Step-wise increases in FT1 expression regulate seasonal progression of flowering in wheat (Triticum aestivum L.)

Flowering is regulated by genes that respond to changing daylengths and temperature, which have been well-studied using controlled conditions; however, the molecular processes underpinning flowering in nature remain poorly understood. Here, we investigate the genetic pathways that coordinate flowering and inflorescence development of wheat as daylengths extend naturally in the field, using lines that contain variant alleles for the key photoperiod gene, Photoperiod-1 (Ppd-1). We found flowering involves a step-wise increase in the expression of FLOWERING LOCUS T1 (FT1), which initiates under day-neutral conditions of early spring. The incremental rise in FT1 expression is overridden in plants that contain a photoperiod-insensitive allele of Ppd-1, which hastens the completion of spikelet development and accelerates flowering time. The accelerated inflorescence development of photoperiod-insensitive lines is promoted by advanced seasonal expression of floral meristem identity genes. The completion of spikelet formation is promoted by FLOWERING LOCUS T2, which regulates spikelet number and is activated by Ppd-1. In wheat, flowering under natural photoperiods is regulated by step-wise increases in the expression of FT1, which responds dynamically to extending daylengths to promote early inflorescence development. This research provides a strong foundation to improve yield potential by fine-tuning the photoperiod-dependent control of inflorescence development.

molecular processes underpinning flowering in nature remain poorly understood. 23 Here, we investigate the genetic pathways that coordinate flowering and inflorescence 24 development of wheat as daylengths extend naturally in the field, using lines that 25 contain variant alleles for the key photoperiod gene, Photoperiod-1 (Ppd-1). We found

Inflorescence architecture measurements 146
Spikelet number was counted for the developing inflorescence at sequential stages. 147 At the double-ridge stage, the spikelet meristem ridge was counted as a spikelet. From 148 the lemma primordium stage onwards, spikelet meristems were clearly visible and 149 counted as a spikelet. For fully emerged inflorescences, both viable and non-viable 150 spikelets were counted. For Ppd-D1a insensitive and ppd-1 lines, data of early 151 inflorescence development are the average ± SEM of at least 3 replicates. For final 152 spikelet numbers, data are the average ± SEM of at least 10 replicates. For ft-b2 153 mutants, spikelet data are the average ± SEM of 5-7 replicates. 154

Seasonal and genetic regulation of Photoperiod-1 156
To investigate the seasonal regulation of the flowering pathway, we first measured the 157 response of Ppd-1 to increasing daylengths in the field. Specifically, we analysed 158 expression of Ppd-B1 and Ppd-D1 in photoperiod-sensitive wild-type plants (cv. 159 Paragon), as these homoeologues contribute the major photoperiod-insensitive alleles 160 that confer early flowering phenotypes of hexaploid wheat in global breeding 161 programmes (Ppd-B1a and Ppd-D1a, respectively) (Beales et al., 2007;Diaz et al., 162 2012; Shaw et al., 2013). Transcripts were measured over a series of photoperiods 163 defined by hourly increases in daylength from winter (9 h light/15 h dark) until late 164 spring (13 h light/10 h dark), and diurnal patterns were analysed to precisely detect 165 the daily peak(s) in gene expression ( Fig. 1; Fig. S1). Ppd-D1 and Ppd-B1 were 166 expressed at comparable levels to each other and they displayed very similar daily 167 expression profiles ( Fig. 1A-B, D-E, Fig. S2). The diurnal rhythm of Ppd-D1 and -B1 168 was maintained across all photoperiods, with transcripts peaking during the day (ZT 169 3-6 h) and at dusk, and dipping during the night (ZT 16-24 h). The consistent diel 170 pattern in relation to dawn and dusk indicates Ppd-1 expression adjusts to the 171 changing daylengths. The amplitude of Ppd-D1 expression was stable across all 172 photoperiods tested, and Ppd-B1 maintained a normalized expression range between 173 0.02 and 0.06, which was slightly higher at 10 and 13 h photoperiods ( Fig. 1A-B, D-174 E; Fig. S2). These results suggest that the seasonal regulation of flowering in field-175 grown wheat is not determined by quantitative changes in Ppd-1 expression. 