FLOWERING LOCUS T paralogs control the annual growth cycle in Populus trees

Populus form when photoperiod falls below a certain followed by establishment of dormancy and cold hardiness At the of the photoperiodic pathway in Populus is the gene FLOWERING LOCUS T2 ( FT2 ), which is expressed during summer and harbors signiﬁcant SNPs in its locus associated with timing of bud set. The paralogous gene FT1 , on the other hand, is hy-per-induced in chilling buds during 3,5 Even though its function is so far unknown, it has been suggested to be involved in the regulation of ﬂowering and the release of winter dormancy. In this study, we employ CRISPR-Cas9-mediated gene editing to individually study the function of the FT -like genes in Populus trees. We show that while FT2 is required for vegetative growth during spring and summer and reg-ulates the entry into dormancy, expression of FT1 is absolutely required for bud ﬂush in spring. Gene expression proﬁling suggests that this function of FT1 is linked to the release of winter dormancy rather than to the regulation of suggested to be associated with callose synthesis and during callose plug formation and removal.

Another paralog is required for bud flush in spring. This function is linked to the release of dormancy rather than to the regulation of bud flush per se.

RESULTS AND DISCUSSION
Populus species have several FT genes The Populus FT1 and FT2 paralogs are the result of the salicoid whole-genome duplication event. 6 A more local FT2 duplication event has been described in European aspen (Populus tremula), 4 but the functional relevance is so far unknown. From here on, we will refer to the duplicated FT2 genes as FT2a and FT2b, where FT2a corresponds to the previously characterized FT2 gene. A 500 kb introgression event in the genome region harboring FT2a and FT2b was recently shown to be strongly associated with local adaptation. 4,7 Phylogenetic analysis revealed that FT2a and FT2b are present in Populus tremula, P. tremuloides, and P. trichocarpa, indicating that the duplication took place before the species separated ( Figures 1A and S1A). We then compared the gene synteny, the genomic regions surrounding the FT-like genes in Arabidopsis, P. tremula, and P. trichocarpa ( Figure 1A). Orthologous genes have a similar arrangement in all three species. The duplication around FT2 seems to have included the orthologous gene of FASCIATA1 (FAS1). However, in P. trichocarpa both FT2b and FAS1 are truncated, while both genes are full length in P. tremula ( Figure 1A).
Expression patterns of FT2 genes are similar but different from FT1 The expression patterns of FT1 and FT2a have previously been established, 3 but nothing was known about FT2b. Thus, we analyzed the expression of all three FT genes in our model species T89 (Populus tremula 3 tremuloides) as well as in field-grown mature Populus tremula (Figures 1B-1D). In our samples, FT2a and FT2b expression was limited to leaves in long days. FT2b was significantly more highly expressed than FT2a ( Figures 1B and 1C), but both followed a circadian rhythm with a peak at the end of the light period ( Figure 1B), similar to Arabidopsis FT. 8 FT2b overexpression also induced a dramatic early flowering phenotype (Figures S1B and S1C) similar to what has been reported for FT1 and FT2a. 1,2 FT1, on the other hand, was exclusively expressed in buds exposed to cold temperatures ( Figures 1C and 1D). In these buds, in situ hybridization revealed that the transcript is broadly present in shoot apex, embryonic leaves, and vasculature in February, but is undetectable in May ( Figure S1D).
FTs are required for vegetative growth Previous attempts to study the role of individual FT genes in Populus trees have been hampered by the fact that due to high levels  of homology it has been impossible to generate gene-specific knockdowns using RNAi or artificial microRNAs. To understand their individual roles, we generated specific knockout mutants for FT1 and FT2 using CRISPR-Cas9 ( Figures S2A and S2B) and subjected the mutant plants to a simulation of the changing seasons to examine their phenotypes ( Figure 2). FT2 has previously been identified as an important regulator of the timing of bud set since RNAi-mediated downregulation of FT expression leads to earlier bud set while FT overexpression prevents growth cessation and bud set. 1 SNPs at the FT2a locus are also very strongly associated with timing of bud set in Populus tremula. 