Variation in floral form of CRISPR knock-outs of the poplar homologs of LEAFY and AGAMOUS after FT heat-induced early flowering

Abstract Plant migration and gene flow from genetically modified or exotic trees to nearby lands or by crossing with wild relatives is a major public and regulatory concern. Many genetic strategies exist to mitigate potential gene flow; however, the long delay in onset of flowering is a severe constraint to research progress. We used heat-induced FT overexpression to speed assessment of the expected floral phenotypes after CRISPR knockout of poplar homologs of the key floral genes, LEAFY and AGAMOUS. We selected events with previously characterized CRISPR-Cas9 induced biallelic changes then re-transformed them with the Arabidopsis thaliana FLOWERING LOCUS T (AtFT) gene under control of either a strong constitutive promoter or a heat-inducible promoter. We successfully obtained flowering in both a male and female clone of poplar, observing a wide range of inflorescence and floral forms among flowers, ramets, and insertion events. Overall, flowers obtained from the selected LFY and AG targeted events were consistent with what would be predicted for loss-of-function of these genes. LFY-targeted events showed small catkins with leaf-like organs, AG-targeted events had nested floral organs consistent with reduction in floral determinacy and absence of well-formed carpels or anthers. These findings demonstrate the great developmental plasticity of Populus flowers during genetically accelerated flowering, which may be of horticultural value. They also provide an informative early view of floral phenotypes and apparent sterility from knockouts of both these gene targets.


Introduction
The movement or gene f low from genetically modified, domesticated, or potentially invasive plant species to wild lands or into sexually compatible relatives is a substantial public and regulatory concern.Many forestry and horticultural species are grown outside of their native range, and some have become invasive due to their ability to propagate within those environments [1].Many strategies have been developed to mitigate potential sexual dispersal and spread, both by conventional and transgenic or gene-editing approaches (reviewed in [2].These include natural cleistogamy, identifying naturally sterile varieties, creation of sterile hybrids, using sterile polyploids, and engineered sterility via transgenic or gene-editing methods.The mechanisms of engineered sterility include pollen or organ ablation, gene suppression by RNA interference (RNAi), or targeted gene editing of essential f lowering genes, among others.
As knowledge of the molecular biology of f lowering has grown, methods that seek to suppress or eliminate the functions of key, highly conserved f loral genes for genetic containment have grown.For example, RNAi directed against the f loral genes LEAFY (LFY) and AGAMOUS (AG) have been at least partly successful in attenuating fertility in apple [3], sweetgum [4], and poplar [5][6][7].Where the goal is complete disruption of gene function, CRISPR knockouts have become a powerful tool for genetic containment.
CRISPR is known to work with high efficiency in almost any animal or plant species, and there are many possible applications in woody species [8,9].Our previous work has established that CRISPR-Cas9 gene editing of f loral development genes is efficient at creating bi-allelic mutations in target genes in both Eucalyptus [10,11] and poplar [12].In the former case, we also demonstrated the use of FT-induced rapid f lowering to speed the evaluation of f loral form in Eucalyptus, for which constitutive overexpression worked well.However, as discussed below this method has low efficiency in poplar, prompting us to employ a heat induction system to stimulate early f lowering.
Many strategies have been tried to obtain faster f lowering in tree species.These include, altering the photoperiod [13], hormone and fertilizer treatments [14], finding and breeding rare naturally early-f lowering individuals [15], RNAi-based suppression of key f loral genes [16,17], and increased expression of key f loral regulatory genes by either viral expression or stable transformation.Viral induced expression has been used successfully to induce early-f lowering in domestic apple, and resulted in formation of viable pollen, a shortened breeding period, and the successful identification of virus-free progeny [18].
Most research on genetically-accelerated f lowering relies on stable transformation, for which a variety of promoter and gene combinations have been employed to overexpress f loral inductive genes (Table 1).Nearly all published reports include successful early f lowering, which indicates the robustness of this approach, and likely the tendency to publish positive results.For work within Populus, different clones showed different degrees of f loral induction, even for the same genetic construct tested at the same location [33,40], (Supplemental Table S1).Sometimes, in addition to the desired early f lowering, there were unintended impacts on vegetative form.For example, the use of the strong viralderived 35S promoter to control target gene expression often resulted in shortened plant stature, as was observed for poplar, plum, and eucalypts [10,21,28,32,41].Therefore, it can be challenging to parse out any vegetative differences if this accelerated f lowering system is used to assess vegetative impacts of genetic containment systems.Our prior study of CRISPR-Cas targeting of LFY in Eucalyptus used two separate populations for this very reason; one set of trees with CRISPR-derived gene edits and Arabidopsis thaliana FLOWERING LOCUS T (AtFT) overexpression was used for evaluation of f loral form, while CRISPR edited trees (in the absence of genetically accelerated f lowering) was used for the analysis of overall juvenile tree form and growth [10].
Here, our main goal was to use genetically-accelerated f lowering to obtain an early view of f loral morphology on male and female clones of poplar that have been CRISPR-mutated at the poplar LFY or AG genes.LFY and AG are known key f loral development genes and are highly conserved across plant species.Our prior work using RNA interference (RNAi) to suppress LFY and AG in poplar showed modification of female f loral form and loss of seed formation under natural f lowering in field conditions [10,21,42].We generated CRISPR-Cas9 constructs targeting either the single LFY or both AG-like genes, and successfully obtained a large number of independent events with bi-allelic targeting of one or both target genes [12].Selected knockout events were then retransformed with FT transgenes.In prior work, we tested a variety of f loral induction constructs to inform our choice of promoter and gene [40], for which a heat-inducible AtFT system was the most efficient.Such an approach should reduce the strong impacts of FT overexpression on vegetative form [19,21,28,29,35].This HSP-AtFT system has been previously used to evaluate male sterility in poplars [22,38].One challenge to heat-induced f lowering is that a change to lower temperatures is needed to obtain viable pollen formation [43].However, lower temperatures can also lead to f loral reversion to vegetative structures, which complicates evaluation of f loral form [40]. Therefore, we focused our efforts on evaluating overall f loral form without attempting to use lower temperatures to achieve viable pollen-especially as both LFY and AG knockouts are expected to lack any pollen production.We report a high rate of successful f lowering induction, and in knockouts we observed the expression of a wide range of f loral phenotypes that are congruent with expectation for these genes and suggest that there will be a high rate of f loral sterility in normally f lowering trees.

