The main effects of population and genotype were very highly significant (
p ≤ 0.001;
Table 2) for all traits evaluated in the study (the no. weeks in flower, width 1, width 2, height, stem length, stem length to 1st branch, proportion of stems unbranched, stem diameter, flower diameter, height:width ratio, semi-ellipsoid volume, eccentricity, circumference, base area, plant stature, boll dehiscence, no. underdeveloped seeds/capsule). This significant variation within and among populations and genotypes of flax is not unexpected, given the diversity of sampling as well as inclusion of many wild species. Most importantly, significant genotypic differences provide for directional selection of traits for creation of CF, OS, and HP ideotypes.
3.1. Floral Characteristics
Traits relevant to CF potential for floral designing and HP potential in landscapes for “flower power” [
38] include flower diameter (mm), weeks in flower, stem length (cm), and stem diameter (mm) (
Figure 1). Since main effects are significant for all CF traits, mean separations of pooled groupings show differences among the wild species as well as OS and CF selections (
Figure 1). Flower diameter, when pooled into species and ideotype groupings, was greatest in
L. grandiflorum and
L. lewisii, followed by CF, OS,
L. altaicum, and
L. austriacum (
Figure 1A). Neither the CF or OS selection population had significantly different flower diameters compared to
L. perenne,
L. austriacum,
L. altaicum, and
L. hirsutum, and both selections had significantly larger flowers compared to
L. baicalense,
L. bienne,
L. pallescens, and
L. usitatissimum. Flower diameter differed greatly among populations on a genotype mean basis; the smallest diameter flowers (8–15 mm) occurred in
L. bienne and
L. usitatissimum (
Figure 2). The significantly largest flowers occurred in three CF accessions (R-2.2.2DT-clone, S-294-3DT-clone, S-272-5DT-clone), two
L. lewisii (Ames 33354, Ames 31364), and one OS genotype (S-293-5DT clone;
Figure 2). On a genotype mean basis, a large range of flower diameters is observed for both selection populations, ranging from ~20 to −30 mm;
L. bienne and
L. usitatissimum were comparable in size and significantly smaller than any of the other populations (
Figure 2). The CF selections alone exhibit a range of genotype mean values which encompasses all genotypes within populations of
L. altaicum,
L. austriacum,
L. grandiflorum,
L. hirsutum,
L. lewisii,
L. narbonense, and
L. perenne. Such wide-ranging variation highlights the opportunities for selection within this population, as well as the challenge of achieving a consistent response to selection within a highly outcrossing species.
Given the shared evolutionary history of
L. bienne and
L. usitatissimum [
5,
39], flower size may not have undergone strong selective pressure over time as
L. bienne gave rise to domesticated
L. usitatissimum. Among breeding populations, while intentional selection for larger flower diameter had little impact overall with the CF selections’ mean flower diameter being slightly (non-significantly) greater than OS selections (
Figure 1A). This may be partially due to CF populations being derived from a single cycle of selection in 2018, whereas three cycles of selection for OS traits have occurred since 2005. The existence of CF genotype means with significantly wider flowers than most OS types (
Figure 2) allows for directed breeding and selection for CF improvement. Likewise, these CF genotypes would be useful in creating HP cultivars.
3.1.1. Flowering
A significantly greater number of weeks in flower was observed for the CF and OS selections compared to all wild species besides
L. grandiflorum and
L. austriacum (
Figure 1B). The shortest flowering periods were observed for
L. altaicum,
L. baicalense, and
L. lewisii, all of which were significantly less than the other populations; the remaining populations were intermediate to these extremes. As with flower diameter,
L. usitatissimum and
L. bienne exhibited similar responses in the number of weeks in flower. On a genotype mean basis, the range of flowering was widespread across all genotypes, ranging from 0 (non-flowering)–12.8 wks. (
Figure 3). A lengthy ~13 wk. flowering period is unprecedented for HP that routinely flowering over a short period of time or reflowering after cutback or deadheading. This demonstrates the potential perennial flax has of becoming an annualized HP for the market [
40]. Non-flowering types indicate the need for a cold treatment for Yr. 2 flowering unless these are bred for Yr. 1 blooming as annualized perennials [
40]. Genotype mean differences also revealed a few key trends regarding the level of variation present within flax populations. Most notably, apart from two OS genotypes, all of the CF and OS selections exceeded the grand mean of 6.9 wk. in flower (
Figure 3). This contrasts with
L. austriacum,
L. perenne, and
L. bienne, which had genotypes in the 0–3 range, but also genotypes with >10 wk. in flower, on average. A distinct cluster of
L. lewisii is observed on the low end of the range and all but two
L. lewisii genotypes were below the grand mean (
Figure 3).
