Background

Genetic improvement of grain yield is important for ensuring food security. Many important agronomic traits, including yield, show continuous phenotypic variation [1]. Grain yield is a complex quantitative trait, and is determined by several components including kernel number and kernel weight [2]. To date, more than 185 quantitative trait loci (QTL) underlying yield components including kernel row number (KRN), kernel number per row (KNPR), and hundred kernel weight (HKW) have been identified across the maize genome using various mapping populations (2012 November update to Gramene database), which has improved our understanding of the genetic basis of maize yield. Therefore, grain size and kernel weight are important targets for artificial selection for high grain yield. For example, QTL associated with GS3 [3] and GW5/SW5 [4, 5] have been selected for grain size in rice. Also, QTL associated with TaGW2-6A Hap-6A-A for grain size [6] and TaCKX6-D1 [7] for grain weight have been identified and artificially selected from within Chinese wheat germplasm. Markers derived from QTL controlling similar components of grain yield in maize will be very useful for marker-assisted selection.

Several genes that control kernel weight or kernel number have been cloned in crops, and some of these genes were found to be involved in carbon and nitrogen metabolism [8–10], protein degradation [11, 12], and hormone metabolism [13, 14]. All of these processes affect the production and export of C- and N-assimilates to the seed, thereby increasing crop yield. C- and N-metabolism are essential to all processes in plants, including reproductive development, but are especially so during grain filling [8]. Among the genes involved in carbon metabolism, rice grain incomplete filling 1, which functions similarly to the invertase encoded by miniature 1 in maize [10], also encodes a cell-wall invertase required for carbon partitioning during early grain-filling [8]. A gene involved in nitrogen metabolism in rice, OsARG, encodes an arginine hydrolase that, when overexpressed, increases grain number per plant under nitrogen-limited conditions, due to increased nitrogen remobilization at the reproductive stage [9].

In addition, degradation of proteins via the ubiquitin/proteasome pathway negatively regulates cell division and grain yield. Another gene, Grain Weight 2 (GW2), which encodes a RING-type E3 ubiquitin ligase, functions in ubiquitin-mediated degradation, and the loss of GW2 function can enhance rice grain width, weight, and yield [11]. Furthermore, ZmGW2-CHR4, which functions in the same manner as GW2, is also associated with kernel width and kernel weight in maize [12]. Although they have potentially positive impacts on yield, alleles at these loci also should be monitored to avoid possible negative effects of introduced germplasm on yield.

As to the influence of plant hormones, brassinosteroids (BRs) and cytokinins (CKs) are useful for controlling grain yield in crops. BRs stimulate the transport of sucrose and other sugars to the endosperm and embryo. Expression of the maize, rice, or Arabidopsis thaliana C-22 hydroxylase that is involved in BR synthesis in stems, leaves, and roots had clear effects on seed weight [14]. In rice, the accumulation of CK in the inflorescence meristems increases the number of grains per panicle. The rice gene Gn1a encodes an oxidase/dehydrogenase (OsCKX2) that functions in the degradation of CKs. An 11-bp deletion in its coding region created a premature stop codon that reduced expression of OsCKX2 and resulted in increased grain number [13]. In wheat, TaCKX6-D1, an ortholog of rice OsCKX2, was significantly associated with 1000-grain weight by linkage mapping and association analysis [7]. Thus, the CK regulatory pathway likely plays a crucial role in grain yield.

There is also evidence for a role of CKs in balancing the source and sink relationship [15]. Accordingly, exogenous applications of CKs increased seed set and yield stability under heat stress in maize [16]. CKs are co-regulated by isopentenyl transferase (IPT) and oxidase/dehydrogenase in plants. The discovery of biosynthetic enzymes IPT1 through IPT10 [17–19] and degradative enzymes CKX1 through CKX12 [18, 20–23] has led to a better understanding of the role of CKs in maize kernel development. Among the IPTs, IPT1 and IPT10 are highly abundant and are constitutively expressed in all organs, but other IPT transcripts show distinct spatial and temporal patterns of expression [18]. ZmIPT2 is specifically expressed in the pedicel, endosperm, and embryo [18, 19]. ZmIPT2 consists of an intronless 966 bp coding region for dimethylallyl diphosphate (DMAPP): adenosine triphosphate (ATP)/adenosine diphosphate (ADP) IPT, which preferred ATP and ADP over adenosine monophosphate (AMP) as substrates for IPT activity [19].

