Understanding the genetic basis of the timing of bud set, an important trait in conifers, is relevant for adaptation and forestry practice. In common garden experiments, both Scots pine (Pinus sylvestris) and Norway spruce (Picea abies) show a latitudinal cline in the trait.
We compared the regulation of their bud set biology by examining the expression of PsFTL2, a Pinus sylvestris homolog to PaFTL2, a FLOWERING LOCUS T/TERMINAL FLOWER 1 (FT/TFL1)-like gene, the expression levels of which have been found previously to be associated with the timing of bud set in Norway spruce. In a common garden study, we analyzed the relationship of bud phenology under natural and artificial photoperiods and the expression of PsFTL2 in a set of Scots pine populations from different latitudes.
The expression of PsFTL2 increased in the needles preceding bud set and decreased during bud burst. In the northernmost population, even short night periods were efficient to trigger this expression, which also increased earlier under all photoperiodic regimes compared with the southern populations.
Despite the different biology, with few limitations, the two conifers that diverged 140 million yr ago probably share an association of FTL2 with bud set, pointing to a common mechanism for the timing of growth cessation in conifers.
In temperate and boreal regions, adaptation to the timing of seasons is critical, as plants which do not stop growing and enter dormancy at the appropriate time are likely to be damaged by low temperatures. Photoperiod at each location is the most reliable cue for the progress of the season, allowing species to adapt the timing of their growth, reproduction and other life history events to the changing seasons (Garner & Allard, 1920; Lagercrantz, 2009; Petterle et al., 2013). In most woody species from boreal and temperate zones, photoperiod affects the duration of extension growth and the time at which buds enter dormancy (Wareing, 1956). The onset of this bud dormancy is usually accelerated by short days (SDs) and prevented or delayed by long days (LDs) in trees such as poplar (Hsu et al., 2006, 2011).
Norway spruce (Picea abies) and Scots pine (Pinus sylvestris) are two economically important tree species with large distribution ranges. They show several similar characteristics with regard to photoperiodic reactions. Their photoperiodic response is most pronounced at the seedling stage, with first-year seedlings showing free growth, which is replaced by predetermined growth in the subsequent years (Ekberg et al., 1979; Repo et al., 2001). Under a natural photoperiod, trees of the two species set buds in late summer in response to the shortening photoperiod, triggering, concomitantly, frost hardening (Clapham et al., 2001a,b; Beck et al., 2004). After a period of cold temperatures in winter, increasing temperatures in spring induce bud burst. Several traits controlled by photoperiod show a clinal variation in the two species (e.g. timing of growth cessation and bud set; Heide, 1974; Dormling, 1979; Hurme et al., 1997; Clapham et al., 1998; Oleksyn et al., 1998).
The two species also show differences in their photoperiodic reactions. Although height growth of Norway spruce seedlings will continue indefinitely under continuous light, Scots pine seedlings will stop growing and form terminal buds sooner or later under all light regimes, which makes it difficult to determine a critical night length for bud set (Dormling, 1973; Ekberg et al., 1979). When maintained under photoperiods longer than the optimum for height growth, seedlings of Scots pine show recurrent flushing, unlike Norway spruce, but the duration of the period between successive flushes is decreased by LDs (Ekberg et al., 1979). Photoperiod exerts a strong control on needle growth from the basal needle meristem of Scots pine, whereas this effect is observed for the apical meristem growth in Norway spruce (Ekberg et al., 1979). There are also differences between the two species in dormancy and the buildup of cold tolerance. Whereas buds of Norway spruce enter rest shortly after transfer to SDs and can attain high cold tolerance with the sole effect of SDs, the buds of Scots pine stay in a quiescent state (before true dormancy) by the influence of SDs and develop true rest and maximum freezing tolerance only after exposure to chilling temperatures (Dormling, 1993; Thomas & Vince-Prue, 1997). Another difference between the two species is their behavior in transfer experiments: it is possible to transfer populations of Norway spruce over much larger latitudinal distances than Scots pine, which does not tolerate long transfers to the north (Eriksson et al., 1980; Skrøppa & Magnussen, 1993; Persson, 1994).
