The Relationships between Plant Developmental Traits and Winter Field Survival in Rye (Secale cereale L.)

Overwintering cereals accumulate low temperature tolerance (LTT) during cold acclimation in the autumn. Simultaneously, the plants adjust to the colder season by making developmental changes at the shoot apical meristem. These processes lead to higher winter hardiness in winter rye varieties (Secale cereale L.) adapted to Northern latitudes as compared to other cereal crops. To dissect the winter-hardiness trait in rye, a panel of 96 genotypes of different origins and growth habits was assessed for winter field survival (WFS), LTT, and six developmental traits. Best Linear Unbiased Estimates for WFS determined from five field trials correlated strongly with LTT (r = 0.90, p < 0.001); thus, cold acclimation efficiency was the major contributor to WFS. WFS also correlated strongly (p < 0.001) with final leaf number (r = 0.80), prostrate growth habit (r = 0.61), plant height (r = 0.34), but showed weaker associations with top internode length (r = 0.30, p < 0.01) and days to anthesis (r = 0.25, p < 0.05). The heritability estimates (h2) for WFS-associated traits ranged from 0.45 (prostrate growth habit) to 0.81 (final leaf number) and were overall higher than for WFS (h2 = 0.48). All developmental traits associated with WFS and LTT are postulated to be regulated by phytohormone levels at shoot apical meristem.


Introduction
Rye (Secale cereale L.) is an annual grass of the Triticeae tribe within the Pooideae subfamily, which also includes barley (Hordeum vulgare L.), hexaploid wheat (Triticum aestivum L.), and several pasture grasses. The cultivated form of rye (Secale cereale ssp. cereal) is mainly grown in North America, Northern and Eastern Europe, Russia, and China, where the grain is used for animal feed or production of bread and alcoholic beverages. Rye can also be grown for biomass used as forage, green manure, or production of bioenergy [1,2]. Cultivation of rye is traditionally performed with open-pollinated breeding populations, but they are gradually being replaced by hybrid varieties producing higher yields due to heterosis effects [3]. Rye has a relatively high drought tolerance due to a well-developed root system and is therefore often cultivated on marginal land that is unsuitable for most other cereal crops [4]. The undemanding nature of rye is an important asset for future development of new varieties that will withstand the effects of climate change.
The growth habit of annual temperate cereals can broadly be divided into winter, facultative, and spring types, which differ in seeding time due to variation in vernalization and photoperiod sensitivities. The most commonly grown rye varieties are autumn-seeded, frost-tolerant winter types, which have a vernalization requirement prior to winter and flower when long days and warmer temperature return in the spring. Facultative types, eties are tall (about 120-150 cm), the photosynthetic area of stem becomes highly significant (60-80% of total) and stem is therefore considered the main producer of photosynthate for grain filling [4]. The top internode (peduncle) elongates last during stem growth and has a major influence on the final plant height (PHT) [41,42]. Tall plants are generally desired for biomass production [1], but less advantageous for grain production as long peduncles can cause lodging with negative effects on grain quality and yield. Reduction in plant height in rye by introgression of dwarfing genes, such as Ddw1 [43], have not resulted in similar yield increases obtained by semi-dwarf wheat [44], which commonly carries various height-reducing alleles [45,46]. Observations in winter wheat support semi-dwarfing genes do not negatively influence LTT [37]; thus, the search for new sources of height-reducing genes for winter rye continues.
In this study, a rye panel of 96 genotypes of diverse geographic regions, growth habits, and winter hardiness levels was analyzed for WFS during five years of field trials. The genotypes were also assessed for LTT (LT 50 values) and six developmental traits to characterize their association with WFS. The results showed that LTT contributed most to high WFS, and several of the plant developmental traits showed very strong association with WFS and LTT.

A Multi-Trait Approach to Study WFS
In this study, WFS in a rye population of 96 rye genotypes (Table 1) was studied during five years of field tests ( Figure S1). To dissect the main trait further, we assessed traits associated with WFS, which is a strategy that can be very informative when analyzing complex traits [47]. Therefore, the WFS studies were accompanied by determinations of LTT, FLN, PGH, days to anthesis (DTA), and PHT, which associate with winter hardiness in certain winter wheat genotypes [14,35,37]. Additionally, top internode length (TIL) and flag leaf area (FLA), which are not commonly related to WFS, were also analyzed. Phenotyping of LTT and the developmental traits is relatively easy to accomplish when compared to WFS, which can vary largely from year to year due to abiotic and biotic stress factors affecting plant survival from the early seedling stage in autumn to spring regrowth.