Optimizing early floral induction
We tested four different constructs for inducing early-f lowering.We achieved f lowering rates ranging from 0% to 48.5% of events (independent gene insertions) within each of the two tested clones (Supplemental Table S2).The promoter and gene combination which gave the highest rate of f lowering for both clones 717 and 353 was HSP:AtFT.Overall, female clone 717 had a lower rate of f lowering than male clone 353.Of the two different Populus tremula x Populus tremuloides clones, we achieved f lowering with clone 353 but not with clone T89.From these findings, we created an experimental plan to induce early-f lowering in selected CRISPR trees in clones 353 and 717 by re-transforming selected events with either constitutive or heat-inducible f loral constructs (Figure 1).

CRISPR-Cas9 event selection for re-transformation with floral-inducing constructs
A main goal of this work was to induce early f lowering in events with known changes to the f loral gene LFY or to both AG genes.We selected a set of previously characterized Cas9 control and CRISPR events in male clone 353 and female clone 717 [12].Events were chosen that had bi-allelic, putatively knockout mutations (encoding a non-functional protein) in LFY or bi-allelic changes in both AG1 and AG2 (Figure 2, Supplemental Tables S3 and S4).All chosen AG events were generated using two sgRNAs and were named DAG events (for double sgRNA in AG).Chosen LFY targeting events either had one sgRNA and were named LIC events, or two sgRNAs and were named DL (for double targeting of LFY) events.For chosen LFY events, the single targeting site was located within Exon 1, while the second sgRNA target site was located prior to the coding region, such that events with large deletions no longer retained the start codon.For the AG events chosen, the sgRNA target sites were located within Exon 1.All selected events were used for two approaches (Figure 1).Male clone 353 was selected for a constitutive f loral induction approach using the strong promoter 35S driving expression of AtFT, allowing for evaluation in a standard greenhouse.Events in both clones were re-transformed with a heat-inducible f loral construct for evaluation in a heat-induction greenhouse.

