Introduction

Genetic control of plant architecture in herbaceous plants

The architecture of a plant defines the spatial arrangement of its individual (aerial) organs. With reference to the biological dogma of form and function being intrinsically tied to each other, studying plant architecture is crucial for the comprehension of plant function. Furthermore, controlling plant architecture has always been an important aim in plant breeding, as manipulating plants to grow to a smaller height and produce shorter branches, while at the same time maximizing yield, bears great economic advantages. This has been proven by the introduction of lodging-resistant semi-dwarf wheat and rice mutants that led to the Green Revolution in the 1960s (Peng et al. 1999) and by the adoption of dwarfing rootstocks in fruit tree breeding in the 1920s (Fideghelli et al. 2003).

Even though plant architecture is influenced by environmental factors such as light penetration, temperature, humidity and soil conditions including nutrient availability, the intrinsic body plan of the plant is genetically determined. During the past few years, significant progress has been made in the disclosure of genes that play a pivotal role in establishing the shape of the plant, mostly by examining herbaceous plants showing an aberrant architecture due to a mutation. This has been extensively reviewed elsewhere (Reinhardt and Kuhlemeier 2001; Wang and Li 2008), so only a few key players and basic concepts which can be applied to most plant species are presented here (Fig. 1). For clarity, Arabidopsis nomenclature is used for genes and proteins unless indicated otherwise.

Fig. 1
figure 1

Major aspects of plant architecture regulation. The shape of a dicotyledonous plant is regulated by some key proteins (left-hand side) and phytohormones (right-hand side) that show promotive (arrows) or inhibitory (bar-headed arrow) effects on shoot apical meristem activity (top), floral meristem activity, inflorescence branching and vegetative branching (bottom) as well as on elongation growth (double-headed vertical arrow) and secondary growth (double-headed horizontal arrow). For detailed explanations see text. ABA abscisic acid, BR brassinosteroid, CK cytokinin, CUC cup-shaped cotyledon, GA gibberellin, IAA indole-3-acetic acid, KNOX knotted-1-like homeobox, LAS lateral suppressor, Phab phabulosa, Phav phavoluta, SL strigolactone, STM shoot meristemless

Full size image

In the shoot apical meristem, a feedback loop between wuschel and clavata controls the balance between stem cell maintenance and organogenesis: wuschel (Laux et al. 1996; Mayer et al. 1998) maintains undifferentiated cells within the central zone and activates clavata 3 (Clark et al. 1995), whose gene product in turn together with Clavata 1 and Clavata 2 (Clark et al. 1993) restricts wuschel expression to a defined region, thus enabling cells in the peripheral zone to undergo differentiation (Brand et al. 2000; Schoof et al. 2000). The products of the knotted-1-like homeobox (KNOX) genes, with shoot meristemless as their best known member, are also responsible for stem cell maintenance (Long et al. 1996; Hay and Tsiantis 2010). Furthermore, KNOX proteins define boundaries of newly formed organs via interactions with cup-shaped cotyledon proteins (Aida et al. 1999).

When lateral organ primordia are initiated, they are arranged in distinct phyllotactic patterns. These patterns are mainly controlled by auxin levels, as the primordia are induced by a local auxin maximum (Reinhardt et al. 2000; Benková et al. 2003). Therefore, genes regulating phyllotaxis are either genes establishing the organization of the shoot apical meristem or genes involved in auxin transport and signaling like pin-formed 1 (Reinhardt and Kuhlemeier 2001). Only mutants of terminal ear 1 (Veit et al. 1998) in maize and of perianthia (Running and Meyerowitz 1996) in Arabidopsis show alterations in phyllotactic patterns without also showing changes in the apical meristem. In the leaf primordia, phan, phabulosa and phavoluta establish the adaxial cell fate (McConnell et al. 2001), while yabby and kanadi promote abaxial cell fates (Siegfried et al. 1999; Kerstetter et al. 2001). Once the leaf has been formed, its shape is controlled by angustifolia in the lateral direction and rotundifolia in the longitudinal direction (Tsuge et al. 1996; Tsukaya 2005); lanceolate as well as KNOX induce the formation of compound leaves (Mathan and Jenkins 1962; Hareven et al. 1996).

In the axils of the leaves, axillary meristems either develop from cells of the shoot apical meristem that have maintained their stem cell identity (Garrison 1955; Sussex 1955) or from differentiated cells that undergo dedifferentiation (Snow and Snow 1942). The initiation of axillary meristems is regulated by revoluta (Otsuga et al. 2001), lateral suppressor (Greb et al. 2003) and branched 1 (Aguilar-Martínez et al. 2007). For the maintenance of stem cell identity, shoot meristemless is induced by Cup-Shaped Cotyledon 1 (Hibara et al. 2003), indicating similar regulatory loops in axillary meristems as in the apical meristem. Axillary meristems generate axillary buds, which can remain dormant or grow out to form lateral shoots (Shimizu-sato and Mori 2001). This is regulated by apical dominance, a concept described below. In addition, the angle at which lateral shoots grow out largely contributes to overall plant architecture. The control of the crotch angle has not yet been thoroughly researched, but recently lazy 1 and tac 1 have been shown to play important roles as a negative and a positive regulator, respectively, of tiller angle in maize (Li et al. 2007; Yu et al. 2007).

