Evidence That the Origin of Naked Kernels During Maize Domestication Was Caused by a Single Amino Acid Substitution in tga1

Abstract

teosinte glume architecture1 (tga1), a member of the SBP-box gene family of transcriptional regulators, has been identified as the gene conferring naked kernels in maize vs. encased kernels in its wild progenitor, teosinte. However, the identity of the causative polymorphism within tga1 that produces these different phenotypes has remained unknown. Using nucleotide diversity data, we show that there is a single fixed nucleotide difference between maize and teosinte in tga1, and this difference confers a Lys (teosinte allele) to Asn (maize allele) substitution. This substitution transforms TGA1 into a transcriptional repressor. While both alleles of TGA1 can bind a GTAC motif, maize-TGA1 forms more stable dimers than teosinte-TGA1. Since it is the only fixed difference between maize and teosinte, this alteration in protein function likely underlies the differences in maize and teosinte glume architecture. We previously reported a difference in TGA1 protein abundance between maize and teosinte based on relative signal intensity of a Western blot. Here, we show that this signal difference is not due to tga1 but to a second gene, neighbor of tga1 (not1). Not1 encodes a protein that has 92% amino acid similarity to TGA1 and that is recognized by the TGA1 antibody. Genetic mapping and phenotypic data show that tga1, without a contribution from not1, controls the difference in covered vs. naked kernels. No trait differences could be associated with the maize vs. teosinte alleles of not1. Our results document how morphological evolution can be driven by a simple nucleotide change that alters protein function.

ALTHOUGH the study of adaptive evolution through natural or artificial selection dates back to Darwin, the genetic mechanisms that drive changes in morphology remain strongly debated (Stern and Orgogozo 2008). Questions surrounding the number of genes (few or many) and types of mutations (regulatory or coding) in these genes have been a focus in this debate (Carroll 2005, 2008; Hoekstra and Coyne 2007). These questions can only be answered through the genetic and molecular dissection of genes that have undergone selective pressure between lineages or within populations. Crop species offer a powerful system for investigating these questions since crops are the products of continuous directional selection to adapt them to the human controlled environment and human needs (Meyer and Purugganan 2013). Research on crop models is further facilitated by the extensive genetic and genomic resources available for them. Moreover, because of the recent divergence of crop species from their wild progenitors, crop-progenitor pairs remain cross-compatible and amenable to genetic analysis.

Maize was domesticated in the central Balsas valley of Mexico ∼9000 years ago from a wild relative called teosinte (Piperno et al. 2001; Matsuoka et al. 2002). Because the morphology of modern maize is drastically different from teosinte species, the progenitor of maize was highly debated until molecular evidence proved the ancestral link (Mangelsdorf and Reeves 1938; Beadle 1939; Doebley 2001). Remarkably, genes controlling much of the morphological difference between maize and teosinte were shown to map to just six regions of the genome or major domestication loci (Doebley et al. 1990). Subsequent studies have sought to elucidate the genetic nature of these loci. Fine mapping of two of these regions, located on opposite arms of chromosome 1, identified causative polymorphisms in the regulatory region of teosinte branched1 and grassy tillers1 (Studer et al. 2011; Wills et al. 2013). These regulatory changes lead to altered plant architecture, and population genetic data indicate that the regulatory regions of these two genes were targets of selection during domestication. Recently, the domestication locus on chromosome 5 was shown to contain multiple factors (genes) rather than a single large effect gene (Lemmon and Doebley 2014).

Another major maize domestication locus, which is located on chromosome 4, controls whether the grains are enclosed in a “fruitcase” as in teosinte or uncovered as in maize (Dorweiler et al. 1993). The fruitcase that encapsulates the teosinte grain is formed from a hardened cup-shaped stem segment (cupule) in which the grain is located and a hardened bract (glume) that seals the grain in the cupule. In maize, the grains are borne naked on the exterior of the ear, and the organs that form the fruitcase in teosinte are redeployed to form the internal central axis of the ear (the cob). The transition from encased to exposed grain greatly facilitated the use of the grain as food. The locus that largely controls this difference has been resolved to a single gene called teosinte glume architecture1 (tga1) (Wang et al. 2005). The maize allele of tga1 disrupts the normal development of the cupulate fruitcase, exposing the grain on the surface of the ear. tga1 encodes a squamosa-promoter binding protein (SBP), a transcription factor family that has been shown to regulate floral development (Klein et al. 1996; Wang et al. 2005). Although tga1 has been identified as the major gene controlling changes in fruitcase development during domestication, the causative polymorphism in tga1 and how this polymorphism affects the phenotype has not been resolved.