4 However, surprisingly, knockout of FT2a expression had no visible effect on vegetative growth or timing of growth cessation/bud set ( Figure S2C). In contrast, knockout of FT2b expression had a clear effect leading to bud set already in long day conditions ( Figure S2D). However, a much more dramatic phenotype was seen in FT2a FT2b double knockouts, which displayed a severely dwarfed phenotype because of bud set even in tissue culture under 23-h-long days, and immediately after transplanting to soil (Figures 2A and 2E), suggesting that the FT2 genes are necessary to maintain vegetative growth. While FT2b appears to have the most important function, presumably linked to higher levels of expression compared to FT2a, the genes are partially redundant. A similarly extreme phenotype has been shown for GIGANTEA (GI) RNAi plants, which display very low FT2 expression. 9 However, in contrast to GI RNAi trees, FT2 doubleknockout plants were not impaired in their bud flush ( Figure 2D), suggesting that the FT2 genes have specific roles in the maintenance of vegetative growth and in the regulation of growth cessation and bud set. Recently, it was shown that CRISPR knockouts of FT2a lead to early growth cessation and inhibition of elongation growth in P. tremula 3 alba. 10 Since the presence of the active P. tremula gene FT2b reported here was not known at that time, the retention of this active gene likely explains the relatively weaker phenotypes reported in these CRISPR lines compared to our double ft2a ft2b lines.
While RNAi-mediated downregulation of FT2 expression had already hinted at their function, 1 only preliminary data regarding the phenotypes of ft1 and ft1 ft2 trees have been described. 11 In our experiments, disruption of FT1 function using CRISPR-Cas9 had no visible effect on vegetative growth or SD-induced growth cessation (Figure 2), confirming that the gene has no function during these processes when it is not expressed (Figures 1B-1D). However, after cold treatment to break dormancy and reactivation at warmer temperatures, ft1 plants were unable to flush their buds and only some plants flushed a few buds several months later ( Figures 2B-2D and 2F). This is a similar but stronger phenotype than the previously described preliminary data in Populus tremula 3 alba. 11 To exclude the possibility of ft1 buds simply having died during the cold treatment, we performed a viability staining, which showed that buds were indeed still alive ( Figure S4A). This shows that FT1 is required to resume vegetative growth after winter. Together, these results show that both FT1 and FT2 are required for vegetative growth: FT1 is required for bud flush and FT2 is required to allow vegetative growth and prevent growth cessation and bud set during summer.
FT2 is graft transmissible while FT1 function is restricted to its place of production We also investigated whether grafting on T89 could rescue the growth defect of FT CRISPR plants ( Figure S3). FT is a mobile graft-transmissible protein in Arabidopsis [12][13][14] and has recently been shown to also be a long-ranged signal in poplar. 15 We grafted both ft1 and ft2a ft2b scions onto wild-type (WT) rootstocks, as well as WT scions on mutant rootstocks. The early growth cessation of ft2a ft2b plants could be temporarily rescued ( Figure S3A) by a WT rootstock. However, as the shoot grew the WT rootstock could no longer support the growth of the ft2a ft2b scion, and it went into growth cessation again. Conversely, WT scions initially grew slowly on ft2a ft2b rootstock but then started to grow normally, presumably because they were now able to produce enough FT2 themselves, which in Arabidopsis typically occurs when leaves turn from photosynthate sinks to sources. 16 Grafting of ft1, however, did not rescue the delayed bud flush phenotype ( Figure S3B). WT parts of the grafts flushed simultaneously as the WT control regardless of their position. ft1 scions, rootstocks, and controls did not flush during the entirety of the experiment. These results indicate that FT1 is acting locally within individual buds. Since FT1 is expressed in embryonic leaves and vasculature ( Figure S1D) and has also been shown to be mobile, it is still possible that FT1 travels locally to the embryonic shoot apex, as suggested earlier. 5

