ZmGA3ox2, a candidate gene for a major QTL, qPH3.1, for plant height in maize


For correspondence (email


Maize plant height is closely associated with biomass, lodging resistance and grain yield. Determining the genetic basis of plant height by characterizing and cloning plant height genes will guide the genetic improvement of crops. In this study, a quantitative trait locus (QTL) for plant height, qPH3.1, was identified on chromosome 3 using populations derived from a cross between Zong3 and its chromosome segment substitution line, SL15. The plant height of the two lines was obviously different, and application of exogenous gibberellin A3 removed this difference. QTL mapping placed qPH3.1 within a 4.0 cM interval, explaining 32.3% of the phenotypic variance. Furthermore, eight homozygous segmental isolines (SILs) developed from two larger F2 populations further narrowed down qPH3.1 to within a 12.6 kb interval. ZmGA3ox2, an ortholog of OsGA3ox2, which encodes a GA3 β-hydroxylase, was positionally cloned. Association mapping identified two polymorphisms in ZmGA3ox2 that were significantly associated with plant height across two experiments. Quantitative RT-PCR showed that SL15 had higher ZmGA3ox2 expression relative to Zong3. The resultant higher GA1 accumulation led to longer internodes in SL15 because of increased cell lengths. Moreover, a large deletion in the coding region of ZmGA3ox2 is responsible for the dwarf mutant d1-6016. The successfully isolated qPH3.1 enriches our knowledge on the genetic basis of plant height in maize, and provides an opportunity for improvement of plant architecture in maize breeding.


During the past century, the increase in maize yield per unit area has mainly resulted from increased planting density, rather than an improvement in yield potential per plant (Duvick et al., 2004). However, high-density planting also involves the risk of lodging, which may reduce grain yield. To avoid lodging, one breeding strategy is to moderately decrease the plant height of hybrids (Tang et al., 2007). Plant height (PHT) is an important yield-associated trait in maize that has been frequently selected during crop domestication and in modern breeding programs. For example the Green Revolution gene, sd1, which is involved in the biosynthesis of gibberellin (GA) (Monna et al., 2002; Sasaki et al., 2002; Spielmeyer et al., 2002), and Rht, which results in defective GA signaling (Peng et al., 1999). When sd1 was introduced in rice and Rht was inserted in wheat, a reduction in PHT was observed (Sasaki et al., 2002; Peng et al., 1999).

The existence of dwarf and semi-dwarf mutants, and the variation in PHT, provide opportunities for improving plant architecture by crop breeding, and permit cloning of PHT-related genes. A few maize PHT genes have been cloned, such as Dwarf3 (D3), which encodes a cytochrome P450 involved in the early stages of GA biosynthesis (Winkler and Helentjaris, 1995), and Anther ear1 (An1), encoding an enzyme involved in the synthesis of ent-kaurene, the first tetracyclic intermediate in GA biosynthesis (Bensen et al., 1995). These findings indicate that GA biosynthesis and GA signaling play a key role in the control of plant height. In fact, many dwarf mutants respond to bioactive GAs (Phinney, 1956; Koornneef et al., 1990; Sakamoto et al., 2004). However, the PHT variation in maize breeding populations is mostly controlled by a set of quantitative trait loci (QTLs) with minor effects. Over the last two decades, genetic dissection of maize PHT by classical QTL mapping using biparental populations has resulted in identification of numerous PHT QTLs (Beavis et al., 1991, 1994; Berke and Rocheford, 1995; Austin and Lee, 1996; Khairallah et al., 1998; Tang et al., 2007; Zhang et al., 2011). Compared with biparental mapping populations, an introgression library has a higher QTL-detecting power because of the absence of other segregating QTLs: the target QTL becomes the major source of genetic variation (Alonso-Blanco and Koornneef, 2000). Recently, a maize introgression library was produced using Gaspé Flint as the donor and B73 as the recipient. Gaspé Flint showed a much lower number of nodes, days to pollen shed and PHT than B73. Four PHT QTLs were detected as major QTLs for node number in the introgression library, indicating that node variation tightly drives variation in PHT (Salvi et al., 2011). Similarly, Bai et al. (2010) detected nine PHT QTLs using an introgression library. A genome-wide association study was also used to identify candidate genes that affect PHT in maize (Weng et al., 2011).

Map-based cloning has been used to isolate genes underlying QTLs in several plant species, including tomato (Frary et al., 2000), rice (Yano et al., 2000; Fan et al., 2006; Xue et al., 2008) and maize (Wang et al., 2005; Salvi et al., 2007). Recently, a QTL for PHT was identified on chromosome 3S in our laboratory using an introgression library (Z3HBIL), and was designated qPH3.1 (Wang et al., 2007; Bai et al., 2010). The QTL for PHT was repeatedly detected in previous studies (Beavis et al., 1991; Veldboom et al., 1994), and was suggested to be associated with dwarf-1 (d1) in diverse populations. Multiple QTLs for PHT were also found within bin 3.03–3.04 of the maize genome (Schön et al., 1993; Ajmone-Marsan et al., 1994; Beavis et al., 1994; Berke and Rocheford, 1995; Kozumplik et al., 1996; Melchinger et al., 1998). In the present study, qPH3.1 was finely mapped by linkage mapping and chromosomal segment substitution mapping, as described by Paterson et al. (1990). A candidate gene for qPH3.1, encoding a GA3 β-hydroxylase, was cloned and genetically validated by candidate gene association mapping and by a genetic assay in the dwarf mutant d1-6016.