There has been important progress to understand the molecular mechanisms that control the photoperiodic induction of flowering, mainly in angiosperms and, especially, in Arabidopsis thaliana (Blazquez, 2000; Suarez-Lopez et al., 2001; Yano et al., 2001; Bohlenius et al., 2006; Imaizumi & Kay, 2006; Turck et al., 2008; Kikuchi et al., 2009; Mohamed et al., 2010; Hsu et al., 2011; Karlgren et al., 2011; Andres & Coupland, 2012; Klintenäs et al., 2012). To measure day length, plants require the circadian clock which regulates internal oscillators (Blazquez, 2000; Jackson, 2009). The plant circadian oscillator is composed of several partially redundant feedback loops, forming a complex gene network of transcription factors and proteins (Mas, 2008). Among them, ZEITLUPE (ZTL) is an important protein turnover regulator which interacts with several components of the clock or the photoperiodic pathway, such as PSEUDO-RESPONSE REGULATOR (PRR) or GIGANTEA (GI) (Makino et al., 2000; Somers et al., 2000; Kim et al., 2007; Baudry et al., 2010).
Downstream of the photoperiodic pathway, there are important floral integrators. One of them is FLOWERING LOCUS T (FT; Kardailsky et al., 1999; Kobayashi et al., 1999). The FT protein acts as a long-distance signal (florigen) between leaves and the shoot meristem (Jaeger & Wigge, 2007; Andres & Coupland, 2012). The expression of FT has been shown to be higher in LDs, with a diurnal pattern peaking in the evening (Harmer et al., 2000; Suarez-Lopez et al., 2001). In the apex, the FT protein contributes to the activation of the meristem identity gene APETALA 1 (AP1; Abe et al., 2005; Wigge et al., 2005; Kaufmann et al., 2010).
FT belongs to the CENTRORADIALIS/TERMINAL FLOWER 1/SELF-PRUNING (CETS) gene family. The family shares homology with phosphatidylethanolamine-binding proteins (PEBPs), present in all major phylogenetic groups, and which are generally proposed to act as regulators of various signaling pathways to control growth and differentiation (Yeung et al., 1999; Chautard et al., 2004; Andres & Coupland, 2012). In plants, members of the family have been shown to be involved in the transition of vegetative to reproductive stage, in architecture elaboration or in the control of the growth and termination of meristems (Bradley et al., 1996; Kardailsky et al., 1999; Lifschitz & Eshed, 2006; Shalit et al., 2009). In Arabidopsis, in addition to FT, the CETS gene family includes five other closely related genes: TERMINAL FLOWER 1 (TFL1, a repressor of flowering), ARABIDOPSIS THALIANA CENTRORADIALIS (ATC), TWIN SISTER OF FT (TSF), MOTHER OF FT AND TFL1 (MFT) and BROTHER OF FT AND TFL1 (BFT) (Mimida et al., 2001; Yoo et al., 2004, 2010; Yamaguchi et al., 2005). The number of copies of CETS genes varies between species. Phylogenetic analyses have shown that the family is split into three main clades: FT-like, MFT-like and TFL1-like (Chardon & Damerval, 2005; Hedman et al., 2009). Studies in other angiosperm species have revealed that the role played by the different members of the gene family is mostly conserved. In poplars, Hsu et al. (2011) showed that FT1 and FT2, the duplicated homologs of FT, play different and complementary roles: reproductive onset is determined by FT1, whereas vegetative growth and inhibition of bud set are regulated by FT2 (bud set and dormancy induction require inhibition of FT2 expression).
There is increasing evidence that the photoperiodic pathway is highly conserved within the plant kingdom (Jackson, 2009; Lagercrantz, 2009; Serrano et al., 2009; Song et al., 2010). Although having diverged from angiosperms some 300 million yr ago (MYA), conifers (gymnosperms) show several similar characteristics in the response to photoperiod. In Norway spruce, four genes with similarity to the CETS family have been isolated: two are similar to the MFT-like clade (PaMFT1 and PaMFT2), whereas two others (PaFTL1 and PaFTL2) are more similar to the FT and TFL1 clades (Gyllenstrand et al., 2007; Hedman et al., 2009; Karlgren et al., 2011). PaFTL2 is the best characterized. It has been shown to control growth cessation and bud set in response to SDs as well as bud burst in response to increasing temperatures (Gyllenstrand et al., 2007; Karlgren et al., 2013). When overexpressed in Arabidopsis, PaFTL1 and PaFTL2 genes repressed flowering, whereas no effect was observed for PaMFT1 and PaMFT2 (Karlgren et al., 2011), indicating that the previously well-characterized PaFTL2 is functionally more similar to TFL1 than to FT. The duplication and divergence between FT-like and TFL1-like clades probably occurred in the angiosperm lineage and the ancestral function would have been more similar to TFL1 than FT (Karlgren et al., 2011). Gymnosperm genes possess a mix of FT and TFL1-like amino acid character states, giving some further support to this suggestion, but alternative hypotheses cannot be totally ruled out (Klintenäs et al., 2012). However, although Norway spruce has been shown to possess two copies of FT/TFL1-like genes (in contrast with angiosperms), only one copy has been suggested for pines (Klintenäs et al., 2012).
