I think one of the leading people in that field would be Marcus Ebi at the ETH in Zurich. I did my post up there, where we were trying to rebuild glycosylation pathways in bacteria. And you can do that. You can actually attach carbohydrates, not of this nature, but of bacterial nature. There's some bacteria that actually put glycans on asparagus, and the most famous one is countulobacter geogeni, the phypatogen. That puts carbohydrates on their own completely different nature, but it never has this folding catalyzing role as you find it in eukaryotes, you see. So there's been co-evolution of these glycans with the chaperones that help in folding proteins, and that has not been implemented in E. coli yet. That would be hard to imagine how to do that. So these kind of proteins don't fold well in E. coli? No. Okay, so quick question over here then, Luis. Nico, if a similar question, what is this protein catalyzing role if you can cut off almost everything? Well, so the clue is that we cut it off after the folding has happened. So we keep everything intact in the ER. They're all folding happens. So you need the ER, and you need all the reactions in the ER? Yes. You cut it off, and then you have a single clip neck, and if there's any contacts between the glycans and the protein surface, it's through that first version. That's why you keep it. So there's been this approach with antibodies to produce other like-cosylated antibodies in E. coli, and to engineer even the FC. So they can be produced. So how you compare your strategy for producing a therapeutic protein in these yields, which are like-cosylation compared to and struggling to produce a form, a variant of the therapeutic protein black in the blood like-cosylation? I mean, that's a good point. It can be done, but the yields of these strains are not very good at this moment because it's not just the carbohydrate which is important. You need all of that sulfide bone formation in the appropriate way to make full monoclonal antibodies. So you can get some molecules out to my feeling there's a long stretch to go before it becomes competitive to you carry out the expression systems for antibodies. Okay, so let's take some questions for this. You have a question about the way that you moved the operands of the synthetic injectosome into the non-pathogenic E. coli. It seems that in the original strain, these are grouped in 35K kilobase pairs in region that you have scattered as single operands being scattered around the E. coli K12 genome. So I was wondering, first, what were the reasons for this source? And second, I'd like to know what's the production of injectosomes if you have been able to measure it and compare it between the synthetic injectosome and the natural injectosome. Okay, so the reason to amplify in different operands was to remove sequences that were not wanted, especially regulatory elements and infectors which we don't want to inject any of those except the ones that we choose. Okay, so we introduce this in the chromosome and we compare the expression of the proteins. The expression is very similar to the one in IPEC using this approach. But when you compare the assembly of injectosome, we produce like between 50% to 70% of the amount that you obtain in IPEC. So it is reasonable and good. It's not better than IPEC but it's difficult to achieve what nature has done in single shot and first shot. But it is very similar to the amount of injectosome produced by under-induction, optimal induction conditions. That bacteria, because they are lining it too, do you think they can adapt to the function you ask? It is a type to domesticate them in some sense. What is the reward for them to do that? Well, I have probably... This is a good point because it means what is the reward in nature? So probably you don't have to... Yeah, so the thing is that you will have to... So once you engineer this, probably there's no reward for the bacteria initially. And the thing is that in the laboratory, there's no competitors. And so it means that you can do all this type of manipulation and it's like an isolated system so the bacteria do not compete with others. However, we always compare the growth and the viability when we introduce this type of contrast and we see that there's no... We do not affect these basic things like growth, rate, or viability. And it happens also with the type III. So bacteria is not growing worse. But in any case, the point is that probably... This is not something that the bacteria acquires like in nature, a trait that is acquired to fit better into the environment. It's something that you will force to express. So probably this bacteria will be adapted only to a niche in which they can really compete with whatever is found there. But you have to make adaptation into engineering strength so they have also the ability to compete with, for instance, the microbiome that is found in your gut. So we have used K-12 because it's a string... It's the one biggest trend that we use more in the laboratories and it's an easy... It's a micro-organism that's easy to manipulate but it's not a good colonizer. So the metabolism is not good for competition in the gut. But the good thing is that you can move these modules toward other strains that are better competitors depending on the application, for instance, if you wanted to colonize your cash-intestinal tract, for instance, that