North European invasion by common ragweed is associated with early flowering and dominant changes in FT/TFL1 expression

North European invasion by common ragweed is associated with earlier flowering, dominant changes in FT/TFL1 expression, and reduced seed production.

Abstract

During the last two centuries, the North American common ragweed (Ambrosia artemisiifolia L.) invaded a large part of the globe. Local adaptation of this species was revealed by a common garden experiment, demonstrating that the distribution of the species in Europe could extend considerably to the North. Our study compares two populations of common ragweed (one from the native range and one from the invaded range) that differ in flowering time in the wild: the invasive population flowers earlier than the native population under non-inductive long-day photoperiods. Experiments conducted in controlled environments established that the two populations differ in their flowering time even under inductive short-day photoperiods, suggesting a change in autonomous flowering control. Genetic analysis revealed that early flowering is dominantly inherited and accompanied by the increased expression of the floral activator AaFTL1 and decreased expression of the floral repressor AaFTL2. Early flowering is also accompanied by reduced reproductive output, which is evolutionarily disadvantageous under long vegetation periods. In contrast, under short vegetation periods, only early-flowering plants can produce any viable seeds, making the higher seed set of late-flowering plants irrelevant. Thus, earlier flowering appears to be a specific adaptation to the higher latitudes of northern Europe.

Introduction

Common ragweed (Ambrosia artemisiifolia L.) is an annual invasive plant with a negative impact on the economy due to its role as an agricultural weed (Weaver, 2001; Essl et al., 2015) and as a source of allergenic wind-dispersed pollen grains (Arbes et al., 2005). As a typical short-day (SD) plant, A. artemisiifolia starts to produce flowers when daily photoperiods drop below a critical threshold (~14 h of light) (Deen et al., 1998; Essl et al., 2015). The first sign of flowering is a pale green bud at the shoot apex, forming the main inflorescence (see Supplementary Fig. S1 at JXB online for a scheme of reproductive development of A. artemisiifolia and images of the plant). This inflorescence grows to become a leafless raceme with dozens of staminate (male) flower heads (Essl et al., 2015). At the same time, pistillate (female) flowers develop at the axils of the upper leaves and at the involucres of the male inflorescences (Essl et al., 2015). Maturity of male flowers is indicated by the thick yellow pollen-coated anthers protruding from the male flowers, while female maturity is indicated by the protrusion of a dichotomous stigma out of each female flower (Essl et al., 2015). Fertilized female flowers develop a fruit that mostly consists of a single seed (Essl et al., 2015).

The well-documented invasion history of A. artemisiifolia makes this species a good model to study eudicot invasiveness. Ambrosia artemisiifolia is native to North America and has successfully spread to all continents except Antarctica during the last two centuries (Gaudeul et al., 2011). The first European record is from France in 1863 (Chauvel et al., 2006), and the first Australian record is from 1911 (Webb, 1987). Although A. artemisiifolia seeds can be naturally dispersed by epizoochory (Rosas et al., 2008), most of its dispersal in the last centuries is probably due to human activity: A. artemisiifolia seeds are routinely detected in imported bird feed and livestock fodder (EFSA, 2010), and its settlement follows anthropic land disturbance (Essl et al., 2009). In Europe, the main infestations can be found in South-East Europe (particularly Hungary, Croatia, and Serbia), although A. artemisiifolia plants were observed from Great Britain to western Russia and from southern Italy to southern Sweden (Buters et al., 2015; Essl et al., 2015). Stable populations, however, can form in northern Europe only if mechanisms evolve to cope with the early onset of low temperatures in autumn. SD plants such as A. artemisiifolia cannot easily spread to higher latitudes. At higher latitudes, temperatures are too low for growth and for seed development soon after day length drops under the critical threshold to induce flowering. Therefore, cold resistance or accelerated flowering is required for successful seed production under these conditions. Changing flowering time is particularly likely to underlie rapid adaptation, because it can cause reproductive isolation. The gene pool of a population in the northern part of the invaded range would be less affected by non-adaptive alleles from the mother population adapted to lower latitudes. Previous common garden experiments have identified a latitudinal cline in flowering time among invasive European A. artemisiifolia populations, indicating that indeed a change in flowering time has occurred to adapt to higher latitudes (Leiblein-Wild and Tackenberg, 2014; Scalone et al., 2016).

In flowering plants, flowering time is regulated by members of the phosphatidylethanolamine-binding protein (PEBP) gene family (Bradley et al., 1996; Kardailsky et al., 1999; Yano et al., 2001; Lifschitz et al., 2006; Blackman et al., 2010). PEBP genes are found in all kingdoms of life, and in plants three major clades exist: FLOWERING LOCUS T-like (FT) genes, TERMINAL FLOWER 1-like (TFL1) genes, and MOTHER OF FT AND TFL1-like (MFT) genes. The MFT clade is the oldest of the three, and MFT-like genes are found in lycophytes, bryophytes, gymnosperms, and angiosperms, where they have a function in seed germination (Karlgren et al., 2011). FT/TFL1 genes are absent in studied lycophytes and bryophytes, but were detected in all studied gymnosperms and angiosperms (Karlgren et al., 2011). Separate FT and TFL1 clades originate in gymnosperms and further diverged in angiosperms (Liu et al., 2016). Both clades contain genes that regulate flowering time (Bradley et al., 1996; Kardailsky et al., 1999), although genes of that family can also have other functions such as regulating bulb or tuber formation (Navarro et al., 2011; Lee et al., 2013), stomatal opening (Ando et al., 2013), inflorescence architecture (Bradley et al., 1996), and flower morphology (Molinero-Rosales et al., 2004).

