The wheat AGL6-like MADS-box gene is a master regulator for floral organ identity and a target for spikelet meristem development manipulation.

The AGAMOUS-LIKE6 (AGL6)-like genes are ancient MADS-box genes and are functionally studied in a few model plants. The knowledge of these genes in wheat remains limited. Here, by studying a 'double homoeolog mutant' of the AGL6 gene in tetraploid wheat, we showed that AGL6 was required for the development of all four whorls of floral organs with dosage-dependent effect on floret fertility. Yeast two-hybrid analyses detected interactions of AGL6 with all classes of MADS-box proteins in the ABCDE model for floral organ development. AGL6 was found to interact with several additional proteins, including the G protein β and γ (DEP1) subunits. Analysis of the DEP1-B mutant showed a significant reduction in spikelet number per spike in tetraploid wheat, while overexpression of AGL6 in common wheat increased the spikelet number per spike and hence the grain number per spike. RNA-seq analysis identified the regulation of several meristem activity genes by AGL6, such as FUL2 and TaMADS55. Our work therefore extensively updated the wheat ABCDE model and proposed an alternative approach to improve wheat grain yield by manipulating the AGL6 gene.

Introduction

Common wheat (Triticum aestivum, AABBDD, 2n = 6x = 42) is a worldwide major staple food crop that provides ~20% calories and 25% protein in the global human diet (http://www.fao.org/faostat/en). Common wheat is a hexaploid plant that arose from cultivated tetraploid wheat (Tricitum turgidum, AABB, 2n = 4x = 28) and goatgrass (Aegilops tauschii, 2n = 4x = 14) at a time coinciding with wheat domestication. Wild wheat was domesticated at every ploidy level leading to cultivated diploid wheat T. monococcum (AA), cultivated tetraploid wheat T. turgidum ssp. dicoccum (AABB), and cultivated hexaploid wheat (AABBDD), i.e. common wheat including bread wheat. Except for a few traits, such as the brittleness of spikes and the grain hardness, tetraploid wheat is often used as a model to study hexaploid wheat due to its relatively simpler genome composition and high sequence similarity (>94% identity) between the two species (Dubcovsky and Dvorak, 2007). The two systems are often used interchangeably during genetic analysis and even for trait improvement (Feng et al., 2017; Fu et al., 2009; Konopatskaia et al., 2016; Krasileva et al., 2017; Qin et al., 2016; Shitsukawa et al., 2009; Uauy et al., 2006).

Wheat bears inflorescences in the form of a spike, with a main axis of two ranks of lateral spikelets directly attaching to the rachis, or so‐called distichous spikes. Such a structure is different to that of rice, which forms secondary branches within its inflorescence. The basic unit of wheat spikes is spikelet which is composed by several florets that are subtended by a pair of glumes (Sreenivasulu and Schnurbusch, 2012). The spike architecture has a large effect on floret development and therefore the final grain number per spike, and hence yield (Gao et al., 2019). Similar to other grass species, a wheat floret is comprised of four whorls of floral organs including lemma and palea (the first whorl), two lodicules (the second whorl), three stamens (the third whorl), and carpel/ovule (the fourth whorl pistil) (Callens et al., 2018; Yoshida and Nagato, 2011). The ABCDE model classifies genes responsible for floral organ identity into A, B, C, D and E classes which are mostly MIKC‐type MADS‐box transcription factors except for the A‐function gene APETALA2 (AP2) (Callens et al., 2018; Chongloi et al., 2019; Schilling et al., 2020). The model was first established in Arabidopsis and has since been largely confirmed in grasses such as rice and maize (Hu et al., 2015; Wu et al., 2017, 2018). In wheat, however, the ABCDE model is not systematically characterized because of lack of floral organ mutants. Until recently, with the availability of the wheat genome sequence (Appels et al., 2018; Jia et al., 2018), wheat floral genes like the AP2 family Q and AP2L2, the AP1/FUL‐like family VRN1, FUL2 and FUL3, and AGL6 (Debernardi et al., 2020; Li et al., 2019; Su et al., 2019), as well as the wheat homolog of ABBERANT PANICLE ORGANIZATION1 (APO1) which affects spikelet number per spike (Kuzay et al., 2019; Muqaddasi et al., 2019; Voss‐Fels et al., 2019), and Grain Number Increase 1 (GNI1), a regulator of floret fertility (Golan et al., 2019; Sakuma et al., 2019), have been identified.

The AGAMOUS‐like6 (AGL6) genes represent an ancient subfamily of MADS‐box genes, present in both angiosperm and gymnosperm plants (Dreni and Zhang, 2016). In maize, two duplicated AGL6 genes ZAG3 and ZAG5 are identified, but only ZAG3 is essential for floral development with different functions in tassel and ear (Thompson et al., 2009). Two AGL6 homologs, OsMADS6 and OsMADS17, were found in rice that arose during ancient whole‐genome duplication events. Between them, OsMADS6 plays a key role in determining the floral organ identity while the role of OsMADS17 is minor and somewhat redundant (Ohmori et al., 2009). Studies based on mutants showed that OsMADS6 works in palea, lodicule and pistil development, as well as in floral meristem indeterminacy (Li et al., 2010; Ohmori et al., 2009; Reinheimer and Kellogg, 2009; Yu et al., 2020). OsMADS6 is also essential for grain size and grain quality (Yu et al., 2020; Zhang et al., 2010).

In wheat, only a few floral genes have been functionally evaluated. More will be studied with the available of genome sequences and mutant resources such as EMS mutants made with the tetraploid cultivar Kronos and the hexaploid cultivar Cadenza (Krasileva et al., 2017). Among wheat MADS‐box genes, three AP1/FUL‐like genes VRN1, FUL2 and FUL3 have been found to play central roles in specifying spikelet meristem (SM) (Li et al., 2019), while homoeologs of other MADS‐box genes such as wheat LEAFY HULL STERILE1 (WLHS1) were functionally differentiated in flowering time determination due to genetic and epigenetic modifications (Shitsukawa et al., 2007). In addition, the expression of TaSEP5 (also name TaPAP2) appeared to be associated with spikelet number in wheat (Gauley et al., 2019; Wang et al., 2017b). Recently, the wheat AGL6 gene TaAGL6 is functionally identified in RNAi plants to be involved in the development of the third and fourth whorls floral organs by regulating the putative B‐class MADS‐box gene TaAPETALA3 (TaAP3) expression, although anther is not one of the major expression domains for TaAGL6 in wheat (Su et al., 2019). The understanding of the wheat AGL6 gene functions remains limited.

