Heterosis for Biomass-Related Traits in Interspecific Triploid Hybrids of Willow (Salix spp.)

Species hybridization is key for the improvement of shrub willow (Salix spp.) bioenergy crops because hybrids often display heterosis for yield. The development of high-yielding genotypes requires numerous broad attempts at hybridization followed by field evaluation and selection for stable performance. Selection of improved shrub willow varieties for use as a bioenergy crop involves evaluation of full-sib progeny in family-based selection trials. Improving the accuracy of evaluation through the use of components of yield would greatly improve the efficiency of selection. Heterosis for biomass yield in intra- and interspecific F1 and F2 shrub willow crosses, made between Salix sections and ploidy, was examined by utilizing a suite of morphological, physiological, and chemical composition traits collected over the course of 12 weeks in the greenhouse and over 3 years in the field. Triploid families generated from diploid S. viminalis and tetraploid S. miyabeana displayed the greatest levels of heterosis for harvestable biomass and biomass-related growth traits in the greenhouse and in the field. While intraspecific S. purpurea diploids exhibited low levels of heterosis for these traits, interspecific diploids produced moderate levels of heterosis in greenhouse experiments. Differences between greenhouse and field trial results can largely be explained by pest damage, which negatively impacted interspecific diploids. Heterosis for the traits that form the basis for biomass yield, including stem growth, foliar, and physiological traits, was quantified, and family-level differences are discussed.


Introduction
Shrub willows (Salix spp., Salicaceae) are vigorous woody perennials that can be harvested to provide feedstocks for biofuel and bioenergy production [1]. Commonly found as pioneer species and in riparian habitats, the range of Salix extends from the arctic plains to the subtropics, with more than 350 described species [2,3]. Salix spp. are dioecious outcrossing perennials that are mostly entomophilous with possible low proportions of anemophily [4,5]. As Salix is particularly amenable to wide hybridization, taxonomic characterization within the genus has been an enduring challenge for botanists and breeders alike [6]. In sympatric populations of Salix, members within the same section will often hybridize, which can generate mixed populations of both pure species and species hybrids [7]. In addition, vegetative clonal propagation can be a significant contributor to population structure in Salix and has been shown to be fairly common in naturalized stands of North American S. purpurea [8]. Beyond the tremendous ecological amplitude, the heterogeneity and adaptive plasticity of Salix delivers a prodigious source of germplasm for genetic improvement.
Modern domestication of Salix traces back to the Swedish geneticist Nils Heribert-Nilsson's early cytological studies of S. viminalis × S. caprea hybrids in the 1920s [9]. Shortly thereafter, willow conservation and breeding were principally led by H.P. Hutchinson and K.G. Scott for nearly 30 years at the Long Ashton Experiment Station in the UK. Since the 1970s, breeders have maintained a goal of producing fast-growing shrub willow bioenergy feedstock cultivars that are high-yielding, genetically diverse, resistant to pests and diseases, and amenable to marginal sites, without competing with food crops [10]. A thorough review of willow botany and breeding can be found in [11].
While most complex quantitative traits, including yield, generally display additive inheritance, deviations from the midparent value in the F 1 can result in either hybrid vigor (heterosis) or hybrid necrosis [12]. Since the Green Revolution, the phenomenon of heterosis has been exploited in crop systems, more than doubling global commodity yields in only a few decades. Plants expressing heterosis are thought to have experienced one or more duplication events in their past [13], and many crop plants are polyploid. There is no unifying model to predict heterosis in interspecific hybrid plants, while much of the body of literature has been centered on the comparison of hybrids generated from inbred or synthetic allopolyploids. Common in Salix, polyploidy is thought to be largely due to allopolyploidization events [14]. Polyploidy via chromosome doubling or wide hybridization can have a positive impact on the accumulation of biomass in hybrids compared to progenitors [15]. In most plant crops, aneuploidy generally corresponds with hybrid necrosis due to the negative impact of sub-optimal gene dosages on protein function and metabolic homeostasis [12].
In order to advance the adaptive capacity of the US agricultural and energy sectors to respond to climate change, plant breeders must develop regionally adapted and sustainable bioenergy crops displaying heterosis. Shrub willow has emerged as a highly sustainable bioenergy feedstock that can directly substitute for fossil fuels with great potential for yield increases through species hybridization. Through careful hybridization, phenotypic evaluation, and selection, substantial improvements have been made in shrub willow biomass yield [16] and lignocellulosic composition (quality) [17]. Species hybridization is a key component in the development of shrub willow bioenergy crops, as hybrids often display heterosis for yield [18]. Interspecific hybridization is also thought to be important in the generation of diverse secondary metabolites, and that selection may favor plants expressing novel profiles [19]. For instance, diploid interspecific shrub willow populations have been implicated in enhanced resistance to pests [20,21] and pathogens [22] and have been attributed to differential expression of metabolites [23,24], resistance genes [21], as well as overall increased vigor [25].
Although heterosis has been observed in intraspecific crosses [26], it is more pronounced in triploids derived from the hybridization of diploid and tetraploid parents [27]. Ever since Nilsson-Ehle (1936) first reported finding naturally occurring "gigas" triploid Populus tremula L. in the forests of Sweden [28], polyploids have garnered substantial interest in tree breeding and research. Previous shrub willow yield trials have shown that elite triploid hybrid cultivars produce greater biomass yield compared to diploids and exceed or are not significantly different from tetraploids [29]. In addition, there is evidence that triploids have the potential to produce more cellulose per unit area because of higher biomass yields and/or cellulose content [17]. However, little is known about the genetic mechanisms responsible for heterosis in triploid willow [30] because there has been no systematic survey of heterosis from triploids.
The inherent time and investment associated with cultivar development in woody perennial feedstocks require efficient population development and phenotypic evaluation methods to significantly increase gains in biomass yield. Evaluation of large populations of progeny will be necessary to make significant gains, which requires early screening to reduce the costs associated with long-term field trials. The families referenced in this study were generated from crosses made between parents of differing ploidy and taxonomic section within the genus Salix, involving diploid S. purpurea and tetraploid S. miyabeana in the Sect. Helix Dumont, and diploid S. viminalis in the Sect. Vimen (formerly Viminella Seringe).
The main objectives of this study were (1) to determine the extent to which eight intra-and interspecific shrub willow families exhibit heterosis for biomass-related traits and biomass yield in the greenhouse and the field over three years, (2) to investigate components of yield to identify those which will serve as important selection criteria to make future selection programs more efficient, and (3) to determine if there is concordance between trials to improve selection efficiency using shorter-term, lest costly greenhouse trials.