While most FT/TFL1 genes function as floral activators or repressors, some have no effect on flowering time (e.g. Lee et al., 2013). FT/TFL1 proteins often act in a non-cell-autonomous manner (Lifschitz et al., 2006; Corbesier et al., 2007; Mathieu et al., 2007; Tamaki et al., 2007; Navarro et al., 2011). In particular, Arabidopsis FT and its rice homologue Hd3a are produced in leaves and move through the phloem to the shoot apex (Corbesier et al., 2007; Jaeger and Wigge, 2007; Mathieu et al., 2007) where they interact with the bZIP transcription factor FLOWERING LOCUS D (FD) and together regulate, among others, the floral integrator gene SOC1 and the floral meristem identity gene FUL to induce flowering (Abe et al., 2005; Wigge et al., 2005; Yoo et al., 2005). Some FT-like proteins act as floral repressors, and the balance of activators and repressors determines flowering time (Higuchi et al., 2013), presumably through competition for FD (Ahn et al., 2006; Ryu et al., 2014). Many FT/TFL1 genes are floral integrators at which multiple flowering pathways converge. In Arabidopsis, rice, barley, and potato, for instance, FT/TFL1 genes are activated by photoperiod via CONSTANS-like (COL) proteins (An et al., 2004; Kikuchi et al., 2011; Abelenda et al., 2016; Nemoto et al., 2016). Similar to FT/TFL1, COL genes are of ancient origin, occurring in all land plants and some algal lineages (Valverde, 2011). It is likely that most plants possess COL–FT/TFL1 modules, but the specifics of the regulation of FT/TFL1 genes differ between species. COL transcript and protein levels are regulated by light in long-day (LD) Arabidopsis, but not in SD rice (Nemoto et al., 2016; Shim et al., 2017). Arabidopsis CO activates FT/TFL1 genes, while the rice homologue of CO can activate or repress FT/TFL1 genes depending on its photoperiod-regulated interaction partners (Nemoto et al., 2016). In addition to photoperiod, Arabidopsis FT is also a main target of the vernalization pathway, which responds to long periods of cold such as during winter (Ream et al., 2012). Similarly, in wheat and barley, FT genes integrate photoperiod and vernalization signals promoting flowering (Ream et al., 2012). In strawberry, FT/TFL1 genes integrate light quality, temperature, and photoperiod (Rantanen et al., 2014; Nakano et al., 2015). In trees and perennial herbs, FT/TFL1 genes may also convey competence to flower in relation to plant age (Böhlenius et al., 2006; Hsu et al., 2006; Wang et al., 2011). Together, FT/TFL1 genes serve as central developmental regulators in many plant species and constitute a major potential target of evolution in the adaptation to particular environmental conditions.

In this study, we asked whether the previously identified A. artemisiifolia populations that flowered differently in a common garden show a similar difference in flowering time in a controlled environment differing only in photoperiod. We studied the genetic basis of the flowering time of A. artemisiifolia by creating an F₁ hybrid population and by studying candidate genes involved in flowering time regulation. We also studied the reproductive output to estimate the evolutionary significance and potential fitness benefits or costs of the early flowering trait.

Materials and methods

Plant material

A previous common garden study (Scalone et al., 2016) identified two A. artemisiifolia L. populations with significantly different flowering times: one early-flowering population from the northern part of the invaded European range (Drebkau, Germany N51°38'21'' E14°11'50''), the ‘invasive population’, and one late-flowering population from a central part of the native range (Lexington, KY, USA N38°01' W84°33'), the ‘native population’.

Seeds from at least 10 different plants in the two field populations were used to generate the respective study populations. For an additional analysis [gene expression at 15 days after germination (DAG)], a new seed batch was generated representing the original genetic diversity of the populations sampled from the wild by performing intrapopulation crosses (native×native and invasive×invasive) of all members of one population with each other. The intrapopulation F₁ seeds were used in the experiment. For tests of the genetic basis for different flowering time, interpopulation F₁ seeds (native×invasive) were used.

For the expression of the A. artemisiifolia FT/TFL1-like (FTL) proteins in Arabidopsis thaliana (L.) Heynh., coding sequences were amplified from cDNA and cloned into pEARLEYGATE100 (Earley et al., 2006). Transformation of Col-0 wild-type plants was performed by floral dip, and seeds were germinated on medium containing phosphinothricin. Plants transformed with empty pMDC160 (Brand et al., 2006) served as the phosphinothricin-resistant control. Seedlings were transferred to soil at 6 DAG. For expression analyses in Arabidopsis, phosphinothricin-grown T₂ plants were used.

Growth conditions

Ambrosia artemisiifolia seeds were stratified on moist filter paper for 4 weeks and subsequently placed in a growth chamber providing 16 h light of 50 µmol m^–2 s^–1 at 27 °C and 8 h darkness at 15 °C to induce germination. After germination, seedlings were transplanted to pots with pumice (0.5 mm<Ø>2.8 mm; Hekla Green, Bara Mineraler, Bara, Sweden), and moved into a climate chamber with SD conditions (12 h light of 280 µmol m^–2 s^–1 at 25 °C; 12 h darkness at 15 °C; humidity at 70%). Plants were watered with nutrient-enriched water (2 ml l^–1 of Wallco växtnäring 53-10-43+micro from Cederroth, Upplands Väsby, Sweden). The position of the plants in the chamber was randomized daily. Arabidopsis plants were grown under LD conditions (16 h light of 220 µmol m^–2 s^–1 at 22 °C; 8 h darkness at 20 °C).

