What makes the enterobacterial plant pathogen Pectobacterium atrosepticum different to its animal pathogenic relatives?

Editor's Notes

• #2: Good afternoon. I’m Leighton Pritchard. I’m a bioinformatician in the plant pathology programme at SCRI in Scotland, and I’m going to talk to you about some of the comparative genomics work we’ve been doing on the plant pathogenic bacterium Pectobacterium atrosepticum.
• #3: Pectobacterium atrosepticum (Pba) is an enterobacterial plant pathogen that causes blackleg and soft rot diseases on potato in temperate climates. The major rotting symptoms of disease caused by Pba are due to plant cell wall degrading enzymes, such as pectinases and pectate lyases. Evertheless, Pba also possesses a type III secretion system and associated effectors (about which much has been said this week) and synthesises phytotoxins. This implies a more stealthy role for Pba in terms of its interaction with the plant host. It doesn’t just exist to blast holes in plant cell walls – there is the potential for subtler interaction
• #4: Pba is also, unlike many of its enterobacterial relatives that invade gut epithelial cells, plant- and not animal-associated. It is thus reasonable to anticipate substantial differences between the genomes of the animal- and plant-associated members of the enterobacteriaceae. It is also one of a number of plant pathogenic, soft-rotting enterobacteria, such as Pectobacterium carotovorum and Dickeya species. These bacteria vary in their host ranges and climate-dependent disease profile. Pba has a relatively narrow host range for disease, and causes symptoms only in temperate climates. The genes differentiating host range, geographical distribution and disease development remain largely unknown, however. In order to identify candidate genes associated with these differences, we have employed Pba-centric genomic and transcriptomic comparative analyses, and I will report some of our results in this presentation.
• #5: Our starting point for genomic comparisons is the sequenced Pba genome. Pba is a fairly typical-looking enterobacterial genome in terms of size, gene count and GC content. One unusual feature however is the large number of putative horizontally-acquired islands, marked with red and green blocks on this circular diagram. These were mostly identified on the basis of the presence of insertion sequences, phage, and aberrant GC content. It has been noted (Kunin et al, 2006) that Pba appears to be a ‘hub’ of bacterial horizontal gene transfer. Genome Overview IMAGE: Circular diagram of genome LOGIC: O Lab-amenable strain 1043 sequenced by Sanger/SCRI collaboration in 2004 O Typical size for free-living enterobacterium, 5Mb with 4472 CDS, and 51% G+C content O There appears to be an unusually large degree of horizontal gene transfer, with 17 putative horizontally-acquired islands
• #6: In order to find out what makes Pba different from other bacteria, we carried out reciprocal best hit analysis using FASTA, taking reciprocal protein matches with 30% sequence identity over at least 80% of the longer sequence. We plotted each of these hits on a circular diagram of the Pba genome. Each inner ring represents a comparison with a different bacterial genome, and each coloured block indicates a reciprocal best hit. Here, we’ve coloured organisms by taxonomic class, and grouped them together on the diagram. Two features stand out immediately in these diagrams. Firstly, radial bands of colour where Pba shares a block of sequences with many other genomes. And secondly, radial ‘gaps’ with no colour, where sequences are unique to Pba, or common with only a few other organisms. Introduction to comparative approach – RBH against other genomes II IMAGE: Circular diagram of genome comparisons LOGIC: o We can wrap the image round to represent a circular bacterial chromosome o The colours mean the same as in the previous diagram, but it becomes clearer where regions of overall similarity and dissimilarity lie