Phenotypic evaluation of SL15 and Zong3

The two parents used to construct introgression library Z3HBIL, Zong3 and HB522, showed distinct differences with respect to PHT. The PHTs for Zong3 and HB522 were 188.9 and 80.1 cm, respectively, in 2007, and 206.8 and 82.2 cm, respectively, in 2008 (Bai et al., 2010). Although HB522 is a short-stature parent, SL15, which contained the introgression segment flanked by ND1 and ND75 from HB522, was significantly taller than Zong3 (203.9 ± 10.8 cm versus 179.7 ± 10.5 cm for SL15 versus Zong3, respectively; Figure 1a,b). Both SL15 and Zong3 have the same number of elongated internodes (six above the ear and seven below the ear), indicating that the observed PHT difference between the two lines is caused by differences in the length of the internodes, rather than the number of internodes. We labeled the internodes under the ear from −1 to −7, and those above the ear from 1 to 6, as shown in Figure 1(a). All internodes of SL15 were markedly longer than those of Zong3, except for the tassel (Figure 1c). Cytological observation revealed that internode cells in SL15 were significantly longer than those of Zong3 (P < 0.01) (Figure 1f), indicating that the longer internodes in SL15 relative to Zong3 are caused by increased cell lengths, not increased cell numbers.

Figure 1.

Plant height and longitudinal sections of the stem at the mature stage.(a–c) Plant height of Zong3 and SL15. (a, b) Plant height of Zong3 and SL15 at the mature stage. The nodes are indicated by white arrows. Scale bar = 50 cm. (c) Comparison of the length of the tassel and internodes of Zong3 and SL15. Asterisks indicate significant differences between Zong3 and SL15 (*< 0.05; ***< 0.001; Student's t test). Values are means ± SD (nZong3 = 48, nSL15 = 63). T, tassel. The internodes above the ear are labeled 1, 2, 3, 4, 5 and 6. The internodes below the ear are labeled −1, −2, −3, −4, −5, −6 and −7. (d, e) Hand-cut longitudinal sections of the stem of Zong3 and SL15, respectively, at the mature stage. Scale bar = 500 μm. (f) Length of the internode cell in Zong3 and SL15. Six slides of each sample were observed, and more than 100 cells were measured for each slide. Values are means ± SD (> 600). Student's t test was used to calculate the P value.

Identification of qPH3.1

To isolate the gene for qPH3.1, 161 F2 individuals were developed by crossing SL15 with Zong3. Each individual was genotyped using eight simple sequence repeat (SSR) markers (two public and six newly developed markers) in the target region flanked by ND1 and ND75. One-way anova was used to test the significance of the phenotypic differences for each marker among the three genotypic classes (the Zong3 homozygote, the heterozygote and SL15 homozygote). The phenotypic distributions of the three genotypes defined by the marker umc2369, which was significantly associated with PHT, are shown in Figure S1. The mean PHT resulting from the homozygous Zong3 allele (172.3 cm) differed significantly from that of the heterozygous allele (186.4 cm) and the SL15 homozygous allele (190.9 cm). Similarly, the mean PHT for the heterozygote differed significantly from that of the SL15 homozygote (< 0.05), indicating that qPH3.1 may be partially dominant.

QTL mapping revealed a QTL for PHT in the target region, which was narrowed down to a 4.0 cM interval, flanked by ND51 and ND35. The additive and dominant effects of the SL15 allele were 9.6 and 4.5 cm, respectively (Table 1). To identify the effects of the residual segment from HB522 on PHT, 245 genome-wide SSR markers showing polymorphism between Zong3 and HB522 were used to screen SL15. Only one introgression segment was detected, near dupssr23 on bin 3.06, and the segment was not markedly associated with PHT. However, qPH3.1 accounted for only 32.3% of the phenotypic variance, which may be attributed to the high environmental sensitivity of PHT, and the effects on PHT of those residual loci that were not found by SSR screening.

Table 1. Detection of qPH3.1 in the F2 population in three experiments
Year n Flanking markersLODADD/APlant height (cm)aR2 (%)
  1. n, population size; LOD, log10 likelihood ratio; A, additive effect; D, dominant effect; ZZ, Zong3 allelic homozygote; SS, SL15 allelic homozygote; ZS, Zong3/SL15 allelic heterozygote. The three classes of genotypes were defined by the marker umc2369. R2, percentage of phenotypic variance explained by a QTL.

  2. a

    Values are means ± SD.

