Synthetic Biology for the Development of New Antibiotics

Top

00:00:11.17 Hello, my name is Eriko Takano and I'm
00:00:13.18 from the University of Manchester.
00:00:15.16 And today, I'd like to speak to you about synthetic
00:00:17.29 biology for the production of high-value chemicals.
00:00:21.02 So as you all know, antibiotic resistance is a very
00:00:24.25 big problem, and it's a worldwide problem.
00:00:27.10 You can see here there's methicillin-resistant
00:00:31.12 Staphylococcus aureus increasing, even enterococci
00:00:35.26 that are resistant to vancomycin has emerged
00:00:38.19 and is on the increase. There are many hospital infections
00:00:42.28 and of course resistance is absolutely increasing.
00:00:46.16 You can see also from this European map
00:00:49.06 as well, that the proportion of MRSA which has close to 50%
00:00:53.29 that there are many countries that do.
00:00:56.17 Now, antibiotics are produced by actinomycetes
00:01:00.13 and they're gram positive, soil dwelling bacteria.
00:01:04.24 You can see spores here, this is scanning EM
00:01:08.11 picture with spores growing up from the soil.
00:01:12.16 Here another picture of streptomyces colony taken
00:01:17.01 from the top with blue blobs, which are all antibiotics
00:01:20.06 that are being secreted out from the cells.
00:01:23.11 Another cross-section here, you can see this
00:01:27.04 is a red antibiotic being produced and this is being
00:01:30.02 retained in the cell. Now you can also see some chemical
00:01:33.26 structures here, and some very simple structures
00:01:37.14 like penicillin to some very, very large and complex
00:01:41.02 structures like daptomycin. So you can
00:01:42.29 see that these antibiotics have a very
00:01:45.29 diverse chemical structure. Now, the first antibiotic
00:01:50.18 to be discovered was penicillin. In those days
00:01:53.01 it was called the antibiotic -- the miracle drug.
00:01:57.12 And in the 1960's, as you can see her, there was a huge
00:02:00.29 peak of antibiotic, novel antibiotics being discovered.
00:02:06.11 But ever since then, there's been a decrease. And even
00:02:09.19 now, the most recent antibiotic to be found was daptomycin.
00:02:15.24 But as you can see, resistance is increasing.
00:02:20.13 So why do we have this decrease? Well
00:02:23.17 some people say that it's because the industry
00:02:25.24 is no longer interested in looking for natural products.
00:02:28.13 Why? Well, it's not very profitable. Every drug
00:02:32.24 takes a lot of money to actually get it through
00:02:36.01 the to the market, but then antibiotics you only take it for
00:02:40.22 two weeks. So of course, it's not as profitable as
00:02:44.08 other drugs, compared to other drugs. Some other people
00:02:47.19 say that well, there's no antibiotics to be found anymore
00:02:51.11 in nature. And we know that's not true.
00:02:53.25 Because recently, we've genome sequenced
00:02:57.01 an actinomyces called Streptomyces clavuligerus.
00:03:00.04 Clavuligerus is a commercial producer of beta-lactamase
00:03:04.22 inhibitor. It's being used currently, as well, together with
00:03:08.20 beta-lactams to induce its activity. Now when we genome
00:03:12.14 sequenced this and analyzed it, we found there's
00:03:15.00 more than 50 secondary metabolite gene clusters
00:03:18.02 there. Now clavuligerus is known to produce
00:03:21.04 4-5 compounds, so all the rest, about 45 of them,
00:03:25.07 they're all either asleep or they're being produced
00:03:29.12 in such small quantities that we just cannot detect
00:03:32.06 them. Now, it's not atypical for clavuligerus because
00:03:37.01 we've recently found a global microbial genome
00:03:39.20 analysis of all secondary metabolite pathways.
00:03:42.14 Now if you look at this green part here.
00:03:45.18 This bar, the height of the bar tells you how many
00:03:48.23 secondary metabolites there are on the genome.
00:03:51.07 Now if you look at the actinomyces, you have
00:03:54.05 quite a few, but if you look at all the other bacterias
00:03:57.26 as well, they have quite a few as well. You can
00:04:00.23 imagine there's a load of secondary metabolite
00:04:03.22 gene clusters out there waiting to be discovered.
00:04:06.17 Now we've done some proof of concept studies, as well,
00:04:09.26 to awake some of these clusters. Here we've deleted
00:04:14.26 this repressor and you can now see a yellow compound