176 To determine how photoperiod-insensitive alleles modify Ppd-1 expression under field 177 conditions, we used a NIL expressing the early flowering Ppd-D1a allele that contains 178 a 2.09 kb promoter deletion (Beales et al., 2007;Shaw et al., 2013). The photoperiod-179 insensitive Ppd-D1a allele altered the diurnal expression of Ppd-D1 from the 10-13 h 180 photoperiods by promoting higher expression late at night (ZT 20 and 24 h), relative 181 to wild-type (Fig. 1A). This difference was particularly significant at 11 and 12 h 182 daylengths; the difference in night-time expression was not detected in the 9 h 183 photoperiod. There were also minor changes in expression of Ppd-D1 during the day 184 in the 13 h photoperiod (Fig. 1A). The insensitive Ppd-D1a allele did not significantly 185 affect the amplitude of Ppd-D1 expression, relative to wild-type, especially during the 186 daytime when Ppd-D1 peaked between 3-6 h after dawn in both genotypes (Fig. 1A, 187 D). Ppd-B1 expression was also affected in the photoperiod-insensitive NIL, the day of the 10-13 h photoperiods, relative to wild-type, and, as expected, no Ppd-202 B1 transcripts were detected (Fig. 1B, D, Fig. S3). These results indicate that Ppd-1 203 plants. In wild-type, the differences in Ppd-D1 expression between field-and 220 glasshouse-grown plants were less dramatic -the transcript peaks were slightly 221 higher in glasshouse-grown plants during the day and the troughs were moderately 222 lower during the night, relative to the field (Fig. 1C). In the ppd-1 NIL, we observed a 223 similar dramatic increase in Ppd-D1 expression that was detected in the field, 224 supporting the suggestion that Ppd-1 forms a self-regulatory feedback loop (Fig. S3). 225 Step-wise increases in FLOWERING LOCUS T1 expression underpin seasonal 226

regulation of flowering 227
To investigate the seasonal progression of major flowering-time genes that act 228 downstream of Ppd-1, we analysed expression of FT1 and VERNALIZATION1 (VRN1) 229 ( Fig. 2, Fig. S4). In wild-type, FT1 expression was very low under 9 and 10 h 230 daylengths and moderately induced between 10 and 11 h photoperiods of late winter, 231 with a 3-fold increase in transcripts ( Fig. 2A-B, D). Under these daylengths, the diurnal 232 expression pattern of FT1 included peaks during the day and at dusk. This pattern 233 continued under 12 h photoperiods with a significant peak in the morning (Zeitgeber 234 Time (ZT) 3 h) and a minor peak after dusk (ZT 16 h). FT1 transcripts significantly 235 increased in spring between 12 and 13 h daylengths, with expression at 13 h 236 photoperiods being approximately 10-fold higher than that of 12 h ( Fig. 2A, D). Under 237 13 h daylengths, FT1 transcripts peaked during the day at ZT 6 h and again in the 238 evening (ZT 13-16 h). The ppd-1 lines unexpectedly displayed a similar diurnal pattern 239 of FT1 expression and near-identical responsiveness to the changes in daylength, 240 relative to wild-type; however, the amplitude of expression was significantly lower than 241 wild-type at all time-points of the 11-13 h photoperiods ( Fig. 2A, D). Conversely, the 242 insensitive Ppd-D1a plants showed dramatically different FT1 transcript activity 243 compared to wild-type, with expression detected much earlier than wild-type during 244 the 9-10 h photoperiods and showing much higher amplitudes (Fig. 2B, D). In the Ppd-245 D1a NILs, we detected low levels of FT1 transcripts under 9 and 10 h photoperiods of 246 winter, before expression increased significantly as daylengths extended at the 247 beginning of spring to 11 h, with levels comparable to that detected in wild-type under 248 13 h (Fig. 2B, D). FT1 expression then settled at 12 h daylengths, before increasing 249 again at 13 h -under photoperiods of 12 and 13 h, the diurnal expression pattern 250 included peaks during the morning (ZT 3-6 h) and evening (ZT 16-20 h). These results 251 show that Ppd-1 is required for robust expression of FT1, and that photoperiod-252 insensitive alleles promote rapid induction of FT1 under shorter daylengths of winter. 