Morphological characterization and genetic analysis of tri-pistil trait, a precisely regulated pistil number mutation in bread wheat

Bread wheat (Triticum aestivum L.) is an important source of nutrients for humans. Therefore, improvement of its yields is essential to feed the increasing world population. The tri-pistil (TRP) trait in wheat has a high potential for increasing yields. We obtained a pure tri-pistil wheat line, 4045, and evaluated its morphological properties. The 4045 wheat line stably produced three independently inherited pistils, which led to 1-3 grains in each oret. Among the three pistils, two lately emerged pistils initiated at late anther primordia stage to early tetrads stage. Genetic analysis revealed that there were TRP penetrance variations among the 11 F 1 populations of 4045. Fine mapping narrowed the single dominant TRP locus to a 97.3 kb region, containing two candidate genes, on the 2DL chromosome. However, further gene sequence, functional as well as comparative genomic analyses ruled out the only two candidate genes. Therefore, TRP is high-likely a unique gain-of-function mutation that does not exist in normal wheat genome. Transcriptome analysis of oral homeotic genes revealed that expressions of the C-class TaAG-2s, which are essential for carpel specication, signicantly increased in 4045, implying that TaAG-2s have played important roles in TRP-regulated tri-pistil formation. This study highlights that TRP leads to a precisely regulated pistil number increase (PRPNI) mutations and proposed a regulatory model of PRPNI pistil architecture. PRPNI offers a novel abnormal pistil development resource for research of oral architectures and potential on crop yield improvement. ne-mapped


Introduction
Bread wheat (Triticum aestivum L.), which is globally consumed, accounts for approximately one-fth of food sources for human beings (Asseng et al. 2020). Due to the constantly increasing global population, there is a need to develop crops with higher yields. However, since the Green Revolution, genetic gain rates in grain yields have been decreasing. Therefore, new resources and breeding approaches are required to improve grain yields (Foulkes et  In 1973, Chen et al. (1983) reported three wheat-grain combinations, three or two tightly combined grains, on the air-drying eld of breeding materials. Seeds were saved and developed into a pure wheat line, referred to as the "tri-grain wheat". Within the past fty years, studies evaluated oral morphologies, genetic make-up, biochemical features, and gene mapping of similar tri-pistil mutants, which were subsequently referred to as "multi-ovary" (MOV), "multiple-pistil" (MP), and "three-pistil" (TP) (Chen et al. 1999; Guo (Li et al. 2020c;Mahlandt et al. 2021;Peng 2003). When crossed with normal lines, all F 1 plants from different cross groups exhibited a tri-pistil trait. Additionally, segregation of the F 2 population exhibited a 3:1 or 1:2:1 ratio, implying that the tri-pistil trait is controlled by a single dominant gene (Mahlandt et Zhu et al. 2019). However, the candidate gene has not been identi ed. Since oral organ architectures in uence plant reproductive success and is a key agronomic factor determining yield, studies on tri-pistil wheat may elucidate on wheat oral establishment mechanisms and inform on viable approaches for increasing wheat yields.
Floral homeotic MADS-box genes, most of which belong to ABCDE class genes, play a critical role in oral architectures of most angiosperms (Callens et al. 2018;Theißen 2001). In owers of monocot plants like rice and wheat, lemma and palea in whorl 1 are speci ed by A and E class protein complexes, lodicules in whorl 2 are speci ed by A, B, and E class protein complexes, while stamens in whorl 3 are speci ed by B, C, and E class protein complexes. In whorl 4, C and E class protein complexes, along with another regulator, Dropping Leaf (DL) specify carpels (pistils), and D and E class proteins specify ovules (Callens et al. 2018; Ito 2019; Shen et al. 2021). The roles of AG-like C class genes or DL in pistil speci cation have not been conclusively determined. In rice double mutants of C class genes (OsMADS3 and OsMADS58), reproductive identity was lost, a palea/carpel-like organ was observed while oral meristem (FM) became indeterminate. In addition, OsMADS3 exerted greater effects, relative to OsMADS58 (Dreni et al. 2011;Sugiyama et al. 2019; Yamaguchi et al. 2006). Knockout of DL resulted in complete transformation of carpels into stamens (Yamaguchi et al. 2004). These ndings show the important roles of C class genes and DL in carpel specify. Dysfunctions of other oral homeotic genes severely affect oral architectures, accompanied by formation of pistillodies that were transformed from other oral organs. Knockout of AP3-like B class gene (OsMADS16) led to pistil-like stamens in rice (Callens et al. 2018;Nagasawa et al. 2003). In double knock-down mutants that were de cient in PI-like B class genes (OsMADS2 and OsMADS4), pistil-like organs transformed from stamens were also observed (Callens et al. 2018;Yao et al. 2008). In a rice Tos17 insertion mutant de cient in the D class gene (OsMADS13), ovule primordia developed into carpelloid structures (Callens et al. 2018;Dreni et al. 2007;Yamaki et al. 2011). Throughout oral establishment, E class gene products function by forming higher-order protein complexes with A, B, or C class proteins (Callens et al. 2018). The AGL6-like genes work similarly to E class genes (Reinheimer and Kellogg 2009;Su et al. 2019). Knockout of rice AGL6-like gene (OsMADS6) and down-regulation of TaAGL6 in wheat induced the generation of pistillodies (Ohmori et al. 2009;Su et al. 2019). Moreover, an increased number of pistils, which had been transformed from stamens, was also observed in about 52% of the owers of the strong mutant of OsMADS32 (cfo1-1), a monocot-speci c MADS-box gene, but not in its weak mutants Sang et al. 2012;Wang et al. 2015). Apart from oral homeotic MADS-box genes, other genes play important roles in pistil emergence and development. In rice, dysfunctions of MFS2, FON1 and FON4 (also known as FON2) resulted in increased number of pistils and stamens. Compared to pistillody, additional pistils in these mutants were distinct, normal and not transformed from other oral organs (Chu et al. 2006;Li et al. 2020b;Suzaki et al. 2004;Suzaki et al. 2006;Xu et al. 2017). In wheat male-sterile mutant dms, 1-6 pistils were observed along with 3 normal stamens (Duan et al. 2015).
In this study, we investigated the morphological and genetic features of tri-pistil wheat 4045, which is a pure tri-pistil wheat line, and ne-mapped the TRP locus, the gene involved in regulation of the tri-pistil phenotype in 4045. The tri-pistil wheat stably produced two extra late-initiated pistils, which emerged at late anther primordia (AM) stage to early tetrads (TS) stage (Feng et al. 2017), without alterations on other oral organs. Genetic analysis revealed that three grains in each 4045 oret are individually inherited while TRP penetrance varied in different F 1 populations. Segregation ratio of F 2 individuals indicated that the tri-pistil trait was controlled by a single dominant gene. Fine mapping narrowed the TRP locus to a ~97.3 kb region, with only two genes annotated. However, further gene sequence, functional and as well as comparative genomic analyseis rejected that the two candidate genes have been the casual genes. Furthermore, RNA-seq revealed that transcriptional abundance of oral homeotic genes that are closely associated with pistil development were signi cantly altered in 4045. Our ndings provide the basis for breeding as a strategy for increasing wheat yield through oral architectural modi cations.

