Some
of the best evidence for environmentally induced epigenetic inheritance comes
from studies of pathogen infection in A.
thaliana. When infected by the common laboratory strain of the bacterial pathogen Pseudomonas syringae (DC3000), A. thaliana plants undergo extensive DNA
methylation changes that regulate defense gene expression. Furthermore, some of
these induced methylation changes can be transmitted to offspring, trans-generationally
‘priming’ offspring for more effective defense responses when they encounter
similar pathogens.
However, plants in nature are typically subject to
simultaneous infection by pathogens that induce different defense responses.
The defense systems activated by different pathogens may even antagonize each
other via hormonal crosstalk. The effects of such co-infection on DNA
methylation patterns and trans-generational defense priming remain entirely
unexplored, as does the extent of host genetic variation for these epigenetic
responses.
To address these issues, we generated A. thaliana lineages with different
histories of bacterial infection across generations. This framework enables
several key determinations, including the specific DNA methylation changes that
are induced in parents by single- versus co-infection, which of these changes
are inherited by offspring, and how inherited methylation changes influence
offspring defense responses when offspring are infected. To date, we have
characterized the genome-wide DNA methylomes of the founding (parental) plants
of these lineages, which were infected by the natural bacterial pathogens Pseudomonas syringae (Michigan strain
NP29.1A) and P. viridiflava (Michigan
strain RMX3.1B),separately and in
combination (i.e., co-infection).
Here we examined the population genetics of the three-copy
R-gene family of Rpp8. Across Rpp8, nucleotide diversity ranged from double to
27x the genomic background rate and amino acid substitutions were 5 to 16x
higher due to intergenic gene conversion (IGC) between the 3 paralogs. Simulation models suggest IGC coupled with
balancing selection to maintain copy number polymorphism drives the high level
of diversity we observe in Rpp8. If we
consider paralogs undergoing IGC as analogous to a single gene, then IGC
between paralogs could effectively create a heterozygous locus in a
predominantly homozygous individual thus establishing reservoirs of variation
for the generation of new R-gene recognition specificities via some sort of
recombination event.
Bootstrap consensus trees for the maximum parsimony phylogenies of the leucine-rich repeat region (LRR) and non-LRR regions of all three Rpp8 paralogs. Clades comprised of alleles from one paralog are boxed. Green, blue, and orange boxes represent P1, P2, and P3, respectively.
a) Phylogeny of non-LRR region (239 parsimony-informative sites out of 1701 sites). b) Phylogeny of the framed LRR region for the same accessions as in (a). There were 236 parsimony-informative sites out of 1019 in this phylogeny
Host-pathogen interactions are often viewed as a conflict between two organisms:
the pathogen thrives if it can exploit its host, and the host thrives if it can
fend off infection. This restricted view, however, overlooks the fact that
organisms in nature are exposed to many pathogens simultaneously, which can lead
to co-infection by multiple pathogen strains or species within each host
[1].
Interactions between co-infecting pathogens can profoundly alter the ability of individual pathogens to invade and exploit a host, with consequences for disease progression and host fitness. For example, we have previously found that pathogen strains lacking key virulence genes proliferate better during mixed infections with more aggressive strains [3]. In general, interactions among pathogens range from antagonistic to beneficial: microbes not only compete for resources but can also mount synergistic strategies to subvert host defenses and better exploit host resources [1,2]. How pathogens maximize their performance during co-infections, and the genes underlying these strategies, is largely unknown.
We are studying co-infections using a genetic model plant species, Arabidopsis thaliana, and one of its
most abundant bacterial pathogens, Pseudomonas
viridiflava. Recent ecological genomics work has revealed that co-infection
is remarkably common in this system: most individual Arabidopsis plants in natural settings harbor multiple virulent P. viridflava strains within their
leaves [4]. Our work uses high throughput assays to answer the following
questions in this system:
Figure 1. Experimental system for plant growth and infections. Above, gnotobiotic Arabidopsis in plant growth (MS) media in 24-well microplates. Below, from left: a lightly, moderately, and heavily diseased plant, 36 hours after infection with P. viridiflava strains differing in pathogenicity
Do pathogen strains differ in performance depending on which other strains are present during co-infections? Our earliest results suggest this is overwhelmingly the case (Figure 1). We are therefore pursuing the genetic basis of these microbe-microbe interactions.
