In previous work, we found that a natural B. rapa plant population in southern California evolved earlier flowering  and increased fungal disease susceptibility  following a 7-year natural drought. A shift in susceptibility for this plant pathogen system could have widespread implications because this fungus is an important agricultural problem for B. rapa crops [29, 30]. Also, we have found widespread infection (22.2% of plants observed to be infected) in all 6 of our B. rapa field populations on the west coast of the USA . In the current study, we compared the genomics of ancestral and descendant plants from one population to elucidate the underlying genetic and molecular basis of this shift in disease susceptibility. This is the first study, to our knowledge, to examine the genetic basis of evolution of susceptibility to pathogens in a natural plant population following drought [32-35]. There are many challenges to studying evolution in natural systems. Using a resurrection genomics approach we were able to both directly measure the disease susceptibility of ancestors and descendants in common conditions, and directly assess evolution by measuring changes in allele frequencies between these ancestral and descendant populations .
EVOLUTION OF NFPR GENES
Using an outlier FST analysis approach, we found evidence for our first hypothesis, that a number of NFPR genes evolved over the course of just 7 years of drought. This is consistent with selection acting on a trait that is controlled by multiple loci, such as defense response to a necrotrophic fungus. This is expected to be more complex than response to a biotrophic pathogen, which in many cases is controlled by a single gene [6, 36]. We did not see significant evolution across the wider NFPR system compared to the rest of the genome. It is possible, given the genetically complex nature of NFPR, that specific causative genes could evolve, underlying a shift in susceptibility, without evolution across the wider system, especially given the short time frame between ancestors and descendants studied. In addition, we found that a substantial proportion of these NFPR genes are JA/COI1 dependent. Studies in the closely related A. thaliana-A. brassicicola model plant pathogen system have identified COI1-dependent genes, part of the JA signalling pathway, as playing a central role in defense against necrotrophic fungal pathogens, suggesting that JA signalling mediates the defense response. JA activity is dependent on the downstream COI1 gene, which encodes an F-box protein involved in proteolysis . JA signalling might play an important role in the evolution of susceptibility in addition to its known role in a plastic response to necrotrophic fungal pathogens.
There are limitations to using an outlier FST analysis to characterize the ultimate molecular or developmental mechanisms by which a population phenotypically evolves (e.g. driving mechanism of evolution). Additionally, our pooling sample size may limit us to identifying alleles with larger changes in allele frequency. However, our outlier FST approach, together with a q-value threshold correction, is effective at characterizing loci that showed significant shifts, and thereby can identify putative candidate genes underlying genetically-based phenotypic shifts observed. This approach has also been successfully utilized in other studies in these populations .
POTENTIAL SIGNATURES OF PLEIOTROPIC EVOLUTION
In previous work, we found adaptive evolution in response to drought (through earlier flowering) also resulted in the evolution of pathogen susceptibility in this population . In exploration of this associated response, in the current study, we also found evidence for our second hypothesis: that a number of genes involved in both drought and pathogen response evolved, suggesting a signature of pleiotropic evolution. In the current study, we considered genes to be candidates for pleiotropic evolution for drought and necrotrophic fungal pathogens if they had been found to be differentially expressed in response to independent drought stress and inoculation with a necrotrophic fungal pathogen, and if in our study they evolved between ancestors and descendants. A subset of five of these genes evolved in our study. Allelic variation in any of these five genes may have simultaneously affected both drought response and NFPR response, i.e. caused pleiotropic phenotypic changes. For this population we found clear evidence that JA pathway genes, which affect both pathogen response, flowering time, and stress resistance [38-40], evolved, suggesting evidence for evolution in pleiotropic genes in this system. These findings agree with other studies which have found that pleiotropy plays a role in shaping multiple ecological traits for plant populations displaying adaptation to local water availability . Furthermore, it is possible that antagonistic pleiotropy was playing a role in the evolution observed . Under this scenario, pleiotropic genes underlying both drought and pathogen response could have evolved under positive selection due to drought, despite negative selection by pathogens, if the negative selection was not strong enough to overcome the positive selection, or if it was not acting in the contemporary environment . Indeed, the increased susceptibility we observed in this population was accompanied by an evolutionary shift to thinner leaves (increased specific leave area) , which could indicate a trade-off between drought escape and disease susceptibility in agreement with the growth and defense trade-off [43, 44].