Population Development
A total of eight full-sib F 1 and F 2 families were generated from crosses between diploid and tetraploid parents representing three Salix species: S. purpurea Sect. Helix (2n = 2x = 38), S. viminalis Sect. Vimen (2n = 2x = 38), and S. miyabeana Sect. Helix (2n = 4x = 76). All parents were locally adapted to the photoperiod and climate. The full-sib F 1 S. purpurea family 82 was generated from a cross between female 94,006 and male 94,001, both collected from naturalized S. purpurea populations in upstate NY. Two F 1 offspring from this cross, female S. purpurea "Wolcott" (9882-41) and male S. purpurea "Fish Creek" (9882-34), were crossed to generate the full-sib F 2 S. purpurea family 317. The female S. purpurea genotype 94,006 was crossed with the male S. viminalis "Jorr" to generate the interspecific diploid family 407, and a cross between S. viminalis 07-MBG-5027 and the male S. purpurea genotype 94,001 generated the pseudo-reciprocal interspecific family 421. Female diploid genotypes 94,006 and 07-MBG-5027 were separately crossed with tetraploid S. miyabeana male 01-200-003 to generate the interspecific triploid families 415 and 423, respectively. Triploid family 430 was generated from a cross between the tetraploid S. miyabeana female 01-200-006 and diploid S. viminalis male "Jorr." Family 425 was meant to be S. miyabeana (01-200-006) × S. purpurea (94,001) to complete the reciprocal triploid cross, but after ploidy was determined (all tetraploid), and F 1 individuals and parents were genotyped, it was determined that tetraploid S. miyabeana "SX64" was the pollen donor due to pollen contamination. We did not include "SX64" in the field trial design because we learned of the pollen contamination after the trial was already planted. All family progeny individuals and their parents were planted as seedlings in nursery beds at Cornell AgriTech (Geneva, NY).

