Barley (Hordeum vulgare) spikes are developmentally switched from two-rowed to six-rowed by a single recessive gene, six-rowed spike 1 (vrs1), which encodes a homeodomain-leucine zipper I class transcription factor. Vrs1 is a paralog of HvHox2 and both were generated by duplication of an ancestral gene. HvHox2 is conserved among cereals, whereas Vrs1 acquired its current function during the evolution of barley. It was unclear whether divergence of expression pattern or protein function accounted for the functionalization of Vrs1.
Here, we conducted a comparative analysis of protein functions and gene expression between HvHox2 and Vrs1 to clarify the functionalization mechanism.
We revealed that the transcriptional activation activity of HvHOX2 and VRS1 was conserved. In situ hybridization analysis showed that HvHox2 is localized in vascular bundles in developing spikes, whereas Vrs1 is expressed exclusively in the pistil, lemma, palea and lodicule of lateral spikelets. The transcript abundance of Vrs1 was > 10-fold greater than that of HvHox2 during the pistil developmental stage, suggesting that the essential function of Vrs1 is to inhibit gynoecial development. We demonstrated the quantitative function of Vrs1 using RNAi transgenic plants and Vrs1 expression variants. Expression analysis of six-rowed spike mutants that are nonallelic to vrs1 showed that Vrs1 expression was up-regulated by Vrs4, whereas HvHox2 expression was not.
These data demonstrate that the divergence of gene expression pattern contributed to the neofunctionalization of Vrs1.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Inflorescence architecture is a major determinant of the components of final grain number per plant at harvest. The flowers of grass plants, such as wheat (Triticum aestivum), barley (Hordeum vulgare), rice (Oryza sativa) and maize (Zea mays), develop on a specialized short branch called a spikelet. The number of fertile spikelets per inflorescence is positively associated with grain yield per plant. Several major quantitative trait loci (QTLs), such as Grain number 1a (Gn1a), which encodes cytokinin oxidase, and WEALTHY FARMERS PANICLE 1 (WFP1), which encodes SQUAMOSA PROMOTER BINDING PROTEIN-LIKE 14 (SPL14), have been cloned in rice (Ashikari et al., 2005; Miura et al., 2010); however, in the Triticeae, none of the major QTLs or genes involved in grain number per inflorescence have been cloned, except for barley six-rowed spike 1 (vrs1) (Komatsuda et al., 2007) and intermedium-c (int-c) (Ramsay et al., 2011).
Barley is one of the most important crops in the Triticeae tribe, which falls within the Pooideae subfamily of the grass family Poaceae. The barley inflorescence, which is called a ‘spike’, is composed of three spikelets (one central spikelet and two lateral spikelets) per rachis node. The spikelet arrangement characteristic of Hordeum (Bothmer et al., 1995) is found in other grass genera, for example, some species of Hilaria (Barkworth, 2002). In the two-rowed spike, only the central spikelet is fertile and produces grain, whereas in the six-rowed spike, three spikelets are fertile and the plant produces three times as much grain as plants with two-rowed spikes. The six-rowed spike is controlled by vrs1, which is located on the long arm of chromosome 2H. The wildtype Vrs1 allele encodes a homeodomain-leucine zipper (HD-Zip) I transcription factor and suppresses the development of lateral spikelets (Komatsuda et al., 2007). Barley collections of morphological mutants and natural variants have been assembled and used as a valuable resource to examine fundamental biological processes (Druka et al., 2011). Loss of function of Vrs1 results in six-rowed spikes by restoring fertility of the lateral spikelets. Three other six-rowed spike loci (vrs2 on chromosome 5H, vrs3 on chromosome 1H and vrs4 on chromosome 3H) have been described in induced barley mutants; recessive alleles at each of these loci enhance the development of lateral spikelets to various degrees and result in seed set depending on spikelet position in the spike (Lundqvist et al., 1997). The intermedium-c (int-c) locus on chromosome 4H, which encodes an ortholog of maize teosinte branched1 (tb1), a major domestication gene that regulates tillering (Doebley et al., 1997), modifies Vrs1 alleles to set grain in some lateral spikelets (Ramsay et al., 2011).