Floral induction in male clone 353 with 35S:AtFT
For our constitutive f loral induction approach, we transformed selected Cas9 and LFY events in male clone 353 with the constitutive 35S:AtFT construct.Our past work showed that this clone had a higher rate of genetically accelerated f loral induction than female clone 717 ( [40], Supplemental Table S2).We successfully obtained many re-transformants for analysis (341 ramets from 30 sub-events and 10 WT controls: Supplemental Figure S1).Of these, no f lowering was observed for the WT control trees, 16.8% of 35S:AtFT f lowering control trees produced f lowers, and no f lowering was observed for any of the 34 AtFT sub-events in the Cas9 event background.Retransformation of two selected LFY events (DL 143 and DL 106) gave 52.2% and 83.3% f lowering by sub-event (Supplemental Figure S1).Microscopy of selected AtFT f lowering control and f lowering sub-events from DL 143 showed the presence of well-formed anthers on the AtFT control catkins, while the DL 143 AtFT subevents had catkins with underdeveloped fuzzy f loral organs, an absence of well-developed anthers, and the presence of many leaf-like structures in the catkins.Due to the typically lower f lowering rate for female clone 717 we did not transform events in this clone with the 35S:AtFT construct, but instead chose to focus on heat-inducible f lowering.

Floral induction of male clone 353 and female clone 717 with HSP:AtFT
Our next approach was to retransform selected events in male clone 353 and female clone 717 with a heat-inducible f loral construct HSP:AtFT.Here we obtained 250 ramets in clone 353 representing 130 sub-events (Supplemental Table S5), and 170 ramets in clone 717 representing 97 sub-events (Supplemental Table S6).Due to space constraints, trees were tested one clone at a time.A female and male clone of hybrid poplar were selected as the initial starting materials.Both clones underwent initial transformation with either a Cas9-only control vector (lacking guide RNAs) or a CRISPR-Cas9 vector (including guide RNAs, called Cas9-only control).These events were then characterized by DNA sequencing of the target sites to identify allelic changes.Selected events from both female clone 717 and male clone 353 underwent re-transformation with the vector HSP:AtFT to obtain doubly transformed sub-events for a heat-induced f loral study.Selected events from male clone 353 also underwent re-transformation with the vector 35S:AtFT to induce f lowering in a standard greenhouse.
We had two adjacent heat-induction greenhouses, designated A and B, and identical sets of trees were present in each greenhouse for each heat induction.For clone 353, the overall f lowering rate was 22.4% of all trees, and this rate was similar between the two greenhouses (Supplemental Table S7).For clone 717, the overall f lowering rate was 34.1%, with greenhouse A showing 49.4% of trees f lowering while greenhouse B showed just 18.1% of trees f lowering (Supplemental Table S8).

Vegetative performance HSP:AtFT
We quantified vegetative performance of trees after heat-induced f lowering by measuring total tree height.For clone 353, we measured trees twice during the heat-induction period, once in March and once in April.Trees were scored as f lowering or non-f lowering to allow for detection of any growth differences associated with f loral onset.We found that grouping trees by initial event showed similar growth between groups, and that f lowering trees grew equally as tall as non-f lowering trees (Supplemental Figure S2).For clone 717, we completed the height measurements once during the heat-induced growth.We also found that groups of these trees showed similar growth, and that f lowering trees grew equally as tall as did the non-f lowering trees (Supplemental Figure S3).