Greenhouse Design
Parent genotypes and randomly chosen progeny from the eight families described above were grown from stem cuttings (20 cm) in 12-L plastic pots with peat moss-based potting mix (Fafard, Agawam, MA) to evaluate growth traits under greenhouse conditions over the course of 12 weeks. Families consisted of 12 progeny individuals and their parents, for a total of 104 genotypes (Table 1). One exception is that the male parent SX64 of the intraspecific S. miyabeana family 425 was not included. The plot was defined as a single cutting planted in a pot, which was arranged in a randomized complete block design with four replicate blocks. Two blocks were located on benches in one greenhouse, with the other two blocks in an adjacent greenhouse set for identical growing conditions. Supplemental greenhouse lighting was provided on a 14-h day:10-h night regimen with a max daytime temperature of 26 °C and a nighttime temperature of 18 °C. Liquid fertilizer (Peter's 15-16-17 Peat-Lite Spe-cial®, Scott's, Marysville, OH) was applied weekly after week four according to manufacturer recommendations.

Field Design
The field trial was established in May 2014 at Cornell Agr-iTech (Geneva, NY) in a randomized complete block design with four replicate blocks of three-plant plots and evaluated for 3 years. To avoid edge effects, S. purpurea parent genotypes "Fish Creek" and 94,006 were planted as border rows along the east and west sides of the trial, respectively, and the north and south ends were buffered by a single row of genotype 94,006. Within-row spacing was 0.4 m, and spacing between rows was 1.82 m. The soil at the field site is Odessa silt loam with a depth to the water table of 25 to 45 cm. For additional site characteristics, see [30,31].

Determination of Ploidy Level
The DNA content (2C-value in pg) of family parents and progeny was determined by flow cytometric analysis using young leaf material harvested from actively growing shoots in greenhouse conditions. Analysis of 50 mg of mature leaf tissue from parental genotypes and selected progeny was performed at the Flow Cytometry and Imaging Core Laboratory at Virginia Mason Research Center in Seattle, WA. A minimum of four replicates of all samples was independently assessed using either the diploid S. purpurea female genotype 94,006 or the diploid S. purpurea male genotype 94,001, and the tetraploid S. miyabeana female genotype 01-200-006 or the tetraploid S. miyabeana male genotype 01-200-003 as internal standards. Diploid and tetraploid parent genotypes from multiple runs were averaged and then divided by the value of the check for that run. This factor was then multiplied by each sample value within the same run as the check. When a genotype was analyzed more than once, the pg 2C −1 values were averaged.

Greenhouse Phenotypes
Starting approximately 7 days after planting (dap), the vegetative phenology stage (PHE) of each plot was scored at 7, 9, 11, 13, and 15 dap. Vegetative phenology was scored as six stages described as follows: stage (0) dormant axillary buds are tightly closed and covered by bud scales; (1) axillary buds begin to swell and change color; (2) generative bud burst with visible leaves; (3) leaves emerge and begin to unfold; (4) unfolded leaves begin expanding; and (5) at least two leaves are fully expanded. Leaf area (LFA) was determined using a portable leaf area meter (Model No. CI-203, CID Inc., Camas, WA). A representative leaf from each plot was scanned, then excised, dried to constant weight at 65 °C, and weighed to obtain leaf dry weight (LFDW). Leaf petioles were excluded from LFA and LFDW measurements. Specific leaf area (SLA) was calculated as LFA/ LFDW (cm 2 g −1 ). Leaf aspect ratio (LFR) is the ratio of the leaf length to its maximum width. Leaf shape factor (LFF) or circularity was corrected so that the shape factor of a circle is equal to one: 4π × LFA/LFP 2 . To provide an indirect measurement of leaf chlorophyll and nitrogen content, leaf color was measured using a Minolta SPAD 502 Chlorophyll Meter (Spectrum Technologies, Inc., Aurora, IL). Four fully expanded leaves sampled from the upper 25% of the canopy were measured and averaged for each plant at 14, 42, and 70 dap. Primary stems were defined as those emerging from dormant axillary buds and ≥ 6 cm in length. Secondary stems were defined as emerging from axillary buds on the primary stem on current-season growth (sylleptic) and were counted as the total number of branches within each plot with a PHE ≥ 3. The length of each stem per plant was measured from the proximal base of the primary stem to the distal inner-whorl of the leaf primordia. The sum of stem lengths for each plot was considered to be the total stem length (TSL), and mean stem length (MSL) per plant was the mean of individual stem lengths. Starting 14 to 70 dap, all primary stems were measured within each plot once a week, totaling nine measurement time points. The diameter of each primary stem within a plot (SDIA) was measured at the base the final week of the study using a digital caliper. Stem diameter measurements were used to calculate the cross-sectional area (SA) of all stems > 20 cm in length. The sum of stem area per plant (SSA) was calculated by summing individual SA per plant. Total stem volume (VOL) was estimated by considering conical stems: SSA × HT/3.
Root biomass was harvested from a subset of 20 pots (2-3 progeny individuals per family) that were selected from the 416 pots as representing a distribution of capacitance readings, ranging from 70.5 to 283.8 nF. In order to predict root dry biomass (RDW), root electrical capacitance (REC) was measured according to the protocol described in [32].
Stem and leaf biomass was harvested separately from each plot, dried in an oven to constant weight at 65 °C, and then weighed to determine total stem dry weight (SDW) and total leaf dry weight (LDW), which sum to aboveground dry biomass (AGB).

