Technologies for engineering the microbiome
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
| Title |
| |
| Title of Series | ||
| Number of Parts | 38 | |
| Author | ||
| License | CC Attribution 3.0 Unported: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
| Identifiers | 10.5446/20533 (DOI) | |
| Publisher | ||
| Release Date | ||
| Language |
Content Metadata
| Subject Area | ||
| Genre | ||
| Abstract |
|
1
3
4
8
9
11
12
13
15
21
23
25
33
35
38
00:00
BiochemistryNitrosamineMan pageBiofoulingFood additiveBohriumCancerSense DistrictSauceModul <Membranverfahren>MashingDiethylstilbestrolCell (biology)Anaerobic organismMemory-EffektMeat analogueGeneGene expressionBase (chemistry)Adenomatous polyposis coliAbiogenesisMilchsäureRiver sourceActive siteProbiotikumRiver mouthInflammationFunctional groupWine tasting descriptorsAntibacterialDigestateLuciferasenGenetic engineeringTool steelBase (chemistry)Anaerobic organismRibosomeSynthetic biologyCheminformaticsPlasmidInduktorSpeciesCobaltoxideSetzen <Verfahrenstechnik>TiermodellThermoformingCommon landSecretionCancerSet (abstract data type)Binding energyBiochemistryWalkingCell (biology)DNS-SyntheseDiseaseActivity (UML)Adenomatous polyposis coliOperonMemory-EffektOrganische ChemieAssetGap junctionPhenobarbitalSerineTransfer RNAEukaryoteErythromycinComputer animation
09:26
LuciferasenRibosomeCalcium hydroxideRNAAdenomatous polyposis coliToll-like receptorMan pageSet (abstract data type)Secondary ion mass spectrometryPan (magazine)FirthAncient Roman units of measurementCell (biology)DarmstadtiumStickstoffatomBockBohriumPhase (waves)Hydroxybuttersäure <gamma->GeneGene expressionCombine harvesterZearalenoneAdenomatous polyposis coliTranscription (genetics)GenomeActive siteSense DistrictRiver sourceCarbon (fiber)Systemic therapyGene expressionLibrary (computing)RibosomeBiomolecular structureHerzfrequenzvariabilitätConstitutive equationRegulatorgenSea levelBinding energyShear strengthTranslation <Genetik>Electronic cigaretteOperonDiseaseActivity (UML)Consensus sequenceTachyphylaxieIce frontInduktorConcentrateBasenpaarungPlasmidSetzen <Verfahrenstechnik>Lac operonDermatansulfatGeneShine-Dalgarno-SequenzLactoseAsymmetric inductionLuciferasenFunctional groupPhosphateProteinDeterrence (legal)Controller (control theory)SunscreenIntergranular corrosionRNAElectronegativityColourantCollisionBearing (mechanical)SulfateEntsafterCombine harvesterActinGum arabicSecretionBase (chemistry)RhamnoseComputer animation
18:47
Memory-EffektBase (chemistry)International Nonproprietary NameChromosomeMan pageCell (biology)ChromateRhamnoseSynthetic oilDarmstadtiumInclusion (mineral)MagnetometerSolutionPolymyxinSet (abstract data type)S-Adenosyl methionineBohriumPsychological traumaStimulus (physiology)NucleaseCollectingRNAHydroxybuttersäure <gamma->Battery (electricity)DNA replicationSystemic therapyLuciferasenRNACollectingResistenzPolymyxinWine tasting descriptorsGeneActivity (UML)Active siteBiomolecular structureDNS-SyntheseTranscription (genetics)GelatinPlasmidKlinisches ExperimentCell growthRiver sourceCarbon (fiber)FructoseGenomeInduktorAbundance of the chemical elementsRing strainAcetoneThermoformingStop codonWaterCell (biology)Growth mediumPhagocyteWursthülleAntibacterialBasenpaarungSchwermetallAntibiotic resistanceMicroarrayFunctional groupButcherToxicityController (control theory)ToxinRecombinasenDerivative (chemistry)KohlenhydratchemieSea levelAsymmetric inductionNucleaseAdenomatous polyposis coliMemory-EffektMultiprotein complexLibrary (computing)AntitoxinRadioactive decayCast ironGene expressionMatchIceElectronic cigaretteActinIronRadioactive tracerCollisionComputer animation
28:08
