[Music] thank you very much Steven I'd like to thank the organizers for their kind invitation to tell you about some of the work that we do in my laboratory Philippe asked if I could speak on the topic of why do we need to understand microbial pathogenesis and I have to say as a researcher that's worked on this topic for the last 25 years that was quite a searching question for me because there are of course some infectious diseases that we have been able to bring under control using relatively simple interventions such as antibiotics and simple vaccines but this audience of course needs no reminding that poultry remain afflicted by diverse pathogens that can constrain productivity and welfare and of course they can also throughs transmitted through the food chain and farm environment Zuno sees that affect people and exerts substantial costs to the economy and society and for some of you in Persians simple interventions that solve the problem aren't available and even where we do have control strategies these can sometimes be limited in their cross protective efficacy partly because pathogens evolve rapidly to escape vaccine mediated protection and they can also exhibit a high degree of antigenic diversity of course we're also seeing the effectiveness of antimicrobials blunted by the emergence of transmissible drug resistance and there are restrictions on antimicrobial use in some countries which will require some more inventive thinking around control strategies for diseases and I also think it's important that we study microbial pathogens it's important to train the next generation of researchers to study avian diseases and zoonosis in the animal host where they are a problem a great deal of published research relies heavily on cell lines and indeed surrogate rodent models and I believe there's a vital need to be able to study these diseases where they exist so in my team formerly at the Institute for Animal Health and now at the Roslin Institute we study a number of bacterial avian and zoonotic pathogens Campylobacter and Salmonella you've been hearing a lot about already this afternoon we also study avian pathogenic e.coli and sheikah toxin-producing e.coli that I won't talk about because they affect ruminants and as we've been hearing these organisms collectively account for around 70,000 laboratory confirmed cases of foodborne illness in the United Kingdom each year it's probably more like 1% of the national population and that comes at a high recurring cost to the national economy and when you scale it in global terms there's a very substantial burden in terms of human health and healthy life years lost to these pathogens and as you heard from Lisa and others this afternoon in certain circumstances and depending on factors both of the microbe and host you can also see disease in poultry associated with some variants of these bugs so I'm going to talk primarily about Salmonella and Campylobacter this afternoon and I won't reap raise all of the background on sound le you've heard a lot about it most infections in warm-blooded animals caused by salmonella enterica subspecies one of which there are thousands of entre genetically distinct serovars and those serovars can vary in their host range in clinical presentation where surveys like typhimurium are largely promiscuous and can therefore cause you notic infections and at the other end of the spectrum there may be host pacific serovars which tend to me or associated with systemic disease such as fennel typhoid as you've been hearing handling of contaminated poultry meat and eggs remains an important risk factor and we've seen the evolution of these organisms punctuated over the years by the emergence of epidemic and drug-resistant variants and so it's a rather moving target we heard from Stefan DeVries this afternoon about inactivated vaccines which are partly credited with reducing the instance of non typhoid else and lettuce in humans this graph shows the pandemic of Salmonella enteritidis flash type 4 and other prevalence Salmonella serotypes during the 1980s and 90s these are data from the UK and the point at which we began to vaccinate broiler breeders and layers and indeed implement other control strategies was somewhat coincident with a major reduction in the incidence of human infections but I should say that it's been rather difficult to discern precisely what contribution vaccines made to this decline because national control programs in the UK mandated the use of those vaccines at the same time as a raft of other control measures Stephan presented I think some compelling evidence for protective efficacy against serogroup B C and D strains with their trivalent vaccine but there are other published studies in the literature which perhaps provide a less compelling picture of their ability to reduce gut colonization and egg contamination and we know that in some countries even where those vaccines are used there continues to be a problem with Salmonella in poultry meat and eggs so I believe this an ongoing need for us to try to devise improve control strategies because we haven't yet solved the problem so another important point to make at this juncture is that not all Salmonella are the same I've mentioned antigenically distinct serovars that can cause either enteric disease or systemic disease but we're also beginning to learn that within serovars of salmonella that we long regarded as promiscuous there are some passo variants which appear to be hosted Apted there are certain types that cause systemic disease in pigeons and apparently nothing else and others that cause systemic disease in pasaran birds and we and others have began to use machine learning to try to look at the blueprints of salmonella typhimurium strains and associate that with their hosts of origin and it does appear through machine learning that there are puffer variants of Salmonella typhimurium that differ in their invasive potential and their host tropism so it remains a challenging organism to try to control because of its diversity now we're greatly aided in the task of trying to control Salmonella by the availability of complete genome sequences and just before I came away