3.1.2. Inflorescence and Stem Traits
The longest mean stem length was observed in
L. pallescens (
Figure 1C). This was significantly longer than all other populations besides
L. lewisii, the CF and OS selections. As such,
L. pallescens could be an excellent source for taller CF and HP types as well as for fiber. Overall, less variation was observed for stem length relative to other traits, with many populations exhibiting overlapping means, such as
L. altaicum,
L. austriacum,
L. grandiflorum,
L. hirsutum,
L. lewisii,
L. perenne,
L. usitatissimum, and OS selections. Visual selection for improved stem length was evident for the CF selections, which had significantly longer stems compared to
L. baicalense,
L. bienne,
L. grandiflorum,
L. hirsutum, and
L. usitatissimum.
Stem diameter is a complex trait affecting CF and fiber and, to a lesser extent, for OS and HP perennial flax. Large diameter stems are stronger, largely due to their greater cross-sectional area, and are, therefore, desirable for CF applications as well as HP types [
41]. Several species in the present study stand out for their increased stem diameters relative to other populations. The largest stem diameters, on average, were observed for
L. usitatissimum and were significantly greater than all other populations besides
L. baicalense and
L. grandiflorum (
Figure 1D). These latter two populations were observed to have larger diameters compared to all except
L. bienne. While
L. pallescens had notably thin stems, it did not differ significantly from the majority of populations tested (
Figure 1D). Both CF and OS populations were very similar for stem diameter, although OS selections were observed to have slightly thicker stems.
3.1.3. Petal Overlap
Petal overlap (“gappiness”) did not differ significantly from a 1:1 χ
2 for
L. altaicum,
L. grandiflorum,
L. pallescens, and OS selections (
Table 3), possibly due to low sample sizes. In cases where petal overlap differed significantly from the 1:1 ratio, only
L. bienne had a majority of flowers with <50% overlap. For populations with a majority of flowers with >50% overlap, the two highest test statistics were observed for CF selections and
L. hirsutum. For
L. austriacum and
L. perenne, even though the majority of flowers had >50% overlap, both phenotypes were observed, and
L. perenne was just under the threshold of significance. Most importantly, the improved ornamental quality of the CF selections relative to the OS selections is demonstrated.
3.1.4. Flower Shape
Bowl shaped flowers were the most frequently observed flower shape and all populations differed significantly from the 1:1:1 χ
2 except for
L. baicalense and
L. bienne (
Table 4).
Linum bienne is the only species tested with an equal distribution of flower shapes. Overall, tube-shaped flowers were the rarest shape, which can generally be considered a positive finding, as such a tightly bound corolla would have little ornamental value. Tube shaped flowers occurred only in
L. austriacum,
L. bienne,
L. usitatissimum, and CF selections [
2]. Bowl shaped flowers, the most desirable ornamental phenotype, were generally the most common, especially among