HD-Zip proteins are composed of a homeodomain (HD), which functions as a DNA-binding site, and an adjacent leucine zipper (Zip), which acts as a dimerization motif. The HD-Zip proteins form a transcription factor family that is unique to the plant clade (Ariel et al., 2007). HD-Zip proteins bind to specific DNA sequences as homodimers or heterodimers through their leucine-zipper domains, and the absence of a leucine zipper abolishes their binding ability (Sessa et al., 1993). These proteins can be classified into four subfamilies (HD-Zip I to IV), according to a set of distinctive features that includes DNA-binding specificities, gene structures, additional common motifs and physiological functions. HD-Zip I proteins mainly regulate the plant's response to abiotic stresses, such as drought, extreme temperatures and osmotic and light stress (Ariel et al., 2007). Recently, grassy tillers1 (gt1), a Vrs1 homolog that regulates lateral branching in maize, was isolated (Whipple et al., 2011). gt1 expression is induced by shading and is promoted by tb1.
In rice and maize, HD-Zip I class genes consist of 14 and 17 members, respectively, and have evolved through a series of gene duplications (Agalou et al., 2008; Zhao et al., 2011). Vrs1 and HvHox2 must have diverged after the separation of Brachypodium distachyon from the Pooideae (Sakuma et al., 2010). The biological function of HvHOX2 in barley is unknown, but is conserved in other cereals. As a mechanism of functionalization in duplicate genes, differentiation of both gene expression and protein function are important (Hughes, 1994; Oakley et al., 2006; Ganko et al., 2007; Hanada et al., 2009). The HvHox2 gene is expressed broadly, whereas Vrs1 is specifically expressed in immature spikes (Sakuma et al., 2010). The HD-Zip domains in HvHOX2 and VRS1 are well conserved. The 300 bp insertion into the third exon of Vrs1 introduced a stop codon; thus, VRS1 lacks a motif that is conserved at the C-terminal end of the HvHox2 orthologs (Sakuma et al., 2010). Based on these data, we hypothesized that Vrs1 may have acquired a novel role by neofunctionalization (Ohno, 1970). It was unclear, however, whether divergence of expression pattern or protein function, or of both of these aspects, accounted for the functionalization of copy gene Vrs1.
To investigate how Vrs1 acquired its current functions, we characterized the functions of HvHOX2 and VRS1 as transcriptional regulators. We then performed a detailed expression analysis to identify differences in expression pattern between HvHox2 and Vrs1. Furthermore, mutant and transgenic analyses were conducted to determine the molecular function of Vrs1 and to decipher the genetic interactions among the six-rowed spike genes.
Materials and Methods
Barley (Hordeum vulgare L.) cvs Golden Promise (GP, JP15923), Hanna (JP15594), Bonus (JP15929), Haruna Nijo (HN, JP67556) and Kanto Nakate Gold (KNG, JP15436), which were used as two-rowed representatives, and Morex (JP46314), Azumamugi (AZ, JP17209) and Hayakiso-2 (HK2, JP18099), which were used as six-rowed representatives, were obtained from the National Institute of Agrobiological Sciences (NIAS) Genebank, Japan. Six induced six-rowed mutants and their original cultivars (Supporting Information, Table S1) were obtained from the Nordic Gene Bank (Alnarp, Sweden).
Transactivation activity assay
Transactivation activity assays were performed using the Matchmaker Gold Yeast Two-Hybrid System (Clontech, Mountain View, CA, USA). To construct the necessary serial vectors, the full-length coding sequences of HvHox2 and Vrs1 and the sequences of various domain truncation products were amplified by PCR (Table S2). The PCR products were digested with EcoRI and BamHI and cloned into pGBKT7, cut with the same restriction enzymes, to fuse them downstream of the GAL4 DNA-binding domain. The resulting plasmids were transformed into the Y2HGold yeast strain and selected on synthetic dropout (SD) medium lacking tryptophan (–Trp). After 2 d of incubation at 30°C, the yeast colonies were diluted to an OD600 of 0.1, and 5 μl of culture was dropped onto SD/−Trp medium with or without the antibiotic aureobasidin A (AbA). These plates were incubated at 30°C until yeast cells had grown to form colonies.
In situ mRNA hybridization analysis
The Vrs1 gene segment comprising part of the 3′-UTR (300 bp) and the full-length cDNA of HvHox2 (1072 bp) were amplified from cDNA isolated from immature barley spikes using specific primers (Table S2). Immature spikes were developmentally staged by observation under a stereoscopic microscope (Kirby & Appleyard, 1981). The PCR product was cloned into the pBluescript II KS (+) vector (Stratagene, La Jolla, CA, USA). Two clones with different insert orientations were linearized with NotI (for Vrs1) or EcoRI (for HvHox2) and were used as templates to generate antisense and sense probes using T3 or T7 RNA polymerase. In situ hybridization was conducted as in Komatsuda et al. (2007).