Although angiosperms and conifers have similar genes, their function differs. In contrast with Populus, in Norway spruce, the expression of PaFTL2 increases before bud set and decreases during bud burst (Gyllenstrand et al., 2007; Karlgren et al., 2011).
As already discussed, Scots pine and Norway spruce share patterns of latitudinal clines, but differ in some aspects of the physiology of bud set. Here, we examined to what degree these conifers share the regulation of the cessation of growth by studying the expression of the Scots pine ortholog of PaFTL2 in relation to growth rhythm. More precisely, we ask: do Scots pine and Norway spruce, two highly diverged conifers, share the same FT/TFL1-like gene family members?; does the expression of PsFTL2 respond to varying natural or artificial photoperiod?; is the expression pattern associated with the timing of growth cessation and bud set and/or bud burst in populations that are known to display different timing of bud set?
Materials and Methods
Three populations were chosen along the previously observed latitudinal cline for timing of bud set: Kolari (northern Finland: 67°11′N, 24°03′E; FIN), Punkaharju (southern Finland: 61°48′N, 29°19′E; FIS) and Radom, Poland (50°41′N, 20°05′E; POL). Seeds collected from eight open-pollinated (mostly half-sibs) families of each population were used in the different experiments.
Experiments under the natural photoperiod
Seeds from the three populations were sown in Haapastensyrjä (southern Finland: 60°37′N, 24°25′E) in a glasshouse under a natural photoperiod (see Table 1 for temperatures) in six blocks with two plants per pot. In each block, one family from the FIN population was included for validation purposes. The family has been studied previously in a larger experiment and its timing of bud set is known. Seedlings (300 in total) were raised from 9 June 2009 and sampled several times until bud set was completed (see Table 2 for sampling dates). To track bud set, height was measured weekly and the seedlings were checked for visible buds from 8 August to 11 November 2009. For each sampling time and each population, two needles were sampled from each plant of the eight families and pooled for RNA extraction. Samples were taken between 08:00 and 11:00 h for all sampling stages.
Table 1. Temperature (in °C) conditions in the glasshouse during the experiment
Temperature was recorded on a minute basis.
Table 2. Sampling stages and dates for bud set and bud burst in seedlings of Scots pine (Pinus sylvestris)
For bud burst, numbers in parentheses are the sum of degrees days (based on temperatures above 5°C threshold) accumulated from the beginning of the artificial bud burst course to the sampling day).
Needles, Fig. 3(g). During light time (14 h light)
7 May 2010 (115)
Needles, Fig. 3(g). During light time (15 h light)
11 May 2010 (170)
Needles, Fig. 3(g). During light time (16 h light)
After bud set, the seedlings were kept in a growth cabinet at 2°C for 5 months and, thereafter, bud burst was forced by increasing the temperature progressively from 2 to 22°C at a rate of 1°C every other day on average, and light length from 6 to 18 h at a rate of half an hour every day on average, over 30 d. Needles were sampled as for bud set at three different times during the bud burst course (Table 2).
Experiments under the artificial photoperiods
We used approximately the same critical day lengths for growth cessation in our experiments as employed by Oleksyn et al. (1998) (4 h night for the FIN population and 7 h for the FIS population). We chose a moderate temperature, as it has been shown that warmer temperatures can override any potential photoperiodic response, probably as a result of the accumulated heat sums (Oleksyn et al., 1998).
Seedlings were thus raised under continuous light and 18 : 14°C (16 h : 8 h) for 3 months in a growth chamber and split into three different growth cabinets with three different photoperiods: 20 h : 4 h (day : night); 17 h : 7 h (day : night); and 8 h : 16 h (day : night). Needles were sampled every 4 h for 96 h (families were pooled as previously) and thereafter the plants were returned to continuous light and bud set was scored 6 wk later.