• #7: We have since extended these comparisons to all available genomes, to generate images such as this. The colour scheme here is that individual reciprocal best hits are marked in a colour indicating the percentage sequence identity to the Pba sequence – red for near 100%, blue for near 30%. We’ve also ordered the comparison rings so that the most similar genomes to Pba (on average) are outermost. Even with 400 comparison sequences, we can still see many large radial gaps, indicating sequences that Pba does not share with many other organisms. We can zoom in on a region to get a closer look… Description of more recent comparisons against many genomes, and introduction to interpreting the images with colours, etc… particularly gaps IMAGE: Circular diagram of comparison against 229 bacteria LOGIC: o We need to familiarise the audience with how to interpret the image for the next few slides o This diagram represents the Pba chromosome, and the reciprocal best hits present in 229 other bacterial genomes, each one represented by a concentric circle o Coloured blocks represent individual RBHs, coloured on a scale from blue to red by % identity (blue = low identity, red = high identity) o The concentric circles for organisms are ordered from most to least similar to Pba from the outer edge (red ones are enterobacteria) o Immediately, we can see regions of radial gaps in the genome, associated with the putative HAIs, and that these gaps get filled in the closer you get to the outer edge o These gaps are informative in terms of Pba’s evolutionary history and its adaptation to a novel functional niche
• #8: In this region we can see the blocks, and their individual colours more clearly. Focusing on this radial gap here, we can see that on either side of the gap the trend is that the genomes that are more similar to Pba also have the most similar sequences. Within the gap however, we can see sequences that are highly similar to Pba, but that lie within organisms whose genomes are, on the whole, not very similar to Pba. We reason that these sequences have therefore potentially been acquired by horizontal gene transfer or, as in the case here, are potentially indicating gene loss. Description of more recent comparisons against many genomes, and introduction to interpreting the images with colours, etc… particularly gaps IMAGE: Circular diagram of comparison against 229 bacteria - zoomed LOGIC: o We need to familiarise the audience with how to interpret the image for the next few slides o This diagram represents the Pba chromosome, and the reciprocal best hits present in 229 other bacterial genomes, each one represented by a concentric circle o Coloured blocks represent individual RBHs, coloured on a scale from blue to red by % identity (blue = low identity, red = high identity) o The concentric circles for organisms are ordered from most to least similar to Pba from the outer edge (red ones are enterobacteria) o Immediately, we can see regions of radial gaps in the genome, associated with the putative HAIs, and that these gaps get filled in the closer you get to the outer edge o These gaps are informative in terms of Pba’s evolutionary history and its adaptation to a novel functional niche
• #9: When considering the origin of HGT or gene loss, we consider the environments in which the organism must survive. Here we have a typical agricultural environment involving animals and plants. I’m told that this is a cow, but I’m no biologist. We can imagine an organism that divides its time between the animal and the surrounding environment. The organism enters the organism at the sharp end, here… eventually leaves through the blunt end, here… and hangs around near plants for a while, before going back into the animal at the sharp end and repeating the cycle. There are many possible trajectories for the genome, depending on circumstances. If the animal-plant cycle persists, we expect the organism to carry functions that benefit it in both environments – as we might for E. coli 0157:H7, for example. If we remove one or other of the environments, let’s say we take the cow out of the picture, then those functions that were previously relevant to survival in the cow may be lost. Also, we might expect that the organism might somehow acquire functions specifically to survive better in the environment. A quick way to do so would be to acquire them directly from organisms that are already successful in that environment, for example by HGT. If we want to find out what has made Pba a successful plant pathogen when many of its relatives are successful animal-associated bacteria that occasionally enter the environment, we can ask: what functions does Pba have that plant-associated bacteria also have, but