Measuring flowering time

Once a day, A. artemisiifolia plants were inspected for signs indicating the start of specific phenological phases (see Supplementary Fig. S1 for a scheme of reproductive development of A. artemisiifolia): (i) appearance of the main male inflorescence bud (start of male floral initiation); (ii) first release of yellow pollen grains (end of male maturation phase); and (iii) first full extension of a dichotomous stigma out of a female flower (end of female maturation phase). The phases are referred to as ‘vegetative-male initiation’ (from germination to male floral initiation; symbol: ♂1), ‘male maturation’ (from male floral initiation to first pollen release; symbol: ♂2) and ‘vegetative-female maturation’ (from germination to appearance of the first elongated pistil; symbol: ♀). Plants with only pistillate flowers were excluded from the analysis. Because data were not normally distributed, the Mann–Whitney U-test (Marx et al., 2016) was used to test for significant differences between means.

Arabidopsis plants were inspected daily for the appearance of the first inflorescence (time to bolting, which was here equated with floral initiation). Upon bolting, both the date and number of (true) rosette leaves were noted.

Other plant measurements

Final plant size characteristics were determined: plant height (length of the main stem, not including the male inflorescence), basal diameter of the main stem, number of branches on the main stem, root weight, and root area (measured using a camera operated by WD3 WinDIAS software, Delta-T Devices, Cambridge, UK). For the last 2 months of the experiment, seeds were collected weekly. For each plant, the number of produced fruits was counted and the total fruit weight was measured. Mann–Whitney U-tests were performed to test for significant differences between groups.

Identifying A. artemisiifolia FT/TFL1 genes

Arabidopsis and Helianthus annuus FT/TFL1 cDNA sequences were used as queries in BLAST searches of expressed sequence tag (EST) libraries (Lai et al., 2012) from two A. artemisiifolia populations (one from the USA and one from Hungary) and four A. trifida populations (two from the USA and two from China). The two FT/TFL1-l gene fragments thus obtained were used for rapid amplification of cDNA ends (RACE) to acquire the full-length sequences. Another partial FT/TFL1 gene fragment reported by Li et al. (2015) was amplified using the published primers and also used for RACE. The three identified A. artemisiifolia FT/TFL1 genes were designated AaFTL1, AaFTL2, and AaFTL3.

RACE

5'- and 3'-RACE were performed using the GeneRacer kit (Invitrogen, Carlsbad, CA, USA), according to the manufacturer’s instructions using cDNA based on RNA from plants of the native population. The optional nested PCRs were always included. RACE products were cloned into pJET1.2 and sequenced. The 5' end of the AaFTL3 transcript could not be obtained, presumably due to low transcript abundance. Ambrosia artemisiifolia FT/TFL1 sequences were compiled from multiple sequencing runs, each base confirmed by at least three independent sequencing reactions. Supplementary Table S1 lists the RACE primers. Supplementary Table S2 lists the obtained A. artemisiifolia FT/TFL1 cDNA sequences with GenBank accession numbers.

Phylogenetic analysis

For phylogenetic analysis, amino acid sequences were obtained by searching the literature for FT/TFL1 proteins with confirmed function. FT/TFL1 proteins from the same species but with unknown function were included. This resulted in a list of FT/TFL1 proteins from 33 species and 16 families. FT/TFL1 proteins were designated as floral activators/repressors if overexpression resulted in earlier/later flowering or if knock-out resulted in late/early flowering, respectively. Sequences were considered as flowering neutral if their overexpression or mutation did not result in a change in flowering time. Sequences of most FT/TFL1 proteins were from NCBI; the remaining sequences were from other databases or studies (Supplementary Tables S3, S4). Sequences were aligned using MUSCLE with the maximum number of iterations set to 16 (Edgar, 2004; Dereeper et al., 2008), and curated using GBLOCKS allowing gaps (Dereeper et al., 2008). Curated sequences were used in a Maximum Likelihood analysis (PhyML) (Dereeper et al., 2008; Guindon et al., 2010), allowing gaps and using the WAG substitution model. Branch confidence was determined by aLRT (Guindon et al., 2010), and branches with a confidence score of <85% were collapsed. The analysis involved 172 amino acid sequences with a total of 132 sequence positions. For additional confirmation, a minimum evolution analysis as implemented in MEGA 7.0.20 (Kumar et al., 2016) was performed. A consensus tree was made of 100 bootstrap replicates, and the branches occurring in <50% of replicates were collapsed.

Prediction of FT/TFL1 protein function

Curated aligned sequences were used to generate consensus sequences to identify amino acid residues unique to functional groups (floral activators, repressors, or flowering neutral proteins) within either the FT clade or the TFL1 clade. The A. artemisiifolia sequences were compared with these consensus sequences. As an additional separate analysis, a score was determined to link protein functionality with the amino acid residue diversity across the entire curated aligned sequences. This involved creation of first a score for each position based on the prevalence of different amino acid residues within the group of activators/repressors and non-activators/non-repressors (for FT-like proteins/TFL1-like proteins, respectively). This is a value between 1 (occurs in all of the functional group) and 0 (occurs in none of the functional group). Then each FT/TFL1 sequence was investigated separately and the aforementioned values were assigned to each position. For each sequence, the values of individual positions were averaged to form a score for the entire sequence. Each sequence then results in an activator/repressor score and a non-activator/non-repressor score (again for FT-like proteins/TFL1-like proteins, respectively), and the ratio between these scores is the value presented (named ‘activator ratio’ and ‘repressor ratio’). Cut-off values were determined empirically to minimize the probability of falsely calling a known non-activator an activator, or a known non-repressor a repressor.

