Retrotransposon-mediated DELLA transcriptional reprograming underlies semi-dominant dwarfism in foxtail millet

Meicheng Zho, Hui Zhi, Xue Zhng, Gunqing Ji,Xinmin Dio,*

aInstitute of Crop Sciences,Chinese Academy of Agricultural Sciences,Beijing 100081,China

bCenter for Agricultural Research Resources, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Shijiazhuang 050021,Hebei,China

Keywords:Retrotransposon Transcriptional reprogramming DELLA Dwarf breeding Foxtail millet(Setaria italica)

A B S T R A C T Retrotransposons account for a large proportion of the genome and genomic variation, and play key roles in creating novel genes and diversifying the genome in many eukaryotic species.Although retrotransposons are abundant in plants, their roles had been underestimated because of a lack of research.Here,we characterized a gibberellin Acid(GA)-insensitive dwarf mutant, 84133, in foxtail millet. Map-based cloning revealed a 5.5-kb Copia-like retrotransposon insertion in DWARF1 (D1), which encodes a DELLA protein. Transcriptional analysis showed that the Copia retrotransposon mediated the transcriptional reprogramming of D1 leading to a novel N-terminal-deleted truncated DELLA transcript that was putatively driven by Copia's LTR,namely D1-TT,and another chimeric transcript.The presence of D1-TT was confirmed by protein immunodetection analysis. Furthermore, D1-TT protein was resistant to GA3 treatment compared with the intact DELLA protein due to its inability to interact with the GA receptor, SiGID1. Overexpression of D1-TT in foxtail millet resulted in dwarf plants,confirming that it determines the dwarfism of 84133.Thus,our study documents a rare instance of long terminal repeat (LTR) retrotransposon-mediated transcriptional reprograming in the plant kingdom. These results shed light on the function of LTR retrotransposons in generating new gene functions and genetic diversity.

1. Introduction

Retrotransposons belonging to the class I group of transposable elements that transpose via a RNA intermediatemediated‘copy-and-paste'mechanism to amplify themselves[1].Retrotransposons are abundant in plants especially wheat and maize, resulting in their huge genomes [2,3]. The long terminal repeat (LTR) retrotransposons that are predominant in plants are considered as an important driver of genetic diversity and phenotype variation in plants. They disturb genes in a variety of ways, including silencing gene expression,altering transcript structure and changing tissue-specific gene expression depending on their insertion position[1].For instance, insertion of an LTR retrotransposon into the promoter of the grape color gene Vvmby1A silenced gene expression and resulted in colorless fruit [4]. Dwarfism in wheat was attributed to a newly identified LTR retrotransposon insertion at the N-terminal of Rht-B1c,although this retrotransposon was removed as an intron during transcription, a 90-bp fragment remained as a part of an exon, resulting in completely dominant dwarf [5]. A particularly striking example of an LTR retrotransposonmediated change in tissue-specific gene expression comes from maize, in which a transposon element inserted into the first exon of b1, which is involved in regulating anthocyanin accumulation, changed its expression site from vegetative tissues to seeds,resulting in dark corn kernels.Subsequently,a large LTR retrotransposon insertion in this transposon element resulted in decreased and variegated expression of b1, resulting in variegated seeds [6]. Besides affecting the transcriptional structure or tissue-specific expression of genes, retrotransposons can cause transcriptional reprogramming, resulting in novel genes and trait diversity.The color of red fruit flesh in blood oranges was shown to be regulated by a Copia retrotransposon-induced chimeric gene,in which the partial 3′ LTR of Copia is fused to the original coding sequence of a MYB transcription factor gene to form Ruby. Recruitment of new regulatory sequences from the 3′LTR of the Copia was shown increase Ruby expression in fruit flesh,conferring a distinctive red color[7].Bs1,chimeric maize gene, was also created by retrotransposon-mediated exon shuffling, and may influence reproductive development [8].These rare instances illustrate the importance of retrotransposons in genetic diversity and trait variation.Knowledge about their functions is still limited, especially the ways in which they mutate genes,because of the rarity of case studies.