What genes, and combinations of alleles, determine co-infection outcomes? This knowledge will help us understand pathogen ecology and evolution. But it also has practical applications: it could improve our ability to model, predict, and prevent disease outbreaks in natural populations.
One of the main advantages of the Arabidopsis-Pseudomonas
study system is its amenity to large-scale experiments (Figure 2). Hundreds
of plants can quickly be inoculated using pipetting robot, infection outcomes (pathogen
abundance) can be monitored by high-throughput DNA sequencing, and new
statistical approaches in two-organism genome-wide association mapping [5] are
poised to identify combinations of alleles in the pathogens’ genomes with
beneficial or deleterious effects on their performance. Although this project
focuses on a single plant and pathogen species, we hope to generate conceptual
and statistical insights relevant to ecological, evolutionary, and biomedical
models of host-enemy interactions.
Figure 2. Pairwise co-infections strongly shape pathogen performance in gnotobiotic Arabidopsis. The abundance of two luciferase-tagged strains of P. viridiflava (strain p13.G4, “b”; strain p25.A12, “c”) were measured 36 hours after co-inoculation with each of 60 different strains from the P. syringae complex, whose phylogenetic relationship is shown in “a”. Abundances of the two focal strains in each pairwise co-infection are expressed relative to their abundance in single infections. Effects on focal strain abundance are expressed as log10 units of photon counts/second.
References
[1] Tollenaere, Charlotte, et al.
“Evolutionary and epidemiological implications of multiple infection in
plants.” Trends in Plant Science (2016).
[2] Abdullah, Araz S., et al.
“Host–multi-pathogen warfare: pathogen interactions in co-infected
plants.” Frontiers in Plant Science (2017).
[3] Barrett, Luke G., et al. “Cheating, trade‐offs and the
evolution of aggressiveness in a natural pathogen population.” Ecology
Letters (2011).
[4] Karasov, Talia L., et al. “Arabidopsis thaliana and Pseudomonas pathogens exhibit stable
associations over evolutionary timescales.” Cell Host & Microbe
(2018).
[5] Wang, Miaoyan, et al. “Two-way
mixed-effects methods for joint association analysis using both host and
pathogen genomes.” Proceedings of the National Academy of Sciences
(2018).
As microbial ecology has advanced
in recent decades, the importance and incredible diversity of microbial
communities has become apparent. However, the processes that determine the
composition of microbial communities remain poorly understood. Determining what
gives rise to a certain community composition may help us manipulate microbial
communities into healthier or more productive forms.
To gather candidates for our
synthetic microbial community studies, we are coordinating two A. thaliana microbial collections: one
from Sweden and one from the Midwestern United States. We are attempting to collect as many microbes
from our samples as possible, creating a permanent “living library” for future
research. We are collecting microbes primarily from internal leaf tissue.
However, collections from the Midwest also include microbes from roots and
siliques.
We are currently processing over
5000 new bacterial and hundreds of fungal isolates (we already hold >6,000
Midwestern bacterial and 50 fungal isolates).
We seek taxa that match hub OTUs that have not been previously cultured
in order test them in controlled growth chamber experiments with sterile plants
and ultimately combine them with other OTUs to form synthetic communities in
which the network of interactions among microbes has been empirically verified.
Such a community will be used to assess
the accuracy of various interaction inference approaches. This evaluation of
our ability to identify microbial interactions is fundamental for our continued
application of network science to microbial communities.
Future
work will expand the application of this experimental community to address
questions and hypotheses from network science and ecology. This may include
topics such as: the importance of competitive interactions in community
stability, and the effect of higher order interactions on community dynamics
and composition.
Seven Arabidopsis Midwestern
accessions in HPG1 were grown in two locations, Warren Woods and the Michigan
Research and Extension Center, for two successive years and sampled monthly
during the growing seasons over the span of two years. The aim was to collect samples for bacterial
microbiome analysis using 16S rRNA from all developmental stages of the plants
to understand how the microbiome changes in space and time.
Figure 1. PCoA showing separation of bacteria from soil, roots, and rosettes (colors) and location (shapes).
We find that the phyllosphere and rhizosphere communities have distinct compositions compared to each other and to the surrounding soil (Figure 1 above). Figure 2 (below) shows the networks constructed for each developmental stage in the roots at two different sites. The taxa richness, and thus the number of members in the network, increased as plant development progressed. An increase in community diversity at later stages can be seen as the number of different types of bacteria represented increases.