Antibodies against VRS1 were obtained by immunizing rabbits with the synthetic peptide C (for conjugation) -X (aminohexanoic acid) VPEWFLA (positions 216–222 of VRS1). Immature spikes at the awn primordium stage were used for immunostaining as described previously (Yamaji & Ma, 2007).
RNA extraction and Absolute Quantitative Real-Time PCR
Total RNA was extracted from immature spikes (ca. 10 spikes per extraction) using TRIzol (Invitrogen, Carlsbad, CA, USA). RNA was quantified using NanoDrop 1000 (Thermo Fisher Scientific, Waltham, MA, USA). To remove genomic DNA contamination, RNA was treated with RNase-free DNase (Takara Bio, Otsu, Japan). First-strand cDNA was synthesized with SuperScript III (Invitrogen) and first-strand cDNA derived from 25 ng RNA was used as a template. Transcript abundances of each gene were measured by quantitative real-time PCR (qRT-PCR) using a StepOne Real-Time PCR System (Applied Biosystems, Foster, CA, USA) and THUNDERBIRD SYBR qPCR Mix Kit (Toyobo, Osaka, Japan) according to the manufacturer's protocols. Primers used for qRT-PCR are listed in Table S2. Each gene fragment was cloned into pCR4-TOPO (Invitrogen) or pBluescript II KS (+) (Table S2). Plasmid DNA harboring each gene fragment was used to generate standard curves for absolute quantification. CT values for each sample were converted into copy numbers using the standard curves. (Fig. S1). The HvActin gene, which showed a stable expression pattern throughout spike development (Fig. S2), was used as the internal control for calculating the relative expression levels of the Vrs1 ((Vrs1 mRNA copy number µg–1 total RNA)/(HvActin mRNA copy number µg–1 total RNA)) and HvHox2 ((HvHox2 mRNA copy number µg–1 total RNA)/(HvActin mRNA copy number µg–1 total RNA)). qRT-PCR analysis was performed at least three times for each sample. Biological replicates of at least three independent RNA extractions per sample were performed.
Transformation of barley
To prepare the RNAi construct, the Vrs1 gene segment (628 bp) was amplified using specific primers (Table S2). The PCR product was cloned in the forward and reverse orientations into each of the pIPKb008 and pIPKb009 vectors in which transgene expression was under control of either the rice Actin promoter or the CaMV 35S promoter, respectively (Himmelbach et al., 2007). The constructs were introduced into Agrobacterium tumefaciens strain AGL-1 and transferred into immature barley (cv Golden Promise) embryos by Agrobacterium-mediated transformation, as previously described (Hensel et al., 2008).
Determination of the transcriptional activation domain
To examine whether HvHOX2 and VRS1 have transcriptional activation potential, we conducted a yeast one-hybrid analysis. The yeast GAL4 DNA-binding domain (GAL4BD) was fused to different segments of HvHOX2 or VRS1 (Fig. 1). If HvHOX2 (VRS1) has transcriptional activation potential, the GAL4BD-HvHOX2 (VRS1) fusion protein bound to the GAL4-responsive promoter would be expected to induce AUR1-C reporter gene expression and the expression would confer strong resistance to the toxic drug Aureobasidin A (AbA) of the host yeast strain Y2HGold. All transformants grew well on SD/−Trp medium, because pGBKT7 contains the TRP1 nutritional marker. Yeast colonies of each transformant were then transferred to SD/−Trp medium containing AbA. Yeast carrying GAL4BD-HvHOX2-full (I) and GAL4BD-VRS1-full (VIII) grew well on SD/−Trp/+AbA medium (Fig. 1), suggesting that both HvHOX2 and VRS1 could act as transcriptional activators, at least in yeast.
To localize the activation domain (AD), nine deletion derivatives of the constructs were tested. The constructs that included the C-terminal region (CTR) (II, V, VI, XI) showed transcriptional activation, whereas those that lacked this region (III, IV, IX, X) did not (Fig. 1). HvHOX2 harbors a motif in the CTR that consists of 19 amino acids (VII) and is conserved in the cereal HvHOX2 orthologs but absent in VRS1. The construct that included only the motif (VII) did not show reporter gene activation (Fig. 1). These results indicate that the transcriptional AD of HvHOX2 and VRS1 is localized to a 60-amino-acid region in the CTR, at least in yeast.