In this article, we show that a single fixed nucleotide polymorphism in the coding sequence of tga1 distinguishes the maize and teosinte alleles. This difference creates an amino acid substitution that changes TGA1 protein dimerization and alters how TGA1 regulates its targets, with the maize allele acting more as a repressor relative to the teosinte allele. We also describe the pleiotropic effects of RNAi lines for tga1 on multiple traits, indicating that tga1 may play a broad role in development, having effects on kernel size and shape as well as plant architecture. Finally, we describe another gene, neighbor of tga1 (not1), that is tightly linked to tga1 and has a high sequence homology with tga1, but for which we observed no differences in phenotypic effect between the maize and teosinte alleles. Our molecular and genetic analyses of tga1 show how a simple amino acid change can alter protein function and thereby drive the evolution of a new phenotypic state.

Materials and Methods

Detailed materials and methods can be found in Supporting Information.

Plant materials

Maize inbred W22 and W22:tga1, an introgression line that contains a teosinte chromosomal segment surrounding tga1 in a W22 background, were used in most experiments (Dorweiler and Doebley 1997). A set of recombinant lines (T249, T1214, T1464, and T2956) derived from a W22 × W22:tga1 F₂ fine mapping population that was previously described (Wang et al. 2005), were also utilized. The W22:tga1-ems line, previously reported in Wang et al. (2005), contains the amino acid substitution Leu5 to Phe5, which is immediately upstream of the single amino acid difference between maize- and teosinte-TGA1. The tga1-ems allele was recovered from an ethyl methanesulfonate mutagenesis and displays ear phenotypes similar to maize lines containing the teosinte allele of tga1. Additional plant materials include the not1-Mu2 stock from the Trait Utilization System for Corn (TUSC) collection and the transgenic tga1-RNAi lines made at the Plant Transformation Facility at Iowa State University.

Protoplast transient assays

Dual luciferase reporter assays were used to determine the repressor function of tga1 in transient protoplast expression experiments. N-terminal sequences of tga1 were fused to a GAL4-DNA binding domain (DBD) and cotransformed with a firefly luciferase gene downstream of two GAL4 binding sites. Firefly luciferase expression was measured and then normalized to a Renilla luciferase internal control. Maize mesophyll protoplasts used for the transient expression experiments were extracted and transformed using a protocol developed by the Jen Sheen laboratory (see Supporting Information).

Protein purification from plant tissue

Protein was extracted from ear tissue using a plant total protein extraction kit (Sigma, St. Louis, MO) following the manufacturer’s instructions. To show that TGA1 forms a dimer in vivo, formaldehyde was used to fix immature ear tissue prior to protein extraction.

Binding assays

Electrophoretic mobility shift assays (EMSAs) for PCR-assisted binding site selection experiments were performed as described previously (Tang and Perry 2003), and for experiments testing the in vitro binding of maize-TGA1 to the not1 promoter, performed as described previously with some modifications (Wang et al. 2004). Chromatin immunoprecipitation (ChIP) experiments with an anti-TGA1 polyclonal antibody were used in to verify that TGA1 bound the not1 promoter in vivo. ChIP assays were performed as described previously (Gendrel et al. 2002; Wang et al. 2002).

Results

Only one fixed nucleotide difference exists between maize and teosinte alleles of tga1

Wang et al. (2005) demonstrated by fine mapping that the causal polymorphism that distinguishes maize and teosinte lies within a 1042-bp segment of tga1 (GenBank: AY883436–AY883460), which includes the first 18 bp of the ORF and 1024 bp upstream of the ATG. With a small sample of maize and teosinte alleles, these authors observed seven nucleotide differences between maize and teosinte: six that are upstream of the start codon and one that is at position 18 of the ORF. To determine if these seven candidate polymorphisms could be narrowed to a smaller number, we assayed a larger sample (20) of teosinte alleles (Figure S1, GenBank: KR261098–KR261108). These data show that the six upstream polymorphisms no longer represent fixed differences between maize and teosinte but rather that some teosintes possess the same nucleotide as maize at each of these six sites. However, a nucleotide difference at position 18 of the ORF (C for maize and G for teosinte), which encodes a Lys6-to-Asn6 substitution from teosinte-TGA1 to maize-TGA1, still remains a fixed difference. Thus, this is the only fixed difference in the causative region that defines the glume architecture difference between maize and teosinte.

neighbor of tga1 is a tightly linked paralog of tga1

Examination of the genomic region near tga1 revealed a gene (AC233751.1_FG002) that shares high nucleotide similarity to tga1. This gene is located only ∼270 kb away from tga1 and thus it was named neighbor of tga1 (not1) (Preston et al. 2012). Comparing the TGA1 and NOT1 proteins from maize inbred W22, they have 92% identity in sequence. not1 also exists in teosinte and a sequence alignment of different alleles of tga1 and not1 is shown in Figure S2.

