• Open Access

Mapping of quantitative trait loci controlling timing of bud flush in Salix


Vasilios Tsarouhas, Department of Plant Biology, Swedish University of Agricultural Sciences, Box 7080, SE-750 07, Uppsala, Sweden. E-mail: vasilios.tsarouhas@vbiol.slu.se


Dormancy release is an important phenological stage, which determines plant growth and survival in northern temperate regions. Spring bud flushing was studied in a Salix pedigree (n=82) derived from a cross between the male hybrid clone “Björn” (Salix viminalis×Salix schwerinii) and the female clone “78183” (Salix viminalis). The timing of bud flush was recorded outdoors in two consecutive years (1998, 1999) and indoor in the spring of 1998. Timing of bud flush was found to be under moderately strong genetic control (clonal mean heritabilities ranging from 0.43 to 0.72). Phenotypic correlations between height growth and bud flushing were negative but non-significant (r=0.1–0.3). Using a Salix linkage map composed of 325 AFLP and 38 RFLP markers, six quantitative trait loci (QTLs) and three unmapped marker loci associated with timing of bud flush were detected. Four QTLs were detected in the field experiment while two QTLs and three unmapped marker loci were identified in the indoor experiment. One QTL associated with indoor bud flushing coincided with one of the QTL detected from the field data. Individual QTL explained 12–24 % of the phenotypic variance. None of the bud flush QTLs coincided with QTLs controlling height growth identified previously in the same pedigree.

Injury due to low temperature is considered one of the most serious problems in the cultivation of agricultural crops and in forestry, affecting yield, quality and survival. After dormancy release in spring, the buds and the newly flushed shoots are particularly vulnerable to low temperatures. Spring frost damage has received more attention lately due to an increased incidence of warm episodes in spring (e.g. warm March followed by a cold April) (Cannell and Smith 1986). According to the European International Phenological Garden records, onset of spring phenophases is now 6 days earlier than in the 1960s (Menzel and Fabian 1999; Beaubien and Freeland 2000). Furthermore, the predicted global warming could induce earlier flowering and bud flushing in plants worldwide. It has been estimated that a 3°C temperature increase would advance bud flushing in Fagus sylvatica by 11 days (Kramer 1996). Therefore, frost injury of plants, due to an earlier dormancy release, could become an even more serious problem in the future for both agricultural and forestry practices.

One strategy that could contribute to the reduction of spring frost damage is the development of late flushing genotypes through selection and breeding. In the last few decades, timing of bud flush in forest tree species has been of practical interest for selection and breeding, especially in northern countries (Eriksson and Lundkvist 1986; Hannerz 1999). Observations from Salix field trials indicated considerable species and clonal differences in time of bud flushing (von Fircks 1994; Rönnberg-Wästljung and Gullberg 1999). These studies also showed that timing of bud flush is not inherited in a simple Mendelian mode, but rather exhibits an additive genetic variation typical of a quantitative trait (Rönnberg-Wästljung 2001). Moderate to strong genetic control in timing of bud flush is well-documented among and within families in conifers (Aitken and Adams 1997) and broad-leafed species like poplars (Bradshaw and Stettler 1995; Howe et al. 2000). Quantitative trait loci (QTLs) for bud flush have also recently been identified in Populus (Bradshaw and Stettler 1995; Frewen et al. 2000) and Pseudotsuga menziensii (Jermstad et al. 2001).

Salix species growing in short rotation systems are extremely sensitive to spring frosts. During the period of bud break, flushing buds and newly flushed shoots cannot tolerate freezing lower than −3°C (von Fircks 1994; Tsarouhas et al. 2001) and thus, there is a great need to characterize genotypes for late bud flushing. One important question is also to what extent bud flushing and growth are under separate genetic control – is it possible to select for late flushing clones with a high growth potential?

The specific objective of the present study was to identify and map QTLs affecting timing of bud flush in Salix using an available AFLP and RFLP genetic map, and investigate to what extent these QTLs affected growth.