Gene expression

Ambrosia artemisiifolia leaf tip and shoot apex samples were collected at 4 h Zeitgeber time (ZT). Arabidopsis seedlings were collected at 8 h ZT. RNA extraction was performed using an RNeasy plant mini kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Genomic DNA was removed with the RNase-free DNase set (Qiagen) following the optional on-column DNase digestion protocol. cDNA synthesis was performed with the RevertAid First Strand cDNA synthesis kit (Thermo Scientific). HOT FIREPol polymerase and Evagreen qPCR Mix Plus (Solis Biodyne, Tartu, Estonia) was used in conjunction with the MyIQ Single-color real-time PCR detection system (Biorad, Hercules, CA, USA) for transcript quantification. For Arabidopsis quantitative reverse transcription–PCRs (RT–qPCRs), PP2AA3 (AT1G13320) was used as the reference gene (Mozgová et al., 2015). For the A. artemisiifolia RT–qPCRs, the following reference genes were used: TUBULIN A (TUA) (El Kelish et al., 2014), TUBULIN B (TUB) (Li et al., 2015), and GAPDH (Hodgins et al., 2013). Because TUB was found to be the most stable across samples (Supplementary Table S5), expression values relative to TUB are shown in the main figures. Expression values relative to the other reference genes are shown in the Supplementary figures. The observed trends were consistent and independent of the choice of reference gene. Ct values of technical replicates (three per sample) were used to calculate the mean normalized expression (MNE) according to Simon (2003). Then the MNEs of all biological replicates were used to calculate the ‘average relative expression’. For most A. artemisiifolia gene expression values, biological replicates were single individuals from the same population, but for the 15 DAG time point and in the leaf versus shoot apex expression analysis, four pools of four individuals each were used. For Arabidopsis transformants, each biological replicate was an independent transgenic line pooling 10 individuals per line. For experiments where all data met normal and equal variance criteria, a one-tailed t-test was performed; otherwise a Mann–Whitney U-test was performed.

Results

An invasive population of A. artemisiifolia from Germany has a shorter vegetative phase

A previous study (Scalone et al., 2016) reported that an invasive A. artemisiifolia population flowered earlier than a native population, when both were grown in the same experimental garden with naturally changing photoperiod (see Supplementary Fig. S1 for images of plants from the two populations). The present study extended this experiment using controlled cultivation conditions and constant SD photoperiods. For the tested native and invasive populations, the lengths of three phenological phases were recorded (Fig. 1A). The plants from the invasive population had shorter ‘vegetative-male initiation’ and ‘vegetative-female maturation’ phases ($\overline{\text{X}}$_n♂1=66.8 d, $\overline{\text{X}}$_i♂1=34.8 d, P<0.0001, Fig. 1B; $\overline{\text{X}}$_n♀=77.5 d, $\overline{\text{X}}$_i♀=40.2 d, P<0.0001, Fig. 1C), but the ‘male maturation’ phase was not significantly different ($\overline{\text{X}}$_n♂2=12.9 d, $\overline{\text{X}}$_i♂2=11.7 d, P=0.064, Fig. 1D). Invasive plants also showed an increased degree of dichogamy, namely delayed development of female after male flowers ($\overline{\text{X}}$_n♂♀=2.2 d, $\overline{\text{X}}$_i♂♀=6.2 d, P<0.0001, Fig. 1E).

Fig. 1.

Flowering phenology of the native, invasive, and F₁ populations. (A) Averages of the different phenological phases together for the native population (red, n=20), the invasive population (blue, n=36), and the F₁ population (purple, n=111). Dashed vertical lines indicate times of sample collection for gene expression analysis. (B) Length of the vegetative-male initiation phase. (C) Length of the vegetative-female maturation phase. (D) Length of the male maturation phase. (E) Degree of dichogamy, expressed by the time between the appearance of the first mature male flower and the appearance of the first mature female flower. Box plots show the first, second, and third quartiles (the box), the 10th and 90th percentiles (the whiskers), and outliers as individual dots. Mann–Whitney U-tests were performed to test for significant differences between groups (ns P>0.05, ***P<0.001; Bonferroni correction was applied with m=12 to the α of individual comparisons to obtain the indicated overall α values).

In addition to these two populations, a third population was investigated: the F₁ of native×invasive crosses. For these plants, all phases were shorter than for the parental native plants ($\overline{\text{X}}$_F1♂1=34.3 d, $\overline{\text{X}}$_F1♂2=9.3 d, $\overline{\text{X}}$_F1♀=42.1 d, P<0.0001, Fig. 1B–D). The F₁ plants had ‘vegetative-male initiation’ and ‘vegetative-female maturation’ phases equally short as those of plants from the invasive population (P=0.59 and P=0.16, respectively, Fig. 1B, C), indicating that the invasive population has contributed (a) dominant allele(s) controlling these traits. The ‘male maturation’ phase of the F₁ plants was shorter not only than in the native population, but also than in the invasive population (P<0.0001; Fig. 1D). The reasons for this pattern could be overdominance of an ‘early’ allele for this trait or genetic heterogeneity in a parent population that is lost in the experimental F₁ plants. Unlike the invasive plants, the F₁ plants did not show increased dichogamy ($\overline{\text{X}}$_F1♂♀=1.6 d, P=0.02, Fig. 1E). In short, there appears to be at least one (over-) dominant allele that makes the invasive plants flower earlier than the native plants by shortening the vegetative phase.