To investigate whether not1 contributes to the glume architecture difference between teosinte and maize, we investigated the glume phenotype of recombinant inbred lines (RILs) carrying different combinations of the tga1 and not1 alleles. RILs that carry tga1-teosinte alleles, such as T249 and W22:tga1, all show teosinte glume architecture regardless of whether they have the maize or teosinte allele at not1 (Table 1). Similarly, RILs that possess the tga1-maize allele, such as W22, T1464, and T2956, all show maize glume architecture regardless of whether they have the maize or teosinte allele at not1. Thus, the teosinte glume architecture phenotype is associated with the tga1-teosinte allele but not related to the not1-teosinte allele.

Table 1. Glume architecture phenotypes of isogenic lines with different genotypes at tga1 and not1.

Genotype	Allele at tga1	Allele at not1	Glume architecture trait
W22	maize	maize	Maize like
T249	teosinte	maize	Teosinte like
T1214	maize	teosinte	Maize like
T1464	maize	teosinte	Maize like
T2956	maize	teosinte	Maize like
W22: tga1	teosinte	teosinte	Teosinte like
W22:tga1-ems	ems	maize	Teosinte like

We investigated the tga1 and not1 gene expression across different lines with RT-qPCR. As shown in Figure 1A, there is no statistical difference in message accumulation between the maize and teosinte alleles of tga1 (ANOVA: P = 0.9467), and tga1 message level is not affected by the genotype at not1. However, the not1 message level is lowest when there are maize alleles at both not1 and tga1, highest when there are teosinte alleles at both genes, and intermediate for the heteroallelic genotypes. These results indicate that the maize allele of not1 accumulates less message than the teosinte allele, and they suggest that maize-TGA1 represses not1 expression compared to teosinte-TGA1.

Figure 1.

tga1 and not1 gene expression and protein accumulation. (A) RT-qPCR results showing tga1 (top panel) and not1 (bottom panel) expression. tga1 expression was not statistically different between genetic stocks (ANOVA: P = 0.9467, n = 10). not1 expresses less when tga1 is a maize allele. Maize-not1 expresses less in W22 than in T249 and EMS. Teosinte-not1 expresses less in T1464 than in tga1. (B) Anti-TGA1 recognizes both TGA1 and NOT1; however, TGA1 and NOT1 cannot be separated with a short gel run performed in duplicate. (C) Teosinte-NOT1 cannot be separated from TGA1, but maize-NOT1 can be resolved under extended gel running conditions. The allelic compliment of each genetic stock is indicated M, maize allele; t, teosinte allele; ems, ems allele.

An antibody was generated using the TGA1 C-terminal protein and was used for a Western blot as part of the initial characterization of tga1 (Wang et al. 2005). On the Western blot, the signal associated with protein samples for genetic stocks with the tga1-teosinte allele were stronger than for stocks with the tga1-maize allele (Wang et al. 2005). However, these stocks not only differ for their tga1 allele but they also differ for their alleles at not1. This situation raises the possibility that the difference in signal strength on the Western blot was due to not1 rather than tga1. If the anti-TGA1 antibody cross-reacts with NOT1, then the Western signal might be a combination of both TGA1 and NOT1.

To investigate this possibility, we performed a Western blot with proteins from comparable developmentally staged ears of our different RILs. The strongest signals were detected from the ears of the RILs that carry a not1-teosinte allele including W22:tga1 and T1464 (Figure 1B). By comparison, line T249, which contains a teosinte allele at tga1 but a maize allele at not1, has a signal intensity on the Western blot that is much less. In addition, the Western blot signals from W22:tga1-ems, which carries a tga1-ems allele and a not1-maize allele, do not show a dramatic difference in signal from W22. These observations suggest that strong signals in Western blots are associated with not1-teosinte instead of tga1-teosinte. Thus, the observation made by Wang et al. (2005) that the tga1-teosinte allele confers greater protein accumulation than the tga1-maize allele does not appear to be correct. Rather, the level of protein accumulation associated with these two alleles appears roughly equivalent.