00:04:19.12 being produced by this mutant. This is the parent
00:04:22.13 here, which normally produces a blue color.
00:04:24.19 And we've also been able to show that this gene
00:04:28.00 cluster is responsible for producing a compound
00:04:31.18 with antimicrobial activity, and now this yellow compound
00:04:35.14 the chemical structure has been elucidated. And you can
00:04:40.00 see, it's a complete novel compound. So if you think
00:04:44.16 all of these potential secondary metabolites, if they were all
00:04:48.08 awakened, we're absolutely sure there should be new
00:04:51.18 chemical structures in there. We can find diversity
00:04:54.22 and we will also be able to find novel
00:04:57.28 antibiotic -- antimicrobial antibiotics. Well, but
00:05:03.02 to awaken all this, if we tried to do it like I did before,
00:05:06.28 deleting one gene, overexpressing a promoter,
00:05:10.14 this takes much too long. We need to be ahead of the game
00:05:14.20 of the antibiotic resistant bacterias. So we need something
00:05:18.06 much more systematic, high-throughput, much faster. And so
00:05:22.28 that's why we want to use synthetic biology.
00:05:24.23 What's synthetic biology, it's to engineer new
00:05:28.24 life forms with unrestrained versatility, which means
00:05:32.13 your imagination is really the limit.
00:05:35.26 And it could be the next industrial revolution.
00:05:38.28 And it's using the concept of engineering, of design,
00:05:43.20 build, and test, and learn, and bringing it together
00:05:48.02 with biology. So here are some examples of synthetic biology,
00:05:52.16 for example, a total synthesis of a functional designer
00:05:55.29 eukaryotic chromosome from yeast. Another example
00:06:00.05 which I personally quite like very much is projects from
00:06:03.25 the iGEM competition. iGEM is the International Genetically
00:06:07.07 Engineered Machine competition. We have this every year,
00:06:11.22 all the undergraduate students from all over the world
00:06:14.20 participate. There are about 300 teams and they make new
00:06:19.23 engineered microbes using standard parts.
00:06:24.13 So here's one example, which is quite nice.
00:06:27.16 In 2013, Heidelberg was the world champion, and what they did
00:06:32.14 was try to recycle gold from electronic waste.
00:06:35.29 Another example would be using bacillus
00:06:39.13 as a biosense for meat spoilage. So bacillus turns blue
00:06:44.02 if the meat's spoiled. Another example, of course, is the
00:06:48.28 famous example from artemisinin. So Jay Keasling's lab
00:06:53.17 has used all these enzymes from yeast and also
00:06:57.12 from plants to produce artemisinin in Saccharomyces
00:07:01.10 cerevisiae, where originally it was produced in
00:07:04.11 a plant. So this is great. You know, being able to produce
00:07:09.08 artemisinin in a non-natural host. But can we do something
00:07:14.28 more than that? Can we take this one level higher?
00:07:17.01 Can we use synthetic biology to produce compounds
00:07:21.12 that nature has never seen before? And to use it
00:07:24.24 to awaken all those antibiotic biosynthesis gene
00:07:28.11 clusters. And we think we can. And in fact, so this
00:07:34.25 is how we can actually do this? Well, first of all
00:07:38.28 we can look at all the genome sequences from all
00:07:42.06 kingdoms of life. Not just microbes, it can be even from humans
00:07:45.25 if it needs to be. We can identify all the secondary
00:07:49.24 metabolite gene clusters. We can even use enzymes that have
00:07:53.25 very special activity. Or even change the enzyme.
00:07:57.16 Redesign the protein specificity, or even the active
00:08:01.11 sites as well. Once we find the enzymes that we require,
00:08:04.25 we can then put them into a gene cluster, and at this point
00:08:09.14 together with promoters and ribosomal binding sites,
00:08:12.00 we want to completely rewrite the DNA.
00:08:15.22 Once we have this gene cluster together, we can put this
00:08:18.29 into a screening host and screen for the novel drug
00:08:22.12 of your choice. And once you've found that novel drug,
00:08:25.18 we then want to put this whole gene cluster into a
00:08:28.10 production host. Because the production host is going
00:08:31.07 to be completely different from a screening host
00:08:33.05 in that the primary metabolism is going to be completely