2007161ND51–ND3512. ± 10.7186.4 ± 11.1190.9 ± 9.232.3
2008617ND51–ND3556.310.84.40.41192.2 ± 8.7207.5 ± 10.5213.9 ± 9.336.6
20092153umc2369–ND37167. ± 11.8150.3 ± 10.8156.6 ± 10.330.0

The effects of qPH3.1 on yield-associated traits, including 100-kernel weight, number of rows of kernels, kernels per row, ear length and ear diameter, were evaluated. No QTL for yield-associated traits was identified in the qPH3.1 region. Phenotypic differences of yield-associated traits among the three genotypes of qPH3.1 were not significant (Table S1), indicating that qPH3.1 controlled only the PHT, rather than yield-associated traits.

Fine mapping of qPH3.1

To fine map qPH3.1, two larger F2 populations, containing 617 and 2153 individuals, were developed in 2008 and 2009, respectively. The two populations were also genotyped using the newly developed SSR markers in the qPH3.1 interval. Previous works suggested that the well-known flowering time QTL, Vgt1, in maize also affects the PHT (Salvi et al., 2002, 2007, 2011). To determine whether qPH3.1 affects flowering time, the effects of qPH3.1 on flowering time traits were evaluated in 2008. The results showed that qPH3.1 was not associated with grain yield and flowering time traits (Table S1). In 2008, qPH3.1 was narrowed down to a 3.3 cM interval (ND51–ND35) (Figure 2a) with a 10.8 cm additive effect and a 4.4 cm dominant effect, which explained 36.6% of the phenotypic variance (Table 1). Similarly, the QTL was further narrowed down to a 2.1 cM interval (umc2369–ND37), with a 10.0 cm additive effect and a 3.7 cm dominant effect in 2009, explaining approximately 30.0% of the phenotypic variance (Table 1). Although the mean PHT showed distinct differences across the three experiments (2007, 2008 and 2009), qPH3.1 was still consistently detected in the three F2 populations, indicating that qPH3.1 is stably expressed in the three populations.

Figure 2.

Map-based cloning of qPH3.1.(a) Location of qPH3.1 on maize chromosome 3, mapped by a F2 population in 2008.(b) Genotypes and phenotypes of the parental lines (Zong3 and SL15) and eight homozygous segmental isolines carrying crossovers around qPH3.1. Lines 670, 348 and 427 were developed from the 2008 F2 population, and the other five homozygous segmental isolines were developed from the 2009 F2 population. Values are means ± SD. n, number of evaluated individuals in each homozygous segmental isoline. Asterisks indicate that the mean phenotypic value of the recombinant is significantly different from that of Zong3 at the < 0.01 level. NS, no significant difference at the < 0.01 level.

In the qPH3.1 interval umc2369–ND37, five new markers were developed and used to identify recombinant genotypes for development of segmental isolines (SILs). Eight homozygous SILs (three from the 2008 F2 population and five from the 2009 F2 population) were developed to represent the probability of single crossover among markers in the QTL interval. The PHT phenotypes of these homozygous SILs were compared with that of Zong3 at < 0.01 level. Recombinants 670, 348, 427, 48-6, 26-6 and 20-5 placed qPH3.1 in a region between ND72 and ND37. The most informative recombinants were 6-2 and 44-3. Two recombinant breakpoints were observed in the ND87–ND88 interval. The homozygous SILs containing SL15 alleles in the ND87–ND88 interval showed significantly greater PHT than Zong3, indicating that (i) the ND87–ND88 interval may contain the gene for qPH3.1, and (ii) the allele from SL15 increases the PHT, while the allele from Zong3 decreases the PHT. For the B73 genome, the ND87–ND88 interval is located on a single bacterial artificial chromosome (BAC) clone, c0160G03, and extends to approximately 12.6 kb (Figure 2b). This sequence has only one ORF (GRMZM2G036340), encoding GA3 β-hydroxylase (hereafter designated as ZmGA3ox2).

Polymorphisms at the ZmGA3ox2 gene

The genomic DNA sequences of ZmGA3ox2 were amplified by PCR. 5′ and 3′ RNA ligase-mediated rapid amplification of cDNA ends (RLM-RACE) were performed to amplify the ZmGA3ox2 cDNA ends. ZmGA3ox2 contains three exons and encodes a 1892 bp full-length cDNA with a 238 bp 5′ UTR and a 505 bp 3′ UTR (Figure S2a). The deduced protein contains 382 amino acids that form a 2OG-Fe (II) oxygenase functional domain from the 207th to 308th amino acid residue. Sequence analysis of a 4778 bp ZmGA3ox2 sequence that included 2736 bp of sequence upstream of the start codon, a 1634 bp coding region and 408 bp of the 3′ UTR (reference B73 sequence) revealed 18 SNPs and nine insertion/deletion polymorphisms (indels) between Zong3 and SL15. Among these, one SNP in exon 1 results in a substitution at the 61st amino acid (alanine to proline for Zong3 to SL15, Figure S2a), which is in a small helix of the N-terminus (Figure S2b). However, this SNP does not affect the 2OG-Fe (II) oxygenase functional domain or the secondary structure of the protein (Figure S2b), suggesting that the substitution may not affect the protein's function. In addition, several deletions in Zong3 relative to SL15 were detected in the promoter and 5′ UTR. Similarly, two deletions in SL15 relative to Zong3 were found in the promoter region, and three polymorphisms were found in the introns and the 3′ UTR region (Figure S2a).