Gibberellic acid (GA) is an important growth-promoting phytohormone that regulates many aspects of plant growth and development including seed germination, heading date,fruit expansion and plant height [9,10], Mutations in GA biosynthesis or signaling pathway lead to dwarfism in diverse plants including rice,wheat,maize and Brassica napus[10-13].The genetic basis for GA signaling has been well established through the identification and characterization of key components such as GA receptor Gibberellin Insensitive Dwarf1(GID1), SLEEPY1 (SLY1)/Gibberellin Insensitive Dwarf2 (GID2)and DELLA [10,14]. Among these regulators, DELLA proteins are master repressors of GA signaling and largely determine plant height. It has been reported that DELLA proteins are encoded by GRAS family genes [i.e., GIBBERELLININSENSITIVE (GAI), REPRESSOR OF GA1 (RGA), and SCARECROW (SCR)], and contain N-terminal DELLA and VHYNP motifs. Deletion or amino acid substitution in the DELLA N-terminal results in GA-insensitive dominant dwarfism in monocot and dicot species [15-17]. For instance, a 17-amino acid deletion from the DELLA domain of GAINSENTITIVE (GAI) in Arabidopsis results in a GA-insensitive dwarf [16]. Constitutive expression of truncated SLR1 with a deleted of N-terminal generates a severe GA-insensitive dwarf phenotype in rice [15,18]. Notably, Reduced Height (RHT)-B1b and-D1b, which contain mutations in N-terminal domains,reduce height in the wheat cultivar Norin 10. These genes were widely used to breed lodging-resistant wheat cultivars with significantly increased grain yield worldwide during the“green revolution” [11]. All these findings demonstrate that the N-terminal of DELLA is critical for its repression of GA action. Further molecular mechanism studies have shown that the DELLA N-terminal is required to interact with the GA receptor GID1 and its consequent degradation [19,20]. In contrast to the dominant-negative effect of DELLA Nterminal mutations on GA and plant height, the loss-offunction of DELLA leads to slender plants with over-elongated stem internodes,for example,slr1 in rice and slender1(sln1)in barley (Hordeum vulgare) [18,21]. These phenotypes resemble the response of GA treatment. All DELLA mutants, including dwarf types and slender types,are unresponsive to exogenous GA treatment, indicating that DELLA proteins are critical regulators in the blocking of GA signal transduction.

Foxtail millet is one of the staple cereals grown in China,India, and other Asian countries [22]. Because of its relative high-density planting and soft stalk, it is prone to lodging especially at the grain-filling stage, resulting in serious reductions in yield and quality [23,24]. Dwarf breeding is an efficient way to prevent lodging and has been achieved for in many cereals such as rice (Oryza sativa) and wheat (Triticum aestivum) during the ‘green revolution' [11,25], but not for foxtail millet. Although there are many dwarf resources in foxtail millet [22,26], the causal genes and the underlying genetic basis of dwarfism are unknown, which hinders the breeding of elite dwarf cultivars. Recently, we isolated and characterized a recessive dwarf gene SiDWARF2 (D2) from a foxtail millet mutant. This gene encodes a cytochrome P450 enzyme, and this mutant plants show moderately reduced plant height without a penalty in the 1000-grain weight [27].Thus, this gene is a candidate for use in a dwarf breeding program. Compared with recessive dwarf genes, dominant dwarf genes have some special advantages in crops hybrid breeding programs and in the use of heterosis [28]. As far as we know, no dominant dwarf genes have been cloned from foxtail millet so far.

Foxtail millet has a relatively small genome (～510 Mb)comprising approximately ～9% protein-encoding genes and＞25% LTR retrotransponsons [29]. However, little is known about the biological functions of those LTR retrotransposons in the genome,especially in creating novel genes and genetic diversity. 84133, a spontaneous mutant of Zhaogu 1 cultivar,was previously shown to harbor a single dominant dwarf gene and to be insensitive to exogenous GA3treatment [22]. Here,we cloned the underlying gene for dwarfism from 84133 and showed that retrotransposon-mediated DELLA transcriptional reprogramming underlies semi-dominant dwarfism in foxtail millet. Our findings indicate that LTR retrotransposonmediated transcript reprogramming might be a key driver for novel gene initiation and genetic diversity. Our results have also identified a candidate gene for a dwarf breeding in foxtail millet, especially for hybrid breeding and the use of heterosis.