L. austriacum and OS selections. Other populations, including
L. grandiflorum,
L. perenne, and CF selections had a relatively large proportion of funnel-shaped flowers.
3.1.5. Flower Color
Flower color proportions for each
Linum species and populations show distinct differences in color distributions for blue, red, and/or white tints, tones, and shades, based on RHS color codes and hexRGB values (
Figure 4) [
34,
35,
36]. The fewest number of flower color codes within a species/population occur in
L. altaicum (
Figure 4A) with 3 codes, and 4 codes occurring in
L. usitatissimum (
Figure 4J). Increasing numbers of colors range from 5 (
L. baicalense,
L. bienne,
L. pallescens,
Figure 4C,D,H, respectively) to even higher occurrences of flower color codes were in the remaining species and populations:,
L. perenne L. austriacum,
L. grandiflorum,
L. hirsutum,
L. lewisii, Selections–CF, and Selections–OS (
Figure 4B,E–G,I,K,L). Flower color ratings were done by multiple, trained researchers. This, coupled with the large numbers of samples to evaluate prevented collection of flowers and placement in standard environmental conditions each day for ratings, meant that these ratings could have minor levels of variation in the tints (hue + white), tones (hue + grey) or shades (hue + black) categorization of each flower color, while the hue would have remained the same. Nonetheless, these initial ratings demonstrate the wide range of variation for flower color present within and among the species and populations for use in future flower color breeding and enhancement of CF and HP ideotypes. Future flower color evaluations need to be conducted with smaller sample sizes to ensure their ratings can be accommodated in a standard light environment. Digital imaging for characterizing flower color, floral patterns and UV signals pre- and post-pollination, may further aid our understanding of plant-pollinator interactions [
37,
42,
43] once a pollinator ideotype is developed for perennial flax.
3.1.6. Styly
The observed style morphs for most species matched previous reports, if applicable, e.g.,
L. bienne,
L. pallescens,
L. perenne, and
L. usitatissimum (
Table 5). Mismatches in observed vs. reported style morphs occurred for
L. austriacum,
L. grandiflorum,
L. hirsutum and
L. lewisii. The style morph for
L. grandiflorum had been identified as polymorphic distylous (
Table 5). However, only homostylous flowers were observed in the present study, but homostylous, pin, and thrum flowers have been detected (
Figure 5). To the best of our knowledge, this is the first known report of a homostylous
L. grandiflorum. Further research into the cross compatibility of the three flower morphs is needed to determine fecundity (seed set) among cross-compatible and cross-incompatible style and stamen length differences [
21,
23]. Similarly,
L. hirsutum is reported to be polymorphic distylous, yet two of the studied individuals produced homostylous flowers, and the population overall supports a 1:1:1 distribution of flower morphs (
Table 5).
The second style morph inconsistency was observed in
L. lewisii. Multiple reports have identified that
L. lewisii is monomorphic approach herkogamous, possessing pin flowers (
Table 5). In dichotomous keys, style morph is one of the main distinguishing features between
L. perenne and
L. lewisii, which are otherwise almost identical in appearance [
23]. For this reason, it was surprising to find that 14/46 observations in
L. lewisii were ‘thrum’ flowers. This either means previous misidentification and/or questions the purity of
L. lewisii germplasm examined herein which will be addressed in future molecular studies. Likewise, such misidentifications have necessitated phenotyping styly morph upon first flowering in every new genotype under evaluation to provide additional confirmation or questioning of species identity.
While the species background of the CF and OS selections is unknown, these populations appear to be polymorphic distylous (
Table 5). Data did not deviate from a 1:1 χ
2, and no homostylous flowers were observed.
Linum altaicum (distylous) and
L. baicalense (distylous and homostylous;
Table 5) have not been previously studied for style morphs. Self and cross pollination studies will be needed to determine linkage with SI (
L. altaicum) and/or SC (
L. baicalense). Additional germplasm needs to be screened before clear styly trait association of the species can be made.
3.3. Plant Dimensions
The CF selections had width 1 dimensions that were significantly greater than all populations except for OS selections, which exhibited similar, though slightly smaller, width 1 values (
Figure 7A). The OS selections were also significantly greater than
L. altaicum,
L. baicalense,
L. bienne,
L. hirsutum,
L. lewisii, and
L. usitatissimum for width 1. The greatest width 1 observed among wild species was in
L. perenne and
L. austriacum, which had comparable mean values that were both significantly greater than
L. baicalense,
L. hirsutum, and
L. usitatissimum. The smallest width 1 belonged to
L. baicalense (
Figure 7A).
The CF selections were also significantly larger than all populations for width 2 except the OS selections, which overlapped in distribution (
Figure 7B).