ZmGA3ox2 is the gene responsible for the dwarf-1 mutant phenotype

We propose that ZmGA3ox2 is the candidate gene for the maize mutant dwarf-1 (d1) for the following reasons: (i) d1 responds to exogenous GA (Phinney, 1956), indicating that it is involved in the GA biosynthesis pathway, (ii) it fails to control the 3-hydroxylation of GA20 (Spray et al., 1984, 1996), and (iii) d1 is located on chromosome 3S (Scanlon et al., 1994). To test this hypothesis, the ZmGA3ox2 was amplified from one d1 mutant (d1-6016) (Figure 3a) using ZmGA3ox2-specific primers. The PCR product of d1-6016 was shorter than that from wild-type B73 (Figure 3b). Sequencing revealed a 2304 bp deletion in ZmGA3ox2 of the d1-6016, including 734 bp of the 5′ upstream sequence and 1570 bp of the coding sequence (with reference to the B73 sequence, Figure 3c). In particular, allelic variation determined by the ZmGA3ox2 gene-specific marker ND88 co-segregated with the PHT phenotype in a backcross population containing 406 individuals (Figure 3d). Furthermore, the 200 kb regions flanking ZmGA3ox2 were analyzed, but no other GA biosynthesis genes were found. The experimental evidence from the co-segregation analysis and sequence variation in the d1-6016 allele supported the hypothesis that ZmGA3ox2 is the candidate gene for d1-6016, and functions to control the PHT.

Figure 3.

Mutation of the d1-6016 allele.(a) Heterozygous +/d1-6016 and homozygous d1-6016/d1-6016 mutants.(b) PCR amplification of ZmGA3ox2 in B73 and the d1-6016 homozygote using primers 4046F and 7762R.(c) The structure of ZmGA3ox2. The gray and open boxes indicate exons and UTRs, respectively. The region deleted in the d1-6016 allele is indicated by dashed lines. Arrows indicate the position of primers 4046F and 7762R. The triangle indicates the position of ND88.(d) Co-segregation between ND88 and plant height in a back-cross population. +/d1-6016 is the heterozygous mutant, which shows the wild-type phenotype, while d1-6016/d1-6016 is the homozygous mutant, which has a dwarf phenotype.

ZmGA3ox2 expression and GA measurement

GA 3-oxidase catalyzes the final steps in formation of bioactive GAs (GA1 and GA4) (Olszewski et al., 2002). In the rice genome, there are two GA 3-oxidase-encoding genes (OsGA3ox1 and OsGA3ox2). Previous works suggest that OsGA3ox1 is specifically expressed in the inflorescence, while OsGA3ox2 is constitutively expressed in all organs to produce GA1, a dominant bioactive GA in the vegetative organs of rice (Itoh et al., 2001; Kaneko et al., 2003). To determine the number of genes encoding GA 3-oxidase in the maize genome, the nucleotide sequence encoding the 2OG-Fe (II) oxygenase functional domain of ZmGA3ox2 was used to search the B73 RefGen_v2 ( sequence using BLAST (Altschul et al., 1990). As a result, another gene (GRMZM2G044358) on bin 6.05 was identified as an ortholog of OsGA3ox1 (hereafter designated ZmGA3ox1), while ZmGA3ox2 is an ortholog of OsGA3ox2 (Figure S2c). Semi-quantitative RT-PCR revealed that ZmGA3ox1 is specifically expressed in the immature tassel of B73, with a similar expression pattern to OsGA3ox1 in rice. ZmGA3ox2 was constitutively expressed in all tissues tested, with similar expression pattern to OsGA3ox2 in rice (Figure 4a). These results indicate that ZmGA3ox2 is a GA 3-oxidase that functions in the vegetative organs, and possibly regulates production of GA1 like OsGA3ox2 in rice. In the d1-6016 mutant, ZmGA3ox2 was expected to be non-functional because of deletion of almost the entire coding region. If the maize genome has no other copy of ZmGA3ox2 or other GA 3-oxidase-encoding genes to compensate for the functional loss of ZmGA3ox2, the d1-6016 mutant will be unable to produce a functional ZmGA3ox2 enzyme. Bioinformatics assays indicated that the maize genome contains only one copy of ZmGA3ox1 and ZmGA3ox2. Expression analysis indicated that normal ZmGA3ox1 transcripts and mutant ZmGA3ox2 transcripts were present in the immature tassel of d1-6016 (Figure 4a). ZmGA3ox1 was not expressed in the roots, stem and immature leaf, but was normally expressed in the immature tassel. However, its expression did not rescue the dwarf phenotype of d1-6016. This finding indicates that ZmGA3ox1 cannot compensate for the functional loss of ZmGA3ox2, and further highlights the crucial role of ZmGA3ox2 during vegetative development. Stem apices are one of the actively growing and elongating tissues in which GAs are mainly produced (Kobayashi et al., 1988). Therefore, the stem apices were collected to determine expression of ZmGA3ox2 and measure the GA1 levels. Quantitative PCR showed that significantly higher levels of ZmGA3ox2 were expressed in the stem apices in SL15 than in Zong3 (Figure 4b). This is consistent with the greater endogenous GA1 levels in SL15 relative to Zong3 (Figure 4c).

