Genetic material.

A set of 28 roses consisting of eight elite roses from the Vineland Research and Innovation Centre’s (VRIC, Vineland, ON, Canada) breeding program, one breeding line (CA60), and 19 commercial rose cultivars that were a mixture of roses from the Explorer Series (George Vancouver, John Cabot, John Davis, John Franklin, Lambert Closse, Martin Frobisher, Quadra, William Baffin, William Booth), floribunda [KORmuse (Salmon Vigorosa), MACivy (Singin in the Rain), Poseidon (Novalis)], hybrid tea [MEInomad (Desert Peace), MEIpierar (Caroline de Monaco), Peace, WEKrigoyelo (Gentle Giant)], grandiflora [MEImouslin (Cardinal Song)], and shrub roses [BAIine (Yellow Submarine), RADrazz (Knock Out)], and that varied in their level of winterhardiness (Table 1) were used in this experiment. All the roses were grown on their own roots.

Table 1.

Genetic material used in the genetic mapping of electrolyte leakage and field winterhardiness of garden roses (Rosa ×hybrida).

View Table

Experimental design and growing conditions.

The plants were propagated from cuttings in early-Winter 2016 using foam strips designed for vegetative propagation of cuttings (Oasis strips; Oasis Grower Solutions, Kent, OH) in the mist house, potted into 7.5-L pots (22 cm height, 22 cm width) in a soilless potting mix (Sunshine Mix 1; Sun Gro Horticulture, Agawam, MA) and grown outdoors. Roses were manually watered daily and fertilized two to three times per week with 20N–3.4P–16.6K and 14N–0P–11.6K fertilizers (Plant-Prod; Master Plant-Prod Inc., Brampton, ON, Canada) at an electrical conductivity (EC) level in the range of 1.2 to 1.6 dS·m⁻¹.

The electrolyte leakage assay consisted of three main steps: 2 weeks’ acclimation of whole plant at 4 °C, freezing treatment of stem segments from acclimated plants, and measurement of conductivity before and after autoclaving of the stem segments. Expt. 1 was replicated between one and three times depending on the genotype (i.e., replications in time) using unique sets of clones for each replication (i.e., true biological replications) (Supplemental Table 1), and it was conducted from May to Aug. 2016.

In addition, 12 of the commercial rose cultivars used in electrolyte assays were planted in 2017 in the field at VRIC (lat. 4311'30.9″N, long. 7923'45.7″W; cultivated Gleyed Brunisolic Gray Brown Luvisol soil with sandy loam) using a completely random design (CRD) with three replications. The 12 cultivars were grown on the rootstock Rosa multiflora. Plants were spaced by 50 cm within rows, and the rows were spaced by 1.50 m. The eight elite roses were planted on their own roots according to CRD with six replications in 2013 in Saskatoon, SK, Canada, and Olds, AB, Canada, as part of the Pan Canadian testing, a network of growers and university partners across Canada who evaluate VRIC’s selected roses.

Phenotyping.

The experiment was conducted according to Dexter et al. (1930) with various modifications to the authors’ original protocol. After the acclimation period, rose stems from acclimated plants were collected, sprayed with type 1 water (Milli-Q; Merck KGaA, Darmstadt, Germany) to nucleate ice formation, and placed into sealable plastic bags. The bags were moved to a programmable freezer in which the temperature dropped from 0 to −20 °C, at a rate of −2.5 °C·h⁻¹. One stem per genotype was removed from the freezer at 0, −5, −10, −15, and −20 °C. The stem removed from the freezer was then cut into 10 1-cm-long stem sections and each section placed into individual 20 mL centrifuge tubes filled with type 1 water (i.e., technical replications). The tubes were shaken overnight at a relatively low speed in a benchtop orbital shaker before conductivity was measured and recorded as R₀ using a traditional EC meter with a probe (Oakton CON 450; Cole-Parmer Instrument Co., Montreal, QC, Canada). The tubes were then autoclaved for 1 h at 120 °C to promote cell rupture and shaken again overnight. Conductivity was measured again and recorded as R_t. The percentage of electrolyte leakage (EL) was calculated as follows: $\text{EL}\left(\text{\%}\right)=100\hspace{0.17em}\times \hspace{0.17em}\frac{Ro}{Rt}$. In addition, field winterhardiness was recorded on a scale from 0 to 5 (0 = no winter damage, 1 = damage on the tips of the canes only, 2 = damage down to the snow line, 3 = damage to the crown with good spring regrowth, 4 = damage to the crown with poor spring regrowth, 5 = dead plant) in Spring 2018 (VRIC) for the 12 commercial rose cultivars, and in the Spring of 2014 and 2015 (Alberta and Saskatchewan) for eight elite roses.

Data analysis.

The percentage of electrolyte leakage was averaged for up to three replications for each cultivar (Supplemental Table 1). The lethal temperature at which 50% of leakage occurred (LT50) was estimated using a logistic regression on electrolyte leakage data according to the formula: $\ln \left(\frac{L}{1-L}\right)=a\hspace{0.17em}+\hspace{0.17em}bX$, where L was the proportion of electrolyte leakage corrected for leakage not due to freezing (i.e., control) and X was the logarithm to base 10 of the temperature. Correlations among electrolyte leakage, LT50, field data, and USDA cold hardiness zone were computed using the Pearson method in the RStudio software (RStudio Team, Boston, MA).

Genetic material.

Two mapping populations and their parental genotypes grown on their own roots were used in this experiment (Table 1). The first mapping population was a tetraploid F₁ rose population created between 2015 and 2016 from the cross between the cold-hardy female parent CA60 and the cold-susceptible male parent Singin’ in the Rain (‘SITR’). The cold-hardy female CA60 originated from the Morden rose breeding program from a cross between RSM 104 and the cold-hardy Canadian Explorer rose ‘Frontenac’ (Ogilvie, 1993). The female parent RSM 104, also named 91/104–1, was a German selection resulting from a cross between CT50–9 (a chromosome doubled clone of 88/124–46, which was selected in 1987 and sawn in 1988) and the cultivar TANca (Caramba). The diploid genotype 88/124–46 is the product of selection after three generations of open pollination from a cross of a tetraploid cultivar (most probably Taunusblümchen) with an R. multiflora selection (T. Debener, personal communication). CA60 has demonstrated exceptional winterhardiness in the field at Vineland, ON, Canada. The second mapping population consisted of 107 individuals and resulted from the 2016-cross between the cold-susceptible female parent ‘HARpageant’ [Easy Does It (‘EDI’)] and the cold-hardy male parent ‘George Vancouver’ (‘GV’). ‘GV’ was bred by Felicitas Svejda and belongs to the Explorer Rose Collection.

Experimental design and growing conditions.