Plant materials and growth conditions
The tri-pistil wheat line, 4045, and other parental wheat lines were developed and preserved in our lab. The blue-aleurone wheat, ZLWM, was provided by Professor Yizhe He (Northwest Agriculture and Forestry University). Plant materials, including parental wheat lines, 11 F 1 populations, 725 F 2 individuals for ne mapping and 3494 F 2:3 plants were grown at the experimental farm of Northwest A&F University in Yangling, Shaanxi, China during the 2014-2015, 2015-2016 and 2016-2017 wheat-growing seasons, respectively. During growth, the wheat lines were regularly watered and fertilized. Then, 101 F 2 individuals for primary mapping were vernalized for 45 days at 4℃ before being transferred to a greenhouse, where plants were grown under a photoperiod of 14-hour light:10-hour dark and temperatures between 18°C and 30°C until maturity. Plants for the two F 3:4 sub-populations were grown in a greenhouse under the same conditions without vernalization as described above, until being sampled for DNA extraction and genotyping.

Tri-pistil phenotype investigation and calculation of penetrance
To establish the tri-pistil phenotype of 4045, F 1 plants from the 11 crosses and 826 F 2 individuals were investigated by dissecting the orets and counting the grains inside. Two spikes for each plant (apart from those tillered on only one plantlet) were dissected oret by oret and different grain types, including regular grains, un-developed pistils and un-lled grains (Fig. 1d), carefully counted. Based on the observations, tri-pistils and double-pistils were considered mutant phenotypes. Penetrance was calculated as (No. of tri-pistil orets + No. of double-pistil orets) / No. of total orets×100%). Phenotypes of recombinants screened out from F 3:4 individuals were examined by visual observation of pistils at owering stages.
Optimal microscopy and Scanning electron microscopy Photomicrographs of oral organs and developing grains were obtained by Olympus stereo microscopy (SZX1603040112). Young spikes at different growth stages, according to Waddington et al. (1983) and Feng et al. (2017), were obtained and xed in formaldehyde-alcohol-acetic acid (FAA), dehydrated through a standard ethanol series, and critical point dried using CO 2 . Dried tissues were coated with platinum and imaged by scanning electron microscopy (JSM-6360LV03040415).