253 In wild-type, the increase in FT1 expression occurs in a step-wise process, with an 254 initial rise at the beginning of spring (11 h) followed by a stronger induction in late-255 spring (13 h; Fig. 2D). The relative differences in FT1 expression associated strongly 256 with flowering-time phenotypes, as the photoperiod-insensitive lines flowered 13 days 257 earlier than wild-type, while the ppd-1 NILs flowered 11 days later (Fig. 2E). 258 As we detected a difference in Ppd-1 expression in field-grown plants relative to 259 glasshouse conditions, we hypothesised that FT1 would display an altered 260 transcriptional profile under the two growth regimes. Contrary to the field, wild-type 261 glasshouse-grown plants expressed FT1 at significant levels under the 11 h 262 photoperiod, which stabilized as daylengths increased to 12 and 13 hours, indicating 263 that 11 h daylengths are sufficient to induce flowering in wheat (Fig. 2C). The ppd-1 264 lines displayed a similar trend with considerable expression at 11 h, which is 265 maintained through to 13 h; however, at 12 and 13 h there are significantly fewer 266 transcripts than wild-type throughout the day and night (Fig. 2C). In the photoperiod-267 insensitive lines, FT1 was induced at the 10 h daylength, with a dramatic increase in 268 expression as daylengths extend to 11 h, which is maintained through to 13 h ( Fig.  269

2C). At all time-points and daylengths, the amplitude of FT1 transcripts is higher in the 270
Ppd-D1a lines than wild-type and ppd-1 NILs -FT1 transcript levels of photoperiod-271 insensitive plants grown at 11 h are comparable to those of wild-type plants at 13 h 272 ( Fig. 2C). A similar diurnal expression pattern of FT1 was detected in all three lines, 273 with one peak detected in the morning (ZT 0-6 h) and another at dusk. The relative 274 differences in FT1 expression are reflected in the flowering-time phenotypes, as 275 photoperiod-insensitive lines flowered 20 days earlier than wild-type, while the ppd-1 276 NILs flowered 11 days later (Fig. 2E). 277 To investigate floral promoting genes that respond to temperature, we examined the 278 transcript levels of VRN1, whose expression increases following exposure to cold and 279 VRN1 expression was significantly higher in photoperiod-insensitive NILs, relative to 285 wild-type, and significantly lower in ppd-1 NILs (Fig. S4). These data indicate the early 286 flowering phenotype of photoperiod-insensitive wheat involves a coordinated increase 287 in expression of floral activating genes, including VRN1 and FT1. 288

Inflorescence development responds dynamically to changes in FT1 activity 289
To investigate the connection between the floral promoting pathway in the leaves with 290 IM development, we first examined developmental progression of inflorescences from 291 field-grown plants in the context of changes in FT1 expression (Fig. 3). In wild-type, 292 the SAM remained vegetative until the beginning of spring at the 11 h photoperiod, at 293 which point it transitioned towards the double-ridge stage ( Fig. 3A-C, Fig. S5). The 294 timing of this transition coincided with the three-five-fold increase in FT1 expression 295 that occurs between the 10 and 11 h photoperiods ( Fig. 2A, D). The IM remained at 296 the double-ridge stage until daylengths extended to 12.5 h, when it transitioned to the 297 glume and lemma primordium stages ( Fig. 3A; Fig. S5). The timing of this second 298 transition coincided with the ten-fold increase in FT1 expression that occurs in leaves 299 between the 12 and 13 h photoperiods of April ( Fig. 2A, D). The IMs remained at the 300 lemma primordium stage until daylengths reached 13.5 h. (Fig. 3A, Fig. S5). Growth 301 and development of the IM then proceeded rapidly beyond this point, with spikelets at 302 the lemma primordium stage forming floret primordia and reaching the terminal 303 spikelet stage when daylengths were 14 h (Fig. 3A, Fig. S5). Interestingly, transition 304 of the wild-type IM from glasshouse-grown plants followed a very similar progression, 305 only that key events occurred at a photoperiod