Figure 4.

Expression assay, endogenous GA level, and exogenous GA application.(a) Expression of ZmGA3ox1 and ZmGA3ox2 in various organs and tissues. Total RNA was isolated from the crown root (CR), primary root (PR), seminary root (SR), lateral root (LR), mesocotyl (CM), shoot (SH), stem (ST), immature leaf (IL), immature ear (IE) and immature tassel (IT) of B73, and immature tassel of the d1-6016 (IT-d1).(b) Quantitative RT-PCR of ZmGA3ox2 at four different stages. Tissues from stem apices of Zong3 and SL15 plants were collected at the four stages, starting with jointing (10-leaf stage) and then the 12-leaf, 14-leaf and 16-leaf stages. The PCR signals were normalized to those of actin (NM_001155179) transcripts. Transcript levels from Zong3 were set to 1 at every stage. Three biological replicates were used for each sample. Values are means ± SD. Student's t test was used to calculate the P values.(c) Endogenous GA1 level in the stem apices at four growth stages. Values are means ± SD. Three replicates were used for each sample. Student's t test was used to generate the P values.(d) Effect of exogenous GA on the plant height of Zong3 and SL15. Plants were grown in the greenhouse, and their heights were measured each week from 2 weeks after emergence. Thirteen to 17 individuals were measured for each group. The plant heights are significantly different at the < 0.01 level where the error bars do not overlap. Zong3 (−), SL15 (−), Zong3 (+) and SL15 (+) represent untreated Zong3, untreated SL15, treated Zong3 and treated SL15, respectively.

Exogenous GA3 treatment

The above-described results indicate that ZmGA3ox2 may be the candidate gene for qPH3.1, and the phenotypic difference between Zong3 and SL15 was ascribed to a difference in accumulation of ZmGA3ox2 mRNA, which in turn results in a difference in the GA level. If that conclusion is correct, exogenous GA should eliminate the phenotypic difference between Zong3 and SL15. Both Zong3 and SL15 respond to exogenous GA3 applied by continuous spraying, and the PHT of the treated individuals was significantly greater than that of the control. The PHT difference between SL15 and Zong3 started from the 7th week after emergence, and increased over plant development. Notably, no significant difference between treated SL15 and treated Zong3 was found during the entire growth period under greenhouse conditions (Figure 4d). These results strongly demonstrate that the PHT difference between SL15 and Zong3 results from the different GA levels in the two lines.

Association mapping for the ZmGA3ox2 gene

To determine the causative sites responsible for the PHT difference between SL15 and Zong3, allelic variation of the 27 sequence polymorphisms identified in SL15 and Zong3 was exclusively tested in 244 inbred maize lines. Alleles in each polymorphism with minor frequency >0.05 were used for association mapping using the mixed linear model (MLM) controlling population structure (Q) and kinship (K) (MLM Q+K). The results revealed that two polymorphisms (S-575 and S-566) in the promoter were associated with PHT variation across two field experiments at the < 0.01 level (Figure 5a,b and Table S2). The SL15 allele of the two associated sequence polymorphisms results in an increased PHT. Detailed DNA sequencing revealed that S-575 represents an indel polymorphism with GCA three-nucleotide repeat motifs, and S-566 represents a GCT motif (Figure 5c). These two sites were in strong linkage disequilibrium (LD) (r2 = 0.53), and formed four main haplotypes in the association panel. The PHT phenotypes of haplotype 1 and haplotype 2 were not significantly different, but the plants were significantly shorter than both haplotype 3 and haplotype 4, which correspond to the genotypes of Zong3 and SL15, respectively (Figure 5d). Similarly, the PHT of haplotype 3 was significantly lower than that of haplotype 4 (Figure 5d), consistent with the PHT difference between Zong3 and SL15.

Figure 5.

Association mapping results and phenotype of four haplotypes.(a) Association between PHT and the polymorphisms in ZmGA3ox2.(b) The gene structure of ZmGA3ox2.(c) Sequences of two sites that are significantly associated with PHT variation.(d) Phenotypic comparison of four haplotypes for plant height in the Sanya 2009 and Wuhan 2010 populations. Different letters indicate statistically significant differences (< 0.01), according to a pairwise t test. Values are means ± SD.