BohriumHydroxybuttersäure <gamma->NucleaseMan pageTrihalomethaneSodium hydrideSilicon carbidePhagocyteMagmaVancomycinAltbierSynthetic oilCHARGE syndromeBis (band)S-Adenosyl methionineElfCalcium hydroxideHost (biology)GenomeEnolSample (material)CigarAttachment theoryGeneMagnetometerSequence alignmentPitting corrosionToll-like receptorDominican RepublicConjugated systemDNS-SynthesePhagocytePlasmidWeinfehlerElectron donorHost (biology)Amino acidVirulenzfaktorMolecular biologyTarget cellRing strainBasenpaarungAnomalie <Medizin>ThermoformingSurface scienceAlpha-1-RezeptorActivity (UML)SolutionHope, ArkansasSetzen <Verfahrenstechnik>MedicalizationAntibodies (film)FaserverbundwerkstoffConsensus sequenceAdenomatous polyposis coliCollectingChromosomeBiomolecular structureGeneSystemic therapySet (abstract data type)DiseaseRNAStereoselectivityGenomeGenotypeFood additiveAntibiotic resistanceTiermodellSurvival skillsDeathSpaltflächeBuffer solutionCell (biology)AcidResistenzAgeingSunscreenIntergranular corrosionWaxNalidixic acidWursthülleAsparagineStorage tankDeterrence (legal)Separation processReducing agentInitiation (chemistry)Binding energyComputer animationLecture/Conference
37:29
Adenomatous polyposis coliVancomycinMan pageBohriumMagmaHost (biology)GenomePhagocyteGap junctionAcepromazineCell cycleMixtureTrihalomethaneFaserverbundwerkstoffCell (biology)CloningPrimer (film)VerdünnerMagnetometerHydroxybuttersäure <gamma->GraukäseCloningPhagocyteWursthülleCombine harvesterFaserverbundwerkstoffCapsule (pharmacy)LipopolysaccharideFaserplatteGenotypeBiomolecular structureN-terminusSaccharomyces cerevisiaeProbiotikumC-terminusSetzen <Verfahrenstechnik>GenomeRiver sourceRing strainSpeciationColloidHost (biology)DNS-SyntheseLegierenSynthetic oilBinding energyOrders of magnitude (radiation)Yeast artificial chromosomeSystemic therapyAdenomatous polyposis coliGap junctionScreening (medicine)GeneChromosomeBaker's yeastTool steelYttrium aluminium garnetArtificial leatherElectronic cigaretteOperonWine tasting descriptorsAnaerobic digestionCancerPlant breedingCell (biology)BackhoeAgeingChemical structureTeaStress (mechanics)Computer animation
46:50
Diagram
Transcript: English(auto-generated)
00:17
Thank you very much to the organizers for giving me an opportunity to speak about what we do in the laboratory of Timothy Loo.
00:25
I apologize, he cannot be here so he sent me instead of him. The Loo Lab is a large operation with many different aspects ranging from biomaterials to eukaryotes in bio with activities in cancer detection and treatment and also production platform.
00:44
It's also involved in biocomputation, so building all of those gates that SynBio is so interested in. What I'm going to talk about is the last aspect that the lab develops, which involves possible applications of SynBio in infectious diseases treatment
01:03
and microbiome engineering, all of that being somehow linked together. The human body is not just eukaryotic cells, it's also a very, very large amount of microbes, trillions of commensal microbes and a very diverse type of microbes. The gut microbiome, for example, is completely different from the skin microbiome
01:27
which is itself completely different even from the head microbiome. And what I'm going to focus on today is the gut microbiome.
01:40
The microbiome, especially the gut microbiome, is heavily involved in a lot of different aspects of health that range from development to immunity, eliciting at the very beginning of life, digestion, mood, behavior and so many other things that are still to be discovered.
02:01
Unfortunately, the tools to manipulate it when some problem is discovered are extremely limited. They are basically limited to wiping it out with antibiotics, possibly introducing pro or prebiotics although their efficiency is heavily debated as they are.
02:22
And once you've wiped it out with antibiotics, maybe replacing it with the healthy microbiome of someone else. And that's about all we have as of today if you want to replace a microbiome. So transplant, change the microbiome with something completely different which may or may not be healthy.
02:45
What we want to do is to have a much more delicate approach to microbiome engineering. We'd like to be able to either selectively add a bug with predetermined functionalities to an existing microbiome
03:03
or possibly remove a bug which is involved in a pathogenic situation. And this requires tools that do not exist as of now. So the idea that we want is there may be a problem somewhere in the gut.
03:22
You design a bug which will then be ingested, travel through the stomach all the way through the intestine, reach the site of inflammation, cancer, you name it, and then elicit a medical action to cure the problem.