I checked enter obeys and there are now a hundred and ninety seven thousand of confirmed genome sequences for Salmonella enterica strains so we have no shortage of genetic blueprints for these bacteria what we do have though is a relative lack of understanding of the role that those genes play in colonization pathology and transmission in particular in farm animal reservoirs where these bacteria can be found we'll let understand little of when wearing at what level they're expressed and because of these two limitations we also find it hard to interpret the impact of genetic variation whether that's in the repertoire sequence or expression of genes to some extent we can predict through homology how the encoded proteins act but for many of the factors we find to be important for infection but we don't understand how they work and then as I highlight in red we have to think how are we going to use information about the function of these genes for some sort of practical game and an important priority for my group is also to beam up to answer these questions with minimal use of animals in experiments and minimum harm so each Salmonella enterica strain has about four-and-a-half thousand genes how do we go about assigning functions to those well we're fortunate some know is a genetically tractable organism it's relatively easy to introduce mutations into the Salmonella genome and then to evaluate the phenotypes of those mutated strains relative either to the isogenic parent or a repaired strain or one in which the mutated gene has been reintroduced in trans and you can do this gene by gene this was a study from my group over 10 years ago now shortly after the Salmonella enteritidis genome was first published you can look at the blueprints and identify on the basis of homology candidate virulence factors shown here are the thirteen ephemeral operons that determine these hair line appendages on the bacterial surface these often mediate interact with host cells and so we chose in a systematic way to mutate each of the memory major ephemeral subunits and to test those mutants one by one and some of them are attenuated and we can partially restore them to virulence by putting the genes back now that gene by gene analysis is powerful but with an organism with four and a half thousand genes that's going to be an incredibly time labor and animal intensive process to assign functions to the genes and it was with that in mind that the Institute for Animal Health some years ago we began to use a method for screening pools of random mutants of Salmonella in farmed animals this is a technique we call signature tagged transposon mutagenesis or STM as you'll see shortly this is now superseded by a more powerful technology but in the mid-1990s to 2000s this was the state of the art briefly what we do is randomly mutate the bacterial chromosome with a transposon it's a kind of jumping gene that inserts almost at random into the bacterial chromosome and we use a panel of transposons that each have a unique DNA sequence this is the signature tag and you can think of it as a kind of bar code when the transposon whether unique tag inserts into the genome of in this case salmonella it gives that Salmonella a unique oligonucleotide sequence that we can scan for there are actually 95 of these but for simplicity just for a shown here in different colors and so we can introduce these different tagged transposons create mutants and then take one mutant with each of the different tags and assemble them into pools and these pools can then be screened in animals and we can compare the composition of the inoculum with a composition of mutants that we recover from these animals we call this the output pool and what we're looking for here is the presence of those organisms by virtue of the presence of the tags they contain and we do that by PCR amplification of the tags and detect them by hybridization and what we hope to find I'm not sure if you can see the laser pointer that there are some mutants that are represented in the inoculum like this one in red which are negatively selected and don't survive in the host and we can infer from that negative selection that the mutant lacked a gene that was important for colonization of the host and we can then find out where that mutation is located by sub cloning and sequencing that transposon flanking regions now we did this for someone out of time from you I mean cattle pigs chickens am i some years ago data are all published now screened a relatively modest library of just over a thousand mutants but have identified over 200 new Salmonella virulence factors at the time including things that appeared to be putatively host specific now this was a very powerful method but it has its limitations it's very subjective comparing the intensity of the hybridization signals between the input and the output it's very time and labor intensive to make and screen the libraries and also even once we've identified negatively selected mutants it's a time intensive process to find out where those mutations were located in the genome and so in the late part of the first decade of this new millennium we started a project with the University of Cambridge to apply what is now the absolute state of the art to functional annotation of bacterial genomes this is called transposon directed insertion site sequencing or trotta's it's sometimes also called TN SEC or transpose on sequencing now don't be frightened by the detail of that it is conceptually very similar to what I've just described we make random transpose on mutants where a transpose on jumps into the Salmonella chromosome and we take pools of mutants and screen them in animals and again we compare the content of the pool that went into the animal with a content of the pool that came out it's just that in the case of trotta's instead of looking for signature tag mutants we use a special kind of sequencing called massively parallel sequencing using the Illumina platform to give us two types of information simultaneously we can tell to the nucleotide where the insertions are located because as we copy out of the transposon we sequence the adjacent genomic DNA and because