Molecular Mechanisms Regulating the Columnar Tree Architecture in Apple

Abstract

The columnar apple cultivar ‘McIntosh Wijcik’ was discovered as a spontaneous mutant from the top of a ‘McIntosh’ tree in the early 1960s. ‘McIntosh Wijcik’ exhibits the columnar growth phenotype: compact and sturdy growth, short internodes, and very few lateral shoots. Classical genetic analysis revealed that the columnar growth phenotype of ‘McIntosh Wijcik’ is controlled by a single dominant gene, Co. This review focuses on the advances made toward understanding the molecular mechanisms of columnar growth in the last decade. Molecular studies have shown that an 8.2 kb insertion in the intergenic region of the Co locus is responsible for the columnar growth phenotype of ‘McIntosh Wijcik’, implying that the insertion affects the expression patterns of adjacent genes. Among the candidate genes in the Co region, the expression pattern of MdDOX-Co, putatively encoding 2-oxoglutarate-dependent dioxygenase (DOX), was found to vary between columnar and non-columnar apples. Recent studies have found three functions of MdDOX-Co: facilitating bioactive gibberellin deficiency, increasing strigolactone levels, and positively regulating abscisic acid levels. Consequently, changes in these plant hormone levels caused by the ectopic expression of MdDOX-Co in the aerial organs of ‘McIntosh Wijcik’ can lead to dwarf trees with fewer lateral branches. These findings will contribute to the breeding and cultivation of new columnar apple cultivars with improved fruit quality.

1. Introduction

Tree architecture influences various aspects of fruit production and orchard management [1], such as planting density, fruit quality and yield, pruning and training, fruit thinning and harvesting, and pesticide spraying. Tree architecture is regulated by four processes: primary growth, branching patterns, flowering location, and meristem and shoot mortality [1]. Recently, molecular mechanisms regulating tree architecture have been extensively studied [2,3,4].

Apple (Malus × domestica Borkh.) trees have been classified under four tree architectural types (types I to IV) according to both the overall tree growth pattern and their fruiting type: type I (columnar type), type II (spur type), type III (standard type), and type IV (tip-bearing type; this differs from the genetic weeping type [5]) (Figure 1A–E) [1]. Columnar apples (type I) display multiple distinct traits: compact and sturdy growth, short internodes, very few lateral shoots, many spurs, and large and thick leaves with high chlorophyll content (Figure 1A,B,F) [6,7,8]. Consequently, the tree grows naturally as a column.

The columnar architecture of apple trees was first discovered in ‘McIntosh Wijcik’. ‘McIntosh Wijcik’ was identified as a spontaneous mutant arising from the top of a 50-year-old ‘McIntosh’ tree in the early 1960s. It exhibits very slow growth, a negligible number of side shoots, compact and upright growth habits, and biennial bearing [11,12]. Genetic analysis revealed that the columnar growth phenotype of ‘McIntosh Wijcik’ is controlled by a single dominant gene, Co [6]. In addition, a possible role of certain modifier genes was suspected because the percentage of compact seedlings obtained from the test crosses was consistently lower than expected [6,13]. Since dominant genes controlling tree architecture in apple are valuable, ‘McIntosh Wijcik’ has been used not only as an important genetic resource to develop compact cultivars for high-density orchards but also as a model system to elucidate the mechanism of apple tree architecture determination [13].

The columnar growth phenotype permits high-density planting, minimal pruning, and mechanical harvesting [2,8]. Columnar apples can also function as space-saving pollinizers in orchards of a single cultivar because apples exhibit self-incompatibility [2,14]. The agronomic characteristics of columnar apple cultivars have been studied from the perspective of fruit production. Owing to the suppressed growth of side branches, columnar apples can be planted at 1 m intervals [11]. Moreover, Inomata et al. [15] compared the fruit productivity of ‘Maypole’ columnar apple trees which were trained in a central leader system at 0.66 m intervals with that of trees trained in a Y-trellis system at 1.14 m intervals. They concluded that a Y-trellis system was more advantageous than a central leader system because apple trees trained in a Y-trellis system maintained a lower fruiting position and displayed higher dry matter and fruit productivity. However, columnar apple trees generally suffer from susceptibility to biennial bearings [11]. Blazek and Krelinova [16] reported biennial bearing in five new columnar apple cultivars, most likely because of the proximately between the Co locus and biennial bearing quantitative trait loci (QTLs) in the shared linkage group [17]. Iwanami et al. [18] proposed a new labor-saving method for processing columnar apple trees using a thinning strategy for biennial bearings.