To date, the allelic diversity of ZmIPT2 and the most favorable allele(s) affecting kernel weight have not yet been reported for maize. The objectives of this study were to (1) examine the sequence diversity of ZmIPT2 in 175 Chinese maize inbred lines; (2) test associations between nucleotide polymorphisms in ZmIPT2 and various yield components including KRN, KNPR, and HKW, and identify further favorable allele(s) for grain yield components; (3) characterize the IPT activity of the protein products of different alleles in vitro; and (4) investigate selection for favorable ZmIPT2 alleles for kernel weight during maize breeding in China.

Results

Phenotypic data

Phenotypic values for individual lines ranged from 7.81 to 18.62 g for KRN (mean = 12.80), from 8.26 to 34.67 g for KNRP (mean = 22.40), and from 12.35 to 47.60 g for HKW (mean = 28.36), as shown in Table 1. KRN was significantly positively correlated with KNPR (P <0.01), while HKW was negatively correlated with KNPR (P <0.05). Analysis of variance (ANOVA) indicated significant phenotypic variation for KRN, KNPR, and HKW among these 175 maize lines (P <0.01). Heritabilities estimated on a per-plot basis were 86.58% for KRN, 84.27% for KNRP, and 86.12% for HKW.

Table 1 Mean squares from ANOVA and correlation coefficients for KRN, KNPR, and HKW
Full size table

Nucleotide diversity in the ZmIPT2coding region

A total of 16 single nucleotide polymorphisms (SNPs) were identified in the ZmIPT2 coding region, with an average of one SNP every 60 bp, and six of these variants resulted in amino acid substitutions (Figure 1A). Seven haplotypes, named Hap_1 through Hap_7, were detected in our association panel based on these 16 SNPs (Figure 1B). A total of 75 (42.9%) inbred lines harbored Hap_1, while 10, 31, 22, 9, 10, and 18 inbred lines carried Hap_2 through 7, respectively. Hap_1 harbored the ZmIPT2-T allele at SNP28 (C-T), but other haplotypes bore the ZmIPT2-C allele. The window covering ZmIPT2 exhibited relatively lower diversity with a value of 0.347, and it was significantly lower than the average diversity value of 0.372 for Chromosome 2 across 175 maize inbred lines (P < 0.01) (Figure 2). Most polymorphisms in ZmIPT2 were found in the region from 360 to 590 bp, (P >0.005) (Figure 3). Among maize subgroups revealed by Xie et al. [24], subgroups Lancaster (Lan) and Lvda red cob (LRC) harbored higher levels of diversity in the ZmIPT2 coding region, compared with the average for all inbreds (Figure 3).

Figure 1
figure 1

Nucleotide diversity and amino acid variants in the ZmIPT2 coding region. (A) The full sequence of ZmIPT2 gene. Sites are labeled according to the position of the corresponding polymorphisms in the aligned sequences, with amino acid alterations indicated in parentheses. The gray box indicates the IPT domain, which starts at position 28 and ends at position 480. (B) The ZmIPT2 genomic sequences grouped into seven haplotypes defined by 16 single nucleotide polymorphisms identified in the alignment.

Full size image
Figure 2
figure 2

Sliding-Window analyses of nucleotide diversity π and Tajima’s D in the association panel, with a sliding window size of 500 SNP markers and step intervals of 50 SNP markers on maize chromosome 2. Blue diamonds, π; Black diamonds, Tajima’s D; Red diamonds, windows containing the ZmIPT2 gene locus. Blue dashes represent the mean π value; Black dashes represent the mean Tajima’s D value.

Full size image
Figure 3
figure 3

Sliding-Window analysis of nucleotide diversity in the ZmIPT2 coding region across six maize subgroups with 100 bp window and 25 bp step size. BSSS, derived from US BSSS and Reid germplasm; Lan, derived from US Lancaster Sure Crop germplasm; LRC, derived from Lvda Red Cob, a Chinese landrace; PA, derived from modern US hybrids in China; PB, derived from modern US hybrids in China; SPT, derived from Si-ping-tou, a Chinese landrace. Combined sample, 175 maize inbred lines.