Linum baicalense was observed to have a significantly smaller width 2 than most species, excluding only
L. hirsutum and
L. usitatissimum. Among the wild species,
L. perenne had the greatest width 2 measurements, on average, being significantly greater than
L. baicalense,
L. hirsutum,
L. lewisii, and
L. usitatissimum (
Figure 7B). Wider plants would be desirable for HP selections with equal widths 1 and 2 to creating spherical plant habits, rather than for either CF (upright) or OS types (funnel form or V-shaped for mechanical cutting) [
1,
3].
The greatest average height was observed in CF selections, followed by OS (
Figure 7C). Both
L. usitatissimum, the CF and OS selections were significantly taller than
L. baicalense,
L. bienne,
L. hirsutum, and
L. lewisii. The remaining species
L. altaicum,
L. austriacum,
L. grandiflorum,
L. pallescens, and
L. perenne were all similar in height (
Figure 7C). This trait would be most advantageous for CF and fiber types, rather than HP or OS.
The integration of all three plant size measurements (width 1, width 2, height) into the calculation of semi-ellipsoid volume (cm
3) illustrates the large differences in the size of the populations tested (
Figure 7D). With this measurement, the CF selections are more than double the overall size of all species populations, besides
L. austriacum and
L. perenne, and they are significantly larger in volume than all populations except for overlapping with OS (
Figure 7D). The small plant volume of
L. lewisii is surprising (
Figure 7D), given its close phylogenetic relationship with
L. perenne and
L. austriacum [
5,
23]. One possibility is that the climate and soil type of Minnesota is unfavorable for this species, even though Minnesota lies on the eastern border of its native range [
45].
Linum lewisii is often found in alpine regions, although it is also known to be a native component of prairie grasslands. The wide range of biomes included in its native range suggests that adapting
L. lewisii to MN for the purposes of developing a native perennial crop may still be feasible, if a collection can be located from a similar region as the target environment.
The OS selections have similarly large semi-ellipsoid volumes (cm
3), at nearly twice the average size of
L. grandiflorum; significantly larger than all populations besides
L. austriacum and
L. perenne (
Figure 7D). This highlights the significant impact that even 1–3 generations of selection can have on vigor and adaptation of perennial flax to a new environment and bodes well for making progress in HP and fiber ideotype selection.
A larger plant volume was expected for
L. grandiflorum given its annual life cycle and reported ornamental value. While it did flower profusely throughout the season (
Figure 3), its volume did not differ significantly compared to most of the other species studied (
Figure 7D). This could indicate that the genotypes tested were poorly adapted to the local environment.
For evaluating ornamental value of potential HP selections, it is also informative to consider the plant base area and circumference, which exclude plant height from the calculation (
Figure 8A,B). Wider plants would ensure soil line coverage in the landscape for HP types. In general, the same relative pattern as a semi-ellipsoid volume is observed except that
L. lewisii and
L. bienne appear larger, potentially indicating that their size comes more from their width than their height. As with volume, the largest base area and circumference is found with CF selections, which overlap with OS types, and both are significantly greater than all other populations (
Figure 8A,B). It is also notable that the visual differences between populations appear greatest for semi-ellipsoid volume (
Figure 7D) compared to base area and circumference (
Figure 3B and
Figure 8A), yet the circumference shows a greater number of significant mean separations between populations. For example,
L. baicalense has a significantly smaller circumference compared to
L. altaicum,
L. austriacum,
L. bienne,
L. grandiflorum,
L. pallescens,
L. perenne, CF and OS selections. A similar outcome occurs in
L. usitatissimum, which has a significantly smaller circumference and base area relative to
L. austriacum,
L. grandiflorum,
L. perenne, CF selections, but a semi-ellipsoid volume which only differs significantly from the CF and OS selection populations. For HP development, CF, OS,
L. perenne, and
L. austriacum would provide the greatest selection potential for both plant base area and circumference.
Ultimately, the choice of plant size measurement must relate to the breeding and selection goals of each ideotype. If the goal is to select for the largest or most compact plants overall, then semi-ellipsoid volume is superior, since it integrates all three size measurements. If the goal is a wide or narrow base, irrespective of height, then base area or circumference are the measurement(s) of choice. Base area may be a superior calculation criterion, as it integrates both widths 1 and 2, rather than using an average width to calculate circumference. However, based on this study, circumference creates greater statistical differences among populations.