Localization of Vrs1 and HvHox2 transcripts
In situ RNA hybridization using a HvHox2 probe (Fig. 2a) revealed signals in the vascular tissue in the rachis at the awn primordium stage (Fig. 2b). At the white anther stage, when floral organ differentiation was completed, signals were localized to provascular tissue in the pedicel (junction between the rachis and spikelet) of two lateral spikelets (Fig. 2d,f). Although a signal was detectable at the awn primordium and white anther stages, HvHox2 expression was not detected at the triple mound stage. No HvHox2 signal was detected in hybridization with the sense probe (Fig. 2c,e,g). We used the Vrs1 deletion (null) six-rowed spike mutant, hex-v.3 (Komatsuda et al., 2007), as a source of immature spikes and therefore did not detect any similar Vrs1 sequences during RNA in situ hybridization with the HvHox2 probe. Owing to relatively low expression (see later), HvHox2 was not readily detected by in situ hybridization. To enhance the HvHox2 signal, we included the entire cDNA sequence of HvHox2 in the probe (Fig. 2a).
On the other hand, in situ RNA hybridization using a Vrs1-specific probe that shows no sequence similarity with HvHox2 (Fig. 2a) revealed strong signals in the lateral florets of two-rowed cv Bonus (Fig. 2h,j). Vrs1 expression was localized to the lemma, palea, lodicule and, most strongly, the pistil in the lateral florets at the white anther stage (Fig. 2i,k). From this stage, floral organs transited from the differentiation to the developmental phase. The in situ hybridization signal was at background values in anther tissue, and Vrs1 was not detectable in glumes. These data showed a highly floral organ-specific pattern of gene expression. Interestingly, rachillae bearing rudimentary florets, which are ultimately suppressed in barley, showed a strong signal (Fig. 2k), comparable to that in the pistil. No Vrs1 expression was detected in the central spikelet meristem during the triple mound and glume primordium stages or in any organs in the central spikelet during the white anther stage. No signal was detected using the sense probe in wildtype two-rowed spikes (Fig. 2l) and the antisense probe in hex-v.3 mutants (Fig. 2m). Thus, the Vrs1 probe used in this in situ hybridization did not cross-hybridize with HvHox2 mRNA or any other similar gene sequence.
Localization of VRS1 in nuclei
Immunostaining with an antibody against VRS1 was performed to investigate the tissue-specific localization of VRS1. VRS1 signals were detected in the lateral spikelets of two-rowed cv Bonus at the awn primordium stage (Fig. 3a,b). No signal was detected in the central spikelet. VRS1 and HvHOX2 were predicted to have nuclear localization signals (‘RRRRRRSAR’ and ‘RPRARRRRRRAAR’, respectively) by the PredictProtein database (http://www.predictprotein.org/), indicating that the two proteins are targeted to the nucleus. To confirm the nuclear localization of VRS1, the immunostained cells were costained with DAPI, which labels DNA. VRS1 was indeed localized to nuclei (Fig. 3c). Only an autofluorescence signal was detected in the rachis vascular tissue in the negative control (Fig. 3d).
The hex-v.3 mutant lacked Vrs1; only autofluorescence was detected in the vascular bundles of the rachis and spikelets (Fig. 3e,f). The anti-VRS1 antibody failed to detect any VRS1 in the central and lateral spikelets and their nuclei in hex-v.3 (Fig. 3g). The signal at the rachis of Fig. 3(g) was autofluorescence, as seen in the negative control that lacked anti-VRS1 antibody (Fig. 3h).
The abundance of Vrs1 transcription is greater than that of HvHox2 in developing spikes
Absolute qRT-PCR was performed to compare the expression levels of Vrs1 and HvHox2 throughout the various developmental stages of the spike (Fig. 4). We analyzed expression in two two-rowed cultivars and two six-rowed cultivars. In Bonus (Fig. 4a), a two-rowed cultivar possessing the Vrs1.b3 allele, Vrs1 transcript abundances ((Vrs1 mRNA copy number µg–1 total RNA)/(HvActin mRNA copy number µg-1 total RNA)) were significantly (P <0.05) higher than HvHox2. At the lemma primordium and stamen primordium stage, the Vrs1 expression was > 20 times that of HvHox2. The expression levels of the two genes were comparable in the green anther and yellow anther stages. A similar expression pattern of Vrs1 and HvHox2 was observed in Kanto Nakate Gold (KNG; Fig. 4b), which is a two-rowed cultivar possessing the same Vrs1.b3 allele as Bonus. Vrs1 mRNA was most abundant from the awn primordium stage to the white anther stage, during which pistil differentiation begins. HvHox2 mRNA was most abundant from the white anther stage to the yellow anther stage.