Full size image

Association analysis of KRN, KNPR, and HKW

Association analysis revealed that SNP28 was significantly associated with HKW across all three environments tested (P <0.05), and explained 5.42% of the phenotypic variation in HKW at Xinjiang in 2007, 3.41% of the phenotypic variation at Beijing in 2008, and 5.46% of the phenotypic variation at Xinjiang in 2008. Significant association of HKW with SNP195, SNP413, and SNP591 was detected in one or two environments (Table 2). However, only SNP413 was found to be associated with KNPR in Beijing (P <0.05), and no SNP was significantly associated with KRN (Table 2). The inbred lines containing the ZmIPT2-T allele had significantly (P <0.05) greater kernel weight than those carrying the ZmIPT2-C allele, as seen in the data from Xinjiang in 2007 and in Beijing in 2008 across this association panel (Figure 4). The P value obtained here supports the strength of the observed association between HKW and SNP28, and improves confidence in identification of the ZmIPT2-T allele in Hap_1 as the most favorable allele for kernel weight in maize.

Table 2 ZmIPT2 polymorphisms associated with KRN, KNPR, and HKW in the association panel
Full size table
Figure 4
figure 4

Hundred-kernel weight (HKW) comparison of groups harboring either ZmIPT2 - T or ZmIPT2 - C . P, P value for t-test comparing two groups carrying different alleles in each of three environments. 07XJ, Xinjiang in 2007; 08XJ, Xinjiang in 2008; 08BJ, Beijing in 2008. White boxes represent the ZmIPT2-T group; grey boxes represent the ZmIPT2-C group.

Full size image

Characterization of ZmIPT2

Both the ZmIPT2-T and ZmIPT2-C protein alleles have the same pI and aliphatic index. However, the transition at SNP28 from cytosine (C) to thymine (T) converts proline into serine, resulting in a decrease in the hydropathicity of the entire ZmIPT2 protein from −0.006 to −0.008, as analyzed using GRAVY. The IPT coding region aligned with the P-loop NTPase superfamily that is characteristic of all NTP-binding proteins from position 28 to 480 with an E-value, or probability of a protein sequence of a given bit score occurring by chance, of 1.3e-06 (NCBI blastp database) (Figure 1A). The 3-D structure predicted by SWISS-MODEL (changed to http://swissmodel.expasy.org/) showed that the spatial structure of ZmIPT2-T and ZmIPT2-C differed due to variation in hydrophobicity between the two proteins (Additional file 1). These kinds of changes could affect the binding of substrates to the enzyme, which is mediated by hydrogen bonds [25]. This mutation and resultant differences in physicochemical properties imply that IPT activity could differ between the protein products of the ZmIPT2-T and ZmIPT2-C alleles.

To further confirm the CK biosynthetic function of these two ZmIPT2 products in vitro, Escherichia coli prokaryotic expression vectors were constructed to express and purify ZmIPT2 as C-terminal His-tagged recombinant proteins. ZmIPT2-T and ZmIPT2-C purified on Talon® columns reached 95-100% purity, as quantified by Coomassie staining after PAGE (Additional file 2). Purified proteins were then used to assay DMAPP:ATP, DMAPP:ADP, and DMAPP:AMP transferase activity. Figure 5 shows the profile obtained from the reaction mixture of purified ZmIPT2 protein incubated with DMAPP and substrates, including ATP, ADP, and AMP. ZmIPT2-T converted AMP to isopentenyl adenosine monophosphate (iPMP) with higher efficiency than did ZmIPT2-C, as indicated by the amount of iPMP after reaction (Figure 5C). The mean AMP consumption by ZmIPT2-T and ZmIPT2-C were 0.20 μmol and 0.07 μmol, respectively (Additional file 3). Similar results were obtained with ATP or ADP as the substrate (Figure 5A, B). ZmIPT2-T consumed more ADP and ATP, with means of 270.2% and 548.4%, respectively, than consumed by ZmIPT2-C. Additionally, ZmIPT2-T consumed much more ATP and ADP than it did AMP, with means of 7.38- and 13.61-fold the amount of AMP consumed, respectively (Additional file 3). These results indicate that the protein product of the ZmIPT2-T allele has higher IPT enzyme activity and prefers ADP as substrate, and so could have a more favorable impact on yield.