FT1 is required for dormancy release
Since dormancy release is a prerequisite for bud flush, it is very difficult to know in which of these interconnected processes FT1 has a role, especially since there are no well-established molecular markers for dormancy release. 17 We therefore performed transcript profiling on WT and ft1 buds at different time points during an artificial growth cycle ( Figure 3) to see at what point in time lack of FT1 expression affected the transcriptome. The analysis showed that the transcriptomes of ft1 mutant plants look like those of WT controls up until 4 weeks of cold treatment (CTW4) ( Figure 3A). After 8 weeks (CTW8), when WT endodormancy is released, the WT had drastically changed its transcriptomic profile, while ft1 seemed to remain in the same stage as at Report CTW4 ( Figures 3A and 3B; Data S1A-S1F). This is also represented in the number of differentially expressed genes between WT and mutant trees at the different time points ( Figure 3C; Data S1G-S1I). After 7 days of warm temperature (LDD7; just before any visible signs of bud flush), the transcriptomes were again more similar, presumably due to a similar response to the temperature increase and the fact that bud flush had not yet started. The most affected Gene Ontology (GO) terms were ''catalytic activity'' for genes downregulated in WT between CTW4 and CTW8, and that remained high in ft1 versus WT at CTW8, and ''binding'' and metabolic process for genes with the opposite pattern of expression (Data S1G-S1I). A closer examination of the expression of genes previously suggested to be associated with the regulation of dormancy or bud set/bud break showed the most consistent changes between WT and ft1 at CTW8 in genes associated with GA metabolism and reception ( Figure 3D). In particular, the GA receptor GIBBERELLIN-INSENSITIVE DWARF (GID) genes are upregulated in WT at CTW8 while they are maintained at lower expression levels in ft1, suggesting a possible role for GA reception in the release of dormancy. Also, PICKLE (PKL), an antagonist of polycomb repression complex 2 whose downregulation has been shown to mediate the ABA-induced plasmodesmata closure and establishment of dormancy in Populus trees, 18 is induced in WT at CTW8 while it remains low in ft1, suggesting a possible involvement in dormancy release ( Figure 3D). It can be speculated that the very large number of genes changing in expression during dormancy release could be indicative of a more general chromatin remodeling releasing a repressed state of a large number of genes. The role of FT1 would then be to release the repressed state and make the genes accessible for later inductive signals.
Our data suggest that FT1 function is required for dormancy release rather than bud flush per se. This was further supported by moving WT and ft1 trees with non-dormant buds back to long days before dormancy was established. Under these conditions, both WT and ft1 trees flushed their buds normally ( Figure S4B), showing that FT1 is only required for bud flush after dormancy release. However, we cannot exclude that FT1 also has a role in post-dormancy-specific bud flush.
Since dormancy release occurs at a similar time as removal of plasmodesmata callose plugs, 5 we wondered how FT1 influences this process. After 12 weeks of short days, both WT and ft1 trees were dormant and had developed frequent electrondense plasmodesmata callose plugs, or dormancy sphincters, in apices ( Figures 4A and 4C), as shown before. 18 After a further 12 weeks of cold treatment, when WT dormancy has been released while ft1 trees are still dormant, no plasmodesmata callose plugs could be found in either WT or ft1 (Figures 4B and 4D). This shows that FT1 has no role in the removal of the callose plugs but is rather acting downstream of or parallel to this process. It also shows that it is not the removal of the callose plugs per se that determines the dormant versus non-dormant state. One possibility is that a local movement of FT1 to target cells in the shoot apex is absolutely required to release dormancy, and that the dormancy sphincters are needed to prevent this movement. However, it is still unclear to what extent opening of the dormancy sphincters is required for dormancy release after cold treatment, or if it is just a consequence of that release. Also, at the peak of its expression FT1 displays a broad expression in the shoot apex, in vasculature, and in the young leaf primordia ( Figure S1), making it unclear if a restriction of FT1 movement is relevant. A better understanding of direct FT1 targets is required to better understand the mechanism for dormancy release. Taken together, these data show that FT1 is required for dormancy release and the concomitant bud flush, even though its specific mode of action remains unknown.
