Supplementary Text
1. Model
(a) Description of resource-consumer model
Agrobacterium tumefaciens, the facultative pathogen and soil microbe, can sustain growth in what we define as three distinct environments: bulk soil (B), rhizosphere (R), and tumor (T). The bulk soil environment (B) only contains general resources (G), available to all plasmid genotypes. The rhizosphere environment (R) contains general resources (G) in addition to the plant-produced resources deoxy-fructosyl-glucosamine, DFG and γ-butyrolactones, GBLs (D), which can only be utilized by pAt+ cells. The tumor environment (T), contains general resources (G), plant-produced DFG and GBLs (D), as well as the infected-plant-produced resources, opines (O) which can only be catabolized by pTi+ cells.
Based on the principles of a basic Monod model for bacterial population growth, the change in population size of each competing genotype is determined as a function of the initial population size (, birth (; determined by the conversion of available environment-specific resources into new cells; representing the concentration of each resource ‘‘ (; general, ; DFG and GBLs, or ; opines) in the pertinent environment ‘’ (; bulk, ; rhizosphere, or ; tumor)), and death (; a constant per capita rate that is equal for all genotypes). For a between genotype comparison of growth and death rates in the conditions of competition see Tables S5 and S6.Although the general, rhizosphere, and tumor resources include multiple forms of carbon that are utilized by cells with varying efficiencies, we find that the qualitative outcomes predicted by the model are robust to changes in resource levels and substitutability. Thus, for the purpose of simplicity, we assume that all resources are substitutable.Growth of each plasmid-bearing genotype is limited by the predicted associated plasmid costs ().
For a description of all terms, see Table S4.
We have established that the cost of the Ti plasmid is very low outside of the disease environment, but when actively engaged in pathogenesis, the costs associated with virulence are significant [1]. Thus, the primary benefits and costs of this plasmid are coupled with, and restricted to, this environment and the presence of opines. This tradeoff is incorporated into the model so that Ti plasmid costs are set to correlate with the relative abundance of opines (i.e. the product of disease) as a Type II functional response.
Whereas the cost associated with the Ti plasmid are minimized when it is not receiving a resource benefit, we found that when in isolation, the At plasmid has a very high cost under carbon-limiting conditions. We know that cells harboring the At plasmid carry specific genes that are likely to provide environment-specific benefits such as catabolism of GBLs and DFG, which are abundant in the rhizosphere[2]. In spite of the high cost of the At plasmid, its large size and prevalence in nature suggest the benefits it confers to A. tumefaciens cells are substantial. As the only known benefits of the At plasmid are associated with carriage of catabolism genes targeted to plant produces resources, we assume that the cost of the At plasmid is the same in all environments and the benefits are restricted to the rhizosphere and tumor, where D resources (GBLs and DFG), are expected to be present in equal abundance.
Competitions performed in the study also demonstrate that the costs of co-inhabitant plasmids are lower than pAt alone. This non-additivity was observed under carbon-limiting conditions when neither rhizosphere or tumor-specific opines resources (D and O, respectively) were present. Further substantiating this non-additivity is that the costly Ti plasmid virulence genes are expressed at lower levels in pAt+ cells (figure 2), suggesting a lower cost for pTi+ cells in the disease environment as well. This non-additivity was incorporated into the model as the modifier allowing us to establish the potential ecological significance of plasmid epistasis on the outcome of the predicted resource-consumer interactions.
The concentrations of available resources (G, D, and O) in each environment (B, R, and T) are calculated by the flow rates into the environment, the conversion of resources into cell biomass, and the rate of diffusion out of the environment. The supply rate () and diffusion rate () are held constant for all resources, where is always assumed to be less than and the rate of resource conversion into cell biomass is genotype dependent.
Concentration of resources ) in the environment ):
(b) Resource-dependent population growth
Population growth of four plasmid genotypes: p- (plasmidless), pAt+, pTi+ and pAt+pTi+, depends on the relative availability of environment-specific resources.
The change in population size of the plasmidless genotype in all three environments depends on the initial population size (), its growth rate (; determined by the conversion of available general resources () into new cells) and the death rate (; a constant per capita rate that is equal for all genotypes). Equations 1.1, 1.2, and 1.3 describe the change in population size for the plasmidless genotype in all three environments: , , and .
Equation 1.1
Equation 1.2
Equation 1.3
Change in population size of the singly infected pAt+ genotype depends on the availability of general resources () and plant-produced resources (), which are present in both the rhizosphere and the tumor environments. Growth of this genotype is also affected by the cost associated with harboring the plasmid (). Thus, in a mixed population of p- and pAt+ cells, plasmidless cells can take advantage of general resources more quickly due to the absence of any plasmid cost. But, because both genotypes are able to use resources, and all resources are equally substitutable, as resources are depleted where resources are present, the pAt+ genotype will increase in population size as the p- genotype declines. Equations 2.1, 2.2, and 2.3 describe the change in population size of the pAt+ genotype in all three environments.