Figure 5
figure 5

Determination of isopentenyl transferase activity of ZmIPT2. A, B, and C depict determination of DMAPP:ATP, DMAPP:ADP, and DMAPP:AMP isopentenyl transferase activity of recombinant ZmIPT2 protein. iPTP = isopentenyl adenosine triphosphate, iPDP = isopentenyl adenosine diphosphate, and iPMP = isopentenyl adenosine monophosphate.

Full size image

Selection in the region of the ZmIPT2locus

A total of 4282 SNPs [26] were used to analyze the effects of selection occurring in this region of Chromosome 2 during selection and breeding. The estimates for π and Tajima’s D in the ZmIPT2 window showed lower values than the mean estimates for those parameters for all of maize Chromosome 2 (Figure 2). Tajima’s D tests of the ZmIPT2 coding region and genomic window identified different artificial selection effects across these six maize subgroups, with the largest effect identified in subgroup PB (Figure 6). Notably, 81% of inbred lines in subgroup PB harbored the favorable allele ZmIPT2-T (Table 3). These results indicate that artificial selection occurred in the genomic region harboring ZmIPT2 during maize breeding.

Figure 6
figure 6

Tajima’s D values for the ZmIPT2 coding region and hundred kernel weights across six maize subgroups. PB, derived from modern US hybrids in China; BSSS, derived from US BSSS and Reid germplasm; LRC, derived from Lvda Red Cob, a Chinese landrace; PA, derived from modern US hybrids in China; Lan, derived from US Lancaster Sure Crop germplasm; SPT, derived from Si-ping-tou, a Chinese landrace. HKW, hundred-kernel weight. 07XJ, Xinjiang in 2007; 08XJ, Xinjiang in 2008; 08BJ, Beijing in 2008.

Full size image
Table 3 Distribution of the ZmIPT2-T allele among six subgroups in 175 Chinese inbred lines
Full size table

Discussion

Grain yield is a complex trait controlled by minor genes

Grain yield is one of the most important and complex traits in maize. Although many studies on the genetic basis of maize yield-related traits have been reported, few favorable alleles for yield candidate genes have been identified because single genes identified in association studies explain only a small proportion of phenotypic variation. In this study, SNP28 in the ZmIPT2 coding region was significantly associated with kernel weight across three environments, explaining 4.76% of phenotypic variation. Relatively small allelic effects were also found for one polymorphism in the ZmGS3 region, which explained 6.29% of variation in Beijing and 7.73% of that in Hainan [27], and another SNP, S40 in the ZmGW2-CHR4 region, which accounted for less than 10% of variation across three environments [12]. Thus, the identification and pyramiding of a large number of favorable alleles will be the key to molecular marker-assisted breeding of high-yielding maize.

Association mapping of complex traits

As an effective way to dissect complex traits and identify favorable alleles or haplotypes, association mapping has been widely used for many maize traits, including flowering time [28], leaf architecture [29], plant height [26], and disease resistance [30–32]. With candidate-gene association mapping, excellent allelic variants of many genes have been identified, such as crtRB1 for β-carotenoid synthesis in maize kernels [33]; Dwarf8 for maize flowering time [6, 34, 35]; ZmGW2-CHR4[36] and ZmGS3[27], which are related to maize kernel size and weight; TaGW2, which is associated with wheat kernel weight and maturity [37]; and other genes related to starch synthesis in rice [38].

Contribution of CKs to grain yield

Plant hormones control many aspects of development, and like other hormones, CKs affect many plant physiological and biochemical processes, including those controlling delayed senescence [39], suppression of auxin-induced apical dominance [40], signaling of nitrogen availability [41], cell division and differentiation [42], and sink strength [15]. IPT proteins, which are involved in synthesis of CKs, are encoded by a multi-gene family in maize [17–19], rice [43], soybean [44], Arabidopsis [45, 46], petunia [47], hop [48] and pea [49]. A favorable IPT allele that resulted in increased grain weight has now been identified in maize using an association mapping approach (Figure 3).