Recently, significant advances have been made toward the understanding of the molecular mechanisms of columnar growth in apple trees. In this review, we describe the genetic, molecular, physiological, and biochemical features of columnar apples elucidated over the last decade.

2. Fine Mapping of the Co Locus

Previously, the Co locus was mapped on linkage group 10 of ‘McIntosh Wijcik’ [19]. However, the causative mutation and Co candidate genes remain unknown. Since the release of a high-quality draft genome sequence of the apple cultivar Golden Delicious by Velasco et al. [20], several research groups have performed fine mapping of the Co locus using novel DNA markers developed from apple genome sequence information (Figure 2A). Firstly, Bai et al. [21] delimited the Co locus between markers C1753-3520 and C7629-22009, and the physical size was estimated to be 193 kb in the ‘Golden Delicious’ genome. Thereafter, Moriya et al. [22] restricted the Co locus between markers Mdo.chr10.11 and Mdo.chr10.15, with a size of 196 kb in the apple genome. Third, Baldi et al. [23] also narrowed down the Co locus between markers Co04R10 and Co04R13, and the determined size was 393 kb in the apple genome. Fourth, Morimoto and Banno [24] defined the Co locus between markers LG10-Co-N and C7629_12936 as 530 kb apart in the apple genome. Finally, Okada et al. [25] further delimited the Co locus between the markers Mdo.chr10.11-2 and Mdo.chr10.13-2, with a distance of 101 kb between them in the apple genome. From these results, the position of the Co locus may be deduced between the markers Co04R10 and C7629-22009, and the physical size of this region (genomic coordinates: 18.51–19.10 Mb) on chromosome 10 of the ‘Golden Delicious’ genome is estimated to be ca. 590 kb.

3. Identification of Mutation in ‘McIntosh Wijcik’

To identify the mutation responsible for the tree architecture of ‘McIntosh Wijcik’, three research groups used a positional cloning approach. Wolters et al. [26,27] sequenced BAC clones encompassing the Co region of ‘McIntosh Wijcik’ (Co/co) and the co region of ‘McIntosh’ (co/co). Similarly, Otto et al. [28] cloned and sequenced the Co region from the columnar cultivar Procats 28 (Co/co) and the homologous genomic regions of ‘McIntosh’ and ‘McIntosh Wijcik’. Okada et al. [25] also sequenced BAC clones containing the Co region of the columnar cultivar Telamon (Co/co) and the co region of ‘McIntosh’. Three research groups have unanimously reported that an 8.2 kb insertion in the Co region is the only genomic difference between the Co and co regions (Figure 2B). Therefore, the columnar growth phenotype in ‘McIntosh Wijcik’ is attributed to this insertional mutation.

The insertion consists of two long terminal repeats (LTRs) of 1951 bp each, a primer binding site (PBS), and a polypurine tract (PPT) required for transposition, but lacks the typical ORFs encoding the group-specific antigen (Gag), reverse transcriptase, integrase, or RNase H [25,28], exhibiting the characteristics of a non-autonomous LTR retroposon (Figure 2C). The 100% sequence identity between the two LTRs indicates that the insertion sequence had translocated recently, supporting that the mutation in ‘McIntosh Wijcik’ occurred approximately 60 years ago [28].