ZmGA3ox2 is responsible for qPH3.1

In rice, OsGA3ox2 has been confirmed to correspond to D18, a known gene controlling elongation of the vegetative shoot (Itoh et al., 2001). As an ortholog of OsGA3ox2, ZmGA3ox2 shows 80% amino acid identity to OsGA3ox2 and has a similar expression pattern to OsGA3ox2 (Figure 4a), suggesting that ZmGA3ox2 may also function to control elongation of the vegetative shoot in maize, which would affect the PHT. This hypothesis was supported by the observation that a deletion in ZmGA3ox2 is responsible for the maize dwarf mutant d1. Sequence analysis found that ZmGA3ox2 in the d1-6016 mutant had a large fragment deletion that resulted in loss of function of ZmGA3ox2 (Figure 3c). Bioinformatics analyses revealed that ZmGA3ox1 in maize bin 6.05 is highly homologous with OsGA3ox1 (Figure S2c). In the d1-6016 mutant, wild-type ZmGA3ox1 is specifically expressed in the immature tassel, and the mutated ZmGA3ox2 transcript was detected in multiple tissues. However, expression of ZmGA3ox1 cannot rescue the dwarf phenotype, demonstrating that the deletion in ZmGA3ox2 is the main cause of dwarfism in the d1-6016 mutant. Quantitative PCR showed that expression of ZmGA3ox2 was significantly higher in SL15 than in Zong3 in the stem apices. The expression difference for ZmGA3ox2 was predicted to result in different GA levels between SL15 and Zong3. This was confirmed when the two plants were sprayed with exogenous GA (Figure 4d). Dai et al. (2007) and Zhu et al. (2006) suggested that GA regulates the length of internodes by affecting cell elongation. The internode cells of SL15 were significantly longer than those of Zong3 (Figure 1d–f). These findings suggest that high expression of ZmGA3ox2 in SL15 results in accumulation of ZmGA3ox2, which catalyzes production of more GA in SL15. The higher level of active GA then causes increased longitudinal elongation of internode cells in SL15, which in turn increases the length of the internodes. They also indicate that ZmGA3ox2 modulates maize PHT in the same way as its ortholog in rice (Dai et al., 2007). Moreover, two polymorphisms were found to be significantly associated with PHT in two field experiments, strongly indicating that the observed associations were real and demonstrating the stability of the gene's function. Taken together, the results from cytology, linkage, expression, physiology, bioinformatics and association mapping support the conclusion that ZmGA3ox2 is the candidate gene for qPH3.1.

Causative loci for qPH3.1

Recently, many QTLs have been cloned in plant species. The allelic variations underlying these cloned QTLs have two possible consequences. First, the protein structure and function may be altered. For example, nucleotide mutations in GS3 of rice lead to premature termination and loss of protein function (Fan et al., 2006). An 11 bp deletion in the coding region of Gn1a in rice creates a premature stop codon and causes a reduction or loss of protein function (Ashikari et al., 2005). Similarly, variations of GW2 in rice include a 1 bp deletion, which results in a premature stop codon and causes reduction or loss of function of the protein (Song et al., 2007). Second, QTLs may cause changes in mRNA accumulation. Vgt1 in maize functions as a cis-acting element that regulates expression levels of the downstream gene (Salvi et al., 2007). The expression of tb1, an important domestication gene in maize, is also regulated by a region 58.7–69.5 kb upstream of tb1. A transposable element inserted in the upstream region acts as an enhancer of gene expression (Doebley et al., 1997; Clark et al., 2006; Studer et al., 2011).

In the coding region of ZmGA3ox2 in both Zong3 and SL15, there are no premature stop codons; however, one SNP causes a substitution of the 61st amino acid. This SNP does not involve the 2OG-Fe (II) oxygenase domain nor is it predicted to alter the secondary structure of protein (Figure S2b), implying that the SNP is not the causative site for qPH3.1. This deduction was supported by the finding that the SNP was not associated with PHT variation (Figure 5a,b). If the functions of protein are not altered, then changes in the expression level of ZmGA3ox2 are responsible for qPH3.1. Association mapping revealed that two sequence polymorphisms, S-575 and S-566, in the promoter region of ZmGA3ox2 were significantly associated with PHT. The SL15 allele for these two sequence polymorphisms is associated with taller plants than the Zong3 allele, indicating that the two associated sequence polymorphisms may function in the regulation of gene expression. This is consistent with the higher levels of ZmGA3ox2 mRNA in SL15 compared with Zong3. These results indicate that the S-575 and S-566 sequence polymorphisms may be the actual functional sites affecting the difference in PHT between Zong3 and SL15. In the association panel, the mean PHT of haplotype 4 (SL15 genotype) was significantly greater than that of haplotype 3 (Zong3 genotype) by approximately 14.1 cm at Sanya in 2009 and 24.4 cm at Wuhan in 2010 (Figure 5d). The additive effect of the SL15 allele at qPH3.1 was 9.6–10.8 cm in the three F2 populations (Table 1). Therefore, the effect of haplotype 4 relative to haplotype 3 was very similar to that determined by association mapping and based on biparental mapping.

qPH3.1 for maize breeding

Maize is a dual-purpose crop, in that the grain can be harvested for food or feed, and maize straw is an important source of lignocellulosic biomass. Identifying useful PHT QTLs is important for genetic manipulation of plant architecture in maize breeding. qPH3.1 was identified using an introgression library (Z3HBIL) from experiments over 2 years (Bai et al., 2010). In the present study, the presence of the QTL was confirmed using three F2 populations of various sizes. Unlike dwarf mutants, which cause severe dwarfism, qPH3.1 modified the PHT by approximately 20 cm (approximately 10% of the total PHT) and did not affect grain yield, yield-associated traits and flowering time. Thus, ZmGA3ox2 may be utilized in maize breeding using marker-assisted selection. The Zong3 allele at qPH3.1 decreases PHT relative to the SL15 allele; therefore, it may be feasible to use lines carrying the Zong3 allele to improve lines carrying the SL15 allele. This may result in increases in grain yield by permitting increased planting density. On the other hand, the allele from SL15 increases plant height, leading to increased plant biomass, for use in synthetic fuels, chemicals and power.