CK content is co-regulated by two families of genes, including isopentenyl transferase and an oxidase/dehydrogenase in plants. In rice, reduced expression of OsCKX2, which encodes the oxidase/dehydrogenase, caused an increase in grain number per panicle, due to CK accumulation in inflorescence meristems. OsCKX2 is expressed strongly in pistils, at an intermediate level in inflorescences and seeds, and at low levels in other organs [13]. TaCKX6-D1, a wheat ortholog of rice OsCKX2, was related to HGW by linkage mapping and association analysis. TaCKX6_D1 was expressed at a high level in seeds and pistils, but at a low level in other organs [7]. One study showed that ZmIPT2 is specifically expressed in pedicel, endosperm, and embryo, and plays a major role in CK biosynthesis for maize kernel development [18, 19]. In this study, however, only SNP413 was found to be significantly associated with KNPR in one of the three experimental environments (Beijing), and no associated SNP was found for KRN (Table 2). CKs coordinately induce extracellular invertase and hexose transporter activity, which are functionally coupled to supply carbohydrates to sink tissues [15]. In this study, a single nucleotide substitution from C to T in the ZmIPT2 coding region was significantly associated with kernel weight. This suggests that manipulation of CK metabolism in reproductive organs could be an effective approach to further increase crop grain yields, by increasing the flow of assimilates from source to sink, or by increasing sink capacity.

In vitro characterization of the ZmIPT2 enzyme showed that it prefers ADP and ATP over AMP as substrates for DMAPP IPT activity (Figure 5, Additional file 3), consistent with the results of a previous study [19]. Six amino acid substitutions, none of which were conservative, occurred in ZmIPT2, compared with 10 other maize ZmIPTs [18], indicating that ZmIPT2, and its specific expression pattern, is conserved and essential for kernel development. Interestingly, the mutation associated with the favorable allele, ZmIPT2-T, was detected in the ZmIPT2 coding region, and found to result in a serine to proline substitution in the conserved IPT domain that caused the ZmIPT2-T enzyme to consume more ATP, ADP, and AMP than the ZmIPT2-C enzyme (Figure 5, Additional file 3). Thus, selection for the molecular marker SNP28 should improve yield by positively influencing maize kernel development via an increase in IPT enzyme activity.

Nucleotide diversity in ZmIPT2across six subgroups

In this study, the coding region of ZmIPT2 was sequenced across 175 Chinese maize inbreds. The coding regions of functional genes tend to be relatively conserved, due to their specificity for and affinity with other types of molecules. Thus, under natural selection, beneficial variants tend to become gradually fixed, while detrimental variants tend to be eliminated [37]. Lower levels of diversity within the ZmIPT2 locus were estimated in subgroups PB and BSSS, compared to the combined group (P < 0.01) (Figure 3), suggesting either that the germplasm base remained remarkably narrow or that strong positive selection at that locus occurred in these subgroups. According to pedigrees, 17 inbreds in subgroup PB were mainly derived from US hybrid P78599. However, subgroup Lan harbored the most diversity in the ZmIPT2 coding region across all these inbreds (Table 1; Figure 3). This subgroup has been identified as having the most genetic diversity in one study using 145 SSR loci randomly distributed across the genome [50].

Selection for grain yield during the breeding process

It is hypothesized that natural selection favored smaller-seeded wild ancestors with a larger number of seeds per plant, earlier maturity, and wider geographic distribution [51]. However, high crop yield associated with larger grain size and weight has been the objective of artificial selection during most breeding programs. For example, the genes GS3 [52] and GW5/SW5 [3, 53] have been proven to be involved in the evolution of grain size in rice. Unlike rice, in wheat and maize, artificial selection is associated with an almost uniform increase in seed or grain size. The allele TaGW2-6A Hap-6A-A, which was associated with larger grain size, increased in frequency under positive selection from 50.0% in the 1950s to current levels of 77.42% [37]. Under strong selection at a given locus, domestication has the potential to drastically increase LD and reduce diversity [4].