Flowering and floral phenotypes of male clone 353 with HSP:AtFT
We tested a total of 250 ramets in male clone 353 representing 121 AtFT transformation and retransformation events (1 AtFT only, 20 AtFT in Cas9, and 100 sub-events from selected CRISPR-Cas9 trees: Supplemental Table S5).No f lowering was observed prior to heat-induction conditions.After heat induction, of the 250 total ramets 56 (22.4%) f lowered, 190 (76.0%) did not f lower, and 4 (1.6%) died (Supplemental Table S7).Whole trees appeared in general good health following heat induction (Supplemental Figure S4).We used photography followed by microscopy of selected ramets to examine f loral morphology (Figure 3).The HSP:AtFT control trees typically produced a terminal catkin at the apex of the tree, as well as numerous axial catkins.These catkins had a perianth cup and structures consistent with developing anthers (Figure 3).Cas9-only AtFT sub-events were similar in overall appearance to AtFT only f lowering control events.Occasionally, carpels were observed on catkins of AtFT control and Cas9 AtFT trees (Supplementary Figure S5).These carpels were typically present in the basal area of the catkin and were surrounded by anthers.Retransformation of AG events 142 and 167 led to production of atypical catkins (Figure 4).
In these re-transformation events, catkins were often elongated and contained multiple layers of f loral organs.Individual f lowers within the inf lorescence often had long pedicels.Clusters of small organs in perianth cups were typically green in appearance and these f lowers lacked well-developed anthers.Retransformation of LFY events 88 and 143 sometimes led to growth of branch-like structures in positions that gave rise to catkins in control events (Figure 5).
Microscopic examination of these structures showed they had many organs that were leaf-like in appearance, with numerous trichomes present and a green coloration (Figure 5).We occasionally observed small regions of reddish colored tissue (Figure 5 panel C) which may be partially developed anthers.
We tested a total of 170 ramets in female clone 717 representing 97 AtFT transformation and re-transformation events (1 AtFT only, 13 AtFT in Cas9-only, and 83 from selected CRISPR-Cas9 trees, Supplemental Table S6).No f lowering was observed prior to heat-induction conditions.After heat induction, of the 170 total ramets 58 (34.1%) f lowered, 107 (62.9%) did not f lower, and 5 (2.9%) died (Supplemental Table S8).After heat-induction trees from this clone appeared a bit stressed, with yellowing of lower foliage (Supplemental Figure S6).Only three of the nine AtFT f lowering control trees f lowered (Supplemental Table S6).Flowering controls typically had a terminal catkin with numerous carpels (Figure 6).
Re-transformation of LFY events 190 and 121 resulted in trees with branch-like structures near the tree apex, which gave these trees a bushy appearance in this location.Microscopic examination of these structures showed small underdeveloped fuzzy organs (Figure 7).
While we scored such trees as "f lowering," these organs had no obvious well-developed f loral structures.Re-transformation of AG events 27 and 316 showed similar outcomes (Figure 8).Here, the trees grew structures that had clusters of f lat replicated organs.Microscopic examination of these structures revealed a wide range of phenotypic diversity.The most striking example can be found in the variety of f loral forms from a single tree in event 316 (Figure 8).Here, the terminal catkin was somewhat normal in appearance, if underdeveloped.Other catkins and f lowers from this same tree showed layers of green, f lat f loral organs; similar f loral structures found in another region of this same trees had a rather ornamental appearance, with layers of pink petal-like organs.This phenotypic variation was also present between trees of the same AtFT sub-event, between trees from the same AG event but different AtFT sub-event, and between trees from differ-ent AG events (Supplemental Figure S7).All the AG knockout trees had replicated f loral organs, but the appearance of the organs, in terms of size and color, was variable.