Molecular methods and expression studies
For each time point and each population, needles were sampled from two plants per family and for all the eight families of each population and pooled. Each data point consisted of three biological replicates. Total RNA was extracted from the needle samples of the different time points of the two different experiments using a cetyltrimethylammonium bromide (CTAB) method (Chang et al., 1993). Quality and quantity were checked with agarose gel electrophoresis and using a NanoDrop ND1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). cDNA was synthesized using the Applied Biosystems High-Capacity cDNA reverse transcription kit (Applied Biosystems Inc., Foster City, CA, USA) following the manufacturer's procedure. In addition to PsFTL2, three other genes were analyzed: GI and PRR1 under the natural photoperiod, plus ZTL under the artificial photoperiods. These genes are putative phenology candidate genes based on studies in Arabidopsis thaliana, and they showed signs of selection (detected by different methods) in our previous studies (see Kujala & Savolainen, 2012, and citations therein). Gene expression was studied by quantitative reverse transcription-polymerase chain reaction (qRT-PCR). A dilution of 1 : 100 was used for the reactions with the SYBR Green method employing a Roche LightCycler® 480 SYBR Green I Master kit and a Roche LightCycler® 480 machine. Two housekeeping genes (Actin and glyceraldehyde-3-phosphate dehydrogenase, GAPDH) were used for expression normalization. The primers of the different genes are presented in Supporting Information Table S1. A ‘no-template control’ (NTC) was included in each run. The PCR efficiency of each gene was evaluated by the sample dilution method. PCR conditions were as follows: initial denaturation at 95°C for 10 min to activate the DNA polymerase, followed by 45 cycles of denaturation at 95°C for 10 s, annealing at the primer-specific annealing temperature for 30 s and extension at 70°C for 30 s. Following the last cycle, the melting curve was determined in the temperature range 57–95°C. A last step of cooling was performed at 40°C for 10 s.
Sequence and phylogenetic analyses
We sequenced the full coding sequence of PsFTL2 in five samples per population (FIN, FIS and POL) from megagametophytes. Sequencing and chromatogram editing were performed as described previously by Kujala & Savolainen (2012). The primers used were also presented in the supplementary data of the same paper, except for an additional primer pair: Fwd (5′-GTTACGGACATTCCTGCTACAAC-3′) and Rev (5′-TCTCATTCTCAAACAGCCTTGATA-3′).
Analysis of divergence was conducted in DnaSP v5.10.01 (Librado & Rozas, 2009).
For phylogenetic analysis, we reused part of the sequences in the phylogenetic tree of Klintenäs et al. (2012) (Fig. 1, most of the sequences in the FT/TFL1-like and MFT-like clades and only A. thaliana sequences from the other clades), together with our own sequences (sequences have been deposited in the NCBI/GenBank database under accession numbers KJ809134–KJ809140) and additional sequences identified in blast searches. Multiple protein (translation from coding sequences) and codon alignments were created using MUSCLE (Edgar, 2004) in Geneious version 6.1 (created by Biomatters, available from http://www.geneious.com). Phylogenetic trees were reconstructed with maximum likelihood using Phyml 3.0 (Guindon & Gascuel, 2003) as a plugin in Geneious. The analyses were run with the default parameters, using the Jones, Taylor and Thornton (JTT) substitution matrix for protein sequences (Jones et al., 1992) and the General Time Reversible (GTR) model for nucleotide sequences (Tavaré, 1986), except BEST was used for topology search. Branch support was obtained by 1000 bootstrap replicates.
To calculate variance components and genetic parameters for the timing of bud set, a linear mixed model was used:
where Yijk is the response of the kth seedling of family i planted in plot ij of the jth block, μ is the overall fixed mean of the population, fi is the random effect of the ith family, bj is the fixed effect of the jth block, pij is the random effect of plot ij and eijk is the random error term.
Relative quantification of gene expression using ΔCT values (CTcontrol − CTPsFTL2), where CT is the threshold cycle and CTcontrol is the geometric mean of the two housekeeping genes, was used to compare between different time points. For the experiment under artificial photoperiods, quantifications were calculated relative to the expression level of the first time point (4 h, under continuous light) as R(xi) = 2ΔCTxi − ΔCT4 h. Quality control was performed using Biogazelle qbasePLUS 2.1 (Hellemans et al., 2007) and statistical analyses were performed with the same program or using SAS statistical package v9.3 (SAS Institute Inc., Cary, NC, USA).