03:40
As I have written here, going from the mouth to somewhere in the gut requires passing through a lot of different environments. And this is where most of the previous applications of microbiome engineering, probiotics and so on failed. Very few bugs are capable of adapting quickly enough and well enough to all those diverse environments
04:00
to actually still be alive when they reach the site where they may do something. So that's one of the challenges. What we expect is that synthetic biology with its capability to design desired functions inside of microorganisms can help with designing microbes that will face this.
04:24
This is a very common description of synthetic biology where it is usually considered to emanate from a concept which leads to a design. The design may be modelled and simulated and there will be none of that today. Then it is constructed, probed, in vitro in some form of isolated culture, measured, possibly evolved if something is not optimal.
04:50
So rare, right? And then if it still fails, maybe go back to the design, go through another model iteratively until you manage to build whatever it is that you really wanted to build.
05:02
And at MIT we are an engineering school so we are very much for that top bottom approach where you have an idea and no matter what you will build it. Lots and lots of work on somehow modifying the microbiome was based on E. coli.
05:23
The vast majority of the probiotics that exist today are either coli or lactic acid bacteria. Coli is slightly better than lactic acid bacteria but none of them really colonise the gut microbiome easily.
05:41
Gram-negative of the type of E. coli are maybe something in the range of 0-1% of the gut microbiome so it's almost negligible. It's not completely useless but very much so. So we are looking for other possible bacteria that would be more suited to creating a probiotic that would be stably maintained in the gut.
06:07
And one such bacterium is a bacteria, it is Theta-Aeota micron. And you will forgive me if I just call it theta, it's way too long. So theta is just like colliogram-negative bacterium.
06:20
It's an obligate anaerobe but it tolerates oxygen pretty well. It just does not grow but at least it does not die. It's a member of a dominant phyla in the microbiome. It accounts for about 30-40% of the total number of species in an overage human gut microbiome.
06:40
It is very abundant up to 10-10 CFU per gram of stool in an overage human microbiome. And is present in a large amount of microbiomes of humans.
07:01
And finally it is relatively amenable to genetic modification. It's not nearly as easy as colli but it's feasible. The problem with B-Theta-Aeota micron is that although it is amenable to genetic modification, there are hardly any tools yet. And suddenly nothing like we can have in colli where we have banks of various inducible or considerative promoters,
07:25
large amounts of plasmids which are compatible so that you can add gazillions of circuits inside of the exact same cell. So the question is how can we transfer all of those different biological computations?
07:45
So those are examples of biological computations that were devised in the lab. So the lab devised various methods for integrating memory inside of bacteria. We also devised ways to make analogue computation. We worked also in digital computing like many others.
08:05
We also devised ways of rewriting DNA directly in vivo inside of the cells. And all of that was developed for colli but how can we transfer all of that into theta when we don't even have a single promoter which is properly described.
08:22
So the first step in doing what we wanted to do was to develop a set of basic promoters, inducible and constitutive, a set of ribosome binding sites, the very basic set of tools that you need to build any circuit into any living organism.
08:44
We started from that simple plasmid developed in the Gordon Lab, PNBU2, which is an integrative plasmid in theta iota micron that integrates here, catalyzes its integration into a serine tRNA. The good thing about that integration inside of a serine tRNA is that there are only two in theta.
09:06
So if you inactivate one with your construct, the other one has to be here. So that limits the possibility of having multiple integrations. And it is selected with erythromycin and it's a suicide plasmid obviously in theta.
09:21
And it has here a nanolook which is a smaller version of luciferase that gives you a readout for anything you clone in front of it. So what we'll do from that PNBU2 sequence is to vary ribosome binding site sequence just in front of nanolook, vary promoter sequence,
09:43
assay luciferase activity in different conditions and see what we can do. And finally, engineer all of that to see if we can change levels. In Coli, the ribosome binding site is well defined. It has a consensus sequence of AGG AGG that matches well to the end of the 16S RNA.
10:03
All of that is good. And if you modify the sequences on both sides of the ribosome binding site or within the ribosome binding site, you can change the level of expression of genes from zero to maximum, which is whatever Coli is capable of tolerating.
10:22
In bacteriologists, the consensus sequence of the ribosome binding site is not that different. It matches to the end of the 16S RNA just as well. And so the idea was, well, if we do exactly the same thing as in Coli, then that should work, right? That looks feasible.
