Evidence for regulation of columnar habit in apple by a putative 2OG-Fe(II) oxygenase



This article is corrected by:

  1. Errata: Corrigendum Volume 207, Issue 3, 928, Article first published online: 17 April 2015


  • Understanding the genetic mechanisms controlling columnar-type growth in the apple mutant ‘Wijcik’ will provide insights on how tree architecture and growth are regulated in fruit trees.
  • In apple, columnar-type growth is controlled by a single major gene at the Columnar (Co) locus. By comparing the genomic sequence of the Co region of ‘Wijcik’ with its wild-type ‘McIntosh’, a novel non-coding DNA element of 1956 bp specific to Pyreae was found to be inserted in an intergenic region of ‘Wijcik’.
  • Expression analysis of selected genes located in the vicinity of the insertion revealed the upregulation of the MdCo31 gene encoding a putative 2OG-Fe(II) oxygenase in axillary buds of ‘Wijcik’.
  • Constitutive expression of MdCo31 in Arabidopsis thaliana resulted in compact plants with shortened floral internodes, a phenotype reminiscent of the one observed in columnar apple trees. We conclude that MdCo31 is a strong candidate gene for the control of columnar growth in ‘Wijcik’.


Tree architecture exerts a considerable influence on fruit yield and directly affects the amount of work required for orchard maintenance, including pruning, training and tree support systems which represent a major cost for apple growers (Rom & Barritt, 1990; Barritt, 1992; Lespinasse & Delort, 1993; Wünsche & Lakso, 2000). Columnar trees such as the apple (Malus × domestica) mutant ‘Wijcik’ have been proposed as an interesting solution for creating high-density orchards despite the columnar habit being linked to undesirable traits such as alternate bearing (Looney & Lane, 1984; Davenport, 2000; Kenis & Keulemans, 2007). ‘Wijcik’ was first identified in the 1960s as a shoot of a normal ‘McIntosh’ tree and exhibits a pronounced columnar phenotype with short internodes and a thick stem. Axillary buds usually develop into spurs rather than lateral branches, resulting in a tree bearing fruits close to the stem (Fisher, 1969; Tobutt, 1985). Nevertheless, the development of long side shoots from spurs also occurs in columnar trees, for instance when the central leader is cut or damaged. In this particular case, some spurs near the top can grow and give rise to shoots exhibiting a columnar phenotype suggesting that lateral buds are under tight apical control (Petersen & Krost, 2013).

The columnar phenotype was shown to be caused by a single, dominant allele named Columnar (Co), mapping on chromosome 10 (Conner et al., 1997). Different studies helped to narrow down the genetic window for the Co region (Bai et al., 2012; Moriya et al., 2012; Baldi et al., 2013). Baldi et al. (2013) identified a candidate region delimited by the SSR markers Co04R10 and Co04R13 which span 393 kb in the homologous region of ‘Golden Delicious’ genome sequence (Velasco et al., 2010), overlapping with the 196 kb region identified by Moriya et al. (2012).

In parallel, others attempted to identify the Co gene using gene expression profiling with RNAseq, which resulted in the identification of several candidates for the Co gene with roles in plant hormone signaling, in particular DELLA proteins involved in the regulation of the gibberellin pathway (Zhang et al., 2012). A similar approach was used to propose a correlation between cytokinins and IAA (indole-3-acetic acid) and the columnar phenotype (Krost et al., 2013; Petersen & Krost, 2013).

Understanding the genetic mechanisms controlling columnar growth habit will help design novel strategies to develop new apple cultivars for high-density planting while avoiding the introduction of undesirable characters linked to the columnar locus.

Thus, in this investigation we compared the sequences of a BAC library containing the Co region of the mutant ‘Wijcik’ with the one prepared from the corresponding wild-type cultivar ‘McIntosh’ and an insertion of 1956 bp was identified in the ‘Wijcik’ genome which was absent in the wild-type.

We analyzed the expression patterns of the six genes identified within the 50 kb region surrounding the insertion and found only one gene (MdCo31) encoding a putative 2OG-Fe(II) oxygenase which was differentially expressed between ‘McIntosh’ and ‘Wijcik’ buds. We then investigated the biological function of this gene by expressing it constitutively in Arabidopsis thaliana and observed that transformed plants exhibited phenotypes reminiscent of the one observed in ‘Wijcik’ providing an evidence for the role of MdCo31 in regulating the columnar-type growth in apple.

