Genetic distance predicts trait differentiation at the subpopulation but not the individual level in eelgrass, Zostera marina

Abstract Ecological studies often assume that genetically similar individuals will be more similar in phenotypic traits, such that genetic diversity can serve as a proxy for trait diversity. Here, we explicitly test the relationship between genetic relatedness and trait distance using 40 eelgrass (Zostera marina) genotypes from five sites within Bodega Harbor, CA. We measured traits related to nutrient uptake, morphology, biomass and growth, photosynthesis, and chemical deterrents for all genotypes. We used these trait measurements to calculate a multivariate pairwise trait distance for all possible genotype combinations. We then estimated pairwise relatedness from 11 microsatellite markers. We found significant trait variation among genotypes for nearly every measured trait; however, there was no evidence of a significant correlation between pairwise genetic relatedness and multivariate trait distance among individuals. However, at the subpopulation level (sites within a harbor), genetic (F ST) and trait differentiation were positively correlated. Our work suggests that pairwise relatedness estimated from neutral marker loci is a poor proxy for trait differentiation between individual genotypes. It remains to be seen whether genomewide measures of genetic differentiation or easily measured “master” traits (like body size) might provide good predictions of overall trait differentiation.


| INTRODUC TI ON
Trait differences within and among species can influence ecological processes at multiple levels, from populations to ecosystems (e.g., Bolnick et al., 2011;Hector et al., 1999;Macarthur & Levins, 1967;Stachowicz, Kamel, Hughes & Grosberg, 2013;Wojdak & Mittelbach, 2007). However, identifying the traits that matter to ecological processes a priori is often challenging and contextdependent (Naeem & Wright, 2003). Furthermore, measuring continuous trait variation among individuals within a species requires extensive effort and may not be practical for assemblages with a large number of taxa. Both of these challenges raise the question of whether measures of genetic distance can be used as proxies for functional divergence, based on the assumption that phylogenies, genealogies, or estimates of relatedness can reflect integrated phenotypic differences among taxa or individuals (Cadotte, Cardinale & Oakley, 2008;Felsenstein, 1985;Harvey & Pagel, 1991;Stachowicz et al., 2013).
Some evidence supports this assumption at the species level.
Similar approaches indicate that both trait diversity and genetic relatedness within species can influence the outcome of ecological interactions (Abbott & Stachowicz, 2016;Dudley & File, 2007;Stachowicz et al., 2013). For example, trait differences could lead to niche partitioning reducing competition among individuals and promoting coexistence (Chesson, 2000); alternatively, trait differences could lead to competitive exclusion if certain traits allow individuals to be competitively dominant in that habitat (e.g., Abbott & Stachowicz, 2016). Relatedness could influence intraspecifc interactions indirectly if it is correlated with trait differentiation (Jousset, Schmid, Scheu & Eisenhauer, 2011;Stachowicz et al., 2013) or directly through kin recognition (Aguirre, Miller, Morgan & Marshall, 2013;Dudley & File, 2007). However, the degree to which genetic relatedness serves as a reliable proxy for trait differentiation remains a key question. Within a species, genetic relatedness may not be tightly correlated with trait differentiation, especially where there is strong selection on ecologically relevant traits. In these cases, phenotypic differentiation often exceeds what might be predicted by genetic distance (McKay & Latta, 2002;Reed & Frankham, 2001). Intraspecific trait differentiation may also be correlated with genetic distance for some traits but not others, depending on the extent to which particular traits contribute to reproductive isolation (Wang & Summers, 2010). Similarly, drift-based models of trait change suggest that the relationship between genetic distance and trait differentiation might be wedge-shaped or saturating rather than linear (Cadotte et al., 2017). Furthermore, different metrics of genetic variation appear to have different effects on the outcome of ecological interactions (Abbott et al. 2017;Hanley, Hughes, Williams, Garland & Kimbro, 2016;Jousset et al., 2011), clouding the mechanistic interpretation of the effects of genetic variation in ecology.
In this study, we ask whether genetic relatedness can serve as a proxy for integrated trait differentiation within a species, at both the level of an individual and a subpopulation, using eelgrass, Zostera marina. Eelgrass is a foundation species that forms extensive meadows in bays and estuaries throughout the northern hemisphere, where it provides critical habitat for fishes and invertebrates, while buffering shorelines from erosion and playing a key role in nutrient cycling ( Figure 1; Williams & Heck, 2001). Eelgrass reproduces sexually, as well as vegetatively, and genotypic diversity varies at scales of meters (1-15 genetically distinct individuals m −2 (hereafter "genotypes") in Northern California, , and among sites and tidal heights (Kamel, Hughes, Grosberg & Stachowicz, 2012;Olsen et al., 2004;Ort, Cohen & Boyer, 2012). Multiple genotypes potentially interact at an even finer scale, because as many as four unique genotypes can grow highly intertwined in a 10 cm by 10 cm area (J. Abbott and J. Stachowicz, unpublished data). Previous work shows that eelgrass genotypes differ in traits such as individual growth rate, nutrient uptake, susceptibility to herbivores, and detrital production (Hughes, Stachowicz & Williams, 2009;Tomas et al., 2011), and that these trait differences can predict assemblage performance . Although much of the differentiation in traits observed in eelgrass across tidal heights is due to phenotypic plasticity (Dennison & Alberte, 1986;Li, Kim, Kim, Kim & Lee, 2013), there can be genetic differentiation in eelgrass growing at different depths (Kim et al., 2017;Ort et al., 2012), suggesting that there may be genetically based trait variation across depths.