2. Materials and methods

2.1. Plant materials and field trials

The plant materials included the foxtail millet dwarf line 84133, which is a spontaneous mutant of Zhaogu 1 cultivar[22], Yugu 1 (i.e., cultivar with the foxtail millet reference genome)[29]and a Residual Heterozygous Line(RHL)population selected from an F6recombinant inbred line population derived from a cross between Zhangai 10 and 84133. In the past 20 years, we have collected many dwarf mutant lines including Zhangai 10 and 84133 from different breeding programs in China to identify genes for the breeding of dwarf foxtail millet[22].Zhangai 10 × 84133 cross was initially used for half-diallel cross assays to investigate the genetic basis and allelism of dwarf genes from different dwarf lines.Then,this cross was developed into a recombinant inbred line due to the high diversity in plant height among the progeny.One line among the progeny,No.136,exactly resembled 84133 and exhibited a clear 1:2:1 segregation ratio in plant height.Subsequently, the No.136 line was developed into the RHL to isolate the gene responsible for dwarfism from 84133. Plants were grown in a field in Beijing in summer or in a greenhouse under long-day conditions (16-h light/8-h dark 28 °C/24 °C) at a light intensity of 100-150 mmol m-2s-1.The plant height of each population(RHL and the F2population derived the Yugu 1 × 84133 cross) was measured for five individuals after heading.

2.2. Map-based cloning of D1

The RHL was an immortalized segregation population that showed a clear segregation ratio in plant height.Therefore,we used this RHL for D1 mapping. First, we developed additional new simple sequence repeat(SSR)markers.Briefly,we downloaded the corresponding sequences from Yugu1 reference genome and identified repetitive microsatellite motifs using SSR HUNTER software.Then specific primers flanking the SSR were designed using Primer 3.0 software. These new SSR markers combined with previously reported ones [30] were used to screen 40 individuals (20 dwarf plants and 20 normal plants) in the RHL population to identify the markers closely linked to the D1 gene.Then,240 RHL individuals were used to perform preliminarily mapping.Finally,seven SSR and single nucleotide polymorphism (SNP) markers were developed in the target region and used to screen recombination events in 1500 individuals to fine-map D1. Gene predictions for the mapped 72-kb region were completed using JGI annotation(http://phytozome.jgi.doe.gov/pz/portal.html#!info?alias=Org_Sitalica).

2.3. Southern blot analyses

The procedures for DNA isolation and Southern blotting were as described by Chao et al. [31]. We digested 20 μg genomic DNA for each blot.The probe used to verify the Copia insertion was designed based on the C-terminal sequence of the D1-TT transcript. The probe used to identify correctly transformed transgenic plants was designed according to the hygromycin gene coding region. Probe labeling and hybridization were conducted using the DIG High Prime DNA Labeling and Detection Starter Kit II (Roche, Basel, Switzerland), according to the manufacturer's instructions. The primers used for Southern blotting are summarized in Supplementary Table S1.

2.4. Candidate gene sequencing and rapid amplification of cDNA ends (RACE) assay

For 84133 or homozygous dwarf (hoD) plants in the RHL,genomic sequences of D1 were obtained by PCR amplification using the primers Della-A1 and Della-S. These primers were also used to amplify the full-length D1 transcript from wildtype. For the RACE assays, total RNA was extracted from 2-week-old seedlings using the TRIzol plus RNA Purification kit(Life Technologies,Carlsberg,CA,USA).The RACE assays were completed using the 3′/5′-Full RACE kit (Takara, Dalian,China), according to the manufacturer's instructions. The primers used for RACE assays are listed in Supplementary Table S1.

2.5. Plasmid construction

To construct Ubi: D1-TT, the open reading frame (ORF) of D1-TT was amplified from 84133 cDNA using the primers C3-F/R.The amplified fragment was inserted into a modified pCAMBIA1390 binary vector downstream of the maize ubiquitin promoter using the In-Fusion kit (Clontech, Palo Alto, CA,USA). For the yeast two-hybrid (Y2H) assay, the bait plasmid was constructed by inserting the ORF of SiGID1 into the pGBKT7 vector between EcoRI and BamHI.To construct the prey plasmid, the full coding regions of D1-WT and D1-TT were each inserted into the pGADT7 vector at the EcoRI site. The primers used to create these constructs are listed in Table S1.

2.6. Yeast two-hybrid assay

For the Y2H assay, Gold yeast (Clontech) was used for the plate growth assay with detection medium (i.e., -Ade/-His/-Trp/-Leu or -Ade/-His/-Trp/-Leu supplemented with 50 mg mL-1X-α-Gal). In some cases, the medium contained 10-4mol L-1GA3or 3 mmol L-13-AT. The Y2H assays were conducted according to the manufacturer's protocols.

