We obtained second-generation S. pennellii ILs (Eshed and Zamir, 1995), Solanum pennellii, and Solanum lycopersicum cv M82 seeds from the Tomato Genetics Resource Center and the lab of Neelima Sinha (University of California, Davis) and Dani Zamir (Hebrew University, Rehovot, Israel). All seeds were prepared and germinated at the Donald Danforth Plant Science Center in St. Louis, MO as described in Chitwood et al. (2013).

Each sample for lipid extraction consisted of five leaflets (terminal, distal lateral left and right, proximal lateral left and right) from a single leaf of each plant. We measured leaf area from all sample leaflets with a flatbed scanner and ImageJ software (Abràmoff et al., 2004). The collected leaf samples were cut into 1 cm² pieces, placed into pre-baked 15 mL clear borosilicate vials (Qorpak), and then dried in a 70°C oven for 48 h. We extracted epicuticular waxes from the dried leaf samples by adding 5 mL of hexane (Omni-Solv HR-GC Hexanes 98.5%, VWR International, LLC) and agitating via pumping with a Pasteur pipette. The resulting total lipid extract (TLE) was collected into a pre-baked 15 mL borosilicate vial; the extraction step was repeated three times, and the three extractions were pooled. We evaporated the TLE to dryness with heat (30°C) under a steady stream of nitrogen gas (FlexiVap Work Station, Glas-Col).

To isolate alkanes for analysis, we performed silica gel column chromatography on the dried TLE of each sample. We transferred the TLE with 50 μL of hexane to a silica gel column (5 cm × 4 mm Pasteur pipette packed at the base with a small amount of pre-baked glass wool, which was covered by a thin layer of pre-baked chromatography grade sand and then filled 3–4 cm high with H₂O-deactivated silica gel). We collected the alkane fraction by eluting with hexane. The polar compounds retained on the silica gel column were eluted with ethyl acetate and archived.

We analyzed the alkane fractions using an Agilent 7890A gas chomatograph (GC) coupled with a 5975C Series Mass Spectrometric Detector system at Washington University in St. Louis. The GC was equipped with an Agilent J&W HP-5 ms column (30 m long, 0.25 mm inner diameter, 0.25 μm film thickness). The GC oven had an initial temperature of 60°C and was heated at a rate of 6°C/min to a final temperature of 320°C, which was held for 20 min. One sample run lasted approximately 65 min. Alkanes were identified by their mass spectra and quantified against an internal standard (n-hexadecane-d₃₄, 98 atom%, Sigma-Aldrich).

Compound-specific isotope analyses were performed at Washington University in St. Louis using a GC coupled via a combustion reactor to a Thermo Delta V Plus isotope ratio mass spectrometer. Isotopic compositions were determined relative to an internal n-alkane standard (n-C₁₈ [−31.1‰]) and are reported as

where R_samp and R_stnd represent the ¹³C/¹²C abundance ratio of the sample and standard, respectively. Delta values are reported in per mil notation (‰) and are expressed relative to Vienna Pee Dee Belemnite (VPDB). All samples were analyzed in triplicate. Externally calibrated sets of standards with n-alkanes (mixture B3 or A5) and fatty acids (mixture F8), provided by A. Schimmelmann (Indiana University), were measured between every fifth sample injection. Analytical uncertainty of reported δ¹³C values ranges between ±0.2 and 0.3‰ (SEM), dependent on the number of analytical replicates, after propagating the uncertainty of replicate analyses and external molecular standards (Polissar and D'Andrea, 2014; data available on GitHub repository at http://github.com/aldbender/13C-heritability).