We further investigated the TGA1 and NOT1 proteins by performing gel electrophoresis for an extended period of time to see if the TGA1 and NOT1 proteins could be separated (Figure 1C). In this analysis, we included an additional not1 allele with a Mu element insertion (not1-Mu2) for which RT-qPCR shows no evidence of a transcript (Figure S3). The RILs with not1-maize all show two protein bands corresponding to TGA1 and NOT1. The not1-Mu2 line is missing the lower of these two bands. Thus, the lower band appears to be NOT1. The RILs with not1-teosinte show a single thick band, which could be comigrating TGA1 and NOT1 proteins. Importantly, a comparison of RIL T249 and W22 shows that the band corresponding to NOT1 is stronger when tga1-teosinte is present than when tga1-maize is present. This observation is consistent with our interpretation of Figure 1B that the stronger protein signal for the W22:tga1 line represents NOT1 rather than TGA1. The combined RT-qPCR and Western data suggest that not1-teosinte is expressed higher than not1-maize and that tga1-maize represses not1 such that the greatest difference in not1 expression is seen between lines that are tga1-maize; not1-maize vs. tga1-teosinte; not1-teosinte (Figure 1, A–C). Again, the statement by Wang et al. (2005) that the tga1-teosinte allele confers greater protein accumulation than the tga1-maize allele does not appear to be correct.

Lys6-to-Asn6 substitution converted TGA1 N-terminal into a repression domain

Based on the maize-teosinte sequence comparison and lack of an expression difference between the maize and teosinte alleles of tga1, our working hypothesis is that the single amino acid substitution (Lys to Asn) at the sixth position of tga1 is the causative site for the loss of teosinte glume architecture during maize domestication. To test if the Lys6-to-Asn6 substitution controls the functional difference between maize- and teosinte-TGA1, we employed a protoplast transient assay system.

We constructed six effectors and two reporter constructs for the transient assays (Figure 2, A and B). We mixed plasmids with different effector and reporter combinations (Figure 2C) and introduced the mixture into maize mesophyll protoplasts by electroporation. After incubating the transfected protoplasts for 16–18 hr, the firefly luciferase activity was normalized to the Renilla luciferase activity for each assay. Transfection of the LD-VP16 effector along with the reporters gave strong firefly gene expression (Figure 2C); however, all other effectors (maize-GD, teosinte-GD, and ems-GD) along with the reporters in absence of LD-VP16 showed no activity in reporter gene expression relative to the negative control effector (GD) (Figure 2C, lightly shaded columns). These results suggest the TGA1 N-terminal domain is not an activation domain for any of the three alleles.

Figure 2.

Protoplast transient dual-luciferase assays show maize-TGA is a transcriptional repressor. (A) Diagram showing effector construct components. (B) Diagram showing reporter construct components. (C) Effectors 1–5 were cotransformed with both reporters into maize protoplasts with and without the activation control (effector 6). Dual-luciferase assay results are shown in relative light units (RLU), y-axis, (±SE; n = 6). Maize-TGA1 (effector 2) shows stronger repression than the repression control (effector 5).

Cotransfection of LD-VP16 with an effector encoding the GAL4-DBD fused to the IAA17(I) (positive control for repression) resulted in repressed firefly expression (Figure 2C). Cotransfection of LD-VP16 with the effector carrying the maize allele TGA1 N-terminal fusion also showed repressed firefly luciferase activity. Interestingly, cotransfection with the TGA1 N-terminal domain from teosinte or ems allele effectors showed similar reporter activities to the GD effector (negative control). These results indicate that the N-terminal domain of maize-TGA1 is an active repression domain, but the N-terminal of teosinte-TGA1 or ems-TGA1 are not active repressors. The results also suggest that the repression activity of TGA1 N-terminal domain is established by just a single naturally occurring amino acid substitution (Lys6 to Asn6) and it was reversed by the ems-induced substitution (Leu5 to Phe5).