In addition to examining evolution following the recent drought, we tested whether NFPR genes were under historical directional selection by looking for genes with a low Tajima's D (<-2). Loci under balancing selection would not have a low Tajima's D. Genome-wide, we found no association between FST and Tajima’s D, and there were no genes that were outliers for both high FST (>0.2) and low Tajima’s D (<-2). This finding concurs with other prior work in this system . This indicates that evolution in response to the drought affected different genes in this population than have historically been the target of directional selection.
TOP CANDIDATE GENES UNDERLYING EVOLUTIONARY SHIFT IN SUSCEPTIBILITY
The most highly differentiated gene in this study was Sec14p-like phosphatidylinositol transfer family protein (Bra030295) (Figure 2A; FST = 0.21; q = 1.75 x 10-5) which has been found to be differentially expressed in Arabidopsis in response to infection with A. brassicicola  and cabbage leaf curl virus . Little is known about the specific function of this gene, although it is thought to play a role in the movement of substances across cell membranes and has been found to be expressed widely in 23 plant structures . The shift in allele frequency at two non-synonymous segregating sites in this gene indicate a potential functional effect, with one of these polymorphisms being shared between ancestors and descendants (potentially indicating a previous shift due to the recurrent droughts in this region), and the other polymorphism found only in the descendant population, indicating a better functional candidate in response to this drought event.
Several of the NFPR genes with the highest level of differentiation between ancestors and descendants demonstrate the pleiotropic nature of the genes subject to evolution in response to the drought in this population. We found that the second most highly differentiated NFPR gene was the COI1 dependent gene ERF4 (Bra038107)(FST = 0.20; q = 7.30 x 10-5). In addition to a demonstrated role in the antagonistic regulation of JA and ET, this gene is also induced by ABA [47, 48]. This is interesting because this gene, along with other genes mediated by ABA that we found to have high FST, including Bra012746 and Bra033968, and Bra011282, have pleiotropic roles in both abiotic and biotic stress response [49-51]. Studies have shown that ERF4 is induced by ABA, and ERF4 suppresses expression of PDF1.2 , a defense effector that is elicited by JA signaling . This gene showed a shift in allele frequency at three non-synonymous sites, and evolution at these sites could have a functional effect (Figure 2B). Two of these polymorphisms were shared between ancestors and descendants (potentially indicating a previous shift due to the recurrent droughts in this region), and the third polymorphism was found only in the descendant population, indicating a better functional candidate in response to this drought event.
Caffeoyl-CoA 3-O-methyltransferase (Bra033968) is an NFPR gene that is COI1 dependent  and also had a significant FST value. It is a lignin biosynthesis gene, which is pleiotropic, playing a role in both drought and pathogen response. Lignification is a primary defense mechanism against pathogens , which works primarily through leaf structure . It has also been shown to be suppressed in Arabidopsis by ABA application, along with other pathogen response genes, resulting in a decrease in lignin accumulation and an increase in disease susceptibility to Pseudomonas syringae pathovar (pv.) tomato . This gene is also interesting because lignin plays a key role in leaf structure, and we found evidence that leaf structure played a role in the increase in disease susceptibility seen , with earlier flowering plants having thinner leaves that displayed greater disease severity.
LIMITATIONS OF THIS STUDY
This study was limited due to it being an observational study focused on one natural population that underwent an evolutionary shift in response to an extended drought. Alternative explanations to pleiotropy include that drought response and pathogen susceptibility were acted on separately, or that if there is a large cost of disease resistance and these pathogens become less common during a drought, then selection could favor reduced resistance due to these costs. The current study design does not allow us to differentiate between these alternative explanations. Experimental evolution studies with additional populations, including a non-drought control population will be important to address some of these findings further. Additionally, epigenetics, especially longer-term stable epigenetics could have played a role in the shift observed, and was not tested in the present study. Such epigenetic effects may act in addition to or in concert with the observed genetic changes, and would not replace or negate the contribution of genetic change. Although seeds from prior generations emerging from the seedbank could have influenced our findings, this effect is likely minimal and would only have caused our analyses to be more conservative because less evolution would be observed than actually occurred . Finally, in addition to pleiotropy, correlated change in these two phenotypes could have been caused by drift, linkage, or by simultaneous change in selective pressure on both flowering time and pathogen load. While we cannot eliminate these from consideration, they likely played a minor role in the observed patterns due to data presented in a previous study showing an adaptive evolution in response to drought, lack of evidence of a bottleneck, and the short 7-year time frame over which the evolution occurred, too short for drift alone to explain the extensive allelic differentiation observed .