Field Phenotypes
During the dormant period after each growing season, DIAs of stems ≥ 0.5 cm were measured at 30 cm from the base of the plant using Masser Racal 500 digital caliper (Masser, Rovaniemi, Finland), and stem number (PSN) was counted for each plant. The SA of each stem was calculated from DIA, and SSA per plant was calculated by summing the areas of every stem. The maximum stem height (HT) of every plot was recorded using a measuring rod (Crain Enterprises, Inc., Mound City, IL).
Physical and chemical wood properties were measured for four replicates of each genotype. Stem segment samples (15.24 cm) were collected in the dormant period after each growing season using sampling methods previously described (Liu et al., 2015) and were stored frozen at − 4 °C until they were processed. The specific gravity of each sample was measured by volumetric displacement (TST om-06, 2006), where the volume of water displaced of submerged samples was weighed to determine stem density (DEN). Following specific gravity determination, stem segments were oven-dried at 65 °C to a constant weight and then roughly milled to a 5-mm particle size with a Retch SM300 cutting mill (Retch, Haa, Germany) and were further comminuted to < 0.5 mm particle size by fine milling with the IKA MF 10.1 knife mill (IKA, Wilmington, NC) for compositional analysis. Approximately 20 mg of each milled biological replicate was analyzed with a Thermogravimetric Analyzer (TGA) Q500 instrument and Universal Analysis 2000 version 4.5A software (TA Instruments, New Castle, DE), as previously described (Serapiglia et al., 2009). Hemicellulose (HCL), cellulose (CLS), lignin (LIG), and ash (ASH) contents were determined as a percentage of total dry biomass for each sample, as previously described in [30].
The dry weight, diameter, and specific area of stem segments sampled for wood composition were measured and referenced as SiDW, SiDIA, and SiSpSA, respectively. At the end of the second growing season, crown diameter (CDIA) was measured using modified Haglöf Mantax forestry calipers (Haglöf Sweden AB, Långsele, Sweden). Stool diameters were measured at 15.24 cm above the soil, which is the average cutting height of a shrub willow harvester. Crown form (FORM) was calculated by multiplying the arc-tangent2 of one-half CDIA and the fixed distance at which CDIA was measured (15.24 cm) by 180/π to obtain the angle of the stem branching relative to the soil.
The sex of progeny individuals was determined by visual observation after forcing 2-3 dormant shoots to flower in greenhouse conditions and confirmed in field trial plots.