Here, we studied the wheat AGL6 gene loss‐of‐function mutants that exhibited dramatic morphological changes in nearly all floral organs, demonstrating a wider role than previously known of this gene in wheat floral organ development. Comprehensive protein‐protein interactions revealed additional protein complex and refined the ABCDE model for wheat floral development. Further interaction of the wheat AGL6 with the G protein γ subunit (DEP1) and subsequent genetic analyses uncovered a role of DEP1 for wheat spikelet meristem development. Overexpression of TaAGL6 increased the final grain number of transgenic plants. Taken together, we propose that the wheat AGL6 gene is a master regulator for floral organ identity and its involvement in spikelet meristem development renders it as a potential target to improve wheat grain yield.

Results

Characterization of wheat AGL6‐like genes

The three homoeologs TaAGL6‐A (TraesCS6A02G259000), TaAGL6‐B (TraesCS6B02G286400) and TaAGL6‐D (TraesCS6D02G240200) of the AGL6 gene in common wheat cv. Chinese Spring (CS) were 98.85% identical in protein sequences (Figure S1a). Orthologs between hexaploid and tetraploid wheat were also highly similar with the tetraploid homoeologs TtAGL6‐A (TRIDC6AG039910) and TtAGL6‐B (TRIDC6BG046780) 100% and 98.79% identical to their hexaploid counterparts TaAGL6‐A and TaAGL6‐B, respectively (Figure S1b). The nearly identical sequences of these homoeologs and orthologs suggest that they play similar functions in tetraploid and hexaploid wheats. Despite this, in CS, there were a few amino acid (aa) differences at the C‐terminus of the three homoeologous proteins. That is, relative to TaAGL6‐D, there were two and eight aa differences TaAGL6‐A and TaAGL6‐B, respectively, a few of them on the critical AGL6‐II motif that have been shown to mediate transcription activation (Ohmori et al., 2009). These sequence divergences may be responsible for different transcriptional activation activities of the three homoeologous proteins (Figure S2).

To study the evolution trajectory of wheat AGL6 genes, we reconstructed a phylogenetic tree using protein sequences from wheat, rice and Arabidopsis and found that the AGL6 subfamily genes were most closely related to the E‐class MADS‐box genes (Figure S3 and Table S1), consistent with the previous report (Becker and Theißen, 2003). Further study of the grass AGL6 genes revealed a trend of copy number increase in grasses with different modes. In rice, the two AGL6 genes OsMADS6 and OsMADS17 appear to be derived from ancient genome duplication that have diverged into two lineages with significant differences in both sequences and functions (Figure S4 and Table S2) (Ohmori et al., 2009). In maize, the two AGL6 genes ZAG3 and ZAG5 were duplicates of the OsMADS6 and no homologs of OsMADS17 was found (Figure S4). Only ZAG3 was reported to function in floret, which displayed defects in both tassel and ear when mutated (Thompson et al., 2009). In common wheat, the three homoeologs were most similar to OsMADS6 and all three were maintained and functional (Figure S4).

Despite the different paths for AGL6 gene evolution, gene expression patterns in the related species seem to be conserved. We found that the expression of TaAGL6 firstly appeared at the Waddington stage 3 (W3) and maintained high till the heading date stage by qRT‐PCR and in situ hybridization (Waddington et al., 1983) (Figure 1a). No or faint signal could be observed at the stage of spikelet meristem development (Figure 1c, d). Expression signals of TaAGL6 became apparent when floral meristems appeared (Figure 1e). In floral organs, TaAGL6 mainly expressed in palea, lodicule and pistil, but low in lemma and invisible in stamen (Figure 1b). Consistently, in situ hybridization detected strong signals at these organs, but not in lemma and stamen (Figure 1f, g, h). These findings to a large extent are consistent with a previous study (Su et al., 2019). Also consistent with earlier reports (Feng et al., 2017; Su et al., 2019), the three homoeologs of TaAGL6 showed similar expression patterns throughout the spike development periods as well as in different floral organs (Figure S1c,d). Conserved expression patterns were also found in tetraploid wheat (Figure S1e). The widespread expression of the wheat AGL6 gene suggests its extensive roles in spike and floral development.

Figure 1

Expression patterns of TaAGL6 in common wheat cv. Chinese Spring. (a) Relative expression levels of TaAGL6 in spikes at different developmental stages by qRT‐PCR. W, Waddington stages as described by Waddington et al. (1983); BS, booting stage; HD, heading date. (b) Expression levels of TaAGL6 in various floral organs at heading date. (c–h) TaAGL6 expression at stage W2.5 (c), W3 (d), W3.5 (e) and W5.5 (f and g). The sense probe was used a negative control (h). sm, spikelet meristem; fm, floret meristem; gl, glume; le, lemma; pa, palea; lo, lodicule; st, stamen; pi, pistil. Bars = 200 μm.

Wheat floret fertility was sensitive to the functional allele number of the AGL6 gene

To study the functions of the wheat AGL6 gene, we obtained two single homoeolog mutant lines from the tetraploid wheat variety ‘Kronos’ mutant population (Krasileva et al., 2017). The two ‘stop‐gain’ mutants, ttagl6‐A (corresponding to the A homoeolog) and ttagl6‐B (corresponding to the B homoeolog), bore mutations that generated premature stop codons that potentially disrupted the functions of the two homoeologous proteins (Figure S5a,b). Significant reduction of TtAGL6 gene expression was observed in both ttagl6‐A and ttagl6‐B plants (Figure S5c), probably due to nonsense‐mediated mRNA decay (Kalyna et al., 2012). No vegetative or spike anomaly were found in these single homoeolog mutants (Figure S6a–d). Despite this, the single homoeolog mutants did show reduced grain number per spike when compared with wild type plants (Figure 2a,b). Further observation revealed significantly decreased three‐grain spikelets and significantly increased two‐grain spikelets (Figure S6e,f). No defects were found in developing ovules, neither in pollen activity (Figure S6g,h).