Ambient greenhouse CO₂ was monitored during growth of the Solanum ILs from December 19, 2013–February 24, 2014, except for the period from January 1–14 when a technical error prevented data collection. For the ILs and cv M82 plants grown simultaneously in the greenhouse, we report the apparent fractionation (ε) to characterize the difference in δ¹³C between source (i.e., atmospheric CO₂) and product (i.e., lipid), which is defined as,

where δ¹³C values are calculated as described in Equation 2. As with delta values, epsilon values are reported in per mil notation. Carbon isotope values of lipids from S. pennellii and cv M82 grown during November 2013 are not reported as ε values because the carbon isotopic value of ambient greenhouse CO₂ was not recorded during their growth period.

Data from all traits measured from the ILs are reported in Supplemental Dataset 3. The R code and data sets used for modeling are available on GitHub at http://github.com/aldbender/13C-heritability. Leaf wax traits measured from the ILs and cv M82 grown simultaneously were modeled using mixed-effect linear models with the lme4 packages (http://CRAN.R-project.org/package=lme4) in R (R Development Core Team, 2015). Before modeling, we compared the measured values against theoretical normally distributed values in a Q-Q plot to check whether the measured values came from a normally distributed population. If a trait deviated from a normal distribution, we transformed the trait by either taking the square root, log, reciprocal, or arcsine of the trait and tested for normality of each transformed population via the Shapiro-Wilk test, using the transformation that resulted in the least deviation from a normal distribution (see Supplemental Dataset 4 for measured trait summaries and selected transformation for each trait). After performing the mixed-effect linear modeling, the normal distribution of residuals in the model was verified. We extracted p-values for significant (p < 0.05) differences between ILs and cv M82 from the models using the pvals.fnc function from the language R package (http://CRAN.R-project.org/package=languageR); these p-values were used to generate the QTL analysis. Broad sense heritability (H²) is defined as the proportion of total phenotypic variance that can be attributed to genetic variation (Futuyma, 1998), as given in the equation:

where V_G is genetic variance and V_E is environmental variance. We determined broad-sense heritability values (H²) using the estimates of variance for genetic and environmental effects (here, this includes the block assignment of each plant and the variance of residual effects) that were derived from the mixed-effect linear modeling (see Supplemental Datasets 5 and 6). We categorized H² values as low (0.0–0.2), intermediate (0.2–0.4), or high (0.4–0.6); a trait with high heritability (closer to 1) generally indicates that genetic factors strongly influence the amount of trait variation.

Hierarchical clustering is used to build nested groupings of traits that are organized based on similarities between sets of trait observations. We performed a correlation analysis of leaf wax traits from this study with traits existing in the phenomics database (Phenom-Networks, www.phenome-networks.com). For each trait in a data set, data were z-score normalized in order to transform all data ranges to a standardized average and standard deviation. Z-scores were averaged across replicates. We performed hierarchical clustering using the hclust function from the stats package in R (R Development Core Team, 2015), clustering by the absolute value of the Pearson correlation coefficient using Ward's minimum variance method. We created a correlation matrix (Spearman) between all traits measured in this study (modeled as described above) and traits from other published studies, as described below. Significance values for correlations were determined and the false discovery rate controlled via the Benjamini-Hochberg method (Benjamini and Hochberg, 1995).

All traits analyzed for hierarchical clustering are divided into five major groups, as determined by the studies from which they are reported and by the phenotype that they measure, following the naming system described by Chitwood et al. (2013). “DEV” refer to leaf morphological and developmental traits as reported by Chitwood et al. (2013). “MET” traits report metabolite levels in the fruit pericarp, as measured by Schauer et al. (2006, 2008). Traits labeled “MOR” as recorded in Schauer et al. (2006, 2008) and Phenom-Networks include traits relevant to yield and morphological measures of fruits and flowers. The “ENZ” traits measure enzymatic activities in the fruit pericarp, as reported by Steinhauser et al. (2011), and “SEED” traits measure metabolite levels in seeds, as derived from Toubiana et al. (2012). Traits described in the present study are termed “WAX” because they relate to leaf waxes.