Genetic map construction and primary mapping
The genomic DNA of 101 greenhouse-grown F 2 plants along with 4045 and ZLWM were extracted by the CTAB method. DNA samples were hybridized on a Wheat90K SNP array containing 81587 SNP markers by China Golden Marker (Beijing, China). Subsequently, polymorphic SNPs between 4045 and ZLWM were selected. Among them, SNPs with a missing rate higher than 5% or didn't follow the Mendelian segregation (1:2:1) were deleted. A total of 3857 eligible SNPs were randomly binned using the Bin function of IciMapping 4.1 (https://www.isbreeding.net/software/?type=detail&id=18). Bin markers were departed into different linkage groups with a logarithm of odds (LOD) threshold score of 5.5, after which markers were ordered and the genetic map rippled with default parameters. Then, the TRP locus was identi ed using the Bip function of IciMapping via the inclusive composite interval mapping (ICIM) method. The genetic linkage map was drawn using the MapChart 2.3 program (https://www.wur.nl/en/show/Mapchart.htm).

KASP Marker development and genotyping
The SNPs used for KASP marker development were obtained from the Wheat 90K SNP array or detected by RNA-seq. ~50 bp anking sequences on both sides of each SNP were uploaded to a polymarker (http://www.polymarker.info/) for KASP primer designing. First, designed markers were detected by DNA samples of 42 F 2 individuals along with parental, heterozygous and empty controls. Markers that could genotype the samples into 3 groups (2 parental type and 1 heterozygous type) were selected and used for genotyping.
The genomic DNA of 725 eld-grown F 2 individuals, 3494 F 2:3 plants and F 3:4 plants were extracted using the CTAB method after which DNA samples were genotyped by developed KASP markers for recombinant screening. The recombinants were re-sampled and re-genotyped to con rm their genotypes.

RNA-seq and bioinformatics analysis
Total RNA of young spikes at W5.5 and W7.5 of 4045 and ZLWM were extracted using the RNAprep Pure Plant Kit (Tiangen Biotech, Beijing, China). FPKMs were calculated at BMKCloud (http://www.biocloud.net/). The two samples of 4045 were mixed and performed full-length RNA sequencing on a Paci c Biosience RS II platform. RNA-seq and full-length RNA sequencing were performed by Gene-Health Biotech (Shijiazhuang, China). Collinearity analysis was conducted by the program at Triticeae-gene Tribe (Chen et al. 2020b). Alignment of nucleic acids and protein sequences were performed using the ClustalX2 program (http://www.clustal.org/) and modi ed using the GeneDoc program (http://www.nrbsc.org/gfx/genedoc/). Floral homeotic MADS-box genes used for expression analyses (Supplemental Table 6) were previously identi ed by Schilling et al. (2020). The transcriptional expression heatmap was created by TBtools software ).

Gene cloning
The genomic fragment of TraesCS2D03G1038400 and its promoter region, cDNA fragment of TaMADS32-B, TraesCS2D03G1038200 and TraesCS3B03G0814800 were ampli ed by KOD DNA polymerase (TOYOBO, Japan) using the primers listed in Supplemental Table 7. Then, fragments were cloned into the entry vector pEASY-blunt zero cloning vector (TransGene Biotech, Beijing, China) and analyzed by Sanger sequencing.