Measurements characterizing plant shape may also prove useful for selecting for ornamental quality and growth vigor. The height to width ratio observed for
L. usitatissimum was significantly larger than all other populations in the study, indicating that this species was much taller than it was wide (
Figure 8C). In contrast,
L. bienne grew much wider than it was tall, with a height to width ratio that was significantly less than
L. altaicum,
L. hirsutum,
L. lewisii,
L. pallescens, and
L. usitatissimum. Anecdotally,
L. bienne was often observed to have nearly horizontal growth, although a few genotypes did grow into an upright cushion shape. The height to width ratio of CF and OS selections was also low relative to the other populations tested, although these populations did not differ significantly from any others besides
L. usitatissimum.
Linum perenne is the wild species with the most similar height to width ratio as the selections. The remaining species had height to width ratios in the range of 0.6–0.8.
The goal of testing eccentricity is to determine which population has the most spherical footprint. This calculation is most relevant to HP uses where compact, ‘cushion’ shaped plants are generally the goal. Eccentricity values approaching zero indicate a nearly perfect sphere, while values approaching one indicate an increasingly elliptic shape. The most spherical, on average, out of all the populations were the CF selections, which had significantly lower eccentricity values than
L. baicalense,
L. hirsutum,
L. lewisii,
L. pallescens, and
L. usitatissimum (
Figure 8D). The same result occurs for the OS selections, except that this population did not differ significantly from
L. hirsutum. Overall, in addition to being larger in size, both selection populations exhibited a more uniform and cushion shaped growth habit which would be desirable for the HP and bedding plant industry. Of all the wild species,
L. perenne exhibited the most uniform growth habit with a mean eccentricity value of 0.5, which was significantly less than
L. baicalense,
L. lewisii, and
L. pallescens. Of these latter three species,
L. baicalense was the most elliptic, due to an average width 1 that was nearly double the average of its width 2.
Stem length to the first branch is an important trait for CF and OS production, enabling ease of stem harvest by hand (CF) or mechanically (OS). The shortest stems, which would be useful for bedding plant production of HP lines, as well as enhanced branching for HP production in cultivated settings, were found in
L. bienne,
L. grandiflorum, and
L. usitatissimum (
Figure 9A). The longer distance from the crown (soil line) to branching occurred in
L. lewisii, which overlapped with
L. baicalense,
L. perenne, and CF/OS selections (
Figure 9A). Thus, selection has been effective in both OS and CF types; similar development using other species will be beneficial.
The proportion of unbranched stems (length to first branch >5 cm/total stem length) were the highest in
L. baicalense (~0.6) and significantly greater than all other species and selected populations (
Figure 9B), indicating that this species has less basal branching than the others. Most other species and populations were ~0.5 with the significantly lowest proportions occurring in
L. bienne and
L. grandiflorum (
Figure 9B). While branching is highly desirable in both HP and CF ideotypes, it depends on where it occurs (
Figure 9A). Lower proportions of unbranched stems are desirable in the HP flax crop ideotype whereas higher proportions would be ideal for the CF ideotype. Crop development using
L. bienne or
L. grandiflorum will focus selection on higher branching capacity.
3.4. Plant Survival
The highest % summer survival (stand establishment to the end of the growing season) was observed for OS selections (93.2%;
Table 6), followed by wild
L. perenne (90.5%) and CF types (88.7%), which confirms that these selections readily adapt to transplanting with high stand establishment.
Linum austriacum,
L. pallescens, and
L. altaicum also had 81.5%, 72.3% and 70.7% survival, respectively (
Table 6). The lowest % summer survival was for annual flax,
L. usitatissimum, at 24.4% (
Table 6), which is a seed-propagated crop (direct seeded). This was not due to senescence, as plants were considered to have survived the summer through seed set. The most likely cause for this high rate of mortality may have been transplant shock, since it is direct seeded in commercial plantings [
46].
Linum baicalense and
L. grandiflorum also showed relatively poor summer survival, suggesting that significant effort would be required to adapt these species to transplanting. However, it would be interesting to study whether the same outcome was observed through direct seeding. For
L. bienne,
L. hirsutum, and
L. lewisii summer survival was between 50–60% (
Table 6), indicating that these are poorly adapted, but may be improved through several cycles of breeding.