WFS for Rye Population Displayed Large Variations between Years
During the five-year field trials performed at Saskatoon, Canada, the snow cover and winter temperatures were highly variable between years (Table S1). The highest WFS was obtained during the 2015/2016 growing season (92.6% average WFS; Figure S1), which provided adequate snow cover and a relatively mild winter. In contrast, survival upon the 2017/2018 winter season was very low (6.8% average WFS) due to deeper and longer cold spells and occasional poor snow cover. The remaining trials in 2014/15, 2016/17, and 2018/19 provide a desired wide distribution of WFS values within the population. Despite the challenges with WFS phenotyping in certain years, most of the WFS data from the trials were significantly correlated ( Table 2) The BLUEs calculated for WFS from all five trials ranged from 0.0 to 92.5% (Table 1) with a mean of 47.7% ( Figure 1A). The correlation value between WFS_BLUEs and the WFS values for the five individual trials were highly significant (p < 0.001) and ranged from 0.59 to 0.91 and were overall higher than observed between the individual trials ( Table 2). Winter types, as expected, showed the highest WFS (WFS_BLUE = 0.0-92.5%; mean = 52.8%), followed by facultative types (20. [4], were placed in the low and very low WFS classes, respectively (Table 1).  1 LTT is defined as negative LT 50 value; 2 BLUE score determined from five WFS trials. ** significance of Pearson correlation coefficient p < 0.01; *** p < 0.001.

Freezing Tests Provided a Good Estimate of WFS Levels for the Rye Genotypes
One component of WFS is LTT built up during cold acclimation prior to winter and this trait was estimated by controlled freezing tests. Cold-hardy wheat cultivar Norstar used as a control in the tests showed an average LT 50 value of −21.4 • C, whereas higher LTT (lower LT 50 values) was noted for 59 out of 96 (61%) rye genotypes tested ( Table 1). Genotypes of the very high WFS class had the highest LTT (LT 50_mean = −25.4 • C) and a gradual decrease in average LTT was seen for the following WFS classes: high WFS (LT 50_ mean = −24.0 • C), moderate WFS (LT 50_ mean = −22.8 • C), low WFS (LT 50_ mean = −20.3 • C) and very low WFS (LT 50_ mean = −17.5 • C) ( Table 1). Like the WFS distribution, the absolute skewness and kurtosis values for LT 50 distribution were <1.0, which was indicative of continuous trait variation within the population ( Figure 1A,B).
The freezing test data for the rye population confirmed winter types developed the lowest LT 50 values (e.g., highest LTT) during cold acclimation (Table 1) (Table 1). Among the perennial genotypes, the Canadian ACE-1 developed for pasture and silage production [49] performed best, but did not accumulate adequate winter hardiness prior to winter to survive well on the Canadian Prairies (WFS_BLUE = 54.0%; LT 50 = −19.4 • C; Table 1). As expected, the lowest WFS and LTT were recorded for spring genotypes (WFS_BLUE ≤ 38.8%; LT 50 ≥ −21.4 • C; Table 1), for which seven out of eight genotypes did not survive the 2017/18 winter (data not shown). Overall, LTT determined for the rye population showed very high correlation with WFS_BLUEs (r = 0.90; p < 0.001; Table 3).