that animal-associated bacteria do not? (and if we’re interested in how animal-associated bacteria persist in the environment, we can ask what they still share) Interpretation of gaps in terms of HGT and niche adaptation – gene acquisition and loss IMAGE: A cow, some grass, and the cycle of the bacterium as it passes from one environment (enteric) to another (the wider environment) LOGIC: o Each bacterial population is under selection pressure to thrive in all the environments to which it is exposed o In order to thrive better in the gut, we expect to see adaptations to the gut o In order to thrive better in the environment, we expect to see adaptations to the environment o In the transition from one environment to the other, we expect to see relative loss of functions advantageous in the ‘old’ environment, and relative acquisition of functions advantageous in the ‘new’ environment o If we assume that Pba is an enteric bacterium that has adapted to the wider environment, then we expect to find that it has acquired novel functionality relative to other, enteric-based bacteria o We might expect the gaps in the Pba diagram to correspond to some of the features that Pba has acquired to help it live outside the animal gut. o Moreover, the organisms with similar genes to those in the gaps may be the originators of that functionality for Pba, elucidating Pba’s evolutionary history
• #10: Here we have another circular diagram showing reciprocal best hits. We’ve clustered the comparison genomes into two groups: plant-associated bacteria and animal-associated enterobacteria. Note that the AAEs are much more similar, on the whole, to Pba than are the PABs We’ve also marked, in red around the outside, the locations of those Pba genes that are more similar to PAB than to AAE. These account for over 10% of the Pba genome, indicating that the scale of acquisition of function from plant associated bacteria may have been very great. Interpretation of gene acquisition and loss as niche adaptation in the context of adaptation of Pba to plant environment. IMAGE: Circular diagram of Pba against AAE and PAB, with genes more similar to PAB than AAE sequences marked LOGIC: o Acquisition of environment-associated function may be by adaptation of existing functionality to the new environment, or by HGT – direct acquisition from other plant-associated bacteria. o If adaptation is by HGT, then we expect to see genes/functions that are found in plant-associated bacteria, but not in animal-associated enterobacteria. O Outer set of rings are AAE, inner set of rings are PAB – overall, Pba is most similar to AAE (seen from red colour) o Nevertheless, we find XX genes that are more similar to PAB than to AAE – indicated by the red markers on the diagram, and they are frequently associated with islands previously suspected to be horizontally-transferred O This implies a role for HGT and the acquisition of genes from PABs
• #11: Returning to the Pba comparisons, we can identify clusters of these genes that appear to be PAB-associated, and note that they are colocated with regions of putative HGT. We find that, in many cases, these genes encode for functions which we would reasonably expect to be useful in plant interactions: Root attachment Coronafacic acid synthesis T3SS and effectors And Nitrogen fixation In silico predictions are all fine, but in this modern age of systems biology, we need to close the loop between prediction and experiment. Locations of some of the genes associated with niche adaptation on the comparative diagram IMAGE: Circular diagram of Pba against AAE and PAB, with genes more similar to PAB than AAE sequences marked. Labels for environmental function fade in LOGIC: o We have established that genes associated with PABs are located in sites we previously saw were potentially associated with HGT, but this might be coincidence. Are there functionalities that seem biologically to be associated with this new environment? o We can identify a number of gene clusters, apparently transferred from PAB, with functionalities that make sense as adaptations in biological terms: e.g. attachment, coronafacic acid and T3SS (P. syringae people should be interested), and nitrogen fixation o But can we trust these annotations? How well does Pba persist in the environment, or manage at causing disease if we knock out some of these genes?