The plants from the CA60 × ‘SITR’ population and its parents were propagated from cuttings between Aug. 2017 and Jan. 2018, and the plants from the ‘EDI’ × ‘GV’ population and its parents were propagated from Jan. to Mar. 2019. The propagation was staggered over a few months to manage a variable rooting success rate, but the age of each cutting was not recorded. The cuttings were propagated in a mist house directly into a soilless potting mix (Sungro Sunshine Mix 1) or using foam strips. The mist house was heated at 22 °C; the temperature varied between 20 and 25 °C but rose to 35 °C during the hottest summer months. The relative humidity was between 60% and 75% and the misting system was activated every 5 min during a 16 h photoperiod. High-pressure sodium (HPS) supplemental light was provided to maintain a 16 h photoperiod. Cuttings were transferred into 0.5-L pots (9.5 cm height, 10 cm width) between 2 and 4 weeks after propagation before being transplanted into 11-L pots (24 cm height, 25 cm width). Roses were kept in 0.5-L pots (9.5 cm height, 10 cm width) for a minimum time of 4 weeks before being transplanted into larger pots. Plants from the CA60 × ‘SITR’ population were grown on their own roots and maintained in the greenhouse from Sept. 2017 to Oct. 2018. Temperature in the greenhouse was maintained at an average of 24 °C, but diurnal variation of temperature occurred with a maximum–minimum temperature range of 30/18 °C (day/night); 400-W HPS supplemental light was provided from Oct. 2017 to May 2018 to maintain a 16-h photoperiod. Plants from the ‘EDI’ × ‘GV’ population were grown on their own roots and maintained outdoors from May to Oct. 2019 because of restricted space in the greenhouse. Average outdoor temperature in 2019 varied from 12 °C in May, 23 °C in the hottest summer months, and 12 °C in October. All plants were watered with the use of irrigation strips and fertilized two to three times per week with 20N–3.4P–16.6K and 14N–0P–11.6K fertilizers (Plant-Prod) at EC level in the range of 1.2 to 1.6 dS·m⁻¹. Only sulfur was applied to control powdery mildew on the plants, no additional chemicals were used. Biocontrol was used to control insect pressure.

A total of 100 genotypes from the population CA60 × ‘SITR’ and 88 from the population ‘EDI’ × ‘GV’, in addition of the parental genotypes CA60, ‘SITR’, ‘EDI’, and ‘GV’, were used in electrolyte leakage experiments. In preparation for the electrolyte leakage experiments, the plants from each mapping population were moved to a greenhouse maintained at an average temperature of 10 °C (i.e., pre-acclimation phase) where diurnal variation of temperature occurred with a maximum–minimum temperature range of 25/8 °C (day/night) with dark for 16 h·d⁻¹ and shading as needed and without supplemental light or fertilization. Plants were then subjected to a 2-week artificial acclimation period in a walk-in cooler at 4 °C (KeepRite Refrigeration, Brantford, ON, Canada) in which two light-emitting diode (LED) tripods [Husky Model DE007; Stanley Black & Decker (Mississauga, ON, Canada); Western Forge (Colorado Springs, CO); Apex Tool Group (Sparks, MD); and Iron Bridge Tools (Shelburne Falls, MA)] and two 30-m metal cage string lights with LED bulbs (E253217 Model LS-100; Ningbo Linsheng Electric Co., Ltd., Zhejiang, China) delivered 10 µmol·m⁻²·s⁻¹ of light at canopy level for 6 h·d⁻¹. Photosynthetically active radiation was determined from quantum sensors (LI-250A; LI-COR, Lincoln, NE). Clones of the F₁ progeny were brought to the cooler over several weeks to conduct multiple replications of the experiments as detailed in the following paragraph, and the plants were randomized in the cooler using the online plant breeding platform Phenome (Phenome Networks Ltd, Rehovot, Israel). Sections that were 10 cm long and of similar diameter were collected at the proximal end of the stem (i.e., experimental unit) from acclimated plants and were subjected to freezing treatments in a programmable freezer. This experiment used a split-plot design in which temperature represented the main-plot factor and genotypes were assigned to the subplots.

Electrolyte leakage experiments on the parental lines, the population CA60 × ‘SITR’ and the population ‘EDI’ × ‘GV’ were conducted separately. Experiments on the parental lines were conducted in Nov. 2019 and replicated three times with unique sets of clones (i.e., true biological replications) with three stem sections per genotype and temperature (i.e., technical replications) and 1 week between each replication (i.e., replications in time). A fourth replication was conducted on CA60 and ‘GV’ only, using a distinct set of clones 2 weeks after the third replication. Clones of the parental genotypes were moved to the pre-acclimation house in Oct. 2019. Experiments on the population CA60 × ‘SITR’ were conducted from Nov. 2018 to Jan. 2019 and replicated five times (i.e., replications in time) using five stem sections per temperature and per genotype (i.e., technical replications) and three unique sets of clones (i.e., true biological replications), two of which were used in two replications and were allowed to de-acclimate in a greenhouse at room temperature for 2 weeks before being reused. There were 2 weeks between each replication, with the exception of 3 weeks between replications 3 and 4. Plants from the CA60 × ‘SITR’ population were moved to the pre-acclimation house in Oct. 2018. Experiments on the population ‘EDI’ × ‘GV’ were conducted from Jan. to Feb. 2020 and replicated three times (i.e., replications in time) with unique sets of clones (i.e., true biological replications), with 2 weeks between each replication. Only one stem section per temperature and genotype was used in the experiment on the ‘EDI’ × ‘GV’ population. Plants from the ‘EDI’ × ‘GV’ population were moved to the pre-acclimation house in Oct. 2019.

Phenotyping.

The electrolyte leakage assay was conducted as described in the previous section (Expt. 1: Electrolyte leakage as a proxy for winterhardiness) with some modifications. The bags containing the rose stems were moved to a programmable freezer in which the temperature dropped from 0 to −50 °C (experiments on the parental lines), from 0 to −20 °C (experiments on the population CA60 × ‘SITR’) and from 0 to −40 °C (experiments on the population ‘EDI’ × ‘GV’), at a rate of −2.5 °C·h⁻¹, with a 12-h nucleation step at −4 °C. The initial stem from the sealable plastic bag that was taken out of the freezer was then cut into three, five or one 1-cm-long stem sections in the experiments conducted on the parental lines, the population CA60 × ‘SITR’ and the population ‘EDI’ × ‘GV’, respectively. The stem sections were cut in a way that excluded buds and thorns. Each section was placed into individual 5-mL centrifuge tubes filled with type 1 water (i.e., technical replications). R₀ and Rt were recorded before and after autoclave as described previously using a compact EC meter (LAQUAtwin-EC-11; Horiba, Ltd., Kyoto, Japan) with 120 μL of solution. Data on the parental lines were recorded at 4 (i.e., control), −10, −15, −20, −25, −30, −35, −40, −45, and −50 °C; data on the population CA60 × ‘SITR’ were recorded at −10, −15, and −20 °C; and data on the population ‘EDI’ × ‘GV’ were recorded at 4 (i.e., control), −10, −15, −20, −25, −30, −35, and −40 °C. The percentage of electrolyte leakage was calculated as described previously. In addition, an index of injury (I) corrected for electrolyte leakage not due to freezing—measured at 4 °C—was estimated for the parental genotypes and the population ‘EDI’ × ‘GV’ by the formula $\text{I}\left(\text{\%}\right)=\frac{\frac{R0}{Rt}\hspace{0.17em}-\hspace{0.17em}\frac{\mathit{Roref}}{\mathit{Rtref}}}{1\hspace{0.17em}-\hspace{0.17em}\frac{\mathit{Roref}}{\mathit{Rtref}}}$ as described by Ouyang et al. (2019).

Data analysis.

The analysis was conducted separately for the parental lines, the population CA60 × ‘SITR’, and the population ‘EDI’ × ‘GV’. Generalized linear mixed models (GLMM) were fitted to the electrolyte leakage data using SAS software (SAS version 9.4; SAS Institute Inc., Cary, NC), specifying a beta distribution and the logit link under the GLIMMIX procedure (Supplemental Material). Temperature was a fixed effect. Genotypes were considered as fixed effects when analyzing only the parental lines, and the best linear unbiased estimates (BLUE) were computed. Genotypes were considered as random effects in the analysis of the mapping populations, and the best linear unbiased predictors (BLUP) were computed. In addition, LT50 was estimated for the parental lines CA60, ‘SITR’, ‘EDI’, and ‘GV’ by fitting a nonlinear regression dosage curve response on the index of injury with nine temperatures (−10, −15, −20, −25, −30, −35, −40, −45, and −50 °C), using the nlmixed procedure in the SAS software. In the same way, LT50 was estimated individually for each sibling of the ‘EDI’ × ‘GV’ population, from the index of injury obtained from six temperatures (−15, −20, −25, −30, −35, and −40 °C). The index of injury was used to compute the LT50s; the BLUPs of electrolyte leakage and the LT50 were used in QTL mapping.

