Study system:
- viruses: Barley yellow dwarf virus PAV (BYDV-PAV) and cereal yellow dwarf virus RPV (CYDV-RPV)
- aphid vector: Rhopalosiphum padi
- host plant: Avena sativa L. cv. Coast Black Oat
- soil inoculum: collected from Cedar Creek Ecosystem Science Reserve and Long Term Ecological Research (LTER) site (CDR)
Soil microbes:
In June 2014, we collected soil cores from a long-term experiment in a successional grassland at CDR (experiment “E001”, www.cedarcreek.umn.edu; Bethel, MN, USA). We collected soils from field A, which was abandoned from agriculture in 1968 and burned annually beginning in 2005. These plots had received annual additions of P, K, Ca, Mg, S, and citrate-chelated trace metals since 1982 and three levels of N fertilizer: 0, 34, or 270 kg N ha-1 yr-1 (see Tilman, 1987 for details). We randomly sampled three plots for each N fertilization rate (54H, 52H, 45A, 40H, 38D, 23D, 22A, 17H, and 8D) and six locations within each 4 x 4 m plot to extract a soil core (1.9 cm diameter and 10 cm deep). In the lab, soil cores were passed through a 4 mm sieve, then twice through a 2 mm sieve to remove coarse debris and roots, and then combined based on their N fertilization rate. Next, we prepared soil microcosms by filling four large, surface sterilized bins with 17 L of potting soil composed of 70% Sunshine medium vermiculite (vermiculite and less than 1% crystalline silica; Sun Gro Horticulture, Agawam, MA, USA) and 30% Turface MVP (calcined clay containing up to 30% crystalline silica; Turface Athletics, Buffalo Grove, IL, USA), saturated with tap water (approximately 5 L for every 20 L of dry soil) and autoclaved at 121oC and 15 psi for 60 minutes to kill the naturally existing microbial consortium. We then mixed 350 mL of field soil from each N fertilization level separately into the bins. Field soil comprised approximately 2% of the bin soil volume. We did not mix field soil into the fourth bin. Lastly, we covered the bins with non-airtight lids and incubated the soil at 25°C for 11 days.
Experimental setup and implementation:
For each of the four soil microcosms, we filled 80 conical plastic pots (3.8 cm diameter x 21 cm depth, 164 ml) with soil mixture and planted one A. sativa seed per pot 4.5 cm from the surface of the soil. Seeds were obtained from the USDA (National plant germplasm system, USDA; USA) in June 2013 and were surface sterilized with 12.5% bleach solution. Then, we haphazardly assigned plants to later receive one of two N supply rates (7.5 μM NH4NO3 was “low N” and 375 μM NH4NO3 was “high N”) and one of four virus inoculations (BYDV-PAV, CYDV-RPV, co-inoculation, or mock inoculation), leading to ten replicates per treatment. Plants grew in a growth chamber containing only healthy plants with a 16:8 h light:dark cycle at 19-20oC under Lumilux high pressure sodium ET18 bulbs for 11 days. Two days after planting, we watered the pots with 30 ml of the modified Hoagland solution corresponding to the plant’s assigned N supply rate. We watered plants with these solutions twice per week until harvest.
When the plants had been growing for 22 days, we used R. padi aphids to inoculate them with BYDV-PAV, CYDV-RPV, both viruses, or to perform a mock inoculation. Rhopalosiphum padi were obtained from Dr. G. Heimpel at the University of Minnesota (St. Paul, MN, USA) and reared on A. sativa in growth chamber conditions described above (except with 28W Ultramax EcoXL lights). BYDV-PAV and CYDV-RPV isolates were obtained from Dr. S. Gray at Cornell University (Ithaca, NY, USA) in January 2013. They were also maintained in A. sativa plants in similar growth chamber conditions (except with 40W cool white light bulbs). We inoculated plants by allowing aphids to feed on either BYDV-PAV- or CYDV-RPV-infected A. sativa tissue in 25 mL glass tubes sealed with corks for approximately 48 hours. Then, we transferred the aphids to 2.5 x 8.5 cm, 118 μm polyester mesh cages secured to one leaf on each experimental plant with Parafilm and bobby pins. Ten aphids were used to inoculate each plant, with 5 carrying each virus for the co-inoculation treatment, five viruliferous (carrying virus) and five non-viruliferous aphids for each single virus treatment, and ten non-viruliferous aphids for the mock inoculation treatment. We allowed aphids to feed on the experimental plants for approximately 96 hours, after which we manually killed all aphids and removed the cages. Plants grew for 19 more days before we took measurements. To estimate N stress through leaf chlorophyll content, we took three measurements per plant with a SPAD-502 Meter (Soil Plant Analysis Development; Konica Minolta, Tokyo, Japan). Then, we harvested and weighed the aboveground biomass, which we stored at -20°C until it was analyzed for virus infection.