00:08:35.26 re-engineered so that it's geared to produce that
00:08:39.02 compound of your choice. So, actually, antibiotics
00:08:45.02 are perfect for using synthetic biology because
00:08:47.24 it's naturally modular. Here's an example of how an
00:08:51.11 antibiotic is produced, in fact, erythromycin, it has
00:08:56.14 three huge open reading frames. Within the open
00:09:00.08 reading frames, you have modules. Within the modules,
00:09:03.18 you have these domains. And each of these circles
00:09:07.14 are the domains that have the catalytic activity.
00:09:10.18 So it's very similar to how fatty acids are being
00:09:13.20 synthesized. You start off with C3 unit, it gets loaded onto
00:09:18.06 the loading module, another C3 unit will be loaded onto
00:09:23.08 the module 1, and another C3 to module 2, so on
00:09:27.06 and so forth, to get this long chain of fatty acids.
00:09:31.04 And in the end, it's cleaved off and cyclized to make
00:09:34.13 this core structure. And of course, you have modifying
00:09:38.02 enzymes to make the erythromycin the final compound.
00:09:42.04 But if you look at this, you can see different levels where
00:09:46.19 modularity occurs in the very high level, in the module
00:09:50.11 levels, and in the domain levels. Which means that
00:09:53.01 by changing, swapping these domains and modules around
00:09:56.22 you can get a lot of chemical diversity. So how can we
00:10:01.08 put these synthetic pathways together? You need
00:10:05.02 ribosome binding sites, promoters, you need several
00:10:09.06 of them. So you need libraries. But of course, not just
00:10:12.29 promoters, ribosomal binding sites, but you also
00:10:15.18 need enzymes, the parts, the bits that's going to
00:10:19.01 be the most important to make your pathway.
00:10:22.11 So we tried to see if we could actually
00:10:24.20 do this. Can we make libraries of enzymes where we can swap
00:10:28.01 around? And to do this, we've taken an antibiotic called
00:10:32.25 calcium dependent antibiotic as an example. This is a peptide
00:10:36.19 antibiotic and it uses a special amino acid, which is called
00:10:40.28 L-hydroxyphenylglycine. Here's the biosynthesis pathway
00:10:44.18 here. Because it's a special amino acid, you need all these
00:10:48.11 genes within the biosynthesis cluster to produce
00:10:51.25 this HPG in the end, and then it's incorporated into
00:10:55.19 the final structure. Now we looked at this enzyme here
00:10:59.16 called Hmo or MdlB, we tried to see if we could find
00:11:04.02 homologs or orthologs. And to see if we could swap them around
00:11:07.23 and will it still make CDA? Indeed we found
00:11:11.26 some homologs, here we found three homologs
00:11:15.01 for Hmo, and then we found some orthologs. So these
00:11:19.07 genes, or enzymes, are not involved in actual
00:11:23.07 HPG biosynthesis, but are involved in mandelate catabolism.
00:11:27.10 So, to understand whether they can actually
00:11:30.22 produce CDA, what we did was to delete the Hmo
00:11:35.14 in the natural producer. And then complement it with all
00:11:39.00 of these genes, to see if they could actually produce
00:11:41.29 CDA. Now CDA's an antibiotic, so here you can see
00:11:45.14 we've done some bioactivity tests and they do.
00:11:49.07 You can see some halos here like this, which means it has
00:11:53.01 activity. Well, this isn't good enough. We don't know if
00:11:56.14 it's actually CDA. So what we've done further is to prove
00:12:00.04 that it is CDA by doing HPLC analysis and LC mass spec.
00:12:05.00 So this tells us that yes, we can make enzyme libraries
00:12:09.14 and we can swap enzymes around to start producing
00:12:11.29 antibiotics. Once you have the enzyme parts, you need
00:12:16.11 to put them together. You need to refactor them, you need to build
00:12:19.11 the pathway. But what's the best way of doing that?
00:12:22.17 Is there an order that we should follow? Is one way
00:12:26.28 better another? To understand this, we decided to use this
00:12:31.07 six gene pathway that makes aloesapnarin II.
00:12:34.22 Now, if you -- this is the natural organization of the genes
00:12:39.16 the six genes, you can see they're mostly coupled,
00:12:43.02 transcriptionally and translationally coupled. But can we
00:12:46.06 uncouple them? What's the best way? You can see some
00:12:49.16 examples here. This would be the most simplified