Discussion
A main goal of this work was to further develop a system for inducing robust early-f lowering in both male and female poplar trees.Our initial tests of f lowering induction with a variety of constructs showed that male clone 353 typically f lowered at a higher rate than female clone 717, which is similar to previous  studies of genetically accelerated f lowering in greenhouse conditions (Supplemental Table S2, [40]).However, our prior work with a large-scale field trial of natural f lowering trees showed that clone 717 f lowered at a younger age and higher rate than clone 353 [44].These findings indicate that clone 353 may be more tractable for genetically accelerated f lowering but that this same clone takes longer to reach natural maturity in field tests than clone 717.We selected heat-induced f lowering as the primary approach to induce early-f lowering in our CRISPR events as this method gave a f loral induction rate of 31.0% in clone 717 and 48.5% in clone 353 and avoids pre-heat changes in tree morphology.
We selected CRISPR events that had been previously characterized regarding allele changes in target genes but had yet to undergo f loral phenotype evaluation [12].Chosen events had bi-allelic changes that disrupted predicted peptide formation and were likely to interfere with LFY or AG function (Figure 2, Supplemental Tables S3, S4).Re-transformation of our selected Cas9-only and CRISPR events was successful for both clones, and we obtained a large number of transgenic shoots harboring both the HSP:AtFT transgene and the initial Cas9 or CRISPR-Cas9 cassette.As event-to-event differences can be substantial for transgenic plants, we decided to maximize the number of retransformation events studied rather than choosing numerous ramets per event.The overall f loral induction rate of our 353 CRISPR study was 22.4%, which was lower than the 48.5% we found for our initial heat-induction study of this clone.Floral induction of 717 CRISPR events f lowered at a rate of 34.1%, which was slightly higher the initial heat-induction test of 31.0%.One possible reason for the lower f lowering rate in clone 353 is the timing of this experiment.Clone 353 ramets underwent the large-scale heat induction in February and March, which is a cool and cloudy time of year in western Oregon, USA.The heat-induction greenhouses are located on the building exterior and as such are inf luenced by the external environment, such as ambient air temperature and sun exposure.By contrast, clone 717 underwent large-scale heat induction starting in August, when outside temperatures are generally very warm (often above 30 • C in afternoons) and clouds are sparse.One advantage to heat-induced f lowering is it avoids the striking branchy and dwarf phenotypes seen when strong constitutive promoters driving FT are used [21,28].With our heat-induced FT system, both clones showed no significant difference in tree size between heat-induced f lowering and non-f lowering trees (Supplemental Figures S2, S3).Observations of overall tree form showed that most CRISPR trees were similar to AtFT f lowering control trees following heat induction (Supplemental Figures S4, S6).One exception was some ramets with CRISPR targeting LFY that developed a bushy appearance at the apex (Figure 7).This extra growth appeared to be due to the presence of branch-like vegetative structures in positions where catkins typically develop.
A primary focus of this work was determining impacts of CRISPR targeting of key f loral development genes on f loral form and fertility.We did not directly assess f loral fertility (e.g.pollen formation and germination), as producing fertile heat-induced poplar f lowers requires highly specific growth conditions [43].We instead focused on f loral morphology changes relevant to whether any gametogenic f loral organs would form at all.We found that, in general, targeting of LFY led to small, underdeveloped f lowers in both male clone 353 and female clone 717 (Figures 5, 7).These f lowers were often highly leaf-like or branchlike, where it could sometimes be a challenge to determine the type of organ being formed.These LFY targeted trees showed an absence of well-formed anthers or carpels, no evidence of male  or female gamete production, and the presence of f lat leaf-like structures.In our male clone, the LFY-mutated trees often formed replicated f loral structures, which was not a predicted outcome given that our field trial of natural f lowering with RNAi targeting of LFY in poplar gave reduced catkin size and very underdeveloped flowers [5,6].However, we observed a similar re-iterative f loral form in eucalypts with CRISPR targeting of the LFY homolog under f loral acceleration with a constitutive promoter [10].As also discussed in the eucalypt study, we believe that it is likely that continued exposure to FT leads to continued f loral meristem activity and thus continued f loral organ formation that would not occur under natural f lowering.
Targeting of both AG1 and AG2 led to f lowers with replicated f loral structures in both our male and female trees, indicative of a loss of f loral determinacy (Figures 4, 8).As AG genes are known to have a role in f loral meristem determination, this phenotype fits what would be predicted for loss of AG function and aligns well with our field observations of RNAi targeting of AG genes in female poplar [7].In our male AG-targeted trees, we observed an absence of anthers.In our female AG-targeted trees, we observed an absence of well-formed carpels.These data are consistent with a homeotic conversion of organ identity due to loss of AG function.For female clone 717, there was an absence of carpels in the CRISPR-targeted trees, which should lessen the ability of these trees to form ovules or wind-dispersible seeds.For male clone 353, there was an absence of anthers in the CRISPR-targeted trees, indicative of reduced potential for pollen formation, which would decrease the possibility of gene f low via pollen grains.Overall, our findings support that targeting of these key f loral genes by CRISPR would dramatically reduce, if not eliminate, the ability of the trees to undergo sexual reproduction.
We observed striking variation in f loral form, both within and between ramets and gene insertion events.Variable f loral form may be of horticultural interest, as a single plant could exhibit great variation in f loral form during the growing season under normal variation in temperature.This phenomenon has been observed in multiple species, with f lowering differing within a single shoot, between shoots of a single plant, or between individuals in a single population [45].This outcome speaks to the developmental plasticity of f lowering onset and organ development due to the variation in insertion sites of FT, different degrees of juvenility (and thus competence to respond to FT induction) depending on plant size and meristem position [40], and the wide variation in heating regimes within and between plants in the greenhouse environment (with typical wide spatial variation in temperature and lighting).In addition, f loral development is considered a highly polygenic trait, with expression dependent on many genes in addition to FT [46].There may be hundreds of genes affecting f lowering that may also have differential responses to high temperature microenvironments.Other laboratories have made similar observations; a similar gradient of f loral form, ranging from well-developed to more vegetative types of f lowers, has been observed in other studies of genetically accelerated poplar trees [47].
Chimerism for knockout mutations may be another cause of variation in f loral form within trees and events.However, initial characterization of CRISPR lines used to select these particular events for re-transformation found a very low level of suspected chimerism and no evidence of off-target changes for a subset of CRISPR sites similar to those targeted [12].A prior study of CRISPR targeting of the 4CL1 gene in poplar, where the phenotype can be assessed by a readily-observable wood color, showed no evidence of phenotype variation after 4 years of cutting back and re-growing shoots of selected transgenic events [48].Therefore, it is unlikely that chimerism can account for a substantial amount of the variation in f loral form that we observed.
We observed bi-sexual f lowers in clone 353 (Supplemental Figure S5).These f lowers were present on both AtFT-only f lowering control and Cas9-only control trees.Such bisexual f lowers on male clone 353 have been documented previously for genetically accelerated f lowering trees in the greenhouse [40].We did not observe bisexual or female f lowers on CRISPR knockout LFY events in 353 as we previously observed in naturally f lowering field-grown RNAi-LFY trees [44].This difference is likely attributable to the system used (RNAi vs CRISPR), as RNAi may not eliminate all target gene expression like CRISPR is expected to do.However, to validate these results, and all the effects of CRISPR knockout on f loral form and fertility reported above, it is essential to test trees under natural maturation and f lowering in the field.Field studies would mitigate possible side effects of heat shock and would allow for f lowering in the context of standard f loral regulators.This work will begin once natural f lowering of CRISPR knockout trees begin in a year or few.