2.7. Recombinant protein and antibody production

To produce the anti-D1-WT antibody the full-length D1-WT in the pGADT7 vector was transferred into the pET28a vector at the EcoR? site.Escherichia coli Transetta(DE3)(Transgen Biotech,Beijing, China) was used to produce the D1-WT-His fused protein. A 1-mL aliquot of precultured cells was used to inoculate 250 mL Luria Bertani medium in a 1-L flask. The culture was grown at 37 °C until the optical density (600 nm)reached 0.4-0.6.Recombinant protein production was induced by adding 0.2 mmol L-1isopropyl-β-D-thiogalactopyranoside.The culture was incubated at 20 °C for 16 h, and then cells were harvested and resuspended in Buffer A (50 mmol L-1Tris-HCl, pH 8.0, 100 mmol L-1NaCl, 10 mmol L-1imidazole,and 0.1% Triton X-100). Cells were lysed by sonication (40 times at 15-20 W for 5 s). The lysate was centrifuged at 11,000 ×g for 30 min, and the supernatants were mixed with 400 μL TALON Metal Affinity Resin (Clontech). Samples were rotated for 2 h at 4 °C. The resin was washed five times with Buffer A,and recombinant protein was eluted five times with 400 μL Buffer A supplemented with 500 mmol L-1imidazole.The eluate was collected and desalted on a PD-10 column(GE Healthcare, Piscataway, NJ, USA) equilibrated with Buffer B(20 mmol L-1Tris-HCl, pH 7.5, 0.15 mol L-1NaCl, and 2 mmol L-12-mercaptoethanol), and then eluted using 1×phosphate-buffered saline(PBS)before analysis by 8%sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDSPAGE). The purified protein was used to produce anti-D1-WT polyclonal antibodies in rabbits (Sangon Biotech, Shanghai,China).

2.8. Immunoblot analysis of DELLA proteins

Crude protein extracts were prepared by grinding young seedlings of hoN and hoD plants from the RHL in liquid nitrogen. Some plant materials were sprayed with 1 × 10-4 mol L-1GA3prior to total protein extraction. An equal volume of 2× SDS buffer (125 mmol L-1Tris-HCl, pH 6.8, 4%SDS,10%glycerol,and 0.01%bromophenol blue)was added to protein extracts, which were then boiled for 5 min. Protein samples were separated by 8%SDS-PAGE and transferred to a polyvinylidene fluoride membrane (GE Healthcare) by semidry blotting.The membranes were incubated in 5%skim milk in PBS Tween-20 (PBST) buffer for 1 h at room temperature,and then in anti-D1-WT antiserum (1:1000 dilution in PBST)for 1 h. Membranes were washed three times with PBST for 15 min each and then incubated in a horseradish peroxidaseconjugated goat anti-rabbit IgG secondary antibody solution for 45 min. Membranes were washed three times with PBST Before detecting peroxidase activity.

3.Results

3.1. Characterization of dwarf mutant 84133 plants

Previously,the foxtail millet 84133 dwarf line was shown to be insensitive to exogenous GA3treatment and to harbor a dominant dwarf gene [22]. Here, we found that the F1generation of the Yugu 1 (Y1) × 84133 cross showed an intermediate plant height compared with its parents (Fig.1A). We analyzed the distribution of plant height in Yugu 1 × 84133 F2and RHL offspring and found that they separated into three categories in a 1:2:1 segregation ratio,(respectively,70-80 cm, 40-50 cm, and 20-30 cm for RHL; and 120-130 cm,65-75 cm, and 30-40 cm for the F2population) (Fig. S1, Table S2). This segregation pattern indicated that 84133 plants carried a semi-dominant dwarfism gene. In each population,plants whose height fell into the first category were defined as normal plants,and those in the two shorter height categories were defined as dwarf plants. The homozygous dwarf (hoD)plants among the RHL offspring or F2population of Y1×84133 were about two-thirds shorter than the corresponding homozygous normal (hoN) plants, while the heterozygous dwarf plants (heD) were approximately 30%-40% shorter than hoN plants (Fig. 1B). All RHL offspring produced the same number of elongated internodes (Fig. 1C), indicating that plant height differences among individuals were the result of shorter internodes. Analyses of tissue sections revealed that the epidermal cells were much shorter in hoD plants than in hoN plants(Fig.1D,E).