A mapping pedigree derived from a cross between the male hybrid clone “Björn” (Salix viminalis×Salix schwerinii) and the female clone “78183” (Salix viminalis), produced by Svalöf Weibull AB, was used in the present study. A set of 92 progenies was analysed. Time of bud flushing was assessed in both field and indoor experiments.

The outdoor site was located at the experimental station Pustnäs outside Uppsala (59°N, 17°E), in a uniform flat field at an elevation of 28 m. Prior to field planting, plants were grown in a greenhouse for 8 weeks. The plantation was established in the summer of 1997 with a spacing of 1.5×1.0 m using a randomized complete block design with single plant plots replicated ten times. Timing of bud flush assessed in the spring of 1998 and 1999 by tri-weekly inspections. Bud flush was recorded when the first unfolded leaf, within the bud primordium, was observed and it was expressed as the number of days to bud flush starting from first of January 1998 and 1999 for the estimates in 1998 and 1999, respectively. In the year 1998 bud flushing was recorded separately in four stem sections: 1) terminal bud (tbfl98); 2) 10 cm below the terminal bud (sect1); 3) from the half of the stem length until 10 cm below the terminal bud (sect2); and 4) from half of the stem length until the stem base (sect3) (Table 1). At the end of the second year of growth (1998) a stem harvest took place. Bud flushing for the year 1999 was recorded in the stems cut approximately 10 cm above the ground. It must be noted that the Salix species studied here form shoot tip abscission instead of a typical terminal bud. Therefore the first axillary bud, below the abscission zone, is here considered as a terminal bud. The height of the tallest shoot was measured at 2 consecutive years (1998 and 1999) in the field. Air-temperatures were recorded from a temperature recorder within the plantation and from the meteorological station in Ultuna, approximately 1 km away from the field trial.

Table 1. Description of the investigated bud flushing traits.
Trait descriptionYearStem ageNumber of clones nTrait abbreviation
Terminal bud flush19981 yr74tbfl98
Bud flush 10 cm below the terminal bud19981 yr82sect1
Bud flush from 10 cm below the top until the middle of the stem19981 yr82sect2
Bud flush from the middle until the base of the stem19981 yr82sect3
Bud flush in cut stools19992 yr82bfl99

Bud flush on the basal stem section19991 yr64inbfl

The time of bud flushing was also recorded for progenies wintered indoors under controlled environmental conditions. Ten cm cuttings from 1-yr old shoots were planted in two litter pots containing 20 % clay and 80 % peat. The plants were arranged in five blocks with two replications within block. After 10 weeks of growth (18/22°C, 16 h photoperiod) the plants were transferred to a glasshouse room, isolated from the daily light and temperature by iso-thermosol curtains, with the following conditions: 10/5°C, 9 h photoperiod and 65 % RH. After another 8 weeks the plants were constantly exposed to 0±1°C for 2 months. At the end of this period the plants were cut at 10–15 cm (depending of the total height) from the base and transferred with pots to 10/5°C, 9 h photoperiod and 70 % RH for the bud flushing evaluation. Bud flush was recorded tri-weekly and the time of bud burst was scored as the number of days from the day the plants entered to the 10/5°C temperature regime.

Analysis of variance (ANOVA) was used to detect differences among clones and to estimate the variance components for the clone effect. All trait values followed a normal distribution. Clonal mean heritabilities were calculated as:


where σc2 and σe2 are the clone and error variance components, respectively, and bc is the coefficient for σc2 from the clone expected mean square as calculated by the standard least squares model of an analysis of variance (JMP 3.0, 1994).

The linkage map utilized for QTL analysis is described by Tsarouhas et al. (2002). In brief, two maps (one for each parent) were constructed according to the “two-way pseudo-testcross” strategy using AFLP and RFLP markers. The map from the hybrid parent (Salix viminalis×Salix schwerinii) included 223 markers and formed 26 major linkage groups, while the S. viminalis map consisted of 153 markers placed on 18 major groups. Both maps cover about 75 %, in average, of the Salix genome.