Equation 2.1
Equation 2.2
Equation 2.3
The change in population size for the singly infected pTi+ genotype is distinct in that growth depends on the availability of both general resources () and opines (). Additionally, growth on opines is directly coupled with tumor-associated costs . Equations 3.1, 3.2, and 3.3 describe the change in population growth for the pTi+ genotype in the and environments.
Equation 3.1
Equation 3.2
Equation 3.3
The change in population size of the pAt+pTi+ genotype depends on the availability of all resource types: general (), DFG and GBLs (), and opines (). Here, the modifier is incorporated to account for the non-additive costs of the two plasmids. Equations 4.1, 4.2, and 4.3 describe the change in population growth for the pAt+pTi+ genotype in each relative environment.
Equation 4.1
Equations 4.2
Equation 4.3
(c)Genotype-dependent resource availability
The concentrations of available resources are a function of the supply, diffusion rate, and use (conversion into microbial biomass). In this model, the supply and diffusion rates for each resource are constant, and it is the composition of the microbial population (i.e. relative genotypes) that varies as a function of available resources.
Thus, the change in concentration of each type of resources depends on 1) the relative abundance of each genotype, and 2) the availability of other resources in that specific environment that are contributing to increases in population size.
Equation 5.1 describes the change in resources in the bulk soil environment.
and =0.
Equation 5.1
Equations 6.1 and 6.2 describe the change in and resources in the rhizosphere. .
Equation 6.1
Equation 6.2
Equations 7.1, 7.2, and 7.3 describe the change in , , and resources in the tumor environment.
Equation 7.1
Equation 7.2
Equation 7.3
2. Supplementary Methods
(a) Curing the Ti plasmid
The Ti plasmid was cured from A. tumefaciens C58 using the approach developed by Uraji et al. 2002, and explained in detail in Morton and Fuqua 2012. The curing vector, pEM112, was constructed by cloning the entire replication region of pTiC58 (repABC) into pNPTS138, which contains a counterselectablesacBgene conferring sucrose sensitivity. We used a stock of Agrobacterium tumefaciensC58 (designated C58-CU) provided by Professor Steven Beer at Cornell University, that is reported to be as close to the original isolate (Cherry gall, Geneva, NY, 1958) as is available. The curing vector was introduced into C58-CU by conjugation from an Escherichia coli S17-1 λpirdonor. Plating on ATGN supplemented with kanamycin selects for A. tumefaciens pEM112 transconjugants. The two plasmids, pEM112 and pTiC58, use the same replication and partitioning machinery, and are therefore incompatible [3, 4]. As a consequence, transconjugants that contain both plasmids give rise to segregants that lack one or the other plasmid. Growth in the presence of kanamycin selects for only those cells that carry pEM112 and have lost pTiC58. These cells were subsequently screened for loss of the Ti plasmid by a lack of AHL production and diagnostic PCR screening for loss of plasmid genes, targeting multiple regions distributed across the Ti plasmid. Subsequently, these pTiC58- derivatives were grown in the presence of 5% sucrose and in the absence of antibiotic to select for cells that had lost the curing vector. These sucrose-resistant colonies were patched onto ATGN supplemented with kanamycin to screen for KmS clones. Sucrose-resistant and kanamycin-sensitive clones were screened for the absence of repABCby PCR.
(b) Curing the At plasmid
Curing of the At plasmid was first attempted using the same method first described by Uragi et al. 2002. The curing vector that was generated, pEM123 contained the repABC region of pAtC58. However, this method alone was ineffective in curing the At plasmid. Exposure to kanamycin, resistance to which was conferred by the expression of the resistance gene carried on the curing vector, instead selected for the co-integration of the At and curing plasmids. The repABC region provided plenty of genetic material for recombination to occur and the result was that these chimeric plasmids would be stably maintained throughout the entire selective process.