Odd-carbon-numbered n-alkanes in S. pennellii and cv M82 ranged from C₂₇ to C₃₅, with C₃₁ being the most abundant, followed by C₃₃ (Table 1; Figure 2). The prevalence of n-C₃₁ among the leaves of cv M82 and S. pennellii in this study matches that reported from the fruit cuticles of the same plants (Yeats et al., 2012).

Branched alkanes with methyl groups at the iso (second) and anteiso (third) positions are present in measurable quantities among both S. pennellii and cv M82. For each carbon chain length, the iso-alkane precedes the anteiso-alkane, which precedes the n-alkane. For example, for alkanes with 31 carbon molecules, the elution order of alkanes would be i-C₃₁, a-C₃₁, and then n-C₃₁. We compiled concentrations of high-molecular-weight iso-, anteiso-, and n-alkanes in order to quantify variations in alkane distributions. Standard deviations (1σ) for leaf wax traits from cv M82 and S. pennellii are provided in the text below and in Table 1.

We provide the reader with the range of each leaf wax trait value from all ILs in the text and in Table 1. A comprehensive database of leaf wax trait values measured from the ILs is available in Supplemental Dataset 4.

We modeled the suite of IL leaf wax trait values to estimate the influence of genetic factors over these traits in the population of the IL plants and cv M82. Broad-sense heritability values (H²; Table 1 and Figure 3) for the leaf wax traits ranged from low (H² = 0.13) to intermediate (0.39). The percentage of anteiso- (H² = 0.39) and iso-alkanes (0.37) were of intermediate heritability, as were the δ_n–δ_iso values for C₃₁ (0.39) and C₃₃ (0.32). The methylation index was also of intermediate heritability (H² = 0.35). Traditional structural traits were also of intermediate heritability: CPI for n-alkanes (H² = 0.31), anteiso-alkanes (0.27), and iso-alkanes (0.22), as well as ACL for n-alkanes (0.31), anteiso-alkanes (0.28), and iso-alkanes (0.26). The ε values of individual alkane molecules were of low heritability (H² = 0.19 for i-C₃₁, 0.18 for n-C₃₁, 0.18 for n-C₃₃, and 0.13 for i-C₃₃).

In order to detect QTL for each leaf wax trait, we determined which ILs had a trait value that was significantly different from the trait value of cv M82. We identified 156 QTL in which a distinction at a significance level of p < 0.05 was detected (Figure 4, Supplemental Table 1). Of these, 33 QTL were transgressive in the direction of S. pennellii (i.e., the trait value was outside of the range between cv M82 and S. pennellii in the direction of S. pennellii) and 30 were transgressive beyond cv M82. Some ILs have trait values outside of the range of the parents (i.e., are transgressive) due to epistasis, or non-linear interactions between genetic loci, which can lead to non-additive effects among the traits of the ILs. Epistasis manifests because of novel combinations of genes within the ILs. The remainder of the QTL had trait values that were intermediate between the bounds of the trait defined by the two parent strains. One trait, δ_n – δ_iso for C₃₃, had no significant QTL.

Solanum pennellii had a far greater methylation index and percentage of iso-alkanes than any IL (Table 1). All methylation traits were of intermediate heritability (Figure 3), indicating that these traits are moderately influenced by genetic controls. ILs 1-1-2, 1-1-3, 1-4, and 3-2 contributed the largest effects to the alkane methylation phenotypes, with QTL for both methylation indices and percentages of iso-alkanes (Figure 4; Supplemental Table 1). The variation among the QTL for methylation traits may be attributed to the variation induced by the genetic loci of the ILs. The correlations in the QTL observed for methylation and for other leaf wax traits (Figure 5; Supplemental Dataset 7) may be informative in generating hypotheses about the physiological function of methylated alkanes. For example, the negative correlations between all methylation traits (i.e., methylation index and percentages of iso- and anteiso-alkanes) with CPI of n-alkanes might indicate a physiological connection between a lower odd-over-even predominance of n-alkanes (i.e., greater concentration of even-numbered n-alkanes) and a higher concentration of methylated alkanes.

