[Music] thanks so much for the organizers it's it's great to be here and yes so I think I've been invited to offer an evolutionary perspective sort of evolutionary doom and gloom perhaps and so and it's been it's been great to learn so much during this meeting so the title so Philips sort of asked me to talk about anti virulence so that's in there one of the levers we can pull but also antibiotics and probiotics I have heard quite a bit about at this meeting and so I wanna ask how we can use these in a way that balances efficacy with evolutionary risks so when I begin with the antibiotic resistance crisis and what I'm showing here is data from humans so this Americans note wrong way emergence and increase in frequency of drug-resistant strains now you know in a one health context this is of course also a problem in agriculture in farming so we have the shared concern with the increasing failure of our old drugs so this is a problem what truly makes this a crisis of course is this emptying of the drug pipeline so right so that's the crisis so what can we do next there's lots written about this lots spoken about this I think we can boil it down into two easy to say but hard to do solutions the first one is to simply open the drug pipeline we can we can seek out new antibiotics or new antimicrobials you know whatever whatever molecule you like as long as it kills or cripples bacteria and if we do this at a rapid enough pace we can keep up with this evolutionary process the old drugs fail but there's always new drugs around the corner the other generic solution is of course to modify the selective pressures on the antibiotics we use to be smarter about how we use antibiotics so this of course begins with not taking antibiotics if there's a violent cause involved and but there's more subtle things we can do to to cut to modify the Selective pressure on resistance so one of the things I want to argue today and you know generally interest in my lab is how evolutionary biology and also ecology can contribute to both of these challenges okay so as an evolutionary biologist my and my you know when I initially became interested in this problem of drug resistance probably about 10 years ago you know I had my evolution hammer so I thought this is an evolutionary problem that's all there is so I thought ok all we have to do is figure out how can we maximize the evolutionary robustness robustness of our interventions which you know one metric of this would be the time until failure the time for example the time until 50% of the chickens you treat you have a failure to resolve the infection but you know speaking to clinicians speaking to farmers speaking to other researchers with other interests I realized the problem is a bit more complicated there are other things we have to worry about we have to worry about the the efficacy of the treatment so if I go inventing some new evolutionary robust therapy it also has to work um it's a trivial point but I think it's it's often missed in my field at least so so we have these two criteria we're interested in and we can maybe think of other dimensions has to be a working treatment and it has to be a robust treatment if we have sort of a canonical antibiotic like drug it could be a bacterial sin it could be an antimicrobial peptide it could be a conventional antibiotic we can to begin to address this problem we built some simple epidemiological models which I won't go into but the the Upshur sort of Luke McNally was leading the modeling here the upshot of this work is that we see this uncomfortable trade-off and it's really quite intuitive so the more effective your antibiotic the more we move to the right along this axis we predict the time until failure of that treatment to get shorter and shorter so the good news is you know with my obsession with evolution which we can devise an evolutionary robust therapeutic the bad news it it will necessarily not work very well so we you know so homeopathic doses of antibiotic will work forever so there's the good news right but if we want things to work we face this trade-off so I'm not going to go in we have you know we're populating this with data it's not not quite ready yet but we're certainly finding support for this constraint but this this map also sort of illustrates the challenge we want to address we want to get up to that question mark up there the question is how do we get there we want treatments that are robust that work and continue to work into the future so we know already a little bit about the ingredients we want to bring so antibiotics remain part of our arsenal we have these incredibly effective drugs we have immunity in immuno competence hosts this is a very obviously a very powerful mechanism that we want to leverage effectively we have other ingredients that we can bring into bear so we're very interested in phages but there are many other sort of novel antimicrobial chemicals that we can bring to bear and then there's a whole menu of adjuvant that we can bring so I'm going to talk a little bit about anti virulence drugs there are anti resistance compounds there are transmission blocking compounds and strategies Diagnostics I'm going to talk about vaccines of course and then we have to ask how do we put all these things together can we combine combined compounds so this is a very popular route let's try to antibiotics rather than one let's try an antibiotic in an adjuvant I also want to talk a little bit more about conditional strategies strategies that change in response to diagnostic information I think this is a very powerful solution actually okay so anti virulence drugs first so anti violence drugs are drugs that disarm rather than kill bacteria so that they stop the expression typically stop the expression or the activity of virulence factors so virulence factors are we can define as non essential components that predict harm so these are things such as a descends toxins EXO enzymes immunomodulators lots of stuff that bacteria do to harm hosts so so in a screen for an anti virulence drug so we have the