10:41
And even the natural Coli ribosome binding site should work well enough. So there shouldn't be too many problems with translation. So that's exactly what we did. We took our PNB U2 plasmid. We cloned a known constitutive promoter from theta in front of it.
11:04
The Shine-Dalgarno was that of that BT1311 gene. And we started making libraries by randomising the sequences around the Shine-Dalgarno and assessing function. So just for reference, this is the kind of thing you can get in Coli
11:21
when you modify the sequence of the ribosome binding site. So you get from very high to absolutely zero with some of the mutants you make. Interestingly enough, when you do exactly the same thing in theta, you go from high, so that red line here is the natural ribosome binding site,
11:43
the one we started from. So we can increase a little bit, maybe half a log. We can decrease some, we can decrease by about two log. But there is no way to make it completely negative. For some reason, you can put any sequence and no matter what you will still get some level of expression.
12:01
Some people suggest that actually the pairing with the 16S RNA is not the major determinant of specificity of the ribosome binding site and that in theta it's more binding to the S1 protein of the ribosome which dictates recognition. So there's a lot less of the base pairing involved
12:22
as recognition of an actual phosphate backbone, ribophosphate backbone, because I don't think anyone knows what it really binds. Nonetheless, we could still get something in the range of about three logs difference between the strongest and the weakest of our ribosome binding site.
12:42
But this large difference in behavior between coli and theta probably explains why a lot of parts that were brought from coli into theta completely failed because there are obviously immense differences in how the two bacteria work.
13:02
Now we decided to look at promoter. So in coli, promoter specificity is dictated by a sigma factor. There are several in E. coli, I'm not going to go into the detail. Each of those sigma factor recognize consensus sequences which are different and so exactly the same way as you could do with the ribosome binding sites earlier
13:23
if you modify either those consensus sequences or the sequences in the middle you can vary the strength of the promoters. Theta has a completely different promoter architecture. There is a single sigma factor which is responsible for all transcription in the genome of the bug.
13:47
And the consensus sequence is pretty different from that of a coli with instead of a minus 10, a minus 7 box and instead of a minus 35, a minus 33 box which does not look very much like that of coli.
14:02
So when you have a circuit that works in coli it's pretty unlikely to work in theta directly because if the ribosome binding sites may possibly work there's no way transcription will ever work. So we started building libraries of constitutive promoters using the randomization strategy that I told you.
14:24
So this is our starting promoter, PBT1311. We also cloned a few other promoters from B theta that were expected to work well. And we also started introducing variability in all those various regions
14:40
highlighted in blue here either replacing those regions with those from other promoters or from completely random sequences. And I'm just showing you a few examples of the promoters we got out. And all together the different constitutive promoters we managed to build span about two logs of magnitude in expression level.
15:02
So when you combine the different transcription levels we're capable of reaching with the different translation strengths that we're capable of doing with the ribosome binding site library we get about five log difference which is about four logs more than whatever was possible before.
15:22
But constitutive expression is still fairly limiting. A lot of the constructs we use for any circuit require being capable of turning things on and off on demand. So we wondered how we could obtain inducible systems. Bacteriodis theta eutemicron, theta,
15:43
has an extremely varied array of carbon source usability. It is capable of degrading a lot more carbon sources than coli and just like coli it senses them and induces the genes for their degradation only on demand when the carbon source is there.
16:01
So there is a large amount of systems that can be borrowed from the genome for sensing at least carbon sources. One such example is the ramnose regulation system. In theta ramnose sensing functions very much like arabinose sensing in coli. So you have a single regulator called RaR that binds to ramnose
16:24
and then activates the ramnose operon promoter PBT 3763. And so if we clone that promoter in front of Nanoluke you can see that while we increase ramnose concentration
16:41
we increase expression of Nanoluke, of luciferase and we get an about 104 fold difference between absence of inducer and inducer which is pretty good. We screen the genome for other types of inducible promoters a lot of those carbon sources are sensed through kinase response regulator systems
17:07
two component systems although those of theta are different from those of coli. And one such example is the chondroitin sulfate response regulator BT 3334 which senses chondroitin sulfate, activates the promoter
17:23
and we got about 60 fold with that system or the arabinogalactan system encoded by PBT 0268 which gives us about 29 fold induction between absence or presence.
17:40
Now we had a number of simple parts we wanted to check to what extent we could build more complicated parts so try to import functions that are not native to theta. And the most simple system to do that is probably the lac system of coli. So we wanted to see if we could get lacI to function in theta.
18:02
So we put lacI under the control of theta promoter, PBT 1311 actually and then created a mutant of the PBT 3111 promoter with lac operator sites at various locations.