4. Exploration of Co Candidate Genes

Because the insertional mutation in ‘McIntosh Wijcik’ was found in an intergenic region instead of the coding region, it was hypothesized that the expression patterns of adjacent genes may be affected rather than the disruption of the coding region of any gene [26]. To identify the Co candidate gene, Wolters et al. [26] analyzed the expression levels of six genes (MdCo27–MdCo32) predicted in the 50 kb region encompassing the insertion sequence using quantitative reverse transcription PCR. They revealed that only MdCo31 was upregulated (14-fold) in the axillary buds of ‘McIntosh Wijcik’ compared to ‘McIntosh’. However, the expression of MdCo31 was not observed in the leaves of both ‘McIntosh Wijcik’ and ‘McIntosh’.

Otto et al. [28] and Petersen et al. [29] mapped the RNA-seq data obtained from shoot apical meristems (non-columnar cultivar A14-190-93 and columnar cultivar Procats 28), leaves (‘McIntosh’ and ‘McIntosh Wijcik’), and primary roots (non-columnar, heterozygous columnar, and homozygous columnar apples) to the Co region. Furthermore, they analyzed the expression levels of five genes located in the vicinity of the insertion, as well as that of the insertion sequence, using RNA-seq and quantitative real-time PCR. The expression levels of MDP0000927098 (ATL5K-like), MDP0000163720 (ACC1-like), and the insertion sequence were elevated in the shoot apical meristems of the columnar apples. Moreover, downy mildew resistance 6-like (dmr6-like) was strongly upregulated in the shoot apical meristems and newly developing leaves of columnar apples. However, it was downregulated in the primary roots of heterozygous columnar apples and remained suppressed in the young but fully developed leaves of both cultivars. At1g08530-like and MDP0000934866 (At1g06150-like) showed similar expression levels in all tissues of both the columnar and non-columnar apples.

Okada et al. [25] mapped the RNA-seq data of shoot apices of ‘McIntosh Wijcik’ and ‘McIntosh’ to the Co region and analyzed differentially expressed genes. Five contigs (12053, 41231, 38029, 44905, and 91071-genes) were upregulated in ‘McIntosh Wijcik’, consistent with the fact that the Co gene is dominant. In contrast, a single contig (18023-gene) was downregulated in ‘McIntosh Wijcik’. Homology searches using BLAST showed that, among the six contigs, the 91071-gene was the most likely candidate for the Co gene. Furthermore, Okada et al. [25] showed that the three genes (MDP0000927098, MDP0000163720, and MDP0000934866) identified previously by Otto et al. [28] and Petersen et al. [29] did not show differential expression between ‘McIntosh Wijcik’ and ‘McIntosh’.

Overall, a common Co candidate gene was identified by three research groups and given different names: MdCo31 [26], dmr6-like [29], and 91071-gene [25]. Therefore, the MdCo31/dmr6-like/91071-gene is the strongest candidate for the Co gene. In this review, the designation ‘MdDOX-Co’ [30] is used instead of the aforementioned names.

5. Expression Analysis of MdDOX-Co

In non-columnar apples, MdDOX-Co was primarily expressed in the roots, whereas no or negligible expression was noted in the shoot apices, axillary buds, and leaves. In contrast, columnar apples expressed MdDOX-Co in the roots, shoot apices, axillary buds, and leaves [26,28,29,31,32,33,34]. In situ hybridization showed that MdDOX-Co is expressed in the growing root tips and lateral root primordium of both non-columnar and columnar apples and in the shoot meristem and leaf primordium of ‘McIntosh Wijcik’ [31]. A negligible difference was observed in the expression levels and patterns of MdDOX-Co between the roots of non-columnar and columnar apples [31]. In addition, the normal tree architecture of non-columnar scions grafted onto columnar rootstocks indicated that the columnar growth phenotype was not transmissible from rootstock to scion [31]. These results suggest that ectopic expression of MdDOX-Co in aerial organs (shoot apices, axillary buds, and leaves) is responsible for columnar growth, whereas its expression in the roots is not associated with columnar growth.