Experimental procedures

Plant materials and population development

SL15 is a chromosome segment substitution line from the introgression library Z3HBIL, which was developed by crossing a recipient line (Zong3) with a donor line (HB522) through four cycles of advanced back-crosses. The introgressed segment in SL15 harbors a QTL for PHT, called qPH3.1. It is flanked by markers bnlg1647 and bnlg1447 on chromosome 3. To determine the precise position of the two ends of the introgressed segment, two new SSR markers, ND1 (near bnlg1647) and ND75 (near bnlg1447), were developed, and were used to mark the boundaries of the introgressed segment, which extends for approximately 2.6 Mb. SL15 was crossed with Zong3, and an F2 population containing 161 individuals was developed and planted in Baoding, China (38°N, 115°E) in 2007. The PHT and five other yield-related traits (100-kernel weight, number of kernel rows, kernels per row, ear length and ear diameter) were evaluated. To fine map qPH3.1, two larger F2 populations, with 617 and 2153 individuals, were developed and evaluated in Baoding in 2008 and 2009, respectively. The yield-related traits mentioned above, and days to tasseling and days to silking, were also evaluated in the 2008 F2 population. The PHT for each individual in the two populations was measured, and the individuals were genotyped using flanking markers and newly developed markers within the QTL interval. The method for development of SILs was that described by Paterson et al. (1990). Individuals containing recombination breakpoints in the QTL interval were selected for self-pollination. Their progeny were screened using molecular markers to identify homozygous recombinants, which were selfed to develop homozygous SILs. The PHT phenotypes of homozygous SILs and both Zong3 and SL15 were evaluated in three experiments [Sanya, China (18°N, 109°E) in 2009; Wuhan, China (30°N, 114°E); Baoding, China (38°N, 115°E) in 2010], using a randomized complete block design with three replications under a normal water and fertilizer regimen. The row length was 3.0 m, the spacing between rows was 0.6 m, and the spacing within rows was 0.25 m. The PHT was recorded from the field surface to the top of the tassel of physiologically mature plants. The tassel length, number of nodes and the length of the internodes were measured for both Zong3 and SL15. Internode 4, which showed the biggest variation (25.8%) among all internodes between Zong3 and SL15, was harvested from mature plants of Zong3 and SL15, and fixed in FAA (63% ethanol, 1.85% formaldehyde, 5% acetic acid in distilled water). Longitudinal sections of the stems were cut using a double-edged razor and suspended in ddH2O. Images of the cells were obtained using an Olympus 1X71 fluorescence microscope (Olympus, The d1-6016 mutant was obtained from the Maize Genetics Cooperation Stock Center (Urbana, IL, USA).

Molecular marker development

BAC sequences of the B73 genome in the region flanked by ND1 and ND75 on chromosome 3 were obtained from the MaizeGDB (Sen et al., 2010). The SSRs were identified using the simple sequence repeat identification tool (Temnykh et al., 2001). Primers were designed using Primer Premier 5.0, with a product size <300 bp. An indel marker, ND88, was developed on the basis of variations in the 3′ UTR of the candidate gene. Table S3 lists the SSRs and indel markers that were used for this study.

Genotyping and QTL mapping

Genomic DNA was extracted from fresh maize seedling leaves. SSR screening was performed as described by Bai et al. (2010). The marker linkage map in the target chromosome region was reconstructed using the Kosambi function of MAPMAKER/EXP version 3.0 (Lincoln et al., 1992). QTL analysis for composite interval mapping was performed using WinQTL Cartographer 2.5 with 1000 permutation tests for the threshold log10 likelihood ratio (Wang et al., 2006). The significance of phenotypic differences for homozygous SILs relative to Zong3 was evaluated by Student's t test.

Candidate gene sequencing and association mapping

Genomic DNA and full-length cDNA sequences of candidate genes from Zong3 and SL15 were obtained by PCR amplification using primers 4046F and 6276R (Table S4). PCR was performed using high-fidelity LA Taq with GC Buffer II (Takara, The primers 7575F and 8823R (Table S4) was used to amplify the region from approximately 1.5–2.7 kb upstream of the start codon. Primers 6107F and 7762R (Table S4) amplified the 1.6 kb promoter and the 5′ UTR. Total RNA extracted from the stem apices was used for RLM-RACE using a FirstChoice RLM-RACE kit (Ambion,, according to the manufacturer's instructions. The primers used for RLM-RACE are listed in Table S4. The purified PCR product was cloned into the pGEM-T Easy vector (Promega, according to the manufacturer's instructions. Three positive clones were sequenced for each sample. Sequence contig assembly and alignment were performed using BioEdit software (Hall, 1999).