The final important decision in a plant’s life is the phase change from vegetative to reproductive growth. For many herbaceous plants, producing a flower and subsequently fruit represents the end of their life cycle. This developmental switch therefore has to be precisely controlled; revoluta, knotted-1-like 1 and erecta are essential players in the regulation of inflorescence architecture (Douglas et al. 2002). In Arabidopsis, five genetically defined but cross talking pathways have been identified in the control of flowering (recently reviewed in Srikanth and Schmid 2011): the vernalization pathway, the photoperiod pathway, the gibberellin pathway, the autonomous pathway and the aging pathway. Focusing on a simplified view of the photoperiod and autonomous pathways, under long days constans (CO) mRNA is stabilized at the end of the day (with the help of phytochrome A and due to the influence of gigantea) and activates the transcription of flowering locus T (FT) and twin sister of FT (Kardailsky et al. 1999; Kobayashi et al. 1999; Samach et al. 2000; Suarez-Lopez et al. 2001; Valverde et al. 2004). FT moves through the phloem into the meristem and, together with flowering locus D, activates apetala 1, fruitful and suppressor of overexpression of constans (Abe et al. 2005; Wigge et al. 2005; Yoo et al. 2005). Floral meristem identity genes like apetala (Irish and Sussex 1990) and leafy (Weigel et al. 1992) transform indeterminate axillary meristems into determinate floral meristems. Leafy directly activates apetala 1 redundantly with FT and also induces the transcription of the homeotic genes agamous and apetala 3 (Busch et al. 1999; Lamb et al. 2002). The antagonist of leafy is terminal flower 1 (TFL1), which favors indeterminate vegetative growth (Shannon and Meeks-Wagner 1991).

While the individual parts of the plant are being built, intercalary meristems mediate elongation growth as well as secondary growth of the main axis and lateral organs. Elongation growth of the main axis occurs via cell divisions and subsequent internode elongation. Internode elongation is mainly hormonally controlled (see below), but regulators of cell division or cell wall remodeling and polyamine signals also influence cell elongation in the stem (Hanzawa et al. 2000; Zhou et al. 2006; Asano et al. 2010; Todaka et al. 2012). The increase in stem diameter is caused by the formation of secondary xylem and secondary phloem by the vascular cambium. This process seems to be regulated in essentially the same way as primary growth because it involves wuschel-related homeobox and clavata 3/ESR-related factors (Schrader et al. 2004; Hirakawa et al. 2010) and requires shoot meristemless and KNOX genes for the maintenance of stem cell identity (Groover et al. 2006; Du et al. 2009; Zhang et al. 2011).

Phytohormones regulating plant architecture in herbaceous plants

For a plant as a sessile organism, it is crucial to establish a stable architecture while at the same time maintaining the possibility to modify the spatial arrangement of its individual parts to adapt to environmental changes. Thus, plant architecture is dynamic and extensive signal integration and crosstalk constantly take place for its precise control. Most long-distance signaling within the plant is accomplished via the highly interconnected network of the different phytohormones, and the regulation of architecture is no exception to this rule (Fig. 1).

Auxins, the most important representative being indole-3-acetic acid (IAA), are involved in nearly all aspects of plant growth and development (reviewed in Overvoorde et al. 2010; Müller and Leyser 2011; Durbak et al. 2012). As mentioned above, auxin levels determine the sites of organ primordia formation (Reinhardt et al. 2000). They are also the predominant mediator of apical dominance, which is defined as the control exerted by the apical portions of the main shoot over the outgrowth of lateral buds (Cline 1991). The high auxin content in the shoot apex regulates ramification because it usually suppresses the outgrowth of lateral shoots so that the latter usually only take over in case the leader is damaged and its influence ceases (Thimann and Skoog 1933; Cline 1991), although outgrowth can also occur under “vigorous” growth conditions or on vigorous rootstocks. The regulation of apical dominance has been extensively discussed elsewhere (Leyser 2005; Rameau 2010; Domagalska and Leyser 2011) and will thus not be discussed in detail. Clearly, the polar basipetal IAA transport within the dominant stem, mediated via the asymmetric distribution of auxin influx carriers of the Auxin Resistant 1/Like Auxin Resistant family (Swarup et al. 2008; Péret et al. 2012; Swarup and Péret 2012) and auxin efflux carriers of the Pin-formed family (Gälweiler et al. 1998; Palme and Gälweiler 1999; Křeček et al. 2009), plays a central role in the establishment of apical dominance (Friml and Palme 2002). The polar auxin transport in the main shoot either hinders the lateral buds from establishing their own auxin flux which then prevents their outgrowth (Li and Bangerth 1999), or it regulates the levels of a second, upwardly moving messenger of bud sprouting (Snow 1929; Sachs and Thimann 1967). Cytokinins (CKs) and strigolactones (SLs) are the most likely candidates for this second messenger: direct CK application to the lateral bud promotes its outgrowth (Cline 1991), whereas SLs act as a branching inhibitor (Gomez-Roldan et al. 2008; Umehara et al. 2008). Furthermore, CK and SL levels have been shown to be regulated by IAA (Nordström et al. 2004; Johnson et al. 2006; Brewer et al. 2009). In turn, CK induces IAA biosynthesis (Jones et al. 2010), and genes involved in the biosynthesis and/or signaling of SLs act as negative regulators of polar auxin transport (Bennett et al. 2006). As a central component of plant development, auxin crosstalks with nearly all other phytohormones (reviewed in Chandler 2009). The interplay between auxin and SLs not only determines sprouting but also stimulates secondary growth (Agusti et al. 2011).

Besides auxin, CKs are essential for plant growth, since they generally are stimulators of cytokinesis (Hartig and Beck 2006). As such, they maintain the activity of the vegetative and floral shoot apical meristem (Riou-Khamlichi et al. 1999), regulate cambial activity (Matsumoto-Kitano et al. 2008; Nieminen et al. 2008) and antagonize the inhibitory effect of auxin on the outgrowth of lateral buds. Contrary to their promotive role for growth of the aerial parts of the plant, they negatively regulate root apical meristem activity and lateral root formation (Werner et al. 2003).