FLN and PGH Values Were Strongly Associated with WFS
Among the rye genotypes FLN_BLUEs varied from 7.7 to 12.4 leaves per plant with a mean 9.6 leaves ( Figure 2 and Figure S2; Table S2). High FLN (>11 leaves) was observed for 11 genotypes of the very high WFS class (19 genotypes), whereas all genotypes within the very low WFS class (19 genotypes) produced less than nine leaves (Table S2). Perennial, facultative and spring genotypes had overall lower FLN than winter types (Table S2). Rye genotypes Musketeer, AC Rifle, AC Remington with the highest FLN (≥12.0 leaves; Table S2) were among the most winter-hardy (WFS ≥ 83%) and developed highest LTT during cold acclimation (LT 50 < −27.0 • C) ( Table 1). In contrast, breeding lines Petkus and Carsten had relatively low FLN (8.0 and 8.6, respectively; Table S2) combined with low winter-hardiness and relatively high LT 50 values ( Table 1). The FLN values determined for the rye population were strongly associated (p < 0.001) with WFS (r = 0.80) and LTT (r = 0.71) values, respectively (Table 3).
For individual plants, PGH was determined by visual scoring of cold-acclimated plants for their growth habit according to a scale ranging from erect (1), intermediate (2), or clearly prostrate (3) ( Figure S3). Data from the four trials generated PGH_BLUE values ranging from 1.0 to 3.1 with an average of 2.1 for the rye population ( Figure 2 and Figure S3). The winter and perennial genotypes showed overall higher PGH scores (mean 2.1) than the facultative and spring types (mean~1.6; Table S2). PGH showed strong correlations (p < 0.001) with WFS (r = 0.61), LTT (r = 0.59), FLN (r = 0.43), and DTA (r = 0.43) ( Table 3).

Delayed Anthesis Time Was Weakly Associated with Higher WFS
The plants that had undergone vernalization under controlled conditions were used to determine DTA, by counting the days from the end of cold acclimation to the start of anthesis. Values ranging from 15.3 to 48.8 days with a mean of 31.4 days were obtained for the rye population ( Figure 2 and Figure S4; Table S2). With the exception of genotype SR4A-S5, the spring lines flowered early (DTA <26 days) and the longest delay to flowering was noted for the perennial genotypes, which all needed at least 34 days to reach anthesis (Table S2). DTA showed relatively weak association with WFS (r = 0.25, p < 0.05), but no significant association with LTT, and was negatively associated with both PHT (r = −0.29, p < 0.01) and TIL (r = −0.25, p < 0.05) ( Table 3). However, a relatively strong association (p < 0.001) was noted between DTA and PGH (r = 0.43; Table 3).  (Figures 2 and S3). The winter and perennial genotypes showed overall higher PGH scores (mean 2.1) than the facultative and spring types (mean ~1.6; Table S2). PGH showed strong correlations (p < 0.001) with WFS (r = 0.61), LTT (r = 0.59), FLN (r = 0.43), and DTA (r = 0.43) ( Table 3).

Higher WFS Was Associated with Genotypes Growing Tall
Data from four trials were used to calculate PHT_BLUEs and TIL-BLUEs ( Figure S5 and S6) and were found to be highly correlated (p = 0.001; r = 0.72; Table 3). The PHT_BLUEs varied from 62.7 to 144.4 cm with a mean of 121.4 cm and variation for TIL_BLUEs ranged from 10.5 to 46.5 cm with a mean of 35.5 cm (Figure 2; Table S2). TIL values constituted 15.6 to 38.3 % of the total height for the rye genotypes. Genotypes of short stature (PHT < 100 cm) were only identified among the spring and winter types (Table S2). Both PHT and TIL showed similar associations with WFS (r = 0.34, p < 0.001 versus 0.30, p < 0.01) and LTT (r = 0.39 versus 0.36, p < 0.001) ( Table 3).

Variation for FLA Did Not Relate to Winter-Hardiness
Field grown plants from four trials were used to determine FLA_BLUEs ( Figure S7) and showed a normal distribution with areas varying from 8.0 to 26.3 cm 2 and a mean of 16.9 cm 2 ( Figure 2; Table S2). FLA showed no significant (p < 0.05) association with any of the other traits studied (Table 3), and thus did not seem to be affected by factors controlling cold acclimation.