Morphological characterization of the Tri-pistil wheat 4045
To elucidate on the tri-pistil trait, spikes, especially orets in 4045, were investigated. And Zhiluowumai (ZLWM), a blue-aleurone (Ba) wheat that is derived from wild blue-aleurone germplasm discovered in Qinling Mountains (China), was used as the normal single-pistil control in this study. At the late lling stage, spike length of 4045 was about 15 cm while its orets were stuffed to bursting due to the 2-3 grains inside (Fig. 1a). At maturity, the spikes were harvested after which 10 of them were randomly selected for further investigations. Compared to ZLWM (Fig. 1c), tri-grains of 4045 had three grains that were tightly combined (Fig. 1b). There was 1 bigger grain and 2 smaller grains (Fig. 1b, d), implying an asynchronous development for the combined grains. However, not all orets produced tri-grains. Florets of the 10 spikes were dissected and analyzed. In the investigated 460 orets, 17% (78 orets) contained 3 grains (Fig. 1d, Supplemental Table 1). However, deeper in the orets, undeveloped pistils stacked to the grains were detected. Besides, un-lled grains, which only contained seed coats without embryos or endosperms, were found in a few of the orets. These two forms of grains were considered to be aborted grains. Therefore, we further divided the orets into 6 types, that is, orets containing; i) Three grains (17%), ii) 1 normal 2 aborted grains (42.6%), iii) 2 normal 1 aborted grain (31.7%), iv) 2 normal grains (1.3%), v) 1 normal 1 aborted grain (5.4%), and vi) 1 normal grain (2.0%) (Fig. 1d, Supplemental Table 1).
The i, ii and iii grain forms, which had been developed from three pistils, accounted for 91.3% of the orets, and were considered to exhibit tri-pistil phenotypes. The double-pistil type (iv and v) accounts for 6.7% while the single-pistil type (vi) accounts for 2% of total orets (Supplemental Table 1). Both tri-pistil and double-pistil are mutant phenotypes. Then, we calculated the penetrance of the mutant phenotype. The 10 spikes had an average penetrance of ~98.0%, among which, the penetrance of ve spikes was 100.0% with the lowest one being 92.1% (Supplemental Table 1). These ndings imply that the mutant phenotype of 4045 was very stable. Notably, double-pistil orets were only present in the top two and bottom two spikelets, which developed later, relative to the middle ones. Single-pistil orets were only found in the top two spikelets. These results imply that all orets are supported to produce three pistils, and aborted pistils/grains in some of the top and bottom orets are likely to occur due to late development. Moreover, due to the aborted grains, the number of (normal) grains per spike did not reach the expected levels (Fig. 1d, Supplemental Table 1). However, the 4045 plants grown in a greenhouse under ideal conditions (like Fig. 1a, which was a spike from green-house grown plants) exhibited a much higher seed set, of which GNPS reached approximately 120. These ndings imply that the TRP trait in 4045 is sensitive to environmental conditions and that grain yield may potentially be better if ideal growth conditions are provided.
To elucidate on the oral 4045 architecture, we investigated the development processes involved in in orescences of 4045 grains. Young spikes of 4045 and ZLWM at different growth stages were analyzed by scanning electron microscopy (SEM). At the double ridge (DR) stage, spikelet meristem (SM) emerged and started to develop (Fig. 2a, e). As it enlarged, glume primordium (GP) was initiated, this stage is hereby referred to as the glume differentiation (GD) stage (Fig. 2b, f). Then, lemma primordium (LP) and FM appeared (Fig. 2c, g). The next stage is the anther primordia (AM) stage, in which the stamen and pistil primordia emerged from FMs with visible anther primordia for some orets (Fig. 2d, h). Prior to anther primordia (AM) stage, there were no signi cant differences between 4045 and ZLWM (Fig. 2, a-h). However, at the tetrads (TS) stage, as young orets started to differentiate to produce immature anthers and unelongated pistils, two extra pistils were markedly observed in 4045, but not in ZLWM (Fig. 2i-l).
Consistent with different sizes of three grains in one oret, the two extra pistils, which were referred to as secondary pistils (SP), were at an earlier developmental stage, compared to rst differentiated pistil (primary pistil, PRP) (Fig. 2k, l). Besides, three pistils were observed in almost all of the orets, except for a few of those at the top or bottom of spikelets, in accordance with the phenomenon that double-grain and single-grain orets only appeared in the top two and bottom two spikelets of 4045. These ndings imply that differentiation of SPs stably occurred at the late AM stage to early TS and later than differentiation of PRPs. The development of the other oral organs in tri-pistil orets was completely normal.
Optical microscopy was performed to investigate 4045 pistil and grain development at heading to lling stages. In conformity with morphological features of young spikes and mature seeds, two smaller-sized extra pistils were observed in 4045, compared to normal pistils in ZLWM at heading stages, while stamens and lodicules did not exhibit any differences (Supplemental Fig. 1). Accordingly, secondary grains that had developed from SPs were smaller, relative to primary grains (Supplemental Fig. 1d-i).
Taken together, the two extra SPs, which had been initiated at late AM stage to early TS stage, were the only altered oral organs in 4045, compared to wild-type wheat ( Fig. 1 and Supplemental Fig. 1). In addition, all oral organs, including stamens and three pistils were normal and fertile. Given that a number of aborted grains were present in the orets, the name "tri-grain" seems inaccurate, compared to "tri-pistil". Similarly, the number of entire pistils (including an ovary, two styles, and two stigmas), not the ovary, was increased and the number was always three, therefore, "multi-ovary", "multi-pistil" or "multigrain" was not the perfect name. Besides, it has not been conclusively determined whether the mutation in 4045 is allelic to the TP identi ed by Peng (Peng et al. 2008). Therefore, we named this phenotype in 4045, which is controlled by the TRP gene, as tri-pistil (TRP).