3.2. Map-based cloning of the causal gene, D1, in 84133

We used a large RHL population derived from a cross of Zhangai 10 × 84133 cross to locate the D1. Based on a linkage analysis using SSR markers, we found that all plants containing the 84133-sourced SSR del037, regardless of their heterozygous or homozygous genotype, exhibited dwarfism,while plants only harboring homozygous Zhangai 10-sourced del037 showed the normal plant height (Fig. 2A), This indicated that marker del037 was closely linked to D1 due to its co-segregation with plant height.Next,240 RHL individuals were screened using SSR markers around del037,and D1 was preliminarily localized between del037 and del366F2R1,which were separated by about 500 kb (Fig. 2B). A higher-resolution analysis with more molecular markers and a larger population positioned D1 within a 72-kb region(Fig.2C).In this target interval, six open reading frames (ORFs) had been annotated previously. The ORFs included those that encoded a hydroxylase (Seita.9G122600 and Seita.9G122700), a putative phospholipase (Seita.9G122800), an amidase (Seita.9G122900), a DELLA protein from the GRAS family(Seita.9G123000),and an EamA-like transporter family protein (Seita.9G123100) (Fig.2D).Of these,DELLA was considered a candidate for D1.

3.3. Structural analysis of D1 in 84133 plants

Polymerase chain reaction (PCR) analysis of the full length DELLA gene (Seita.9G123000) revealed that an 8-kb fragment was amplified only from the genomic DNA of 84133 and RHL hoD plants, in addition to the normal 2-kb product amplified from all genotypes (Fig. 3A). Sequencing of the 2-kb fragment confirmed that it was Seita.9G123000, which has been annotated to encode a DELLA protein. Further phylogenetic analysis of the DELLA gene in related species revealed that Seita.9G123000 is a foxtail millet ortholog of SLR1 in rice, and Rht in wheat(Fig.S2).

Comparisons of the 2-kb and 8-kb fragment sequences revealed an insertion of a 5496-bp Copia retrotransposon,271-bp downstream of the Seita.9G123000 initiation codon in dwarf individuals.Sequence and structural analyses indicated that the inserted Copia retrotransposon comprised two 636-bp long terminal repeats (LTRs), a primer binding site (PBS), a polypurine tract (PPT), and elements encoding mobilityrelated proteins (Fig. 3B). Copia was flanked by 5-bp direct repeats of CTACA, representing the target site duplication(TSD) generated after Copia insertion. A BLAST search of the panicoid repetitive elements databases (http://www.repeatmasker.org/cgi-bin/WEBRepeatMasker) with the Copia DNA sequence revealed strong similarity between the retrotransposon and Copia-59 in maize.

Southern blot analyses were conducted using the coding region of Seita.9G123000 as a probe to confirm the Copia insertion.The results showed that,compared with Y1 plants,the 84133 plants contained a larger band(marked with*in Fig.3C) with the expected size of Copia, which confirms the insertion of Copia retrotransposon (Fig. 3C). Additionally, the 84133 plants contained two copies of DELLA, in line with our PCR amplification results in which 2-kb and 8-kb bands were simultaneously observed (Fig. 3A). This indicated that the DELLA gene might have undergone a duplication event,possibly related to the Copia insertion, as reported previously[23].

Fig.1- Characterization of dwarf line, 84133.(A)Morphology of Y1(right),F1 individual derived from cross between Y1 and 84133(center),and 84133 mutant(left).(B)Plant height of RHLs and F2 progenies derived from cross between 84133 and Y1.Bars represent±standard deviation(n = 5). Asterisks represent statistically significant differences(Student's t-test, **P ＜0.001).Homozygous dwarf(hoD),heterozygous dwarf(heD),and homozygous normal(hoN)plants in RHL and F2 population.(C)Stems divided into internodes,with same number of internodes among hoN,heD,and hoD plants of RHL.(D)Longitudinal sections of last stems of mature hoN(left)and hoD(right)in RHL.(E)Statistical analysis of internode epidermal cell length in hoD and hoN plants of RHL(n = 5 cells)(Student's t-test, **P ＜0.001).