In Azumamugi (AZ), a six-rowed cultivar, the Vrs1 transcript abundance was much greater than that of HvHox2 (Fig. 4c). AZ possesses the vrs1.a1 allele, which encodes a truncated VRS1 protein (Komatsuda et al., 2007). The Vrs1 expression level in Hayakiso 2 (HK2), another six-rowed cultivar, was one-fifth of that in Bonus and one-third of that in AZ (Fig. 4d). HK2 possesses the vrs1.c allele, which does not alter the deduced amino acid sequence encoded by Vrs1 (Saisho et al., 2009). These results suggest that the reduced Vrs1 expression level found in HK2 was not sufficient to suppress lateral spikelet fertility, and thus produced fully fertile lateral spikelets. The HvHox2 expression level did not differ significantly between AZ and HK2. These data show that the production of six-rowed spikes in HK2 can be ascribed to the reduced Vrs1 transcript abundance in this cultivar.
Restoration of lateral spikelet fertility by Vrs1 RNA interference
To demonstrate the quantitative function of Vrs1, we transformed the two-rowed barley cv Golden Promise with a Vrs1-specific hairpin-RNA interference (RNAi) construct under the control of the rice Actin1 promoter or the CaMV 35S promoter. We obtained 33 independent T0 plants that showed a developed awn in the lateral spikelets using the constructs driven by both of the promoters. In the case of the construct driven by the rice Actin1 promoter, three T0 plants (BG87/1E18, BG87/1E35 and BG87/2E4) produced fertile lateral spikelets that yielded grains (Table S3). We selected seven independent T1 lines that had shown phenotypic change in the T0 generation for further analysis. All seven T1 plants harboring the RNAi-expression cassette exhibited an elongated awn in the lateral spikelets (Table S3). Grains were produced in only three T1 lines, namely BG87/1E18 (0–38% fertility in the lateral spikelets; spike-to-spike variation exists), BG87/1E35 (0–14% fertility), and BG87/2E04 (0–5% fertility), indicating that grain productivity in the lateral spikelets was inherited by the T1 plants (Fig. 5a–e). No phenotypic changes were observed in organs other than lateral spikelets, in agreement with Vrs1 being expressed only in the lateral spikelets (Table S3).
To evaluate the correlation between Vrs1 mRNA level and lateral spikelet development, we carried out qRT-PCR using RNAs isolated from immature spikes at the white anther stage. T1 plants derived from the BG78/3E02, BG87/1E18, BG87/1E35 and BG87/2E04 lines showed a Vrs1 expression level approximately half that of the wildtype Golden Promise or nontransgenic plants derived from segregation of T1 families (Fig. 5f). The levels of Vrs1 expression in T1 plants derived from BG78/1E01, BG78/1E02 and BG78/2E02 lines and the wildtype were almost the same. Furthermore, the level of HvHox2 expression showed no significant change between wildtype and transgenic plants, suggesting that Vrs1 was the sole gene targeted by Vrs1 RNAi (Fig. 5g). These results indicate that the level of Vrs1 expression and the development of lateral spikelet awns were negatively correlated (r = −0.55), supporting the hypothesis that Vrs1 expression suppresses the development of lateral spikelets in a quantitative manner.