18:21
And we build actually all possible combinations of one, two or three operators. Interestingly, they all worked. They did not work great. There is a limited dynamic range in those promoters but what's interesting is that by modifying the relative position of those lac operators
18:44
we could change the sensitivity of those promoters. So that gave us the possibility to induce at various levels of IPTG. Finally we tested whether all of those promoters that have different inducers but remember they're all sugars or derivatives of sugars
19:04
were orthogonal to each other. That is if the inducer of one system would induce another promoter. And what you can see on that graph is that there is induction when you combine the inducer of a given system with its corresponding promoter but absolutely not any induction with the other inducer.
19:23
So they're perfectly orthogonal. So now we have a good library of parts for transcriptional control and we wanted to see if we could go a little farther in complexity. And what we've done is to see if we could replicate memory integration inside of Theta.
19:44
The idea behind that memory integration is pretty simple. It uses the capacity of integrators to flip pieces of DNA between inverted recognition sites for that given integrators. And we used a system developed by our neighbours in the VOID lab
20:02
where they built an array of different integrase recognition sites along with all of the integrators that specifically recognize each of those sites. We only cloned four of them as a test trial and checked what happened if we would transform Theta carrying this on its genome
20:24
with plasmids constitutively expressing the integrase. And what these gels tell you is that in the presence of the corresponding integrase you get the expected inversion so everything works great.
20:43
We then checked briefly if it worked with an inducible integrase and to make a long story short, it does. It does actually beautifully much better than in coli because we have absolutely no activity of the recombinase when we don't add any inducer which is rare of coli promoters.
21:04
The vast majority of coli promoters have some leakiness, especially in those systems. Still pushing further in complexity, we wanted to check if we could benefit from all of the advances of CRISPR-Cas. So I'm expecting everyone is familiar with CRISPR-Cas.
21:23
But in case not, very briefly, Cas9, the natural Cas9 is a nuclease which requires a small RNA to direct itself to the appropriate sequence in the DNA that it then cleaves. Here we're using a dCas9 which is a deactivated Cas9 so the active site of the nuclease were modified so that it can no longer cut DNA
21:46
but it can still land on it and stay there. So what we were hoping with a dCas9-based system was that we could perturb expression from genes. The first system we built was pretty simple.
22:01
It used an IPTG-inducible dCas9. We would clone whatever guide RNA we need to direct dCas9 to given sequences within the promoter or the coding sequence of Nanolook and then we would measure the ciferase activity as a function of IPTG.
22:21
And what this graph, or this graph, they both show the same thing, show you is that it works beautifully when you add IPTG then luciferase activity goes down indicating that the dCas9 really does land here and does prevent transcription. It's a little surprising though that even the single guide RNA is targeting
22:45
the coding sequence of luciferase still had an activity. Again, this is a big difference with Coli. In Coli, dCas9 really only works if you target within the promoter and as soon as you get outside of the promoter it quickly sees working.
23:07
We then wanted to check again with that system if we could target those genus genes. These two genes here are involved in polymixin B resistance so we wanted to check if by using the dCas9 targeted system
23:22
we could inactivate those genes and thus decrease resistance to polymixin and that's indeed what happens. We also use that exact same system but targeting some of the carbon source usability operands of θ and you can see here that when we induce
23:44
the dCas9 system with IPTG then growth on fructose stops pretty abruptly indicating that the targeting works as expected. Everything tends to work pretty easily in the lab in culture medium.
24:03
Very often when you go to a more natural environment things stop working. So we wanted to see to what extent all of those parts that we had built would still work within some form of natural environment and we chose mice because they're easy to manipulate.
24:21
So what we did was to treat mice with tratomycin to strip them off of most of their microbiome then remove antibiotics, colonize with various θ strains that would not have seen the deconstructs and then treat those microbiomes with inducers as needed,
24:43
collect the stools and then measure luminescence and measure the presence and abundance of θ by qPCR. The first system we tried was a very simple arabinogalactan induced luciferase system just to see if anything would happen whatsoever and what you can see here is that this represents times when there is arabinogalactan
25:06
in the water of the mice and you can see that there is a sharp increase in luciferase activity in the stool as we induce and it goes back to baseline very quickly after we remove inducer. So it seems to work well.