wild-type here it grows well on a plate and damages the host this host that looks like it's sleeping is supposed to be sick okay whereas if you knock out a virulence factor or use of anti violence drug it grows well in vitro but the mouse is fine so that's the sort of a classical screen and so this is great for the pipeline it gives us lots of new molecular targets so that's certainly exciting in terms of resistance evolution at least the initial sort of comments on this this you know and they're sort of reviews and papers about this this is going to be great because we're not killing bacteria there won't be a resistance problem so in this review we went through this a little more carefully and actually a lot of reason for concern but we also mapped out one avenue that we believe is it's going to be more evolutionarily robust okay so there's what we believe is a more evolutionarily robust strategy is to target bacterial virulence factors that are that are cooperative now this is a strange thing to talk about for microbes microbes do a lot of stuff collectively cooperatively and so we predict that by targeting bacterial cooperative activities that underpin virulence we can diminish or even reverse selection for resistance so to give you a sort of a concrete example of what this looks like I want to talk about sedara force here we go so this is an example of a cooperative virulence trait so Siddhartha's are molecules that many bacteria produce to scavenge for iron so the host is a knight typically in iron limited environment and so they produce these molecules these siderophore molecules so an individual cell will produce a molecule that will scoot out into the environment by intuitive ferric ion and then this this this sedara for iron complex can be taken up by that cell but also by other individual cells so basically you have an individual cost and a collective benefit from the perspective of the bacteria so this is an example of microbial cooperation and so we predict that this will be more robust as a therapeutic target because if you have a mutant that pops up it has a different sedara for a different way of scavenging iron it's paying the cost of that innovation but the benefit will be distributed among other cells and so selection is works much less effectively with this this context okay so you knockout sedara Falls you suppress virulence so we use Pseudomonas quite a lot as was working collaboration with Roth coming Lee's lab actually so you can see in black the wild-type survivorship of an acute infection in these Galleria wax moths so it's just an injection they all die you knock out genetically knock out the iron scavenging apparatus you see an attenuation okay so we use gallium as a as an anti virulence drug in this context so gallium is is a transition metal like iron it binds with higher affinity to the siderophores so basically adding gallium you titrate out the efficacy of these siderophore molecules so we see this in the end the figure to the top left there so the the increase in gallium supplementation low doses that are very well tolerated by many hosts so in the dotted line there we see if you add gallium to two bacteria in an iron-rich environment where the iron acquisition environment mechanism has been disabled there's very little effect but if you add gallium in an iron limited environment to the wild-type you super suppress their ability to grow so so it's working in that level and then in the in vivo experiments we see something interesting so actually in the red what you see is you see more attenuation than the genetic knockout which actually surprising normally the cat you know that the chemical suppression of a trait doesn't quite recapitulate the genetic knockout here we see an excess effect which is unusual and the greatest excess was for the intermediate dose and our handle on this just a sort of a window into the mechanism is as you increase gallium you have this behavioral feedback in the organism the organism responds to the iron limitation by making more sedara for you see there's positive feedbacks you see this increasing slope but if you had a lot of gallium there the negative feedback kicks in and they give up so what we're seeing here we think this intermediate dosing is the bug getting caught in this regulatory trap okay so this is all very well we see some effect in vivo this evolution proof challenge we then did some experimental evolution so I always recommend if you have a new new therapeutic do experimental evolution see how it fails very important to understand how it fails and how you can mitigate that process but here's a case that did not fail over 12 days so twelve days is not very long so what we have here is the in black is the control treatment so this is just the integral of daily growth it's a measure of how well Pseudomonas is growing in this environment and then we have the level of growth for intermediate and higher doses of gallium so we were suppressing Pseudomonas and over 12 days of serial passage Pseudomonas could not solve this problem so it's robust over 12 days not very long but we can benchmark this against equivalent control by our existing antibiotics and so I think this is the fun part of this result so you can let you can introduce you could be controlled with gentamicin panel be in about six days Pseudomonas had solved the problem of gentamicin easy no problem at intermediate and higher doses ciprofloxacin similarly converges to the control in about six days no problem then we gave students a hard challenge we gave it two drugs so we gave intermediate and high doses of the combination and again you see this rapid evolutionary solution whereas the gallium was able to maintain suppression so I think this is an interesting result that it sort of supports this claim that targeting collected behaviors of cooperative behaviors is more evolutionarily robust the problem as I see it though is back to this this this efficacy challenge so we have these two axes so we can beat the constraint we can get into the white space but frankly if I was