Genetic analysis of TRP
The three pistils in tri-pistil wheat are distinct from each other (Fig. 2, Supplemental Fig. 1). Combined to delay-developed SPs, we hypothesized that the three pistils in each oret of 4045 were independently inherited. To validate this hypothesis, we crossed 4045 with ZLWM (Fig. 1c, Supplemental Fig. 2b) and the Ba phenotype was used as an indicator to evaluate the segregation of tri-pistil wheat offsprings. Since the Ba phenotype is determined by the genotype of the sperm and two polar nuclei, segregation can be determined by the seed color of F 1 plants. As expected, grains in orets of F 1 plants exhibited different colors (Fig. 3). In some of the orets, three grains appeared in three colors (white, light blue, and dark blue), which was attributed to different numbers of Ba genes in fertilized polar nucleus (Fig. 3b). These ndings indicate that the three pistils in each oret of 4045 independently underwent meiosis and double fertilization.
For further characterization of genetic features of TRP, 10 more wheat lines were crossed with 4045 after which 10 spikes of different F 1 plants from each cross were investigated. There were variations in penetrance of F 1 among the 11 crosses (Supplemental Table 2). Among the 10 crosses, apart from 4045 × ZLWM, penetrance was very low and in three of them, TRP was completely recessive (Supplemental Table 2 Table 2), F 2 individuals with penetrance rates between 0-92% were considered heterozygous. Supplemental Fig. 3 shows that the ratio of offspring phenotypes was in accordance with SNPs that did not adhere to mendelian segregation laws, or whose missing rate was higher than 5%, a total of 3857 SNPs were nally used for genetic map construction. Based on separating patterns, eligible SNPs were assigned into 1615 bins. Through linkage analysis, a high-density linkage map spanning 4564.94 cM was constructed using the 1605 bin markers (Supplemental Table 3, Dataset 1 spreadsheet 1). To obtain accurate mapping results, two strategies were used to investigate the phenotypes of two spikes from 101 F 2 individuals. The rst strategy only contained three values, wild type, tri-pistil type and heterozygous, investigated by counting and classifying the orets; the second strategy involved calculation of TRP penetrance for each F 2 plant. As expected, a locus on chromosome 2D-3 between IWB29964 and IWB23687 was detected for all 4 TRP traits (Fig. 4a, Supplemental Table 4), which explains the 60.5-86.6% phenotypic variations with LOD values from 26.6 to 44.1 (Supplemental Table 4). Given that TRP was a single dominant gene, additionally, the 5A locus was only detected in one of the replicate penetrance traits (Supplemental Table 4). Thus, the 5A locus may have been a pseudo-locus, since it Mapping results were consistent with previous studies (Mahlandt et al. 2021;Yu et al. 2020).
Collinearity between the genetic map in the mapping interval and physical map (IWGSC Refseq V2.1) was not perfect (Fig. 4a, b). To verify the primary mapping result and ne-map the TRP locus, we extended the mapping interval from IWB29964-IWB23687 to IWA5206-IWB23687 and developed KASP markers for SNP IWA5206, IWB29964, IWB32175, and IWB23687 (Fig. 4a, b). The development of KASP markers for IWB32175 and IWB23687 was successful, but not for IWB29964 and IWA5206. To obtain more SNP resources and gene expressions, young spikes at Waddington5.5 (W5.5) and Waddington7.5 (W7.5) stages of 4045 and ZLWM were sampled, after which RNA-seq was performed. Using SNPs identi ed by RNA-seq, an SNP at 10055 bp downstream of IWA5206, located on the same gene as IWA5206, was successfully developed into a KASP marker, named IWA5206-10. Therefore, IWB23687, IWB32175, and IWA5206-10 markers were used to screen for recombinants among the 725 F 2 progenies of 4045 × ZLWM. A total of 42 recombinants were identi ed, by which the TRP locus was mapped at the interval between IWA5206-10 and IWB32175 (Fig. 4b). Subsequently, IWA5206-10 and IWB32175 were used to screen 3494 F 2:3 plants derived from heterozygous F 2 individuals, and 19 new recombinants between IWA5206-10 and IWB32175 were identi ed. To enrich marker densities of mapping intervals, we developed 9 additional KASP markers (Fig. 4c, Dataset 1 spreadsheet 2). A total of 23 recombinants between IWA5206-10 and IWB32175, including 4 from F 2 and 19 from F 2:3 , were genotyped with the 9 newly developed markers after which their phenotypes were investigated. Based on genotypes and phenotypes, the 23 recombinants were divided into 5 haplotypes (A-E) and mapping interval was nally narrowed to a 97.3 kb region, anked by IWA5206-10 and TRP_SNP167 markers (Fig. 4c, Dataset 1  spreadsheet 2). To con rm the ne-mapping result, two recombinants from haplotype C and D, by which the TRP locus was mapped to the IWA5206-10-TRP_SNP167 region, were selected to generate F 3:4 sub-populations for conformality. The F 3:4 individuals were genotyped with two anking markers. As expected, segregation was detected for IWA5206 (heterozygous in F 2:3 generation) in both F 3:4 subpopulations, however, it was not detected for TRP_SNP167 (homozygous in F 2:3 generation) (Supplemental Table 5), con rming the authenticity and accuracy of the ne-mapping result.