PCR-assisted binding site selection

To assay whether the Lys6-to-Asn6 substitution affects the DNA binding activity of the protein and to better understanding TGA1 function, we performed PCR-assisted binding site selection to determine the binding site of TGA1. Using EMSAs, we observed that both maize- and teosinte-TGA1 can shift up two bands but their patterns are opposite (Figure 3). Maize-TGA1 shifted up a stronger upper band while teosinte-TGA1 shifted up a stronger lower band. However, sequence consensus of DNA fragments derived from these shifted bands showed that they all contain a GTAC motif. This is consistent with previous reports that SBP-domain proteins bind to GTAC (Birkenbihl et al. 2005; Kropat et al. 2005). These results indicate that the Lys6-to-Asn6 substitution did not affect the binding site specificity of TGA1 for the GTAC motif, but likely affected the configuration of TGA1 binding to DNA.

Figure 3.

Binding site selection for maize-TGA1 and teosinte-TGA1. α-³²P-labeled DNA fragments that contain 26 randomized nucleotides were incubated with maize or teostine-TGA1. After five rounds of EMSAs, the shifted DNA bands were excised, and the consensus binding motif was obtained. Both maize-TGA1 and teosinte-TGA1 proteins bind to DNA fragments with a GTAC core motif. For each allele, 50 ng, 100 ng, and 200 ng of TGA1 protein are shown (left to right).

TGA1 binds to the promoter of not1 in vitro and in vivo

We isolated the promoter segments of not1-maize and not1-teosinte alleles (Figure S4, GenBank: KR261109 and KR261110). Interestingly, both sequences share a conservative region that contains the GTAC motif. We took partial sequence surrounding this GTAC motif to make a probe for EMSA (Figure 4A). Our results show this sequence from the not1 promoter can bind maize-TGA1 in vitro (Figure 4B). Furthermore, the binding activity of maize-TGA1 to this DNA fragment was abolished totally with the GTAC motif mutated to CTAC (Figure 4B). These results suggest that TGA1 binds to the not1 promoter at the GTAC motif.

Figure 4.

TGA1 binds the not1 promoter in vitro and in vivo through the GTAC core motif. (A) The not1 promoter DNA probe sequence containing the GTAC motif. (B) EMSA showing that maize-TGA1 binds to the not1 promoter in vitro through the GTAC motif. The not1 probe shifts up two bands with maize-TGA1 protein, but the band shifting is abolished when the GTAC core is mutated to CTAC. (C and D) Chromatin immunoprecipitation (ChIP) confirms that TGA1 binds the not1 promoter in vivo. (C) RT-PCR amplification of the not1 promoter region containing the GTAC core motif. Input, total input chromatin DNA before precipitation; ChIP, chromatin DNA precipitated with anti-TGA1 antibody; PI, DNA precipitated with preimmune serum; mock, no antibody or serum added. (D) RT-qPCR showing the relative enrichment of the not1 promoter in the anti-TGA1 ChIPs (±SE; n = 7).

A ChIP assay using an anti-TGA1 antibody was performed to confirm the in vivo interaction between TGA1 and the not1 promoter. Because anti-TGA1 can also cross-react with NOT1, we used tissue from plants with the not1-Mu2 allele, which is a null mutant for not1. The ChIP-PCR results showed that the DNA fragment of the not1 promoter region was enriched in the anti-TGA1 ChIP population as compared to the input control (Figure 4C). Furthermore, ChIP populations generated using non- or preimmune serum did not show specific selection of the not1 fragment. To quantify the degree of enrichment, we also performed qPCR, which shows the not1 promoter region was enriched ∼4.8-fold in the anti-TGA1 population (ChIP = 4.68 ± 0.27; input control = 0.97 ± 0.11) (Figure 4D).

Maize-TGA1 forms more stable dimers than teosinte-TGA1

The DNA probes used for EMSA in Figure 3 and Figure 4 are different, but they both showed doublet bands. Thus, the double banding pattern is not likely related to the DNA probe sequences. To determine if the doublet was caused by protein dimerization, we performed additional EMSA using tagged and tag-free versions of TGA1 that are expected to migrate differentially during electrophoresis and thus be distinguishable. In addition to these two full-length versions of TGA1, we also used a shortened version of TGA1 with the N terminus removed.

As shown in Figure 5A, all the full-length TGA1 proteins, including tag-maize-TGA1, tag-teosinte-TGA1, tag-ems-TGA1, and maize-TGA1, shifted up two bands. The upper band is the major band when using tag-maize-TGA1 or maize-TGA1 while the lower band is the major band when using tag-teosinte-TGA1 or tag-ems-TGA1. These results suggest that TGA1 exists as dimers and monomers dynamically, and that maize-TGA1 tends to form more stable dimers.

Figure 5.