Statistical Analysis
All statistical analyses and plotting were performed within the open-source statistical computing environment, R [33].
For quantitative traits listed in Table 2, a Shapiro-Wilk test was conducted to detect a significant departure from normality. For non-normal data, the boxcox function was used to maximize the Shapiro-Wilk W statistic by computing log-likelihoods for the parameter (λ) of the Box-Cox power transformation, such that either a single-parameter y − 1 ∕ or two-parameter y − 2 1 − 1 ∕ 1 power transformation was applied. For repeated measurements of quantitative traits, HT, MSL, and TSL, growth rates were determined using Gompertz 3-parameter function: ce −e −a(t−b) , whereas ordinal PHE and PSN growth rates were determined using the following 3-parameter logistic function: c∕ 1 + e −a(t−b) , where a is the growth rate, b is the inflection point, c is the asymptote, and t is time (in dap).
Tests for an association between quantitative traits were done using Pearson's product-moment correlation coefficient (r) at a confidence level of 95%. Correlations between ordinal and quantitative trait pairs were tested using Spearman's rank correlation coefficient, whereas Kendall's rank correlation was used to test ordinal trait pairs. To correct for multiple comparisons, a Bonferroni correction was applied by multiplying P-values by the number of pairwise comparisons. Wilcoxon rank-sum test was used to test whether two sample distributions differed by ploidy and sex, and family-level comparisons were performed using Tukey's range test.
Variance components for the greenhouse trial were estimated with lmer in the package lme4 [34] using the restricted maximum likelihood (REML) method for the following model: where y ijk is the observed value, is the overall mean, i is the effect of genotype i, j is the effect of block j, and ij is the random error, which is assumed independent and identically distributed. Field trial dimensions were 388.6 m × 36.6 m ( Supplementary Fig. S1), which introduced spatial variation not easily accountable by block alone. Thus, to account for spatial variation in the field trial, following the approach outlined in [35], spatial trends (row and column) in the field trial were modeled as two-dimensional Penalized (P)splines, using SpATS and SAP functions (n.seg = (16, 64), tolerance = 1 × 10 -6 ) in the SpATS package [36].
Midparent heterosis (MPH) was calculated as the percent deviation of the F 1 progeny mean relative to the mean of the parents. In this case, genotype (clone) was fixed in the linear model, so MPH represents the average deviation from these estimates. To compare family parent deviations in each trial, field-and greenhouse-collected traits were categorized into modes of inheritance by comparing the deviation of the mean values of the F 1 to the respective female (P1) and male (P2) parents, using an absolute deviation assignment threshold of > 25%. For the reason that the male parent "SX64" of family 425 was not present, only the deviation of the F 1 from the female parent 01-200-006 is discussed.

Harvestable Biomass in the Greenhouse Trial
Of the 104 genotypes harvested in this trial, total aboveground biomass (AGB) ranged from 119.6 to 51.1 g (Fig. 1a, b). The greatest yielding genotype was a triploid hybrid, S. viminalis × S. miyabeana 12X-423-043, and the least yielding genotype was a diploid S. purpurea, 10X-082-078. The greatest mean AGB was from the S. viminalis × S. miyabeana triploid family 423 (100.3 ± 3.8) and the S. miyabeana × S. viminalis triploid family 430 (98.7 ± 3.4), followed by the S. viminalis × S. purpurea diploid family 421 (92.3 ± 3.8). All other families were not significantly different from one another, with a family mean AGB ranging from 78.9 to 82.5 g.

Biomass-Related Stem Growth in the Greenhouse
Growth measurements taken at the end of the study included the length and diameter of every stem, as well as primary and axial stem number. The greatest plot HT after 84 dap in the greenhouse was observed for the triploid hybrid, S. viminalis × S. miyabeana 12X-423-034 (2.1 ± 0.05), whereas the genotype with the lowest HT was the intraspecific tetraploid, S. miyabeana 12X-425-106 (1.2 ± 0.09). Total stem length (TSL) was calculated as the sum of the length of each shoot > 20 cm per plot and ranged from 6.03 to 1.95 m. While TSL of diploids (3.89 ± 0.08) and triploids (3.88 ± 0.10) was not significantly different (Wilcoxon P = 0.85) at 84 dap, tetraploids showed significantly less TSL (3.47 ± 0.13) than diploids (Wilcoxon P = 0.02) and triploids (Wilcoxon P = 0.01

Sex Ratio Bias and Sex Dimorphism
While the selection of individuals from each family was completely random, all families had a greater proportion of female progeny, of which five of the eight families were significantly sex-biased (χ 2 P < 0.05) ( Table 1). Female-to-male sex ratios ranged from 1.13 to 4.56. Notably, the triploid S. miyabeana × S. viminalis family 430 was all female. Sex ratio bias is not uncommon in willows (primarily femalebiased) [37][38][39][40][41][42][43] but is less frequently found in poplars (primarily male-biased) [44]. Whether influenced by environmental factors, unique mechanisms of sex determination, or sex chromosome dosage, the genetic basis of biased sex ratios in Salix reported here and elsewhere is not known.
While sex dimorphism in S. purpurea is well documented [45], this is the first report of sex dimorphic phenotypic expression in triploid Salix hybrids or tetraploid S. miyabeana. Nevertheless, larger populations may be required to validate these findings, as strong female bias could exaggerate sex differences if males are not well represented (e.g., 82 females versus 18 males in family 425). With regards to population improvement, modification of sex ratios or sexspecific selection in high-performing crosses could directly exploit sex linkage. Furthermore, genetic mapping within these interspecific crosses, as was done in intraspecific S. purpurea [31], would validate purported sex linkage.