Figure 2. Bacterial networks sampled from A. thaliana roots by developmental stage. WW vegetative not sampled.
Bacterial networks also show more modularity in their
structure as plant development progresses. Relative to random networks of the
same size, networks from later developmental stages in both tissues were more
modular than the networks from earlier developmental stages. There is more
analysis that can be done on the modules present in the plant and soil networks
to determine what variables in the data (microbe relatedness, site, or year)
can best explain the patterns in community structure.
Previous studies on plant microbial networks identified sets
of fungal or bacterial taxa as “hubs” because they were exceptionally well
connected in inferred interaction networks. It is posited that this small set
of microbes has outsized influence on phyllosphere and rhizosphere communities
through interactions. However, in this dataset we find that the bacteria
identified as hubs based on their connections in the network varied across
plant development in both the phyllosphere and rhizosphere. This suggests the
influence of a hub microbe may not be predictable across different tissues and
developmental stages in plants.
We are investigating the importance of the microbiome and the holobiont in evolution. To test this, we are experimentally evolving the model plant, Arabidopsis thaliana in conjunction with a synthetic microbial community. Plant genetic diversity is supplied with a set of A. thaliana recombinant inbred lines. The synthetic microbial community is composed of bacteria, fungi, and other eukaryotes. These microbes were isolated from the tissues and rhizosphere of A. thaliana growing in the field.
Our current understanding of how polymorphism is maintained relies on models of obligate pairwise species interactions but at least half of all plant pathogens have multiple hosts. This raises the possibility that pathogens drive convergent evolution in co-occurring plants. We propose to test this hypothesis by studying co-occurring Brassicaceae plant species, and how shared plant pathogens can potentially maintain ancient balanced polymorphism of resistance genes in plants. Arabidopsis thaliana has long been the plant model for genetics. We will focus on a set of approximately 180 natural plant populations in the southern France (Midi-Pyrenees) that contain A. thaliana, as well as two closely related weedy Brassicaceae; Cardamine hirsuta and Erophila verna. Our major aim is to unravel the genetic architecture and evolutionary dynamics behind all the R genes shared among these three species. We do this by sequencing all R genes in natural populations of co-occurring C. hirsuta, A. thaliana and E. verna plants. This project has 4 specific aims.
1. Reconstruction of R gene evolution. We will isolate DNA and perform R-gene enrichment sequencing (RENSeq) on 60 natural populations of co-occurring Arabidopsis thaliana, Erophila verna and Cardamine hirsuta, collected in Southern France. After orthologous genes have been detected among the three species, several statistical approaches can be applied to study the evolution of those R genes. We propose to explain evolutionary dynamics observed in these R genes through functional characterization from an ecological perspective (component 2), physiological costs of maintaining resistance in the absence of disease (component 3) and genomic and functional constraints of these R genes (component 4).
2. Functional Ecology of homologs. Shared pathogens are likely driving some of the R gene evolution dynamics we have observed in the past in A. thaliana (Karasov, Kniskern, et al., 2014a; Karasov, Horton, et al., 2014). Pathogen effectors will be isolates in the co-occurring pathobiomes with effector enrichment sequencing (PATHSeq) and metagenomics. We propose a two-tiered approach to test how shared pathogens and their effectors might shape R gene evolution. Transient expression on R- avr protein pairs of a small set of divergently evolving R genes will be performed to study what effectors interact with different homologs, and ancestral protein reconstruction on these R genes will be performed to understand how neofunctionalization of homologs could lead to potential differences in Avr recognition.
3. Physiological burden of resistance. Fitness trade-offs exist for carrying the resistance allele in absence of disease, as demonstrated for RPM1 (Stahl, Dwyer, Mauricio, Kreitman, & Bergelson, 1999) and RPS 5 (Karasov, Kniskern, et al., 2014a) although we also demonstrated an exemption with RPS2 (MacQueen, Sun, & Bergelson, 2016). We will test the costs of carrying functional R alleles (candidates of interest derived from first two components of this grant) by creating isogenic lines in all three species, and test fitness effects of disease resistance in climate cell and common garden experiments in the field.
