Identi cation and characterization of a natural polymorphism in FT-A2 associated with increased number of grains per spike in wheat

Key message We discovered a natural FT-A2 allele that increases grain number per spike in both pasta and bread wheat with limited effect on heading time. Abstract Increases in wheat grain yield are necessary to meet future global food demands. A previous study showed that loss-of-function mutations in FLOWERING LOCUS T2 (FT2) increase spikelet number per spike (SNS), an important grain yield component. However, these mutations were also associated with reduced fertility, offsetting the beneficial effect of the increases in SNS on grain number. Here, we report a natural mutation resulting in an aspartic acid to alanine change at position 10 (D10A) associated with significant increases in SNS and no negative effects on fertility. Using a high-density genetic map, we delimited the SNS candidate region to a 5.2-Mb region on chromosome 3AS including 28 genes. Among them, only FT-A2 showed a non-synonymous polymorphism (D10A) present in two different populations segregating for the SNS QTL on chromosome arm 3AS. These results, together with the known effect of the ft-A2 mutations on SNS, suggest that variation in FT-A2 is the most likely cause of the observed differences in SNS. We validated the positive effects of the A10 allele on SNS, grain number, and grain yield per spike in near-isogenic tetraploid wheat lines and in an hexaploid winter wheat population. The A10 allele is present at very low frequency in durum wheat and at much higher frequency in hexaploid wheat, particularly in winter and fall-planted spring varieties. These results suggest that the FT-A2 A10 allele may be particularly useful for improving grain yield in durum wheat and fall-planted common wheat varieties. Supplementary Information The online version contains supplementary material available at 10.1007/s00122-021-03992-y.


Introduction
Wheat is a global crop of major economic value and nutritional importance as it provides around 54 20% of the calories and protein consumed by the human population 55 (http://www.fao.org/faostat/en/#data/FBS). However, with ever changing environmental 56 conditions and the rising human population, it is critical to increase wheat grain yield to meet 57 future demands. Yield is a multifaceted trait that can be partitioned into several yield 58 components, including spikes per unit of area, spikelet number per spike (SNS), grains per 59 spikelet, and grain weight. Several genes have been identified that affect these grain yield hexaploid wheat delays the transition to reproductive growth and increases SNS (Finnegan et al. 83 2018). 84 In addition to FT1, wheat has at least five FT-like paralogs designated as FT2 to FT6 (Lv et al. 85 2014), which have some overlapping functions but also varying degrees of sub-functionalization 86 (Halliwell et al. 2016;Lv et al. 2014). FT2 is the most similar paralog to FT1 (78% protein 87 identity), but the two genes still exhibit marked differences in transcription and protein 88 interaction profiles. Whereas the FT1 protein interacts with five out of the six wheat 14-3-3 89 proteins tested so far, FT2 failed to interact with any of these members of the Florigen Activation

119
Analysis of the exome capture data generated by the WheatCAP project using the assay 120 developed by NimbleGen (Krasileva et al. 2017 In parallel, we backcrossed the A10 allele into Kronos for three additional generations (Kronos 145 *5/Gredho), and then selected BC4F2 NILs homozygous for the A10 and D10 alleles using the 146 FT-A2 molecular marker. The BC4F3 seed was increased in the greenhouse in 2020 and the 147 BC4F4 grains were used for a second field experiment at UCD in 2021 that used small plots (four    (Table S1) amplify a fragment of 705 bp. After digestion with the restriction 165 enzyme ApaI, the fragment amplified from the D10 allele remained undigested, whereas the 166 fragment amplified from the A10 allele was digested into two fragments of 448 and 257 bp. 167 We used this marker to determine the frequency of the D10A mutation in 89 T. urartu, 82 T.   High resolution genetic map 181 We developed a high-resolution map of the KxG population in two phases. In the first phase, we 182 identified two BC1F3 plants from the KxG BC1F2 head rows, H2 and D12, which were 183 heterozygous for FT-A2 candidate region. From these heterozygous lines we generated large 184 segregating Heterogeneous Inbred Families (HIF) populations to identify recombination events 185 within the FT-A2 candidate region. Phenotype screens of these recombinants were space-planted 186 at least three inches apart in a completely randomized design. To generate additional markers in 187 the candidate gene region, we developed markers for 11 genes on both sides of FT-A2 covering a 188 region of ~10 Mb using the exome capture sequence data from Kronos and Gredho (Table S1).