Figure 2

Dosage effects of the tetraploid wheat AGL6 gene on grain number per spike. (a) The grain number per spike was positively correlated with functional AGL6 allele numbers. Capital letters represent wild type alleles while lower‐case letters represent mutant alleles. Bar = 1 cm. (b) Statistic analysis of grain number per spike between WT and mutant plants. ttagl6 is a double‐homoeolog mutant. Error bars indicate SD (n = 10). (c) Total expression levels of the TtAGL6 gene in WT and various AGL6 mutants. Structures of mutant organs under microscope (d–k). (d, e) Stamens in WT (d) and double‐homoeolog mutant ttagl6. (e). Asterisks indicate stamens. (f, g) WT (f) and ttagl6 (g) spikelets with lemmas and paleas removed showing the arrangement of stamens in florets. Asterisks indicate stamens. (h, i) Pistils in WT (h) and ttagl6 (i). (j, k) Longitudinal sections of the WT pistil (j) and a defective pistil (k). ca, carpel; ov, ovule; ss, spike‐like structure. Bars = 1 mm in (d–i) and 500 μm in (j, k). (l–s) Scanning electron microscopy (SEM) observation of pistil developmental between WT (l–o) and ttagl6 (p–s) at stage W5.5 (l, p), W7.5 (m, q), W8.5 (n, r), W9.5 (o, s); Arrowheads indicate spikelet‐like structures. Bars = 50 μm in (l), 200 μm in (m, n, p, q and r), and 500 μm in (o, s). (t, u) Expression pattern of floral organ development‐related genes in WT and ttagl6 pistils. qRT‐PCR analyses were performed in three biological replicates. Error bars indicate SD Student’s t‐test, *P < 0.05, **P < 0.01.

We then created homozygous AGL6 double‐homoeolog‐mutant lines (DHM, aabb). We firstly backcrossed the mutant lines with the wild type Kronos once using homozygous single‐homoeolog‐mutant (SHM, aaBB and AAbb). We then crossed the BC1 progeny (Aa and Bb) and studied the floret fertility in a DHM line of the BC1F2 progeny by selfing BC1F1 plants (AaBb). We found that DHM plants were completely sterile, while plants with heterozygous alleles (aaBb or Aabb) were partially fertile with significantly reduced grain number per spike (P < 0.01) (Figure 2a,b). Overall, the phenotypic severity, especially the grain number per spike, was correlated with the copy number of mutant alleles (Figure 2c), indicating gene dosage sensitivity of the AGL6 functions in wheat. A more detailed observation of DHM lines revealed that most floral organs were deformed, with a significant number of stamens carrying two anthers on one filament (or so‐called double‐anther stamen) (Figure 2d–g, and Figure S7h,i). Meanwhile, all pistils lost their identities, with additional abnormal spike‐ or spikelet‐like structures being surrounded by carpelloid structures (Figure 2h–s, Figures S6g and S7j‐n). To gain insights into possible molecular mechanisms, we detected the expression levels of orthologs of other rice MADS‐box genes in the ABCDE model in DHM lines. We found that TtAG2 and TtAG1, respective orthologs of rice OsMADS3 and OsMADS58 of C‐functions, were both significantly down‐regulated (Figure 2t, Figures S3 and S8). So were the two orthologs to the rice D‐class genes OsMADS13 and OsMADS21, two direct targets of the rice OsMADS6 gene FACTOR OF DNA METHYLATION LIKE 1 (OsFDML1) (Figure 2t). On the other hand, the expressions of the A‐class genes VRN1 and FUL2 were dramatically increased (Figure 2u and Figure S8). Together, these data showed that ALG6 malfunction had caused mis‐expression of multiple key genes for floral organ development.

Moreover, DHM lines showed additional severe morphological changes in the outer whorls of wheat florets (except for lemma) that have not been reported for this gene before (Figure 3). In these lines, paleas were converted into separated or integrated lemma‐like structures with developed awns (Figure 3a, b and Figure S7a–c, o–r). Extra organs were often observed near the first whorl organs, such as defective floret (Figure 3c), glume‐like (Figure 3d and Figure S7d) and awn‐like (Figure S7e) structures. Meanwhile, nearly all lodicules were lost (Figure 3i). A few remaining were extensively elongated (Figure 3e, f), while others became palea‐like (Figure S7f, g). Ectopic lodicule‐anther mosaic organs appeared near the positions of the stamens between the second and third whorls (Figure 3g). Scanning electron microscopy (SEM) did not detect any difference in DHM lines at the stage W3.5 (Waddington stage 3.5) when compared with those of the wild type (Figure 3k,o). But later at stages W4 to W5, DHMs developed more stamen/lodicule‐anther primordia, and formed indeterminate meristems at the position of carpels, some of which eventually developed into indeterminate spikelets or even spikes (Figure 2k and 3p–r and Figure S6g). These data indicate that TtAGL6 plays an important role in the determinacy of wheat florets (Figure 3j). To confirm the homoeotic changes of paleas to lemma‐like organs, we checked the expression patterns of the homolog of the rice lemma maker gene DROOPING LEAF (DL) in the wild type and DHM lines (Yamaguchi et al., 2004). Indeed, TtDL was highly expressed in palea‐converted lemma‐like organs (Figure 3s). In lodicule‐anther mosaic tissues, the rice B‐class gene homologs TtPI2, TtPI1 and TtAP3 were significantly down‐regulated and the rice C‐class gene homologs TtAG2 and TtAG1 were significantly up‐regulated (Figure 3t and Figures S3, S8). These data showed that the wheat AGL6 gene was required for proper expression of many floral genes and played as a master regulator for floral organ development in wheat.