Analysis of candidate genes in the mapping interval
Only two genes, TraesCS2D03G1038200 harboring IWA5206-10 and TraesCS2D03G1038400 harboring TRP_SNP167 (Supplemental Fig. 5b, c), were annotated in the genome reference of Chinese Spring (CS), IWGSC Refseq V2.1 (Zhu et al. 2021), within the mapping interval (Fig. 5). Based on transposable elements (TEs) annotation of the CS genome, TEs accounted for about 75% of the mapping region (Fig.  5a). To obtain more information about this mapping region, we analyzed the collinearity between the 11 available wheat genomes along with genomes of Aegilops tauschii (D genome) and Thinopyrum elongatum (E genome) (Fig. 5b). The mapping region in the 11 wheat genomes was divided into two haplotypes (~97 kb type and ~324 kb type), however, no extra genes were annotated (Fig. 5b). In the A. tauschii genome (Aet V4.0), at ~103.5 kb downstream of the TraesCS2D03G1038200 homolog, a singleexon gene (AET2Gv21024800) encoding an unknown protein was annotated (Fig. 5, Supplemental Fig.  4a). However, its structure is atypical with the absence of start and stop codons (Supplemental Fig. 4a).
Since AET2Gv21024800 has been annotated as a high-con dence class gene ), we aligned its coding sequence (CDS) and protein sequences to CS databases. Seven proteins exhibited a high identity (>80%) with AET2Gv21024800 (Supplemental Fig. 4b), consistent with CDS alignment to the CS high-con dence CDS database (Dataset 1 spreadsheet 3). In contrast, alignment to CS low-con dence CDS database detected 381 hits (Dataset 1 spreadsheet 4). Since AET2Gv21024800 showed high similarity to only a part of 4 out of 4 wheat proteins, these ndings imply that AET2Gv21024800 might be a non-functional pseudogene. To elucidate on the 4045 transcriptome, we mixed the RNA of young spikes at W5.5 and W7.5 of 4045, after which we performed full-length RNA-seq (Pacbio). Therefore, the CDS of AET2Gv21024800 was aligned to full-length transcript library, however, no hit was detected. Consistently, expressions of the 7 wheat homologs was not detected in RNA-seq (Supplemental Fig. 4c). These ndings imply that AET2Gv21024800 is not the TRP candidate.
Thereafter, we evaluated two anking genes of the mapping region. Transcriptional analysis by RNA-seq and structure analyses were performed. TraesCS2D03G1038200, which consists of 28 exons (Supplemental Fig. 5b), encodes an ATP binding cassette C member 2 (ABCC2) transporter family protein.
Transcriptional differences of TraesCS2D03G1038200 and its homologs in A and B sub-genomes between 4045 and ZLWM were not signi cant (Supplemental Fig. 5a). Six SNPs were detected in the encoding region, however, none of them resulted in amino acid alterations (Supplemental Fig. 5b). ABCC2 is involved in defense molecule accumulation, heavy metal transport, tolerance to biotic and abiotic stresses, as well as transport of other metabolites in plant species such as Arabidopsis and saffron. However, it has not been conclusively determined whether ABCC2 or members of the ABC transporter family contribute to ower architecture development (Behrens et  The expression of TraesCS2D03G1038400 was markedly elevated in the two samples of 4045, especially at the W7.5 stage (Supplemental Fig. 5a). Compared to the CS sequence, an SNP, C138T, was detected at codon 46 (Supplemental Fig. 5c) in 4045, however, it was not associated with any amino acid alteration.
Moreover, a single nucleotide insertion was detected at 535-bp upstream of the start codon, which might have led to increased transcriptional abundance. TraesCS2D03G1038400 encodes a remorin family protein. Moreover, it has a 75.6% identity with its rice homolog, OsGSD1 (Os04g0620200) (Supplemental Fig. 5d). Enhanced expression of OsGSD1 was associated with reduced grain setting, carbohydrate accumulation in leaves, and lower soluble sugar contents in phloem exudates of rice gsd1-D activation mutant. Any abnormalities in the stamen and pistil of gsd1-D were not detected (Gui et al. 2014). As a plant-speci c protein family, the signi cance of remorin family proteins in plant oral establishment has not been established (Gouguet et al. 2021).
In addition to structure and expression analysis of the two anking genes, a recent study reported that there were no signi cant differences at protein levels for both TraesCS2D03G1038200 and TraesCS2D03G1038400 between tri-pistil and normal wheat (Li et al. 2020c).
These results imply that TraesCS2D03G1038200 and TraesCS2D03G1038400 are not TRP candidates. Tri-pistil wheat, including 4045, may carry a new haplotype that is different from normal wheat (Fig. 5) with a gain-of-function mutation in the mapping region, leading to generation of the tri-pistil phenotype.