Taken together, our data show that FLOWERING LOCUS T genes are indispensable for the correct regulation of the annual growth cycle, being critical both for the growth arrest and bud set in the fall and for bud flush in the spring, i.e., the start and stop of the growing season. They evolved from a whole-genome duplication, but despite their sub-functionalization, they still share a common role as growth promoters and are able to induce a shared set of genes when ectopically expressed. 3 As a consequence, both FT1 and FT2 ectopic overexpression leads to early flowering and prevention of growth cessation. [1][2][3] This suggests that the sub-functionalization is primarily driven by changes in the regulatory elements of the two genes, leading to completely contrasting expression patterns. This is reminiscent of the situation in sugar beet where two FT paralogs have evolved complementary expression patterns to regulate the yearly growth cycle. However, in this case there is also a neo-functionalization leading to one of the paralogs acting as a repressor instead of an activator of flowering. 19,20 Despite the previous focus on FT2a, it seems that in Populus tremula 3 tremuloides, FT2b can act redundantly in promoting growth and is of even greater relative importance, since knockout of any FT2 alone led to no (ft2a) or a slight (ft2b) growth phenotype ( Figures S2C and S2D) while double knockout almost completely prevented growth by an immediate trigger of the SD response ( Figure 2).
There are now several examples of how sub-functionalized FT paralogs have evolved within different species to regulate completely different aspects of plant development and growth, usually in response to photoperiodic cues. This includes, for instance, the regulation of flowering and tuberization in potato, 21 bulb formation in onion, 22 and short-day vernalization providing competence to flower in Brachypodium. 23 Besides the regulation of flowering, FT homologs also appear to act as general growth regulators in tomato 24 and maize. 25 Interestingly, recent work has shown that Gentiana trifolia, a herbaceous perennial, has two FT genes where one gene is expressed during the growing season to regulate flowering while the other is expressed in underground overwintering buds to release them from dormancy. 26 CRISPR knockouts of this latter gene lead to a reduced frequency of, and delay in, bud break. This is a very similar situation to what we have shown in Populus  Report trees and suggests that this type of FT sub-functionalization and control of dormancy release might be a more general feature of perennial plants.
Although FT1 has previously been suggested to be the FT paralog controlling flowering, 3 the situation in poplar is clearly different from the one in Arabidopsis. FT1 is not expressed in leaves and seems to be under no circadian control. Furthermore, it is specifically induced by low temperatures, which repress FT expression in Arabidopsis. 27 It is not known which transcription factors activate FT1 expression. SHORT VEGETATIVE PHASE-LIKE (SVL) has been shown to attenuate its induction, 28 but simple downregulation of a repressor seems insufficient to explain the FT1 hyper-induction. Our data clearly show that FT1 has another function, besides potentially flowering, in promoting vegetative growth after winter. To what extent FT1 and FT2 are actually needed to control flowering remains an open question.
Taken together, we show here that the FT paralogs in Populus trees have sub-functionalized to control major developmental transitions during the annual growth cycle, being required to prevent premature growth cessation and bud set in the fall and to induce bud flush in the spring, most likely by releasing the trees from winter dormancy. These are all critical aspects in a tree's ability to adapt to growth in different climates such as those experienced at different latitudes and altitudes, or as a result of climate change.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
This work was supported by grants from the Swedish Research Council, the Knut and Alice Wallenberg Foundation, Kempe Foundation, and the Swedish Governmental Agency for Innovation Systems (VINNOVA). We thank SNIC/ Uppsala Multidisciplinary Center for Advanced Computational Science for assistance with massively parallel sequencing and access to the UPPMAX computational infrastructure. We also thank Agnieszka Ziolkowska and the Umeå Core Facility for Electron Microscopy (UCEM) for help with electron microscopy.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
Hybrid aspen (Populus tremula x tremuloides) clone T89 was used as experimental model.