Figure 3

Double‐homoeolog mutation of the tetraploid wheat AGL6 gene caused diverse changes in floral organs and primordia development. (a, b) Scanning electron microscopy (SEM) observation of paleas in wild type (WT) (a) and the double‐homoeolog mutant ttagl6 (b). Bars = 500 μm. (c, d) SEM observation of ectopic floral (c) and glume‐like organ (d) in ttagl6. Bars = 200 μm in (c), 500 μm in (d). (e, f) SEM observation of normal lodicule in WT (e) and elongated lodicule in ttagl6 (f). Bars = 500 μm. (g) SEM observation of ectopic lodicule‐anther mosaic organ in ttagl6. Bar = 200 μm. (h, i) SEM observation of the florets with lemmas and paleas removed in WT (h) and ttagl6 (i). Bar = 500 μm. (j) Diagrams of floral structure for WT and agl6 wheat plants. (k–r) SEM observation of WT (k–n) and ttagl6 (o–r) of florets at stages W3.5 (k, o), W4 (l, p) W4.5 (m, q) and W5 (n, r). Asterisks indicate positions of stamen or lodicule‐anther primordia. Arrows indicate the fourth whorl primordia. Bars = 50 μm. (s) Expression levels of TtDL in the WT and ttagl6 paleas. (t) Expression of B‐class genes TtPI2, TtPI1, TtAP3 and C‐class genes TtAG2, TtAG1 in WT lodicules and ectopic lodicule‐anther mosaic organs in ttagl6. qRT‐PCR analyses were performed in three biological replicates. Error bars indicate s.d. Student’s t‐test, **P < 0.01, *P < 0.05. gl, glume; le, lemma; pa, palea; pi, pistil; st, stamen; lo, lodicule; da‐st, double‐anther stamen; l‐a, lodicule‐anther mosaic organ; le‐l, lemma‐like organ. I, ectopic glume‐like organ or abnormal floral.

TaAGL6 interacted extensively with proteins involved in floral organ development

Phylogenetic analysis showed that the AGL6 subfamily genes were closely related to E‐class MADS‐box genes. They were shown to act as the ‘bridge proteins’ that enable the formation of the multimeric complexes of ABCDE proteins in ‘quartet model’ of floral organ development (Rijpkema et al., 2009; Theißen and Saedler, 2001). To address the role of AGL6 in wheat ABCDE model, we performed a pair‐wise protein‐protein interaction assay among wheat MADS‐box proteins using the yeast two‐hybrid (Y2H) system. We found that TaAGL6 formed homodimers in yeast (Figure S9a), which was further confirmed by BiFC (Bimolecular Fluorescent Complimentary) and LCI (Firefly luciferase complementation imaging assay) assays (Figure S9b, c). TaAGL6 also formed heterodimers with A‐class proteins FUL2 and AP2L5 (Q), B‐class proteins TaPI1, TaPI2 and TaAP3, C‐class protein TaAG2, D‐class protein TaSTK1, and E‐class proteins WLHS1, TaSEP6, TaSEP4, TaSEP3 and TaSEP5 (Figure 4a, Figure S10a and Table S1).

Figure 4

Protein interaction network of the wheat TaAGL6. (a) Yeast two‐hybrid assays of TaAGL6 and B‐class proteins (TaPI2, TaPI1 and TaAP3) with representative A‐, B‐, C‐, D‐ and E‐class MADS‐box proteins. Yeast cells with different construct combinations were grown on selective media without Trp and Leu (SD/‐TL) and were tested for interactions on selective media without Trp, Leu, His and Ade (SD/‐TLHA). Pictures were taken after 3 days of incubation at 30 °C. Empty was used as negative control. (b) Yeast three‐hybrid assays showing higher order protein complex by TaAGL6, TaAG2 and TaSTK1. TaAGL6 gene driven by the Met‐repressible promoter (pMET25–TaAGL6) was expressed on selective media without Trp, Leu, His, Ade and Met (SD/‐TLHAM) but was not expressed on the SD/‐TLHAM media supplemented with 1 mm Met. (c) Interaction of E‐class proteins with other MADS‐box proteins not present in (a). (d) A summary of protein interactions and regulatory roles of TaAGL6 in wheat floral development. Lines with diamond arrows indicate protein‐protein interaction. Lines with arrows at the end indicate positive regulation and with a perpendicular line indicate negative regulation. Lines in golden colour were reported by Su et al. (2019). Purple lines were the results of the current work. Phenotypes regulated by AGL6 include palea (pa), lodicule (lo), stamen (st), carpel (ca)/ovule (ov) and floral meristem (FM) determinacy indicated by solid lines. The dotted line indicates no regulation on lemma. (e) An updated ABCDE model for wheat floral development. Specification of floral organs by AGL6‐participated protein complexes. AGL6 forms complexes with A‐ and E‐class proteins to specify the first whorl organs and with A‐, B‐ and E‐class proteins to specify the second whorl organs. It works with B‐, C‐ and E‐class proteins to specify stamen identity, with C‐ and E‐class proteins to specify carpel development, and with C‐, D‐ and E‐class proteins for ovule identity.

To gain a more extensive view of the wheat MADS‐box protein interactome, we further expanded our pair‐wise interaction assay among homologs of the rice A‐, B‐, C‐, D‐ and E‐class MADS‐box proteins. As shown in Figure 4, B‐class MADS‐box proteins TaPI1, TaPI2 and TaAP3 interacted with each other; they also interacted with A‐, C‐ and E‐class proteins, but not with D‐class proteins. No interactions were observed between C‐class and D‐class proteins (Figure 4b); but E‐class proteins formed both homodimers and heterodimers with A‐, B‐, C‐ and D‐class proteins (Figure 4a,c and Figure S10c). Moreover, the domestication protein Q, which recently was found to interact with AP2‐like proteins for floret development (Debernardi et al., 2020), formed homodimers as well as heterodimers with VRN1, FUL2, FUL3, TaPI1, TaPI2, TaAP3, TaAG1, TaAG2 and TaSTK1, respectively (Figure S10b). Our detailed protein interaction analyses thus extensively updated the wheat ABCDE model (Figure 4d,e).