The female and male genetic maps of CA60 and ‘SITR’ were created from data previously generated through genotyping by sequencing [GBS (Rouet et al., 2019)] to include all individuals for which cold hardiness data were available. The female and male maps of ‘EDI’ and ‘GV’, respectively, were created from newly generated GBS data following a two-way pseudo-test cross-mapping strategy, using R-package ASMap (Taylor and Butler, 2017), as described by Li et al. (2014) and applied on roses in Rouet et al. (2019). The four parental genetic maps were used as a framework for QTL mapping. A total of 83 and 80 individuals were used to identify QTL associated with electrolyte leakage from the CA60 × ‘SITR’ and ‘EDI’ × ‘GV’ populations, respectively. For the ‘EDI’ × ‘GV’ population, 76 individuals were also used to identify QTL associated with LT50.

QTL were initially detected by interval mapping (IM) using the R/qtl2 package (Broman et al., 2019) and performing a genome scan by Haley-Knott regression (Haley and Knott, 1992). Then, results from IM were compared with a multiple QTL mapping (MQM) approach using the R implementation of Ritsert Jansen’s MQM method (Arends et al., 2010) (R version 3.5.3; R Foundation for Statistical Computing, Vienna, Austria), and additional QTL were detected. Both forward stepwise and backward elimination strategies were used to identify QTL underlying the trait in MQM. The forward stepwise approach was conducted by setting the major QTL detected in IM as cofactors. Variance components used in QTL modeling were estimated by the restricted maximum likelihood (REML) approach. Significance cutoff was determined by performing 1000 permutations and the fitqtl function from the package R/qtl (Broman et al., 2003) was applied to estimate the percentage of variation explained by the main QTL. Bayes intervals were obtained with 95% confidence for the QTL. Only the results from MQM were reported. The relative position of the single-nucleotide polymorphism (SNP) markers in the reference genome R. chinensis ‘Old Blush’ V2 were used to identify QTL intervals.

Genetic material.

The two mapping populations described previously, CA60 × ‘SITR’ and ‘EDI’ × ‘GV’, and their parental lines were used in this experiment. For both populations, the genotypes used in field trials overlapped with the genotypes used in electrolyte leakage experiments. In addition, four cultivars that were a mixture of hybrid teas and roses from the Explorer Series were used as controls in this experiment: cold-sensitive cultivars MEIpierar (Caroline de Monaco) and WEKrigoyelo (Gentle Giant), and cold-hardy cultivars Frontenac and Nicolas (Table 1). The mapping populations were grown on their own roots. The parental lines ‘SITR’ and ‘EDI’, and the controls were obtained from a local nursery and grown on R. multiflora rootstock. The parental genotype ‘GV’ was grown both on its own roots and grafted on R. multiflora rootstock (Table 1).

Experimental design and growing conditions.

The population CA60 × ‘SITR’ was propagated from cuttings using foam strips in the mist house between Aug. 2017 and Jan. 2018 at the same time as the cuttings intended for electrolyte leakage experiments. The propagation was staggered over a few months to manage a variable rooting success rate, but the age of each cutting was not recorded. A total of 101 F₁ progeny and nine control and parental genotypes were planted in two locations in Canada in early June 2018: Elora, ON, Canada (lat. 43°38'27.4″N, long. 80°24'03.2″W; USDA zone 5b; cultivated Brunisol soil with silt loam) and Saskatoon, SK, Canada (lat. 52°07'21.5″N, long. 106°36'41.8″W; USDA zone 3b; Dark Brown Chernozemic soil). Roses were planted at each location using a randomized complete block design (RCBD) with five replications designed with the online plant breeding platform Phenome, the experimental unit being the rose bush. Plants were spaced by 60 cm within rows, and the rows were spaced by 1.50 m. Black heavy duty woven polypropylene landscape fabric and 46,000 kg of organic light-colored mulch, which was applied on top of the landscape fabric, were used to control weeds in Elora, and only mulch was used in Saskatoon. Plants were pruned in Summer 2019 to remove all the dead canes. The population ‘EDI’ × ‘GV’ was propagated from cuttings into a soilless potting mix in the mist house from Jan to Mar. 2019 at the same time as the cuttings intended for electrolyte leakage experiments. A total of 96 F₁ progeny and four parental genotypes were planted in late July. 2019 in Elora using an RCBD with three replications. Plants were spaced by 1 m within rows to accommodate their spreading architecture, and the rows were spaced by 1.50 m. Organic light-colored mulch was used to control weeds. The population CA60 × ‘SITR’ was assessed over two seasons (Winters 2019 and 2020), and the population ‘EDI’ × ‘GV’ was assessed over only one season (Winter 2020). In addition, climate data for Saskatoon were retrieved from the Saskatchewan Research Council Website at the Saskatoon Climate Reference Station (Saskatchewan Research Council, 2021), and climate data collected at the Elora Research Station were retrieved from the Environment Canada Website (Environment Canada, 2021).

Phenotyping.

Field winterhardiness was assessed as two components: winter damage and spring regrowth as described by (Vukosavljev, 2014). Winter damage was a measurement of cane dieback using the following formula: $\text{WD}\left(\text{\%}\right)=100\hspace{0.17em}\times \hspace{0.17em}\frac{ld}{lt}$ with WD being the percentage of winter damage, ld the length of dieback and lt the length of the whole cane. Three stems per plant were measured and the average value for each plant was used as the rating for winter damage. Regrowth was estimated with the following formula: $\text{RG}\left(\text{\%}\right)=100\hspace{0.17em}\times \hspace{0.17em}\frac{ln}{lt}$, with RG being the percentage of regrowth, ln the length of the new shoot, and lt the initial length of the whole cane. Regrowth value for each genotype was calculated as the average regrowth of three branches. For both winter damage and regrowth, the three longest stems from the crown were chosen. Therefore, the height of the three tallest stems that were emerging in the spring were used in reference to the overall height of the three tallest stems going into the previous winter to estimate regrowth. The number of newly emerging shoots from multiple axillary buds was not recorded.

Winter damage was measured in early spring when the leaf buds started to grow actively. Spring regrowth was measured during the second week of June 2019 and first week of July 2020 in Elora and during the last week of July 2019 and 2020 in Saskatoon. Plants that died through Winter 2019 were assigned 100% winter damage for Winter 2019 and missing data for winter damage 2020, regrowth 2019, and regrowth 2020. Plants that died in Spring 2019 because of planting, drought, or pest pressure were assigned missing data for winter damage and regrowth across years. Plants that died through the Winter 2020 were assigned 100% winter damage in 2020 and missing data for regrowth 2020. This was done to ensure that the timing of a plant’s death was accurately captured and that the ability of the meristem to resume growth in the spring even on highly damaged plants was well represented. Winter damage and regrowth data were collected in Elora in 2019 and in 2020 and in Saskatoon in 2019 and in 2020 for the population CA60 × ‘SITR’ and in Elora in 2020 for the population ‘EDI’ × ‘GV’. Pest-induced defoliation [black spot (Diplocarpon rosae), rose slugs (Endelomyia aethiops), and sawflies (Hymenoptera)] was also recorded in Sept. 2019 in Elora on a scale from 0 to 4 (0 = no defoliation and intact leaves, 1 = minor damage to the leaves mainly caused by caterpillars, 2 = minor defoliation with leaves left all over the rose bush, 3 = leaves left only on the tips of the stems, 4 = no leaves left, severe defoliation).