25:22
We then tested the CRISPR system in a similar system so it's exactly the same system as I presented earlier. It's just a lot more complicated so there are a lot more places where things could go wrong. Nonetheless, it still works okay. We're probably reaching the border of what no longer works
25:41
but here again we have IPTG addition and you can see that when we add IPTG, luciferase activity drops which is exactly what we would expect when it stays perfectly flat in the system that doesn't have all of the regulatory systems. So here I have described an additive system
26:05
to add a programmed microbe inside of the microbiome. The problem is doing it still requires a pretty heavy treatment with antibiotics to strip off everything that is not really desired and replace it with what we want.
26:20
So we are now focusing our efforts on trying to build systems to selectively remove a given member of the microbiome and replace it with what we want, a subtractive approach to microbiome engineering. Our first attempt at that was using again Cas9,
26:41
this time the real Cas9, the one that cuts DNA and was aimed at targeting antimicrobial resistance gene in a natural environment. So the idea was pretty simple. You have delivery vehicle, in this case phage M13 because it's very easy to engineer. That M13 delivers a piece of DNA that has the Cas9,
27:05
the trace RNA which is required for Cas9 functionality and most importantly arrays of guide RNAs that direct Cas9 to their target and the target may be in the genome or in plasmids and in most cases in bacteria when you make a double strand cut
27:21
the cell just dies. It's particularly true on plasmids because most plasmids have toxin-antitoxin systems and if you make a double strand cut very quickly the antitoxin degrades and the toxin kicks in, killing the cell.
27:40
So results can be seen here. Here you have all of the control strains. That means they do not have the target for the Cas9 system and you see that there is absolutely no toxicity. So if the target for the guide RNA is not here, nothing happens. However you can see here for example that in a strain that has NdM1
28:03
the phage that targets the NdM1 sequence kills the cells. Here you have a similar system targeting another type of antimicrobial resistance genes and exactly the same thing happens. Of course it is very specific to the sequence
28:22
and you can even mix the phages together and target both at the same time if you want. You can also direct the Cas9 cleavage to the chromosome. This is what we've done here. We've isolated a mutant of our test strain EMG2
28:42
that is Naledixic acid resistance, so it has a mutation JAR-A. This is a single base pair mutation and the guide as you can see is extremely specific. It kills the mutant, it does not kill the wild type. What we've done next was to see if we could counter select
29:06
virulence factors. So here we've been targeting the Intamin gene that we've heard about yesterday I believe which is a major virulence factor of enterohymorrhagic coli. Once again with the proper guide RNAs
29:28
we managed to eradicate the target cells with our engineered phages. We've also looked at whether or not these engineered M13 phages
29:41
could help with survival in the case of infection. We've used a waxworm model here and we're following death of the worms after being injected with either a non-sensitive cell, so bacteria that cannot be targeted by our engineered M13 or cells that can be targeted with the Intamin targeting phage.
30:05
What you can see is that if you don't infect the waxworms, they don't die. If you infect them with a phage which does not target the strain that we are also putting in the worms, they die.
30:22
If you treat them with SM buffer such as NE buffer, they die. But if you treat them with the Intamin targeting phage which is the one that can kill off the bacteria that we're injecting they survive significantly better. There is one major flaw with that project.
30:40
M13 is wonderful for molecular biology. It's small, it's easy to engineer. You can do a lot of things with it. The problem is it's absolutely F dependent. It's a conjugated plasmid and F is extremely rare in nature. So this is a very nice system in the lab, absolutely useless in real life.
31:01
So we have to find other delivery vehicles. The system can work but we need other delivery vehicles. Phages are probably one of the most efficient systems for delivering DNA to bacteria. Conjugation requires way to intimate contacts to be useful again
31:20
in the context of a microbiome where you have billions and billions of other bacteria. So the probability of encounter close enough between your donor bacteria and your recipient bacteria is way too low. So classically if you want more phages you isolate them from nature and you end up with a wide variety of phages that may or may not be strictly lytic
31:45
that may or may not integrate in the genome that may or may not possess virulence factors and you assemble them into a cocktail of phages that collectively target all of the possible strains that you need to target.
32:03
The problem with that is that for every phage you add to your cocktail you need a whole lot of tests to make sure that everything works the way you want it to work. So the approach that we are trying to use is what if we could transform phages into some form of antibody where we keep everything that makes it a phage,
32:25
that makes it a good delivery vehicle and change what makes it target a given type of strain so that you can develop a set of phages originating from the exact same chassis but targeting different bacteria. Maybe a graphical visualization will make it clearer.