FT/TFL1-like genes in pines
We obtained the full coding sequence of PsFTL2 in Scots pine and used version 1.01 of the Loblolly pine genome assembly (http://www.pinegenome.org/pinerefseq/) (Neale et al., 2014) to search for pine FT/TFL1-like genes. We found two distinct genes corresponding to PaFTL1 and PaFTL2, as opposed to Klintenäs et al. (2012). The two gene structures are similar between Loblolly pine and Norway spruce with regard to the exons, but there are differences in the length of the introns, especially for FTL2 (Fig. 1, Supporting Information Fig. S1). The putative FTL2 protein in Scots pine and Loblolly pine has 173 amino acids (AA), as FTL2 in spruce, whereas the putative FTL1 protein in Loblolly pine has 172 AA, as FTL1 in spruce (Fig. S2). Protein and nucleotide identity matrices suggest the existence of two different genes (Tables S2–S4). We calculated synonymous and non-synonymous substitution rates between our studied putative PsFTL2 and FTL1 and FTL2 from other conifers. Synonymous divergence between Norway spruce and pines at FTL2 was Ks = 0.33, and that between Norway spruce PaFTL1 and Loblolly pine PtFTL1 was Ks = 0.17. The synonymous divergence estimates between the two duplicates FTL2 and FTL1 from pine or spruce were much higher, Ks ~ 2, compared with the mean genomewide Ks of 0.175 between pines and spruce (Chen et al., 2012b) (Table S5). These results suggest that our PsFTL2 is a putative FTL2, distinct from FTL1. Finally, with most of the loci used in the phylogenetic analysis by Klintenäs et al. (2012), we rebuilt a phylogenetic tree including Scots pine FTL2, MFT1 and MFT2-exon4, as well as Loblolly pine FTL1 and FTL2. Pine FTL1 grouped with spruce FTL1, and likewise for FTL2. The support was c. 50% for the node at which the FTL1 and FTL2 clades split, but, within each clade, there was very high support for the split between spruce and pines (Fig. 2). Without any functional characterizations, these verifications suggest that pines also have two FT/TFL1-like genes, and that the duplication pre-dates the Picea/Pinus divergence.
Experiments under the natural photoperiod in southern Finland
Figure 3 shows photographs of the bud growth dynamics in Pinus sylvestris. In late summer, freshly formed buds become visible and move progressively into the dormant state in the autumn and winter. With increasing temperatures in the spring, swollen buds precede bud burst (opening of bud scales) and the growth of new needles.
Timing of bud set
Half-sib families from three populations covering different latitudes of the distribution range of Pinus sylvestris were raised in a common garden experiment in Haapastensyrjä (southern Finland) from the beginning of June until bud set. As expected, northern populations set bud earlier than southern ones (Table 3, Fig. 4) and populations differed significantly (ANOVA with Tukey's comparison test). The estimated additive genetic variances were much lower in the two northern populations than in the southernmost Polish population.
Table 3. Variance components and genetic parameters for timing of bud set of three populations of Scots pine (Pinus sylvestris)
σf2, predicted between-family variance; σBP2, predicted within-family–between-plot variance; σWP2, predicted within-plot variance. VA = 4σf2, VP = σf2 + σBP2 + σWP2. CV is obtained as the square root of the variance divided by the mean.
FIN, Finland north; FIS, Finland south; POL, Poland.
PsFTL2 expression under the natural photoperiod
The results of expression at five different times showed an overall increase in the expression of PsFTL2 before the completion of bud set, with a sequential increase in expression following the increasing proportion of bud set (Fig. 5). The northernmost population in this study, FIN, reached the highest expression level before the others, in accordance with the shortest time to bud set of the population (Table 3). Unlike PsFTL2, expression of the other studied putative phenology candidate genes, PsGI and PsPRR1, did not show clear divergence patterns among the populations; no statistically significant differences were observed between FIN, FIS and POL at most of the sampling times (Fig. S3).