Data Analysis.

Descriptive statistical analyses were first conducted on field data for the two mapping populations using the RStudio software. The data were then modeled with a three-dimensional productivity contour map using SAS software (version 9.4) and the spline method in the G3GRID procedure to investigate the presence of nonrandom error distribution. A GLMM was fitted to field data using the SAS software, and the GLIMMIX procedure was used to conduct a genotype × environment (GE) interaction analysis with up to four environments (Elora 2019, Elora 2020, Sask 2019, Sask 2020). The variances of regrowth and winter damage for the field trial CA60 × ‘SITR’ with multiple environments were partitioned into random effects genotype, environment, GE, and block (environment). When analyzing environment separately, the variances of regrowth and winter damage for the field trial ‘EDI’ × ‘GV’ and the field trial CA60 × ‘SITR’ were partitioned into random effects: Block and Genotype. The linear predictors were respectively: η_ij = η+ E_i + a_j +(Ea)_ij + b(E)_ki, where η was the intercept, E_i the environment effect, a_j the genotype effect, (Ea)_ij the GE interaction, and b(E)_ki the bloc effect nested within environment, and η_ij = η + b_i + a_j, where η was the intercept, a_j the genotype effect, and b_i the block effect. A type I error of 0.05 was used to determine the significance of tests in this analysis. Random effects were assessed with a likelihood test for covariance parameters. Winter damage data across environments and from single environments Elora 2020 and Sask 2019 were fitted to a beta distribution with a logit link function and Laplace interval estimation method. Winter damage data from single environments Elora 2019 were fitted to a normal distribution with the identity link function. Regrowth data were fitted to a lognormal distribution with the identity link function. The best fitted model was chosen based on the analysis of residuals and the fit statistics such as the AIC. The models were surveyed to determine if setting a heterogeneous covariance structure was necessary. BLUP estimates were computed for each genotype in each environment.

Heritability estimates were obtained for each trait in single environments by estimating the genetic variance components from the completely random mixed model. In the case in which the data fitted a normal or lognormal distribution in a completely random model, REML was used to compute and retrieve the variance estimates. Broad sense heritability used the following equation: ${\text{H}}^2=\hspace{0.17em}\frac{\text{Vg}}{\text{Vg}\hspace{0.17em}+\hspace{0.17em}\left(\frac{\text{Vr}}{\text{r}}\right)}$, where H² was the broad sense heritability, V_g the genetic variance, V_r the residual variance, and r the number of replicates or blocs in the RCBD trial (Gitonga et al., 2014). Broad sense heritability was obtained from REML-derived estimates whenever the data were about normally distributed, as advised by Nakagawa and Schielzeth (2010). However, if the data fitted a beta distribution with a logit link function, the latent-scale heritability was obtained from the GLMM-based estimates (Nakagawa and Schielzeth, 2010). Under this scenario, the residual variance was calculated as follows: $\text{Vr}\hspace{0.17em}=\varphi \hspace{0.17em}+\frac{{\text{π}}^2}{3}$, where $\varphi$ was the scale parameter computed by the GLMM and represented the over dispersion of the data under the latent scale, and $\frac{{\text{π}}^2}{3}$ represented the variance for the logistic distribution. The latent-scale broad sense heritability was calculated by the following formula: ${\text{H}}^2=\frac{\text{Vg}}{\text{Vg}\hspace{0.17em}+\hspace{0.17em}\left(\frac{\varphi +\frac{{\text{π}}^2}{3}}{\text{r}}\right)}$. In addition, the ratio $\frac{\text{Vge}}{\text{Vg}}$ was calculated to quantify the size of the GE interaction relative to the genetic variance in mixed models using multiple environments (Gitonga et al., 2014).

Environment-metric preserving genotype plus GE interaction (GGE) biplots were generated using the BLUPs of field winter damage and regrowth to visually characterize simultaneously genotype by environment relationships using RStudio and the packages GGEBiplot GUI (Frutos et al., 2014). The GGE-biplot model uses the principle of singular value decomposition (SVD) to decompose genotype (G) and GE effects into two or more components represented on the graph axis. GGE-biplots were generated using the column metric preserving option (SVP = 2), tester-centered (Center = 2) and unscaled G + GE (Ling et al., 2021). BLUPs of winter damage and regrowth were also used in QTL mapping. A total of 83 and 86 individuals were used to identify QTL in the population CA60 × ‘SITR’ and in the population ‘EDI’ × ‘GV’, respectively. The presence of QTL was investigated and reported as previously described (Expt. 2: QTL mapping of electrolyte leakage). Finally, pathogen-induced defoliation data recorded in Elora in 2019 were averaged for each genotype across five blocks and used in correlation analysis with BLUPs of winter damage and regrowth and in mapping. Climate data were visualized in RStudio using the package ggplot2 (Wickham, 2016). The monthly number of freeze-thaw cycles were calculated as the number of days where the maximum temperature was higher than 0 °C and the minimum temperature was below −1 °C.

Relationship between electrolyte leakage and field winterhardiness.

Correlations between the BLUEs and the BLUPs of electrolyte leakage and index of injury, LT50, and field data were computed using the Pearson method in the RStudio software.

Linkage maps.

The parental maps for CA60 and ‘SITR’ were slightly modified from the previous published maps (Rouet et al., 2019) to include all individuals for which electrolyte leakage and winterhardiness data were available. The newly created genetic map for CA60 included 140 individuals and spanned 31 linkage groups with 814 simplex SNP markers across 1863 cM. The length of the linkage groups varied from 6 to 150 cM with a mean interval distance between markers of 2.3 cM. The newly created genetic map for ‘SITR’ included 130 individuals and spanned 29 linkage groups with 660 simplex SNP markers across 1773 cM. The length of linkage groups varied from 8 to 161 cM with a mean interval distance between markers of 2.7 cM. Some homologs were missing, and others were fragmented and were represented by more than one fragment. In addition, rearrangements in marker order were observed in comparison with the reference R. chinensis ‘Old Blush’ homozygous Genome v2.0 (Raymond et al., 2018). The parental maps for ‘EDI’ and ‘GV’ were created from newly generated GBS data. A total of 5211 simplex SNP markers that were heterozygous in ‘EDI’ and homozygous in ‘GV’ were identified, and 2940 simplex SNPs that were homozygous in ‘EDI’ and heterozygous in ‘GV’ were identified. The ‘EDI’ female map included 97 individuals and spanned 29 linkage groups with 685 simplex SNP markers across 2093 cM. The length of linkage groups varied from 6 cM to 139 cM with a mean interval distance between markers of 3 cM. The ‘GV’ male map included 97 individuals and spanned 24 linkage groups across 1799 cM with 489 simplex SNP markers and a mean interval distance between markers of 3.7 cM. The length of linkage groups varied from 6 to 149.5 cM.

Calibration with parental lines.