approximately one hour prior, relative 306 to the field, to coincide with the earlier induction of FT1 expression (Fig. 3B). The IMs 307 transitioned to double-ridge and lemma primordium stages when daylengths were 10 308 and 11.75 h, respectively (Fig. 3B). These results show that the timing of the transition 309 for wild-type IMs to the double-ridge stage in the field coincides with an initial rise in 310 transcripts increased significantly (Fig. 3A, Fig. S5). The IMs then proceeded rapidly 316 to the glume and lemma primordium stages when daylengths were 11 and 12 h, 317 respectively, without the developmental pause observed in wild-type, which coincided 318 with the high expression of FT1 under these photoperiods (Fig. 2, 3A). The IMs arrived 319 at the terminal spikelet stage when daylengths were 12.75 h. The IMs of ppd-1 NILs 320 transitioned to the double-ridge stage at the 11 h photoperiod, proceeding to the 321 lemma primordium and terminal spikelet stages when daylengths were 14 and 15 h, 322 respectively (Fig. 3A, Fig. S5). Development of the ppd-1 inflorescences therefore 323 closely followed changes in FT1 expression in leaves, which initiated between the 10 324 and 11 h photoperiods, before increasing again at 13 h. In both the Ppd-D1a 325 photperiod-insensitive and ppd-1 NILs, the influence of Ppd-1 on spikelet number was 326 detected between glume primordium and terminal spikelet stages, with fewer spikelets 327 forming in the insensitive NIL and more developing in ppd-1, relative to wild-type (Fig.  328   3C). This trend in spikelet number for each genotype was also observed at maturity, 329 with photoperiod-insensitive lines producing shorter inflorescences with fewer 330 spikelets than wild-type, while ppd-1 NILs formed significantly longer inflorescences 331 with more spikelets (Fig. S6). Taken  in Ppd-D1a NILs earlier than wild-type, as determined by time since germination 351 (daylength) (Fig. 4A). These genes were expressed between the 10-11 h 352 photoperiods in Ppd-D1a NILs, but not in wild-type until the 12-13 h photoperiods, NILs, relative to wild-type. For the ppd-1 NILs, expression of these genes was inverted 357 relative to changes observed in the insensitive NILs, with transcripts detected at later 358 photoperiods than wild-type, at daylengths corresponding to the delay in FT1 induction 359 and IM development observed in these plants. The amplitude of AP1-2, AP1-3, VRN1, 360 SEP1-6, SOC1, MADS6 and FT2 transcripts unexpectedly reached the same 361 maximum level in wild-type as in Ppd-D1a NILs, demonstrating that meristem identity 362 genes are not expressed at higher levels in photoperiod-insensitive plants (Fig. 4). 363 Conversely, transcripts were significantly lower in the ppd-1 NILs, showing Ppd-1 is 364 required for robust expression of genes that promote spikelet and floret development 365 (Fig. 4). GNI1 transcripts spiked to higher levels in Ppd-D1a NILs, relative to wild-type 366 and ppd-1, at the green anther stage. Based on the seasonal shift in expression peak 367 detected for these genes in the insensitive and ppd-1 NILs, relative to wild-type, we 368 hypothesised that their expression was changing in relation to the developmental 369 stage rather than daylength. To test this hypothesis, we normalised gene expression 370 according to developmental stages including vegetative, double-ridge, lemma 371 primordium, terminal spikelet, white anther and green anther (Fig. 4B). Following 372 normalization, transcript levels were almost identical in Ppd-D1a NILs, relative to wild-373 type, supporting the conclusion that photoperiod-insensitive alleles advance 374 expression of meristem identity genes without increasing transcript levels (Fig. 4B). 375 An exception to this trend was FT2, which was higher in Ppd-D1a NILs at the lemma 376 primordium stage, relative to wild-type, in which transcript levels did not increase 377 significantly until terminal spikelet (Fig. 4B). In ppd-1 NILs, transcripts of meristem 378 identity genes were much lower and peaked at later developmental stages, relative to 379 wild-type (Fig. 4B). These field-based results show insensitive alleles provoke 380 accelerated but equal expression of meristem identity genes in IMs, relative to wild-381 type, and that Ppd-1 is required for timely and robust induction of genes that promote 382 spikelet and floret development. The expression of meristem identity genes for all 383 three genotypes is dynamically linked with the activity of FT1 in the leaves. 384