Early flowering is associated with a cost in reproductive output

Plants with a shorter vegetative phase have less time to build up resource-gathering organs for the production of seeds, so early flowering can be expected to decrease the reproductive output. We asked whether this was the case for the plants of the studied invasive population. First of all, the early flowering invasive plants had a lower final plant height ($\overline{\text{X}}$_n=16.7 cm, $\overline{\text{X}}$_i=9.1 cm, P<0.0001), main stem basal diameter ($\overline{\text{X}}$_n=6.7 mm, $\overline{\text{X}}$_i=2.5 mm, P<0.0001), number of branches on the main stem ($\overline{\text{X}}$_n=27, $\overline{\text{X}}$_i=9, P<0.0001), root area ($\overline{\text{X}}$_n=36.5 cm², $\overline{\text{X}}$_i=9.4 cm², P=0.0002), and root weight ($\overline{\text{X}}$_n=905 mg, $\overline{\text{X}}$_i=189 mg, P<0.0001) compared with the native plants (Supplementary Fig. S2). Consequently, the invasive plants produced 75% fewer fruits than native plants (P<0.0001; Fig. 2). This agrees with previous correlations of plant size with seed production (Dickerson and Sweet, 1971; Fumanal et al., 2007). Although average fruit weight was slightly higher for plants from the invasive population, this difference was not significant (P=0.07). These results indicate that accelerated flowering in the invasive population is associated with a reproductive cost.

Fig. 2.

Reproductive output. (A) Number of fruits produced per plant for the native population (red, n=15 plants, average: 381 fruits per plant), and the invasive population (blue, n=27 plants, average: 91 fruits per plant). (B) Average fruit weight for the native population (red, n=15, average: 5.5 mg) and the invasive population (blue, n=27 plants, average: 6.3 mg). Box plots show the first, second, and third quartiles (the box), the 10th and 90th percentiles (the whiskers), and outliers as individual dots. Mann–Whitney U-tests were performed to test for significant differences between groups (ns P>0.05, ***P<0.001; Bonferroni correction was applied with m=17 to the α of individual comparisons to obtain the indicated overall α values).

A. artemisiifolia expresses three FT/TFL1-like genes

The transition to flowering is usually initiated by changes in gene expression. Here, we asked whether the observed differences in flowering time in invasive and native A. artemisiifolia populations are also associated with differential gene expression. Because of its important role in flowering time control, we focused on the FT/TFL1 family of A. artemisiifolia. Little was known about FT/TFL1 genes in A. artemisiifolia before, and we identified two members of this family using A. artemisiifolia and A. trifida EST libraries (Lai et al., 2012). The partial sequence of a third member was reported by another laboratory (Li et al., 2015). The partial sequences of A. artemisiifolia FTL1 and FTL2 were completed using 5'- and 3'-RACE. Full-length AaFTL3 could not be obtained, presumably due to its low expression. However, the available sequence implies that AaFTL3 is either dysfunctional or has a radically different function (see Supplementary Table S2 for the cDNA sequences of the three A. artemisiifolia FTL genes including GenBank accession numbers). Ambrosia artemisiifolia FTL1 and FTL2 were translated in silico, aligned with FTL protein sequences from other species, and used in a phylogenetic analysis. The resulting phylogenetic tree (Fig. 3) distinguishes the MFT clade (grey) from its FT and TFL1 daughter clades. The gymnosperm FT clade (yellow) and gymnosperm TFL1 (orange) can be distinguished from the angiosperm FT clade (green) and the angiosperm TFL1 clade (red) (see Supplementary Fig. S3 for a linear version of the same tree with all proteins labelled with name and accession number). Ambrosia artemisiifolia FTL1 belongs to the (angiosperm) FT clade and FTL2 to the (angiosperm) TFL1 clade. AaFTL1 clusters together with the floral activator H. annuus FT4 (Blackman et al., 2010) (H. annuus is the closest relative of A. artemisiifolia in the phylogenetic analysis, belonging to the same tribe Heliantheae within the Asteraceae family) and four Chrysanthemum sp. proteins (also from the Asteraceae family), two of which are floral activators (Oda et al., 2012; Fu et al., 2014) and the other two without known function. This clustering with floral activators suggests that AaFTL1 could also be a floral activator. The second identified A. artemisiifolia FTL, AaFTL2, clusters together with the floral repressor C. seticuspe AFT (Higuchi et al., 2013) and H. annuus BROTHER OF FT AND TFL1 (BFT) (with unknown function), suggesting floral repressor activity. Another phylogenetic analysis (minimum evolution) confirmed the aforementioned clustering of AaFTL1 and AaFTL2 (Supplementary Fig. S4).

Fig. 3.

Phylogenetic tree of plant PEBPs. Cladogram of 172 PEBP sequences from 33 species belonging to 16 families of plants, inferred using a maximum likelihood method (Phyml); branches with <85% support were collapsed. Shading indicates the major clades: MFT (grey), gymnosperm FT (yellow), gymnosperm TFL1 (orange), angiosperm FT (green), angiosperm TFL1 (red). Branch colour indicates floral regulator potential: floral activators are green, floral neutral proteins are yellow, floral repressors are red, and proteins with unknown function in relation to flowering time are coloured black. The positions of the A. artemisiifolia AaFTL1 and AaFTL2 are indicated.