Interaction of TaAGL6 with the G protein β and γ subunits

In rice, MADS‐box proteins, such as OsMADS1, interacted with G‐protein β and γ subunits that were involved in panicle development (Huang et al., 2009; Liu et al., 2018; Utsunomiya et al., 2011). We found that common wheat homologs of the rice G‐protein β and γ subunit genes TaRGB1 and TaDEP1 were expressed at the early stage of spike development, partly overlapping with the expression domain of TaAGL6 (Figure 5a, Figure S11a). Y2H assay showed that TaAGL6 and TaRGB1 strongly interacted with TaDEP1 respectively (Figure 5b, Figure S11b), while LCI assays showed strong luminescence signals between nLUC‐TaAGL6/cLUC‐TaDEP1, nLUC‐TaAGL6/cLUC‐TaRGB1 and nLUC‐TaRGB1/cLUC‐TaDEP1 co‐expression samples (Figure 5c, Figure S11c).

Figure 5

TaAGL6 interacted with TaDEP1. (a) Expression analysis of TaDEP1 in spikes at different stages as described in Figure 1. (b) Interaction of TaAGL6 and TaDEP1 in Y2H assay. Yeast cells growing on selective media without Leu, Trp, His and Ade (‐LTHA) indicate positive interactions. Empty was used as negative control. (c) Firefly luciferase complementation imaging (LCI) assay confirmation of TaAGL6 and TaDEP1 interaction in N. benthamiana leaves. nLUC or nLUC‐TaAGL6 was co‐transformed into N. benthamiana leaves with cLUC or cLUC‐TaDEP1. After infiltration, luciferase activity was detected in injected leaves sprayed with 1 mm luciferin. Empty vector (EV) was used as negative control. (d) Spikes of wild‐type (left) and single‐homoeolog TtDEP1 mutant ttdep1‐B. Bar = 1 cm. (e, f) TaDEP1 mutation (ttdep1‐B) on spikelet number per spike (e) and grain number per spike (f). Error bars indicate SD (n = 20). Student’s t‐test, *P < 0.05, **P < 0.01.

To investigate domain(s) responsible for TaAGL6 and TaDEP1 interactions, LCI assays were performed using protein truncation constructs. As shown in Figure S12, we found that the I and K domains were required for physical interaction between TaAGL6 and TaDEP1, similar to the interaction between MADS‐box proteins (Kaufmann et al., 2005; Liu et al., 2018). We then studied the functional domains of TaDEP1, i.e., the GGL motif (G protein γ‐like) and the vWFC1 (von Willebrand factor type C1) and vWFC2 domains (Figure S13a). We showed strong reconstituted activity in nLUC‐TaAGL6/cLUC‐TaDEP1vWFC1 and nLUC‐TaAGL6/cLUC‐TaDEP1vWFC2 co‐expression samples, whereas activity of luciferase was hardly detected in nLUC‐TaAGL6/cLUC‐TaDEP1GGL co‐expression sample (Figure S13b), indicating that the vWFC domains of TaDEP1 were essential for the TaDEP1‐TaAGL6 interaction.

The involvement of TaAGL6 and TaDEP1 in spikelet meristem development

We generated three TaAGL6 RNAi lines (RNAi‐6, ‐7, ‐10) in common wheat cv. Fielder and found that RNAi‐10 line showed a significant reduction in spikelet number per spike and grain number per spike while the two numbers in the remaining two were not significant, but did have a trend of reduction (Figure S14). Total AGL6 transcripts were reduced by 49.2% and 39.7% in RNAi‐10 and RNAi‐6, respectively. On the other hand, TaAGL6‐D overexpression in Fielder produced elongated spikes with dense spikelets (Figure 6c and Figure S15). Accordingly, spikelet number per spike and grain number per spike were significantly increased (Figure 6d–f). It must be noted, however, that the flowering time of the three selected T3 lines, OE‐1, OE‐22 and OE‐6, were delayed 21, 14 and 6 days, respectively (Figure 6a,b). This was consistent with scanning electron microscopy (SEM) observation that the terminal spikelet, the mark of the completion of spikelet meristem development, appeared at W4.0 (the stage when stamen primordia began to development) in OE plants, 0.5 Waddington phase unit (W4.0 vs. W3.5) later than those of the wild type Fielder whose terminal spikelet appeared at W3.5 (Figure 7a–h and Figure S16a). Indeed, in these OE plants, the wheat florigen VRN3 was significantly down‐regulated in leaves (Figure S15b), explaining the late‐flowering phenotypes in these plants.

Figure 6

Phenotypic analyses of common wheat transgenic lines overexpressing TaAGL6. (a) Transgenic plants overexpressing TaAGL6 (OE) flowered later than the wild type Fielder. Bars = 10 cm. (b) Significantly delayed flowering time of TaAGL6 OE plants. (c) Spike morphology of Fielder and TaAGL6 OE plants showing additional spikelets in the latter. Bar = 1 cm. (d) Statistical analysis of spikelet number per spike of Fielder and TaAGL6 OE plants. (e) More grains were harvested from TaAGL6 OE plants relative to the wild type. Bar = 1 cm. (f) Statistical analysis of grain number per spike of Fielder and TaAGL6 OE plants. Error bars indicate SD (n = 20). Student’s t‐test, *P < 0.05, **P < 0.01.

Figure 7

Overexpression of TaAGL6 slowing down spikelet meristem phase transition. (a–h) Scanning electron microscopy (SEM) of young spikes from Fielder (a–d) and TaAGL6 OE‐1 line (e–h) showing the later appearance of the terminal floret in OE plants. fm, floret meristem; gp, glume primordium; lp, lemma primordium; stp, stamen primordium; ts, terminal spikelet. The number at the bottom represents the spikelet number per spike. Bars = 500 μm. (i) A heatmap of DEGs in TaAGL6 OE plants showing the up‐regulation of inflorescence meristem identity genes and down‐regulation of floret meristem identity genes. FW‐1, ‐2, ‐3, and OE‐1, ‐2, ‐3 represent inflorescence tissues collected at the W2.5, W3, W3.5 stages respectively in Fielder and TaAGL6 OE‐1 lines for RNA‐seq. (j) Spikelet meristem activity genes found in DEGs in both wheat AGL6 and rice DEP1 mutants.