Previous studies have shown that a deletion or insertion in the DELLA N-terminal can lead to dwarfism[5,15].However,in a transcriptional analysis, no bands were detected for 84133 plants or hoD individuals using a specific primer spanning the entire ORF of DELLA, even with a longer extension step (Fig.S3).This suggested that its transcription might be interrupted by the Copia insertion, as observed in Drosophila [32]. We conducted 3′ and 5′ RACE-PCR to amplify the putatively truncated transcripts. As expected, a 657-bp chimeric transcript (coding region) was obtained by 3′ RACE assay. This chimeric transcript consisted 271 bp from the 5′ end of the DELLA gene and 386 bp of the partial LTR sequence from the 5′portion of the inserted Copia element (Figs. 3D, S4). The 5′RACE-PCR also successfully amplified a 1557-bp truncated DELLA transcript (coding region) lacking the conserved domains at the 5′ end, including the DELLA and TVHYNP motifs (Figs. 3D, S5). The putative transcription start site was mapped to the original coding sequence adjacent to the 3′LTR(Figs. 3D and S5). Therefore, we concluded that the 3′ LTR of the Copia retrotransposon may provide regulatory sequences for initiating the expression of the truncated transcript.Here,we named the short chimeric transcript D1-cT (chimeric transcript), the truncated transcript D1-TT (truncated transcript), and the full-length wild-type transcript D1-WT(wild-type transcript).

Fig.2- Map-based cloning of D1.(A)Genotyping of normal tall plants and dwarf plants in RHL population using SSR marker,del037.Parents(Zhangai 10 and 84133),served as controls.(B) D1 was roughly-mapped to small region between SSR loci del366F2R1 and del037 using 240 individuals derived from heterozygous RHL plants.These two markers are separated by about 500 kb according to Phytozome reference genome sequence(http://www.phytozome.net).Thick bar represents genomic region,D and N refer to dwarf and normal phenotypes, respectively.Number of recombinants between adjacent markers is indicated above linkage map.White and black bars represent homozygous Zhangai 10 and 84133 genotypes,respectively,gray bar corresponds to heterozygous genotype.(C)High-resolution linkage map of D1.D1 was fine-mapped to 72-kb interval between si332 and si188 based on genotype assessment of 1500 RHLs.(D)Annotation of candidate region.Arrows indicate putative genes predicted in Foxtail Millet Genome Annotation Database(http://www.phytozome.net).White arrow represents D1(Seita.9G123000).

A previous report showed that DELLA transcription is controlled by a negative feedback loop in response to GA content [33]. Consistent with that report, we found that an exogenous GA3treatment significantly up-regulated the expression of D1-WT. In contrast, D1-cT and D1-TT were unaffected by GA3treatment (Fig. 3E), indicating intact regulatory elements including upstream and downstream elements are required for the transcriptional negativefeedback response of D1.

Next, we determined the tissue-specific expression of D1-TT and wild-type D1-WT in panicle,leaf,internode and root by RT-PCR, and found that both were highly expressed in these tissues without obvious differences in their patterns (Fig. S6).This finding indicated that the newly recruited LTR as well as original promoter of D1 may drive ubiquitous expression, at least among these analyzed tissues. To further confirm that DELLA transcription was disrupted by the Copia retrotransposon insertion, we conducted an immunoblot analysis using a specific anti-D1-WT polyclonal antibody,which was prepared using the full-length D1-WT as the immunogen (Fig. S7). In the hoD of the RHL, we detected a band that was 10 kD smaller than the corresponding band detected from hoN, this smaller band corresponded the expected size of the D1-TT protein, and its abundance was not affected byGA3treatment, even after 4 h, compared with that of D1-WT in hoN(Fig.3F).These results indicated that the truncated DELLA protein(i.e.,D1-TT)was present in 84133 and hoD plants of the RHL. Collectively, our findings clearly demonstrated that the Copia retrotransposon insertion resulted in reprogramming of the DELLA transcription.

3.4. D1-TT determines the dwarfism of 84133

The abundance of the D1-TT protein was resistant to GA3treatment, indicating this truncated protein modulates the GA-insensitive dwarfism of 84133. To test this idea, we overexpressed D1-TT in Y1.The transgenic plants exhibited extreme dwarfism and produced broad, dark-green leaves similar to those of 84133(Fig.4A).Overexpression of D1-TT in rice also led to dwarfism (Fig. 4B). All transgenic events were confirmed by Southern blotting (Fig. S8). These findings confirmed that D1-TT determines the dwarfism of 84133.