Conclusions
Heat-induced rapid f lowering enabled us to observe strong effects on f loral form and fertility from knockouts of the LFY and AG genes in poplar.The marked variability in f loral form we observed shows the striking developmental plasticity of f lowering onset and morphology and suggests possible new means to create f loral diversity of horticultural value.

Peptide predictions and alignments
The previously obtained allele sequence information [12] was used to obtain predicted peptides using an online translation tool (www.expasy.org/translate).Predicted LFY peptides were aligned to compare WT LFY alleles of both 717 and 353 with mutants in those clones.Peptide alignments were created using Clustal Omega [49].

Vector construction, event selection, and plant transformation
To create the HSP:AtFT vector, the FT gene from Arabidopsis thaliana was cloned into the vector pCAMBIA under control of a heatshock inducible promoter (HSP, see supplemental Figure S8).This vector was used to transform WT control and the selected Cas9 control and CRISPR events.Five LFY mutants were selected for re-transformation, two in clone 353 and three in clone 717 (See Supplemental Table S3).Four AG double mutants were selected for re-transformation, two in clone 353 and two in clone 717 (see Supplemental Table S4) These mutants were previously identified and characterized regarding their genetic changes in the target genes [12].Mutants were selected for having predicted lossof-function changes to both alleles of either LFY or both AG1 and AG2.Cas9-only control trees were selected as having no changes in target genes.Allele-specific sequence information for these events can be found in Supplemental File S1.Transformation occurred using standard organogenic techniques with hygromycin selection [50].A total of 130 AtFT sub-events in clone 353 and a total of 97 AtFT sub-events in clone 717 were selected for heat induction (see Supplemental Tables S5, S6).Re-transformed events were confirmed by genotyping with the primers HSP FT-F01 5 -AGTGAAGGCATCGTATCAAGC-3 and HSP FT-R01 5 -CGCGGGATATCACCACTTTG-3 .Confirmed events were propagated by shoots to obtain multiple ramets per event.Small rooted ramets were transplanted to soil and grown in a greenhouse in the absence of purposeful heat induction, daytime temperature of 25C, nighttime temperature of 21C, with 16 hours of light and 8 hours of dark.Prior to heat induction, trees were trimmed then singled to retain one main shoot.Trees with a height of 40 cm or more were used for heat induction.

Heat induction
Two greenhouses were used for growth and heat induction of selected ramets, these were designated greenhouse A and greenhouse B. Trees were randomized within each greenhouse such that each greenhouse was an experimental block.Heat-induction greenhouses included supplementary fans and heaters to aid in air circulation and quick heating.Plants were well watered prior to the start of daily heat induction.The daily heat induction goal was 43C from 11:00 a.m. to 5:00 p.m., with a 2-degree heat increase for cloudy days.At 5:00 p.m. the temperature was set to 29 • C, this was reduced to 23 • C at 10:00 p.m., and increased to 29 • C after 6:00 a.m.A graph over average hourly temperatures per greenhouse and a view of one set of 717 trees prior to the onset of heat induction is shown in Supplemental Figure S9.