We then searched and obtained the Kronos mutant library for DEP1 mutants. One EMS line (Kronos #4262) was found to carry a premature stop codon on the B sub‐genome homoeolog (hence ttdep1‐B) (Figure S17a–c). In addition to the reduced plant height, significant reduction in spikelet number per spike and hence the smaller grain number per spike in homozygous SHM ttdep1‐B plants were observed, although spike length was unaffected (Figure 5d–f and Figure S17d,e). We studied the downstream genes in the rice gain‐of‐function mutant dep1 as reported by Liu et al. (2018) and found that spikelet meristem genes MADS55, SEP5 and EATB were regulated by both AGL6 and DEP1 in wheat (Figure 7j, Tables S4, S5 and Figure S19). These data suggest that wheat AGL6 and DEP1 all work in spikelet meristems that may contribute to inflorescence development and affect final wheat yield.

Regulation of meristem activity genes by TaAGL6 for both inflorescence and floral development

We performed RNA‐seq analyses of young spikes of TaAGL6‐D OE plants and Fielder (WT) at W2.5, W3 and W3.5 stages. ‘Maintenance of meristem identity’ functions genes were found to be enriched among Cluster 6 genes by Gene Ontology (GO) analysis of differentially expressed genes (DEGs) (Figures S16b and S18). Specifically, genes for inflorescence meristem identity such as TaTAW1, TaMADS22 and TaMADS55 were up‐regulated and genes for floral meristem identity such as FUL2 (an ortholog of the rice A‐class gene OsMADS15) and FUL3 (an ortholog of the rice A‐class gene OsMADS18) were down‐regulated (Figure 7i and Figure S19). We then detected the expression levels of these genes in RNAi lines and found opposite expression patterns, i.e. TaMADS55 was down‐regulated while FUL2 was up‐regulated (Figure S19), supporting that TaAGL6 may closely regulate these meristem genes, a mechanism for AGL6 genes in determining wheat yield‐related traits such as spikelet number per spike. Whether AGL6 directly regulates these genes, or work together with DEP1 in wheat, needs further investigation.

Discussion

AGL6‐associated protein interactions enriched the wheat ABCDE model

In rice, the AGL6 subfamily gene OsMADS6 has an E function working in all four whorls of floral organs where it acts as a cofactor for A, B, C and D class proteins to form higher order complexes (Li et al., 2010, 2011; Ohmori et al., 2009). A second AGL6‐like gene OsMADS17 showed redundant and hence minor functions in floral organ development relative to OsMADS6 (Ohmori et al., 2009). In wheat, TaAGL6 has been shown to play an important role in determining stamen development (Hama et al., 2004; Murai et al., 2002; Su et al., 2019). We show here that, like in rice, the wheat AGL6‐like gene plays a much broader role by involving in determining the identity of all four whorls. Two A‐class proteins FUL2 and FUL3 can form protein complex with TaAGL6 and E‐class proteins that may be required for the first whorl of floral organ development, and TaAGL6, as well as FUL2, Q and E‐class proteins may form heterodimer with B‐class proteins TaPI1, TaPI2 and TaAP3. The B‐class proteins per se do not form homodimers. Although we did not find interaction between C‐ and D‐class proteins, the fact that they form complex via TaAGL6 as a bridge, as has been partly reported in other species, is as expected (Seok et al., 2010).

Expanded functions of the AGL6 gene in wheat

Although the wheat AGL6 genes and the rice OsMADS6 have similar expression patterns and their mutants have overlapping floral organ phenotypes, they did show differences in the two species. In tetraploid wheat, all paleas of AGL6 DHM plants were converted to two lemma‐like organs. Indeterminacy of stamens was observed with extra anthers on the same filamentous, or so‐called double‐anther stamens, which was not reported in rice. Moreover, many pistils were transformed into spikelet‐ or spike‐like structures within carpelloid structures. These observations indicate that wheat AGL6 may play an even wider role in floral organ development in wheat than in rice. Additional interaction protein with TaAGL6 such as the domestication gene Q and novel interaction with TaAG2 and TaSTK1 support additional functions of the AGL6 gene in wheat. We speculate that the different copy number of the AGL6 gene and their different evolutionary paths may contribute to such differences in different species.

The dosage sensitivity for wheat AGL6 gene functions in floret fertility

The genus Triticum has several related species with the genome of various levels of ploidy and high similarity, such as diploid T. urartu (AA), T. turgidum (AABB) and common wheat (AABBDD). Sequence analysis in multiple plant species showed that plants tend to increase the copy number of AGL6‐like genes in their genomes, either by retaining copies from ancient genome duplication such as in rice or by relative newer genome duplication such as in maize or polyploidization such as in wheat (Dreni and Zhang, 2016). In light of the high sequence similarity among wheat homoeologs and those of related species, the increase in gene copy number may potentially increase gene expression levels (Schiessl et al., 2019). Wheat AGL6 genes showed gene dosage effect in their functions for floret fertility. Single‐homoeolog mutation in tetraploid wheat did not display clear abnormality in floral organs, but significant reduction in grain numbers. We speculated that the lack of pollen vigour pollens may underlie such an observation. Although the double homoeolog mutations caused severe phenotypic changes and complete sterility, knockdown of AGL6 by RNAi in common wheat caused a trend of reduction in spikelet meristem numbers, which was significantly reversed when AGL6 was overexpressed. These data strongly suggest that wheat AGL6 gene is involved in spikelet meristem development and may contribute to increasing grain number per spike in wheat.

It is interesting that our OE plants showed delayed flowering time while those from a previous report flowered earlier (Su et al., 2019). A few factors may contribute to such differences: 35S promoter vs. Ubi, homoeolog B instead of homoeolog D, bombardment transformation method vs. Agrobacteria approach, and different hosts Kenong 199 vs Fielder used. Further study of AGL6 functions, at both protein level or at the expression level, should provide more insight into the precise functions of this key transcription factor in wheat.