Transcript abundances of Vrs1 and HvHox2 in nonallelic six-rowed spike mutants
To determine whether the other known six-rowed spike loci regulate the expression of Vrs1, we examined Vrs1 expression in vrs2, vrs3, vrs4 and int-c mutants. These mutants are nonallelic to the vrs1 locus (Lundqvist et al., 1997). Immature spikes at the awn primordium stage were used for RNA extraction. Expression levels of Vrs1 in these mutant lines and their original wildtype cultivars were compared, and the null hypothesis was that expression level is the same in the mutant and the wildtype plant. The largest and most statistically significant (P =0.0046) difference was observed between vrs4 and the wildtype (Fig. 6a). The transcript abundance of Vrs1 in vrs4 mutants was 1/12 of that of its two-rowed wildtype parent. This result suggests that the wildtype Vrs4 allele positively regulates Vrs1 expression in wildtype two-rowed barley. A smaller, but nonetheless statistically significant, difference was observed between vrs2 and wildtype plants, suggesting that the wildtype Vrs2 allele positively regulates Vrs1 expression. The hex-v.44 mutant was allelic to the vrs1 locus, which was caused by truncation of the VRS1 protein by a premature stop codon (Komatsuda et al., 2007). We included hex-v.44 as a negative control; however, for reasons unknown, the transcript abundance of Vrs1 in hex-v.44 was one-half that in the wildtype. No significant differences in Vrs1 expression were observed between the wildtype and vrs3 mutant or between the wildtype and int-c mutant. HvHox2 and HvActin expression levels were not significantly different between these mutants and the wildtype plants (Fig. 6b,c). These data indicate that the expression of HvHox2 was not regulated by Vrs1 or any of these six-rowed spike genes.
HvHOX2 and VRS1 function as a transcriptional activator
HD-Zip proteins are transcriptional factors and contain several kinds of modular domains such as HD, Zip, and AD. The nuclear localization of VRS1 in this current study supports the hypothesis that VRS1 functions as a transcriptional regulator. Among the induced six-rowed spike mutants, the mutation site was concentrated in the HD (Komatsuda et al., 2007), suggesting that the importance of HD as a DNA-binding domain. Among the 17 members of HD-Zip I proteins in Arabidopsis thaliana (Henriksson et al., 2005), six members (i.e. ATHB1, AT3G01470; ATHB5, AT5G65310; ATHB6, AT2G22430; ATHB7, AT2G46680; ATHB12, AT3G61890; and ATHB16, AT4G40060) act as transcriptional activators (Aoyama et al., 1995; Lee et al., 2001; Himmelbach et al., 2002; Wang et al., 2003; Henriksson et al., 2005), whereas Medicago truncatula HB1 encoding HD-Zip I functions as a transcriptional repressor (Ariel et al., 2010). In this study, we revealed that both HvHOX2 and VRS1 act as transcriptional activators. The AD is located in a 60-amino-acid segment at the CTR. Several six-rowed spike mutants that disrupted the AD of VRS1 have been found (Komatsuda et al., 2007), supporting the functional importance of the CTR. Recently, Arce et al. (2011) showed that a conserved motif at the CTR of the Arabidopsis HD-Zip I protein plays an important role in activating downstream genes. Additional support of the importance of the CTR was provided by Hofer et al. (2009). This group found that the homeotic tendril-less mutation in garden pea (Pisum sativum) was caused by a 12-amino-acid truncation of the HD-Zip I protein at the CTR. We only used yeast cells in this study; however, several papers have reported similar results using yeast and plant cells (Sablowski et al., 1994; Ohta et al., 2000; Depege-Fargeix et al., 2012). Although VRS1 lacks a conserved motif of 19 amino acids at the CTR end, our data suggest that this deletion does not affect the activation activity. In conclusion, HvHOX2 and VRS1 have both maintained the ancestral function of transcriptional activation.
Neofunctionalization of Vrs1 by modifying expression
The spatial and temporal regulation of homeobox gene transcription is essential for patterning the animal body axis and plant organ formation (Sentoku et al., 1999; Noordermeer et al., 2011). In this study, we revealed that HvHox2 mRNA localized to the vascular bundles of the rachis or pedicel, whereas Vrs1 mRNA localized to lateral rachillae and florets, and particularly to pistils. The spatial differentiation of gene expression was the most essential factor in the neofunctionalization of Vrs1 from its paralog, HvHox2. The qRT-PCR data also showed that the expression patterns of Vrs1 and HvHox2 are temporally and quantitatively distinct.