of the massively parallel way in which the sequencing works the number of times that we return those specific sequence reads then becomes a measure of the abundance of that mutant in a pool so in this hypothetical example of an in vivo screen there are two genes a and B which were each mutated three times the insertion sites are denoted by these red pins and when we apply this massive parallel sequencing approach we can see to the nucleotide where those insertions are located within those genes so returning sequence reads at these positions tells us where the insertions are located but because of this massively parallel nature of the sequencing the number of reads is a measure of the abundance of the mutant and so in this hypothetical example gene B was mutated three times and there is an equivalent number of reads before and after screening this library in a hypothetical animal now that tells us that this gene is not under selection in vivo by contrast there's another mutant here in gene a in fact three different mutations at these positions the sequence reads in this instance are returned in the inoculum but they are largely missing in an output pool from this hypothetical experiment that tells us these mutants were present in the inoculum but they're missing after infection and therefore this must be an important gene for survival in the host now this has proven to be a fantastically powerful approach for assigning functions to Salmonella genes with minimal animal use the data have published some five years ago now and I won't go into all of the details but we've made a comprehensive library of Salmonella mute and screened about eight and a half thousand of those in cars pigs and chickens the Salmonella chromosome is here in yellow the red lines around that denote the position of all the transpose and insertions around the chromosome and the three strains plasmids in this strain and we've screened this library of about eight and a half thousand mutants in carbs pigs and chickens by the oral route it's the same library just screened in three separate hosts and remarkably this methodology was able to give us both the identity and the phenotype for 90% of all the mutants that we screened so that's phenotypes for about seven thousand seven hundred different mutants and it represents about two thousand seven hundred different Salmonella genes so well over half the Salmonella chromosome that you can mutate so if you calculate Fitness scores by which I mean the log twofold change in the number of sequence reads between the input and the output and then plot all of those Fitness scores from the lowest to the highest you end up with plots like this where we have a fitness score plotted here for all seven thousand seven hundred or so of the different Salmonella mutants for which we were able to get the identity in phenotype and we see these s-shaped curves in all the species and what it tells us is that the vast majority of mutations that we can make in the terminal a typhimurium genome have no effect the Fitness score approximates to zero and therefore these mutants are neither positively or negatively selected in the host and this indicates a great degree of functional redundancy in the zonal a genome there are many genes and pathways that can be removed without having a negative effect on bacterial colonization but at the tails of the distribution you can see some that are strongly negatively selected and that's interesting information because that tells us which genes have been mutated lead to a loss of virulence now obviously you have to have clever statistical approaches to know how to approach numerical data of this kind I won't go into the data it's all published but basically we can apply statistics to find a cutoff in the Fitness score that will describe attenuation with some degree of statistical reliability there's actually a bimodal normal distribution of the fitness scores if you look at this population the vast majority belong to this larger population with a fitness score that approximates to zero but at lower fitness scores there's another population that are negatively selected and where these cross we can say what an appropriate fitness score is to say that plays a significant role in infection and when you apply a cutoff like that you end up with numbers such as these where there are around 600 different Salmonella typhimurium genes that play a significant role in colonization of the guts of chickens pigs and calves these are obviously fantastic targets for the design of control strategies that might be fitting inhibitors or developing subunit vaccines or using mutants in those genes as live attenuated vaccines but interestingly what we found doing this analysis is there are also smaller subsets of genes that pay an apparently host specific role there are some that are important in chickens that don't seem to play a role in carve some pigs and of course that's vital information if you're trying to devise a control strategy because something that works for one animal will not necessarily work for the next among the host specific factors that we identified in poultry were lots of genes and pathways related to anaerobic metabolism and that probably speaks to the distinct ecology of salmonella typhimurium in the avian gut where they're much more likely to be found in the anaerobic lumen whereas in mammals they're more often associated with the oxygenated mucosa because in those species they're far more invasive this is roughly how the data looks in practice we have a vast spreadsheet of the insertion sites for nearly 8,000 mutants and we can see to the nucleotide where those insertions located and what effect they have this is colored in different shades here blue indicates a strong negative selection red would be song enrichment yellow or orange really not one or the other and clearly there's a cluster of genes here that play an important role in colonization across all three hosts but if you come out into an integer NIC region the effect is lost so we tile through the Salmonella genome at high density map the important regions and can see effects in this case that are conserved but in this case that