In an attempt to destabilize the At plasmid, a deletion in the core sequence of the plasmid stability gene, repA, was generated. Interestingly, the result of the allelic replacement process yielded clones that harbored both the wild-type and mutant forms of the plasmid. Due to plasmid incompatibility and the inherently less stable quality of the repA- plasmid, the mutant plasmids were quickly lost from the population. The difficulty in curing the At plasmid suggested that either some genes on the At plasmid were essential, or that a toxin-antitoxin (TA) system was preventing survival of any pAt- variants that arose. The At plasmid carries several putative toxin-antitoxin systems that were identified by RASTA-Bacteria (Rapid Automated Scan for Toxins and Antitoxins in Bacteria). One pair of adjacent genes, Atu5112 and Atu5113, exhibit homology to the established TA system, HipA and HipB, respectively. HipA is a protein kinase that, in the absence of neutralization by HipB, phosphorylates EF-Tu to inhibit protein synthesis and induce cell dormancy [5]. A genome-wide screen for essential genes in A. tumefaciens C58 recovered very few isolates with transposon insertions in Atu5113, the putative antitoxin, providing support for this being a functional TA system (Curtis and Brun, unpublished). In an attempt to remove this selection for maintenance of the At plasmid (dormancy of any pAt-cured derivative), the predicted toxin, Atu5112 was deleted. The curing plasmid was subsequently introduced into the ΔAtu5112 mutant, and positive transformants were subjected to heat shock (42°C for 60 seconds) before passaging for two more days in the presence of kanamycin (to select for preferential maintenance of the curing vector). This combinatorial approach proved to be effective and approximately 5 out of 500 screened colonies had lost the At plasmid: a curing rate quite low relative to the Ti plasmid (~1% versus 85%). For further confirmation of plasmid loss, all five clones were screened by PCR using primers specific to regions distributed around the At plasmid. They were also tested for pAt-conferred bclactivity by spotting putative pAt- clones on a petri plate containing growth media, an AHL-producing strain, an AHL-reporter strain, Xgal, and salicylic acid (inducer of bcl expression). Strains that exhibit bcl activity are able to catabolize the AHLs produced by the AHL-producing strain that is growing in the media. The presence of the AHL-reporter strain in combination with Xgal results in the development of a blue color in the media wherever AHLs are present. Catabolism of these AHLs results in a loss of blue color, or a clear zone, around any bcl-expressing pAt+ strain.
(c) Testing isogenicity of derived strains
To determine if plasmid curing, conjugation, and additional genetic manipulations had resulted in unintended genetic changes impacting the fitness of the derived wild-type (doubly-infected strain), we competed the derived ERM89 (pAt+pTi+) strain with the original C58-CU strain (same methods and analyses as all other competitions) and found no significant difference in relative fitness (WC58-CU = 1.0025, SE= 0.005, p = 0.6020).
(d) Generating isogenic strains
To control for the possibility of mutations that could have arisen during the curing process, isogenic strains were generated by reintroduction of the At and Ti plasmids into the plasmid-free derivative, ERM52. Markers conferring resistance to gentamycin and ampicillin were introduced by allelic replacement into the Ti and At plasmids of C58-CU, respectively. For the Ti plasmid, aacC1 was cloned into the intergenic region between the two convergent genes, Atu6112 and Atu6113. For the At plasmid, bla(conferring resistance to ampicillin) was inserted between Atu5196 and Atu5197 (also convergent). These plasmids were then conjugativelytransferred to ERM52 to generate pAt+, pTi+, and pAt+pTi+ derivatives (ERM73, ERM76, and ERM77). To limit marker effects during competition, the aacC1 and bla genes were subsequently removed by the same allelic replacement strategy.
(e) Measurement of virulence gene induction
Strains ERM89 (pAt+pTi+) and ERM66 (pTi+) were transformed with the reporter plasmid, pSW209Ω (S. C.Winans, Cornell University) which carries a PvirB::lacZ fusion (PvirB from pTiA6). Cells were grown approximately 24 h in Induction Broth (pH 5.6, 50 μM phosphate, and 200 μMacetosyringone) [50]. When cultures reached mid-log phase of growth, they were assayed for β-galactosidase activity as described [46]. Activity is presented in terms of Miller Units; a quantitative measure of specific activity that accounts for gene-expression mediated β-galactosidaseactivity, normalized to growth. The equation for calculating Miller Units () is as follows:
, whererepresents absorbance of o-nitrophenol, represents optical density of the culture, is the time of the reaction, and is the culture volume. Each treatment was carried out in triplicate.
Supplementary References
1.Platt T.G., Bever J.D., Fuqua C. 2012 A cooperative virulence plasmid imposes a high fitness cost under conditions that induce pathogenesis. Proc R Soc Lond Ser B-Biol Sci279(1734), 1691-1699. (doi:10.1098/rspb.2011.2002).
2.Baek C.-H., Farrand S.K., Park D.K., Lee K.E., Hwang W., Kim K.S. 2005 Genes for utilization of deoxyfructosyl glutamine (DFG), an amadori compound, are widely dispersed in the family Rhizobiaceae. FEMS Microbiol Ecol53(2), 221-233.
3.Uraji M., Suzuki K., Yoshida K. 2002 A novel plasmid curing method using incompatibility of plant pathogenic Ti plasmids in Agrobacterium tumefaciens. Genes & Genetic Systems77(1), 1-9. (doi:10.1266/ggs.77.1).
4.Cevallos M.A., Cervantes-Rivera R., Gutierrez-Rios R.M. 2008 The repABC plasmid family. Plasmid60(1), 19-37. (doi:10.1016/j.plasmid.2008.03.001).
5.Hansen S., Vulic M., Min J.K., Yen T.J., Schumacher M.A., Brennan R.G., Lewis K. 2012 Regulation of the Escherichia coli HipBA Toxin-Antitoxin System by Proteolysis. PLoS One7(6). (doi:10.1371/journal.pone.0039185).