QTL analysis was performed with composite interval mapping as described in Tsarouhas et al. (2002) using clonal means as phenotypic values. The proportion of the phenotypic variance explained by each QTL (σq2) was estimated from the variance component of an ANOVA model by simultaneously fitting all the detected QTLs (the marker closest to the QTL) for the trait. The total phenotypic variance explained by all the QTLs for each trait was determined from the R2adj (coefficient of determination) of the ANOVA model.


Phenotypic traits in the field

The S. viminalis parent was late flushing as compared to the S. viminalis×S. schwerinii parent in all experiments (Fig. 1). This, however, should be expected since the geographical origin of the S. schwerinii clone was further north than that of S. viminalis. It is well established in forest tree species that genotypes originating from northern latitudes release dormancy earlier than southern ecotypes in common garden tests (Pauley and Perry 1954; Farmer and Reinholt 1986).

Figure 1.

Distribution of phenotypes for some of the investigated bud flushing traits. Axes x and y represent time (days) from the 1st of January (tbfl98, bfl99) and number of progenies (n) respectively. The phenotypic values for the parents B (clone Björn: Salix viminalis×Salix schwerinii) and V (clone 78183: Salix viminalis) are indicated by arrows.

The estimated temperature sums above 5°C (Ts) required to release dormancy in our pedigree were 56 and 66 for the years 1998 and 1999 respectively (Fig. 2), which is considerably less than other tree species growing in Sweden. For instance, it has been estimated that the Ts required for bud burst in 3–5 yr-old Norway spruce seedlings in central Sweden is approximately 120, which corresponds to middle May (Hannerz 1999). This confirms earlier reports showing that short rotation coppice willows in Sweden are among the earliest leaf flushing tree species (von Fircks 1994) and highlights the high risk of spring frost damage.

Figure 2.

Temperature sums above the threshold 5°C at Pustnäs, Uppsala (Sweden), over the years 1998 and 1999. Vertical lines indicate the mean of the timing of bud flush in the years 1998 and 1999.

The time needed for plants to release bud dormancy depended on the bud position on the stem. Buds on the base of the stem during the spring of 1998 flushed significantly later than buds on other sections, while bud burst on cut stools in 1999 flushed later than any section of the stem in 1998 (Table 2). Although the difference in time of flushing between the two years could be attributed to the temperature dissimilarities (Fig. 2), the Ts for bud flush was significantly higher in 1999 than in 1998 (Table 2). Assuming that the most important factor influencing bud break in northern regions is air temperature (Hänninen 1990), willow buds on cut stems appear to require higher temperature to release dormancy than buds on stems. It has been suggested earlier that bud break on stems and bud break on stumps are controlled by different genes (Sennerby-Forsse and Zsuffa 1995; Rönnberg-Wästljung 2001). The phenotypic correlations between stem sections were relatively low and decreased with the distance between them (Table 3).

Table 2. Means, variance components for clone effect (σc) and clonal heritabilities (H2) for the analysed traits. *, ** significantly different at p<0.001, 0.0001 respectively, in the one way ANOVA test.
TraitMean (days)Mean (Ts)2Earliest clone (days)Latest clone (days)σc2H2
  1. 1 Column values followed by different letters are significantly different at p<0.05 (Student's t test for each pair of means; JMP 1994).