TaAGL6 may be applicable in grain yield improvement in wheat

A subtle reduction of grain number per spike and a significant increasing in spikelet number and also grain number per spike encouraged us to further investigate the role of AGL6 in spikelet meristem in wheat, since the effect of OsMADS6 on inflorescence architecture was not reported (Duan et al., 2012; Li et al., 2010; Ohmori et al., 2009). Despite this, several genes have been recently reported to contribute to inflorescence development in rice, including the G‐protein γ subunit DEP1 that works to enhance meristematic activity and branch production (Huang et al., 2009), the G‐protein β subunit gene RGB1 that causes dwarfism and a reduced size of panicle (Utsunomiya et al., 2011), and TAW1 that manipulates inflorescence development by enhancing branch meristem (BM) activity, and delaying spikelet meristem (SM) specification through its downstream genes OsMADS22 and OsMADS55 (Yoshida et al., 2013).

Our work provides a substantially updated ABCDE model in wheat to establish a better understanding of the molecular mechanism of wheat floral development. The lack of functionality of AGL6 leads to low floret fertility, even sterility, and appropriate manipulation of AGL6 expression contributes to increased spikelet number and grain number per spike. We speculate that cooperated functions of AGL6 and the G protein DEP1 participate in spikelet meristem development via regulation of downstream meristem genes, such as FUL2 and TaMADS55, which calls for further investigation. The duo‐regulation on spikelet meristem and floral meristem by AGL6 requires precise coordination among the genes involved. Further study of the function and molecular mechanism of the wheat AGL6 gene and its related complex will provide clues to modify spikelet and spike architecture and hence to improve wheat yield.

Methods

Plant materials and growth conditions

Plant materials used in this study included the bread wheat (Triticum aestivum L.) cv. Chinese Spring (CS) and cv. Fielder, the tetraploid wheat variety Kronos (Triticum turgidum ssp. durum), and Nicotiana benthamiana. CS was used for gene sequence amplification and expression studies. Fielder was used for transgenic plant production. Kronos and its EMS mutants were used for phenotype analysis. Wild type heat plants were grown in the experimental field (39°57′N, 116°19′E) in Beijing. N. benthamiana plants were used for Agrobacterium‐mediated transient expression and protoplasts transient expression analyses and were grown in a growth chamber under the long‐day condition (16 h light/8 h dark) at 24 °C.

Phenotypic characterization of tetraploid wheat mutants

Kronos mutants (M4 lines) from a sequenced EMS (Ethyl Methane Sulphonate) mutant population of the tetraploid spring wheat variety Kronos were kindly provided by Professors Jiajie Wu and Daolin Fu of Shandong Agriculture University, Shandong, China (Krasileva et al., 2017). A line Kronos#2244 (henceforth ttagl6‐A) had an early stop codon in the A subgenome homoeolog (TtAGL6‐A) causing a point mutation at the 267th nucleotide of the ORF (Open Reading Frame). The second EMS mutant #309 (henceforth ttagl6‐B) also had a premature stop codon caused by a point mutation at the 418th nucleotide of the ORF of the B homoeolog (TtAGL6‐B). For DEP1, the Kronos#4262 mutant bore a premature stop codon at the 275th nucleotide of its B‐homoeolog ORF and was named ttdep1‐B. The fragments containing the mutation sites were amplified from Kronos and mutants by PCR amplification and the products were directly sequenced for verification. Relevant primer sequences were shown in Table S3. To obtain ttagl6 double‐homoeolog mutant lines (DHM, aabb), the A and B mutants were backcrossed once to Kronos to reduce background mutations. Heterozygous hybrids (AaBB and AABb) were then crossed to generate BC1F1 plants (AaBb) which were used to generate complete sterile DHM lines for phenotypic observation.

Constructs and wheat transformation

To generate TaAGL6 overexpression transgenic lines, the pUbi:TaAGL6 construct was developed from the full‐length ORF of TaAGL6 with a 6× myc tag at the N terminal and cloned into the reconstructed binary vector pCAMBIA3300 containing a maize ubiquitin promoter. The TaAGL6‐RNAi construct was developed using a 340 bp region from the C‐terminus and 3′‐UTR of TaAGL6‐D that was 95.0% and 91.6% identical to TaAGL6‐A and TaAGL6‐B, respectively, and was expected to target all three homoeologs. The trigger fragment was cloned into the intermediate vector pMWB006 and then the destination vector pMWB111. The pUbi:TaAGL6 and TaAGL6‐RNAi constructs were used to transform wheat cv. Fielder by the Agrobacterium‐mediated transformation method (Wang et al., 2017a). DNA of the transformed plants for positivity verification was extracted using a Plant Genomic DNA Kit (TIANGEN, Beijing, China). Primers referred to in this study were shown in Table S3.

Quantitative real‐time PCR

Total RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, California, USA) and cDNA was synthesized with the PrimeScript RT reagent Kit (Takara, Tokyo, Japan) according to the manufacturer’s instructions. qRT‐PCR assays were performed with SYBR Premix EX Taq (Takara) on an ABI 7300 real‐time PCR system in a total volume of 20 μL. Wheat GAPDH was used as an endogenous control to normalize expression levels of different samples. method was used for calculating genes expression (Livak and Schmittgen, 2001). All experiments had three independent biological replicates, each consisting of three technical repetitions. Primers used were listed in Table S3, and relevant genes were listed in Table S6.

In situ hybridization

Digoxigenin‐labelled antisense and sense probes were prepared with in vitro transcription from PCR products of TaAGL6 using homoeolog‐specific primers with T7 promoter sequence (TAATACGACTCACTATAGGG) (Table S3). Probe for TaAGL6 was 304‐bp. The promoter sequence with the forward primer transcribed a sense probe, and the promoter sequence with the reverse primer transcribed an antisense probe. Fresh wheat material embedding, in situ hybridization and immunological detection of hybridization signals were performed as previously described (Liu et al., 2013, 2014). Images were obtained using a ZEISS AXIO IMAGER Z2. Primers used were shown in Table S3.

Scanning Electron Microscopy

Young spikes from transgenic and wild‐type wheat were first fixed in 2.5% glutaraldehyde overnight at 4 °C. They were then dehydrated through a series of ethanol solutions, dried in a carbon dioxide critical‐point dryer, and coated with platinum. The samples were observed and photographed using a scanning electron microscope (Hitachi SEM SU8010, Tokyo, Japan) at an accelerating voltage of 10 kV.