really sick when my kids were sick I don't want to give them gallium I want an antibiotic I want a bacteria Seidel antibiotic to take out the bacteria I want something effective so we're still left with this challenge of improving efficacy so I think one key move here is the use of conditional strategies so this is David McAdams he's an economist at Duke in the US and we've been chatting for a while now about the use of conditional strategies to address the antibiotic resistance challenge and so what got us thinking is the use of point-of-care resistance diagnosis which is of course well certainly in biomedicine is gaining a lot of attention and traction as a sort of a flow for the appropriate treatment of patients in a healthcare setting so this is great news for treatments if you have a rapid diagnostic of course if you're resistant to drug one you get antibiotic two if you're not resistant you get antibiotic one this sounds like a great idea for patient care the challenge we wanted to address is what does this mean for Public Health can we use how effective is this kind of intervention going to be to improve public health and broadly the resistance crisis and so we can ask this is in a one health context will rapid Diagnostics help to manage infections in a farm setting in a one health setting I think the logic carries over but bear with me is I talk about humans and and an actually respiratory infections just to really go off piste for this meeting okay so let's think about this sort of the the status quo so what we have here is a schematic of a population of hosts okay this is a schematic of an epidemiological model for these boxes represent different classes of people or chickens so the black the black s box are the individuals that are not infected they're susceptible to infection but they're not infected and then the brown boxes represent individuals that are in fact it okay and that are going to undergo treatment with an antibiotic the different sub so they're all infected they all present with the same phenotype this brown sick phenotype sort of demanding attention but the numbers the subscripts refer to their resistances so i0 is the pan susceptible genotype I too is resistant to drug - I want to drug one I 1/2 is pan resistant that's obviously problematic so under the current sort of status quo you're sick you get the frontline drug which is drug one okay this will shorten the infection of drug one susceptible strains and will reward the drug-resistant strains this is what we know this is just selection for resistance but now if we have this this magic box in the farmer in the clinic where we can rapidly diagnosing individuals so now the veterinarian or the clinician has these x-ray goggles can discriminate these different genotypes and and give appropriate antibiotics to to the genotype so obviously i 0 will get drug 1 I too will get drug 1 I 1 will get drug 2 etc that's all straightforward what we're suggesting in top though is we can make additional behavioral interventions to put the thumb on the scale to differentially penalize resistant strains so if you carry a resistance if your I to or I won yes you'll get an appropriate drug but we could also administer some transmission control measure if you're a kid maybe you're gonna stay home from school if you're a chicken then I'm sure there's other things you can do to limit transmission I want to this is the pan resistant case so for humans this is where we can do heavier measures of discovery and contact tracing isolation I think in a farm context is a culling I imagine might be more of an option okay so we can so we again we've built these epidemiological models and we can play with the parameters so for a generic acute infection the parameter space the behavior of the model looks something like this we've got diagnostic delay in days and we've got cost of resistance on the y-axis the blue space is the MAGIX parameter space where we can select against resistance so this resistance crisis is going into reverse so at the top there you see there's no resistance Diagnostics to get into that space without any resistance Diagnostics we have to assume that resistance is super costly that's a 50% cost right so it's really impossible that's not that's not not plausible but what we find is with resistance Diagnostics even cost free resistance can be selected against if we if we have a sufficiently rapid diagnostic and you notice the timescale is not it's not good you know in the clear clinic we're talking about one-hour diagnostic results for this public health concern we can be a little slower and it can still work so that was the most optimistic part of my talk now sort of add complexities and challenges and one of the major challenges which I it's common to animal to various houses the problem of carriage states most of the pathogens we worry about spend a lot of time in in humans and other animals in an asymptomatic carriage state so this really complexify z-- the model makes it more interesting so now we have these two genotypes they spend a lot of time under the radar in carriage so you're not so you have it you have a potential pathogen you're not actually infected you're not expressing the symptoms and this introduces this problem a bystander selection so while you know I'm walking around carrying a bunch of potential or opportunistic pathogens in my guts and my nasal pharynx and wherever if I take an antibiotic I'm going to penalize susceptible strains without recognizing that effect necessarily okay so this problem bystander selection was recently flagged by Christie and Teddy Jan 2 and Mark lip stitch we've been working with in Harvard and so this is just a sum of their data actually so what we're looking at is the rate of bystander exposure so this is the the rate at which you're a Europe you're a potential pathogen the rate at which you're exposed to antibiotics due to some other cause and then this is the rate at which you are the focal pathogen and you you therefore face antibiotics because of your dastardly deeds in a host and you