In addition to cytokinesis, cell elongation is essential for growth to take place. Elongation growth is mainly mediated via gibberellins (GAs) and brassinosteroids (BRs) (reviewed in Phinney 1985; Müssig 2005), which stimulate internode elongation (Yamamuro et al. 2000; Dayan et al. 2012; Li et al. 2012). Ethylene signaling has been shown to influence internode elongation in rice adapted to deep water conditions (Hattori et al. 2009; Qi et al. 2011), and IAA contributes to elongation growth in pea (McKay et al. 1994). Low GA levels in the shoot apical meristem together with the high CK/IAA ratio favor the maintenance of stem cells, whereas high GA and low CK/IAA ratio induce the formation of lateral organs (Shani et al. 2006). GAs and BRs also promote flowering (Langridge 1957; Wilson et al. 1992; Domagalska et al. 2010). This effect is antagonized by abscisic acid (ABA), which generally has an inhibitory role on plant growth (Milborrow 1967).

The remaining three groups of phytohormones are ethylene, jasmonates and salicylic acid; however, they only play a minor role in the regulation of plant shape, except under stress conditions (Xu et al. 1994; Zhang and Turner 2008; Sehr et al. 2010). In contrast, the past few years have shown that other molecules like microRNAs (miRNAs) highly influence plant architecture (Chen et al. 2010; Jiao et al. 2010). It is often not yet clear at which points and in which way phytohormone or miRNA pathways and the key genes for the regulation of plant architecture converge. The input from several signaling pathways is probably integrated to fine-tune a few final developmental switches.

Approaching tree architecture

To date, our understanding of plant architecture establishment as described above is mostly based on results obtained from the herbaceous model organism Arabidopsis thaliana and the economically important cereals such as rice or maize. In contrast, research on tree architecture has long been mainly descriptive (e.g., Ceulemans et al. 1990; Costes et al. 2006). However, during the past few years, the genus Populus (including poplars and aspen) has emerged as a model for the study of tree architecture and physiology, supported by the sequencing of its genome, which was the first tree genome to be completed (Tuskan et al. 2006; Wullschleger et al. 2013). As the number of available genome sequences of perennial plants continues to increase (Jaillon et al. 2007; Velasco et al. 2010; Shulaev et al. 2011; Verde et al. 2013), the unraveling of the regulation of tree architecture has been greatly accelerated. Most of the genes and mechanisms described in the previous sections have homologs and equivalents in woody plants. Sometimes two or more orthologs to each Arabidopsis gene can be found due to gene duplication, and these paralogous genes might show different expression patterns and fulfill slightly different roles. For instance, Populus has several wuschel and shoot meristemless orthologs, which in addition to regulating stem cell maintenance in the SAM also play a similar role in the vascular cambium (Schrader et al. 2004; Groover 2005; Groover et al. 2006; Bao et al. 2009). With regard to the regulation of flowering, Populus has two orthologs of FT (see below), TFL 1 (Mohamed et al. 2010) and agamous (Brunner et al. 2000), respectively, and in apple, two leafy orthologs with different expression patterns and thus possibly different functions can be detected (Wada et al. 2002). Functional differences of certain genes in Arabidopsis and trees have been reported, especially in the case of heterologous transformation (Gocal et al. 2001; Flachowsky et al. 2010).

Trees and herbaceous plants also share all key concepts of phytohormone regulation. In trees, IAA is involved in the control of apical dominance and apical control, acting in combination with CKs (Wilson 2000; Cline and Dong-Il 2002). For perennial plants, “apical dominance” only refers to the decision between outgrowth or bud formation of the current year’s axillary meristems (yielding sylleptic branches), whereas the term “apical control” is used to describe the influence of apical parts of the tree on the growth of lateral shoots and of previously dormant buds in subsequent years (yielding sylleptic and proleptic branches) (Brown et al. 1967; Cline 1997). Apical dominance and apical control are dynamic in time and can be modified in response to environmental effects. IAA also regulates secondary growth, which is more pronounced in trees because their longevity combined with the indeterminate growth of plants leads to a higher average plant size and biomass compared with annual plants, necessitating the reinforcement of the plant body. In this context, the formation of a radial auxin gradient with a peak in the cambium and the adjacent first few layers of xylem cells and its synergistic action with GA in wood formation have been intensively researched (Uggla et al. 1996, 1998; Eriksson et al. 2000; Israelsson et al. 2005; Björklund et al. 2007; Nilsson et al. 2008; Mauriat and Moritz 2009; Han et al. 2011; Chen et al. 2013). Furthermore, GAs affect elongation growth (Han et al. 2011; Elias et al. 2012) as well as flowering of perennial plants (Zawaski et al. 2011; Randoux et al. 2012). They also control seed dormancy together with ABA (reviewed in Graeber et al. 2012). ABA regulates responses to abiotic stresses, especially drought (Li et al. 2004; Popko et al. 2010; Ji et al. 2013).

To gain an advantage in the competition for light and nutrients during their first few years of life and to build up constructional and photosynthetically active organs before the formation of reproductive structures, most trees undergo a juvenile phase in which they cannot be induced to flower (Hackett 1985). This phase can last several years; for instance, poplar and apple flower for the first time after 7–10 and 4–8 years, respectively (Hackett et al. 1985; Hsu et al. 2006). The transition from the juvenile to the adult phase is regulated by the CO/FT regulatory module, similar to the photoperiod pathway of the transition from the vegetative to the reproductive phase in Arabidopsis described above. FT transcription gradually increases during the juvenile phase (Böhlenius et al. 2006), and Populus plants overexpressing the FT homologs FT1 or FT2 as well as plum plants transformed with poplar FT1 show an early flowering phenotype (Böhlenius et al. 2006; Hsu et al. 2006, 2011; Srinivasan et al. 2012). Downregulation of the Populus TFL homologs Populus Centroradialis 1 (PopCEN1) and Populus Centroradialis 2 (PopCEN2) leads to precocious maturity (Mohamed et al. 2010). In addition, miR156 and miR172 contribute to the control of vegetative phase change (reviewed in Huijser and Schmid 2011).