Foliar and Physiological Traits
Vegetative phenology recorded at 7, 9, 11, and 13 dap demonstrate significantly faster budbreak, leaf expansion, and early shoot development in triploids, compared with diploids and tetraploids (Fig. 3a). Diploids, while significantly lagging triploids and tetraploids early on, were not significantly different from tetraploids by 13 dap. For each plot, three SPAD readings were taken at 14, 42, and 70 dap to assess the chlorophyll content or nitrogen status of fully expanded leaves from the upper 30 cm of the canopy. A significant interaction was identified for SPAD by time and ploidy as well as time and family. While the initial SPAD reading at 14 dap showed marginally greater values for triploids and tetraploids, the opposite trend was found for later readings (Fig. 3b). Diploid genotypes showed significantly greater SPAD readings at 42 dap than triploids and tetraploids, which were not significantly different. By 70 dap, differences in SPAD readings by ploidy were not as great as those taken at 42 dap. This was also observed for SPAD readings taken in the field trial, where higher ploidy was inversely correlated with greater SPAD readings.
Given that all plots received the same fertilizer and application rate, it may be that the nitrogen status of leaves is concentrated to less leaf area in diploids, given the significantly greater SPAD values in diploids, compared with those of triploids and tetraploids. Furthermore, LFA and LDW of triploid F 1 individuals showed primarily additive inheritance, and on the basis of ploidy, triploids were intermediate to diploids and tetraploids for the same traits. Yet, under controlled environmental conditions, higher ploidy tended to result in greater leaf area and biomass, but a lower leaf nitrogen status. One possible explanation for this is that polyploid willows are more efficient in the production of low resource leaves [46]; perhaps by focusing available nutrient resources to a rapidly emerging canopy, rather than uniformly along the stem, as is likely the case in diploids. Triploid genotypes with an intermediate LDW:SDW ratio could indicate more efficient partitioning of photoassimilate from leaves to sink organs. While this attribute would surely stimulate the rapid accumulation of biomass, diploids could benefit in nutrient-scarce environments by sustaining growth rates, whereas higher ploidy levels would likely show an overall reduction in their growth rate.