Phylogenetic analysis

Sequences used in this study were downloaded from Ensembl Plants (http://plants.ensembl.org/index.html). Protein sequences were aligned using ClustalX within MEGA6 and a phylogenetic tree was constructed using the neighbour‐joining method based on 1000 bootstrap replicates (Tamura et al., 2013). Accession numbers for genes used for phylogenetic analysis were listed in Table S1.

Yeast two‐hybrid and yeast three‐hybrid assays

Full‐length ORFs of proteins shown in Table S3 were amplified from CS. The vectors for yeast two‐hybrid (Y2H) assays were generated using pGBKT7 (bait) and pGADT7 (prey). All the target genes were cloned into the pGBKT7 and pGADT7 vectors, respectively. The Y2H assay was carried out according to the manufacturer’s instructions of Matchmaker GAL4 Two‐Hybrid System (Clontech, USA). Briefly, specific bait and prey plasmids were co‐transformed into yeast strain AH109, and spread on selective medium SD/‐Trp/‐Leu (SD/‐TL). Transformed colonies were then dropped on selective medium SD/‐Trp/‐Leu/‐His/‐Ade (SD/‐TLHA) for protein‐protein interaction verification. Primers used here were listed in Table S3.

For yeast three‐hybrid (Y3H) assay, the pBridge vector (Clontech, Palo Alto, California, USA) was used according to the manufacturer’s instructions. TaSTK1 was subcloned into the pBridge vector MCS I (multiple clone site) resulting in pBridge‐TaSTK1 and the TaAGL6 was subcloned into the pBridge‐TaSTK1 vector MCS II under the control of the Met‐repressible pMET25 promoter, resulting in pBridge‐TaSTK1‐proMET25‐TaAGL6 vector. The pBridge‐TaSTK1‐proMET25‐TaAGL6 and AD‐TaAG2 vectors were co‐transformed into AH109 cells. The colonies were streaked on the selection media without Trp, Leu, His, Ade and Met (SD/‐TLHAM) supplemented with or without Met.

Firefly luciferase complementation imaging (LCI) assay

LCI assay was performed to verify protein–protein interactions in N. benthamiana (Chen et al., 2008; Sun et al., 2013). Briefly, LCI assay vectors were developed using full‐length ORFs of target genes by cloning a series of deleted fragments of TaAGL6 or TaDEP1 into the KpnI/SalI‐digested and KpnI/BamHI‐digested p1300‐35S‐cLUC vectors by homologous recombination respectively. The constructed plasmids were introduced into A. tumefaciens strain GV3101 that was mixed with the agrobacterium containing the p19 silencing vector. This mixed agrobacterium solution was used to co‐infiltrate N. benthamiana leaves. After 36–48 h of infiltration, injected leaves were sprayed with 1 mm luciferin (Promega, Madison, Wisconsin, USA). LUC signal was detected using NightSHADE LB 985 (Berthold, Bad Wildbad, Germany). Each assay had three independent repetitions. All the primers referred here were shown in Table S3.

Conflict of interest

The authors declare no conflict of interest.

Author contributions

L.A., L.X. and M.L. designed the project; K.X. and W.F. performed most of the experiments; G.S., T.S., J.M. and S.G. aided in experiments; G.J. and W.Z. helped in RNA‐seq data analysis. Y.X. and W.K. helped in generating the transgenic materials. M.J., L.D., W.Y., Z.Y. and F.X. provided valuable advices about the project. K.X. and L.A. drafted the manuscript. L.X. and M.L. revised the manuscript.

Figure S1 Wheat AGL6 protein sequence comparison and expression patterns.

Figure S2 Transcriptional activation function of TaAGL6 proteins.

Figure S3 Phylogenetic analyses of wheat, rice and Arabidopsis MADS‐box proteins.

Figure S4 Phylogenetic analyses of the AGL6 subfamily proteins from grass species.

Figure S5 Characterization of two single‐homoeolog EMS mutants of the AGL6 gene from tetraploid wheat cv. Kronos.

Figure S6 Additional phenotypic analyses of reproductive organs of TtAGL6 single‐ and double‐homoeolog mutants.

Figure S7 Additional phenotypic changes in the double‐homoeolog mutant ttagl6.

Figure S8 Expression patterns of floral identity genes in glume and floral organs.

Figure S9 Homodimerization of TaAGL6 proteins.

Figure S10 Additional pairwise interactions of wheat MADS‐box proteins, Q and E‐class proteins.

Figure S11 Interaction of TaRGB1 with TaAGL6 and TaDEP1.

Figure S12 Luciferase complementation imaging (LCI) assays showing the necessity of TaAGL6 I and K domains for AGL6‐TaDEP1 interaction.

Figure S13 Luciferase complementation imaging (LCI) assays showing the necessity of TaDEP1 vWFC domains for TaDEP1‐TaAGL6 interaction.

Figure S14 Characterization of TaAGL6 RNAi lines.

Figure S15 Molecular characterization of TaAGL6‐D OE transgenic lines.

Figure S16 Effect of TaAGL6 overexpression on wheat plant development.

Figure S17 Characterization of the ttdep1‐B mutant.

Figure S18 K‐means cluster analyses of genes differentially expressed (DEGs) at least at one developmental stage between the wild type and the OE line.

Figure S19 Comparative expression analyses of indicated genes in TaAGL6 overexpression (OE) and RNAi transgenic plants.

Click here for additional data file.

Table S1 The correspondence of Arabidopsis, rice and wheat MADS‐box gene orthologs.

Table S2 The GenBank accession numbers of the AGL6 subfamily genes.

Table S3 List of primers used in this study.

Table S4 List of DEGs in the enriched GO term ‘regulation of transcription’ in the dep1 mutant.

Table S5 List of DEGs in the enriched GO term ‘regulation of transcription’ in TaAGL6‐OE lines.

Table S6 List of genes and their locus IDs studied in this work.

Click here for additional data file.