can compete with other bacteria that are found there because otherwise they will not be able to do the function that you want to do. So that's a good point because it has to be taken into account that these bacteria, the intention is to use them in vivo. They will need to compete in the environment and then, at the end, they should be able to survive in a hostile environment. So you have to introduce traits also to improve the metabolism toward the environment which they lead, for instance. I have a question. We are modeling the cellular process. It's a chemo-ductic movement. Now, if you introduce an adhesive, not natural part, does it guarantee that this chemo-ductic movement, the flakela, is stopped or is it an independent information which it gets or they are no coupling? So, yeah. Apparently there's a coupling between adhesion and flakela. Yes, normally it is. But in our strain, it is. So once the bacteria attach to a cell, the flakela stops rotation. So in fact, we require rotation of flakela, a good flakela, to see a good adhesion. If we don't have a good flakela, flakela, flakela, and adhesion is reduced. It's something that I haven't shown. So it somehow, I mean, the bacteria somehow needs this movement also to make good conduct with the adhesive. But it is connected. Somehow it is connected, the adhesion, with the stop on the flakela, even with this system. This is what I'm asking. Naturally it is, but in your case it could be that it doesn't give the chemistry, which maybe is necessary. Yeah, if you see in the video, and if you see how the bacteria, so this K12 strain, it swims like crazy. So it moves around. What when they attach to a cell, they stop. You see clearly, only with your eyes, they stop. No, no, with this. So you see it. You see the entire thing and then you see it stops. So this somehow, the flakela has re-configured to position its non-rotate anymore. It's a signal. To signal the GMB or? Okay, from the video, beautiful, beautiful video is actually amazing. The bacteria seem to be also interacting with each other as if there was some sort of biofilm sort of state being induced by the adhesion or the initial set of bacteria. Did you observe that? Do you start inducing quorum sensing? No, we haven't test that. We haven't test if there's quorum sensing induced or we haven't put a promoter that responds to that. We have done some transpectomic analysis to see what are the genes that are induced soon after the adhesion, but what I mean soon means, like in the first 20 minutes. Just what happens in these first 20 minutes and the genes that we have found are more related to oxidative stress for instance. So we haven't seen anything that indicates and also with the lactose simulation. So it seems that it's more a metabolic thing related to the contact with our cells to more cells than quorum sensing. It's at every stage, but maybe if we wait days, then maybe you can see this type of induction in hours, but not certainly in days. I have a question here. I was wondering, one of the possible applications would be that you target the cancer cell and the bacteria would destroy the cell? Yes. And if you do so and if you're successful, the bacteria could learn, if you do so there's plenty of nutrients that can grow very well and then attack other cancer cells hopefully or other cells, is there any mechanism to stop them? All right. Yeah, well in principle, this is like, that's the point of introducing all these engineering non-pathogenic bacteria. So you can control more or less the antibiotic sensitivity of your bacteria. And in fact, one of the reasons why we introduce everything in the chromosome and remove all antibiotic resistance is that our strain is sensitive to all antibiotic. So it's a way to guarantee that you can always treat the individual, for instance, with antibiotics, with the bacteria. But usually they concern, although this is what the people think, so you have to somehow control. Usually what worries me is how I ensure that the bacteria kills all two more cells, for instance. That's totally more, less likely that this happens than the bacteria spreads and generate a systemic infection that you can control. And of course, you can always induce the toxic gene under certain environmental signals. So you, for instance, can induce with a molecule that is given systemically, and you see any adverse effects. You can always uncover the production of the toxin simply by not giving this inducer. So there's different options. So I don't think the problem would be to control the micro. Probably that will make it more efficient and be sure that you will be able to heal what you want. So you showed that you could use your type III secretion system to inject your nanobodies, for instance. Did you have any specific application in mind for doing this? Okay, so, yeah. There's nanobodies selected against what is selected by, for instance, a group of hand-get emits, again. In fact. Who, I've been talking to him because there's, of course, the option to target specifically some signaling cascades involved in self-proliferation or other processes with nanobodies. So you could, for instance, inject these nanobodies against this type of molecules, like P503B or kinases that are involved in cell survival and then block the activity. So it is an option. But I'm considering not only the antibody molecules for delivery, as I said, so