Electrolyte leakage experiments on the four parental genotypes, conducted to establish the range of freezing temperatures that would be further implemented to screen their segregating progeny, were highly repeatable (Table 2). However, the electrolyte leakage did not increase from −30 to −50 °C in the third replication (data not shown). The data on CA60 and ‘GV’ in the fourth replication were in the same range as the first and second replications (Supplemental Fig. 1). Therefore, the third replication was removed from the analysis most likely because of a technical issue with the freezer. Only the first two replications were retained for the statistical analysis. The mixed model analysis indicated that the electrolyte leakage and the index of injury were genotype-dependent (Supplemental Table 2). In addition, there were significant differences between the amount of electrolyte leakage and the LT50 of the parental genotypes (Fig. 4, Supplemental Table 3).

Response of four Rosa ×hybrida genotypes, CA60, Singin’ in the Rain [‘SITR’ (‘MACivy’)], Easy Does It [‘EDI’ (‘HARpageant’)], and ‘George Vancouver’ (‘GV’), to freezing treatment from −10 to −50 °C. Sensitivity to freezing is displayed as the BLUE estimates of the index of injury (I) [electrolyte leakage (EL) corrected for leakage not due to freezing damage (i.e., control)], estimated from two replications of EL assays.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

View Full Size

Response of four Rosa ×hybrida genotypes, CA60, Singin’ in the Rain [‘SITR’ (‘MACivy’)], Easy Does It [‘EDI’ (‘HARpageant’)], and ‘George Vancouver’ (‘GV’), to freezing treatment from −10 to −50 °C. Sensitivity to freezing is displayed as the BLUE estimates of the index of injury (I) [electrolyte leakage (EL) corrected for leakage not due to freezing damage (i.e., control)], estimated from two replications of EL assays.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

Response of four Rosa ×hybrida genotypes, CA60, Singin’ in the Rain [‘SITR’ (‘MACivy’)], Easy Does It [‘EDI’ (‘HARpageant’)], and ‘George Vancouver’ (‘GV’), to freezing treatment from −10 to −50 °C. Sensitivity to freezing is displayed as the BLUE estimates of the index of injury (I) [electrolyte leakage (EL) corrected for leakage not due to freezing damage (i.e., control)], estimated from two replications of EL assays.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

Table 2.

Correlation among three replications (Rep) of electrolyte leakage experiments conducted separately on the Rosa ×hybrida genotypes CA60, Easy Does It [‘EDI’ (‘HARpageant’)], Singin’ in the Rain [‘SITR’ (‘MACivy’)], and ‘George Vancouver’ (‘GV’), the mapping population CA60 ×’SITR’, and the mapping population ‘EDI’× ’GV’. Pearson coefficients of correlation are given in the table and significant correlations at α = 0.01 are given by *.

View Table

CA60 × ‘SITR’ population.

Electrolyte leakage experiments conducted on 100 F₁ progeny and their two parental genotypes over three temperatures (−10, −15, and −20 °C) were highly repeatable (Table 2), and the BLUPs of electrolyte leakage were obtained from the mixed model analysis (Supplemental Table 4). Although the LT50 could not be calculated in this experiment without the index of injury, the amount of electrolyte leakage for CA60 and ‘SITR’ at −20 °C (59% and 66%, respectively) were comparable to the values found in the experiment on the parental lines (58% and 71%, respectively). Within the population, as the experimental temperature decreased, the amount of electrolyte leakage increased from 28% (−10 °C) to 54% (−20 °C) (Supplemental Fig. 2). The amount of electrolyte leakage ranged from 41% to 74% at −20 °C, suggesting the existence of large unidirectional transgressive segregation (Table 3). Three QTL associated with electrolyte leakage were identified (Table 4, Supplemental Figs. 3 and 4). A QTL for electrolyte leakage at −10 °C explaining 22% of the phenotypic variance mapped to LG3.H1 at 15 cM in the CA60 female map. A QTL for electrolyte leakage at −20 °C explaining 15% of the phenotypic variance mapped to LG2.H1 at 99 cM in the CA60 female map. A QTL for electrolyte leakage at −20 °C explaining 13% of the phenotypic variance also mapped to LG2.H5 at 15 cM in the ‘SITR’ male map.

Table 3.

Evidence for transgressive segregation for electrolyte leakage (EL), field winter damage (WD), and field spring regrowth (RG) in the Rosa ×hybrida mapping populations CA60 × Singin’ in the Rain [‘SITR’ (‘MACivy’)] and Easy Does It^TM [‘EDI’ (‘HARpageant’)] × ‘George Vancouver’ (‘GV’) and their parental lines, which differ in their hardiness level.

View Table

Table 4.

Quantitative trait loci (QTL) associated with electrolyte leakage (EL), field winter damage (WD), field spring regrowth (RG), and defoliation in two Rosa ×hybrida mapping populations CA60 × Singin’ in the Rain^TM [‘SITR’ (‘MACivy’)] and Easy Does It^TM [‘EDI’ (‘HARpageant’)] × ‘George Vancouver’ (‘GV’).

View Table

‘EDI’ × ‘GV’ population.

Electrolyte leakage experiments were conducted on 88 F₁ progeny over a wide range of sub-zero temperatures from −15 to −45 °C with a measurement taken at 4 °C for a control. The experiments were highly repeatable (Table 2), the BLUPs of electrolyte leakage were obtained from the mixed model analysis (Supplemental Table 4), and the LT50s were obtained for each genotype (Supplemental Table 5). Although the parental genotypes ‘EDI’ and ‘GV’ were not included in this experiment, inferences on their LT50s could be made based on the experiments conducted on the parental genotypes (Supplemental Table 3). Within the population, the amount of electrolyte leakage increased from 47% (−15 °C) to 78% (−45 °C) as the experimental temperature decreased (Supplemental Fig. 2) without evidence for transgressive segregation (Table 3). Four QTL for electrolyte leakage and LT50 were identified (Table 4, Supplemental Figs. 5 and 6). A QTL for electrolyte leakage at −20 °C explaining 18% of the phenotypic variance mapped to LG6.H1 at position 45 cM in the ‘EDI’ female map. QTL for electrolyte leakage at −20 °C and for LT50 mapped to the same position on LG7.H1 at 135 cM in the ‘GV’ male map. The two QTL explained 14% and 16% of the phenotypic variance respectively. A QTL for electrolyte leakage at −35 °C mapped to LG7.H3 at 20 cM in the ‘GV’ male map, and it explained 22% of the phenotypic variance.

CA60 × ‘SITR’ population.

The CA60 × ‘SITR’ population was evaluated over 2 years at two locations with very different climates. Winter temperatures in Saskatoon (USDA hardiness zone 3b) were colder than in Elora (USDA hardiness zone 5b). Temperature is Saskatoon reached −40 °C during the colder months of the first winter with minimal snow cover. In addition, sub-zero temperatures were recorded for up to 40 consecutive days in Saskatoon between February and Mar. 2019, while sub-zero temperatures were recorded for less than 10 consecutive days in Elora. Environmental conditions also varied at the same location across years, the number of consecutive days with sub-zero temperatures was reduced in Sask 2020 in comparison with Sask 2019, but the number of freeze-thaw cycles increased (Supplemental Fig. 7). Although the winter temperatures were consistent in Elora across years, severe disease pressure occurred in Fall 2019.