Bi-Plot PCA Supported WFS Is Primarily Determined by Developments at SAM during Cold Acclimation
In a PCA bi-plot analysis, the first two PCs determined 67.5% of the total variations for traits studied in the rye population ( Figure 3). PC1 accounted for 48.1% of the total variation and was mainly associated with WFS, LTT, FLN, PGH, and DTA. PC2 with 19.4% share of total variation was mainly associated with DTA. The PHT and TIL vectors indicated associations with both PC1 and PC2. WFS, LTT, FLN, PGH, and DTA vectors were closely spaced and directed in the same orientation on the bi-plot (Figure 3), which suggested high association as confirmed by the correlation analysis (Table 3). A near right angle between DTA and PHT/TIL vectors in the bi-plot suggested a negative association ( Figure 3) that was also supported by the correlation data (Table 3). However, the correlation test did not support a negative association between FLA and TIL as indicated by the bi-plot. Winter genotypes were positioned in all four quadrants ( Figure S8), with all of the very high WFS class positioned in the second and third quadrants (Figure 3). Genotypes with low or very low WFS class, including most spring lines, were positioned within the first and fourth quadrants. Perennial genotypes were clustered in the middle, and most facultative genotypes were positioned on the lower half of the plot ( Figure S8). FLN, PGH, and DTA vectors were primarily associated with the winter-hardy genotypes in second quadrant, whereas PHT and TIL vectors were mainly directed towards a subgroup of the winter-hardy genotypes positioned in the third quadrant. This subgroup included genotypes from Northern Europe such as Anna, Pearl, and Ponsi, which are relatively tall (>133 cm; Table S2). In contrast, most of the winter-hardy Canadian genotypes primarily positioned in the second quadrant appeared to rely more on PGH and FLN traits for their high WFS (Figure 3).

Heritability Estimates Show Genotype Had High Influence on Traits Analyzed
All traits studied showed significant (p < 0.001) effects of genotype (G), environment (E), and genotype x environment (G × E) interactions as revealed by an ANOVA analysis ( Table 4). The environmental factor was largest for WFS, due to the large variations in winter conditions during the trials ( Figure S1). Estimations of broad-sense heritability (h 2 ) generated values ranging from 0.45 to 0.84 for the different traits (Table 4)

Rye Population Studied Provided a Wide Variation of Winter-Hardiness Levels
Cultivation of winter cereals at Northern latitudes is highly dependent on plant survival during winter, which can drastically affect seed or biomass yields at maturity. Thus, studies on WFS are important for future expansion of winter rye growing areas and production volumes, but also for development of other winter cereals with improved winter hardiness. Winter rye is unique among the temperate cereal crops as it has adapted well to Northern climates, allowing cultivation in areas not suitable for wheat. Thus, the most winter-hardy rye genotypes in the study hold valuable information regarding winter survival during very cold winters in the northern latitudes in Asia, Europe, and North America. In contrast, the tender genotypes in the population were useful for identification of traits not associated with high WFS. A large amount of the frost tolerance variation among cereals depends on copy number and allele differences at the VRN1 and FR-2 loci, respectively, based on studies of diploid and hexaploid wheat [50,51] and barley [52,53]. Fr-R2 is confirmed as a major frost hardiness locus in rye [54,55], but very little is known about other loci affecting WFS or LTT in rye. In wheat, interactions between VRN1 and Fr-A2 alleles modulate cold-induced CBF gene expression that is critical for induction of COR genes and development of high LTT [20]. Thus, the poor WFS displayed by spring rye types in this study is likely due to the lack of vernalization requirement combined with the inability to induce COR gene expression to high levels when exposed to low temperatures in the autumn [56,57].

Efficiency of the Cold Acclimation Process Was a Major Factor for WFS
In the freezing tests, the LT 50 values were determined on plants cold acclimated under constant exposure to 4 • C temperature, low light intensity, and a short-day cycle. These growth room conditions do not include the frequent variations in light intensity and temperature that naturally occur in the field, which constitute environmental signals that can prime plants to increase their cold acclimation [58]. Such triggering events may include a short cold spell prior to a longer period of cold or exposure to slightly subzero temperatures, initiating a second hardening step [30,59]. Despite the differences in cold acclimation conditions under natural and controlled conditions, the determined LTT corre-lated relatively well with WFS determined from individual trials (r = 0.54-0.82; p < 0.001) and particularly well to BLUEs determined for WFS (r = 0.90; p < 0.001; Table 2). Thus, the efficiency of the cold acclimation process in the autumn was the major contributing factor to WFS in this study. Previous field tests of winter wheat genotypes performed in the Saskatoon region also demonstrate a very high correlation (r = 0.95) between LTT and WFS [37]. Saskatoon winters typically have low humidity levels, thin snow cover, and overall unfavorable conditions for snow mold infections, which can have a large negative impact on WFS [32,60]. In locations with long-lasting snow cover and humid conditions at ground level, the correlations between WFS and LTT are expected to be considerably lower than observed in this study.