Numerous studies have revealed correlations between n-alkane chain lengths and climatic variables such as temperature and humidity, as well as to the plant sources of the n-alkanes (e.g., Bush and McInerney, 2013 and references therein). Given this correlation, one might expect ACLs to be different in plants that are adapted to different hydrological regimes (e.g., S. pennellii and cv M82) and for ACLs to be of low heritability. Instead, we observed nearly identical ACLs between S. pennellii and cv M82 (Table 1) and ACLs that are of intermediate heritability (Figure 3), suggesting a high degree of genetic control over their alkane chain-length distributions. Alkane distributions are potentially very conserved traits among these Solanum species. If this pattern generalizes to other groups, it suggests that variations in sedimentary ACLs may be due to the changes in plant assemblages rather than phenotypic plasticity of individual lineages.

The ε-values for the four primary alkanes in this study had low broad-sense heritability values (Figure 3). A previous study into the bulk carbon isotopic composition of Arabidopsis thaliana grown in controlled growth chambers measured high heritability for bulk leaf δ¹³C values (H² = 0.67; Easlon et al., 2014). The lower heritability found among δ¹³C of wax in this study may reflect more variable environmental conditions in the greenhouse relative to a growth chamber, more variable response to the greenhouse environment, or a biological difference between Arabidopsis and Solanum. It may also reflect that waxes are a more environmentally sensitive phenotype than bulk leaf δ¹³C. Broad-sense heritability is specific to the population and environment, thus the difference between these two results is not unexpected.

We observed that alkanes derived from S. pennellii were more depleted in ¹³C than those from cv M82 (Figure 2), with a consistent offset in δ¹³C values of about 3‰. The average daily δ¹³C_CO2 values within the greenhouse varied by 1.7‰ during the end of the growth period of the parent lines (December 19–January 3; Supplemental Figure 2, Supplemental Dataset 3). Our sampling procedure collected young leaf material simultaneously from all plants, minimizing potential differences imposed by timing of growth. The isotopic difference between the two parent strains is likely to derive from biological variables that are not yet quantified. Possibilities for this difference include timing of wax synthesis, differences in WUE, or other responses to light, water, or nutrients. Examination of differences among the ILs can be informative. For example, WUE relates to stomatal conductance (Farquhar et al., 1989; Easlon et al., 2014). However, we did not find any significant correlations between δ¹³C in this study and previous estimates of stomatal density (Chitwood et al., 2013).

Every plant in this study produced iso-alkanes that were enriched in ¹³C relative to n-alkanes, expressed here as δ_n–δ_iso. The enrichment of iso- over n-alkanes was treated as a quantitative trait, and was of intermediate heritability (H² = 0.38 for C₃₁ alkanes and H² = 0.32 for C₃₃ alkanes; Figure 3). The variability of isotopic enrichment of iso-alkanes was more heritable than the apparent fractionation. The stronger genetic control on the difference between alkanes suggests a mechanism that relates to biosynthesis. Reddy et al. (2000) noted no apparent differences in δ¹³C values between normal and branched alkanes in their study of four species of Micromeria. However, the enrichment pattern observed in this study is consistent with that reported by Grice et al. (2008), who recorded that iso-alkanes were enriched by 0–1.8‰ over n-alkanes in Nicotiana tabacum (tobacco) plants. This enrichment pattern is attributed to different biosynthetic precursors for iso- versus n-alkanes (i.e., valine for iso- and pyruvate for n-alkanes; Kolattukudy, 1969; Grice et al., 2008). Levels of valine and pyruvate have previously been measured from both the fruits (Schauer et al., 2006, 2008) and seeds (Toubiana et al., 2012) of the S. pennellii ILs; however, these traits do not significantly correlate with any WAX traits in this study.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.