CA. Photograph by Jessica Abbott
Our previous analyses of the covariance between genetic relatedness and trait differentiation in eelgrass suggested that the two were uncorrelated (Abbott & Stachowicz, 2016) or positively correlated, with more closely related individuals having more divergent traits . However, these analyses used only six to eight genotypes collected from three sites at similar tidal heights and covered a narrow range of pairwise relatedness. The main goal of previous studies was to test the influence of genotypic richness or relatedness on intraspecific interactions, rather than to assess rigorously the relationship between genetic relatedness and trait differentiation. A larger, spatially stratified sample of genotypes provides a more rigorous test, while also allowing us to test the level of genetic and functional differentiation at the subpopulation level. Unfortunately, despite differentiation in important physiological and growth traits, these genotypes are not readily distinguishable morphologically, highlighting the importance of developing genetic proxies for ecologically relevant trait differentiation. Here, we quantify the genetic relatedness of 40 eelgrass genotypes collected from different sites and tidal heights within Bodega Harbor, CA, and measure a range of phenotypic traits in these individuals related to resource acquisition, morphology, growth rate, and competitive ability. We use these data to assess how trait variation is distributed among genotypes, sites, and tidal heights, and whether measures of genetic distance (both at the individual and subpopulation/site level) are reliable proxies for trait differentiation that could ultimately be used to predict ecosystem functioning better than simple genotypic diversity.

| Genotype collection
In May 2012, we collected 20 eelgrass ramets harvested at 2-m intervals along a 40-m transect at each of three tidal heights (high intertidal, low intertidal, and subtidal) at four sites within Bodega Harbor, CA. We collected an additional 20 ramets along a 40-m transect from small (<4 m 2 ) patches of eelgrass near the entrance to the harbor, which are all subtidal (see Supporting information Appendix A, for a map and GPS coordinates of collection locations). Bodega Harbor is a shallow harbor located 64 km northwest of San Francisco (38°19′25″N, 123°02′52″W). The harbor is about 5 km 2 in area, much of which is mudflats at or above mean lower low water (MLLW) (NOAA Nautical Chart 18643, "Bodega and Tomales Bays"). Eelgrass grows in Bodega Harbor from about 0.25 m above MLLW in the intertidal to 3 m below MLLW in the deepest parts of the harbor adjacent to a dredged channel. The five eelgrass collection sites are distributed throughout the harbor, between 0.45 and 3.2 km apart. We transported the 260 eelgrass ramets collected from these sites to the Bodega Marine Laboratory (~2 to 4 km), where we trimmed the ramets to a single shoot with 3 cm of rhizome and 30 cm of leaf length and planted them in 11.4 cm diameter by 9.5-cm high-plastic flowerpots. We placed all pots in a single common garden flow-through seawater tank; we randomly assigned the pots to an initial position and rotated the pot position weekly. We collected leaf clips from each ramet for genetic analysis.