TGA1 forms a dimer in vitro and in vivo. (A) An EMSA showing that maize-TGA1 forms more stable dimers than teosinte- or ems-TGA1. Removing the N-terminal domain of TGA1 abolished the dimerization of TGA1, shown in lane3 and 4. A mixture of maize-TGA1 proteins with/without tags for EMSA confirmed TGA1 forms homo and heterodimers (lane11 and 12). (B) Protein from formaldehyde cross-linked ears produce two bands on a Western blot at ∼50 kDa and 100 kDa, while protein from unfixed ears only have the lower band.

In Figure 5A, one can also see that the DNA–protein complexes with tag-maize-TGA1 (version 1) run slower in the gel than the complex from tag-free maize-TGA1 (version 2). When we mixed tag-maize-TGA1 and tag-free maize-TGA1 together in the same binding reactions, we found that a novel intermediated band was detected. We interpret this new band as a heterodimer of tag-maize-TGA1/tag-free maize-TGA1 proteins. Intriguingly, a truncated TGA1 with 103 amino acids removed from the N terminus only shifts up a single band. These results indicate that TGA1 can form dimers and that the N terminus of TGA1 is necessary for dimerization. We conclude that the Lys6-to-Asn6 mutation, which is located in the TGA1 N terminus, can affect the dynamic ratio between dimers and monomers.

To investigate whether TGA1 forms dimers in vivo, we performed a Western blot assay using formaldehyde cross-linked ear tissue from not1-Mu2 plants. Formaldehyde can covalently preserve protein dimers in vivo, thus prevent the dimer disassociation during protein preparation and electrophoresis (Sang et al. 2005). As shown in Figure 5B, only a single band was detected at ∼50 kDa from nonfixed tissue, and this band is approximately the size of TGA1 as monomer. However, when using cross-linked tissue, an additional band, which is approximately twice the weight of the TGA1 monomer, was recognized by the anti-TGA1 antibody. These results suggest that maize-TGA1 forms homodimers in vivo, which is consistent with our in vitro assay data.

tga1/not1 loss-of-function plants via RNAi and their phenotype

To understand the function of tga1, we screened the available maize mutant collections for tga1 loss-of-function alleles without success (see Materials and Methods). We did not recover any mutant alleles for tga1; however, we found 5 Mu insertions in not1. The not1-Mu2 allele is a null mutation lacking detectable expression (Figure S3), although we did not observe any phenotypic difference between homozygous not1-Mu2 plants and wild-type sibs in a segregating population.

We attempted to generate transgenic tga1 overexpression plants by putting tga1-maize or tga1-teosinte under a constitutive rice Actin1 promoter. Surprisingly, no tga1 overexpression transgenic plants could be generated with either the maize or teosinte allele. tga1 contains a miRNA156 target site, and thus miRNA156 could inhibit the overexpression. To rule out this possibility, we built two new constructs with the miRNA156 site in tga1 mutated synonymously and placed them under a maize Ubiqitin promoter. However, we still could not produce transgenic plants and there was difficulty even obtaining transgenic callus with these vectors. Other control constructs that were transformed side by side worked fine as reported to us by Dr. Kan Wang at the Plant Transformation Facility, Iowa State University. These results suggest that the overexpression of tga1 inhibits some aspect of plant development and thus plant regeneration during the transformation process.

We next used an RNAi-based approach to generate tga1 loss-of-function plants. Because RNAi is achieved by overexpression of a hairpin structure (Kusaba 2004), it is likely not efficient enough to generate a complete knockout of tga1 mRNA, especially during plant regeneration. Using this approach, we were able to recover tga1-RNAi transgenic plants. Five of the tga1-RNAi events were grown to maturity and crossed to maize inbred W22. Then, 60 progeny plants from each cross were grown and the segregation of transgene was identified by a Basta leaf painting assay. Progeny groups for four events showed an approximate 1:1 ratio for Basta resistance:susceptible plants, suggesting that these four events have single insertions.