Sequence composition predicts that A. artemisiifolia expresses one activator and one repressor of flowering

According to the phylogenetic analysis, the original angiosperm FT/TFL1 proteins presumably had a floral repressor function, since the gymnosperm FT/TFL1 proteins are floral repressors (Karlgren et al., 2011; Liu et al., 2016). Because of the ubiquitous presence of floral activators in the FT clade, it is reasonable to assume that the early angiosperm FT became a floral activator, and that later reversions occurred in some lineages. Because many mutations can cause loss of function, it should not be surprising that this seemed to have occurred multiple times in both the FT and TFL1 clade. To predict more precisely the function of the A. artemisiifolia FT/TFL1 proteins, a closer study of the amino acid sequences was required. Previous studies have identified critical functional regions of the FT/TFL1 proteins (Hanzawa et al., 2005; Ahn et al., 2006; Danilevskaya et al., 2008). Most strikingly, a single amino acid residue can make the difference between a floral activator and a floral repressor (Hanzawa et al., 2005). The floral activating function of Arabidopsis FT and the repressive function of TFL1 can be interchanged by swapping a histidine residue (at position 88 for Arabidopsis TFL1) for a tyrosine residue (at position 85 for Arabidopsis FT) (Hanzawa et al., 2005). AaFTL1 possesses the tyrosine residue that confers the activating function to Arabidopsis FT (Fig. S5) and AaFTL2 possesses the histidine residue that confers repressive function to Arabidopsis TFL1 (Supplementary Fig. S6), supporting the notion that AaFTL1 is a floral activator and AaFTL2 a floral repressor.

For Arabidopsis FT and TFL1, this residue is important for their function in the regulation of flowering, but it is not the only position that determines whether a FT/TFL1 protein functions like a floral activator or repressor. This is exemplified by Arabidopsis BFT, which has floral inhibitory activity, but possesses the ‘activator’ tyrosine residue that Arabidopsis FT has. To predict protein functionality more accurately, a sequence analysis was performed to determine what other amino acid residues can distinguish a floral activator from a floral repressor. This analysis was performed within either the FT-clade (for AaFTL1) or the TFL1-clade (for AaFTL2), to avoid creating a bias for positions that reflect evolutionary history rather than function. Within the FT-clade, there are activators, repressors, and flowering neutral proteins, and a consensus sequence was made of each of these three groups. The same procedure was followed within the TFL1-clade. For each consensus sequence, positions unique to that consensus (e.g. positions in the FT-activator consensus that do not occur in the FT-repressor consensus or FT-neutral consensus) were selected. Then, the A. artemisiifolia sequences were analysed at the selected positions. Within the FT-clade, AaFTL1 possesses a higher proportion of residues unique to the consensus activator (10/13), than to the neutral consensus (2/16) or to the consensus repressor (1/18) (Supplementary Fig. S5). AaFTL2 possesses two residues unique to repressors (2/9) and none unique to the neutral consensus (0/1) (Supplementary Fig. S6). In short, this analysis also indicates that AaFTL1 is a floral activator and AaFTL2 a floral repressor.

As an independent test for likely AaFTL1 and AaFTL2 function, we calculated scores that provide quantitative measures of sequence similarity for a test sequence to a group of proteins with shared function (see the Materials and methods for details). Unlike using consensus sequences, this analysis retains amino acid residues that are present in >50% of the sequences. For proteins in the FT-clade, the average activator ratio (i.e. similarity to flowering activators divided by similarity to non-activators) is significantly different between activator and non-activator groups ($\overline{\text{X}}$_act=1.16, $\overline{\text{X}}$_non=1.01, P<0.001), showing the usefulness of this score for the prediction of FT/TFL1 function (Supplementary Fig. S7A). AaFTL1 has a ratio of 1.18, which is similar to the average for known activators and well outside of the non-activator distribution, indicating that AaFTL1 has a floral activator function (Supplementary Fig. S7A). For TFL1-like proteins, the two functional groups have significantly different average repressor ratios ($\overline{\text{X}}$_rep=1.00, $\overline{\text{X}}$_non=0.97, P<0.001) (Supplementary Fig. S7B). AaFTL2 has a repressor ratio of 0.99, which is slightly lower than the average for repressors but higher than any non-repressor. Using an empirically determined cut-off to call as many proteins correctly as possible (0.98, calling only one true repressor falsely a non-repressor), AaFTL2 is called as a repressor, as is the closely related HaBFT. Together, the used sequence analysis methods consistently predict that AaFTL1 is a floral activator and AaFTL2 a floral repressor.

To establish experimental proof for AaFTL1’s ability to promote flowering and for AaFTL2’s ability to repress flowering, we transformed 35S:AaFTL1, 35S:AaFTL2, and empty vector controls into Arabidopsis. The expression of the constructs was confirmed by RT–qPCR (Supplementary Fig. S8A). In addition, the expression of two direct targets of Arabidopsis FT, SOC1 and FUL, was measured (Supplementary Fig. S8B). We observed that SOC1 expression was increased in both overexpressors, and that FUL expression was down-regulated in the AaFTL2 overexpression line. This was unexpected as we hypothesized that FTL1 and FTL2 would simply have opposite effects on FT targets. Part of the explanation may be the heterologous expression system together with the evolutionary divergence between Arabidopsis and A. artemisiifolia, or the use of the strong ectopic 35S promoter. Nevertheless, as SOC1 and FUL form heterodimers to regulate the floral transition (Balanzà et al., 2014), the observed expression changes predict accelerated flowering of AaFTL1 lines and delayed flowering of AaFTL2 lines. These predictions are consistent with the observed flowering time. AaFTL1-expressing plants flowered earlier than control plants, on average after 23 d instead of 31 d (P<0.0001), forming fewer leaves (11 instead of 17 leaves, P<0.0001; Fig. 4). AaFTL2-expressing plants flowered significantly later than the control plants, on average after 39 d (P<0.0001), forming more leaves (25 leaves, P<0.0001, Fig. 4). Together, functional tests and sequence-based predictions firmly established that AaFTL1 and AaFTL2 are floral activators and repressors, respectively.