210
Natural variation in FT-A2 211 We used exome capture data deposited in the T3 database (https://triticeaetoolbox.org/wheat/) to 212 explore the natural polymorphisms in FT-A2. We identified an A to C SNP at position  (Table 1). D10 was also the only allele 217 detected in all the other grass species we analyzed including Lolium perenne (AMB21802), 218 Oryza sativa (XP_021310907), Zea mays (NP_001106251), and Panicum virgatum (APP89655), 219 indicating that D10 is the ancestral grass allele. The Chinese Spring reference genome carries the 220 derived A10 allele, but in this study, we describe the change from the ancestral to the derived 221 allele rather than relative to the reference genome. 222 We also screened a collection of 417 T. turgidum ssp. durum accessions with a CAPS marker for 223 the D10A polymorphism (see Material and Methods) and found that only 0.7% carried the A10 224 allele ( Table 1). Two of the three accessions with the A10 allele were from Oman (PI 532239 = 225 'Gredho' and PI 532242, 'Musane and Byaza') and the other one was from Turkey (PI 167718), 226 suggesting that the A10 allele is almost absent from modern Western durum germplasm. 227 We detected a higher frequency of the A10 allele (56.5 %) among 705 T. aestivum ssp. aestivum 228 lines (Table 1). This overall frequency was similar to that detected in a worldwide collection of  Table 1.  Table 2). The strongest effect on SNS was detected for 250 PPD-A1, but a significant effect was also detected for FT-A2 (Table 2), with plants homozygous 251 for A10 showing 6.4 % higher SNS than those homozygous for D10 allele ( Table 2). The   (Table 3). The results from this experiment were 265 consistent between two independent pairs of BC1F3-5 sister lines (H2-14 and H2-23, Table 3).

266
The experiments in Tulelake (Northern California, spring planting) using BC1F3-5 sister lines 267 from family H2-14, also showed increases in SNS (4.0%), GNS (5.4%), grains per spikelet 268 (1.7%), and grain yield per spike (10.5%) associated with the A10 allele. However, the 269 magnitude of the differences between the FT-A2 alleles was smaller than those observed at the 270 2020 UCD experiment under fall planting. Only the differences in SNS were statistically 271 significant in Tulelake (Table 3).

272
For the 2021 UCD experiment using sister BC4F2:4 lines, we had more grains available and we 273 were able to use small plots (1.1 m 2 ) as experimental units, with 12 replications per genotype.

274
Lines with the FT-A2 allele headed on average 0.8 d later than those with the D10 allele (P = 275 0.0252) and showed significant increases in SNS (5.7 %, P = 0.0011) and GNS (6.3 %, P = 276 0.0168, Table 3). In this experiment we did not detect significant differences in grains per 277 spikelet (P= 0.7919). We observed a negative correlation between average GNS and grain  (Table 3). We harvested the complete plots and measured grain yield per plot and the 282 average yields of the two genotypes were almost identical: D10 = 1,254 ± 26 and A10 = 1,251 ± 283 32 g (P = 0.9103).

285
The A10 allele has a positive effect on SNS and spike yield in winter wheat 286 To analyze the effect of the D10A FT-A2 alleles in winter wheat, we used phenotypic data 287 available from 358 F5-derived RILs from the cross between soft-red winter wheat lines LA95135 locations. The A10 allele was also associated with an average 5.8% increase in GNS in the two 295 locations where this trait was measured, but the differences were significant only for the Pla19 296 location (2.7 more grains per spike, P < 0.001, Fig. 1c). The A10 allele was also associated with 297 a 1.2% increase in the number of grains per spikelet but the differences were not significant ( Fig.   298 1d). The differences in SNS were associated with a significant 4.6% increase in average spike 299 yield associated with the A10 allele in two out of the three tested locations (P < 0.001, Fig. 1e).

300
To delimit the QTL for SNS in this population, we performed ANOVAs for markers flanking 301 FT-A2 (Table S2) (Table 4). 319 In addition to the molecular marker for the FT-A2 D10A SNP and the two flanking markers, we 320 developed eight more KASP and CAPS markers in the candidate region (Table S1) (Table 4). Average SNS were as expected, with the lines homozygous for the A10 allele having 328 1.3 more spikelets on average than the lines homozygous for the D10 allele.

329
The phenotype of the critical recombinant line #18-5 with the closest distal recombination event 330 to FT-A2 was validated in a separate experiment in Davis in 2021 (Table S3). In this experiment, 331 control lines showed highly significant differences in SNS (P < 0.0001) confirming that the 332 differences in SNS were detectable in this experiment. By contrast, there was no significant  (Table S3). Taken 335 together, these results confirmed that the causal gene for the 3AS QTL for SNS was proximal to 336 the marker located at CS RefSeq v1.0 coordinate 120,227,651 (Table 4). 337 We identified an additional line (BC1F4 H2-18 #28-4) with a closer recombination event to FT-338 A2 in the proximal region between FT-A2-R1 and 3A-125.4, but we did not have enough grains 339 to evaluate it with the other lines listed in Table 4. We planted a separate field experiment at 340 Tulelake in the spring of 2020, in which we included homozygous sister lines #28-4-1 and #28-341 4-3 that were fixed for either the Kronos or Gredho alleles in the segregating proximal region 342 ( Table 5). As an additional control, we included sister lines derived from plant #17-2 (Table 4) 343 that were either homozygous for the FT-A2 D10 (#17-2-18) or A10 allele (#17-2-22, Table 5).

344
These two lines showed highly significant differences in SNS (P < 0.0001, Table 5) confirming 345 that it was possible to detect differences between the two FT-A2 alleles in this experiment. By 346 contrast, there was no significant difference between the H2-18 #28-4 recombinant sister lines, 347 13 confirming that the candidate gene was still linked to FT-A2 (Table 5). Based on this result, we 348 established a closer proximal flanking marker (3A-125.4), and reduced the candidate region for 349 the 3AS QTL to a 5.2 Mb interval between coordinates 120,227,651 and 125,402,254 (Table 5).  (Table   363 S2), so we ruled out R872H as the causal polymorphism for the SNS phenotype.