Estimates of the parameters π and Tajima’s D in this study showed that artificial selection took place at the ZmIPT2 locus (Figure 3). However, only Hap_1, among seven haplotypes examined here, harbored the ZmIPT2-T allele (Figure 1), and this favorable allele was associated with increased grain weight in two of the three experimental environments (Figure 4), suggesting positive selection for higher grain weight. Similarly, a 95.7% reduction in nucleotide diversity across the GS3 locus occurred among rice accessions carrying the A allele for GS3 [52]. The current study found that the ZmIPT2-T allele was not evenly distributed among these six maize subgroups, occurring with the highest frequency in subgroup PB, in which 81% of the tested lines harbor the favorable allele (Table 3). The lowest Tajima’s D value was also found for subgroup PB (Figure 6) indicating that distinct positive selection pressure has taken place at the ZmIPT2 locus in this subgroup. Selection for the favorable allele ZmIPT2-T within each subgroup should be continued in order to improve maize yield.

Conclusions

A favorable allele, ZmIPT2-T, was associated with kernel weight in Chinese maize germplasm. ZmIPT2-T was shown in vitro to have higher IPT activity than ZmIPT2-C with ADP, ATP, or AMP as substrate. Artificial selection at the ZmIPT2 locus was detected with Tajima’s D tests across six subgroups of Chinese maize germplasm, with the most frequent favorable allele identified in subgroup PB. This favorable allele could be used in molecular marker-assisted selection for improvement of grain yield components in Chinese maize breeding programs.

Methods

Experimental design and statistical analyses

In this study, a total of 175 maize inbred lines were used, which were subdivided into six subgroups [24], including 29 BSSS lines derived from US BSSS and Reid germplasm; 55 PA (Partner A) lines derived from modern US hybrids in China; 21 PB (Partner B, harboring the distinct heterosis reaction of PA) lines derived from modern US hybrids in China; 18 Lan lines derived from US Lancaster Sure Crop germplasm; 27 LRC lines derived from Lvda Red Cob, a Chinese landrace; and 25 SPT lines derived from Si-ping-tou, a Chinese landrace. All lines were planted at Xinjiang in 2007 and 2008, and at Beijing in 2008 to measure grain yield components, including KRN, KNPR, and HKW. A randomized complete-block design was employed with three replications of 20 plants of each line per location planted in 4.5-m rows, 0.6 m apart. Normal maize agricultural practices were carried out. At maturity, all ears were harvested manually and dried to grain moisture of 13%. KRN was scored as number of rows per ear, KNPR was scored as total kernels in a row per ear, and HKW was measured on 100 randomly selected kernels per ear. The average phenotypic values of each plot over three replications in each environment were used for final analysis.

Descriptive statistics and analysis of variance (ANOVA) for KRN, KNPR, and HKW were obtained using the program PROC GLM in SAS software version 9.13 (Copyright © 2009 SAS Institute Inc., 2009. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc., Cary, NC, USA). The SAS program PROC CORR was used to quantify the relationship among the three traits. Broad-sense heritability was estimated based on plot as h2 = σ g 2/ (σ g 2 + σ gl 2/n + σ e 2/nr), where σ g 2is genetic variance, σ gl 2 is genotype-by-environment interaction, σ e 2 is error variance, r is the number of replications, and n is the number of locations. The estimates for σ g 2, σ gl 2, and σ e 2 were acquired from ANOVA [5].

DNA isolation, PCR amplification and DNA sequencing

Genomic DNA was extracted from maize leaves using the CTAB method [54]. Primers ZmIPT2-5′ (5′-ATCATCAAGACAATGGAGCACGGTG-3′) and ZmIPT2-3′ (5′-CGTCCGCTAGCTACTTATGCATCAG-3′) were designed based on the published ZmIPT2 cDNA sequence (Accession number EU263126). ZmIPT2 genomic sequences were amplified from each of 175 inbred lines in 25 μL reaction mixtures containing 20 ng of genomic DNA, 0.2 mM of each dNTP, 1 μM of each primer, 2 × GC buffer (2 mM Mg2+), and 2.5 U TransTaq High Fidelity DNA polymerase (Transgen Biotechnology Corporation, Beijing, China). Touchdown PCR was applied as follows: 94°C for 2 min (one cycle);94°C for 30 s, 65°C for 45 s, and 72°C for 90 s (5 cycles, annealing temperature reduced 1°C per cycle);94°C for 30 s, 60°C for 45 s and 72°C for 1 min 30 s (30 cycles); 72°C for 7 min, then the reaction was then held at 4°C. The products were sequenced (ABI3730) at the public laboratory of the National Key Facility of Crop Gene Resources and Genetic Improvement, Institute of Crop Science, Chinese Academy of Agricultural Science.