As shown in Figure 6, tga1-RNAi plants present some interesting phenotypes. The tga1/not1 loss-of-function plants have much longer lateral branches (Figure 6, A–C). A few lateral branches even have secondary ears developed along them (Figure 6C). The glumes of tga1-RNAi plants are enlarged, which is a characterization of W22:tga1 glumes (Figure 6, F, G, J, and K). However, these glumes are relatively paperish and less thick, less hard, and less polished. The kernels from the tga1-RNAi plants are narrower (Figure 6I) compared to control kernels (Figure 6H). Furthermore, they have a pointed tip, which makes the ear “prickly” (Figure 6M), while the control ear is smooth (Figure 6L). Interestingly, the tga1-RNAi plants have prop roots developed at the first four to six nodes of the stalk, while the control plants only have two nodes with prop roots (Figure 6, D and E). This collection of phenotypes suggests that tga1/not1 not only have function in glume architecture, but also function in juvenile growth, lateral branch formation, and ear development. The fact that the tga1-RNAi plants display some characteristics of the tga1-teosinte glume phenotype suggests that the Lys6 to Asn6 from teosinte to maize is a gain-of-function mutation.

Figure 6.

Phenotypes of tga-RNAi plants. (A–C) tga1-RNAi plants have longer lateral branches. (A) A nontransgenic control plant with a short lateral branch. (B) A typical tga1-RNAi plant from event 4 with a long lateral branch. (C) Lateral branches with leaves removed; top one is from nontransgenic control; the others are from tga1-RNAi plants, which have long shanks and some even have secondary ears. (D and E) A typical tga1-RNAi plant (E) has prop roots on more nodes than a control plant (D). (F–M) The ear and kernel phenotypes. G, I, K, and M are from tga1-RNAi. F, H, J, and L are controls. Compared to the control, tga1-RNAi ears have enlarged glumes (G and K) and narrow shaped seeds (I). Furthermore, each tga1-RNAi kernel is pointed (M) on the surface, which results in a spiky ear.

Quantitative traits associated with tga1-RNAi

Statistically significant associations were identified between the tga1-RNAi transgene and some quantitative traits related to plant and ear architecture. We compared plants with and without the tga1-RNAi transgene as determined by a Basta painting assay. We analyzed 30 resistance (TG = transgene present) and 30 susceptible (NTG = transgene absent) plants each from segregating F₂ families for four events. Associations between the phenotypes and tga1-RNAi transgene genotype were tested using T-tests.

As shown in Figure 7, the tga1-RNAi transgene is associated with statistically significant effects on lateral branch number, length of the uppermost lateral branch, the length of the blade of the first husk leaf, and the number of nodes with prop roots. The results suggest that TGA1/NOT1 represses the growth of lateral branches in length and numbers. The association of some ear traits with tga1-RNAi was also observed (Figure 7). Glume length was significantly increased when TGA1/NOT1 were knocked down/out. In contrast, ear diameter and the weight of 50 kernels decreased in tga1-RNAi plants. We also measured some additional ear traits, such as ear length and kernel row number; however, there were no statistically significant associations between these traits and the tga1-RNAi transgene.

Figure 7.

Quantitative effects associated with the tga1-RNAi lines. Trait values were plotted on the y-axis and dots represent the value range for each phenotype. Diamonds are centered on the mean values for the traits. Data from four events are presented with TG signifying plants that contain the tga1-RNAi transgene and NTG for nontransgenic plants. The P-value for each T-test is displayed in the upper right corner of each graph.

In summary, these results suggest that TGA1/NOT1 have broader functions beyond controlling glume architecture in maize and teosinte. Specifically, they may also affect traits such as seed shape, seed weight, ear and lateral branch morphology, as well as the juvenile seedling phase of plant growth.

Discussion

A previous population genetic analysis of tga1 identified seven fixed nucleotide differences between maize and teosinte using a small sample of accessions (Wang et al. 2005). Here, we used a larger sample size of maize and teosinte, which revealed that six of these seven nucleotide sites are polymorphic within either maize or teosinte (Figure S1). With this larger sample, only a single fixed difference was identified between maize and teosinte within tga1. This single site encodes an Lys6-to-Asn6 substitution between teosinte and maize at the sixth amino acid position from the N terminus of the protein. This observation makes the Lys6-to-Asn6 substitution the most likely candidate for the causative difference between maize and teosinte in tga1.

We tested the functional consequences of the Lys6-to-Asn6 substitution using protoplast transient assays (Tiwari et al. 2004). These assays revealed that this single amino acid substitution alters the function of the TGA1 protein. The Lys6-to-Asn6 substitution gives the maize-TGA1 a strong repressor function as seen by reporter expression levels that are comparable to the repressor control (Figure 2). Furthermore, this amino acid substitution also increases dimerization of TGA1, as shown by gel shift assays (Figure 5). The increased dimer formation and repressor function of the maize-TGA1 are likely concomitants. Arguably, the most well-characterized repressor, the lac repressor (LacR) in Escherichia coli, forms a stable dimer of dimers, which then produces a functional tetramer through weak associations (reviewed in Lewis 2005). Amino acid substitutions that weaken dimer formation of LacR eliminate repressor function. Furthermore, a single amino acid substitution has been shown to block dimerization (Dong et al. 1999; Spott and Dong 2000). Thus, the observed correlation of dimerization and repressor strength reported here is consistent with functional studies in other systems.