• #12: We consider two genes in a region of the Pba sequence that is similar to PAB, but not AAE, and is also part of a putative HAI One of those genes is a homologue of P. putida AggA, which contributes to root adhesion in that organism This is a plot of the recovery of Pba – WT, and ECA3266 and aggA knockouts from a range of plant roots. For each knockout, and each plant, we recover significantly less Pba from plant roots than the wild type, supporting the functionality of root adhesion. Notably, adhesion occurs on more plants than just potato – the only one on which it causes disease – implying a wider reservoir, and maybe even a more benign lifestyle for Pba than was previously suspected. Here, then we have evidence supporting the acquisition of functionality relevant to a plant-associated lifestyle for Pba. This work was carried out mostly by Sonia Humphris, and was presented at a poster that has now been taken down, but you should be able to still catch her about if you’re interested. Adherence to roots I IMAGE: Plot of colony-forming units per millilitre against plant (potato) LOGIC: O The aggA gene is one of those in an HGT region, and found in PAB but not AAE, and contributes to root attachment in Pseudomonas putida. Pba aggA knockouts are expected to show less root adhesion than wild-type. O aggA- knockouts show no significant reduction in root adhesion on potato aggA complemented
• #13: Another region on the genome with genes similar to PAB but not AAE, and that is found in HAI is a region containing coronafacic acid synthesis genes. This work was mostly done by Michael Ravensdale, whose poster is still up outside the Auditorium Sirene and, again, if you’d like to hear more about it, just catch him at some point. Coronafacic acid is a precursor of coronatine, which is a Pseudomonas phytotoxin. cfa synthesis gene knockouts are attenuated in virulence, as seen by the reduction in lesion length. So here we also have evidence supporting the horizontal acquisition of virulence function, as a feature of Pba’s plant-associated lifestyle Coronafacic acid IMAGE: Plants with WT and cfa- Pba LOGIC: o Knocking out cfa genes attenuates Pba virulence on potato Several cfa knockouts – picture representative; needs ligase
• #14: We haven’t just made in silico comparisons. We have designed an 11k Pba microarray (and a similar Dda array), and have challenged it with both Dda and Pcc genomic DNA, in order to investigate what makes Pba different from other soft-rotting plant pathogenic enterobacteria. The genome sequence for Dda is also available for download from Nicole Perna and Jeremy Glasner’s group’s ASAP database, in Wisconsin, which gave us the opportunity to calibrate our transcriptomic and genomic comparisons using the complementary analyses for Dda and Pba. Hui Liu has a poster up describing some work using these arrays that you might be interested in seeing. Comparison with Pcc and Dda I IMAGE: Pcc bugs, Dda bugs, microarray slide LOGIC: O We can see broad trends in niche adaptation from comparisons against large numbers of comparator genomes, but differences in infection profile also exist between Pba and its close relatives Dda and Pcc O We have constructed an 11k Pba microarray, and challenged it with Dda and Pcc genomic DNA O The Dda sequence hasn’t been published, but is available at ASAP, and reciprocal best hit analysis of the Pba genome against this sequence provides a complementary analysis, and a check on the hybridisation results
• #15: When we do whole-genome Southern hybridisations, we can see that in general Pba and Pcc are much more similar to each other than either is to Dda. Indeed, an in silico analysis identifies only 20 probes from the Pba genome that make exact matches to the Dda genome. However, over 900 Dda probes hybridise strongly to the array, and 3200 hybridise in total. Some Dda sequences with over 90% aa identity to their Pba counterparts hybridise very weakly This illustrates a general point that gDNA hybridisations are not very robust. At best they indicate only a set of divergent and potentially absent sequences in the comparator genome, so we proceed with the Dda reciprocal bet hit comparison, rather than the array hybridisations. Comparison with Pba and Dda II IMAGE: Pba-Pcc-Dda Southerns LOGIC: O Pba is phylogenetically more closely related to Pbc than to Dda, so we would expect that Pbc would hybridise better to the microarray than Dda does, and this is borne out by the Southern blots, and by the hybridisation data. O Indeed, it proved impossible to carry out our original aim to construct a microarray that featured common probes to Pba and Dda genes, as the genomes are so dissimilar. An in silico analysis shows exact matches of Dda genomic DNA to only 20 probes on the Pba array. Common bands = rRNA/16S? Not entirely sure.
• #16: This is the Eye of Sauron. The diagram is similar to those you saw before, with PAB and AAE reciprocal best hit comparisons marked. We’ve also added rings indicating those Pba sequences whose probes don’t hybridise to Pcc gDNA, and that don’t make reciprocal best hits to Dda. This is about 25% of the genome in each case. We’ve also marked 30 islands that we’ve been interested in that appear not to have counterparts in either Pcc or Dda. We’re going to focus in on this region first… Comparison with Pba and Dda IV IMAGE: Pba-Dda-Pcc circular diagram LOGIC: O When we examine the comparisons between Pba, Dda and Pcc, we identify over 30 clusters on the genome of functional significance where genes from Pba appear to have no counterparts in one, other, or either of Dda or Pba – indicated by the outermost red bars. O The previously-identified HAIs are mostly associated with these Pba-associated regions O One of these regions is the SPI-7 homologous region, containing the coronafacic acid synthesis genes. We have already shown this to be involved in virulence, and know that the genes have no homologues in Dda. It remains to show that the failure of Pcc to hybridise to the probes for these genes on the Pba array implies that the genes are absent, or highly divergent.