00:12:53.07 way of having one promoter, one ribosomal binding site,
00:12:56.10 for one gene. But is this the best way of transcribing,
00:13:00.22 translating this pathway? Maybe it's not. But if you start
00:13:05.18 to think about all the combinations, it's a bit too much to do.
00:13:08.13 So we went back to nature to see if there are some rules.
00:13:11.18 And indeed, we did find some rules. Here are five different
00:13:15.16 pathways, which are all similar in that they produce this compound
00:13:19.26 here. What we found, in fact, was that there were two genes,
00:13:24.18 these light green two here. They're always transcriptionally
00:13:28.18 and translationally coupled. And we further found out that
00:13:32.17 if we uncouple them, this pathway just does not work.
00:13:36.13 So, we always know that this ketosynthase and this
00:13:40.01 chain length factor always have to be transcriptionally
00:13:42.28 and translationally coupled. So now we can start looking
00:13:46.01 at different combinations, see which organization of
00:13:49.25 the genes, the promoters, and the ribosomal binding
00:13:52.22 sites are the best way to transcribe and translate
00:13:55.29 the pathway. Okay, so just rewriting DNA is not about
00:14:02.11 synthetic biology. We can think about other things, about
00:14:04.26 the cell itself. So here, we're thinking about spatial
00:14:09.07 control of biosynthetic pathways. What do I mean by that?
00:14:12.10 Well, how about thinking about trying to put synthetic
00:14:15.24 protein scaffolds? Or you can make compartments
00:14:19.16 within the cell. And in both of these cases, it's enhancing
00:14:23.06 enzyme activity. We can also think about a bigger spatial
00:14:28.00 control. How about microbial consortia? We can have one
00:14:31.11 cell producing a specific compound for the other cell.
00:14:35.25 For example, you could have lignin, filamentous fungi produce
00:14:41.12 using lignin to produce glucose and E. coli, which
00:14:44.20 produces biofilm using that glucose. And you could grow
00:14:48.11 them together. Synthetic bacterial organelles are something
00:14:52.11 that we're very interested in. Especially using bacterial
00:14:55.17 microcompartments. These are made from proteins,
00:14:58.28 not from lipids and they can be found in E. coli
00:15:02.17 and some gram negative bacteria. What's so good about
00:15:05.12 them? Well, if you think that you don't have these BMCs
00:15:09.06 and you don't encapsulate them into these BMCs,
00:15:11.24 all the pathways you could have degradation of substrates,
00:15:15.08 you could have toxic intermediates which can damage
00:15:19.04 the cell. Once these pathways are embedded into the
00:15:23.06 BMCs, the substrates will no longer have competition
00:15:26.24 it will not be degraded. Even if it produces a toxic
00:15:30.12 intermediate or even end compound, it will no longer
00:15:33.23 affect the cell. So you can produce much more.
00:15:36.01 So we're quite interested in this BMC and pursuing
00:15:39.08 further with this. How about temporal control? Of course
00:15:43.14 this is one of the most major things that a lot of the groups
00:15:46.12 are doing. We can have very fast temporal control, for example,
00:15:51.13 allosteric control. Or just-in-time expression, where
00:15:55.18 the genes are only transcribed exactly when they're needed.
00:15:59.14 One could also think about using signaling molecules
00:16:03.00 to synchronize molecular clocks, but we know
00:16:06.06 how heterogeneous even single cells can be,
00:16:09.12 by using signaling molecules we could make absolutely
00:16:11.28 sure that all the products are being produced at the same
00:16:15.12 time in the cells. One could also use signaling molecules
00:16:19.14 to adjust pathway expression. For example, I
00:16:23.16 talked to you earlier about how antibiotics have
00:16:26.13 these huge open reading frames, those were the core
00:16:29.05 enzymes. And you need the core enzymes first, so
00:16:32.06 you can use the signaling molecules to, for example,
00:16:34.17 express these core enzymes first and then another
00:16:38.12 signaling molecule to express the modifying genes.
00:16:41.17 We are in fact working on these kind of signaling
00:16:45.10 molecules. These are called gamma butyrolactones,
00:16:48.13 which are found in streptomyces and are involved