Perennial plants of temperate and boreal regions need to adapt to seasonality and develop a strategy to survive the winter period with its unfavorable growth conditions. For this purpose, most trees initiate bud set in late summer and then undergo a period of bud dormancy (for a recent comprehensive review, see Cooke et al. 2012). From a physiological point of view, dormancy has been divided into three phases: paradormancy, caused by inhibitors in leaves and terminal buds, ecodormancy, due to unfavorable environmental conditions, and endodormancy, caused by inhibitors within the bud itself (Lang 1987; Lang et al. 1987). Later it has been redefined independently of external and internal stimuli as the inability of a meristem to resume growth under favorable conditions (Rohde and Bhalerao 2007). This can be applied to the apical meristem as well as to the axillary meristems and the cambium. While the cambium is sheltered by the bark, the SAM and the axillary meristems are enclosed in buds, providing protection in the winter period. Buds are highly important for both vegetative and reproductive growth of trees since they are in fact undeveloped shoots. While most trees produce vegetative buds that develop into vegetative shoots and flower buds that develop into flowers, some species such as apple and pear produce vegetative and mixed buds, the latter of which can develop into leafy shoots as well as flowers (Mimida et al. 2009). In these species, a mixed unit (“bourse”) containing vegetative and floral organs can occur. A bourse can subsequently form a sylleptic axillary shoot (“bourse shoot”) that can finally develop into a short or long shoot (Costes and Guédon 2002). The decision whether a bud turns into a flower/fruit or shoot is controlled by various factors such as cultivar, rootstock, shoot growth and phytohormones (Hoad 1984; Buban 1996; Koutinas et al. 2010).

Cessation of apical elongation growth and bud set are the first steps to winter dormancy. Most plants induce these processes in response to shortening of day length (Heide 1974; Junttila 2007), whereas species of the Rosaceae family such as apple react to lower temperatures (Heide and Prestrud 2005; Heide 2008). Species with a strictly determinate growth pattern, in which the terminal bud contains all preformed primordia and internodes for the subsequent growth period, show autonomous control of bud set and growth cessation with almost no influence of environmental changes (Junttila 1976). Dormancy release occurs in answer to the fulfillment of a chilling requirement and/or a longer photoperiod (Murray et al. 1989; Heide 1993; Junttila and Hänninen 2012). The longer and the colder is the chilling period, the lower are the time and temperature sum to bud burst (Junttila and Hänninen 2012). The master switch in the regulation of onset as well as release of seasonal arrest is again the CO/FT module sensing day length in relation to the circadian clock (for a recent review on the circadian clock, see Farrè et al. 2012). Additionally, the circadian clock might be able to sense temperature in Arabidopsis (Edwards et al. 2006; Gould et al. 2006). In Populus, FT2 controls growth cessation, bud set and dormancy induction, and FT1 is expressed during the chilling period to induce the transition from the vegetative to the reproductive phase (Böhlenius et al. 2006, Hsu et al. 2006, 2011; Rinne et al. 2011). Furthermore, overexpression of CEN/TFL1 in Populus causes delayed bud break and altered chilling requirements (Mohamed et al. 2010). Aintegumenta-like genes (regulators of cell division) and dormancy-induced MADS-box genes are downstream targets of CO/FT (Karlberg et al. 2011; Yamane et al. 2011). Additional downstream effects related to dormancy induction are the upregulation of genes associated with cold hardiness and drought, defense, carbohydrate synthesis and transport, cell wall biosynthesis or modification as well as RNA metabolism and chromatin modification/remodeling (Ruttink et al. 2007; Park et al. 2008; Ko et al. 2011). In contrast, the transition from dormancy to active growth is characterized by the induction of flowering pathways, RNA metabolism and protein biosynthesis and transport (Larisch et al. 2012). It has been found that SAM cells are symplastically isolated during the dormant period due to the formation of a callose block at the plasmodesmata (Rinne and van der Schoot 1998; Rinne et al. 2001; Rinne et al. 2011). However, it is not yet clear whether there is a causal relationship between the cellular isolation and the activity–dormancy cycle. Phytohormone levels are also altered by the circadian clock in response to dormancy. A short photoperiod causes downregulation of Gibberellic Acid 20 oxidase and correspondingly decreases GA levels, which induce growth cessation (Eriksson and Moritz 2002). ABA has a central role in seed dormancy and has long been thought to mediate bud dormancy as well (Knox and Wareing 1984). Its precise role in bud dormancy is still unclear, but it seems to be involved in the control of bud development and maturation, as ABA insensitive 3 overexpressing plants develop defective buds, but normal dormancy (Ruttink et al. 2007). Ethylene might crosstalk with ABA to regulate bud dormancy induction and bud morphology (Ruttink et al. 2007). The decreased cell division capacity of the cambium during winter dormancy has been shown to be accompanied by decreased auxin sensitivity (Schrader et al. 2004; Baba et al. 2011). Furthermore, there might be epigenetical (Santamaria et al. 2009, 2011), miRNA (Wang et al. 2011) and metabolism (sugars, energy status, redox state and reactive oxygen species) aspects of the regulation of winter dormancy (Halaly et al. 2008; Ophir et al. 2009), but research on these topics is still underway.

It would be of great value to gain an even deeper insight into the control of plant architecture in trees, with the possibility to transfer the new knowledge to other plant species. The columnar growth phenotype of apple is a natural mutation with the potential to reveal a key player in the regulation of tree architecture since it is dominantly inherited. As columnar apple trees show a thicker main stem with shorter internodes than apple trees with a standard growth habit and since many short fruit spurs emanate from the main stem of columnar apple trees at a narrower crotch angle than the long lateral shoots of standard apple trees (Fig. 2), its causative gene, Columnar (Co), seems to influence nearly all aspects of plant architecture. At the same time, it could equip apple breeders with an important tool for the generation of new cultivars with high economic importance. Therefore, this review will briefly address the history and development of the columnar growth habit and then focus on the different approaches that are currently being taken to identify Co and its function.