Developments at SAM Were Closely Associated with WFS and LTT
The rye genotypes Musketeer, AC Rifle, AC Remington with the highest FLN (≥12.0 leaves; Table S2) were among genotypes with highest WFS (WFS ≥83%) and highest LTT (lowest LT 50 < −27.0 • C) ( Table 1). A low FLN demonstrated by breeding lines Petkus and Carsten (8.0 and 8.6, respectively, Table S2) suggested a short cold acclimation process underlies low WFS and LTT for these genotypes. FLN in the rye population was strongly associated (p < 0.001) with WFS (r = 0.80) and LTT (r = 0.71), respectively (Table 3). Thus, the majority of the variation for WFS among rye genotypes related to differences in the length of the cold acclimation period is similar to observations made in other cereals [56,61]. The switch to inflorescence meristem identity at SAM coincides with an up-regulation of VRN1 expression [9], and VRN1 allele difference is one of the factors determining duration of the vegetative phase in hexaploid wheat [51]. The formation of leaf primordia at the peripheral flank of SAM occurs at auxin maxima see review [62], a process modulated by the relative levels of cytokinin and gibberellin, which display antagonistic effects at SAM [63]. The phytohormone methyl jasmonate (MeJA) is also proposed to affect floral transition time in wheat [64]. Thus, floral transition is strongly affected by phytohormone levels.
The shoot curvature at the base of the crown was not permanently induced by cold as the rye plants reverted back to erect growth habit when growth resumed at normal temperature and long-day conditions. An altered negative gravitropism response resulting in asymmetric distribution of auxin in the shoot was proposed to cause PGH based on studies of gravity persistent signal (gps) mutants in Arabidopsis [65]. Like FLN, PGH is also associated with VRN1 or closely linked Fr-1 locus on chromosome 5A according to early studies of winter wheat [66]. Later studies implicated sensitivity to photoperiod in addition to vernalization sensitivity as the two major factors controlling PGH during juvenile growth in wheat [67]. A role for vernalization requirement was indicated in the study, where winter and perennial types with vernalization requirements developed a stronger PGH than spring and facultative types (Table S2). A role for phytohormone involvement in PGH is supported by studies in barley, for which recessive alleles at the sdw1/denso locus associated with gibberellin biosynthesis induced early prostrate growth in addition to semi-dwarf growth and delayed flowering [68,69]. As phytohormone levels are strongly implicated in the determination of floral transition time and prostrate growth habit during cold acclimation [62,64,70], the influence of VRN1 and/or CBF alleles on genes controlling phytohormone metabolism at SAM are likely to underlie some of the differences in FLN, PGH, and LTT observed for the rye population.

WFS and LTT Were Associated with PHT and TIL
The strong link between LTT and PHT (r = 0.59; p < 0.001), suggested PHT is affected by early events at SAM prior to floral transition. The intercalary meristems, from which leaf initials, axillary buds, and internodes are formed, are laid down during the early stages of SAM development [42], but stem elongation is paused until vernalization, temperature, and photoperiod requirements are fulfilled [71]. A signal transported from the shoot apex is suggested to control internode elongations starting with basal internodes elongating first and the peduncle last [42]. Gibberellin concentrations have a role in stem elongation by promoting increased cell division and cell elongations at the intercalary meristems of the stem [42,46]. The timing of stem elongation in the spring for winter type is important as early elongation can lead to frost kill due to exhausted LTT at this stage. Like FLN and PGH, a major locus for stem elongation in hexaploid wheat is located close to the VRN-A1 locus on chromosome 5A [71]. PHT and TIL, which showed lower correlation with WFS than FLN and PGH (Table 3), appeared to be related to WFS for only a subgroup of the rye population (Figure 3).