and the general result is these organisms for humans at least are seeing a lot more antibiotic as bystanders as then they are as the focal pathogen now I have no idea how the parameters stack up in an animal context but I'd be very interested to know okay so so let's look at strep pneumo so this respiratory bug which has a very significant human pathogen and so we have carriage in the nasopharynx so playing with the model what we see is as we extend the duration of carriage this sort of blue space that is this the you know the the wonderful problem solved space disappears and we move into a sort of normal service of using antibiotics we select for resistance because of this increasing dominance of bystander selection they're seeing most of their antibiotic when they're in carriage okay so there is you know we do have a actually a salute the logical response to this is simple the implementation is more difficult we need carriage Diagnostics we need to be able to differentiate these genotypes these C 0 and C 1 D genotypes there that's the the resistant and susceptible genotypes in carriage we need to be able to see and act differentially on that carriage information so if we're able to do that again this is logically straight forwards but hard to do for able to do that then we can solve this problem so actually there have been experiments or sort of clinical trials in Sweden that have done effectively essentially what we're proposing so this was actually some time ago in Sweden where they went into into nurseries they swapped kids and they would impose heightened transmission control if they found a kid with a resistant genotype of strep pneumo in their nasal pharynx what they did was the kid would just was sent home for the rest of the week so basically they were removed from the population and this penalized the resistance strain so we can take some of this data and sort of scale it up and what we find is this now if we have so this is the rate of carriage Diagnostics are on the x-axis and this is hat the extent to which we can penalize the transmission of the resistant strain during carriage on the y-axis so if we have no carriage Diagnostics we're stuck in this white space resistance is going up year after year but it only takes a little bit of carriage Diagnostics to get into this blue space so for their parameters for you know our estimates of the parameters for strep pneumo it will take annual surveillance and with only a 20% reduction in in the the relative transmission of the resistant strain and carriage to deal with this issue ok there we go so I want to talk about another sort of so this is conditional on the resistance genotype that's one kind of conditionality so I'm sort of begin to talk about some other conditionalities I know that this is a while in Walder tuft in my lab in MD actually in my lab and so we've been talking more and more about conditioning treatments on the severity of symptoms which is very simple idea and they're sort of impetus for this was a review in The Lancet Infectious Diseases which was damning about all of the you know the menu of novel therapeutics it basically said I think this was the title right novel therapeutics are inadequate for serious infections so basically what they they went through phage therapy anti-violence drugs a whole host of you know you know you know your favorite new trick doesn't work and the benchmark they set was incredible was in my view very high and for a reason they set the bench as sepsis so this is you know the case of invasive you know bacteremia with all the inflammatory signals that you're about to die okay and this is the context where you have you know up to 50% chance of mortality antibiotics are central effect you know and and and and effective if you get them soon enough so that was the benchmark and those saying that none of the other therapies are anywhere close to this so I sort of really a bit of a thought experiment was just okay but let's think about how we actually use antibiotics is there absolutely essential in this you know last line Hospital context but that's a sliver of our youth that's Hospital use in its entirety as that sliver sepsis or you know life-threatening infections is a sliver of that sliver of course we use most of our antibiotics in the community and what I'm not showing here is the other half of the pie of antibiotic use in agriculture so we use it mostly for mild self resolving conditions really getting people back to work a couple days earlier and so the kind of presentations this was a sort of a global survey it's UTIs pharyngitis otitis things that if you're healthy you will resolve in a week or so and so the sort of implication from this thought experiment is is that we should use lower efficacy drugs for mild infections if you're immuno competence and we in and the other various risks are addressed in too in order to preserve these and antibiotics that we still have for life-threatening cases so we can go back to the map of these two desirable quantities and this helps a lot the assumption in this figure is that we use all of these different drugs with equal intensity the picture will change if we taper off our use and the most effective drugs we use very rarely that will extend their lifespan that will push that contour backwards but it means we're going to be pummeling this sort of medium efficacy window and this is where I think these anti virulence drugs which I sort of went off for a bit because they're not very effective this is what I think they may play a role okay so I want to sort of switch to one of the empirical contexts I'm you know my lab is most interested in but I think there's a strong I hope there's a parallel to microbiome research generally so we're interested in chronic infections for a bunch of reasons so chronic infections are just complicated intractable in many cases that's why we've chronic infections because medics can't can't treat them they can't effectively clear the infection these are infections that just do not resolve that's their complex in multiple ways