| Genetic analysis
We delineated genotypes and estimated relatedness using 11 microsatellite loci selected from a pool of >30 loci developed specifically for Z. marina (Abbott & Stachowicz, 2016;Oetjen, Ferber, Dankert & Reusch, 2010;Oetjen & Reusch, 2007;Reusch, 2000;Reusch, Stam & Olsen, 1999). We identified a total of 219 unique genotypes from the 260 ramets we collected. We estimated the relatedness of all genotype pairs with a regression-based measure of the number of shared alleles, calibrated by the frequency of those alleles in the population, using the program STORM (Frasier, 2008). We estimated relatedness both using all genotypes from Bodega Harbor as a whole (allele frequency based on all genotypes and relatedness of all possible pairs estimated) and with genotypes from different subpopulations (based on F ST ) independently (allele frequency determined for each subpopulation separately and relatedness between pairs within subpopulations estimated).

| Population structure
To estimate the degree of genetic structure among sites and among tidal heights, we calculated Weir & Cockerham's F-statistics (F ST ) using ARLEQUIN 3.5.1.3 and tested for significance by 10,100 permutations of the data (Excoffier, Laval & Schneider, 2005). We used all 219 unique genotypes originally collected from the different sites in Bodega Harbor for these analyses, not just the 40 on which we measured traits.

| Trait measurements
From the pool of 219 unique genets, we selected 40 genotypes to ensure that we included (a) a wide range of pairwise relatedness values and (b) genotypes from all tidal heights and sites (see Supporting information Appendix B, for multilocus genotypes and site/tidal height information). We transplanted these 40 genotypes into 3.79-L plastic flowerpots and grew them in an outdoor common garden flow-through seawater tank for the duration of trait measurements. We rotated pot position weekly to avoid position effects.
The common garden tank was 4.5 m long and 1 m wide and held approximately 3,800 L of seawater. Seawater flowed into the tank via 10 inflow valves that were distributed along the length of the tank with a combined seawater flow rate of approximately 16 L/min. We then measured a range of performance and resource acquisition traits for these 40 genotypes. We measured these traits only on new shoots produced while the plant was in the common garden, and not until shoots had acclimated to the common garden conditions for at least 10 months and had produced a minimum of three new shoots. Figure 2 summarizes the experimental design and timeline for genotype collection, propagation, and trait measurements.

| Morphology and production
We harvested five shoots of each genotype from the primary common garden, and standardized each to a common module size of 3 cm of rhizome and 30-cm shoot length, then planted them individually in 11.4 cm diameter by 9.5-cm high-plastic flowerpots. We did not standardize other traits within each module such as the initial number of leaves per shoot, shoot width, rhizome diameter, or maximum root length because these would have required significant damage to the modules that would have likely affected their survival and performance. The pots were placed in a second identical outdoor common garden in May 2014 at the beginning of the growing season. We started a second common garden so that we could retain the original genotypes in the primary common garden, while also destructively harvesting shoots for growth and morphological measurements. After a growth period of 10 weeks, we harvested the plants and measured shoot width and length, number of leaves, total rhizome length, rhizome diameter, maximum root length, new shoots produced, and the biomasses of roots, rhizomes, new shoots, and the terminal shoots. Because the transplanted modules used in the experiment came from plants which had been growing in a common garden for 2 years prior to the transplant, any differences among modules belonging to different genotypes are unlikely to be driven by differences in the environment in which they were raised.
One week prior to harvesting plants, we punched holes in terminal shoots to measure leaf growth rate (Williams & Ruckelshaus, 1993).