FT2 contributes to the termination of spikelet development 385
To further investigate the role of the initial induction of FT1 expression in leaves and 386 its relationship to IM development, we analysed the molecular and physiological 387 effects of maintaining plants at 10 h daylengths of late winter. We hypothesised that 388 the inceptive rise in FT1 expression in wild-type promotes the initial stages of 389 inflorescence development, and the second stronger induction is required to proceed 390 to later reproductive stages. To test this hypothesis, plants were grown in a glasshouse 391 under natural photoperiods until the daylength was 10 h, before being shifted to a 392 moving bench that maintained a fixed short-day 10 h photoperiod (Fig. S7). Wild-type 393 plants maintained at 10 h progressed to the lemma primordium stage on the same day 394 and produced the same amount of spikelets as plants maintained under natural 395 photoperiods, by which time the daylength had surpassed 12 h (Fig. 5A). However, 396 these plants stalled at the lemma primordium stage for an extra 30 days before arriving 397 at the terminal spikelet stage and produced more spikelets compared to plants 398 maintained under natural photoperiods (24 ± 0.3 vs 29 ± 0.5 spikelets, respectively; 399 Fig. 5A-C). The delay in inflorescence development coincided with the developmental 400 stage at which the second stronger induction of FT1 expression occurred ( Fig. 2A, D). 401 This delay was also observed in ppd-1 NILs, which produced more spikelets and 402 transitioned to the terminal spikelet stage significantly later than plants maintained 403 under natural photoperiods (25.4 ± 0.2 vs 28.3 ± 0.3 spikelets; Fig. 5B-C). The Ppd-404 D1a NILs transitioned to the terminal spikelet stage more rapidly than wild-type, but 405 there was a slight delay relative to natural photoperiods, as plants maintained at 10 h 406 produced more spikelets than glasshouse-grown plants (21.5 ± 0.2 vs 25.8 ± 0.4; Fig.  407 5B-C). Wild-type and ppd-1 NILs flowered later under the fixed 10 h photoperiods than 408 plants grown under natural daylengths, but the photoperiod-insensitive NILs flowered 409 at the same time (Fig. S7). In addition to increasing spikelet number, the 10 h 410 photoperiods altered inflorescence architecture of wild-type and the ppd-1 NIL by 411 forming elongated basal internodes immediately prior to ear emergence, as occurs in 412 ppd-1 mutants grown under constant long-days ( Fig. 5D; Fig. S7) (Shaw et al., 2013;413 Boden et al., 2015). The insensitive Ppd-D1a lines did not produce these elongated 414 internodes. These results indicate that the initial induction of FT1 is sufficient to 415 promote transition of the IM to the lemma primordium stage; however, the second 416 higher induction of FT1 is required to progress development of the IM to the terminal 417 spikelet and later stages. This delay is overridden in Ppd-D1a insensitive NILs, most 418 likely due to the higher expression of FT1 (Fig. 2). 419 To investigate genes that contribute to the progression of inflorescence development 420 beyond the lemma primordium stage, we investigated the role of FT2. We selected 421 FT2 because its expression increased dramatically in wild-type inflorescences 422 between the lemma primordium and terminal spikelet stages, and it was expressed 423 earlier in Ppd-D1a NILs and later in ppd-1 plants (Fig. 4). Expression of FT2 was much 424 lower in IMs at the glume primordium stage of plants shifted to the 10 h photoperiod, 425 relative to plants maintained in natural photoperiods, suggesting development of the 426 IM to the terminal spikelet stage is associated with robust expression of FT2 (Fig. 6A). 