PsFTL2 and bud burst
To check whether, in Scots pine, warming temperatures result in the repression of expression of PsFTL2 and the induction of bud burst as in Norway spruce (Gyllenstrand et al., 2007), bud burst was artificially forced in dormant seedlings by increasing the temperature progressively. Table 4 depicts the proportion of bud burst in each population. Concomitant with an increase in bud burst over three sampling points (Table 4), a decrease was observed in PsFTL2 expression, in contrast with the findings during the progress of bud set. However, the effect was much less marked than that of bud set (Fig. 5). For instance, at the first sampling time (BB1), the FIS population had already reached 83% of bud burst, whereas the proportions were much lower for the other populations. We would expect FIN, the northernmost population, to reach a high level of bud burst first. However, the levels of PsFTL2 expression decrease between different sampling times followed the proportions of bud burst. The decrease in the expression level for bud burst was higher when the differential between the two sampling times was also higher, and very little decrease in the expression was observed when there was almost no increase in the proportion of bud burst (see Fig. 5 and Table 4).
Table 4. Proportion of plants with visible bud burst in the three populations of Scots pine (Pinus sylvestris)
BB1–BB3, the three sampling times. FIN, Finland north; FIS, Finland south; POL, Poland.
PsFTL2 expression under different artificial photoperiods
The above expression pattern under the natural photoperiod was compared with expression at three different day lengths (8, 17 and 20 h of light). Needles were sampled every 4 h for 96 h, and the plants were then returned to continuous light and bud set was recorded 6 wk later. The first buds in the FIN population appeared c. 2 wk after the end of the photoperiod treatments. For all populations, PsFTL2 expression was highest in short day lengths (Fig. 6). The two LD treatments (20 h : 4 h and 17 h : 7 h) did not induce much bud set in the Polish population (Tables 5, 6).
Table 5. Proportion of visible buds in seedlings of Scots pine (Pinus sylvestris) 6 wk after treatment of the different photoperiod (light : dark) regimes
20 h : 4 h
17 h : 7 h
8 h : 16 h
FIN, Finland north; FIS, Finland south; POL, Poland.
Table 6. Average expression levels for FLOWERING LOCUS T/TERMINAL FLOWER 1-like 2 (FTL2) in the three populations of Scots pine (Pinus sylvestris) under the three photoperiod (light : dark) treatments
20 h : 4 h
17 h : 7 h
8 h : 16 h
Expression levels averaged over time points are relative to the first time point, when plants were still under continuous light. FIN, Finland north; FIS, Finland south; POL, Poland.
The expression levels were all significantly different between populations when using the average levels over time (Table 6). When efficiently induced, PsFTL2 showed a diurnal rhythm with expression peaks often occurring after the night periods (Fig. 6). Each population showed differences between photoperiods, and the populations reacted differently to the photoperiods. The most striking rhythm was found in the northern FIN population in the shortest days, with clear peaks and deep valleys. In addition, under the shortest day conditions, even the lowest daily expression in FIN was elevated (compared with the starting point), unlike in FIS. Another observation was that PsFTL2 expression increased much more quickly in FIN under all photoperiods compared with the other populations. Taken together, the results showed that the expression of PsFTL2 was more highly induced in the FIN population under our experimental conditions. Furthermore, for all three populations, we observed a significant correlation between the expression level of PsFTL2 under the different photoperiod treatments and the proportion of bud set (Fig. 7). Among the other candidate genes, only PsPRR1 in FIS showed a significant correlation, in the opposite direction to PsFTL2 (Fig. S4). We also compared the expression patterns of PsFTL2 and PsGI for six time points during the first 48 h of the experiment under artificial photoperiods. Except for the 17 and 20 h photoperiods in FIN, the expression patterns of PsGI were similar between the populations, with the same phase, and the peaks of the different photoperiods did not systematically correspond to the peaks in PsFTL2 expression (Fig. S5).
FT/TFL1-like genes in pines
To make sure that our PsFTL2 gene is the ortholog of the characterized PaFTL2, we performed sequence and phylogenetic analyses and found that, contrary to the observation of Klintenäs et al. (2012), our results suggest that pines also have two copies of FT/TFL1-like genes. We found a high synonymous divergence between the duplicates within each species (Ks ~ 2), whereas the Ks values for each of the genes, between the two species, were much lower (0.17 and 0.33 for FTL1 and FTL2, respectively; Fig. S6). This suggests that the duplication of conifer FT/TFL1-like genes pre-dates the separation of spruce and pine lineages (mean Ks = 0.175) (Chen et al., 2012b), which occurred around 140 MYA (Savard et al., 1994; Gernandt et al., 2008; Buschiazzo et al., 2012). Despite the ancient duplication event, the exon structure between FTL1 and FTL2 of pine and spruce is very similar, unlike the introns, suggesting a constraint on the evolution of the exons. Whether the existence of two copies of FT/TFL1-like genes is a general feature of all conifers is still to be clarified. Ginkgo biloba seems to have just one copy, but available sequence data are limited. Furthermore, the recently published spruce genome (Nystedt et al., 2013) presents four additional putative FT/TFL1-like genes. Whether they are all functional is an open question.