| Nutrient uptake rate
We measured biomass-specific leaf nitrate uptake and root/rhizome ammonium uptake rates using two-compartment chambers similar to the design in Terrados and Williams (1997). Eelgrass plants can absorb multiple forms of nitrogen in all their tissues; however, because nitrate is most available in the water column and ammonium in the sediments we measured nitrate uptake in leaf shoots and ammonium uptake in roots/rhizomes at ambient concentrations.
We collected eelgrass shoots from the primary common garden of 40 genotypes, cleaned them of all sediment, epiphytes, and invertebrates, and cut their rhizomes to 3 cm the day prior to allow wound healing. The roots and rhizomes of each shoot were compartmentalized from the leaf shoots by inserting the shoot through a slit in a watertight rubber stopper that was then inserted into a 40-ml opaque plastic chamber filled with 35 ml of nitrogen-free artificial F I G U R E 2 Diagram summarizing the work flow of the experiment from genotype collection to trait measurements. Text boxes describe each step, and any traits measured during a step are listed in parentheses in the text Fluorometer May 2012: 260 eelgrass ramets collected from Bodega Harbor and planted in a common garden Aug. 2012: 40 eelgrass genotypes selected for trait measurements and transplanted to larger pots in a common garden Jul. -Sept. 2013: measured nutrient uptake rates using shoots from the common garden (nitrate and ammonium uptake rates) Feb. -Mar. 2014: measured photosynthetic performance of shoots growing in the common garden ( , , P s , and rETRmax) May 2014: initiated tenweek morphology and production trial in a second common garden (shoot width and length, number of leaves, rhizome length, rhizome diameter, maximum root length, new shoots produced, leaf growth rate, and the biomasses of roots, rhizomes, new shoots, and the terminal shoots) Summer 2013: dried and ground eelgrass shoots collected from the common garden for measurement of phenolic content (% dry mass phenols) seawater spiked with ammonium to 100 μM using a 5 M NH 4 -N stock solution prepared with ammonium sulfate. We then placed the root/rhizome chambers with shoots inserted into 2-L clear acrylic cylindrical chambers filled with 1 L of nitrogen-free artificial seawater spiked with nitrate to 40 μM using a 2 M NO 3 -N stock solution prepared with sodium nitrate. Sixteen acrylic chambers were seated in a water bath with water circulating through a chiller to keep chamber seawater temperature between 10 and 12°C (Bracken, Jones & Williams, 2011). To provide sufficient water flow to prevent mass transfer limitation of uptake, we attached submersible pumps to each chamber via inflow and outflow pipes. Full spectrum quartz halite lamps surrounding the water bath provided the chambers with photosynthesis-saturating light (~700 μmol photons m −2 s −1 ).
We took water samples from the root/rhizome chambers and shoot chambers prior to the start of the experiment, then sampled the shoot chambers every hour for 4 hr, at which time we detached the root/rhizome chambers from the shoot chamber, removed the shoots, and took a final sample. We analyzed the shoot chamber samples (nitrate) using a Lachat 8000 series flow injection autoanalyzer and root/rhizome chamber samples (ammonium) using a Beckman Coulter DU640 spectrophotometer (Koroleff, 1976). After removing plants from the chambers, we divided them into shoots, roots, and rhizomes, and dried them at 60°C for at least 48 hr and then weighed each to obtain biomass-specific uptake rates.
We ran uptake trials with between 10 and 14 genotypes per day and measured all genotypes each week for 9 weeks. Within a week, we randomly assigned eelgrass genotypes a day and position in the water bath. Some genotypes did not yield enough shoots in the common garden for the full set of nine replicates (one genotype had only enough shoots for three replicates, but most had seven to nine successful replicates).

| Leaf phenolic content
Leaf phenolic content may be correlated with herbivore feeding preference on eelgrass (Buchsbaum, Valiela & Swain, 1984;Tomas et al., 2011;Vergés, Becerro, Alcoverro & Romero, 2007). We analyzed total phenolic content using approximately 4 mg of dried, ground leaf material from each genotype (pooled from three leaves) following a modified Folin-Ciocalteu method (see Bolser, Hay, Lindquist, Fenical & Wilson, 1998). We extracted phenolics with 2 ml of 80% methanol for 24 hr, and then quantified them with a spectrophotometer using caffeic acid as a standard. Ferulic and caffeic acids are two of the most abundant phenolics in Z. marina (Quackenbush, Bunn & Lingren, 1986;Vergeer & Develi, 1997), and previous work showed that caffeic, ferulic, or gallic acids standards for eelgrass phenolic content from shoots collected in Bodega Bay produced similar results (Tomas et al., 2011).