are host specific this is Group one hydrogen is one of those pathways involved in anaerobic metabolism seems to play a more important role in the chicken than it does in other hosts now this sort of analysis has been really powerful because even though it's now two decades since Salmonella was first sequenced around 40 percent of so of Salmonella genes are still recorded as hypothetical or unknown function and what an approach like this can do is at least tell us what role do those genes play in natural animal hosts and this is just an example of part of the Salmonella genome where there were mutations which are the arrows on the black lines then these are negatively selected again by the blue color here in all three hosts when they occur in these genes donated in blue here so you can see positions where mutations have no effect and their mutations which have a strong effect and this was a cluster of genes about which we previously knew nothing but clearly it plays an important role in farm animals interestingly not the laboratory Mouse so it really tells us how powerful it can be to do this sort of functional annotation in the relevant animal hosts I've mentioned some of the ways in which we can exploit this data whether it's devising subunit vaccines or inhibitors taking live attenuated vaccines I think to be fair the problem is where do we start there are hundreds of genes that play important roles in cows pigs and chickens and deciding which of those is going to offer the best route to control is pretty challenging particularly for live attenuated vaccines where we have to balance protection against persistence and pathology and getting that magic balance of attributes right is pretty challenging so these very same principles that I've mentioned of screening pools of mutants for phenotypes in my group we've now begun to apply to screen pools of wild-type strains I mentioned that we have blueprints now for hundreds of thousands of Salmonella but what we don't have are phenotypes in relevant animals and to test strains one by one it's just not possible so what we've begun to do is use this mass parallel sequencing approach to follow what happens to wild-type strains when we screen them in pools now we can do that with Salmonella because there are some regions of the chromosome that are naturally variable and if we train this deep sequencing on these polymorphic alleles we can see which strains are present by virtue of diagnostic nucleotide polymorphisms in that region and then the number of times we see that specific sequence read gives us the abundance of the strain so in this case we were interested in whether all of these serovars were virulent in fact this is an experiment we did with Zoe T's in cattle but it's just to demonstrate the concept what we can do is validate that this method reads out faithfully through sequencing what we know to be the composition of pools that we've assembled in the laboratory so we can spike pools in the laboratory with this composition where each of the colored boxes denotes the relative abundance of pools that we've made and the correlation between the sequencing and the known composition is very strong in fact there's no statistical difference in any scenario that we tested but when we take some of these pools and put them into animals we see winners and losers so this pool at top here it was what went down the neck of the animal and then you can see there are some serovars that outgrow the other there are some that we included here as control strains that we know to be a virulence in this species they faithfully disappear there was some that we knew to be more systemically virulent they expand in the liver so this sort of approach can give us a lot of information about the relative phenotypes of strains but in a single animal here instead of having to test each of these strains separately and one might think about using this sort of approach when we're trying to devise cross protective strategies instead of testing our vaccine or treatment against every different possible variant separately maybe we can start think about using pools of strains and showing they're all equally affected now just before I move on to talk a little bit about host genetics I just wanted to use this one side to explain how difficult the challenge of functional annotation can be with bacteria when there is a great deal of genetic diversity in the pathogen I've spoken so far about Salmonella typhimurium and broadly speaking many of them look alike but if you try that with avian pathogenic e.coli you get a very different picture altogether there's a vast amount of genetic diversity we were involved some years ago in sequencing strains from dominant serogroups and sequence types this is just an example of an e coli oh 78 strained from sequence type 23 a very dominant lineage when we sequenced its genome it was different from the next available APEC genome by 1100 chromosomal genes and four plasmids with nearly 250 kilobases of different genetic information and among all of those genes are very different virulence factors and this is a problem when we're doing functional annotation because we tend to use a small number of prototype strains that are well studied around the world and they just might not reflect the huge diversity that really exists out there in nature so an important cautionary tale so um for the last 10 to 15 minutes and 30 change gears and talk a little bit about Campylobacter and I won't go over the detail of this slide it was very nicely covered earlier by Lisa and and others suffice to say it's a problem that we get as humans through handling or consuming contaminated poultry meat and it's not under control there are a number of candidate therapies and treatments that have been suggested and some of these are partially effective but maybe they can only reduce the burden of Campylobacter in the gut by one to two orders of magnitude in many cases fitted together as a combination of approaches who knows they might solve the problem but for the moment we don't have effective licensed vaccines and there are relatively few proven reliable strategies that are currently being adopted