PsFTL2 and bud set
The timing of bud set is known to be under photoperiod control (Oleksyn et al., 1992) and to display a latitudinal gradient in conifers such as Scots pine (Hurme et al., 1997; Oleksyn et al., 1998; García-Gil et al., 2003; Serrano et al., 2009). As expected, the northernmost FIN population stopped growth and set bud first, and the southernmost POL population last. This confirms the pattern of several earlier studies in similar conditions, which validates our experimental setting.
PsFTL2 increased in all populations preceding the induction of bud set under natural and artificial photoperiods. Under artificial photoperiods, our results also showed that a shorter day length induced higher expression of FTL2 and was more efficient in inducing bud set. However, populations differed and the northernmost population was clearly responsive to even short night lengths. Under a natural photoperiod, the maximum expression level of PsFTL2 was higher for FIN and FIS than for POL in the early stages of bud set, but later all showed similar expression levels. This might reflect a time lag in the expression between northern and southern populations, which is correlated with the time to bud set (shorter for northern populations and longer for southern ones). In Norway spruce, Qamaruddin et al. (1995) observed that, when raised under continuous light, the Romanian population needed four cycles of 16 h night to reach a high level of bud set completion, whereas only one cycle was efficient in the northern population, and the northern population reached the same degree of dormancy after three nights of 16 h as the southern population after eight nights of 16 h. Using the same populations, Gyllenstrand et al. (2007) observed that the more southern Romanian population reached the same level of expression in twice the time of the Arctic population. The photoperiod of the original site plays a predominant role in the response observed in common garden experiments (Partanen & Beuker, 1999).
To compare the expression pattern of PsFTL2 with that of other putative phenology candidate genes that had been tested previously for selection (Kujala & Savolainen, 2012), we also analyzed the expression of PsGI, PsPRR1 and PsZTL under the same conditions. These genes are presumably located upstream of the PsFTL2 gene (Andres & Coupland, 2012). The results showed that, overall, these genes did not display similar clear divergent expression patterns between populations over time as did PsFTL2. However, it is likely that subtle differences in expression of these genes are sufficient to induce important differences in the expression of PsFTL2. Furthermore, the lack of differential expression does not guarantee that there is no role, because effects may be exerted through existing genetic polymorphisms, as suggested by the allele frequency cline at PRR1 (Kujala & Savolainen, 2012). In addition, because only one time point per day analyzed for the populations under the natural photoperiod might not be sufficient to accurately assess a potential diurnal change in expression, we also compared the expression patterns of PsFTL2 and PsGI for six time points during the first 48 h under artificial photoperiods. These data show clear differences in the patterns of PsFTL2 among populations, whereas the between-population differences were much smaller for PsGI expression and were not similar to those of PsFTL2. For instance, PsGI expression in FIS and POL had seemingly the same patterns (same phase), but the peaks were higher for the 20 and 17 h day lengths than for the 8 h day length, whereas only the latter day length induced a significant increase in the expression of PsFTL2. Although they need to be taken with caution (because time points are every 8 h in this comparison), these results suggest that the difference in the expression of PsFTL2 is probably a result of mechanisms associated with the circadian clock, but these mechanisms seem to differ from those described in Populus (Bohlenius et al., 2006). More experiments are needed to confirm this observation.
In addition to its importance through expression studies, our previous work revealed an allele frequency cline in PsFTL2 (Kujala & Savolainen, 2012). In Norway spruce, previous studies have also revealed allele frequency clines in PaFTL2 (Chen et al., 2012a). The analyses suggested that one single nucleotide polymorphism (SNP) in the promoter of PaFTL2 might affect the divergent expression patterns observed between genotypes.