Fig. 2
figure 2

Comparison of the plant architecture of standard and columnar type apple trees. a Apple trees with standard growth habit (variety A14) have long lateral branches with a wide crotch angle and usually require staking. b By contrast, columnar apple trees (variety Goldcats) do not require wood stakes and show compact growth. c The pillar-like growth of columnar trees (here variety A73-75-97K) is due to short fruit spurs and few longer lateral branches emanating at a narrow crotch angle and growing almost parallel to the stem. Pictures were taken at the Geisenheim University in early fall 2011 (a, b) and early spring 2013 (c)

Full size image

History and development of the columnar growth habit

In the 1960s, researchers at East Malling in Kent were trying to revolutionize fruit tree breeding by the induction of mutations in apple (Malus × domestica) and the subsequent examination of the resulting phenotypes, focusing on spur type trees. The spur type growth habit was first recorded in the 1920s and is characterized by the formation of numerous short fruit spurs instead of large side branches (Quinlan and Tobutt 1990). However, the spur type is a fruiting type rather than a growth habit (Fideghelli et al. 2003), and an extreme of a continuous variation rather than a distinct heritable trait (Looney and Lane 1984), so in targeted breeding experiments, almost none of the spur type varieties transferred the desired phenotype to their progeny to a significant extent (Lapins 1969).

Being trained to watch out specifically for spur type mutants, in 1961 grower Anthony Wijcik spotted a very compact and spurry limb sport atop a 50-year-old McIntosh tree at the Summerland Research Station, British Columbia, which arose as a result of a spontaneous somatoclonal mutation (Fisher 1969, 1995). Fruits from this sport reached maturity slightly later than that of the rest of the tree and had a less intensive color (Fisher 1995). Vegetative propagation of this sport, later called “McIntosh Wijcik” (commercially also known as “Starkspur Compact Mac”), and subsequent crosses with plants showing a normal growth habit demonstrated that almost 50 % of its progeny showed the compact phenotype (Lapins 1969, 1974; Lapins and Watkins 1973), which was later referred to as columnar growth habit. A number of similar sports were found on top of other aging McIntosh trees, e.g., in the Bendig orchard at Summerland, but were not propagated (Looney and Lane 1984; Fisher 1995).

Using McIntosh Wijcik as a parent, thousands of crossings performed at East Malling yielded six more columnar apple varieties until 1991: Telamon (Waltz), Trajan (Polka), Tuscan (Bolero), Obelisk (Flamenco), Charlotte (Hercules) and Maypole (Tobutt 1994). These columnar apple trees of the first generation, known as the “Ballerina” trees due to their commercial names, were susceptible to scab, showed biennial bearing and their fruits were not competitive with those of the most popular commercial apple tree varieties such as Golden Delicious, Jonagold or Gala. Breeding approaches in Canada, the USA, China, Korea, Belgium, Lithuania, Russia, Great Britain, Germany and France have since produced columnar varieties like Arbat, Moonlight and Goldlane that bear fruits of higher quality than the original columnar varieties and are resistant to scab and other common diseases (Gelvonauskienë et al. 2006). However, breeding apples by conventional crossing is time-consuming and cost-intensive. Apples are highly heterozygous and self-incompatible (Hegedűs 2006; Newcomb et al. 2006), and many agronomically important traits are under polygenic control, which constrains the production of varieties with a specific combination of advantageous traits and increases the need for large progenies of crossings (Velasco et al. 2010). Furthermore, apples have a long juvenile period so that fruit quality can only be assessed after several years (Hackett 1985). Thus, the demand for methods of early detection of the growth phenotype and more efficient creation of new columnar apple cultivars has accelerated research of this interesting phenotype and its molecular cause.

Phenotype characteristics of columnar type apple trees

The compact growth of columnar type apple trees is based on their very thick and upright stems with almost no difference in diameter between the top and the bottom and short internodes, overall looking like a sturdy cordon (Fig. 2b) (Tobutt 1985, 1994). They produce short fruit spurs rather than long lateral branches (Fig. 2c). Rarely, the axillary buds do develop into long lateral shoots which then grow almost parallel to the stem at a very narrow crotch angle (Hemmat et al. 1997; Bai et al. 2012). The development of long side shoots is favored if the central leader is damaged, in which case two to three spurs near the top grow to about 50 cm of length (Tobutt 1985; Watanabe et al. 2006), which implicates that the lateral buds are under tight apical control. The new lateral shoots also show the columnar habit (Kenis and Keulemans 2007). If a number of lateral buds of 1-year-old branch sections are removed, then the reaction is that the more buds are removed the more likely the remaining ones are to grow out (Looney and Lane 1984), indicating a competition between individual spurs. Columnar trees have less sylleptic shoots and thus show higher apical dominance. They exhibit a lower level of acrotony and develop more proleptic shoots, which is in line with the hypothesis of the buds being under higher apical control than those of trees with standard growth habit (De Wit et al. 2000). Even though shoots of columnar apple trees grow longer during one vegetative season than shoots of normal trees (Watanabe et al. 2004), the central leader of McIntosh Wijcik grows to only about 55 % the size of a standard McIntosh tree, and many of the other columnar and spur type trees are also smaller than normal trees (Lane and Looney 1982; Kelsey and Brown 1992). However, since this does not apply to all varieties, it is likely that columnar growth habit and dwarfing are two distinct traits that segregate independently (Eaton and Lapins 1970).