I think there's many other alternatives to the natural effect of these pathogens to certain functions in the cell host. How big of a protein can you deliver? And you mentioned that IPTG, you thought was a problem, as an industry, could you say why it's a problem and what you look down to you? No, IPTG is fine in the lab, but it cannot use it even once. It's not allowed, because it's a carcinogen. So you have to use other molecules for induction. I mean, you can do it, but it's not something you should do because it's molecular, it's toxic. So usually the people, there's other groups that also are using arabinous, for instance, as a sugar that can be used. And other types of antibiotics like tetracycline, and oxycycline, which are induced to that. And even you have promoters that are induced by molecules like salicylate. So you can induce your genocenter with aspirin. So what I mean is that there's options to molecules to be used. And I think there was another question. The size of the protein. So more than the size, what matters is the folding of the protein. So for instance, this tear is a large protein. It's almost 90 kilodaltons protein. And it can be easily translocated, very efficiently. GFP is shorter. It's only 30 kilodaltons, but GFP cannot be translocated through this complex particle. It falls previously and it blocks the translocation through the injective zone. So it depends more on the folding of the protein than on the size. Or on the unfoldability? Or on unfoldability. Because there are 80 paces that are associated to these injective zones. Unfold the proteins through the translocation process. So they also have a function of folding the protein. So the protein probably should not fall rapidly in the cytosol, otherwise it will not be efficiently translocated. So how worried should it be to re-design bacteria, exchanging genetic bacteria with bowel populations? Do you think, for instance, if your bacteria acquire resistance against antibiotics, it could be more difficult to maintain? Yes. So that's another classical question. Let's say there are no concerns. No, it's true, it's true. It's a good point. There is the possibility, once you release micro-organisms to the environment, that it will change this is something that bacteria has done during the evolution very efficiently. And so there's ways to try to avoid this, to reduce the resistance. So you can engineer the bacteria also so that they are not so competent for recombination. Or that you can try to eliminate various ways to reduce the genome size and to reduce eliminating transphosomes and also other integrative elements from the genome, which are not probably needed for your final strain. So by doing this type of thing that are feasible now, of course you can change even the DNA and introduce a completely senile molecule that will not be able to recombine or acquire genes from outside. But you can, things that are easy to do are simply, for instance, removing mobile elements, transphosomes, and make the strain less prone to recombination and to accept foreign DNA from these kinds of things. Thank you for your question. I was just wondering if you used these systemically, presumably, they would be seriously immunogenic. Was life as part of your gut flora or something? Could you give them to treat intestinal tumors? Yeah, so the idea although the people are using the systemic administration more as an experimental model, but in fact with Salmorella they have been using it even in clinical trials and there was a problem in the way that the dose that it was required to have a significant colonization was already toxic for endotoxin production also with LPS and so on. So there's been some work in human with bacteria. So I don't think probably systemic administration is the way to go. So we need more in the applications in which niches in which bacteria already exist and are more accessible for bacteria. Like the gastrointestinal tumors or inflammation in the gastrointestinal tumor or the platter, urinary platter also. And while not the skin or certain places in which you can really access easily with bacteria. So you can target specific cells in your skin or in the gut and not thinking on a systemic administration. Also for diagnostic applications, maybe you can think in the bacteria. So they can monitor the presence of cells tumor cells or molecules in the gut and it tells you if you have a disease with this type of approach. So this last question here. I was just wondering if there was if there was any phenotype tendency to form biofilms if so what is the history of biofilms? With when bacteria is fixed to the surface is there any tendency to form biofilms? With strain, we haven't analyzed really our strain is able to form biofilms or to induce quorum sensing molecules. So but it's clear that you can maybe target bacteria not only against tumor cells but also against antigens found in other bacteria including maybe biofilms. So you go maybe think of bacteria that are able to attack pathogens or biofilms formed by other bacteria. So because antibodies are easy to produce against different antigens so you could really think on ways to target your bacteria to maybe a biofilm so I don't know if that explains your question or not. Alright, thank you very much I think we need to stop there now. So we are going to have a longer break. Can we all be back here at 3.5?