Plant growth measurements, floral scoring, and floral imaging
Ramet size was measured as stem height (soil level to plant apex) using a meter stick.All ramets were screened for the presence or absence of catkins or catkin-like structures.Ramets with at least 1 catkin or catkin-like structure were scored as f lowering.Events were scored as f lowering if at least 1 ramet from that event was scored as f lowering.Whole plants were photographed using a Canon Rebel XSI digital camera as were whole catkins.Detached catkins were imaged using a Keyence digital dissecting microscope.

Figure 1 .
Figure 1.Experimental overview.A schematic representation of the overall workf low.A female and male clone of hybrid poplar were selected as the initial starting materials.Both clones underwent initial transformation with either a Cas9-only control vector (lacking guide RNAs) or a CRISPR-Cas9 vector (including guide RNAs, called Cas9-only control).These events were then characterized by DNA sequencing of the target sites to identify allelic changes.Selected events from both female clone 717 and male clone 353 underwent re-transformation with the vector HSP:AtFT to obtain doubly transformed sub-events for a heat-induced f loral study.Selected events from male clone 353 also underwent re-transformation with the vector 35S:AtFT to induce f lowering in a standard greenhouse.

Figure 2 .
Figure 2. Locations of gene edits in LFY and AG genes and predicted peptides in LFY.(A) Guide RNA target sites in LFY are shown with scissors.Boxes show exons, thin lines show non-coding regions.Green arrows indicate start codons.Note that the first target site in LFY occurs prior to exon 1. (B) Guide RNA target sites in both AG1 and AG2 are shown with scissors.(C) Predicted LFY peptides in clone 717.Events 190 and 13 have a single guide RNA designed to target Exon 1.Note that while the start of the LFY peptide is conserved, the single deletion mutations cause a frame shift and early stop codon.Green and orange arrows indicate first methionine (M) in WT and edited alleles, respectively.Blue star shows the first predicted stop codon in the edited alleles.(D, E) Possible predicted peptides for the 120 bp deletion mutation in event 121.This event lacks the start codon but generates an in-frame deletion.(D) Use of the second in-frame M for LFY results in a peptide prediction missing the first several amino acids but retaining the c-terminus.(E) Use of the first in-exon M results in a very short peptide that shows little similarity to LFY and terminates early.Green and orange arrows indicate first M in WT and edited alleles, respectively.Blue star shows the first predicted stop codon in the edited alleles.Note that all alignments begin with the first M in LFY and show the N-terminus of the protein.LFY events in clone 353 have identical mutations to the 717 events shown here.

Figure 4 .
Figure 4. Male ag f lowers from clone 353 knock-out lines.Inf lorescences and close-up view of catkins from DAG double biallelic event 142 (A, C-F) and DAG double biallelic event 167 (B, G).Note the absence of anthers and f lat leaf-like appearance of the f loral organs.

Figure 5 .
Figure 5. Male lfy f lowers in a CRISPR knock-out line.Example of inf lorescence (A) and catkins (B, C) from DL event 143, tree 221.

Figure 7 .
Figure 7. Crown and f loral morphology of female lfy mutant in clone 717.(A) Apexes of f lowering control trees typically had a larger terminal catkin and numerous axillary catkins.By contrast, lfy knockout mutants such as event 190 tree 39 (B-D) had a very different appearance.(B) The tree apex had numerous projecting branch-like growths in positions typically occupied by catkins.(C) Examination of these structures revealed numerous, fuzzy, underdeveloped potential catkins.(E) Dissection of a single "catkin" showed an absence of carpels or other internal organs.

Figure 8 .
Figure 8. Intra-ramet and intra-catkin variation in f loral form in female clone 717.All images are from a single tree of knockout DAG event 316.(A) Highly-variable catkin morphology was observed for the apex of DAG event 316 AtFT sub-event 23, ramet 50.(B-I) Microscopy of individual catkins showed phenotype diversity between and within catkins from the same tree.Individual f lowers showed loss of well-formed carpels, presence of extra whorls of f loral organs, and presence of f lattened organs.

Table 1 .
Constructs used to induce early f lowering in trees by stable transformation.
C. sinensis L. Osbeck × P. trifoliata L. Raf. C. sinensis Osb.× P. trifoliata L. Raf.Examples of work in Populus species are shaded in grey.+ flowered in vitro.++ plants flowered once then stopped.a 1 event flowered b clone 353 flowered while clone 717 did not flower