While the number of leaves per shoot is similar between normal and columnar apple trees (Lee and Looney 1977), the total leaf area is greater in columnar seedlings at 3 years of age (Zhang and Dai 2011), and the leaves themselves also show some differences. Leaves of columnar apple trees are dark green and very thick with long petioles and usually have a serrate or crenate margin (Lapins 1969; Tobutt 1988a, b, c, d; Sarwar et al. 1998). This is a characteristic of most other spur type growth habits as well (Liu and Eaton 1970). Microscopic examinations and detailed measurements have demonstrated that the leaves of columnar apple trees have a thicker palisade parenchyma as well as a greater dry weight and chlorophyll content (Gelvonauskis et al. 2006; Zhang and Dai 2011). This results in a higher net photosynthetic rate and transpiration rate of columnar compared with normal type apple trees (Zhang and Dai 2011).

The diameter of xylem vessels is bigger in shoots and roots of columnar than standard apples and the number of xylem vessels is also higher in roots of columnar trees (Zhang and Dai 2011). This together with the higher photosynthetic rate explains why these trees can produce high yields of fruit despite their compact growth, at least when they are grown on typical commercial rootstocks such as M9: they are able to efficiently transport water and minerals from the soil up through the stem and to produce a higher amount of sugar compounds. Regarding the number or width of phloem vessels, no differences were found (Zhang and Dai 2011).

Another characteristic of columnar apple varieties is that they often show frost and drought resistance, which might be due to their Canadian origin (Jacob 2010).

The columnar growth habit can be detected as soon as 2 weeks to 2 months after germination (Lee and Looney 1977; Meulenbroek et al. 1998). However, this early examination is often erroneous. A reliable verification of the phenotype is possible after about 2–3 years (Blazek 1992; Baldi et al. 2012). Even then, classification can be difficult because there is not always a clear distinction between different growth phenotypes and many intermediate types exist (Hemmat et al. 1997; Kim et al. 2003; Ikase and Dumbravs 2004; Moriya et al. 2009; Baldi et al. 2012). In addition to the age of the plants, the proportion of progeny with an intermediate growth habit resulting from a cross with a columnar parent seems to be dependent on the columnar variety used (Table 1) as well as on the growth conditions (Tobutt 1985; Brown et al. 2004). The columnar growth habit is still evident when columnar plants are grafted on rootstocks obtained from apple varieties with standard growth. However, the choice of rootstock influences tree height, stem diameter and shoot number (Gelvonauskienë et al. 2006). This can either be caused by the lack of expression of columnar-specific genes in the roots of normal type rootstocks, or by the general influence of the rootstock on plant growth characteristics like tree height, trunk diameter, number of shoots and flowering onset, which can be observed to act upon standard type apple trees after grafting (Seleznyova et al. 2008). As for normal apple trees, dwarfing rootstocks such as M9 are efficient tools for controlling the height and shoot number of columnar apple trees (Gelvonauskienë et al. 2006).

Table 1 Results of crosses with columnar cultivars
Full size table

Most of the phenotype characteristics of columnar apple tree varieties provide economic benefits, which is why it has always attracted the attention of breeders and subsequently researchers. Columnar trees can be planted only 0.5–1 m apart in orchards of about 10,000 trees per hectare (Tobutt 1994). They require no staking due to their thick and upright stems and only need to be pruned to control their height. Flowering for the first time takes about 4 years after germination; they have a lifespan of about 20 years and could pay back for the expenses of planting after about 4 years. Furthermore, mechanical harvesters could be used in orchards of this kind, which would save additional cost and labor. It has also been proposed that columnar apple trees be used as space-saving pollinators for conventional orchards or as ornamentation in gardens or streets (Tobutt 1985).

Mapping and analyzing the Columnar gene region

The columnar growth habit represents a class of fruit tree architecture on its own (Fideghelli et al. 2003), so it would be highly interesting to determine its molecular basis. Lapins (1969) first proposed that the columnar growth habit could be attributed to the dominant allele of a single gene, Columnar (Co). Since crosses between a columnar and a non-columnar parent usually yield less than 50 % columnar progeny (Table 1), Lapins (1976) suggested that one or two modifier genes might be involved. In contrast, Blazek (1992) deduced that the columnar growth habit might be a double recessive trait. However, the latter hypothesis can be rejected because all commercially available columnar cultivars have been found to be heterozygous for Co (Tian et al. 2005) and crosses between two columnar cultivars have yielded only up to 75 % columnar progeny (Lapins 1976; Meulenbroek et al. 1998). It has been proposed that the deficiency of columnar type trees in the progeny might be caused by a negative influence of Co or a linked gene on the viability of pollen, seeds or emerging seedlings (Meulenbroek et al. 1998). Baldi et al. (2012) found the lack of columnar F1 plants to be more pronounced for grafted trees than for trees grown on their own roots, so they concluded that columnar plants might be lost during and/or shortly after the grafting process, which might be amplified when dwarfing rootstocks (in their case M9) are used.

At present, the identity and function of Co as well as its gene product or the type of mutation are still unknown. Similar growth phenotypes have been found for other Rosaceae species such as peach and sour cherry, but their genetic background is either unclear or is distinct from apple (Scorza et al. 2002; Schuster 2009). Since Co is dominant, it is rather unlikely that it is a loss-of-function allele, unless it has a dose-dependent effect. It might carry an amino acid-changing single nucleotide polymorphism (SNP) that causes a gain-of-function in a protein. It might also be a small RNA or a transposon insertion/deletion. Since columnar trees still show their compact growth even when grafted on normal type rootstocks (Fisher 1995), the Co gene product probably exerts its effect in the shoot rather than working up from the roots. Interestingly, overexpression of the Arabidopsis leafy gene in apple trees causes the plants to develop a columnar growth habit (Flachowsky et al. 2010), so Co might somehow be involved in the developmental switch from indeterminate vegetative to determinate reproductive growth. However, these are mere speculations and different scientific approaches are needed to form well-founded hypotheses.