Materials and Methods

Plant material and growth conditions

Apple (Malus × domestica Borkh.) trees were grown and maintained at the orchard ‘Giaroni’ belonging to the Fondazione Edmund Mach (FEM) at the Istituto Agrario di San Michele all'Adige (IASMA) located in Italy (latitude 46.181539°, longitude 11.119877°). ‘McIntosh’ and ‘Wijcik’ trees (Fisher, 1969) were obtained from the department of fruit trees and woody plant species at the University of Bologna (DCA-UNIBO, Italy) and grafted on M9 rootstock. The cross between ‘Wijcik’ and ‘Golden Delicious’ was performed using ‘Wijcik’ pollen giving rise to a segregating population of 103 plants (Supporting Information, Fig. S1). Axillary buds from 6-yr-old ‘Wijcik’ × ‘Golden Delicious’ trees and 2-yr-old ‘McIntosh’ and ‘Wijcik’ trees were harvested at the end of March and leaves from the same trees in May. Arabidopsis plants Col-0 used in this study were grown in GS90 soil (Manna Italia) and raised at 22°C (16 h of 100 μmol m−2 s−1 light, 8 h of dark).

BAC library construction and sequencing

The ‘McIntosh’ BAC library was prepared by Amplicon express (Pullman, WA, USA) from leaf material using the same method as described for the ‘Wijcik’ BAC library (Baldi et al., 2013). The SSR markers used in Baldi et al. (2013) and novel ones indicated in the Supporting Information, Table S1, were used to screen the BAC libraries as described in Methods S1. BAC plasmids were purified using the Plasmid Midi kit (Qiagen) and sequenced on a GS FLX Titanium platform (Roche). The reads were assembled using Newbler Assembler software (Roche) and MIRA assembler (Chevreux et al., 1999) using the ‘Golden Delicious’ genome as a reference. The resulting contigs were joined together using Sanger sequencing. The assembled sequences of ‘McIntosh’ and ‘Wijcik’ (NCBI Accession number KF530875 and KF530876) were compared by alignment using SSAHA2 (Ning et al., 2001) and visualized using IGV (Robinson et al., 2011). Glimmer HMM was used to identify ORFs in the ‘Wijcik’ Co region sequence (Majoros et al., 2004). Gene predictions and full-length cDNA sequences from the ‘Golden Delicious’ Co region were used as a reference. Predicted genes in the Co region were annotated by performing a tBLASTx search against the Arabidopsis protein database (Rhee et al., 2003) and by carrying out a conserved protein domain search using the NCBI's Conserved Domain Database (Marchler-Bauer et al., 2011). A BLAST search in GeneBank databases for the 1956 bp insertion sequence was performed. Matches were found only on the apple and pear genome (Wu et al., 2013). Apple and pear genome contigs containing the insertion were aligned using MEGA 5.1 (Tamura et al., 2011).

RNA extraction and qRT-PCR analysis

Total RNA was isolated from apple buds and leaves using the Plant spectrum kit (Sigma-Aldrich) and from Arabidopsis leaves using TRI reagent (Sigma-Aldrich). cDNAs were synthesized from 1 μg of total RNA, previously subjected to DNAse treatment, using the superscript VILO™ cDNA synthesis kit according to the manufacturer's instructions (Life Technologies, Carlsbad, CA, USA). Real-time PCR reactions were performed using SYBR green chemistry (Platinum SYBR Green qPCR SuperMix-UDG, Life Technologies) with the primer pairs listed in the Supporting Information, Table S1, as described in Methods S1. Transcript levels were normalized to the reference genes Actin (Li & Yuan, 2008) and Md_4592:1:a (Botton et al., 2011) for apple, and TIP41-like (Kutter et al., 2007) for Arabidopsis (Supporting Information, Methods S1). Data were analyzed using the comparative Ct method (Pfaffl, 2001).

Binary vector construction and plant transformation

MdCo31 was amplified from total RNA of ‘Wijcik’ using the SuperScript One-Step RT-PCR System (Life Technologies) and the primers indicated in Table S1 and cloned into the pENTR/D vector using the pENTR-D/TOPO cloning kit (Life Technologies). pENTR/D-MdCo31 was recombined with the pH2GW7 vector (Karimi et al., 2002) using the Gateway LR Clonase II kit (Life Technologies) to give rise to pPro35S:MdCo31. Arabidopsis transformants were produced by the floral dip method using Agrobacterium tumefaciens strain GV3101 carrying the T-DNA binary vector pPro35S:MdCo31 (Clough & Bent, 1998).