32:43
Let's say that these are all the strains you need to target to treat any given disease. They all have a different envelope, they may have different sets of genes inside of them that may provide defenses against phages. Traditionally you go into a sewer plant or whatever natural environment suits you,
33:02
gather a few liters of the effluent and pretty easily you find phage against pretty much any bacteria. So that's what you see here, you find phages and they may infect a single bacterium or infect several and it's pretty unpredictable.
33:20
Then if you want to use them in any kind of industrial setting they need to be tested one by one for specificity, for stability in terms of storage, for their biology, do they have virions factors, are they lysogenic or strictly lytic. Then if it is for a medical application you also need to test safety, efficacy, delivery,
33:42
all of that one by one before you finally think about putting them all together and have a product which obviously because bacteria where they are they evolve so that cocktail will need to be updated probably twice yearly, so twice a year you need to go through all of that for your new cocktail.
34:01
That just cannot work. So what if instead of that we use a single chassis, a single phage which we have decided is a good starting point, make all of those testings so that we know everything about it, everything we can. And for some of the phages we already have this is almost there.
34:24
And then simply grab from all of those natural phages or all of the DNA banks that we have the sequences that we need to make that single chassis infect all of the strains. Hopefully then limiting the amount of testing you need to make to get it approved for any kind of medical application.
34:43
Obviously at this point this is just a project. But actually the FDA at least is pretty interested in it and we have some fair amount of hope that this can be a viable solution to phage therapy in western worlds.
35:01
So let's test it. The phage we decided to use for initial testing of that ID is the family of T7. The reason why we decided to use T7-like phages is because they are very widespread in nature.
35:20
They are very easy to use in the lab. They are well studied and understood. They have a relatively short genome with few activities so not too many chances for surprises. And most importantly they are extremely host independent. Outside of their receptor at the surface of the bacterium the only other gene they need from their host
35:46
is the siroredoxin which is an adjuvant to their own RNA polymerase to make it more processive. So there is hope that it can actually work in a wide variety of bugs.
36:02
The other reason is bioinformatic. If you align all of the tail fiber genes which is the major host determinant in T7 from all of the sequence T7-like phages together what you find is that the first about 150 amino acids are pretty conserved
36:20
whereas the rest is not conserved at all. The consensus to explain that is that that 150 amino acid and terminus is involved in binding the tail fiber to the phage because the rest is different because it infects different bacteria. So it needs to recognize different receptors.
36:41
So one of the ideas is can we start from T7 or T3 and graft inside of either one of them any of those other tail fibers thus changing host range. Summarized here. So you take phage A that infects bacterium A, phage B that infects bacterium B.
37:02
What happens if you graft the tail fiber gene of phage B inside of bacterium phage A? Does it go after bacterium B, bacterium A or something completely different? To start simply we started with T7 and T3.
37:22
They are both caulifragous, they are both well studied, they are both understood but they don't both recognize the exact same receptor. While T7 binds somewhat deeply in the LPS of coli, T3 binds at the very top. So that in principle, and it's been demonstrated in some cases,
37:45
exchanging the gene should change the phenotype. And we can indeed see that here while T3 and T7 infects B strains perfectly well
38:01
as well as regular K12 cloning strains, T3 here does not infect MT1655 or BW25113 with T7 dots. So we now have a screen if we change the GP17 genes to...
38:20
I'm going to skip because time is running out. So now how are we going to do that? Because engineering phages is much more complicated than it seems. Traditionally it's done with all the replacement systems. To make it short because I'm running out of time, this rarely works. It's long and painful. So what we've developed is a system that uses the gap repair system of Saccharomyces cerevisiae
38:47
to completely reconstruct the genomes inside of yeast before putting them back to life inside of coli. So the system is pretty simple. You isolate the genome of whatever phage you want to play with.
39:01
You need a yeast artificial chromosome which will make the final construct replicate inside of yeast. You transform either the genome plus the YAG or various PCR products that span the entire length of the chromosome of the phage with ends homologous to the yeast artificial chromosome
39:20
and yeast by some miracle reassembles that into a fully assembled genome. You extract that yeast artificial chromosome, transform it into bacteria and if you've done your design properly you get functional phages which you can then maintain, grow, play with. Obviously in some cases it's not going to work
39:40
but at least compared to other replacement methods we have the yeast clone. We can at least go back, sequence regions, check if it didn't work because something happened during the PCR, there is a mutation and the phage is dead or if that's simply our design which is wrong and we need to go back to the drawing board. Interestingly enough that works with a lot of different phages
40:02
so we've tested it with a number of different T7-like phages and again, long story short, it works with all of them independently of what host they normally infect. And we built from T7 a T7 that has the T3 tail fibre and a T7 that has just the C-terminal part of the T3 tail fibre.