• #17: This is the region of Pba (on the top) containing the cfa synthesis genes we saw earlier. The bottom row is the corresponding region of Pcc. The blue bars indicate reciprocal best hits, and we can see that the two regions are mostly syntenous. However, the region containing the cfa synthesis genes is missing entirely in Pcc, though there has been some interesting expansion with other genes in that region. In this case at least, then, the absence of Pcc hybridising gDNA does appear to indicate that the region is absent in that organism. SPI-7 comparison between Pba and Pcc IMAGE: Pba vs Pcc SPI-7 regions, in ACT LOGIC: O We have an annotated Pcc BAC sequence spanning the SPI-7 region from a collaboration with the Sanger Institute O Pcc regions = resistance to ox. Stress? Being investigated
• #18: Now we consider a second region, making no hybridisation to Pcc, but having RBH matches to Dda, and to other PAB. This is found in an HAI. It encodes genes that are homologous to those for nitrogen fixation in other organisms. This analysis in particular enabled us to make some predictions: WT Pba fixes N WT Dda fixes N WT Pcc doesn’t fix N If we knock out the nif genes in that region, the mutant Pba will not fix N Comparison with Pba and Dda V IMAGE: Pba-Dda-Pcc nif region LOGIC: O This diagram indicates a number of features: That there is a region of putative horizontal gene transfer from PAB to Pba This region does not have counterpart genes in AAE That part of this region is Pba-specific, in that genes are shared with Dda, but not Pcc O The genes in yellow are the key genes – they are nif genes, contributing to nitrogen fixation, including nifA. They are present in Pba and Dda, but not Pcc, which allows us to make some predictions: WT Pba and Dda both fix nitrogen WT Pcc does not fix nitrogen Pba nifA knockouts should not fix nitrogen
• #19: And these are the results. On the right is an Azobacter control that really does fix N On the left we see three WT strains of Pba that fix N, and the nifA mutant that does not On the right we see a Dda strain, and another Dickeya strain, that both fix N In the middle we see five Pcc strains, including the one we used for our comparisons, Pcc193, that do not fix N, and one that does. This confirms our predictions made by comparative genomics and transcriptomics. Comparison with Pba and Dda VI IMAGE: Nitrogen fixation graph LOGIC: O This graph shows the nitrogen fixing ability of a number of Pba, Pcc and Dda strains, with an Azobacter control - Three WT Pbas fix nitrogen - Two WT Ddas fix nitrogen - Only one of six WT Pccs fixes nitrogen O This validates our predictions from comparative genomics Looking to see if NF occurs on roots/in environment as well as in vitro Did we complement nifs? No.