they're spatially structured they're multi species that polymicrobial and we have this problem of poor antibiotic efficacy so I'm flagging here this these are issues for microbiome generally so these are the kind of themes that my lab are particularly interested in sort of by the behavioral dynamics of bugs in these complex situations the evolutionary dynamics the spatial patterning what I want to focus on for the remainder of my talks is this issue of community dynamics okay so we have multi species the empirical context that we work with is we have some great colleagues in the emory atlanta cystic fibrosis clinic and so people with cystic fibrosis of this genetic condition end up with lifetime lung infections and typically the the leading cause of death is Pseudomonas is attributed to Pseudomonas aeruginosa okay and other pathogens to play key roles so so one of the sort of really you know sort of eye-opening results in this context is that doctors are truly flying blind so we're looking at here is the the futility of resistance diagnostic so I began my talk saying what we need is great Diagnostics well it depends on the context if you're one of these chronic infection contexts or a microbiome context in this in essence then resistance Diagnostics are pretty much useless what we're seeing here is the treatment of individuals that tested positive for Pseudomonas aeruginosa the the y-axis is a measure of the health outcome it's a measure of lung function but it's you know generalized this is a measure of the health outcome of the treatment sometimes it goes up sometimes it goes down the unit of analysis is the patient or the units of analysis is the exacerbation but what you can see is whether it goes up or goes down has nothing to do with whether the Pseudomonas you pulled out of that patient was resistant or susceptible to the drug you treated them with so to me this is really striking there's lots of hypotheses we could explore here what we're really pursuing is that as there as there is a role of polymicrobial interactions there's more than one pathogen in that in a host and there's a whole bunch of other microbes too so we work with the CF clinic led by Arlene statisti Sanko we work closely with joiner goldberg also at emory and so we've been gathering cystic fibrosis sputum samples and various health metrics the sort of map how this project first of all we're interested in the association between the state of this the microbiome and the health of the patient so we do the kind of sort of Association studies so here we're just coloring patients and their microbiomes by the severity of their symptoms when red very low lung function so on the cusp of a lung transplant in green these are a pretty normal lung function actually we also have an experimental way to probe the system here so I'll talk more about this we have experimental model communities that we can Petare with antibiotics and probiotics and any therapy we want our goal one of our goals is to build a mathematical model of the community to ask how these species interact with each other so then we have a map of health you know what's the healthy state what's a diseased state we have a mechanistic model of how the community functions and then putting this together this in principle will lead us towards optimizer or at least informed interventions and so the this is sort of a map of the project it's really early days but I just thought I'd share this with you so on this sort of health mapping issue one of the tools we use is machine learning this is Conan who's been leading the machine learning so we're trying to predict lung function you know predict a health outcome from the micro to the microbiome content and so you do this kind of analysis you end up with a bunch of sort of weightings or scores for the different microbes and so first of all we have this sanity check that Pseudomonas this notorious number one pathogen is a negative predictor of health so we we passed the sanity check so pathogens are bad that's that's comforting right and but we see a bunch of bugs that are predictive of improved outcomes or improved health so but of course what we don't have here is causality they could be biomarkers and that's very important right so it could be these could be biomarkers of health although you can just get someone to breathe into a pipe and you figure out how well their lungs doing but we have this sort of you know the possibility there may be probiotics there may be acting in a way that is is is resisting colonization by these more severe pathogens and so very Anela is one of the ones that I want to keep your eye on so it should address this problem of causality we build an experimental model which i think is actually important for this kind of work so what we have here is that some advantages granted for the the CF microbiome it's relatively species poor so about 12 species account for about 90% of the reads we find in a CF sputum sample we can recapitulate much of the physiology in vitro so we use a synthetic sputum which has been extensively benchmarked by my colleague Marvin white Lee basically it's lime it's got its got the nutritional composition of sputum we add new sins we add DNA so it also has the viscosity and bugs grow as these little sort of characteristic aggregates so they're not sitting on the side they're not making conventional biofilms but they are spatially structured okay and so we take two broad strategies ones that top-down strategy will take sputum so this is microbiome this complex community from a patient will culture it into this synthetic sputum and we can trial new interventions we can trial antibiotics probiotics and we can see how the community responds to these perturbations but we can also take a bottom-up approach where we reuse defined communities so again so these are our players these are the actual bugs we use so in red are the pathogens that clinical microbiology cares about in the context of cystic fibrosis so Pseudomonas her marvelous staph a Croma back tubercles area these are all you know on the