Winter damage, which was the first aspect of field winterhardiness, varied across environments (Fig. 5A–D). As expected, cold-susceptible genotypes suffered higher damage than the hardy control and parental genotypes (Fig. 5E–H). The data from Sask 2020 were highly skewed toward extreme winter damage and not informative, as 90% of the population exhibited more than 95% winter damage (Fig. 5D); therefore, Sask 2020 was dropped from the rest of the winter damage analysis. The GE interaction was significant for the three remaining environments with a high $\frac{\text{Vge}}{\text{Vg}}$ ratio ($\frac{\text{Vge}}{\text{Vg}}=1.33$), where V_ge is the variance associated with the GE interaction; therefore, the BLUPs of winter damage were obtained separately for each of these environments (Supplemental Table 6). GGE-biplots were generated using the BLUPs of winter damage to further visualize the relationships between the environments and characterize the nature of GE interaction. The two principal components PC1 and PC2 of the GGE-biplot explained 71.51% and 20.8% of the GGE variation, respectively, meaning that the two sources of variation G plus GE explained 92.31% of the total phenotypic variance (Fig. 6). The angle between the environment vectors associated with Elora 2019 and Sask 2019 was acute, meaning that the environments were positively correlated without crossovers GE patterns; however, the vector associated with Elora 2019 was longer than that of Sask 2019. Therefore, Elora 2019 was more discriminative than Sask 2019, and the nature of the GE interaction was mainly a change in magnitude due to the severity of the climatic conditions in Saskatoon. Furthermore, Elora 2019 and Elora 2020 were distant on the GGE-biplot, meaning that the two environments discriminated the genotypes differently based on winter damage with minor patterns of GE crossovers (Fig. 6). These observations were supported by the Pearson’s correlations (Table 5). The heritability of winter damage varied by location and year; it was high in Elora 2019, but it was low in Sask 2019 and Elora 2020, ranging from 0.13 to 0.83 (Table 6). Noticeably, pest-induced defoliation that occurred in Elora 2019 was positively and highly correlated with winter damage in Elora 2020 (Table 5). Overall, Elora 2019 was the least severe environment, with a population mean of 51% winter damage, whereas Sask 2019 was the most severe environment with a population mean of 81% winter damage (Table 3). There was evidence for transgressive segregation beyond both parental values (Table 3), with 4% of individuals that were as hardy as the hardiest parent CA60 and 30% that were as sensitive or more sensitive than the tender parent ‘SITR’ in Elora 2019.

Distribution of field winter damage (WD) in the Rosa ×hybrida mapping population CA60 × Singin’ in the Rain [SITR (‘MACivy’)] in four environments: Elora, ON, Canada in (A) 2019 (Elora 2019) and (B) 2020 (Elora 2020), Saskatoon, SK, Canada in (C) 2019 (Sask 2019) and (D) 2020 (Sask 2020), and distribution of WD among the parental and control genotypes used in (E) Elora 2019, (F) Elora 2020, (G) Sask 2019, and (H) Sask 2020. WD was a measurement of cane dieback in the field using the following formula: $\text{WD}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ld}}{\text{lt}}$, with WD being the percentage of winter damage, ld the length of dieback, and lt the length of the whole cane. Data on three canes per plant were collected and averaged. The parental and control genotypes can be divided into cold-hardy [‘Frontenac’ (Fr), ‘George Vancouver’ (GV), GV grown on its own roots (GV1), ‘Nicolas’ (Ni), and CA60 (60)] and non–cold-hardy genotypes [Gentle Giant (GG), Caroline de Monaco® (CM), Easy Does It (EDI), and SITR] relative to their U.S. Department of Agriculture (USDA) cold hardiness zones.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

View Full Size

Distribution of field winter damage (WD) in the Rosa ×hybrida mapping population CA60 × Singin’ in the Rain [SITR (‘MACivy’)] in four environments: Elora, ON, Canada in (A) 2019 (Elora 2019) and (B) 2020 (Elora 2020), Saskatoon, SK, Canada in (C) 2019 (Sask 2019) and (D) 2020 (Sask 2020), and distribution of WD among the parental and control genotypes used in (E) Elora 2019, (F) Elora 2020, (G) Sask 2019, and (H) Sask 2020. WD was a measurement of cane dieback in the field using the following formula: $\text{WD}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ld}}{\text{lt}}$, with WD being the percentage of winter damage, ld the length of dieback, and lt the length of the whole cane. Data on three canes per plant were collected and averaged. The parental and control genotypes can be divided into cold-hardy [‘Frontenac’ (Fr), ‘George Vancouver’ (GV), GV grown on its own roots (GV1), ‘Nicolas’ (Ni), and CA60 (60)] and non–cold-hardy genotypes [Gentle Giant (GG), Caroline de Monaco® (CM), Easy Does It (EDI), and SITR] relative to their U.S. Department of Agriculture (USDA) cold hardiness zones.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

Distribution of field winter damage (WD) in the Rosa ×hybrida mapping population CA60 × Singin’ in the Rain [SITR (‘MACivy’)] in four environments: Elora, ON, Canada in (A) 2019 (Elora 2019) and (B) 2020 (Elora 2020), Saskatoon, SK, Canada in (C) 2019 (Sask 2019) and (D) 2020 (Sask 2020), and distribution of WD among the parental and control genotypes used in (E) Elora 2019, (F) Elora 2020, (G) Sask 2019, and (H) Sask 2020. WD was a measurement of cane dieback in the field using the following formula: $\text{WD}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ld}}{\text{lt}}$, with WD being the percentage of winter damage, ld the length of dieback, and lt the length of the whole cane. Data on three canes per plant were collected and averaged. The parental and control genotypes can be divided into cold-hardy [‘Frontenac’ (Fr), ‘George Vancouver’ (GV), GV grown on its own roots (GV1), ‘Nicolas’ (Ni), and CA60 (60)] and non–cold-hardy genotypes [Gentle Giant (GG), Caroline de Monaco® (CM), Easy Does It (EDI), and SITR] relative to their U.S. Department of Agriculture (USDA) cold hardiness zones.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

Environment-metric preserving biplot representing genotype plus genotype × environment interaction (GGE) and generated from best linear unbiased predictors (BLUPs) of (A) field winter damage (WD) and (B) regrowth (RG) in the Rosa ×hybrida mapping population CA60 × Singin’ in the Rain [SITR (‘MACivy’)]. The length of the environment vectors provides information on the ability of an environment to discriminate among genotypes, whereas the distance between environment markers and the angles between the environment vectors are associated with the correlation between environments. The GGE-biplot model uses the principle of singular value decomposition (SVD) to decompose genotype (G) and genotype × environment (GE) effects into two components represented as principal components (PC1 and PC2) on the graph axis. WD is represented over three environments (Env): Elora, ON, Canada in 2019 (Elora 2019) and 2020 (Elora 2020), and Saskatoon, SK, Canada in 2019 (Sask 2019). RG is represented over four environments: Elora 2019, Elora 2020, Sask 2019, and Saskatoon in 2020 (Sask 2020). WD was a measurement of cane dieback in the field using the following formula: $\text{WD}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ld}}{\text{lt}}$, with WD being the percentage of winter damage, ld the length of dieback, and lt the length of the whole cane. RG was estimated with the following formula: $\text{RG}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ln}}{\text{lt}}$, with RG being the percentage of regrowth, ln the length of the new shoot, and lt the initial length of the whole cane. For both WD and RG, data on three canes per plant were collected and averaged.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