| Photosynthetic rate
We evaluated the photosynthetic performance of each genotype using a Diving-PAM ® (Pulse Amplitude Modulated) fluorometer (Walz, Germany) to measure maximum quantum yield (potential photosynthetic efficiency, F V /F m ) and rapid light curves (RLC), which determine the effective quantum yield as a function of irradiance and can be used to assess light adaptation (Ralph & Gademann, 2005;Williams, Carranza, Kunzelman, Seema & Kuivila, 2009). First, we dark-acclimated the outer leaves of each shoot for 30 min by placing a Waltz 4 mm opaque leaf clip 20 cm from the sediment surface on a leaf cleaned of epiphytes, then we immediately took maximum quantum yield and rapid light curve (RLC) measurements. RLCs comprised eight incremental steps of actinic light irradiance from 30 to 1,129 PAR (μmol photons m −2 s −1 ), and the resulting yield measurements were converted into a relative electron transport rate (rETR) using the following equation: where ΔF∕F � m is the effective quantum yield, ΔF is the difference between background fluorescence F and F at each PAR increment, F ′ m is the maximum fluorescence, 0.5 assumes that photons absorbed are equally distributed between photosystems I and II (Genty, Briantais & Baker, 1989), and AF is the standard absorption factor (0.55) for seagrasses (Durako, 2007).
To compare RLCs among genotypes we used curve fitting methods outlined in Ralph and Gademann (2005) to estimate characteristic parameters for each curve including: α, initial slope of the curve (rate of increase in photosynthesis with increasing light in light-limited region of the RLC or photosynthetic efficiency); β, slope of the curve where yield declines (strength of photoinhibition); and P s , which is a scaling factor used to determine the maximum relative electron transport rate (rETRmax). We fit each curve to a double exponential decay function (Platt, Gallegos & Harrison, 1980) using the "nls" function in the stats package in R 3.0.2.

| Trait distance metric
Using the data from our trait measurements, we calculated trait distances between all possible genotype pairs using standard methods (e.g., Petchey & Gaston, 2002). We used the data for the 17 traits that varied significantly between genotypes (all traits except for number of leaves and some photosynthesis parameters) to create a trait matrix in which trait values were standardized to have a mean = 0 and variance = 1. We then used the R "dist" function (R 3.0.3) to produce a Euclidean distance matrix of the multivariate trait distances between all genotype pairs for the 17 traits (hereafter referred to as "multivariate trait distance"). We did this for all pairs of genotypes within Bodega Harbor as a whole, as well as separately for genotype pairs within subpopulations, which were defined a priori by a sampling site being significantly differentiated from other sites based on F ST . As a second metric, we used a principal components analysis (PCA) to account for correlations among traits, and used the principal component scores for PC1 and PC2 to calculate the Euclidean distance between all pairs of genotypes (individual PC scores), and among genotypes from different sites or tidal heights (mean PC scores) in two-dimensional trait space rETR = ΔF∕F � m * PAR * 0.5 * AF, (hereafter "PCA trait distance"). We performed the PCA using scaled and centered (mean = 0 and variance = 1) trait values and used a correlation matrix to calculate principal components. The analysis was performed in R 3.0.3 using the prcomp function, which conducts the PCA using singular value decomposition of the data matrix. Separately, we assessed correlations among traits using pairwise regression (lm function from the stats package in R 3.3.3; R Core Team 2017).

| Statistical analysis
We tested for variation among genotypes in each of the measured traits using separate ANOVAs in the "car" package in R. We analyzed the relationship between genotypic pairwise relatedness and trait

| Trait variation among genotypes
Seventeen of the 21 traits differed among the 40 genotypes we  (Table 1).
We also analyzed trait differences as a function of the sites and tidal heights from which genotypes were collected to assess whether genetically based trait variation measured in a common garden was nonrandomly distributed in the field (i.e., evidence for local adaptation or environmental filtering). Out of the 21 traits, we found that

| Pairwise trait distance versus relatedness
The

| Trait versus genetic differentiation across sites and tidal heights
Using the position in trait space for each genotype represented in  (Table 3).