Fig. 4.

Transgenic expression of AaFTL1 and AaFTL2 in Arabidopsis. Box plots of the duration of the vegetative phase in days to bolting (A) and number of rosette leaves at bolting (B) of T₁ plants containing pMDC160 (control, yellow), and of plants expressing the coding sequences of AaFTL1 (green) or AaFTL2 (red) under the constitutive 35S CaMV promoter. Box plots show the first, second, and third quartiles (the box), the 10th and 90th percentiles (the whiskers), and outliers as individual dots. Mann–Whitney U-tests were performed to test for significant differences between groups (***P<0.001; Bonferroni correction was applied with m=4 to the α of individual comparisons to obtain the indicated overall α values).

FT/TFL1 gene expression is associated with flowering time in A. artemisiifolia

Having confirmed the functionality of AaFTL1 as a floral activator and of AaFTL2 as a floral repressor, the expression of these A. artemisiifolia FT/TFL1 genes was investigated to determine whether interpopulation differences in expression level correlate with the differences in flowering time. The AaFTL1 and AaFTL2 transcript levels were determined at several different time points in A. artemisiifolia plants (Fig. 5; Supplementary Fig. S9). At the first time point measured (15 DAG), the invasive population already had higher transcript levels of the floral activator AaFTL1 than the native population (P=0.029; Fig. 5). The expression further increased at 31 DAG in the invasive population, but not in the native population (P=0.006). Only at 46 DAG did AaFTL1 expression levels in the native population reach the levels in the invasive population without much further change until 61 DAG (p_46DAG=0.44, p_61DAG=0.52). AaFTL2 expression was initially similar in the two populations (P=0.686), but increased in the native population at 31 DAG and stayed higher until at least 61 DAG (p_31DAG=0.028, p_46DAG=0.006, p_61DAG=0.028). The third FT/TFL1 gene was also investigated: its expression level was similarly low in both populations (see Supplementary Fig. S9). The subset of plants from both populations used for the expression analysis did not significantly differ in phenological characteristics from the larger populations used to generate Fig. 1 (Supplementary Fig. S10), indicating that the gene expression results are representative for the whole population. Repressive FT/TFL1-like proteins can have anti-florigenic function, being expressed in leaves and moving to the shoot apex to repress flowering (Foucher et al., 2003; Harig et al., 2012; Higuchi et al., 2013), but they can also be mostly maintainers of meristem indeterminacy, expressed at specific regions in the shoot apex to prevent conversion of the shoot apical meristem to a floral meristem (Shannon and Meeks-Wagner, 1991; Bradley et al., 1996; Foucher et al., 2003). AaFTL1 and AaFTL2 expression was therefore compared between A. artemisiifolia leaves and shoot apices. The results established that both genes are expressed in both leaves and shoot apices, but that expression levels of both genes were several orders of magnitude higher in leaves than in apices (Supplementary Fig. S11). This result suggests that AaFTL1 may represent canonical florigen function in A. artemisiifolia while AaFTL2 may be an anti-florigen similar to the closely related CsAFT instead of a meristematic maintainer of indeterminacy.

Fig. 5.

Expression of AaFTL1 and AaFTL2. Data points indicate average expression values for AaFTL1 (A) and AaFTL2 (B) of plants from the invasive (blue) and native population (red), respectively. The expression shown here is based on four (native) or 10 (invasive) biological replicates, each with three technical replicates. The data for the 15 DAG time point were obtained in a separate experiment with four pools of four plants each as biological replicates. Error bars show the SE. AaTUB was used as a reference gene. Mann–Whitney U-tests were performed to test for significant differences between populations (ns P>0.05, *P<0.05, **P<0.01). The vertical dashed lines indicate the average male floral initiation time for the invasive (blue) and the native population (red), respectively.

Discussion

Common ragweed plants introduced to Germany flower earlier than plants from the native range. The invasive plants have a shorter vegetative phase, while the male maturation phase remains unchanged. This early flowering trait comes with a cost in reproductive output. Plants from the invasive population produce 75% less fruits, and consequently less seeds, but without showing a significant increase in the size of the individual fruit. Because fruit development in A. artemisiifolia is frost sensitive, reduced fruit set of early flowering plants is a fitness disadvantage only in places with long vegetation periods such as the original range of the native population (Supplementary Fig. S12) and in invaded southern Europe. In contrast, under short vegetation periods, only early-flowering plants can produce any viable seeds, making the higher seed set of late-flowering plants irrelevant. In our study, the invasive plants flowered ~30 d earlier than the native plants, and one might wonder whether this difference is great enough to produce seeds successfully where the native population cannot. Our study was conducted under flowering-inducing SD conditions at optimal growth temperatures, showing the earliest times of flowering in the two populations. In the field, this difference becomes more pronounced at higher latitudes, presumably due to longer day lengths in the summer time (triggering flowering later), and lower temperatures (extending all phenological phases due to slower growth). The difference is ~30 d at Osijek (latitude N59°48'55'') and ~70 d in Uppsala (latitude N45°31'16'') (Scalone et al., 2016). After pollination, seeds can take 4–6 weeks to develop (Essl et al., 2015), which means that by early flowering, invasive plants could complete seed development in the common garden before the native plants have even started to release pollen. This indicates that indeed early flowering could mean the difference between some progeny or no progeny at all.