6. Phenotypes of Transgenic Plants Overexpressing MdDOX-Co

To investigate the effects of MdDOX-Co on phenotype, transgenic plants overexpressing MdDOX-Co were generated. Arabidopsis plants overexpressing MdDOX-Co displayed compact plants with dark green leaves and short floral internodes [26,35]. Moreover, tobacco plants overexpressing MdDOX-Co showed decreased plant height and internode length, thick and wrinkled leaves with high chlorophyll content, and delayed flowering [25,30,32,33,34]. Similarly, apples overexpressing MdDOX-Co exhibited short internodes and suppressed upward growth [25]. These results confirm that MdDOX-Co expression leads to short plant height and internode length in various plants, and the short internodes of columnar apples result from the upregulated expression of MdDOX-Co.

7. Characterization of MdDOX-Co

MdDOX-Co encodes a putative 2-oxoglutarate-dependent dioxygenase (DOX) consisting of 339 amino acids that belongs to the DOXC class of the DOX superfamily. The members of this family participate in various oxygenation/hydroxylation reactions in plants [25,26,36]. Transient expression in Arabidopsis cells demonstrated that the MdDOX-Co-GFP fusion protein was specifically localized in the cytoplasm [34]. Phylogenetic analysis of DOXs classified MdDOX-Co into the DOXC41 clade along with hyoscyamine 6β-hydroxylase, which catalyzes the hydroxylation of hyoscyamine, and Hordeum vulgare iron-deficiency-specific clones 2 and 3, involved in the biosynthesis of mugineic acid [25,36,37,38]. Because these phytochemicals are apparently unrelated to tree architecture, the mechanism by which MdDOX-Co influences the columnar growth phenotype remains unclear.

8. Functions of MdDOX-Co and the Mechanisms of the Columnar Growth Phenotype

Recently, three functions have been proposed for MdDOX-Co:

(1) MdDOX-Co causes bioactive gibberellin (GA) deficiency.

Okada et al. [30] closely examined tobacco plants overexpressing MdDOX-Co, which showed a typical dwarf phenotype (short plant height and internode length, wide leaf shape, and dark green wrinkled leaves) and were similar to the GA- or brassinosteroid (BR)-deficient/signaling mutants. GA is a plant hormone involved in stem elongation, seed germination, and flowering [39], whereas BR is related to cell elongation, cell division, and cell differentiation [40]. The dwarf phenotypes of transgenic tobacco plants were restored to traits similar to the wild-type plants after the application of gibberellin A₃ (GA₃), but not after the application of brassinolide [30,34]. Similarly, ‘McIntosh Wijcik’ showed dwarf traits (short main shoot and internode length), which were partially reversed by GA₃ application. Interestingly, GA₃ treatment of young apple trees also increases the number of lateral branches and reduces the number of flower buds [30]. Furthermore, reduced endogenous concentrations of bioactive GAs (GA₁ and/or GA₄) were observed in transgenic tobacco plants and columnar apples compared with those in wild-type tobacco plants and non-columnar apples, respectively [30,34]. These results suggest that ectopic expression of MdDOX-Co in the transgenic tobacco plants and the aerial organs of columnar apples causes a dwarf phenotype, which is mediated by bioactive GA deficiency, and that GA plays an important role in controlling apple tree architecture by promoting vegetative growth (shoot and internode elongation and lateral branch formation) and inhibiting reproductive growth (flower bud formation).

Recently, Watanabe et al. [35] demonstrated that recombinant MdDOX-Co metabolizes GA₁₂, GA₉, and GA₄ to GA₁₁₁, GA₇₀, and GA₅₈, respectively (Figure 3). In addition, exogenous application of GA₁₂ to Arabidopsis plants overexpressing MdDOX-Co produced GA₁₁₁, but not GA₇₀, GA₅₈, GA₉, or GA₄. They also confirmed a lower efficiency of conversion of GA₁₁₁ to GA₇₀ by recombinant Arabidopsis GA 20-oxidases than that of the conversion of GA₁₂ to GA₉. These data indicate that the conversion of GA₁₂ to GA₁₁₁ by MdDOX-Co blocks the pathway of production of biologically active GAs (GA₄ and GA₅₈).