DTA Showed Strongest Association with PGH Displayed during Cold Acclimation
The development of flowers in Triticeae species with fulfilled vernalization requirement starts when photoperiod and temperature requirements are met, which occurs in the spring for winter cereals [72]. At this transition, the up-regulated VRN1 and absence of VRN2 expression in leaves leads to induced expression of VRN3 [73]. VRN3 is the cereal orthologue of Arabidopsis FLOWERING LOCUS T (FT), a transmissible florigen signal promoting flower development [74]. Besides variation for VRN alleles controlling vernalization requirement, allelic differences for PHOTOPERIOD1 (PPD1), circadian clockrelated earliness per se, early maturity, and gibberellin-regulated genes are some of the factors modifying flowering time in cereals (see review [75]). Like many developmental traits associated with WFS, gibberellin levels regulate development of inflorescence into flowers during long days [76]. Thus, the negative correlation between DTA and PHT (and TIL) observed for the rye population could be due to gibberellins stimulating both early flowering (low DTA) and stem elongation (high PHT/TIL). The strong association between DTA and PGH (r = 0,43; p < 0.001; Table 3) suggested genes affecting phytohormone levels at SAM during cold acclimation could also determine flowering time in the spring. Since the developmental traits induced during vernalization and associated with WFS are influenced by plant growth regulators [70], genes affecting phytohormone metabolism at SAM are of prime interest for future WFS studies.

Plant Material and Seed Production
A panel of 96 rye genotypes previously partially characterized for WFS [48] was used in the study ( Table 1) intensity, and 50% humidity to allow for seven weeks' cold acclimation. Acclimated plants were then transplanted into six-inch pots containing soil and slow-release fertilizers as described above and grown to maturity in a greenhouse set at 17/7 h light/dark regime, 1500 µE m −2 s −1 light irradiance, 23/18 • C day/night temperatures, and 50% relative humidity. During the active growth, liquid fertilizer (0.36 g L −1  was applied once per week. Pollination control was performed by placing five plants per genotype within a 1.2 m 3 pollination bag just prior to anthesis as described [77]. Seeds harvested from mature spikes were stored at room temperature until use.

Field Trials for Determination of WFS
Field trials of 96 rye genotypes (Table 1) were conducted at the University of Saskatchewan Experimental Farm, Saskatoon, Saskatchewan, Canada (52 • 10 N, 106 • 30 W, 457 m alti-tude) as previously described [48]. Climatological data were collected during the trials (Table S1).

Freezing Tests for Determination of LTT
To determine the LTT (negative LT 50 values), freezing tests were performed on the 96 rye genotypes. Cold-hardy winter wheat cv. Norstar was used as internal control. In this assay, plants were grown to four-leaf stage in a growth chamber and cold acclimated for five weeks at 4 • C as described above. A total of 60 cold-acclimated plants per genotype were then transplanted into five six-inch pots with 12 plants per pot and acclimatized in the dark to −3 • C for 12 h in an EPZ-4H test chamber (ESPEC North America Inc., Hudsonville, MI, USA). Thereafter, the chamber temperature was reduced at a rate of 2.0 • C h −1 . When the first test temperature was reached, one pot per genotype was removed from the freezing chamber and transferred to a 4 • C growth chamber for thawing. The remaining pots in the freezing chamber were removed one by one upon every 1.5 • C additional decrease in temperature. For genotypes with highest LTT, the five freezing test temperatures ranged from −24 to −30 • C, whereas test temperatures ranged from −13.5 to −19.5 • C for genotypes with the lowest LTT. The selected test temperatures were predetermined based on WFS data and small-scale freezing tests. Frost-exposed plants were maintained at 4 • C, 8/16 h day/night cycle, 120 µE m −2 s −1 light intensity, and 50% humidity for 20 h before being trimmed to one-inch height and transferred to a growth room set at 20/18 • C day/night temperature, 16/8 h day/night cycle, 50% humidity, and 250 µEm −2 s −1 light irradiance. The plants were fertilized once per week with NPK 20-20-20 (35 g L −1 ) during watering. After two weeks of regrowth, plant recovery was rated for each of the 60 plants per genotype using a scale from zero to five, where zero indicated no regrowth and five represented full regrowth from all tillers. Average survival scores obtained at each test temperature were plotted against freezing test temperature to generate a kill curve. The freezing temperature at which 50% of the plants survived was determined as the LT 50 value. The freezing tests were performed twice to determine an average LT 50 value for each genotype.