The binding site of TGA1 was determined using EMSA and shown to be GTAC (Figure 3). Both the maize- and teosinte-TGA1 were used; however, no difference in binding site preference was observed. Thus, the amino acid substitution that leads to repressor function is likely the mechanism by which plant morphology is altered, not differences in binding site specificity. The tga1 binding motif was found in the promoter of not1 and TGA1 binds this site both in vitro and in vivo, suggesting that not1 and tga1 may function in the same pathway. This result begins to unravel a potential cascade of gene expression changes that accompanies the alteration of a major domestication gene.

tga1 was discovered based on the specific effects of the maize vs. teosinte allele on the development of the teosinte fruitcase and maize cob (Dorweiler et al. 1993). Although these alleles differ for the Lys6-to-Asn6 amino acid substitution, both alleles are expressed at comparable levels (Wang et al. 2005). We assayed the broader effects of tga1 by using an RNAi construct to reduce or eliminate tga1 gene expression. Maize lines expressing an RNAi construct targeting tga1 displayed pleiotropic morphological effects on several branching and kernel traits (Figure 6 and Figure 7), which had not previously been associated with this domestication gene (most recently Brown et al. 2011). With regard to branching, these RNAi lines likely remove the repressive function of TGA1/NOT1, allowing the outgrowth of axillary branches. The effects on kernel shape and size may be related to the fact that the kernel resides within the fruitcase in teosinte, and thus, fruitcase and kernel development are coordinately regulated by tga1. Whether the maize vs. teosinte allele of tga1 affects any of these additional traits is unknown, but such effects have not previously been reported (Dorweiler and Doebley 1997).

An open question is whether the effects of the RNAi construct on kernel and branching traits result from a knockdown of tga1 or not1. The RNAi construct was generated using tga1 sequence, but given the sequence similarity between tga1 and not1, it is likely that the RNAi construct targets both genes. However, given that the Mu insertions in not1 did not produce the morphological phenotypes seen in the RNAi lines, we infer that the phenotypes observed in the RNAi lines are either attributable to tga1 or a redundant function of tga1 and not1 and thus can only be observed when both are knocked down. However, without a null tga1 allele, we cannot show conclusively that these phenotypes are specific to tga1. These inferences about tga1 vs. not1 function are further complicated by the binding of TGA1 to the not1 promoter. While multiple morphological phenotypes are observed in the RNAi lines, the single amino acid substitution fixed during domestication seems to be specific to the ear traits reported previously (Wang et al. 2005).

Both tga1 and not1 belong to the SBP family of transcription factors. Members of this family have been shown to regulate meristem development, and manipulations of these regulators have produced both plant architecture and ear phenotypes (Chuck et al. 2010, 2014). While there is a clear homolog for tga1 in other grass species, the tga1/not1 duplication occurred at the base of the Zea genus, evident by the absence of this duplication in other lineages (Preston et al. 2012). The presence of the tga1/not1 duplication in maize may have facilitated the subfunctionalization of tga1/not1 such that tga1 alone controls the fruitcase/cob in teosinte/maize while tga1 functions in a redundant manner with not1 to regulate plant architecture traits. This would explain the phenotypes observed in the RNAi lines, which likely target both tga1/not1, when no phenotype was present in not1 mutant plants. Furthermore, various tga1 alleles have only been reported to display ear phenotypes (Wang et al. 2005; Brown et al. 2011). This hypothesis is further supported by work on the ortholog of tga1 in rice (LOC_Os08g41940; OsSPL16), which has pleiotropic plant and inflorescence phenotypes (Wang et al. 2011, 2012).

In this article, we have shown that an amino acid substitution in tga1 is the causal variant that underlies the origin of the naked grains of maize as compared to the covered grains of teosinte. Although the predominant mechanism for morphological evolution may be alterations in gene expression (Carroll 2008), changes in protein function are also involved as shown here. Investigation of how protein evolution contributes to the evolution of new morphological forms enhance our understanding of how adaptions arise.