Okada et al. [30] and Watanabe et al. [35] proposed that the bioactive GA deficiency caused by the ectopic expression of MdDOX-Co in the aerial organs of ‘McIntosh Wijcik’ can lead to early growth cessation and short internodes in both the main and side shoots. This possibly resulted in the formation of a dwarf tree with very few lateral branches and numerous spurs. Consequently, the synergistic action of these pleiotropic traits manifests as a columnar tree form.

(2) MdDOX-Co increases strigolactone (SL) content.

SLs are plant hormones implicated in the inhibition of bud outgrowth and shoot branching [41]. To study the relationship between SLs and the columnar growth phenotype, Sun et al. [32] characterized the expression profiles of SL biosynthesis- and signal transduction-related genes in columnar and non-columnar apples. Expression levels of the major genes involved in SL biosynthesis, including the MORE AXILLARY GROWTH genes (MdMAX3-1 and MdMAX4-4) and a DWARF gene (MdD27-1), were higher in both buds and shoots of columnar apples than in the corresponding tissues of non-columnar apples. Expression level of the DWARF gene MdD53-4, which represses SL signal transduction, was lower in columnar apples. In addition, tobacco plants overexpressing MdDOX-Co showed a higher expression level of NbMAX3 and lower expression level of NbD53 than wild-type plants. Because the SL content in columnar apples was higher than that in non-columnar apples, it may be inferred that MdDOX-Co increased the SL content, weakened the inhibition of SL signal transduction, and inhibited lateral branching in columnar apples.

(3) MdDOX-Co positively regulates abscisic acid (ABA) biosynthesis and enhances salt tolerance.

ABA is a stress-response hormone that inhibits shoot and lateral bud elongation [33]. The ABA content in shoots of ‘McIntosh Wijcik’ was significantly higher than that in shoots of ‘McIntosh’, and the ABA content of 35S:MdDOX-Co transgenic apple calli was also higher than that of wild-type (non-transgenic non-columnar apple) calli [33]. Furthermore, the expression levels of the major ABA biosynthesis genes, MdNCED6 and MdNCED9 (MdNCED6/9), were significantly higher in the ‘McIntosh Wijcik’ and 35S:MdDOX-Co apple calli than in the ‘McIntosh’ and wild-type calli, respectively. Sun et al. [33] further indicated that MdDOX-Co forms a protein complex (MdDOX-Co–MdDREB2) with transcription factor MdDREB2. MdDREB2 directly binds to cis-elements in the MdNCED6/9 promoters, thereby functioning as a transcriptional activator. The expression levels of MdNCED6/9 and the ABA content were higher in transgenic apple calli co-overexpressing MdDOX-Co and MdDREB2 than in transgenic plants overexpressing MdDOX-Co or MdDREB2 independently [33]. These results suggest that the MdDOX-Co–MdDREB2 complex promotes ABA biosynthesis by upregulating the expression of MdNCED6/9. Thus, MdDOX-Co plays a positive role in ABA biosynthesis, and the higher ABA content in columnar apples may lead to two effects: (1) ABA directly reduces the elongation of lateral buds, resulting in a few lateral branches, and (2) ABA suppresses the effect of GA, with consequent inhibition of branch and internode growth, ultimately leading to a dwarf phenotype [33].

Sun et al. [42] also found that the expression level of MdDOX-Co increased remarkably in ‘McIntosh Wijcik’ under salt stress. In addition, shoot cultures of ‘McIntosh Wijcik’ exhibited higher salt tolerance than those of ‘McIntosh’. Transgenic tobacco and apple calli overexpressing MdDOX-Co also displayed enhanced salt tolerance, higher superoxide dismutase activity, and lower malondialdehyde levels than wild-type plants under salt stress [42]. Thus, MdDOX-Co was concluded to confer salt tolerance.