Nucleotide diversity and selection in breeding process

DNA sequences of all lines were analyzed using DnaSP Version 4.00 [55]. The Sliding-Window analysis of nucleotide diversity was performed with window size set at 100 bp and steps of 25 bp. The parameter π was estimated as the average proportion of nucleotide differences between all possible pairs of sequences in the sample [56]. A total of 4282 SNPs on maize Chromosome 2 from the MaizeSNP50 BeadChip (Illumina, Inc.) were used to estimate π and Tajima’s D using TASSEL 3.0 software [26, 57].

Association mapping of KRN, KNPR and HKW

To exclude the effect of population structure on association mapping, population structure and kinship information for 175 lines were analyzed using STRUCTURE version 2.3 and SPAGeDi software with SNPs selected from a collection of 5000 SNPs identified in a previous study and selected based on physical position and minor allele frequency [26]. Association mapping between three yield-related traits and the nucleotide diversity of the ZmIPT2 coding region was performed using TASSEL 3.0 [57] with a mixed linear model (MLM) controlling both population structure (Q) and relative kinship (K).

Characterization of ZmIPT2

The 3D protein model of ZmIPT2 was predicted using the online process SWISS-MODEL Automated Mode (http://swissmodel.expasy.org/). Physicochemical properties of ZmIPT2 proteins, including isoelectric point (pI), aliphatic index, and grand average of hydropathicity (GRAVY) were analyzed using ProtParam (http://web.expasy.org/protparam/). These kinds of changes could affect the binding of substrates to the enzyme, which is mediated by hydrogen bonds [25]. The conserved domain was used to query the smart web program (http://smart.embl-heidelberg.de/).

To characterize the enzyme activity of the products of these two ZmIPT2 alleles (the T variant and the C variant), prokaryotic expression and enzyme activity determination of ZmIPT2 were carried out in vitro. The gene fragments harboring ZmIPT2-T or ZmIPT2-C were amplified using gene-specific primers with appropriate NdeI and NotI restriction site extensions, (5′-GGCATATGGAGCACGGTGCCGTCGC-3′ and 5′-CCGCGGCCGCTCATCATGCATCAGCCACGGCGGTGA-3′) from the ‘Ye478’ (Hap_1) and ‘Guan17’ (Hap_2) inbred lines, respectively. The PCR products were digested with NdeI and NotI, and then cloned into pET30b (Novagen, Darmstadt, Germany) at these same restriction enzyme sites. The recombinant plasmids were integrated into BL21 (DE3) E. coli competent cells (Tiangen Biotech Co., Ltd, Beijing, China). The cultures were incubated at 37°C with shaking until log phase at OD600 = ~ 0.4. Expression of the IPT protein was induced by incubation with 1 mM IPTG at 37°C for 3 h, followed by purification using the HisTALON™ Gravity column Purification Kit (Clontech Laboratories, Inc., CN) according to the manufacturer’s guidelines. Sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis analysis (PAGE) was performed to ensure purity, and protein density was measured using the Macro-BCA Assay Kit (Gragen Life Science Inc., CN).

Purified protein was used to determine DMAPP:ATP, DMAPP:ADP, and DMAPP:AMP IPT activities. Each purified protein (~0.13 mg/ml) was incubated in a reaction mixture containing 12.5 mM Tris–HCl (pH 7.5), 37.5 mM KCl, 5 mM MgCl2, 1 mM DMAPP (Sigma-Aldrich Co., St. Louis, USA), and 1 mM ATP (Sigma-Aldrich Co., St. Louis, USA), ADP (Sigma-Aldrich Co., St. Louis, USA) or AMP (Sigma-Aldrich Co., St. Louis, USA) for 2 h at 30°C. The reaction was stopped by boiling for 5 min. The reaction products were separated by reversed-phase HPLC (Shimadzu LC-10AT series HPLC system with a SPD-10AVP UV–vis detector) with a C18-ODS2 column (Kromasil), according to the following program: 20 mM KH2PO4 for 15 min, linear gradient of 0% acetonitrile, 20 mM KH2PO4 to 20% acetonitrile, and 4 mM KH2PO4 over 30 min. UV absorbance was monitored at 280 nm.