METHOD DETAILS
Plant material and growth conditions Hybrid aspen (Populus tremula x tremuloides) clone T89 was used as WT control and all genetic modifications were done in this background. Plants were cultivated on ½ Murashige and Skoog medium under sterile conditions for 4 weeks or until they had rooted (max. 8 weeks). After transfer to soil, plants were grown in growth chambers in LD (18h light, 20 C/ 6h dark, 18 C) and with weekly fertilization (10 mL NPK-Rika S/plant). Illumination was from 'Powerstar' lamps (HQI-T 400W/D BT E40, Osram, Germany) giving an R/FR ratio of 2.9 and a light intensity of 150-200 mmol m -2 s -1 . To induce growth cessation, plants were moved to SD (14h light, 20 C/ 10h dark, 18 C) for up to 15 weeks and fertilization was stopped. For dormancy release, plants were treated with cold (8h light, 4 C/ 16h dark, 6 C) for 8-10 weeks and then transferred back to LD for bud flush. In both SD and LD, previously published bud scores 32 were used to assess effects on bud development (set/flush). For year-around gene expression analysis, a ca. 40-year-old local (Umeå , Sweden) aspen tree was sampled once a month around midday (May to August leaves, buds from September to April).

Design and cloning of CRISPR constructs
Escherichia coli strain DH5a was used for amplification of all plasmids, which were then confirmed by sequencing (Eurofins). GreenGate entry and destination vectors 33 were acquired from Addgene. Potential sgRNAs for target genes were identified with E-CRISP (http://www.e-crisp.org/E-CRISP/). They were introduced into entry vectors by site-directed mutagenesis PCR. The final vector (containing promoter, Cas9 CDS, terminator, two sgRNAs and resistance cassette) was assembled by GreenGate reaction (150 ng of each component, 1.5 mL FastDigest buffer, 1.5 mL of 10 mM ATP, 1 mL 30U/mL T4 ligase and 1 mL Eco31l in a 15 mL reaction) in 50 cycles of 5 min restriction/ligation at 37 C and 16 C, respectively, followed by 5 min 50 C and 5 min 80 C. All reagents were purchased from Thermo Scientific.
Design and cloning of 35S pro :FT2b-GFP To create the 35S pro :FT2b-GFP fusion gene, the coding region (CDS) of FT2b was amplified from cDNA and cloned into the pGREEN-IIS destination vector 33 to C-terminally fuse it in frame to GFP under control of the 35S promoter. The final construct was transformed by electroporation into Agrobacterium tumefaciens and into ft2a ft2b mutant hybrid aspen trees. All plasmids were propagated using the Escherichia coli strain DH5a and verified by sequencing. GreenGate entry and destination vectors were obtained from Addgene. Primers used for plasmid construction are listed in Table S1.

Generation of CRISPR-Cas9 lines
Vectors with different combinations of guide RNAs (Table S1) were transformed into Hybrid aspen using a standard protocol. 34 At least 30 individual transgenic lines from each transformation were screened for target gene deletions using PCR (Table S1; Figure S2A). For each gene (FT1, FT2a and FT2b) at least two independent lines with homozygous, biallelic deletions were initially characterized for growth alterations before selecting one line for deeper analysis. All deletions were confirmed to occur at, or within a few nucleotides of, the expected PAM sites. Except for the ft2a ft2b double mutant lines, where both genes were confirmed to be homozygously deleted, target sites in FT2b were sequenced in FT2a CRISPR constructs and vice versa to exclude ''off target'' effects.

Grafting experiments
Scions of soil-grown plants were grafted onto rootstocks after 4 weeks (ft2a ft2b) or 5 weeks (ft1) in the greenhouse (18h light, 20 C/ 6h dark, 18 C). Scions were between 5 and 10 cm long and had no developed leaves, while the rootstock was decapitated ca 10 cm below the apex and kept its leaves. ft2a ft2b grafts were kept in these conditions until the end of the experiment, while ft1 grafts were transferred to SD (8h light, 20 C/ 16h dark, 18 C) after 5 weeks. After 10 weeks of SD treatment, plants were subjected to cold treatment as described above and returned to warm temperatures after 2 months. 5-8 plants per graft combination was used and 4 selfgrafted control plants per mutant line and wild type.
RNA extraction and quality assessment Poplar leaves were ground to fine powder, of which 100 mg were used for RNA extraction with CTAB extraction buffer 35 (2% CTAB, 100 mM Tris-HCl (pH 8.0), 25 mM EDTA, 2M NaCl, 2% PVP). The samples were incubated at 65 C for 2 min and extracted twice with necessary to make the probes are described in the following protocol https://kramerlab.oeb.harvard.edu/files/kramerlab/files/ in_situ_protocol_corrected-2.pdf?m=1430323911. The same protocol was also used as reference for the proper ISH experiment, with some minor changes. The hybridization temperature was set at 40 C, and the washes were performed at 50 C; for the tissue permeabilization we used 10 mg/mL of Proteinase K acting for 30 min. Sections (8 mm thick) were mounted on glycerol and visualized at Leica DMi8.
Viability staining ft1 buds were taken 15 weeks after the end of cold treatment, stained with 3.6 mM fluorescein diacetate solution (Sigma-Aldrich) and photographed under a stereomicroscope (Leica DMi8). ft1 buds from before cold treatment served as positive control, while the negative control was buds kept at -80 C for 3 days prior to staining.

Transmission electron microscopy
Both WT and ft1 apical buds were collected after growth in 12 weeks of short photoperiod and subsequently 12 weeks of cold treatment. Apical bud samples from three biological replicates were then fixed overnight in 4% paraformaldehyde and 2.5% glutaraldehyde in 1 M Caccodylate buffer (pH 7.2); post-fixed for 2h in 1% OsO 4 in water, dehydrated, infiltrated and embedded in Spurr's resin (TAAB Laboratories Equipment Ltd, England). Ultra-thin sections of 70nm thickness were stained with uranyl acetate and lead citrate and examined with the Thermo Scientific Talos L120C transmission electron microscope.

QUANTIFICATION AND STATISTICAL ANALYSIS
The statistical details of experiments can be found in the corresponding Figure legends.