| D ISCUSS I ON
A growing number of studies use genetic dissimilarity as a proxy for ecological differentiation (e.g., Cadotte et al., 2008;Stachowicz et al., 2013;Violle et al., 2011), often assuming a positive correlation between the two, yet there are few critical tests of the shape or strength of this relationship. We assessed whether neutral genetic differentiation among individuals and subpopulations of the eelgrass, Z. marina, at 11 microsatellite loci predicts their ecological differentiation with respect to a number of traits associated with resource acquisition and growth strategy. We measured these traits in a common outdoor tank, and found considerable variation in traits among eelgrass genotypes that was distributed among individuals, sites, and tidal heights. Genetic relatedness was uncorrelated with either differentiation in particular traits or multivariate trait distance indices. However, the magnitude of genetic differentiation at the site level did predict site-level trait differentiation, although our sample size was relatively small (five sites). Thus, the correlation between genetic relatedness and trait differentiation among eelgrass genotypes appears to be scale dependent, just as has been found for correlations between phylogenetic distance and trait distance among species (Cavender-Bares, Keen & Miles, 2006;Peay, Belisle & Fukami, 2011).
The lack of correlation between the genetic relatedness and phenotypic trait distance between two individuals corroborates our findings from a much smaller sample of six genotypes (Abbott & Stachowicz, 2016). Selection intensity and rates of evolutionary pairwise genetic relatedness at a particular set of loci and overall phenotypic similarity. As genomic methods become more accessible, genomewide measures of genetic distance may prove to be better proxies for trait differentiation (Cadotte et al., 2017), but such data are as yet not available for eelgrass.
At larger scales, we found greater trait differentiation across sites than across tidal heights for traits measured in a common garden. The lack of consistent differences among plants collected from different tidal heights in Bodega Harbor suggests that observed trait differences across tidal heights in the field are primarily plastic responses to environmental conditions (see also Dennison & Alberte, 1986;Li et al., 2013). When measured in a common garden, the only trait that differed significantly among plants collected across tidal heights was rooting depth, with the deepest roots in the high intertidal, potentially an adaptation that enhances anchoring in the higher wave action of the intertidal. The lack of a correlation between genetic differentiation or relatedness and trait differentiation across tidal heights is not surprising given the limited genetic and phenotypic differentiation among tidal heights (Table 3). Other studies have found significant genetic differentiation across tidal heights in Z. marina at other locations (Kim et al., 2017;Ort et al., 2012), raising the possibility that genetically based trait differences in eelgrass from different tidal heights might occur in other populations.
The relationship between phenotypic and genetic differentiation that we observed at the site (or subpopulation) level could be explained by a combination of selection and genetic drift.
Environmental differences across sites could create varying   (Losos, 2008), complicating the use of genetic proxies for overall differentiation in ecologically relevant traits (Cadotte et al., 2017). Thus, the predictive value of genetic relatedness for trait differentiation, both within and among species, likely only applies over a restricted range of genetic distances, which may vary among species.
Predicting pairwise trait differentiation among specific individuals remains an elusive challenge, despite the importance of such data for understanding competitive interactions and ecosystem functioning. Genetic data from a broader sample of the genome (e.g., SNPs) or at loci that actually control the measured traits could provide a stronger genetic proxy for trait differentiation at the amongindividual scale, but such data are currently unavailable for eelgrass.
Independent of its relationship to traits, genetic relatedness in eelgrass influences the outcome of interactions, possibly as an indicator of the intensity of kin interactions (Abbott et al., 2017). Ultimately, as for interspecific comparisons (Cadotte, Albert & Walker, 2013), genetic and trait differentiation may provide complementary information about ecological interactions and outcomes (Jousset et al., 2011;Abbott et al., 2017). Alternatively, the strong positive correlations among many traits (Table 1) suggest that measuring relatively few simple traits might be a simpler proxy for overall trait differentiation. A better understanding of the mechanistic reason for these correlations would help assess the extent to which this is possible (Peiman & Robinson, 2017).

ACK N OWLED G M ENTS
We thank Brenda Cameron, Stephanie Kamel, and Krista Viglienzoni who helped develop microsatellite markers and genotyped samples. We also thank Larson Ankeny, Erica Pollard, Danielle Kronk, and Olivia Rhoades for field assistance and Eliza Oldach and Josh Gevertz for their assistance in measuring eelgrass traits. Comments from the editor and two anonymous reviewers on a previous draft improved the manuscript. This work was supported by NSF grant OCE-12-34345 to J.J.S, R.K.G., and S.L.W.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
JMA, JJS, SLW, and RKG conceived and designed the study. JMA and KD collected data for the study with assistance from SLW. JMA, KD, and JJS analyzed the data. JMA lead the manuscript preparation. All authors contributed critically to the drafts and gave final approval for publication. Service Award for her efforts in marine conservation. Susan was a pioneer in marine biology, paving the way for women to follow in her footsteps. She was a caring and dedicated mentor who used her success to support and empower her students. In 2009, she was honored by the UC Davis Consortium for Women and Research as an outstanding mentor for championing inclusion and diversity in the sciences. Susan's passion was inspiring, and the tremendous impacts of her mentorship, research, and leadership will continue her legacy.