Multivariate Analysis
Traits that were non-informative in biomass yield predictions, but correlated with informative predictors, may still prove to be of relative importance, particularly when assayed in additional environments or pedigrees. While many of the traits listed in Table 2 were strongly correlated with those important for biomass yield, some pairs tended to be more autocorrelated (Supplementary Fig. S3), as they were repeated measurements or components of the same trait. The greenhouse-collected traits showing the strongest correlations with SDW were VOL (r = 0.73), REC (r = 0.78), and LDW (r = 0.62), all grouping closely. Following these traits, later stem length measurements were most correlated to SDW, whereas PHE at 11 and 13 dap and ASN at 56 dap were most correlated with LDW. Although weakly, both SDW and LDW had inverse relationships with the inflection point of HT.
Phenological stages (PHE) recorded at 11 and 13 dap in the greenhouse were strongly correlated with SPAD at 14 dap, LFR, LFL, and TSL at 14 and 21 dap, and the growth rates of HT, MSL, TSL, but not STN. Early PHE measurements at 7 and 9 dap were positively correlated with 2C-value and LFR. At 13 and 15 dap, PHE was only weakly correlated with SPAD at 42 dap. All PHE measurements were inversely correlated with LFW, SLA, and the ratio STN:AGB. Leaf dimensions LFL, LFA, and LFP, as well as the biomass ratio LDW:SDW were strongly and positively Boxplot distributions depict the median and interquartile range (IQR ± 1.5) of (a) phenological stages (PHE) at 7, 9, 11, and 13 days after planting (dap) as well as (b) SPAD values by ploidy. Asterisks above or below boxplots of diploids (beige), triploids (cyan), and tetraploids (dark grey) denote significant differences at a Wilcoxon P < 0.05*, < 0.01**, and < 0.001***. The 15 dap stage is not shown because there were no significant differences by ploidy correlated (r > 0.5) with 2C-value, whereas both SPAD at 42 dap and SLA were inversely correlated (r < − 0.5). Besides showing positive correlations to the ratios SDW:TOT and STN:AGB, SLA was the solitary trait to be inversely correlated with nearly all growth traits. Along with SLA, SDW:TOT and STN:AGB were most negatively correlated with leaf dimensions, as well as SPAD at 42 dap and STN measurements. Root electrical capacitance (REC), SSA, TSL at 28 dap, and PHE at 11 and 13 dap were highly correlated with SDW and have been shown to account for a large proportion of the variance (R 2 = 0.69) in multiple linear regression [32].
Repeated measures have commonly been used to identify growth patterns in response to environmental factors or treatments. Here, the treatments under consideration were pedigree and ploidy level. While the inherent differences between factors tend to inflate the actual differences (e.g., diploids versus tetraploids), relative growth rates (RGR) act as a standardized measure of growth and offer more impartial comparisons. Early vegetative PHE measurements showed that triploids are faster to break bud and grow at faster rates compared with diploids and tetraploids. However, over time, diploids maintained more linear growth rates compared with those of triploids, which leveled off approximately 8 weeks after planting. This could be due to increasingly limited space in the pots of triploids, as they exhibited both greater above and below-ground biomass at the termination of the study, especially for individuals with S. viminalis as one of the parents, as described in [32]. It may be that S. viminalis crosses have a higher propensity for accumulating root mass than intraspecific or interspecific crosses of S. purpurea and S. miyabeana. Using the REC phenotyping method, genetic mapping of this trait could improve our understanding of root development and response to drought among willow crosses. Marker-assisted selection (MAS) for REC could subsequently be used to improve biomass yield in the case that field evaluations correlate well with those in the greenhouse.
For field-collected biomass growth traits, all were strongly correlated across years ( Supplementary Fig. S4). However, foliar traits were positively, but more weakly correlated between years, besides leaf ratio (LFR) (r = 0.65, p < 0.001) and leaf shape factor (LFF) (r = 0.35, p < 0.001). Crown form (FORM) measurements for all 3 years measured were most inversely correlated with biomass stem growth traits (e.g., SA, HT, VOL, and STN) as well as wood density (DEN). Wood chemical composition traits were also strongly correlated in the field trial ( Supplementary Fig. S5). Both LIG and ASH content were inversely correlated with CLS and HCL content. Wood density (DEN) was only positively correlated with CLS content, but inversely with HCL and ASH content. Overall, genotypes with higher ploidy levels tended to have greater ASH content, especially tetraploids (3.1 ± 0.1%), compared with both intra-and interspecific diploids (1.6-2.1%). However, the opposite trend was observed for HCL content. Not significantly different from one another, F 1 and F 2 S. purpurea families had the highest HCL content. Only tetraploid S. miyabeana family 425 had significantly lower mean CLS content (37.8%) compared to the other families, which did not significantly differ (39-40.2%).

Concordance of Heterosis for Common Greenhouse and Field Traits
Field-collected traits for the same families resulted in similar levels of heterosis for common traits collected in the greenhouse trial (Figs. 4 and 5). For instance, families 423 and 430 showed the greatest MPH for HT, SA, and VOL for both years in the field trial as well as in the greenhouse trial. While the interspecific diploid families 407 and 421 showed marginal levels of MPH for the same traits, it was not observed in the field trial as a result of a high incidence of potato leafhopper and Japanese beetle feeding on individuals with an S. viminalis background, also described in [21]. Yet, this was not observed for triploid families with a diploid S. viminalis parent, which suggests some level of resistance in the hybrids. Of the 2 years in the field trial, foliar traits were shown to be the most variable between years (Fig. 5, Supplementary Fig. S6). This was due to differences in precipitation during the 2015 and 2016 growing seasons. In 2015, growing conditions were nearly optimal throughout the season, but 2016 suffered a long stretch of mid-summer drought. Although individuals did not dramatically differ in MPH for most trait rankings between growing seasons, foliar trait MPH variation can be assessed by SLA. However, families 421, 423, and 430, which all had an S. viminalis parent, did show greater MPH for foliar traits in both the greenhouse and field trials. Remarkably, the triploid individuals, 12X-423-060 and 12X-423-110, had an average LFL > 40 cm, which was nearly two-fold greater than the better parent.