Results and Discussion

BAC clones covering the genomic Co region of ‘Wijcik’ and ‘McIntosh’ were identified by screening the respective BAC libraries with specific markers indicated in Table S1. The BAC ends of all positive clones in coupling phase with Co were sequenced and used to anchor the corresponding BACs to the ‘Golden Delicious’ genome creating a minimum tiling path of BAC clones spanning the complete Co region, except for a small portion of 13 kb that was not covered by the ‘McIntosh’ BAC library (Fig. 1a). The selected BAC clones were sequenced using 454 with an average coverage of 100× resulting in a first assembly composed of 25 contigs for the complete Co region of ‘Wijcik’ and only six for the corresponding region in ‘McIntosh’. The gaps between contigs and the 13 kb segment not covered by the ‘McIntosh’ BAC clones were sequenced by Sanger sequencing. Thus, the sequences of the ‘Wijcik’ and ‘McIntosh’ Co region were covered by two contigs each, named WiC1/WiC2 and McC1/McC2, respectively (Fig. 1a). In both genotypes the two contigs could not be joined as they were separated by a long sequence of repetitive DNA (Fig. 1a).

Figure 1.

BAC screening, sequence assembly and identification of ‘Wijcik’ insertion. (a) The location of apple (Malus × domestica) Co region delimited by SSR markers Co04R10 and Co04R13 on chromosome 10 of ‘Golden Delicious’ is shown at the top of the figure. Just below, an enlargement of the Co region is shown, indicating the position (on the ‘Golden Delicious’ genome) of the SSR markers used to screen the ‘McIntosh’ and ‘Wijcik’ BAC library. The BAC clones used to construct contigs covering the Co region in ‘McIntosh’ (blue) and ‘Wijcik’ (red) and their locations on the homologous region from ‘Golden Delicious’ are displayed. The assembly of the BAC clones covering the Co region resulted in two contigs in ‘McIntosh’ (‘McC1′ and ‘McC2′ in blue) and ‘Wijcik’ (‘WiC1’ and ‘WiC2’ in red). The location of the insertion in ‘Wijcik’ is indicated on McC1 and on WiC1. Primers used to detect the presence of the insertion in ‘Wijcik’ and ‘McIntosh’ and the names of the corresponding amplification products are indicated on the ‘Wijcik’ insertion. (b) Specific amplification of the insert sequence from genomic DNA of ‘Wijcik’ (‘Wi’) and ‘McIntosh’ (‘Mc’). The different amplification products 5′CR, CR/TC and TC/3′R refer to the fragments amplified using primer pairs indicated in (a).

The sequences of WiC1/WiC2 and McC1/McC2 were compared. Forty-three genes, named MdCo1 to MdCo43, were predicted in the Co region encoding proteins and enzymes with diverse functions (Supporting Information, Table S2). As ‘Wijcik’ originated from a somatic mutation in ‘McIntosh’ (Fisher, 1969), we expected the two sequences to be almost identical, apart from the mutation that led to columnar growth. Indeed, after correction of sequencing errors, the only difference found was an insertion of 1956 bp in ‘Wijcik’ (Fig. 1a). This insertion falls within the smaller putative Co region described by Moriya et al. (2012), who used segregating populations from different parents, in agreement with our results.

In order to confirm the presence of the insertion in the ‘Wijcik’ genome and to exclude possible artifacts that could have occurred during BAC library construction or during the assembly of 454 reads, selective amplifications were performed on genomic DNA of ‘McIntosh’ and ‘Wijcik’ using specific primers indicated in Table S1. When primers spanning the genomic flanking regions and the insert sequence were designed at the 5′ and 3′ ends of the insertion, specific PCR products of 586 bp (5′ CR) and of 319 bp (TC/CR 3′) respectively, were obtained only in ‘Wijcik’ (Fig. 1b). When amplification was performed using primers designed in the core region (CR/TC), a PCR product of 931 bp was obtained in ‘McIntosh’ as well, suggesting the presence of similar elements at other locations in the apple genome (Fig. 1b). A BLAST search against the apple reference genome confirmed that more than 250 sequences with similarities to the ‘Wijcik’ insertion are distributed throughout the Malus chromosomes (Supporting Information, Fig. S2). Comparison of these sequences by alignment revealed that the ‘Wijcik’ insertion has a conserved sequence of c. 800 bp at the 5′ end, followed by a less conserved TC-rich region and a highly conserved stretch of 160 bp at the 3′ end (Fig. 1a, Supporting Information Fig. S3). Therefore the ‘Wijcik’ insertion might be a novel type of mobile DNA element, lacking the terminal inverted repeats (TIR) that are normally present in transposons (Bennetzen, 2000). Moreover, it seems to be conserved only in Pyreae since similar sequences were found in Pyrus bretschneideri, but not in other Rosaceae genera for which whole genome sequences exist (Fig. S3). The fact that both ends are highly conserved suggests that the insertion might occur through a yet unknown transposition-based mechanism.