A subset of 244 inbred lines (Table S5) from a maize association mapping panel (Yang et al., 2011) were used for candidate gene association mapping. The PHT phenotypes were evaluated using a randomized complete block design with two replications under two environments (Sanya, China, in 2009; Wuhan, China, in 2010). Three overlapping regions of ZmGA3ox2 in the 244 lines were amplified using primer pairs 7575F and 8823R, 6107F and 7762R, and 4046F and 6276R (Table S4). PCR products were directly sequenced. An initial alignment was performed using the multiple sequence alignment program MUSCLE (Edgar, 2004). MEGA version 5 was then used to refine the alignment manually (Tamura et al., 2011). Sites with allelic frequency >0.05, and showing polymorphisms between SL15 and Zong3, were used for subsequent analysis. Association mapping was performed with TASSEL 3.0 using an MLM Q+K model (Yu et al., 2006; Bradbury et al., 2007).

Analysis of the d1-6016 mutant

The primers 4046F and 7762R (Table S4) were used to amplify ZmGA3ox2 in the d1-6016 mutant. The d1-6016 mutant was first isolated from the L289/I205 single cross material which was used for the Bikini tests (Anderson, 1951). The purified PCR product was cloned into pGEM-T Easy vector (Promega), and five positive clones were sequenced. A back-cross population containing 406 individuals was developed by crossing +/d1-6016 with d1-6016/d1-6016, which was then genotyped using a gene-specific marker (ND88).

Prediction of the structure and functional domain of ZmGA3ox2 and phylogenetic analysis

The ZmGA3ox2 protein sequences of SL15 and Zong3 were separately submitted to SWISS-MODEL ( to predict the secondary structure and functional domains using the domain annotation tool. The nucleotide sequence encoding the 2OG-Fe (II) oxygenase functional domain of ZmGA3ox2 was used to search the B73 RefGen_v2 sequence ( The deduced amino acid sequences of GA 3-oxidase in rice and maize were aligned using MEGA version 5 (Tamura et al., 2011) by the neighbor-joining method; the bootstrap values were produced using 10 000 replications (Tamura et al., 2011).

Semi-quantitative RT-PCR assay and quantitative RT-PCR analysis

Tissues from stem apices (immature tassel removed) of Zong3 and SL15 plants were collected at four growth stages: the 10-leaf (jointing) stage, and the 12-leaf, 14-leaf and 16-leaf stages. The collected tissues were immediately frozen in liquid nitrogen and stored at −80°C. Total RNA was extracted using TRIzol reagent (Invitrogen, according to the manufacturer's instructions. First-strand cDNA was synthesized from total RNA samples using a PrimeScript 1st strand cDNA synthesis kit (Takara). Actin (NM_001155179) was used as an internal control. The primers used for quantitative RT-PCR analysis are listed in Table S4. Semi-quantitative RT-PCR was performed to analyze mRNA levels for ZmGA3ox1 and ZmGA3ox2 in various tissues and organs, including the crown root, primary root, seminary root, lateral root, mesocotyl, shoot, stem, immature leaf, immature ear and immature tassel of B73. Actin (NM_001155179) was again used as an internal control. Primers 374F/784R and 4046F/4221R (Table S4) were used to amplify the ZmGA3ox1 and ZmGA3ox2 transcripts, respectively.

Exogenous GA3 treatment and endogenous GA quantification

Exogenous GA3 was applied as described by Evans and Poethig (1995). Greenhouse-grown Zong3 and SL15 plants were treated twice a week in the evening with applications of 10 μg GA3 by pipetting 100 μl of a solution of GA3 (100 μg ml−1, plus 0.1% Tween-20 in distilled water) into the apical whorl of the plants. Control plants were treated in the same way with 100 μl of distilled water containing 0.1% Tween-20. The first treatment was applied 10 days after emergence. The heights of seedlings and plants were measured from the soil surface to the highest point of seedlings or plants once a week (Touzet et al., 1995).

The samples for the ZmGA3ox2 expression assay were also used for quantifying endogenous GA1 in Zong3 and SL15. Tissue samples of approximately 0.5 g were collected from plants at the 10-leaf, 12-leaf, 14-leaf and 16-leaf stages; each line was represented by five independent individuals. The samples were immediately weighed, frozen in liquid nitrogen, and later lyophilized. The GA level was analyzed using an LC/MS system (LCQ Deca MAX, HPLC-ESI-MS, Thermo-Finnigan, as described by Gou et al. (2010).


We acknowledge the Maize Genetics Cooperation Stock Center for providing seeds of the mutant d1-6016. We are also grateful to Xiangning Jiang (Beijing Forestry University, College of Life Science and Biotechnology, Beijing, China) for the analysis of phytohormones, and Zhijie Liu (Huazhong Agricultural University, Wuhan, China) for reviewing this manuscript. The work was supported by grants from the National Basic Research Program of China (2011CB100100), the Science Foundation of the Ministry of Agriculture of China (2011ZX08009-001) and the National Hi-Tech program of China (2012AA10307).