Midparent Heterosis and Patterns of Inheritance
Our data show that relative to diploid and tetraploid genotypes, triploids exhibit greater levels of heterosis for biomass yield and correlated growth traits, especially for crosses made between Salix sections (Fig. 4, Supplementary  Fig. S2). The interspecific diploid family 421 and interspecific triploid families 423 and 430 have S. viminalis as one of the parents, and all showed high levels of MPH (%) for total SDW and RDW, as well as RGR and early TSL and PHE measurements. However, MPH for total biomass or stem growth measurements were not observed in interspecific S.  The log 2 difference of the respective female and male parent from the family progeny was used to assess patterns of inheritance for stem growth and foliar traits in each family (Fig. 6). Diploid intraspecific S. purpurea families showed the most conserved inheritance for all greenhouse and field traits, with dominant and underdominant patterns most common for stem growth traits. Reciprocal interspecific diploid families 407 and 421 exhibited strong patterns of dominance, almost exclusively in the direction of the S. viminalis parent. In family 421, both greenhouse-and field-collected foliar traits deviated substantially from both parents. In addition, underdominant stem growth was only detected for field phenotypes. While stem growth traits in triploid family 415 primarily displayed both conserved and S. miyabeana (P2) dominant inheritance, most foliar and stem growth traits were conserved or additively inherited, reflected in Fig. 2.
Total harvestable biomass of triploid families 423 and 430 exceeded both midparent and better parent values (Fig. 2) and had the greatest number of stem growth traits with nonadditive inheritance. Family 423 showed strong S. viminalis (P1) dominant inheritance for greenhouse stem growth traits, comparable to diploid family 421, yet for stem growth traits collected in the field, S. miyabeana (P2) dominant inheritance was predominant. This conflict was also observed for the reciprocal S. miyabeana × S. viminalis triploid family 430, whereby dominant patterns of stem growth in the greenhouse were primarily in the direction of the S. viminalis parent, but in the direction of the S. miyabeana parent in the field. Stem density (DEN) in these two triploid families substantially deviated from Fig. 6 Family inheritance patterns of field and greenhouse traits. Each point in (a) represents a single trait plotted as the percent difference of the hybrid (F 1 ) mean value from respective female (P1) and male (P2) parent mean values. Greenhouse phenotypes are rep-resented by circles and field phenotypes, by triangles. Each point is colored according to the phenotypic group (see legend). Combined parent deviation plots (b) of greenhouse (GH) and field traits and respective family-level inheritance class proportions the S. viminalis parent and may explain the environmental effect on dominant phenotypic expression for stem growth traits. As noted, both S. viminalis parents are susceptible to potato leafhopper [21], which can dramatically defoliate and stunt plants, often killing apical meristems, leading to lower biomass quality and yield. Thus, dominant resistance to potato leafhopper in triploid hybrids, provided by the S. miyabeana parent, most likely accounts for contrasting dominance deviations of stem growth traits between studies.
The heterozygous nature of Salix prevents a genetically uniform F 1 , and both hybrid vigor and hybrid necrosis can be represented by siblings of the same cross. Here, we demonstrate that the mean trait values of triploid shrub willow hybrids derived from diploid and tetraploid parents of different species exhibit both dominant and transgressive phenotypic expression. Most notably, this was observed for biomass and stem growth traits for crosses between Salix sections Helix and Vetrix.

Conclusion
The results outlined here corroborate consistent findings that triploids produce greater levels of heterosis for biomass yield compared to that of diploids. Heterosis for many of the extensive biomass-related traits collected in the greenhouse also showed heterosis in the field, with consistently greater total stem volume (most associated with biomass yield) among triploid individuals. Most nonadditive traits in interspecific crosses displayed dominant patterns of inheritance, especially stem growth traits, yet the proportion of nonadditive field traits was markedly lower compared with those measured in the greenhouse. While crosses between Salix sections Helix and Vetrix outperformed their progenitors in this trial, additional evaluation of triploid families with diverse parent backgrounds would aid in broadly defining heterotic groups and benefit triploid cultivar development. The genetic basis of heterosis in willows is not well-understood, and further work on characterizing this phenomenon will support community efforts in build a toolkit for improving this sustainable, fast-growing bioenergy crop.