3.5. D1-TT could not interact with SiGID1 in absence or presence of GA3

The formation of the GA-GID1-DELLA complex is known to be critical for DELLA degradation and GA signal transduction[10,14].D1-TT accumulated to high levels and was resistant to GA3treatment (Fig. 3F), suggesting that it could not interact with GID1. To test this hypothesis, we performed yeast Y2H assays. First, we identified SiGID1 by sequence homology BLAST searches against the foxtail millet genome database using OsGID1 as the query, Seita.3G246300 showed high homology with OsGID1, and was annotated as a GA receptor.We performed Y2H assays using SiGID1 (Seita.3G246300) as bait, and D1-WT or D1-TT as prey, to determine the affinity among these proteins in foxtail millet. As expected, D1-TT could not bind to SiGID1, regardless of whether GA3was present (Fig. 5). This result implied that D1-TT still accumulated in the dwarf 84133 plants in GA3treatment because it could not interact with SiGID1. Interestingly, D1-WT strongly interacted with SiGID1 in a partially GA3-independent manner (Fig. 5). Partially GA-independent GID1 proteins have also been described in Arabidopsis and soybean[20].

4. Discussion

4.1. Retroposition created novel functional genes

Retroposition events mediated by retrotransposons have long been recognized as being essential for the development of novel genes and genome evolution [25]. However, there is currently little direct evidence for the generation of novel functional genes by retrotransposon activity in the plant kingdom.

Previously, foxtail millet 84133 was identified as a spontaneous dwarf mutant in Inner Mongolia, China [22]. Here, we showed that a Copia retrotransposon insertion into DELLA is responsible for dwarfism of 84133. Although the Copia is widely found in the foxtail millet genome in BLAST searches,we have not observed similar spontaneous mutants during our 20-year breeding program in Northern China. Thus, Copia activation and insertion is an accidental event.Such an event may have been triggered by the lower temperature for foxtail millet grown in in Inner Mongolia, China, because low temperatures have been shown to activate transposons [34],resulting in gene or genome instability. Similarly to the mutation in 84133 plants, the Tom Thumb, the wheat dwarf mutant harboring a retrotransposon-inserted Rht-B allele(Rht-B1c)originated from Tibet,China,where the temperatures are low for wheat growth.Thus,low temperatures may be one of the key factors to affecting activation of transposons, and consequently,genetic diversity.

Fig.5-Interaction between SiGID1 and D1-WT or D1-TT.Yeast two-hybrid assays conducted using SiGID1 as bait and D1-WT or D1-TT as prey.Gold yeast strain was grown on the-His/-Ade/-Leu/-Trp selection medium supplemented with X-α-Gal with or without GA3 or 3 mmol L-1 3A-T. All clones were grown for 72 h.

Although retrotransposons have been shown to mutate genes in a variety of ways [1], the reprogramming and shuffling of genes by retrotransposon insertion have rarely been reported. In Drosophila melanogaster, a retrotransposon insertion caused the shuffling of nearby sequences to produce chimeric and truncated transcripts with novel functions [32].The production of red fruit flesh in blood oranges is regulated by the tissue-specific and cold-dependent expression of a chimeric retrogene, in which the partial 3′ LTR of a Copia retrotransposon is fused to the coding sequence of a MYB transcription factor gene to form Ruby. Changes in Ruby expression levels in blood oranges are caused by the recruitment of new regulatory sequences from the 3′LTR of the Copia retrotransposon [7]. Similarly, the D1-TT transcript analyzed in our study also originated from Copia-mediated gene reprogramming (Fig. 3D). It lack DELLA and TVHYNP motifs compared with the wild type and potentially recruited the 3′LTR of the retrotransposon as a new promoter, as supported by the observation that D1-TT was unresponsive to GA3treatment at the transcriptional level (Fig. 3E). Thus, we suggest that D1-TT is a retrogene. Our finding provides a rare example of a retrotransposon that induces gene reprogramming leading to a trait variation.

Our PCR amplification and Southern blotting results showed that the 84133 genome contains two copies of DELLA(Fig.3A,C),although it is still unknown where the duplicated gene (without the Copia insertion) is located and how and when it developed. Nevertheless, an interesting study demonstrated that a Ty1-copia retrotransposon insertion led to Rht-D1b duplication in wheat, possibly through unequal chromosome crossing over [35]. Therefore, it is likely that the D1 duplication in the 84133 genome is closely related to the Copia insertion,further studies are required to explore this idea.

