Local adaptation of both plant and pathogen: an arms-race compromise in switchgrass rust

Abstract

In coevolving species, parasites locally adapt to host populations as hosts locally adapt to resist parasites. Parasites often outpace host local adaptation since they have rapid life cycles, but host diversity, the strength of selection, and external environmental influence can result in complex outcomes. To better understand local adaptation in host-parasite systems, we examined locally adapted switchgrass (Panicum virgatum), and its leaf rust pathogen (Puccinia novopanici) across a latitudinal range in North America. We grew switchgrass genotypes in 10 replicated multiyear common gardens, measuring rust severity from natural infection in a 'host reciprocal transplant' framework for testing local adaptation. We conducted genome-wide association mapping to identify genetic loci associated with rust severity. Genetically differentiated rust populations were locally adapted to northern and southern switchgrass, despite host local adaptation to environmental conditions in the same regions. Rust resistance was polygenic, and distinct loci were associated with rust severity in the north and south. We narrowed a previously identified large-effect quantitative trait locus for rust severity to a candidate YELLOW STRIPE-LIKE gene and linked numerous other loci to defense-related genes. Overall, our results suggest that both hosts and parasites can be simultaneously locally adapted, especially when parasites impose less selection than other environmental factors.

Keywords: BLUP; GWAS; biofuel; coevolution; fungal disease; microbial ecology; rust; switchgrass.

Fig. 1

Collection locations (circles), and planting sites (squares) for the replicated switchgrass diversity panel. Coloring for circles indicates switchgrass population membership based on shared single‐nucleotide polymorphisms, and coloring for squares indicates the geographic region used for this manuscript. From north to south, the sites are Brookings, SD; Kellogg Biological Station, MI; Fermilab, IL; Lincoln, NE; Columbia, MO; Stillwater, OK; Overton, TX; Temple, TX; J.J. Pickle Research Campus, TX; and Kingsville, TX.

Fig. 2

Comparing two methods for testing parasite local adaptation, a parasite reciprocal transplant (AC) and a host reciprocal transplant (BD; this study). In this example scenario, two host populations (switchgrass plant icons) and two parasite populations (circular leaf rust icons) are differentiated between two sites, North and South. (a) Traditional parasite reciprocal transplant would introduce northern and southern rust into switchgrass populations at each site. (b) A host reciprocal transplant would introduce northern and southern switchgrass in a common garden to endemic rust populations at each site. (c) Proof of parasite local adaptation via the local–foreign comparison would require greater fitness for southern rust over northern rust on southern switchgrass, and greater fitness for northern rust over southern rust on northern switchgrass. This is indicated by greater rust severity from southern rust (red boxes) in the south site, but greater rust severity from northern rust (blue boxes) in the north site. (d) Proof of parasite local adaptation would require the same fitness advantage as in a parasite transplant: northern rust over southern rust on northern switchgrass, and greater fitness for southern rust over northern rust on southern switchgrass. However, this test would be within host populations rather than within sites, as indicated by swapping the transplant site to the legend from the x‐axis. Boxplots shown in (c, d) are shown for demonstration and do not include real data; they are drawn to depict outliers as points, 1.5 × the interquartile range as whiskers, the 25^th and 75^th percentiles as upper and lower box limits, and the median as the center line. This figure was created in BioRender (BioRender.com/8nb3jl8).

Fig. 3

Puccinia novopanici leaf rust population genetics across nine sites. (a) The y‐axis indicates the latitude of the collection site, and the x‐axis is the first principal component of a principal component analysis of 2.4 million single‐nucleotide polymorphisms, indicating genetic similarity. Site codes correspond to: B – Brookings, SD; C – Columbia, MO; F – Fermilab, IL; K – Kingsville, TX; L – Lincoln, NE; M – Kellogg Biological Station, MI; P – J.J. Pickle Research Campus, TX; S – Stillwater, OK; T – Temple, TX. (b) Puccinia novopanici sori under field conditions in Kingsville, TX. Photograph by Acer VanWallendael. (c) Mean nucleotide diversity (Pi) across 10‐kb windows in the P. novopanici genome. Positions are shown by mapping location in the Puccinia triticina genome.

Fig. 4

Rust score variation across switchgrass populations, space, and time. (a) Site and year variation in rust infection on switchgrass. Points shown indicate the area under the disease progression score across 8 wk for each plant in each year. Sites are shown on the x‐axis; from north to south, the sites are BRKG (Brookings, SD), KBSM (Kellogg Biological Station, MI), FRMI (Fermilab, IL), LINC (Lincoln, NE), CLMB (Columbia, MO), OVTN (Overton, TX), TMPL (Temple, TX), PKLE (J.J. Pickle Research Campus, TX), and KING (Kingsville, TX). 2019–2020 data could not be collected at OVTN (Overton, TX) and rust was not found in BRKG in 2020–2021. (b) Genetic subpopulations vary in rust severity. Points indicate mean scaled rust severity (area under the disease progress curve (AUDPC)) across sites and years. The Gulf genetic population has a subdivision between genotypes originating inland and those by the coast. Boxplots show outliers as points, 1.5 × the interquartile range as whiskers, the 25^th and 75^th percentiles as upper and lower box limits, and the median as the center line. (c) Phenotypic principal component analysis biplot for major traits across all sites and years. Points are colored by subpopulation.

Fig. 5

Evidence for rust local adaptation to switchgrass across two regions. Switchgrass genomic principal component analysis showing susceptibility to endemic rust in northern (a) and southern (b) sites. Rust severity best linear unbiased predictor (BLUP) was measured in northern and southern sites. Negative scores indicating genotype resistance are shown as black, positive scores indicating genotype susceptibility are colored. (c) Rust BLUPs by host population and region. Blue boxes indicate samples planted in northern common gardens, red boxes indicate samples planted in the southern gardens. Asterisks indicate regional differences; Dunn's test P < 0.0001 for both. Boxplots show outliers as points, 1.5 × the interquartile range as whiskers, the 25^th and 75^th percentiles as upper and lower box limits, and the median as the center line.

Fig. 6

Genome‐wide association study (GWAS) of disease severity best linear unbiased predictors in northern (a) and southern (b) regions. Vertical lines indicate the positions of candidate loci found in previous experiments (VanWallendael et al., 2020, 2022a). The green line on chromosome 2N indicates a microbiome structure – associated GWAS outlier, and blue lines indicate rust resistance quantitative trait loci Prr1 (Chr03N) and Prr2 (Chr09N). Red horizontal lines indicate a Bonferroni cutoff. Red arrows indicate points that were also identified as outliers in population‐specific GWAs.

Fig. 7

Three terpenoid cyclase genes were expressed almost exclusively in switchgrass leaf tissue from northern upland switchgrass varieties grown in northern sites in 2016. These three genes are linked to large genome‐wide association study outliers on chromosome 1K. AP13 and WBC are southern genotypes in the Gulf population; DAC and VS16 are northern genotypes in the Midwest population. Boxplots show outliers as points, 1.5 × the interquartile range as whiskers, the 25^th and 75^th percentiles as upper and lower box limits, and the median as the center line.