worldwide so can you do this childís methodology with Campylobacter I've mentioned how powerful it can be for Salmonella well actually when we tried it it was very much more difficult to get to work with Campylobacter jejuni in the chicken and again working with the University of Cambridge over a decade ago we actually showed that if you take two wild-type strains of Campylobacter that are identical to each other except for the presence of a short DNA barcode actually even those two wild-type strains don't compete with each other in an entirely predictable way the population dynamic is unstable and it's unpredictable sometimes you could put them in different ratios and it would be the minority member that wins out of the two now this is really difficult if you want to screen mutants because it suggests that there are processes at play that govern colonization that aren't related to whether or not you have a mutation in a particular gene I should say that Stephan DeVries who spoke earlier has some great work suggesting that you can partially solve this problem by screening pools of mutants in larger numbers of chickens and then pooling the outputs from all of those birds before you analyze it and the reference to the paper is there if you're in Tristan you can take a gene by gene approach with Campylobacter of course it's a much smaller genome maybe just over a third the size of Salmonella we've done that and Matt found roles for flagella capsule and like oscillation systems but also importantly metabolism it's told us some things about what compiler bactin needs to grow in the chicken intestines and one might imagine that that could be useful later in trying to find manipulations of the metabolome and microbiome a long way off but a starting point for analysis what about vaccinations well in my laboratory we've over the years reported statistically significant protection with either subunit or vectored vaccines full control of Campylobacter so these might have subunits of CG edge and I made in a recombinant Foreman injected into the birds or will express those factors in live attenuated Salmonella we do see protective effects but they're rather modest they're often only one to two orders of magnitude and they're often too late in the life of a modern broiler bird to be useful often we see them seven or eight weeks after primary vaccination on the day of hatch it's too late what we and I think other laboratories find is a lack of consistency both between replicates of a given trial that we may conduct but also a lack of comparison between different laboratories testing a similar strategy this is partly because there isn't a single accepted regimen for a trial so people often use slightly different doses lines and and and so on but I think there is an underlying problem with inherent variation in the vaccine induced response and how Campylobacter may or may not be affected by that even where we have seen protection we know very little about the immunological basis of the protective effects that's partly because of a paucity of tools to actually understand the functional immunology whether by ablation or adoptive transfer or other approaches um we do have a project looking at the like a conjugate vaccines of various types but it's too early to talk about that so the final strategy that I'm going to talk about relates to whether or not we can harness any heritable variation that exists in birds to Campylobacter and at the Institute for Animal Health and now at the Roslin Institute we have housed unique inbred lines of bird that exhibit heritable differences in resistance not just to Campylobacter but to other bacterial viral and protozoan pathogens and these are an incredibly useful resource to understand the genetic and immunological basis of protection one of the things that we can do where there is heritable resistance or susceptibility is we can cross those two lines together and then we can look at the genetic makeup of the progeny of those crosses and whether or not that affects their resistance to colonization and we've been doing that over the years for Campylobacter jejuni using these two lines 6 and N and as you can see from this study they exhibit quite a profound difference in colonization by CGI it's often around 2 to 3 orders of magnitude but what we observe is that this is independent of bacterial strain and we've reliably seen it both with our populations and populations held elsewhere so you might think our problem solved mark why can't we just eat these birds well of course they're inbred they're a lair line and they'd be absolutely useful in any useless in any sort of production context but as a genetic resource for understanding the genomic architecture of the resistance trait then they're incredibly powerful and previously at the Institute for Animal Health we studied a backcross population and working with collaborators at INRA we perform studies in an advanced into cross population of these two lines I'm not going to go too much into what we found that data published in the references at the bottom but we've measured the Campylobacter burdens in the seeker of these animals and then we've sampled genetic variation at tens of thousands of positions along the chicken genome we're looking here for single nucleotide polymorphisms at specific positions and if you plot out all of these nucleotide variants and then look at their statistical association with the measured trait in our case compiler back to burden in the seeker then you can see on these so-called Manhattan plots that there are some positions in the compiler in the Salmonella genome where you go above a magic threshold of statistical association between variation at these positions and resistance to Campylobacter these are so-called quantitative trait loci or QTL and in these populations we found for genome-wide significant QTL for Campylobacter resistance now we've been fortunate to partner with a vision to take this the next step forward and ask is there a heritable component to resistance to Campylobacter in commercial broiler populations we've been studying inbred lines and it's been useful but how relevant is it to a modern meat bird