9. Modifier Genes and Other Genes Involved in the Columnar Growth Phenotype

Dougherty et al. [43] identified two recessive loci (c2 and c3) that can suppress the columnar growth phenotype, and c2 appeared to have a more prominent effect in younger (2-year-old) trees than in older (8-year-old) trees. The c2 locus is located on chromosome 10 and the c3 locus is located on chromosome 9, and these two loci were suggested to repress the columnar growth phenotype through additive gene interactions. Trees with a repressed columnar growth phenotype (phenotype, standard growth; genotype, Coco) showed a drastic reduction in the expression level of MdDOX-Co.

Transcriptome analyses also revealed that many differentially expressed genes were associated with the columnar growth phenotype. Zhang et al. [44] identified 5237 differentially expressed genes between newly developing shoots of columnar and non-columnar apples using RNA-seq. Among the 5237 unigenes, the gene ontology functional annotation and KEGG pathway database identified 287 unigenes related to plant architecture formation. Among the 287 unigenes, 106 were GRAS transcription factors, suggesting that GRAS transcription factors play an important role in regulating the architecture of apple trees.

Similarly, Krost et al. [45,46] compared the transcriptomes of shoot apical meristems of columnar apple ‘Procats 28′ and standard apple ‘A14-190-93′ using RNA-seq. Genes that were categorized into cell wall modification, transport, and protein modification were upregulated in the columnar apple, whereas genes that were grouped into light reactions, mitochondrial electron transport, lipid metabolism, cell wall, DNA synthesis, RNA processing, and protein synthesis were downregulated in the columnar apple. Furthermore, 16 plant hormone-associated genes were differentially regulated: indole-3-acetic acid (6 genes), cytokinin (3 genes), ABA (3 genes), BR (2 genes), GA (1 gene), and jasmonic acid (1 gene). Their regulation probably leads to an increase in the endogenous bioactive indole-3-acetic acid, cytokinin, BR, GA, and jasmonic acid in the columnar apple [46].

Otto et al. [28] also mapped the RNA-seq data of leaves from ‘McIntosh’ and ‘McIntosh Wijcik’ to the ‘Golden Delicious’ genome. A total of 5751 of 9961 unigenes were differentially expressed between ‘McIntosh’ and ‘McIntosh Wijcik’. Genes that are involved in secondary metabolism (especially lignin and terpene biosynthesis), metabolism and/or signaling of plant hormones (auxin, jasmonate, and ethylene), glutathione-S-transferases, and proteins that are involved in defense or stress reactions (such as pathogen recognition receptors and heat shock proteins) were upregulated in leaves of ‘McIntosh Wijcik’. In contrast, genes that are associated with photosynthesis, protein biosynthesis, and nucleotide metabolism and enzymes managing the redox state (such as thioredoxin, dismutase/catalase, and peroxidase) were downregulated in leaves of ‘McIntosh Wijcik’.

10. Marker-Assisted Selection (MAS) Systems for Selecting Columnar Apples

‘McIntosh Wijcik’ serves as an important genetic resource for breeding columnar apple cultivars and has been used in several breeding programs at the East Malling Research Station that were initiated in the early 1970s [47]. Consequently, four columnar cultivars were released for amateur gardeners: an ornamental apple, ‘Maypole’ (‘McIntosh Wijcik’ × ‘Baskatong’), and three dessert apples, ‘Telamon’ (‘McIntosh Wijcik’ × ‘Golden Delicious’), ‘Trajan’ (‘Golden Delicious’ × ‘McIntosh Wijcik’), and ‘Tuscan’ (‘McIntosh Wijcik’ × ‘Greensleeves’) [12]. However, outstanding commercial columnar cultivars with good fruit quality have not yet been developed, and ongoing breeding programs are aimed at producing such cultivars.