00:16:51.01 in antibiotic production. And we'd like to use this
00:16:54.20 and produce new regulatory circuits we can use
00:16:59.19 for synthetic biology. Okay, we've talked a lot about
00:17:04.02 putting synthetic pathways together. Of course about
00:17:07.18 building libraries of parts, promoters, ribosome binding
00:17:11.20 sites. But actually, how do we make these libraries? Do we
00:17:15.26 just go out and hunt it by hand? Absolutely not.
00:17:19.06 So what we've done is to design softwares
00:17:22.17 to actually find secondary metabolite gene clusters.
00:17:25.18 It's called antiSMASH and it's already on our version 3.
00:17:29.18 What it does is rapidly detect and annotate secondary
00:17:33.14 metabolite biosynthesis gene clusters. Here's a snapshot
00:17:36.28 of what it would look like. You can put in a whole genome,
00:17:40.02 you can put in parts of the genes that are in the sequence,
00:17:43.13 you'll find the software -- the software will find all the
00:17:47.11 possible secondary metabolite gene clusters. In this case
00:17:50.20 they found 25, and then it will show you the open reading
00:17:54.08 frames and it can even deduce the chemical structures,
00:17:59.29 possible chemical structures. Another software we've
00:18:02.22 developed is called multigene BLASt, and in this case
00:18:06.08 we're not just looking for antibiotic synthesis gene clusters.
00:18:09.13 Anything gene clusters that are in an operon and conserved,
00:18:13.12 we can look for them. You can see here how these operons
00:18:16.29 are all found in different strains. And not everything
00:18:20.05 is conserved, but you can still find them.
00:18:22.28 Another software we've developed is called Pep2Path.
00:18:27.12 This is combining the antiSMASH together with
00:18:33.01 any mass spec data that you have on peptides. So
00:18:37.06 basically what we're doing is, if you have some peptide
00:18:40.05 data, we can actually find the antibiotic biosynthesis
00:18:43.17 cluster that you're looking for. We've also gone off
00:18:47.16 to do some more modeling as well, so in this case
00:18:50.17 we've found some comparative metabolite modeling.
00:18:53.13 This is a constraint-based model, which is just using
00:18:56.19 the genome sequence. And what we've done here is actually
00:19:00.04 looked at all these different actinomyces species
00:19:02.29 and asked which species would be the best to
00:19:07.01 express all these different classes of antibiotics?
00:19:09.28 White colors or the lighter colors mean that it's a good
00:19:14.24 host, the darker colors mean it's a very bad host
00:19:18.23 for expression. Now up here in this box, they are all
00:19:22.08 streptomyces, remember that I said at the beginning
00:19:24.08 that streptomyces are the natural producers for antibiotics.
00:19:28.11 And if you look at the streptomyces species three here,
00:19:31.20 they're not the best host for all of the classes.
00:19:35.02 now if you go further down here to the mycobacterium
00:19:38.23 species, these are natural mycobacterium species.
00:19:41.12 And you can see it's completely white for all throughout.
00:19:45.00 Which means it's a very good host for all
00:19:48.14 kinds of antibiotics. Of course this is in silico
00:19:51.22 analysis, so we need to prove this. And we'd like to
00:19:54.17 do that very much, because this could be a very
00:19:57.06 good chassy for expression. We've also done
00:20:01.02 some more modeling, some on the regulatory networks
00:20:04.22 to show that in fact, these gamma-butyrolactones have
00:20:08.13 regulatory circuits that are a bistable switch
00:20:12.14 for gene antibiotic production. We've also gone off to
00:20:17.06 use the metabolite model to understand how the flux
00:20:21.02 are being used. So in this case, we combine the
00:20:25.17 metabolite model to transcription analysis. We had
00:20:28.11 transcriptome data from a low producer of
00:20:31.24 beta-lactamase inhibitor and a high producer
00:20:34.18 of beta-lactamase inhibitor, and looked to see which pathways
00:20:38.17 were redundant or essential? And in fact, all the green pathways
00:20:43.12 here are basically redundant to produce lots of these
00:20:47.23 beta-lactamase inhibitors. Which means we can now
00:20:51.20 actually minimize the metabolite pathway and also
00:20:56.02 redirect flux so that it produces a lot of these
00:20:59.23 beta-lactones. Now, having done all this. We've made