Transcriptional analysis of oral homeotic genes in 4045
Floral homeotic genes play vital roles in oral establishment. Altered expressions of these genes are associated with abnormalities in oral organs. To investigate the relationship between TRP and oral homeotic genes, their expressions were analyzed by RNA-seq of young spikes at W5.5 and W7.5 stages.
At the W5.5 stage, pistil differentiation in normal wheat is considered complete (Fig. 2i, j) (Waddington et al. 1983). However, in 4045, PRPs nished the differentiation while the late-developed SPs were still at stage W4.5-W5.5 (Fig. 2k, l). Therefore, differences in expressions of pistil/carpel formation-related genes between 4045 and ZLWM were supposed to be detected at W5.5. At the W7.5 stage, both PRPs and SPs completed pistil differentiation. For some genes, different expression patterns were detected in different sub-genomes (Fig. 6). Since orthologs in wheat sub-genomes usually exhibit high similarities and share redundant functions, we analyzed the total expressions of these ortholog genes without normalization (Supplemental Fig. 6). Levels of TaAG-A2 and TaAG-B2 in 4045 at W5.5 were signi cantly elevated, however, the increase in TaAG-D2 levels was not signi cant. These differences were not detected at W7.5 (Fig. 6). TaAG-2s are orthologs of rice OsMADS3, which is transiently expressed during the initiation stages of stamen, carpel and ovule primordia (Yamaguchi et al. 2006). Therefore, upregulation of TaAG-2s indicated that they exerted their functions at W5.5 and contributed to SPs formation when PRPs had nished differentiating. Moreover, overexpressions of OsMADS3 did not result in alterations of pistils (Kyozuka and Shimamoto 2002), suggesting that TaAG-2s may be involved in TRP-regulation, rather than directly determine SPs formation. However, at both stages, signi cant differences in TaDLs, an important pistil determinator, were not detected (Fig. 6). Although down-regulated expressions for TaDL-A and TaDL-B were detected in 4045, they were largely supplemented by TaDL-D (Supplemental Fig. 6). B class TaAP3-1s and D class TaSTK-1s, rice OsMADS16 and OsMADS13 orthologs, respectively, were downregulated in 4045 (Fig. 6). During ovule development, OsMADS13 acts as a repressor of the carpel pathway (Osnato et al. 2021). On this account, the decrease in TaSTK-1s may moderate the suppression of carpel development and facilitate SPs formation. A possible reason for downregulation of TaAP3-1s is that it might be directly/indirectly regulated by TRP. Furthermore, several TaSEP1s were differentially expressed between 4045 and ZLWM (Fig. 6), however, their total expressions were complemented by other homologs (Supplemental Fig. 6). TaMADS32-B1 was downregulated ~5-6 fold in 4045 at both stages (Fig. 6), whereas total expressions were comparable (Supplemental Fig. 6). Its encoding sequence was analyzed and no alterations were found, indicating that the functions of TaMADS32s were not signi cantly affected in 4045.