Holliday et al. (2008) also studied an FT/TFL1-like gene whilst monitoring autumn (between late summer and early winter) gene expression in needles of Sitka spruce. The gene was strongly down-regulated between the first (end of August) and subsequent time points. The gene is actually an FTL1 gene. However, its potential association with the induction of bud set was not studied. In Norway spruce, the expression of PaFTL1 in needles was higher in August than in December and May, but this change in expression over time was not as significant as in male cones (Karlgren et al., 2011). In a recent experiment, the role of PaFTL1 was further clarified, the results showing that, in addition to its putative role in male cone development, it possibly represses meristem activity and the formation of needle primordia during the summer, and that PaFTL1 acts in concert with PaFTL2 to control perennial growth in Norway spruce (Karlgren et al., 2013).
PsFTL2 and bud burst
PsFTL2 expression decreased with an increase in the proportion of bud burst. This agrees with findings in Norway spruce, where PaFTL2 expression decreased steadily during bud burst (Gyllenstrand et al., 2007). The change in expression level was less marked in Scots pine than in Norway spruce. It is often difficult to artificially control bud burst to display a progressive and differential timing among populations, because nearly all population flush buds within a short time frame.
The few previous studies in Scots pine have shown that populations are also differentiated in the timing of bud burst, with populations from cool, higher altitude locations, higher latitudes and short growing seasons flushing buds earlier (Beuker, 1994; Salmela et al., 2013). In Norway spruce, PaFTL2 expression declined earlier in the early flushing clone than in the late flushing one. In Scots pine, the relationship between changes in expression and bud burst were less clear. One contributing factor may have been that the seedlings from the northernmost population (FIN) were smaller and more difficult to score. Nevertheless, overall, we observed an opposite trend in the expression of FTL2 between bud set and bud burst for all the populations.
Temperature has been shown to be the main factor driving bud burst in many tree species, including conifers (Hannerz, 1999; Sogaard et al., 2008; Vitasse et al., 2009; Gauchat & Paques, 2011; Fu et al., 2012). More studies, especially under natural conditions, are needed to further explore the genetic basis of bud burst in Scots pine.
Comparing Scots pine and Norway spruce
Albeit with some limitations for bud burst, our results showed that the expression patterns of FTL2 are shared in Scots pine and Norway spruce, and FTL2 homologs are putative candidates for the control of growth rhythm in the two species. Several phenology aspects of both species are similar: bud set induced by short photoperiods with short timing to the north and long to the south, and bud burst mainly controlled by temperature. However, we know that the photoperiod response of Scots pine differs in many respects from that of Norway spruce, as described in the Introduction. In the aspects studied here, the two species behaved very similarly, and so the reasons for the differences are still open. Future experiments would allow us to better understand the role of PsFTL2 control of growth rhythm and to identify any potential features that might be caused by the observed differential photoperiodic reactions between Norway spruce and Scots pine.
CETS genes seem to be conserved between conifers, as suggested by the phylogenetic analysis. We showed that PsFTL2 is putatively involved in growth rhythm in Scots pine, as has been shown for Norway spruce. Karlgren et al. (2011, 2013) previously showed that PaFTL1 and PaFTL2 have divergent expression patterns, but probably act in concert to control perennial growth in Norway spruce. The expression of PsFTL1 in Scots pine is yet to be studied, but we would expect that the patterns would be similar to that of Norway spruce.
In Norway spruce, divergent expression of PaFTL2 among populations in relation to bud phenology and allele frequency cline has been shown. In Scots pine, an allele frequency cline suggesting clinal selection has been shown previously for PsFTL2 (Kujala & Savolainen, 2012), and here we showed divergent expression patterns among populations, which has not been seen for other putative phenology candidate genes tested. Taken together, these results suggest that there is an array of evidence pointing to FTL2 as an important gene involved in the control of phenology in both Norway spruce and Scots pine.
We thank Aleksia Vaattovaara, Soile Alatalo, Hannele Parkkinen and Riitta Jokela for their help in sampling and laboratory work at the University of Oulu. We thank Sirkku Pöykkö and Merja Lahdenpää for their help during experiments at the METLA station of Haapastensyrjä. We are grateful to David L. Remington (University of North Carolina, Greensboro, NC, USA) for his valuable comments on the manuscript. We also thank three anonymous reviewers who contributed to improve the manuscript. This work was funded by the EU project Noveltree (FP7 211868) and Biocenter Oulu.