00:21:05.17 our cell, we've put our pathways in, it's supposed to work
00:21:10.12 perfect. But of course, it doesn't. As with any
00:21:14.08 engineered products, even cars or even computers,
00:21:18.06 sometimes it just doesn't work very well. You need
00:21:20.29 to fine tune it. And we use metabolomics as a
00:21:25.04 debugging routine. And especially the metabolomics that we use
00:21:29.11 is using LCMS method. So here, we're using an untargeted
00:21:36.14 metabolite analysis, in fact, here are two growth curves.
00:21:40.00 One has been induced with antisense RNAs,
00:21:43.11 so it stops growth. While the other is continuing
00:21:46.22 growth. And we took samples from all these time points
00:21:50.06 and compared what kind of metabolites are being
00:21:53.06 accumulated. Now of course, with biology, everything
00:21:58.05 tends to be a little bit difficult to look at. But we did
00:22:02.01 a lot of replicates, as you can see here, we have
00:22:04.18 biological replicates, we have five different time points
00:22:08.13 on the growth curve, we have two LC columns, plus
00:22:11.13 positive and negative ionization modes with
00:22:14.28 three technical replicates for each of them.
00:22:17.17 But by doing all these replicates, we've been able
00:22:20.22 to see a trend. Here you can see two compounds, which
00:22:25.06 are immediately responding to the induction of
00:22:28.22 these antisense RNA. While here, another compound here
00:22:33.02 is not immediately responding, but rather responding to
00:22:36.24 the stop of growth. So, we found lots of compounds
00:22:41.18 actually metabolized, that seems to be going up and down.
00:22:44.23 But remember, what we perturbed is this little
00:22:47.17 blue spot here. It's just this enzyme. So why
00:22:51.23 is it that it's making so much perturbation in other
00:22:55.07 metabolites? We're not really sure why.
00:22:57.18 But one thing we do understand from this
00:23:00.12 metabolomics analysis is, metabolomics is a
00:23:03.18 great tool to look at the cell as a whole. What's
00:23:07.11 happening within the cell. So it's a great debugging
00:23:10.07 routine. So I hope I've been able to show you today
00:23:14.10 that what kind of tools are needed for antibiotic
00:23:18.12 discovery and design. First, of course you need all
00:23:21.25 these different pathways, we could engineer the
00:23:24.29 chassies, we need to have regulatory circuits,
00:23:28.06 we need to also control genes not just transcription
00:23:31.13 and translation, we need lots of computational
00:23:34.20 softwares doing modeling to help us understand
00:23:37.20 where we should do the next experiment. And
00:23:41.25 last, but not least, the debugging is a very, very
00:23:45.04 important tool as well. And in fact, if you look at
00:23:48.29 this, we can put them into the build, design, and test.
00:23:54.05 And going back again to the original slide where I
00:23:56.29 showed the synthetic biology, using synthetic biology
00:24:00.26 this concept, the three concepts. And in fact,
00:24:04.07 these tools are not just for antibiotics.
00:24:06.29 We can use them for any high-value chemicals.
00:24:10.24 All these tools, using design, build, and test concept.
00:24:15.23 Thank you very much.

This Talk

Speaker: Eriko Takano

Audience:

Researcher

Recorded: June 2015

Talk Overview

Antibiotic resistance is a growing problem worldwide. To address this problem, Dr. Eriko Takano and her colleagues are generating methods for the development of new antibiotics using a synthetic biology approach. By performing genome analysis on many microbes, they can identify genes encoding novel biosynthesis pathways that may produce antibiotics. These gene clusters can be transferred to pre-engineered bacterial hosts to optimize the development of new antibiotics. Takano’s lab has developed software systems to search for gene clusters, as well as to model, analyze, optimize and debug antibiotic production.

Speaker Bio

Eriko Takano

Eriko Takano is a Professor at the University of Manchester, where she is Co-Director of the Manchester Synthetic Biology Research Centre SYNBIOCHEM. Takano studied pharmacy at Kitasato University in Tokyo before moving to the UK where she receiving her PhD in the School of Biological Sciences at the University of East Anglia and the John Innes… Continue Reading