Discussion
Floral organ development in uences grain yields of cereal crops, including wheat (Williams and Sorrells 2014). Male oral organs of wheat produce large amounts of pollen whereas only one pistil harbors one ovule in each oret per spike. On this account, abnormalities in pistils have signi cant effects on wheat yields. Such abnormalities or variations provide non-substitutable resources for understanding oral architectures and increase the yields. A series of anomalous-developed-pistil mutants have been identi ed, which have helped elucidate on molecular mechanisms involved in pistil establishment. Based on previous studies, we divided anomalous-developed-pistil mutants into three types: pistillody type, inde nite-oral-organ-number-increase (IFONI) type and accurately-regulated-pistil-number-increase (ARPNI) type (Fig. 7). The classical ower development ABCDE model (Fig. 7a)  In contrast, tri-pistil wheat normally produced all oral organs, except for the generation of a xed number (three) of pistils (Fig. 2, Supplemental Fig. 1). Apart from delayed emergence and development, SPs were normal, fertile, and capable of producing grains, thereby increasing grain numbers (Fig. 1A, Fig. 2, Supplemental Fig. 1 and Supplemental Table 1). Based on the characteristics of tri-pistil wheat, we de ned this pistil mutant type as precisely-regulated-pistil-numberincrease (PRPNI) type. PRPNI mutants provide a novel resource for exploration of oral development and open opportunities for manipulation strategies to increase crop yields.
Studies have aimed at identifying the causal gene(s) of tri-pistil wheat. It has been postulated that the causal gene is located on 2DL (Mahlandt et al. 2021;Peng et al. 2008;Wang et al. 2009;Yang et al. 2017;Yu et al. 2020;Zhu et al. 2019). Using multiple molecular marker systems, Yu et al. (2020) mapped the Pis1-2020 locus to the 7.32 Mb region of 2DL (Supplemental Fig. 7). Subsequently, Mahlandt et al. (2021) physically located Mov-1 to a 1.1 Mb region by combining radiation hybrid (RH) and recombination-based mapping approaches (Supplemental Fig. 7). However, the locus has not yet been ne-mapped. In this study, using a total of 4320 progenies of 4045 × ZLWM, we narrowed the mapping region down to a 97.3 kb interval (Fig. 4, Supplemental Fig. 7). To obtain a robust result, genotypes of all recombinants were double examined and the nal mapping result validated using F 3:4 sub-populations (Supplemental Table 3). Pursuant to comparisons of recent mapping results, all tri-pistil loci were placed on an approximately 30 Mb region of chromosome 2DL (Supplemental Fig. 7 on available wheat genome assemblies, only two anking genes within the TRP mapping region were annotated (Fig. 5). Along with an extra gene annotated in A. tauschii genome, three genes were considered potential candidate genes. Nevertheless, through gene structure and expression analysis, previous studies provided su cient evidence for rejecting the hypothesis that the three genes are casual genes of TRP (Supplemental Fig. 4, Supplemental Fig. 5). These analyses suggested that the tri-pistil phenotype may be caused by a gain-of-function gene that does not exist in reference wheat genomes. Furthermore, we discussed that the PRPNI type mutant is novel germplasm (Fig. 7 (Fig. 7d). Brie y, TaDLs and C Class genes specify pistil formation. E class genes in combination with TaAGL6s positively regulate C class genes by forming heterodimers with them. A class genes and C class genes, B class genes and TaDLs, antagonistically regulate the expressions of each other. TaSTK-1s, D class genes that specify the identity of ovules, are also involved in pistil development, even though their expressions are restricted to ovules (Callens et al. 2018;Shen et al. 2021). The TRP regulates the pistil development in an unknown way and leads to the formation of PRPNI pistils. The PRPNI provides a novel direction to investigate wheat oral organ development. Investigation of the regulatory mechanisms of TRP will elucidate on the processes and pathways involved in pistil development.
GNPS is a key component of wheat grain yield. To improve GNPS, special germplasms with high GNPS were investigated. For instance, introgression or translocation of Agropyron cristatum 6P chromosome into common wheat signi cantly increases GNPS to ~120 (Wu et al. 2006;Zhang et al. 2016). This germplasm has been successfully utilized in breeding of Pubing series wheat cultivars. The null mutation of WFZP-A along with deletion of WFZP-D led to development of a triple spikelet (TRS) phenotype and increased the GNPS (Du et al. 2021). Similarly, the tri-pistil wheat, 4045, also produces high GNPS (Fig. 1a, Supplemental Table 1). In addition, our unpublished data showed that kernel weight of 4 F 1 populations increased more than 2 folds, compared to the normal parental wheat line, which was also found by Mahlandt et al. (2021). These observations imply the enormous potential of TRP in enhancing yields and development of wheat hybrid-breeding. Therefore, investigation of tri-pistil wheat is of great interest for basic research on oral developments and applications on crop yield increases. We systematically investigated the morphologic features of TRP and ne-mapped the TRP locus, thereby opening opportunities to understand and utilize the excellent germplasm. Figure 1 Spike and grain morphologies of the tri-pistil wheat, 4045, and blue-aleurone wheat ZLWM.   Segregation of blue aleurone trait in single orets of 4045 × ZLWM F 1 plants.

Figures
Page 23/25 (a) An example of spikes of F 1 plants. (b-e) Different color combinations were observed in tri-grains. Trigrain b comprised white, dark blue and light blue (from left to right. Note that the white grain looked bluish because the seed was very thin and the black background went through it) grains; c, d and e comprised one white, two dark blue; one light blue, two white; and one white, two light blue grains, respectively. Panels show the grains that were dissected and enlarged from the three seeds of each oret in the circles. Bar for the spike = 1 cm.

Figure 4
Mapping and ne mapping of the TRP locus.