• #20: In summary then, we asked what differentiated Pba from its close and distant relatives, and have found that the answers are, at least in part: HGT activity The apparent acquisition of verified plant-associated lifestyle functions: coronafacic acid synthesis, root adhesion, and N fixation. The differences between Pba and closely-related genomes comprise about 25% of the Pba sequence, so there is still plenty left to investigate. And we now have some new questions to ask: Doing the reciprocal comparison – what has Pba lost in respect to AAE? What does it not have that its closer relatives have? And, particularly interesting for food safety, what does Pba have in common with animal/human pathogens that might permit them to persist in the wider environment? Conclusions IMAGE: None LOGIC: O In silico comparative genomics against large numbers of bacteria reveals information about overall niche adaptation O Comparative genomics, both in silico and experimental, against close relatives reveals information about species/strain-specific differences O Pba, in adapting to an environmental niche, has acquired nitrogen fixation and root adhesion capability, and a number of virulence-related functionalities associated with host defence manipulation O Some of these functionalities are absent in other soft-rotting plant pathogens O This graph shows the nitrogen fixing ability of a number of Pba, Pcc and Dda strains, with an Azobacter control - Three WT Pbas fix nitrogen - Two WT Ddas fix nitrogen - Only one of six WT Pccs fixes nitrogen O This validates our predictions from comparative genomics
• #21: Acknowledgements
• #22: Introduction to comparative approach – RBH against other genomes I IMAGE: Linear diagram of genome comparisons LOGIC: o Particularly given the number of putative HGT events, we thought it interesting to identify, where possible, the most closely-related genes to Pba genes in other organisms o We identified reciprocal best hits using FASTA, and plotted them on a diagram: blocks indicate where Pba genes are present in other organisms each row represents a different organism. blocks are coloured by the phylogenetic class of the organism in which the match is found
• #23: Problems of visualisation IMAGE: GenomeDiagram flowchart LOGIC: o The data generated for each comparison of Pba against multiple organisms is considerable: 229 bacterial comparisons 185970 RBH 23Gb of data 24h on a 50-node cluster o Visualisation of this much data was a problem, so we developed the GenomeDiagram software to integrate with BioPython and produce publication-quality, poster-sized images (some of those at ICPPB or IEW last year will have seen some of the output)
• #24: Comparison with Pba and Dda III IMAGE: Pba-Dda hyb/RBH plots LOGIC: o The hybridisation data on the last slide only identifies gDNA that is not highly-divergent from Pba, so a failure of Dda gDNA to hybridise does not imply that the Pba gene is absent in Dda, for example. o Dda is still functionally-similar to Pba, and makes 2910 reciprocal best hits. o We don’t have the Pcc genome sequence, so can’t say how many false negatives come from the microarray data o As we already have hybridisation data, we can explore the relationship between hybridisation intensity and RBH identity o As you might expect, the relationship is not entirely clear, but there is a general tendency for stronger hybridisations to correspond to better RBHs, though this tendency is not absolute. o Nevertheless, very few probes with raw hyb scores of over 2000, or normalised scores of greater than 1 fail to make RBHs. This allows us to estimate the number of Pcc RBHs to be…. XX
• #25: We can ask if this kind of pattern is found in other bacterial comparisons, and we find that sometimes it is, and sometimes it isn’t. The human uropathogenic E. coli CFT073 has many radial gaps, perhaps not so many as Pba, but it certainly looks similar. However, the bovine pathogen Pasteurella multocida does not. These patterns of gaps in the comparison data are not present in all comparisons, but are present in some other comparisons IMAGES: The whole-genome comparisons of uropathogenic E.coli and P. multocida. LOGIC: o The uropathogenic E. coli image shows that the pattern of gaps, potentially indicating large-scale HGT, is not unique to Pba. o The P. multocida image, which has fewer gaps, indicates that putative HGT, visible in this way, is not the norm for all genome comparisons o This implies that we can pick up differences between the evolutionary histories of organisms in this way
• #26: Again, we can ask whether this is usually the case. This is a similar comparison of Salmonella Typhi CT18 – a chromosome and two plasmids - against PAB and AAE. Again we’ve marked the locations of genes that are more similar to PABs than to AAEs, and the number isn’t nearly as great as for Pba. However, there are hotspots in one of the plasmids. This is a story with implications for the acquisition of niche-adaptive function, but I don’t have time to go into detail about it, so we’ll just note it and move on. These kinds of gaps and HGT also vary between plasmid and chromosome IMAGE: Salmonella image from the Annual Reviews paper LOGIC: o HGT is not confined to the genome, and may have hotspots on plasmids o note that Pba SCRI1043 doesn’t carry a plasmid