working over a number of years with great support from a virgin we've performed a genome-wide Association study with over 3,000 commercial broilers we've got high-quality phenotypes again in terms of sequel see a few of Campylobacter and high-density snip genotypes for about 90 percent of the birds that we looked at so 2,700 birds in this population analysis we've accounted in the G was model for sex and seasonal sampling effects and again we've identified genome-wide significant QTL what was really cheering is that some of these are the same as those that we found in the progeny of crosses from the inbred lines so that tells us something about how relevant those lines still remain to mapping resistance determinants we've since gone on using Scottish government funding to perform some RNA sequencing of cecal mucosa of birds at the extremes of colonization phenotype these were selected to have the resistant or susceptible genotypes and correspondingly to have low or high Campylobacter burdens and remarkably when we do this we find very few differentially transcribed genes in the seeker of these resistant or susceptible birds if you apply a very strict cutoff it's only four or five and remarkably three of them fall within the QTL intervals so it may be that there's something within those regions of the genome it may be at the expression level rather than the sequence level that underlies this differential resistance and we're working at the moment with a virgin to further dissect that so there is a heritable component to compile a back to resistance but is it going to be useful well the estimates of heritability are relatively modest maybe only 11% of the trait is explained by bird genetics and the rest is everything else that's a relatively low component to selectively breed on but it nevertheless is possible to do that we predict without compromising other breeding goals including feed conversion body weight and gut health indices unfortunately the resistant associated variation that we looked at was already quite prevalent in the commercial populations that we studied but it may be but these information will be useful in guiding selective breeding decisions and in other populations and lines so just going to touch very briefly on whether or not there might be a component for the microbiota here clearly host resistance is partly determined by the bird line but how much of that is bird genetics and how much of that might be distinct microbiota I won't go too much into how these studies were done we've heard a lot about microbiome analysis today so you'll be familiar with the methodology what we've done here is try some flora transplants between line six and and we've done this using flora from donor birds that were three weeks old the age at which Campylobacter resistance was previously reported and we've given that donor flora either in a homologous way line six into line six line n into line N or a heterologous way by swapping them over and then after two attempts at doing the flora transplant we've challenged at day 21 with Campylobacter judge and i and what you may just be able to see here is that the resistant line stays resistant whether it had flora from line six or liner and the susceptible line stays susceptible no matter where the flora came from and that might not be surprising when you look at the micro flora from the donor birds where although there was a slight but non significant difference in the community structure when we analyzed the 16s amplicon types by principal component analysis it's not significant and if you look at the phylum and family level we really can't see anything striking about the microbiota in these two lines of birds that are quite evidently different in resistance you can track these things over time as well one of the most striking things here is actually just how dominant ecoli is despite being a minor constituent of the donor microbiota when you look shortly after colonization this is just a day ecoli by far the most dominant of the proteobacteria there and then there's rapidly as a return to some other times you can see some lines specific operational taxonomic units this is something from the genus or solo Spira but you have to look pretty hard to find these effects this is one that appears somewhat more prevalent in the resistant line but of course discerning cause and effect when we look at these associations it's pretty challenging so that's all I'm going to say I'm going to try to keep us broadly to time I hope I can convince you that we do still need to study microbial pathogenesis because for many avian and zoonotic pathogens simple interventions are not fully effective or at the moment they're ineffective and as I mentioned we live in a post genomic era now where we have no shortage of information on what pathogens look like but we have relatively little information on what they encoded genes do during infection and also how the blueprints of these strains relate to their virulence in target hosts and hopefully I can convince you that there are some fantastic new approaches that rely on deep sequencing to tackle that these same genome scale mutagenesis approaches are now becoming feasible for eukaryotic cells in the same way as you can target every gene in Salmonella you can do that now with CRISPR gecko libraries in cell lines and that's going to be a very powerful technology for looking for host genes that may determine responses to pathogens instead of doing or perhaps as well as doing genome-wide Association studies the gap I see that we really need to address in studying host-pathogen interactions is addressing the gap we have with tools to do functional immunology with the bird immune system we really could do with some edited and transgenic chicken lines that will be fantastically powerful resources for looking at the role of different components of the avian immune system in protective responses so that's all I'm going to say I've got very many people to thank from both Roslyn and formerly the Institute for Animal Health my various collaborators and funders and my team and I thank you for your attention be happy to take any questions [Applause] [Music]