  2. 2 Temperature sum above 5°C from January 1st. For trait abbreviations see Table 1.

tbfl98120 a 155 a11512522 **46
sect1120 a56 a11912511 *44
sect2121 a56 a12012516 **54
sect3123 b59 b12012728 **68
bfl9913066 c12213610 **43

Table 3. Clonal phenotypic correlations among the investigated traits and number of QTL identified. Bold number indicates significant (p<0.01) Pair-wise Pearson product-moment correlations (JMP, version 3.0, 1994). LR denotes the likelihood ratio which, is equivalent to a LOD score after multiplying by 0.2117. Statistical significant thresholds determined by permutation tests implemented by the MAP MANAGER QTXb17.0 software (Manly et al. 2001). The phenotypic effect (%) for all QTL of each trait was estimated from the R2adj values of the ANOVA model. For trait abbreviations seeTable 1.
Traittbfl98 sect1 sect2 sect3 bfl99 inbflN of QTLLR (Threshold)R2adj (%)
tbfl980.490.480.310.170.10214.5, 14.6 (13.7)21.6
sect1 0.800.500.100.12 
sect2  0.680.050.31 
sect3   0.220.42 
bfl99    0.56213.3, 17.2 (12.9)22.3
inbfl     513.6–21.4 (13.2)48.7

The clonal mean heritabilities for the different stem sections and years ranged from 0.43 to 0.68. In general, timing of bud flush in the present study was under moderate genetic control compared to other studies in Salicaceae. In an F2 family of hybrid poplars (Populus trichocarpa Torr. and Gray×Populus deltoides Barr), broad sense heritabilities for bud flush ranged between 0.80–0.94, depending of the method of estimation (Howe et al. 2000). Similarly, Bradshaw and Stettler (1995) reported broad-sense heritability for bud flush to be 0.98 in an F2 family of hybrid poplars. The lower clonal mean heritabilities in the present study could be attributed to the type of the cross and the origin of the parents. We anticipate higher levels of additive genetic variance if we had studied an F2 of more diverse parents. The rapid temperature increase during the spring 1998 (Fig. 2) may also have influenced the precision of the clonal variation assessment. Difficulties in detecting clonal variation under drastic spring warming have been reported for Belgian poplar clones tested in Sweden (Ilstedt 1996). The estimates of clone effect and heritability for timing of bud flush were higher in the low stem sections suggesting that the release of bud dormancy in the basal stem may be less influenced by the environmental conditions than the upper parts.

Time of bud flushing indoors

Plants in the indoor experiment required, in average, seventeen days to release dormancy, which is equivalent to 42 Ts (2.5×17=42). This however, cannot be directly compared to the field experiment due to different wintering conditions. Timing of bud flush is strongly affected by the hardening conditions as has been reported for Norway and Sitka spruce seedlings (Malcolm and Pymar 1975; Dormling 1982). Even though winter conditions differed between indoors and outdoors experiments, significant correlation was observed between timing of bud flush indoors and outdoors (on the lowest stem section and on stumps, Table 3). This supports observations in forest tree species (Hannerz 1999) showing that time of bud flushing exhibits low G×E interaction. The difference between the earliest and the latest clone was largest (23 days) in the indoor experiment, indicating that temperature fluctuations in natural spring conditions, i.e. warm spells, may obscure some of the genetic variation of bud dormancy release.

The relationship between time of bud flushing and height growth in the field

The phenotypic correlations between height growth and time of bud flushing were negative but low in the field experiment (Table 3). This would suggest little risk of reducing a yield component i.e. stem height, through selection for late flushing in willows. However, more data is clearly necessary to test this tentative hypothesis. The genetic correlation between height and time of bud flushing has been estimated fairly strong (−0.42) during a single year in forty factorial (8×8) families of S. viminalis (Rönnberg-Wästjung and Gullberg 1999).

The correlation between bud flushing and growth is also affected by the frequency and severity of frost events on the experimental site. In a frost prone site, for example, late flushing genotypes may have superior height growth performance due to reduced incidence of frost damage. In the present study there was no frost damage during bud flushing or during the early growing season. Experimental trials, comparing clones and families of Norway spruce in Sweden, have shown that height growth is in general negatively correlated with early bud burst, although a positive correlation has been observed in one trial in northern Sweden (Hannerz et al. 1999). Low risk of reducing growth throughout genotypic selection for late flushing has been suggested for Picea sitchensis seedlings and Pinus contorta (Cahalan 1981; Cannell et al. 1985).