Winter survival also varied greatly among the populations tested (
Table 6). The greatest percent winter survival was observed for the OS selections (97.4%), which confirms that these genotypes are USDA Zone 4 (Z4) hardy and well adapted to MN winters.
Linum perenne exhibited similar winter hardiness (94.3%), followed by CF selections at (91.5%;
Table 6). The species
L. altaicum,
L. austriacum, and
L. hirsutum all had winter survival > 70%, which suggests that they have the potential to be fully Z4 hardy following additional selection. Subsequent research has shown variation in species and CF/OS selections for winter survival comparisons of controlled freezing tests with field survival [
24], indicating that perennial flax is selectable for CF, OS, HP and fiber ideotypes. The low winter hardiness observed for
L. lewisii was surprising given its wide distribution over alpine and plains regions, which stretches as far north as Alaska and Canada [
45]. This result may be due to the relatively poor vigor of the genotypes tested, as evidenced by their small semi-ellipsoid volume (
Figure 7D) and poor summer survival (
Table 6). Likewise, year to year differences in snow cover could have impacted the survival of the species. Of the species native to the steppes of Asia and Siberia [
16],
L. pallescens performed better than
L. baicalense; the latter of which exhibited only 6.9% survival (
Table 6), indicating that
L. baicalense is only marginally hardy to USDA Z4. Despite its specific epithet,
L. bienne behaved more like an annual in Z4, flowering and setting seed in the first year of growth and exhibiting only 0.5% winter survival (
Table 6). As expected, the annual species
L. grandiflorum and
L. usitatissimum had 0% winter survival (
Table 6).
3.5. Trait Correlations
The number of weeks in flower had highly significant positive correlations with several traits, including height, width 1, width 2, semi-ellipsoid volume, circumference, and base area (
Table 7). Of these, the highest correlation coefficient observed was for width 2 (r = 0.432). Additionally, there was a significant (
p ≤ 0.05) correlation with flower diameter. Weeks in flower showed highly significant negative correlation with eccentricity; significant (
p ≤ 0.01) negative correlation with height to width ratio; significant (
p ≤ 0.05) negative correlation with stem diameter (
Table 7).
Not surprisingly, height showed a highly significant positive correlation with other size measurements, including width 1, width 2, height to width ratio, semi-ellipsoid volume, circumference, and base area (
Table 7). Of these, the highest correlation coefficient was observed for semi-ellipsoid volume (r = 0.692). A highly significant positive correlation was also observed between height and stem length (r = 0.506). This correlation coefficient would be expected to increase as progress is made in selecting for upright growth habit. A significant (
p ≤ 0.01) positive correlation between flower diameter and height was observed, as well as a highly significant negative correlation between height and eccentricity.
Similar to height, both width 1 and width 2 had highly significant positive correlations with all other plant size measurements, with the exception of height to width ratio and eccentricity, both of which displayed highly significant negative correlations (
Table 7). The highest correlation coefficients observed for both widths 1 and 2 were for circumference (r = 0.969 and r = 0.968, respectively). This is not surprising, as the circumference measurement was calculated using the average of widths 1 and 2. Additionally, there were highly significant positive correlations between widths 1 and 2, stem length, and flower diameter; as well as negative correlations between widths 1 and 2 and stem diameter.
Height to width ratio had highly significant negative correlations with circumference and base area (
Table 7), indicating that, as height increased, width also tended to decrease. There was also a significant (
p ≤ 0.05) negative correlation with semi-ellipsoid volume. A significant (
p ≤ 0.01) positive correlation was observed between height to width ratio and eccentricity, as well as a significant (
p ≤ 0.05) positive correlation with stem length.
There was a highly significant negative correlation observed between semi-ellipsoid volume and eccentricity (
Table 7), possibly due to larger selection populations, which also had the lowest eccentricity values. Semi-ellipsoid volume also displayed highly significant positive correlations with base area, circumference, and stem length, as well as a significant (
p ≤ 0.05) positive correlation with flower diameter (
Table 7). A significant (
p ≤ 0.01) negative correlation between semi-ellipsoid volume and stem diameter was observed.