Collection of Phenotypic Data for Developmental Traits
The developmental traits FLN, PGH, PHT, TIL, and DTA were assessed on plants cold acclimated in a growth room and grown to maturity in a greenhouse. Each trait was examined in four separate trials with five plants per genotype in each test. To determine FLN, plant leaves were labeled numerically as they developed from the primary stem and the number of the final flag leaf on the main stem was recorded as FLN. PGH was rated for fully cold-acclimated plants by visual scoring of three different growth habits: (1) erect, (2) intermediate, or (3) prostrate. The number of days from the end of cold acclimation and first anther extrusion was recorded as DTA. At maturity, the length of the three longest stems per plant were measured from the soil surface to the top of the head with awn length excluded. The average PHT and TIL were recorded for the genotype. Flag leaf area (FLA) was determined on field-grown plants upon full extension of inflorescence and performed on three plants per genotype with five flag leaves sampled per plant. A LI-3000A Portable Area Meter connected to LI-3050A Transparent Belt Conveyer instrument (LI-COR Inc., Lincoln, NE, USA) was used for the FLA measurements.

Statistical Analyses
The phenotypic data collected were tested for normality using the Minitab 19 Statistical Software (Minitab, LLC, State College, PA, USA). Analysis of variance (ANOVA) was performed by the general linear model using the GEA-R software (Genotype x Environment Analysis with R for Windows) version 4.1 (CIMMYT Research Data & Software Repository Network El Batan, Mexico). Mean sum of squares from ANOVA were applied to calculate heritability (h 2 ) for each trait [78]. To determine the overall trait score for each genotype, the Best Linear Unbiased Estimates (BLUEs) were calculated for all the phenotype data collected [79]. The calculation was conducted using the statistical analysis software META-R (Multi Environment Trial Analysis with R) version 6.04 (CIMMYT Research Data & Software Repository Network, El Batan, Mexico). Correlation analyses and principal component analysis (PCA) were conducted using RStudio package version 3.5.1 software [80]. The first two components were used to create bi-plot illustrating the relationships between genotypes and measured traits.

Conclusions
This study of WFS over five years combined with studies of LTT and six developmental traits showed WFS was almost entirely determined by the cold acclimation process. Thus, variation in WFS among the rye genotypes could to a large extent be explained by differences in the number of leaf initials produced at SAM during cold acclimation, factors inducing PGH in the shoot, and possibly also factors controlling early internode development at SAM. FLA and DTA are traits developed in the spring in winter types and showed no or low association with WFS, but variation for the traits is expected to impact agronomic performance and grain yield. The factors underlying the highly correlated traits FLN, PGH, and maybe also PHT are important to understand development of LTT in winter cereals. A moderate to high heritability displayed by the developmental traits indicated genetics had a large influence on trait values. This also suggests that the highly heritable developmental traits can be used in the selection of lines with high WFS among autumn-seeded rye. With high heritability estimates, good segregation for trait values within the population, and available genome sequence assembly [81], the requirements needed for a genome-wide association study are met [82]. Further genetic studies are expected to reveal the genetic basis for WFS and associated developmental traits for future utilization to enhance WFS in other winter cereals and expand their cultivation in temperate regions.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/plants10112455/s1. Figure S1: Frequency distribution for WFS data obtained for rye population in five field trials. Correlation analysis is presented in Table 2; Figure S2: Frequency distribution and correlation analysis for FLN data obtained for rye population in four studies; Figure S3: Frequency distribution and correlation analysis for PGH data obtained for rye population in four studies. Phenotypes scored for PGH is shown; Figure S4: Frequency distribution and correlation analysis for days to DTA data obtained for rye population in four studies; Figure S5: Frequency distribution and correlation analysis for PHT data obtained for rye population in four studies; Figure S6: Frequency distribution and correlation analysis TIL data obtained for rye population in four studies; Figure S7: Frequency distribution and correlation analysis for FLA data obtained for rye population in four field trials; Figure S8: PCA bi-plot of rye genotypes based on PCA1 and PCA2 components and vectors of traits analyzed in the study. Genotypes with winter (W), spring (S), facultative (F), and perennial (P) growth habit are indicated; Table S1: Local climatological data during consecutive growing seasons 2014 to 2019; Table S2: Development trait data for different growth habit classes of rye.