The ‘Wijcik’ insertion occurred in an intergenic region and therefore did not disrupt the coding region of any gene. Moreover, the insertion itself does not contain any open reading frame, raising the question on how the insertion causes the columnar phenotype of ‘Wijcik’. One possible explanation is that the insertion influences the expression of neighboring genes. To verify this hypothesis, we tested by quantitative reverse transcription polymerase chain reaction (qRT-PCR) the expression of all genes located 25 kb upstream and downstream of the insertion. Unlike standard trees, most of the buds present on columnar stems develop into short spurs instead of lateral branches, suggesting that regulation of the columnar phenotype takes place at an early stage of bud development. For this reason, young buds at the green tip stage were selected for expression analysis and leaf tissue was used as a control. Among the six genes located within the 50 kb region containing the insertion (Fig. 2a), only MdCo31 (NCBI Accession number KF562006) was found to be differentially expressed, showing a 14-fold induction in ‘Wijcik’ buds compared to ‘McIntosh’, whereas the mRNA levels in leaves of both genotypes were negligible (Fig. 2b). This suggests an involvement of MdCo31 in controlling the columnar phenotype. In addition, we checked the expression of all transcription factors predicted inside the complete 386 kb Co region (MdCo4, MdCo16, MdCo25 and MdCo26) (Table S2), as they could have a potential role in plant development. No differences between ‘McIntosh’ and ‘Wijcik’ trees were found (data not shown), therefore excluding a role for such genes in controlling columnar phenotype.

Figure 2.

Expression levels of apple (Malus × domestica) genes around the insertion region. (a) ‘McIntosh’ and ‘Wijcik’ genomic regions containing the insertion and structure of MdCo31. The conserved DIOX-N superfamily and 20G-Fe(II) oxygenase domains together with the predicted exons are shown (gray boxes). The position of the predicted genes within the region is indicated by arrows. (b) Gene expression analysis by qRT-PCR of the predicted genes within the candidate region in bud and leaf material from ‘McIntosh’ and ‘Wijcik’. Gene expression in ‘Wijcik’ (red bars) is normalized against the expression of the corresponding gene in ‘McIntosh’ (blue bars). Error bars, ± SD for three biological replicates. Samples for which the expression level is below the limit for reliable quantification are indicated by nd (not detected).

The increased expression of MdCo31 in axillary buds of ‘Wijcik’ is in agreement with the dominant inheritance of the columnar phenotype. Moreover, the fact that the expression of MdCo31 seems to be bud-specific is in accordance with the supposed function of controlling shoot development. Additionally, a strict correlation between the presence of the insertion and an increased expression of MdCo31 in columnar trees compared to standard trees of a segregating population derived from a ‘Golden Delicious’ × ‘Wijcik’ cross confirmed that high expression of MdCo31 in buds is a general feature of columnar trees and not a peculiar feature of ‘Wijcik’ (Fig. 3). The 5′ CR and TC/CR 3′ PCR products could be used as molecular markers in marker-assisted selection for columnar-type growth in apple. Such markers could be used to predict columnar growth in segregating progenies with absolute accuracy, and would enable apple breeders to identify plants carrying the Co locus from ‘Wijcik’ even when offspring of ‘McIntosh’ were involved in the cross.

Figure 3.

Detection of insertion and MdCo31 expression in standard and columnar progeny apple (Malus × domestica) trees. Expression of MdCo31 in buds of standard (blue bars) and columnar apple trees (red bars). Error bars, ± SD for three biological replicates. Specific amplification of the insertion produced a PCR product (5′CR) for columnar progeny trees only. The CR/TC fragment is amplified in all trees and shown as a control for the PCR reaction.