View Full Size

Environment-metric preserving biplot representing genotype plus genotype × environment interaction (GGE) and generated from best linear unbiased predictors (BLUPs) of (A) field winter damage (WD) and (B) regrowth (RG) in the Rosa ×hybrida mapping population CA60 × Singin’ in the Rain [SITR (‘MACivy’)]. The length of the environment vectors provides information on the ability of an environment to discriminate among genotypes, whereas the distance between environment markers and the angles between the environment vectors are associated with the correlation between environments. The GGE-biplot model uses the principle of singular value decomposition (SVD) to decompose genotype (G) and genotype × environment (GE) effects into two components represented as principal components (PC1 and PC2) on the graph axis. WD is represented over three environments (Env): Elora, ON, Canada in 2019 (Elora 2019) and 2020 (Elora 2020), and Saskatoon, SK, Canada in 2019 (Sask 2019). RG is represented over four environments: Elora 2019, Elora 2020, Sask 2019, and Saskatoon in 2020 (Sask 2020). WD was a measurement of cane dieback in the field using the following formula: $\text{WD}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ld}}{\text{lt}}$, with WD being the percentage of winter damage, ld the length of dieback, and lt the length of the whole cane. RG was estimated with the following formula: $\text{RG}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ln}}{\text{lt}}$, with RG being the percentage of regrowth, ln the length of the new shoot, and lt the initial length of the whole cane. For both WD and RG, data on three canes per plant were collected and averaged.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

Environment-metric preserving biplot representing genotype plus genotype × environment interaction (GGE) and generated from best linear unbiased predictors (BLUPs) of (A) field winter damage (WD) and (B) regrowth (RG) in the Rosa ×hybrida mapping population CA60 × Singin’ in the Rain [SITR (‘MACivy’)]. The length of the environment vectors provides information on the ability of an environment to discriminate among genotypes, whereas the distance between environment markers and the angles between the environment vectors are associated with the correlation between environments. The GGE-biplot model uses the principle of singular value decomposition (SVD) to decompose genotype (G) and genotype × environment (GE) effects into two components represented as principal components (PC1 and PC2) on the graph axis. WD is represented over three environments (Env): Elora, ON, Canada in 2019 (Elora 2019) and 2020 (Elora 2020), and Saskatoon, SK, Canada in 2019 (Sask 2019). RG is represented over four environments: Elora 2019, Elora 2020, Sask 2019, and Saskatoon in 2020 (Sask 2020). WD was a measurement of cane dieback in the field using the following formula: $\text{WD}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ld}}{\text{lt}}$, with WD being the percentage of winter damage, ld the length of dieback, and lt the length of the whole cane. RG was estimated with the following formula: $\text{RG}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ln}}{\text{lt}}$, with RG being the percentage of regrowth, ln the length of the new shoot, and lt the initial length of the whole cane. For both WD and RG, data on three canes per plant were collected and averaged.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

Table 5.

Correlation between best linear unbiased prediction estimates (BLUPs) of winter damage (WD) in three environments (Elora 2019, Elora 2020, and Sask 2019), defoliation, and BLUPs of regrowth (RG) for four environments: Elora, ON, Canada in 2019 (Elora 2019) and 2020 (Elora 2020), and Saskatoon, SK, Canada in 2019 (Sask 2019) and 2020 (Sask 2020) for the Rosa ×hybrida mapping population CA60 × Singin’ in the Rain^TM ['SITR' (‘MACivy’)], and between BLUPs of WD and RG for the Rosa ×hybrida mapping population Easy Does It^TM [‘EDI’ (‘HARpageant’)] × ‘George Vancouver’ (‘GV’). Significant correlations are given by *, **, and ***, indicating significant correlation at P = 0.05, P = 0.01, and P = 0.001, respectively.

View Table

Table 6.

Heritability and variance estimates for field winter damage and regrowth for the Rosa ×hybrida mapping populations CA60 × Singin’ in the Rain^TM [‘SITR’ (‘MACivy’)] and Easy Does It^TM [‘EDI’ (‘HARpageant’)] × ‘George Vancouver’ (‘GV’).

View Table

The second aspect of field winterhardiness in this study was spring regrowth. Regrowth in the population and the control and parental genotypes did not vary across environments as winter damage did (Fig. 7A and B). Although ‘SITR’ and CA60 did not differ in Elora 2019, ‘SITR’ had inferior potential for regrowth in the other environments compared with CA60 (Fig. 7B). The mixed model analysis of the four environments indicated significant GE interaction and high $\frac{\text{Vge}}{\text{Vg}}$ ratio [$\frac{\text{Vge}}{\text{Vg}}=3.09$ (Supplemental Table 7)]. The BLUPs for regrowth were obtained separately for each environment (Supplemental Table 7) and visualized with a GGE-biplot (Fig. 6). The two principal components PC1 and PC2 explained 44.41% and 28.9% of the GGE variation, respectively, meaning that G and GE explained 73.31% of the total phenotypic variance (Fig. 6). The right angle between the vectors associated with Saskatoon and the vectors associated with Elora indicate that Saskatoon and Elora environments were not correlated for regrowth. Moreover, the angles between the vectors associated with Sask 2019 and Sask 2020 and between Elora 2019 and Elora 2020 were acute, indicating that regrowth was positively correlated at each location across years. This was corroborated by the Pearson’s correlations (Table 5). Elora 2019 and Sask 2019 had the longest vectors, meaning that they were more capable of discriminating among genotypes than Elora 2020 and Sask 2020 (Fig. 6). The heritability of regrowth was moderate to high, with the highest estimate obtained for Elora 2019 (Table 6). Overall, regrowth was higher in 2019 than in 2020 at both locations, and there was evidence for transgressive segregation beyond both parents (Table 3). Winter damage and regrowth were not correlated in most environments, which suggested that regrowth and winter damage were inherited separately. However, regrowth and winter damage were highly correlated in Elora 2020 (Table 5). Noticeably, pest-induced defoliation recorded in Elora 2019 was highly correlated with regrowth in Elora 2020 (Table 5).

Distribution of field regrowth (RG) in four environments: Elora, ON, Canada in 2019 (Elora 2019) and 2020 (Elora 2020), and Saskatoon, SK, Canada in 2019 (Sask 2019) and 2020 (Sask 2020) in (A) the Rosa ×hybrida mapping population CA60 × Singin’ in the Rain [SITR (‘MACivy’)] and (B) among Rosa ×hybrida genotypes CA60, Caroline de Monaco [Cdm (‘MEIpierar’)], Easy Does It™ [EDI (‘HARpageant’)], ‘Frontenac’ (Fr), Gentle Giant™ [GG (‘WEKrigoyelo’)], ‘George Vancouver’ (GV), GV grown on its own roots (GV1), ‘Nicolas’ (Ni), and SITR. RG was estimated with the following formula: $\text{RG}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ln}}{\text{lt}}$, with RG being the percentage of regrowth, ln the length of the new shoot, and lt the initial length of the whole cane.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

View Full Size

Distribution of field regrowth (RG) in four environments: Elora, ON, Canada in 2019 (Elora 2019) and 2020 (Elora 2020), and Saskatoon, SK, Canada in 2019 (Sask 2019) and 2020 (Sask 2020) in (A) the Rosa ×hybrida mapping population CA60 × Singin’ in the Rain [SITR (‘MACivy’)] and (B) among Rosa ×hybrida genotypes CA60, Caroline de Monaco [Cdm (‘MEIpierar’)], Easy Does It™ [EDI (‘HARpageant’)], ‘Frontenac’ (Fr), Gentle Giant™ [GG (‘WEKrigoyelo’)], ‘George Vancouver’ (GV), GV grown on its own roots (GV1), ‘Nicolas’ (Ni), and SITR. RG was estimated with the following formula: $\text{RG}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ln}}{\text{lt}}$, with RG being the percentage of regrowth, ln the length of the new shoot, and lt the initial length of the whole cane.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