It is difficult to differentiate between columnar and non-columnar apples phenotypically until the seedlings are 2–3 years old [48]. Therefore, MAS systems, which can be applied to several-week-old seedlings, facilitate the selection of columnar apples. Conventional SSR markers linked to the Co gene were unable to distinguish the Co allele from the original wild-type co allele of ‘McIntosh’ [22,25]. Hence, DNA markers specific to the Co allele have been developed based on the insertion polymorphism. Wolters et al. [26] and Otto et al. [28] performed PCR with primer pairs spanning the left or right borders of the 8.2 kb insertion and obtained a fragment of the expected size, specifically from the genomic DNA of columnar cultivars, which was absent in non-columnar cultivars. However, these DNA markers could not discriminate between heterozygous columnar apples (Co/co) and homozygous columnar apples (Co/Co) when a single PCR was performed. Okada et al. [25] and Cmejlova et al. [48] conducted PCRs using three primers designed from the insertion polymorphism and simultaneously identified homozygous columnar apples (Co/Co), heterozygous columnar apples (Co/co), and non-columnar apples (co/co) through a single PCR. These Co allele-specific PCR products are useful DNA markers in the MAS of columnar apples for segregating progenies even when ‘McIntosh’ or its offspring are used as a parent because it can distinguish the Co allele from the original wild-type co allele [25,26].

11. Conclusions and Perspectives

Extensive research in the last decade has successfully identified the causative mutation (insertion of an 8.2 kb LTR retroposon) and gene (MdDOX-Co) for the columnar growth phenotype. Interestingly, MdDOX-Co is involved in GA metabolism [30,34,35], but it is categorized into a different phylogenetic clade (DOXC41) from those containing key enzymes associated with GA biosynthesis and degradation: GA 3-oxidase (DOXC3), GA 20-oxidase (DOXC7), C₁₉-GA 2-oxidase (DOXC12), C₂₀-GA 2-oxidase (DOXC13), and GA 7-oxidase (DOXC22) [25,35,36]. This suggests that MdDOX-Co is a unique enzyme involved in GA metabolism. Furthermore, MdDOX-Co is associated with an increase in SL and ABA biosynthesis [32,33]. Therefore, how MdDOX-Co affects the columnar growth phenotype remains to be elucidated, and more detailed analyses are needed. It is important to investigate whether MdDOX-Co is an enzyme with multiple functions or whether the multiple functions result from crosstalk among plant hormones. In addition, how modifier genes (c2 and c3) reduce the expression level of MdDOX-Co needs to be elucidated in future studies.

Columnar apples have labor-saving properties for apple growers and fit well with sustainable high-density planting, resulting in higher yields per hectare. However, an important problem is that columnar apple cultivars with superior fruit quality have not yet been developed [48]. The columnar growth phenotype can also easily be combined with other desirable characteristics (e.g., disease resistance) by crossing because the columnar growth phenotype is controlled by the dominant Co gene [47]. Recent Co allele-specific DNA markers enable the efficient selection of seedlings with a columnar growth phenotype and will facilitate the breeding of new columnar apple cultivars with good fruit quality and other desirable characteristics.

Another important problem is the development of new cultivation systems suitable for the columnar apples, as the tree architecture of columnar apples is completely different from that of standard apples (Figure 1). New cultivation systems include mechanization, which leads to labor savings. Mechanical harvesters that can take advantage of the flat arrangement of branches on columnar apples are particularly promising. As a promising cultivation method to control tree size, exogenous GA₃ application is an efficient way to optimize tree height in columnar apples [30]. GA₃ application in the juvenile phase will promote tree growth in columnar apples, whereas the cessation of GA₃ application in the adult phase will maintain a desirable tree form that hardly needs pruning and training. Thus, the creation of a series of high quality new columnar apple cultivars, as well as innovative cultivation systems involving mechanization, may dramatically revolutionize future apple industry as in the case of the invention of high-yielding semi-dwarf varieties of wheat and rice that led to the “Green Revolution” [49].