OsHox14, the rice HvHox2 ortholog, is expressed in all organs (Sato et al., 2010). Arabidopsis ATHB21 (AT2G18550), which is phylogenetically closest to HvHox2 (Sakuma et al., 2010), is also expressed in all organs (Henriksson et al., 2005). These data indicate that the expression pattern of HvHox2 is conserved, whereas Vrs1 has undergone neofunctionalization. The localization of HvHox2 mRNA in the vascular bundle is analogous to that of rice and Arabidopsis HD-Zip genes. In rice, OsHox4, which encodes HD-Zip I and is phylogenetically close to HvHox2 (Sakuma et al., 2010), was expressed in vascular bundles (Agalou et al., 2008). Furthermore, OsHox1, encoding HD-Zip II (Scarpella et al., 2000), and OSHB1 to OSHB4, encoding HD-Zip III proteins (Itoh et al., 2008), were also localized to vascular tissue. The HD-Zip III class genes in Arabidopsis, REVOLUTA/INTERFASCICULAR FIBERLESS1(IFL1), ATHB8, PHAVOLUTA/ATHB9, PHABULOSA/ATHB14 and CORONA/ATHB15, are well-characterized developmental regulators of the vascular bundles (Baima et al., 1995; Talbert et al., 1995; McConnell et al., 2001; Green et al., 2005; Prigge et al., 2005). The expression profile of these HD-Zip genes suggests that HvHox2 may play a particular role in vascular bundle development.
Phenotypic analysis of the vrs1 mutant and expression analysis suggest that VRS1 suppresses pistil development. Vrs1 is expected to regulate processes in cell division and elongation, and not in organ identity, as the floral organs in the lateral spikelets were fully differentiated. Vrs1 expression was not detected during anther development. Anthers in lateral spikelets showed various degrees of development, from rudimentary to fully developed (i.e. producing fertile pollen; Table S4). In the case of KNG, functional anthers develop and produce fertile pollen that is capable of outcrossing; however, pistils never develop in the lateral spikelets.
The transcript abundance of Vrs1 was greater than that of HvHox2 during floral organ differentiation. The quantitative nature of Vrs1 function was demonstrated by RNAi silencing and analysis of a natural variant with reduced Vrs1 expression; the stronger the expression of Vrs1, the greater the percentage of sterile lateral spikelets. These data strongly suggest that the differentiation of Vrs1 function might be a result not only of spatial changes in expression, but also of quantitative changes in gene expression. Casneuf et al. (2006) demonstrated that duplicated pairs that arose from small-scale duplication events tend to have a greater level of expression divergence than do pairs from larger events, such as whole-genome duplication. In tomato, the duplication and transposition of IQD12 lead to the novel up-regulation of its descendant locus, SUN, resulting in an elongated fruit shape (Xiao et al., 2008). Furthermore, overexpression of IQD12 in transgenic plants is sufficient to confer the elongated fruit shape, indicating that the expression change alone can explain the transition to a novel fruit shape. Thus, we concluded that differentiation of the gene expression pattern contributed to the neofunctionalization of Vrs1.
Vrs4 up-regulates Vrs1
Vrs1 expression is predicted to be regulated by additional six-rowed spike genes and we previously proposed that these genes act in a common network (Pourkheirandish & Komatsuda, 2007). Based on the drastic reduction in Vrs1 expression in the vrs4 mutant, we suggest that VRS4 works as a transcriptional activator to induce or enhance Vrs1 transcription in the lateral florets of wildtype two-rowed barley. This possibility is supported by the observation that most lateral florets in the vrs4 mutant are well developed and set grains (70% fertility in spike). The vrs4 mutant spikes showed two to four additional fertile spikelets at each rachis node, resulting in a total of five to seven fertile spikelets per rachis (Lundqvist et al., 1997; Forster et al., 2007). VRS4 likely suppresses the differentiation of these additional spikelets in the wildtype. This suppression of the additional spikelets may not be mediated by VRS1, because any vrs1 mutants alone were not sufficient to differentiate additional spikelets. Although Vrs1 expression was up-regulated by Vrs4, HvHox2 expression was not. These data also support the findings that Vrs1 is expressed at higher levels than HvHox2 and is localized to lateral florets. The regulatory region of Vrs1 was likely modified after the duplication of the ancestral HD-Zip I gene, such that Vrs4 now controls its expression.
We thank Yoshiaki Nagamura, Seiichi Toki, Hiroaki Saika and Naoki Sentoku (NIAS) and Shigeo Takumi (Kobe University) for their help and advice, and Daisuke Saisho (IPSR), Thorsten Schnurbusch (IPK) and Jun-Ichi Itoh (University of Tokyo) for useful discussions. We are grateful to Cornelia Marthe (IPK) for their excellent technical assistance. This research was funded by the Ministry of Agriculture, Forestry, and Fisheries of Japan (Genomics for Agricultural Innovation grant no. TRG1004 to T.K.) and the Japan Society for the Promotion of Science (Research Fellowship for Young Scientists to S.S.).