Highly significant negative correlations were found between eccentricity, base area, and circumference (
Table 7), potentially for the same reasons as correlation of eccentricity and semi-ellipsoid volume. Interestingly, there was also a significant (
p ≤ 0.05) positive correlation between stem diameter and eccentricity, indicating that more elliptic plants tended to have thicker stems.
Circumference had a highly positive correlation with base area (r = 0.970;
Table 7), indicating that these are largely duplicative metrics for plant size irrespective of height. Both circumference and base area also showed highly significant positive correlations with stem length and flower diameter, and highly significant negative correlations with stem diameter. There was also a significant (
p ≤ 0.01) negative correlation between stem diameter and flower diameter observed. Finally, stem length had a significant (
p ≤ 0.05) negative correlation with stem diameter, and a highly significant positive correlation with flower diameter. As CF selection progresses, an even higher correlation between these traits should be expected. The degree of positive correlation between stem length and flower diameter could potentially be a good method of measuring progress towards the CF ideotype [
1].
3.7. Ornamental Potential of Wild Flax Species
As noted by Cullis [
14] and Fu [
15], there is a lack of formal research characterizing the ornamental potential of wild flax species for CF and HP use. Aside from taxonomic descriptions [
6], only one other study compared trait values of wild
Linum species [
18]. This study evaluated five perennial (
L. austriacum,
L. hirsutum,
L. narbonense,
L. perenne,
L. thracicum) and six annual (
L. angustifolium,
L. bienne,
L. hispanicum,
L. crepitans,
L. grandiflorum,
L. pubescens) for a suite of traits. The results of Poliakova and Lyakh [
18] are compared with the present study in cases where there was overlap between the traits and species tested, namely
L. austriacum,
L. hirsutum,
L. perenne,
L. bienne, and
L. grandiflorum. Flower diameters observed herein differed slightly from those reported by [
18], particularly for
L. grandiflorum and
L. hirsutum, but also for
L. austriacum,
L. perenne and, to a lesser extent,
L. bienne. In general, the flower diameter measurements by Poliakova and Lyakh [
18] exceed those reported in the present study (
Figure 1A). Plant heights reported herein were lower (
Figure 7C) while the plant width values were greater (
Figure 7A,B) than those reported [
18]. Most of the germplasm tested by Poliakova and Lyakh [
18] was sourced from the N.I. Vavilov Research Institute of Plant Industry (VIR) and the All-Russian Research Institute for Flax (VNIIL), with the exception of
L. austriacum and
L. hirsutum, which were collected from their native range in the southern steppe of Ukraine. Several years of breeding work on
L. grandiflorum at Zaporozhye National University to develop cultivars with different flower colors and shapes were also performed [
18,
51]. These studies also differed in the age of germplasm tested (three years of growth) in comparison with the present study for year one trait values. Altogether, these differences in accession origin, testing environment, and age of plants may explain differences among Poliakova and Lyakh [
18] and the present study. Still, the results raise questions about the range of genetic and phenotypic diversity captured by wild flax collections at VIR, VNIIL, USDA-GRIN, and GRIN-CA. We anticipate examining this diversity with molecular analyses of single nucleotide polymorphisms (SNPs) in these flax species and populations.
The best wild species to incorporate into CF breeding varies, depending on the trait(s) of interest. For flower diameter and stem length,
L. lewisii is the best overall (
Figure 1A,C). However, the short flowering period of
L. lewisii is a significant drawback, as stem yield would low if the plants only flower for 4 wk. throughout the summer (
Figure 1B). Given these constraints,
L. austriacum is the most promising species overall, as it has large flower and stem diameters (
Figure 1A,D), the longest flowering time of any wild perennial (
Figure 1B), and relatively long stem length (
Figure 1C). As previously discussed,
L. pallescens has some favorable CF traits, but it lacks adequate flower and stem diameters (
Figure 1A,D).
Linum grandiflorum also has CF potential, as it has large flowers, a long flowering time, acceptable stem lengths, and thick stems (
Figure 1A–D). Selection among these species to enhance current CF genotypes with superior postharvest vase life [
2] will continue to improve the domestication of CF perennial flax for the market.