40:26
Remember that only the N-terminus is conserved. We had no idea if the T3 tail fibre would be able to bind the T7 capsid by itself. So here is again all of those phages plated on BL21 which is a common host for them
40:48
and then tested against a selective host. You can see that while T7 still grows on MG1655 the reconstructed T7 has exactly the same phenotype
41:01
but the T7 with the T3 tail fibre no longer grows on MG1655 and conversely for the T3 mutants. So the idea works, if you change the tail fibre genes you can change host range. We then wanted to see if we could cross species barrier. Here we resorted to a completely synthetic approach to making the genomes.
41:24
It turns out that that phage R that infects Yersinia only has three-point mutation difference with the T3 gene 17, so the tail fibre genes. So instead of sourcing out gene R, amplifying the whole GP17
41:40
we just decided to order a piece of DNA that corresponded to that region where they are the mutations and then go through the exact same assembly process to make a T3 phage with the R C-terminus of the tail fibre. And you can see here that while T3 does not grow on Yersinia
42:01
T3R grows perfectly fine on Yersinia and kills it extremely efficiently in vitro. We wanted to go even farther and try something a little more difficult. We had tried K11 here. Sorry, it's nowhere.
42:21
K11 is Klebsiella phage. Klebsiella is very different from Coli in that they are all capsulated in sharp contrast with the laboratory Coli which all have a rather rough phenotype. So a capsule is something that a normal tail fibre cannot bind to. It's a thick layer of mucus surrounding the cell and that prevents the phages that recognise LPS from actually finding it.
42:44
So we wanted to see if we could graft onto T7 something that would allow it to go through that capsule and infect Klebsiella. So our first attempt was to simply replace the tail fibres of T7 with those from K11. But that did not work.
43:01
We don't know why, but it did not work. So we started trying out all of the possible combinations and it turns out that if we replace the whole tail the tail being composed of GP11, 12 and 17 then we get a phage that infects Klebsiella just as wild type. So this is T7 on Coli, T7 on Klebsiella.
43:24
You see it does not infect, there are no plaques. And here we have K11 wild type. It does not infect Coli but infects Klebsiella. And if you bring the tail of K11 into T7 then you get a phage that grows on Klebsiella. And conversely, if you bring the T7 tail inside of K11
43:43
you get a phage that grows on Coli. So that tells us that although tail fibre swapping is a method that can work it's not always sufficient. And that points out to the extreme lack of knowledge we have in the structure-function relationship in phages.
44:02
Because there is no reason why just cloning the C-terminus of K11 GP17 would not work. There is no bioinformatic way to predict that. Still it did not work. So we are here hoping for, as Patrick was saying
44:23
maybe more predictive tools, maybe more bioinformatics. Maybe it's just that the tools that we have are not sufficient. What I want to point out, and that will be the end and I will almost be on time is that we used those engineered phages in synthetic bacterial consortia
44:40
to check if they could work just as well as well-typed phages in removing their specific bacteria in the context of some simple microbiome. This is all in vitro. We are doing animal experiments right now but they're not at all ready. So the idea is pretty simple. First of all, you mix in approximately equal amount probiotic colloid NISL,
45:06
the target strain for K11, the target strain for R, Yasenia, and then you try the various phage combinations and see what happens after a short while. And what you can see is that even in 30 minutes when you add K11 onto that synthetic consortia,
45:26
Klebsiella is completely annihilated. So on a pie chart like that, it's difficult to see but we have an about five order of magnitude decrease. So there is a very tiny croissant here. It's just not visible.
45:41
And the same thing is true if you use the engineered phage T7K11GP111217. Although just as we could see earlier, it's not quite as efficient. It's probably good enough but it's not quite as good. But if you give it a little more time, it still works. And if you mix T7 which targets Yasenia and the phage target in Klebsiella,
46:06
then you can completely remove the two pathogens in about an hour. That being said, I would like to acknowledge members of the lab. As I said, the Lu lab is a big operation, lots of people.
46:21
So particularly Hiroki Ando who worked with me on all of the phage engineering work. Rob Sitoric who worked on the M13 Cas delivery. Fahim and Mark who worked on the bacterioteus aspect of things along with our collaborators from the Chris Voigt lab.
46:42
And all of the other members of the lab and our funding sources. Thank you very much.
Recommendations
Series of 5 media
Series of 6 media
Series of 8 media
Series of 4 media
Series of 7 media