427 To test the role of FT2 genetically, we analysed two independent lines containing 428 missense mutations in FT2 of the B genome (FT-B2; G45D and R150C), which are 429 predicted to be deleterious for protein function (PROVEAN scores of -5.1 and -6.8) 430 Chan, 2015). The ft-b2 mutants produced more spikelets than their wild-type 431 NILs, indicating progression of the IM to the terminal spikelet stage was delayed (Fig.  432   6B-D). The ft-b2 mutants flowered later than wild-type, even under extreme long-day 433 photoperiods (Fig. 6E) increasing again as they extend to 13 h (Fig. 7). alleles cause FT1 to be expressed earlier in the season and to higher levels, relative 477 to wild-type, with the initial levels comparable to those detected in wild-type later in the 478 year (Fig. 7). The photoperiod-insensitive allele significantly affects In addition to a role regulating FT1, the higher levels of Ppd-D1 transcripts in ppd-1 501 lines indicates that Ppd-1 may influence its own expression through a self-regulatory 502 feedback loop. This potential function of Ppd-1 is supported by the interaction detected 503 between Ppd-1 homoeologues in the photoperiod-insensitive Ppd-D1a NILs, and is 504 consistent with Ppd-B1 expression being higher in Ppd-D1a mutants containing splice 505 site mutations (Boden et al., 2015). 506 The process of flowering involves communication of FT1 protein from the leaves to 507 the SAM, from which reproductive development is promoted through expression of 508 meristem identity genes (Shaw et al., 2013). In Arabidopsis and rice, overexpression spikelet identity genes being expressed at lower levels in ft-b1 mutants, we 513 hypothesised meristem identity genes would be more highly expressed in Ppd-D1a 514 NILs, relative to wild-type (Beales et al., 2007;Shaw et al., 2013;Boden et al., 2015). 515 Surprisingly, while photoperiod-insensitive lines accelerated meristem identity gene 516 expression to occur earlier in the season, the overall amplitude of transcripts was 517 identical to wild-type. In ppd-1 lines, induction of meristem identity genes was delayed, 518 and transcript levels were lower than those detected in wild-type, consistent with the 519 reduced activity of FT1 in these lines. The step-wise increase in FT1 expression in 520 leaves aligned strongly with the seasonal up-regulation of meristem identity genes and 521 progression of IM development, with the first induction of FT1 facilitating the vegetative 522 to double-ridge transition and the second rise promoting advancement to later stages 523 (Fig. 7). The accelerated peak in meristem identity gene expression in photoperiod-524 insensitive lines coincided with arrival of IMs at the terminal spikelet stage earlier in 525 the season. These data indicate that the regulation of inflorescence development and 526 spikelet number by Ppd-1 is not determined by the absolute level of meristem identity 527 gene expression, but by the timing at which the peak occurs. The advanced induction 528 of FT1 in the photoperiod-insensitive line potentially explains its ability to advance to 529 terminal spikelet when daylengths were maintained at 10 h, relative to wild-type and 530 ppd-1 that stalled at the lemma primordium stage. In addition to regulating spikelet 531 number, photoperiod-insensitive alleles also reduce floret fertility; the increased 532 expression of GNI1 at later stages in Ppd-D1a NILs, relative to wild-type, may explain 533 the decrease in fertile florets (Prieto et al., 2018;Sakuma et al., 2019). Taken together 534 with the seasonal analysis of FT1 expression, these results indicate that inflorescence 535 development is intimately connected with the activity of floral signals generated in 536 leaves, which dynamically respond to increasing daylengths. 537 Our analysis identified FT2 as a key regulator of spikelet development in hexaploid 538 wheat. Based on the ft-b2 mutants producing more spikelets, and the significant 539 increase in FT2 expression between the lemma primordium and terminal spikelet 540 stages, we propose FT2 helps determine spikelet number by promoting transition of 541 the IM to the terminal spikelet stage. This conclusion is consistent with analysis 542 performed in tetraploid wheat, which showed FT2 is expressed strongly in the 543 developing IM and that loss-of-function ft2 alleles delay flowering and increase spikelet