A search for domains contained in the MdCo31 protein revealed two significant matches: a non-haem dioxygenase in morphine synthesis N-terminal (DIOX-N) motif and a 2-oxoglutarate and Fe(II)-dependent oxygenase (2OG-Fe(II) oxygenase) domain (Fig. 2a). Members of the gene family containing these domains catalyze the oxidation of organic substrates and were shown to be involved in the biosynthesis of ethylene, flavonoids, gibberellins, and defense against downy mildew (Prescott & John, 1996; van Damme et al., 2008). Interestingly, a poplar tree with a columnar-like habit called STUMPY was identified during an activation tagging screen that caused a constitutive expression of a GA2-oxidase involved in gibberellin catabolism, resulting in reduced levels of gibberellin (Busov et al., 2003). Alternatively, flavonoids have been shown to inhibit auxin transport, a mechanism that is thought to be required for axillary bud outgrowth and involved in the high apical dominance typical of columnar trees (Brown et al., 2001; Prusinkiewicz et al., 2009; Petersen & Krost, 2013). Some insights towards the possible role of gibberellin in conferring a columnar-like phenotype were already proposed by two recent transcriptome studies, however, none of the described candidate genes corresponds to MdCo31 (Zhang et al., 2012; Krost et al., 2013).

Although a role of MdCo31 in gibberellin metabolism or flavonoid biosynthesis could possibly explain the phenotype of ‘Wijcik’, it does not provide in itself a causal link between the overexpression of MdCo31 and the columnar habit. Transformation of apple, and fruit trees in general, requires a long regeneration time. Moreover, phenotypes related to tree architecture are visible only after several years. Therefore, we chose to use Arabidopsis in order to characterize the role of MdCo31 in plant development. The coding sequence of MdCo31 was expressed in Arabidopsis thaliana in a constitutive manner. Thirty-four hygromycin-resistant transformants were obtained and three independent T3 segregating lines named Pro35S:MdCo31#1 to #3 were selected for further analysis (Supporting Information, Fig. S4). All three lines expressed MdCo31 RNA at high levels and showed a rather normal leaf phenotype but later, at maturity, displayed short inflorescence stems (Fig. 4–a–c). In all three lines the observed marked reduction in overall height was primarily due to a c. three-fold shortening (< 0.01) in floral internodes, a phenotype reminiscent of the one observed in columnar plants (Fig. 4d). Overall, our results strongly support the hypothesis that MdCo31 is responsible for the columnar habit.

Figure 4.

Phenotypes of Arabidopsis plants overexpressing MdCo31. (a) Top view of rosettes of 4-wk-old Arabidopsis Col-0 and three independent Pro35S:MdCo31 lines. Bar, 2.5 cm. (b) Side view of the same Arabidopsis plants 3 wk later. White arrows indicate two consecutive floral nodes for each plant. Bar, 5 cm. (c) Expression level of MdCo31 in Pro35S:MdCo31 lines compared to Col-0. Error bars, + SD. (d) Floral internode length. The average internode distance was calculated for 10 inflorescence stems for each Arabidopsis line. Error bars, + SE of the mean. Asterisks indicate a significant difference from Col-0 (Student's t test, < 0.01).

In summary, our work describes the identification of a 1956 bp insertion in the Co region of ‘Wijcik’ that correlates with the overexpression of MdCo31, coding for a putative 2OG-Fe(II) oxygenase. The overexpression of MdCo31 in ‘Wijcik’ buds provides a plausible explanation for the dominant gain-of-function of the Co gene. However, how the insertion influences the expression level of MdCo31 remains unclear and needs to be investigated in the future. Nevertheless, a further line of evidence supporting the involvement of MdCo31 in controlling plant internode length was provided by transformation experiments using Arabidopsis. Taken together, the results presented in this paper let us conclude that MdCo31 is a very reliable candidate gene for controlling the columnar habit in apple.


The authors thank Pierluigi Magnago for providing and maintaining the ‘McIntosh’ and ‘Wijcik’ trees as well as the segregating population. The authors also thank the FEM sequencing platform at CRI for sequencing the BAC clones. They thank the Autonomous Province of Trento (PAT) for financial support to A.S.A. (GBPF GF) and the GMPF PhD school for funding the scholarship of P.J.W. They also thank the Technological Top Institute Green Genetics in collaboration with Inova Fruit BV for funding H.J.S. The authors thank Daniel James Sargent, and Nada Šurbanovski for their critical comments