364
TraesCS3A02G143600 encodes a short peptide (104 amino acids) with a polymorphism in 365 Kronos that generates a premature stop codon (S59*, RefSeq v1.1 3AS 125,094,949 C to A). 366 However, the predicted protein in Gredho also seems to be truncated since it is much shorter     To explore if this lack of recombination in the central region was caused by a structural 452 rearrangement, we used the sequenced genome of the tetraploid variety Svevo (Maccaferri et al. 453 2019) that showed the same SNPs as the Kronos exome capture across the candidate gene region.

454
Since Gredho showed very few polymorphisms with CS across the candidate gene region, we 455 compared the genomes of CS (A10) and Svevo (D10) in this region. In Svevo, we found 456 orthologs to the 28 high confidence genes present in CS, with the exception of 457 TraesCS3A02G142500 that was present in the correct position and strand in Svevo (100% 458 identical over all its length) but was not annotated. All the genes were in the same orientation in 459 CS and Svevo, and the total length of the region was similar in both species (5.2 Mb), suggesting 460 that no major structural rearrangements occurred in the candidate gene region.

461
Finally, we did a BLAST comparison of all the Svevo genes to a Kronos scaffold assembly from 462 the Earlham Institute, U.K. and were able to detect 27 of the 28 genes with 100% identity. The show that the positive effect of the A10 polymorphism on SNS were translated into positive 486 effects on GNS in both the winter wheat population (Fig. 1e) and in the spring NILs (Table 3). 487 These results suggest that the A10 allele is not associated with negative effect on fertility. This 488 hypothesis was further supported by the higher number of grains per spike observed in the lines 489 carrying the A10 allele in the different field experiments, although the differences were 490 significant only in the two Kronos NILs evaluated in the field in 2020 (Table 3). These results 491 provide a good example of the value of using natural variants selected by breeders to identify 492 mutations that optimize specific traits with limited negative pleiotropic effects. 494 It was encouraging to see that the positive effect of the A10 allele on SNS and GNS was 495 expressed in both winter (Fig. 1) and spring wheats (Table 3), and among the latter in both spring 496 and fall planted spring wheats. However, the magnitude of the increases in SNS, GNS and spike 497 yield associated with the A10 allele varied among experiments, suggesting that the effects of this 498 FT-A2 polymorphisms on these traits are modulated by the environment. We also observed 499 variable effects of the A10 polymorphisms on grain weight. Whereas no significant effects were 500 detected for this trait in the experiments performed in UCD and Tulelake in 2020, we detected a 501 significant reduction in grain weight in field experiment performed at UCD in 2021, which offset 502 the gains in GNS (Table 3).  (Table 3).

533
(2021) suggested that differences in FT-A2 transcript levels may contribute to the differences in 534 DTH, but more precise mapping of the QTL will be necessary to support this hypothesis. The FT-A2 alleles show contrasting frequencies in durum and common wheat, with the A10 546 allele present in less than 1% of the durum accessions and in 56% of the common wheat varieties 547 analyzed in this study (Table 1). We currently do not know if the A10 allele originated in the few 548 durum accessions carrying this allele in Oman and Turkey, or if these represent later 549 introgressions from hexaploid to tetraploid wheat. Either way, since the appearance or transfer of 550 the A10 allele to common wheat, its frequency increased rapidly suggesting that it was favored 551 by breeders in common wheat breeding programs.

20
The low frequency of the A10 allele in durum wheat could be a result of an hexaploid wheat 553 origin combined with lack or infrequent transfers of genes from hexaploid to tetraploid wheat. planted locations will be necessary to test this hypothesis.

571
The high frequency of the A10 allele in the winter wheats and fall-planted spring wheats 572 provides additional evidence that this allele has positive effects in those regions. However, as the 573 frequency of the A10 allele increases, the number of varieties that can benefit from its 574 introgression decreases. By contrast, the A10 allele is almost absent from modern durum wheat 575 breeding programs, and may represent a good opportunity to benefit a large proportion of the 576 germplasm in the durum wheat programs. To facilitate the testing and introgression of the A10 577 allele into durum wheat breeding programs, we deposited the Kronos NIL with the A10 allele in 578 the NSGC (PI 699107). Kronos, is a modern durum wheat variety with excellent pasta quality, 579 which makes it a better donor parent than Gredho.

580
Our preliminary results suggest that the A10 allele may be more beneficial in fall planted than in 581 the spring planted durum wheat programs, but additional experiments are necessary to test this 582 21 hypothesis. It will be also interesting to investigate the combined effect of the A10 allele with 583 alleles from other genes that also result in increases in SNS such as Wapo-A1b (Kuzay et al. 584 2019) and the Elf3 allele from T. monococcum (Alvarez et al. 2016).

585
In summary, the genetic information provided in this study, together with the previous mutant 586 information, provides strong evidence that FT-A2 is the causal gene for the differences in SNS,

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.