During the past 15 years, several research groups have tried to determine the chromosomal location of Co using marker analyses so that we can now conclude that it is located on chromosome 10 within the region of 18.51–19.09 Mb (Fig. 3a). At the same time, the genetic linkage map of apple was designed and constantly refined, providing more markers for coupling analyses (Maliepaard et al. 1998; Liebhard et al. 2002, 2003; Silfverberg-Dilworth et al. 2006; Wang et al. 2012). The different research groups used PCR-based markers, mainly random amplification of polymorphic DNA (RAPD), amplified fragment length polymorphism (AFLP) or simple sequence repeats (SSRs), the former two sometimes being converted to sequence characterized amplified region (SCAR) markers.

Fig. 3
figure 3

Molecular markers in the Co gene region. a Co has recently been mapped on chromosome 10 at 18.52–19.09 Mb based on the apple genome sequence (Velasco et al. 2010). b Before this public

ation, Co had already roughly been mapped to 17.0–19.5 Mb. Markers used are shown at the location they have been mapped to via BLAST searches against the apple genome (Velasco et al. 2010). Marker names are in bold, recombination frequencies are in italics, corresponding publications are designated by colors. */** indicates recombination frequencies found for different varieties. Horizontal lines above the markers in (a) designate the newly delimited Co regions by Bai et al. (2012); Moriya et al. (2012) and Baldi et al. (2012)

Full size image

Before the publication of the apple genome sequence (Velasco et al. 2010), Co was mapped to linkage group (LG) 10 of the apple linkage map and its location was gradually narrowed down to 17.0–19.5 Mb based on seven pivotal markers (Fig. 3b). The very first Co-linked marker, SSRCo (Hemmat et al. 1997), whose locus is also amplified by the COL primers designed by Gianfranceschi et al. 1998 and used for linkage analysis by Kenis and Keulemans (2007), seems to be at a distance of at least 20 cM from Co. Markers in closer proximity to Co are UBC8181000 (Zhu et al. 2007) and SCAR216 (Tian et al. 2005), followed by Hi01a03 (Moriya et al. 2009), SCAR682 (Tian et al. 2005) and CH03d11 (Fernández-Fernández et al. 2008; Liebhard et al. 2002; Tian et al. 2005). Different values for recombination frequencies and genetic distances of these markers to Co were found by different research groups depending on the apple varieties and the number of individuals used in their linkage studies, but their sequential arrangement remained roughly the same. In a comprehensive marker analysis involving segregating progenies of three crosses and a high number of columnar plants of different varieties, SSR markers CH03d11 and Hi01a03 were found to be the most tightly linked markers flanking Co on either side with maximum genetic distances of 7.4 and 1.4 cM, respectively (Moriya et al. 2009). This and two other studies also detected linkage of marker CH02a10, identical to SSRCO F/R (Tian et al. 2005; Bendokas et al. 2007; Moriya et al. 2009), and a fourth study identified SCAR marker SCB82670 as linked (Kim et al. 2003). However, basic local alignment search tool (BLAST) searches (Altschul et al. 1990) against the annotated apple genome (Velasco et al. 2010) show that the sequences of CH02a10 and SCB82670 map to chromosome 3 at about 30 Mb. SCB82670 was later shown to amplify a fragment from the paternal non-columnar parent of the columnar variety used by Kim et al. (2003), so that it cannot be linked to Co (Tian et al. 2005; Fernández-Fernández et al. 2008; Moriya et al. 2009). In contrast, the reason for co-segregation of CH02a10 despite the lack of physical coupling remains unclear. It might indicate a genomic region of major importance for maintaining the viability of columnar plants and is thus covered by a selective sweep as proposed by Krost et al. (2013). Alternatively, the genomic contig might have been incorrectly assembled to chromosome 3.

Since the Golden Delicious genome sequence was released in 2010 (Velasco et al. 2010), the possibility of generating sequence-based markers has significantly simplified and accelerated the fine-mapping of the Co region (Fig. 3a). Bai et al. (2012) developed 88 SSR markers based on genomic apple contigs originating from chromosome 10 around the target region as well as from two unanchored contigs which they identified as being located within this region via a synteny approach using the peach genome. They used four segregating progenies as well as 290 columnar selections to evaluate the quality of 18 already published markers and the new SSR markers. 47 plants showing recombination between SCAR682 and Hi01a03 as well as one double-recombinant were used for screening with the new markers, and 6 key recombinants finally served to delimit the Co gene region to 193 kb between markers C1753-3520 at 18.90 Mb and C7629-22009 at 19.09 Mb. Marker C18470-25831 (19.0 Mb) was found to co-segregate with Co. The newly delimited region contains 20 annotated genes and 7 predicted genes, 3 of which code for homologs of Lateral Organ Boundaries Domain transcription factors in Arabidopsis (Majer and Hochholdinger 2011) and were thus considered the most likely candidates for Co (Bai et al. 2012).

Moriya et al. (2012) also developed SSR markers based on the Golden Delicious genome sequence and used 1,000 F1 individuals from 31 populations for linkage analysis. They delimited the Co gene region to 196 kb, a similar size as in Bai et al. (2012). However, the markers developed by Moriya et al. (2012) flanking this region are located at 18.76 Mb (Mdo.chr10.11) and 18.96 Mb (Mdo.chr10.15). Additionally, markers Mdo.chr10.12 (18.79 Mb), Mdo.chr10.13 (18.83 Mb) and Mdo.chr10.14 (18.89 Mb) co-segregated with Co.