QTL mapping

Four quantitative trait loci controlling timing of bud flush were identified for plants growing under field conditions (Fig. 3). Two of the QTLs were located in the male map (S. viminalis×S. schwerinii) and explained 12–16 % of the phenotypic variance while the remaining two QTLs were found on the female map (S. viminalis) and accounted 13–14 % of the phenotypic variance (Fig. 3).

Figure 3.

Position and phenotypic effects of QTLs for bud flushing traits in the S. viminalis (clone 78183) and S. viminalis×S. schwerinii (clone Björn) maps (Tsarouhas et al. 2002). Bars to the right indicate genomic area in which the likelihood ratio (LR) exceeds the significant for the QTL threshold and vertical lines on the left define the confidence interval for each QTL based on bootstrap resampling (MapManager-QTXb17.0). The two numbers on the right of the bars show the LR value and phenotypic effect σq2 (in parenthesis) respectively. Map units (cM) shown on the left side of each linkage group were calculated by the Kosambi mapping function. The designation A and B in the unmapped marker loci denote the allele origin, S. viminalis and S. viminalis×S. schwerinii respectively. For trait abbreviations see Table 1.

Two QTLs and three unmapped marker loci associated with the timing of bud flush were detected in the indoor-experiment. The QTLs from the indoor study explained 6–16 % of the phenotypic variance. Both indoors and outdoors experiments corroborate the involvement, at least in part, of few genes of large effect for bud flushing. One QTL associated with indoor timing of bud flush shared common intervals with a QTL detected under field conditions (Fig. 3). Frewen et al. (2000) have reported nine QTLs for bud flushing accounting 5.9–16.8 % of the phenotypic variance in a Populus interspecific family. Thirty-three QTLs for timing of bud flush, each explaining a relatively small proportion of the phenotypic variance (2.4–11.5 %), have also been reported in an intraspecific population of Pseudotsuga menziensii (Jermstad et al. 2001).

None of the bud flush QTLs coincided with QTLs controlling height growth identified in the same pedigree (Tsarouhas et al. 2002). Furthermore, molecular markers flanking the detected QTLs for bud flushing were tested for association with the corresponding annual values of height. This analysis showed non-significant associations with the likelihood ratio (LR) ranging from 0.2 to 6.0 (or LOD <1.3) (data not shown). This suggests that time of bud break and height may in part be under independent genetic control in Salix.

The relatively small population size used in this study only allowed detection for QTLs with moderate to large effects (Beavis 1994). This means that QTLs with smaller effects may be present in both indoor and field conditions. Furthermore, the map did not have full coverage and some QTLs may have gone undetected also for that reason. Therefore, the current study detected a minimum number of QTLs involved in timing of spring bud flush for the investigated Salix pedigree.

Concluding remarks

Salix clones important for biomass production are among the earliest leaf flushing tree species in Sweden and they, therefore, have a high risk of spring frost damage. However, the highly significant clonal variation found for spring phenology suggests that there is a potential for selection to improve early bud flushing in Salix.

Several QTLs affecting the timing of bud flush were also identified. Molecular markers flanking these loci as well as other QTLs affecting e.g. growth could potentially be used for a more efficient selection of plants with high growth potential and late bud flushing. A prerequisite for such a marker assisted breeding program, is more information on the stability of the detected QTLs over environments and years.


The authors are grateful to Svalöf Weibull AB for the pedigree production. We especially express our gratitude to Dr. Stig Larsson for his invaluable and continuous support with plant material throughout the study. The technical assistance in the field by Mr Urban Pettersson, Mrs Ewa Winkler and Mrs. Kristin-Sophie Mellsjö is greatly acknowledged. This study was financially supported by the Swedish National Energy Administration.