Plants from the studied invasive population not only were early flowering but also had increased dichogamy. It is thought that the self-incompatibility of A. artemisiifolia is leaky (Friedman and Barrett, 2008; Essl et al., 2015) and the architecture of the inflorescences and release of pollen onto lower tissue guarantee that pistils will encounter self-pollen, so additional blocks preventing inbreeding may be adaptive. For invasive species, however, it is thought that selfing is beneficial as populations are initially small and mating partners are rare (Baker, 1955, 1974). In fact, male and female mature phases still overlap considerably in invasive A. artemisiifolia, and effects on selfing efficiency may thus be limited. It is possible that increased dichogamy in invasive A. artemisiifolia is a by-product of earlier flowering rather than an independent adaptation.

Crossings of plants from both populations showed that the alleles underlying the shortening of the vegetative phase and of the male maturation phase are dominant, which (when adaptive) generally spread faster in a population than recessive alleles. The common garden experiment (Scalone et al., 2016) indicated that the shorter vegetative phase already allows for stable populations of invasive A. artemisiifolia north of the current distribution. Distribution modelling using phenological data of this invasive population confirms that this trait allows A. artemisiifolia to expand north without the need for additional climate change (Chapman et al., 2017). In addition, the shorter male maturation phase of F₁ plants indicates that the gene pool allows for additional phenological changes that potentially lead to adaptation to even higher latitudes.

The early-flowering trait of the analysed invasive A. artemisiifolia plants is accompanied by an earlier activation of a floral activator and lower expression of a floral repressor. It is not certain whether both these differences contribute to the difference in flowering time, but it is plausible: at 46 DAG, the activator AaFTL1 is expressed equally highly in plants of both populations, although plants from the native population flower only 3 weeks later (on average at 67 DAG). This delay may be caused by the transient increase of the expression of the predicted floral repressor AaFTL2 at ~46 DAG, preventing floral initiation by AaFTL1. At 61 DAG, the level of the repressor has decreased again, allowing for flowering even in the native population. Thus, the observed differences in expression of FT/TFL1 genes can explain the differences in flowering time between the native and the invasive A. artemisiifolia populations, and there is no evidence suggesting additional differences in FT/TFL1-independent pathways.

The expression of the two A. artemisiifolia FTL genes differs in three ways between the two populations: the invasive population has (i) a higher baseline expression of the floral activator AaFTL1 as well as (ii) an earlier up-regulation of that gene, and (iii) shows no considerable induction of the floral repressor AaFTL2. The causal mutation(s) for differences in flowering time are likely to affect both AaFTL1 and AaFTL2 expression but, because flowering was accelerated under both LD and SD conditions, it is likely that photoperiod-independent input to the FT/TFL1 floral integrators is altered in the invasive A. artemisiifolia population. Because of the large number of possible candidates, genetic mapping will be more appropriate than candidate gene approaches to identify causal mutations for early flowering in the invasive A. artemisiifolia population.

We noticed that the native population had larger variability in flowering time (time to male floral initiation) than the invasive population (SE of 4.6 and 0.9 d, respectively). The earliest plant from the native population flowered at 36 DAG, right in the middle of the distribution of flowering time for plants from the invasive population. This indicates that the alleles conferring early flowering in the studied invasive population could have pre-existed in the ancestral native population. Indeed, flowering time varies among different populations in the northern part of the USA, and there is an unidentified genetic component underlying this variation (Stinson et al., 2016). The invasive population is from a location similar in terms of the length of the growth period (when temperatures are >5 °C) to the northernmost border of the A. artemisiifolia distribution in America, for instance in Toronto (Supplementary Fig. S12). Wind dispersal of pollen can facilitate exchange of alleles between distant A. artemisiifolia populations, introducing alleles for early flowering even to more southern populations, from which the European populations may have originated. Future studies on the nature and distribution of alleles causing early flowering in A. artemisiifolia together with genome-wide scans for signs of selective sweeps will establish what impact pre-existing alleles had on the invasive success of A. artemisiifolia. Despite the aforementioned arguments for the change of flowering time as a common method of adaptation, it may be warranted to study other invasive A. artemisiifolia populations to determine how widespread this mechanism is.

Supplementary data

Supplementary data are available at JXB online.

Fig. S1. Overview of A. artemisiifolia phenology.

Fig. S2. Final plant characteristics.

Fig. S3. Linear version of the phylogenetic tree shown in Fig. 3.

Fig. S4. Phylogenetic tree based on the minimum evolution method.

Fig. S5. Alignment of functional consensus sequences and A. artemisiifolia FTL1.

Fig. S6. Alignment of functional consensus sequences and A. artemisiifolia FTL2.

Fig. S7. Activator and repressor ratios.

Fig. S8. Gene expression in transgenic Arabidopsis plants expressing AaFTL1 and AaFTL2.

Fig. S9. Average relative expression of A. artemisiifolia AaFTL1, AaFTL2, and AaFTL3.

Fig. S10. Comparison of flowering phenology of total and sample populations.

Fig. S11. AaFTL1 and AaFTL2 expression in leaves and shoot apices.

Fig. S12. Day length and vegetation period at the localities of the two studied A. artemisiifolia populations.

Table S1. Primers used in this study.

Table S2. A. artemisiifolia FT/TFL1 cDNA sequences.

Table S3. List of FT-like proteins with known and predicted function.

Table S4. List of TFL1-like proteins with known and predicted function.

Table S5. Standard errors of gene expression values for different reference genes as established by RT–qPCR.