In a previous study, searches of rice and maize genome databases identified two paralogs of SLR1, SLR1-like1, and -2,which also lack the N-terminal conserved regions such as the DELLA and TVHYNP motifs,and weakly repress GA responses[36]. However, the evolutionary relationship between DELLA and these SLRLs is still unclear. Here, D1 shuffling mediated by the Copia retrotransposon may provide a direct source for the origination and development of DELLA-like genes and supports one possible scenario in which the DELLA gene evolved first, and then SLRLs accidentally evolved through retrotransposon insertion and consequent duplication.

4.2.84133 plants are a potentially valuable resource for hybrid breeding and utilization of heterosis in foxtail millet

Dwarf breeding has been successfully used to prevent stem lodging and significantly increase grain yield in rice and wheat[37].Identification of suitable dwarf mutants and genes is critical for breeding elite dwarf cultivars. For instance, the introduction of two famous dwarf genes, semi-dwarf1 (sd1) in rice and Rht in wheat have generated ideal dwarf varieties[37].However, dwarf breeding has not been successful in foxtail millet. In the past 20 years, we have collected many dwarf mutant lines in China to identify valuable resources for dwarf breeding[22].Recently,we cloned a recessive dwarf gene,D2,in a callus-based regenerating mutant. The D2 encodes a cytochrome P450 enzyme involved in the biosynthesis of phytohormones such as GA and brassinolide. The d2 plants show a mild reduction in plant height without a penalty in 1000-grain weight, indicating potential for dwarf breeding[27].Here,we isolated a semi-dominant dwarf gene,D1,from 84133 plants. A phylogenetic tree based on DELLA in related species showed that D1 is a foxtail millet ortholog of SLR1 in rice and Rht in wheat.Mechanism analysis demonstrated that a retrotransposon insertion induced D1 reprogramming (Fig.3D),and the resultant N-terminal-deleted DELLA(D1-TT)was resistant to GA treatment due to its inability to interact with the GA receptor SiGID1, resulting in dwarfism (Figs. 3F, 5).Taken together, these findings indicate that D1 has a conserved function in the regulation of GA signaling and plant height,similarly to those of SLR1 or Rht.

Previous studies showed that partially or completely inhibited interactions between TaGID1 and the mutant Rht result in semi-dominant or dominant dwarfism in wheat,correspondingly [5]. Based on this fact, it was concluded that different degrees in the affinity between TaGID1 and the mutant Rht determine semi-dominant or dominant gene action on plant height. However, our results clearly showed that a complete lack of interaction between SiGID1 and D1-TT results in only semi-dominant dwarfism in 84133 plants(Figs.5, 1A). This result indicated that there is no direct correlation between the affinity of the GID1-DELLA complex and the semi-dominant or dominant dwarfism in foxtail millet.Other DELLA gene mutants have been identified in various species,including d8 in maize, rga-Δ17 in Arabidopsis, and ds-1 in Brassica napus [11,38,39]. In those mutants, deletion of the DELLA domain (i.e., d8 and rga-Δ17) or complete inhibition of the interaction between GID1 and DELLA(i.e., ds-1) also leads to semi-dominant dwarfism, consistent with our results. In our opinion, the semi-dominant action may be caused by differences in the amount of mutant DELLA proteins between heterozygous and homozygous dwarf plants.

Dominant or semi-dominant dwarf genes have special advantages in crop hybrid breeding programs because only one parent needs to carry the semi-dominant or dominant dwarf gene. Thus, the parental genetic background is not limited. A recent study showed that rice plants harboring heterozygous Slr1-d6, a new allele of SLR1, were equivalent in height to the standard sd1-containing plants, but with a 25%grain yield increase[28].Heterozygous D1 plants were only 30%-40% shorter than wild-type plants, a much milder phenotype than that of the homozygous dwarf plants, in which plant height was reduced by 60%-70%(Fig.1A,B).This indicated that D1 is potentially suitable for hybrid breeding and utilization of heterosis in foxtail millet. Notably, heterosis is present and important for grain yield increases in foxtail millet.84133 plants may serve as a direct target and valuable resource for dwarf breeding especially in a hybrid breeding program.

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2018.12.008.

Conflict of interest

The authors have declared that no conflict of interest exists.

Acknowledgements

This study was supported by the National Natural Science Foundation of China (31871634, 31500985).

Data availability

The sequences of Copia retrotransposon and D1 have been deposited in GenBank under the accession number KX098474 and KX098475,respectively.