4. Genomic and functional constraints. Does any given R gene recognize a set of effectors, or only one? Do they function as direct recognition proteins, or are they guarding other plant receptors to initiate a defence response? Are they located in tandem repeats, and where in the genome? Both genomic location and genetic architecture can influence the evolution of a gene. Single copy homologs have less freedom to diversify, and genes in recombination hotspots are more likely to undergo rapid evolution. In this component, we are taking a closer look at the genomic and functional constraints of R gene homologs through bio-informatics approaches and functional characterization of protein/protein interactions to establish roles of different R genes. We will perform a phenotype free joint GWAS on geographically linked plant and P. syringae pairs to gain insights in the genomic scale at which this diffuse evolution across different species occurs.
infected crop cultivars from the ongoing adaptation experiment
ABSTRACT
Genetic variation is fodder for evolution, and microbial plant-pathogens have it in spades. The Pseudomonas syringae genome is characterized by many rare “accessory” genes that co-occur with “core” genes found in all individuals. In fact, accessory genes outnumber core genes 2:1, even though accessory genes are not essential for survival. Moreover, there is tremendous variation in the gene content of P. syringae; isolates from different crop species, for example, differ in gene content by ~32% (Karasov et al. 2017). Whether these strain-specific genes have adaptive potential remains unknown; they may simply be a consequence of high rates of mutation and lateral gene transfer, even if purifying selection to remove deleterious variants is strong. Another, not mutually exclusive possibility is that accessory genes are maintained by positive selection as pathogens adapt to alternative hosts. Indeed, local adaptation has been hypothesized to explain the presence of rare alleles in P. syringae, which causes major agricultural loss in multiple crop species each year. To address these hypotheses, I have paired a set of P. syringae isolates with their original hosts of isolation. I first test for local adaptation by comparing the in planta fitness of each isolate in its own, and in each other’s, native host. Next, I ask to what degree strain-specific genes influence adaptive patterns by using Tn-seq to track the in planta gene frequencies of each pathogen over the course of infection in each host. From this combination of experiments, we will learn to what extent host ecology influences genome evolution and virulence in P. syringae; this is important not only to inform our understanding of the selective process, but also to fields concerned with the emergence and spread of infectious disease.
During the last two decades, scientists achieved a better understanding of the molecular basis of host-parasite co-evolution. However, many studies focused on the interaction of the genetic plant model species Arabidopsis thaliana and the highly pathogenic but non-specific tomato pathogen Pseudomonas syringae pv. tomato DC3000.
The Bergelson lab studies the interaction of A. thaliana and one of its highly abundant bacterial resident, P. viridiflava . We previously identified broad-scale natural variation in resistance phenotypes towards two distinct clades of P. viridiflava . While some genotypes of A. thaliana show little signs of disease or low bacteria titer, others suffer from severe hydrolysis of leaf tissue.
In a collaboration with Fabrice Roux, Joy Bergelson and Madlen Vetter, we currently identify and confirm the genetic loci underlying strain-specific and general defense mechanisms of A. thaliana against its natural pathogen P. viridiflava.
Araki, Hitoshi, Hideki Innan, Martin Kreitman, and Joy Bergelson. 2007. “Molecular Evolution of Pathogenicity-Island Genes in Pseudomonas Viridiflava.” Genetics 177 (2): 1031–41. https://doi.org/10.1534/genetics.107.077925. Cite
Traw, M Brian, Joel M Kniskern, and Joy Bergelson. 2007. “SAR Increases Fitness of Arabidopsis Thaliana in the Presence of Natural Bacterial Pathogens.” Evolution; International Journal of Organic Evolution 61 (10): 2444–9. https://doi.org/10.1111/j.1558-5646.2007.00211.x. Cite
Araki, Hitoshi, Dacheng Tian, Erica M Goss, Katrin Jakob, Solveig S Halldorsdottir, Martin Kreitman, and Joy Bergelson. 2006. “Presence/Absence Polymorphism for Alternative Pathogenicity Islands in Pseudomonas Viridiflava, a Pathogen of Arabidopsis.” Proceedings of the National Academy of Sciences of the United States of America 103 (15): 5887–92. https://doi.org/10.1073/pnas.0601431103. Cite
The outcome of host-microbe interactions is influenced by host genetics and interactions among bacterial community members. Previous studies described the bacterial community associated with Arabidopsis thaliana in the field. Using controlled greenhouse experiments we now aim to characterize how endophytic species composition influences plant-pathogen interactions. We furthermore seek to identify host genetic loci underlying the putative control of bacterial community composition.