Detection of B/CYDV infection:
To extract total RNA, we ground approximately 50 mg of leaf tissue in a bead-beater with a copper BB and 1 ml of TRIzolTM Reagent (InvitrogenTM, Thermo Fisher Scientific, Waltham, MA, USA) per the manufacturer’s instructions. We then purified RNA from the cellular components following the extraction protocol published by Lacroix et. al. (2014). We re-suspended the purified RNA in nuclease-free water and stored the samples at -20°C until performing the reverse transcription polymerase chain reaction (RT-PCR). We used a nanodrop spectrophotometer (Thermo Fisher Scientific) to quantify the concentration of RNA within each sample and then performed a multiplex RT-PCR assay to isolate and amplify BYDV-PAV and CYDV-RPV nucleic acids as published previously (Deb & Anderson, 2008; Lacroix et al., 2014). We combined 5 μl of each PCR product with 2 μl of 6X loading dye (Genesee Scientific, El Cajon, CA, USA) and loaded the samples and 100 bp DNA ladder (Apex Bioresearch Products, North Liberty, IA, USA) into an Agarose-1000 gel (Invitrogen, Thermo Fisher Scientific) stained with 2% SybrSafe (Invitrogen, Thermo Fisher Scientific). After 25 minutes at 120 V, we observed the gel with a UV-light EZ doc system (Bio-Rad Laboratories, Hercules, CA, USA) to detect bands at 298 bp and 447 bp, indicating the presence of BYDV-PAV and CYDV-RPV, respectively.
Statistical analyses:
We assessed the effects of the experimental treatments on the infection incidence of BYDV-PAV and CYDV-RPV (i.e., the proportion of plants infected out of those inoculated) using binomial (logit-link) generalized linear regressions with virus infection as a binary response variable and long-term N-enriched soil treatment (non-inoculated, ambient N, low N, or high N), N supply (binary variable), whether the plants were co-inoculated (binary variable), and their interactions as independent variables. The intercepts represented singly inoculated plants grown in non-inoculated growing medium with low N supply. We tested the effects of N supply and soil treatment on co-infection incidence using an analogous procedure. Samples with an infection inconsistent with the inoculation treatment were removed from analyses. Inconsistent infections likely arose from small aphids escaping cages during the inoculation period and occurred in 31 of 229 plants (Table A2). Treatment sample sizes in the final dataset ranged from seven to ten.
To assess the effects of the experimental treatments on the A. sativa plants, we used linear regressions with log-transformed biomass and log-transformed chlorophyll content as response variables and long-term N-enriched soil treatment, N supply, successful inoculation treatment (mock, BYDV-PAV only, and CYDV-RPV only), and their interactions as the independent variables. Therefore, we omitted plants from analyses that were unsuccessfully inoculated, either because the intended infection was not detected or because an unintended infection was detected (Table A2). Co-infected plants were omitted from analyses due to limited sample sizes. The chlorophyll values used in the model were the averages of three measurements per plant. The intercepts represented mock-inoculated plants grown in non-inoculated growing medium with low N supply. Treatment sample sizes in the final dataset ranged from three to nine.
All regressions described above were fit using Bayesian models with the brms package in R version 4.0.2 (Bürkner, 2017; R Core Team, 2020). Models had three chains of 6000 iterations each with a 1000 iteration discarded burn-in period. Gaussian distributions with a mean of zero and a standard deviation of ten were used as prior distributions for intercepts and coefficients (very weakly informative). We used a half Student’s t-distributions with three degrees of freedom, a location of zero, and a scale of ten as the prior distribution for the residual standard deviations (Bürkner, 2017). We assessed model fit by ensuring that r-hat values were equal to one, that the three chains were well mixed, and that simulated data from the posterior predictive distributions were consistent with observed data. In the results, we present point estimates with 95% highest posterior density intervals based on posterior samples of model coefficients in brackets.
To evaluate the effect of sample size on the probability of detecting an effect with quantile-based 95% credible intervals that omit zero, we simulated 1000 datasets of the same sample sizes and with the mean effect size measured in the experiment. We drew simulated values from a normal distribution with its mean equal to the model estimate for the treatment and its standard deviation equal to the overall model-estimated standard deviation. We fit regressions to the simulated datasets and calculated the number of times the 95% credible intervals of the variable of interest omitted zero (Kurz, 2019). We repeated the analysis with multiple sample sizes. We performed this analysis for the effects of CYDV-RPV infection on log-transformed plant biomass, where the mean difference was -0.23, the sample sizes were 8 (mock-inoculated, low N supply, non-inoculated growing medium ) and 6 (CYDV-RPV infected, low N supply, non-inoculated growing medium), and the regression was a normal linear regression with infection status as the independent variable.
References:
Bürkner P-C. 2017. brms: An R package for Bayesian multilevel models using Stan. Journal of Statistical Software 80: 1–28.
Deb M, Anderson JM. 2008. Development of a multiplexed PCR detection method for Barley and Cereal yellow dwarf viruses, Wheat spindle streak virus, Wheat streak mosaic virus and Soil-borne wheat mosaic virus. Journal of Virological Methods 148: 17–24.
Kurz AS. 2019. Bayesian power analysis: Part I. Prepare to reject H0 with simulation.[WWW document] URL https://solomonkurz.netlify.app/post/bayesian-power-analysis-part-i/. [accessed 19 April 2021].
Lacroix C, Seabloom EW, Borer ET. 2014. Environmental nutrient supply alters prevalence and weakens competitive interactions among coinfecting viruses. The New Phytologist 204: 424–433.
R Core Team. 2020. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing.
Tilman D. 1987. Secondary succession and the pattern of plant dominance along experimental nitrogen gradients. Ecological Monographs 57: 189–214.