Distribution of field regrowth (RG) in four environments: Elora, ON, Canada in 2019 (Elora 2019) and 2020 (Elora 2020), and Saskatoon, SK, Canada in 2019 (Sask 2019) and 2020 (Sask 2020) in (A) the Rosa ×hybrida mapping population CA60 × Singin’ in the Rain [SITR (‘MACivy’)] and (B) among Rosa ×hybrida genotypes CA60, Caroline de Monaco [Cdm (‘MEIpierar’)], Easy Does It™ [EDI (‘HARpageant’)], ‘Frontenac’ (Fr), Gentle Giant™ [GG (‘WEKrigoyelo’)], ‘George Vancouver’ (GV), GV grown on its own roots (GV1), ‘Nicolas’ (Ni), and SITR. RG was estimated with the following formula: $\text{RG}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ln}}{\text{lt}}$, with RG being the percentage of regrowth, ln the length of the new shoot, and lt the initial length of the whole cane.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

QTL analysis for winter damage and regrowth was conducted separately for each environment because of the existence of significant GE interaction (Table 4, Supplemental Figs. 3 and 4). A QTL for winter damage in Elora 2019 mapped to LG2.H1 at 55 cM in the CA60 female map. This QTL alone explained 16% of the total phenotypic variance. A major QTL for winter damage in Elora 2019 mapped to LG2.H5 at 17 cM in the ‘SITR’ male map. A minor QTL for winter damage in Elora 2019 mapped to a fragment of LG5.H1 at 10 cM in the ‘SITR’ male map. The two-QTL model explained 30% of the phenotypic variance. A QTL for winter damage in Sask 2019 mapped to LG6.H1 at 20 cM in the ‘SITR’ male map and it explained 17% of the phenotypic variance. A QTL for winter damage in Elora 2020 mapped to LG1.H3 at 10 cM in the CA60 female map, explaining 24% of the phenotypic variation. No QTL for winter damage in Sask 2020 were detected. In addition, a QTL for regrowth in Elora 2019 mapped to LG7.H1 at 40 cM in the CA60 female map; it explained 14% of the phenotypic variance. A QTL for regrowth in Elora 2020 mapped to LG1.H3 at 20 cM in CA60—10 cM away from the QTL for winter damage in the same environment—and explained 37% of the phenotypic variance. A QTL for regrowth in Elora 2020 mapped to LG6.H4 at 60 cM in the ‘SITR’ male map, and it explained 15% of the phenotypic variance. A QTL for regrowth in Sask 2019 mapped to LG6.H2 at 20 cM in the CA60 female map, and it explained 18% of the phenotypic variance. A QTL for regrowth in Sask 2019 mapped to LG6.H1 at 20 cM in the ‘SITR’ male map, and it explained 18% of the phenotypic variance. No QTL for regrowth in Sask 2020 were detected. A QTL for defoliation that occurred in Elora in 2019 mapped to LG1.H3 at 15 cM in the CA60 female map and explained 36% of the phenotypic variance.

‘EDI’ × ‘GV’ population.

Winter damage and regrowth were recorded in one environment (Elora 2020). Winter damage was distributed around a mean of 58% with minimum damage of 12% and maximum damage of 92%, and regrowth was distributed around a mean of 89% with minimum regrowth of 53% and maximum regrowth of 148% (Fig. 8, Table 3). The control genotypes CA60 and ‘SITR’ and the parental lines ‘EDI’ and ‘GV’ differed, but their respective ratings were similar to the CA60 × ‘SITR’ multisite field trial (Table 3). The female parent ‘EDI’ exhibited substantial winter damage (85%), close to total dieback to the ground, whereas the male parent ‘GV’ presented minimal winter damage [34% (Fig. 8)]. Regrowth of ‘GV’ appeared greater than ‘EDI’ [96% and 70%, respectively (Table 3)]. There was evidence for transgressive segregation beyond both parents for both winter damage and regrowth (Table 3), with 7% of individuals that were as hardy as the hardiest parent ‘GV’ and 11% that were as sensitive or more sensitive than the tender parent ‘EDI’. The genotypic variance was significant for both winter damage and regrowth (Supplemental Tables 6 and 7). Both winter damage and regrowth were heritable (Table 6), but regrowth and winter damage were not correlated for the ‘EDI’ × ‘GV’ population (Table 5). QTL for winter damage and regrowth mapped to different linkage groups (LGs) of the rose genome (Table 4, Supplemental Figs. 5 and 6). A QTL for winter damage in Elora 2020 mapped to LG6.H2 at 10 cM in the ‘EDI’ female map; this QTL explained 15% of the variability observed. A QTL for winter damage in Elora 2020 also mapped to LG5.H2 at 65 cM in ‘GV’. This QTL explained 14% of the phenotypic variance. In addition, a QTL for regrowth mapped to LG2.H2 at 30 cM in the ‘GV’ male map and explained 14% of the observed phenotypic variation. A QTL for regrowth mapped to LG5.H2 at 50 cM in the ‘EDI’ female map and explained 14% of the phenotypic variance.

Distribution of (A) field winter damage (WD) and (B) regrowth (RG) in the Rosa ×hybrida mapping population Easy Does It [EDI (‘HARpageant’)] × ‘George Vancouver’ (GV) and in Rosa ×hybrida genotypes CA60, EDI, GV, and CA60 × Singin’ in the Rain™ [SITR (‘MACivy’)] at Elora, ON, Canada, in 2020. WD was a measurement of cane dieback in the field using the following formula: $\text{WD}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ld}}{\text{lt}}$, with WD being the percentage of winter damage, ld the length of dieback, and lt the length of the whole cane. RG was estimated with the following formula: $\text{RG}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ln}}{\text{lt}}$, with RG being the percentage of regrowth, ln the length of the new shoot, and lt the initial length of the whole cane. For both WD and RG, data on three canes per plant were collected and averaged.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

View Full Size

Distribution of (A) field winter damage (WD) and (B) regrowth (RG) in the Rosa ×hybrida mapping population Easy Does It [EDI (‘HARpageant’)] × ‘George Vancouver’ (GV) and in Rosa ×hybrida genotypes CA60, EDI, GV, and CA60 × Singin’ in the Rain™ [SITR (‘MACivy’)] at Elora, ON, Canada, in 2020. WD was a measurement of cane dieback in the field using the following formula: $\text{WD}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ld}}{\text{lt}}$, with WD being the percentage of winter damage, ld the length of dieback, and lt the length of the whole cane. RG was estimated with the following formula: $\text{RG}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ln}}{\text{lt}}$, with RG being the percentage of regrowth, ln the length of the new shoot, and lt the initial length of the whole cane. For both WD and RG, data on three canes per plant were collected and averaged.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure

Distribution of (A) field winter damage (WD) and (B) regrowth (RG) in the Rosa ×hybrida mapping population Easy Does It [EDI (‘HARpageant’)] × ‘George Vancouver’ (GV) and in Rosa ×hybrida genotypes CA60, EDI, GV, and CA60 × Singin’ in the Rain™ [SITR (‘MACivy’)] at Elora, ON, Canada, in 2020. WD was a measurement of cane dieback in the field using the following formula: $\text{WD}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ld}}{\text{lt}}$, with WD being the percentage of winter damage, ld the length of dieback, and lt the length of the whole cane. RG was estimated with the following formula: $\text{RG}\left(\text{\%}\right)\hspace{0.17em}=\hspace{0.17em}100\hspace{0.17em}\times \hspace{0.17em}\frac{\text{ln}}{\text{lt}}$, with RG being the percentage of regrowth, ln the length of the new shoot, and lt the initial length of the whole cane. For both WD and RG, data on three canes per plant were collected and averaged.

Citation: Journal of the American Society for Horticultural Science 147, 4; 10.21273/JASHS05189-22

• Download Figure