watch list but then we have a bunch of other bugs that are largely orally derived that we find in some abundance in in the lung of these people so we can play with these in different ways these communities and when we're also developing a phage fire we're interested in phages so if you have a library of phages interacting with Pseudomonas and happy to chat about that later so some of the things we can do with this kind of a system we can do pairwise experiments we can ask how one individual affects another so you do this in different ways and this will give us a network of interactions so here we have we see some facilitation going on but we see mostly competition and so VAE Anela over there is actually inhibiting Pseudomonas so you know back to this yes it's a marker of Health but there may be some functional interaction going on there we can then do experiments where we put these are multi way experiments we put me we assemble communities with multiple species so in this case so on the left here we're seeing a diverse community at time zero so these different panels are different replicates we had multiple replicas just showing you three replicates here and we passage we do subculture in this sputum medium so every two days they've got subcultures and what we're seeing we're losing a lot of the pathogens and we're seeing dominance by a lot of these oral bugs these oral anaerobes in particular so we're dominating by VAE Anela prevotella and some Hamas in the absence of any perturbation but now we kick this system with antibiotics and you see some unfortunate outcomes so turbot mice in not a lot happens but Mara turbo and then some of the combinations you see dominance by pathogens and you see an interesting alternate states so if you go back to the control what's what took very pleasing to my eye at least as these replicates are really on a tram line there's a very repeatable dynamic to this system once you start throwing antibiotics you see these alternate states popping up so meropenem we're seeing we're seeing a dominance by staph tobramycin in meropenem we go towards a Croma bacter or Bourke algeria both pretty terrible actually for in a patient context okay so so that you know the the summary here is you hit this community with antibiotics and you rewarded with more pathogens and more drug-resistant pathogens okay so you know said at the beginning of this section we're interested in building mathematical models of these communities so we use it actually sort of baseline model is a generic community ecological model so a generalized lotka-volterra model we can take the data I just showed you in the previous slide use that data very rich dataset lots of information in the time series of that data to infer species species interactions so this is showing the model inference and so we have the effect of the actor on a recipient the diagonal is here the effect on yourself so red is a negative affects everybody in hibbott their own growth via nella is this sort of universal inhibitor it's the inference from the data Pseudomonas is actually facilitating some others again but we have more more work to do to validate these interactions and so that's ongoing okay and when we validate this model so this is this is really a holding a holding place you know the goal is to understand the trajectories through this this complex space what are the community assembly rules how did you get from a healthy state to a dysbiotic state and and can we use probiotics and antibiotics and other interventions as levers to shape transitions from one state to an alternate state so these are the sort of challenges that were interested in and again I'd you know stress that this lung microbiome this infection microbiome is relatively species poor this is is certainly more difficult in more complex gut contacts I I don't want to shy away from that okay so I think I'm about to wrap up so what I've talked about a lot today are conditional interventions well the first thing I talked about was that antibiotic efficacy is desirable or treatment efficacy it doesn't have to be a conventional antibiotic you know I tend not to believe in this time it's different if you're killing pathogens there will be an evolutionary response there'll be an ecological and evolutionary response and we see this trade-off between efficacy and the other desirable quantity of something continuing to work evolutionary robustness so I talked about anti virulence drugs you know chasing after cooperative behaviors they can be disproportionately robust we can do better than the equivalent control by an antibiotic but there is a question over there Africa see and then I talked about conditional strategies tied on to point-of-care resistance Diagnostics and then in this simple model where as soon as you acquire a pathogen it music immediately causes symptoms and then you can treat it then you can get both you can get efficacy and evolutionary robustness and actually antibiotics become a kind of a form of renewable resources you use you use antibiotic to to drive susceptibility to antibiotic one but you know back in the real world we have a whole sort of litany of complications we have carriage this is opportunistic pathogens going under the radar not causing symptoms for a period of time but still undergoing selection by antibiotics we have chronic infections we have serious life-threatening infections they all require broader diagnostic inputs and a menu of levers including probiotics antibiotics and novel therapeutics whatever they are certainly the logic still holds but there are technical and strategic challenges as we move into these more complex arenas okay so I am about to just thank the lab so thank everyone in the lab who's done the work I presented and some key collaborators so Luke in Edinburgh and Ralph particularly in Zurich for the for the gallium stuff so that the gallium experiments were all done in Ralph's lab and then folks with their diagnostics and and the Emory folks were the the CF stuff and I'll be happy to take any questions [Applause] [Music]