In a third study, Baldi et al. (2012) first used three adult segregating progenies (301 F1 plants in total) and the early published markers to roughly define the Co gene region and then refined it with two large populations treated as a single segregating progeny of 1,250 individuals which they subjected to genotypic mapping with seven newly designed SSR markers. Co was localized between co-mapping markers Co04R10 and Co04R11 on one side and Co04R13 on the other side, which yielded a genomic region of 0.56 cM, corresponding to 393 kb at 18.52–18.90 Mb. A thorough open reading frame analysis revealed 36 potential genes within this region, several of which code for transcription factors of the MYB, basic helix-loop-helix and AP2/ERF classes, members of which play roles in the regulation of plant architecture. The target region overlaps with the region identified by Moriya et al. (2012), but does not span the chromosomal location predicted for Co by Bai et al. (2012). Furthermore, marker C18470-25831, which Bai et al. (2012) found to be linked to Co, showed three recombinants in the mapping population of Baldi et al. (2012). Possible reasons for the different locations of the Co region boundaries are that the three research groups used different apple genotypes which possibly have distinct recombination frequencies and that the marker defining the right border of Bai et al. (2012) originates from one of the contigs which was unanchored in the genome project, so its precise location is not well-founded. There might also have been some difficulties in the phenotypic classification of individual plants. Taken together, Co is most likely located between 18.76 and 18.9 Mb, a region comprising approximately 30 annotated genes, none of which have so far been identified as having a profound influence on plant architecture.

Detailed analyses of the Co target region are already underway: Baldi et al. (2012) constructed a bacterial artificial chromosome (BAC) library based on genomic DNA extracted from leaves of McIntosh Wijcik with an average insert size of 145 kb and found ten BAC clones originating from the Co target region, of which a minimum of four clones is needed to span the entire chromosomal section of interest. These clones will enable a comparative analysis of the columnar and the non-columnar allele within the next months.

(Otto et al. 2013) also constructed BAC libraries using genomic DNA of the heterozygous columnar cultivar Procats 28. They obtained more than 100,000 clones with an average insert size of 27 kb, 37 of which could be assigned to chromosome 10 at 18.0–19.0 Mb. Assembly of their sequences yielded two metacontigs of 590 and 190 kb in size. Comparison of these sequences to the Golden Delicious reference has already been used to create four Indel-based markers linked to Co and is ongoing to detect more differences between the genomic organization of columnar and non-columnar apple trees.

Alleles showing consistent differences between columnar and non-columnar cultivars can be expected to be further analyzed and prepared for transformation in the near future.

Analysis of quantitative trait loci in the Columnar gene region

To gain a better understanding of the many distinct but interconnected factors that influence plant architecture, mapping of QTLs associated with plant growth on progenies of columnar trees have been carried out alongside marker analyses. Several research groups have found clusters of QTLs associated with plant architecture on LG 10 in the vicinity of Co. For 172 juvenile apple trees of the cross McIntosh Wijcik × NY 75441-58, internode number in year 1 and year 2 of their life, the stem base diameter increment in year 1 and year 3 as well as branch number and internode length in year 1 were associated with regions located on LG 10 at roughly the locus of the marker P459z (Conner et al. 1998). The terminal bearing (Tb) locus correlating with branching habit and influencing vegetative bud break (Lawson et al. 1995) also maps close to this region (Conner et al. 1998); however, it is distinct from Co. Based on the investigations of a Telamon × Braeburn progeny comprising 257 individuals, a QTL cluster for a wide range of phenotype characteristics such as total growth increment, total branch number and branch length, internode length, main axis growth rate and main axis height increment was detected on LG 10 (Kenis and Keulemans 2007). The authors deduced clustering of different genes or a pleiotropic effect of a single gene, preferring the latter. They also found the Co gene influence on branch length (apical control) to be more pronounced than its influence on branch number (apical dominance).

Using the same mapping population, loci for plant architectural traits and QTLs associated with fruit quality were both found on the same linkage group as Co (Davey et al. 2006; Kenis et al. 2008). This includes QTLs for fruit flesh weight, flesh l-ascorbic acid content, soluble sugar, firmness and acidity. Some of these QTLs account for up to 60 % of the phenotypic differences observed. A possible theory is that this specific area on LG 10 controls aspects of plant growth and development and these then have pleiotropic effects which affect fruit quality traits (Kenis et al. 2008). This led Moriya et al. (2009) to the conclusion that Co might be tightly linked to genes conferring low fruit quality, and thus it would be desirable to use gene technology for the production of new columnar cultivars because classical breeding approaches would in most cases transfer Co in combination with these unpopular traits.

In summary, the Co gene region on LG 10 seems to be pivotal for plant growth and development, either due to the influence of one gene (possibly Co) or due to the combined action of several genes in a cluster. In any case, changes in growth habit would most likely be accompanied by alterations in the phytohormone levels of the plant.

Phytohormone levels in columnar apple trees

Several attempts have been made to correlate the columnar growth habit with changes in phytohormone levels. Unfortunately, phytohormone concentrations are difficult to measure and they show extensive variations depending on the age of the plant, the season and the environmental factors. Additionally, it is challenging to decide whether changes in hormone levels are a cause or a consequence of the columnar growth habit. Only four hormone groups have been investigated so far in the columnar apple, and direct connections between the results and the phenotypic manifestation are scarce.

Since columnar apple trees form short fruit spurs rather than long lateral branches, it was suggested that they might have stronger apical dominance and apical control than normal type apple trees and thus a focus has been put on the IAA levels in columnar plants. Measurements of IAA content with the indole-pyrone fluorescence method indicated that in shoot tips, free IAA levels correlated positively with the degree of compactness. Additionally, axillary buds of columnar apple trees have a higher ratio of free IAA to total IAA than normal type apple trees due to a lower level of conjugated IAA (Looney and Lane 1984). However, when the polar auxin transport is blocked by application of the inhibitor triiodobenzoic acid, branch attachment angles and the numbers of spurs on both genotypes were increased, which contradicts the hypothesis of stronger apical dominance in columnar apple trees (Looney and Lane 1984). The authors concluded that in McIntosh Wijcik, an IAA-hydrolyzing enzyme might be active at the time when lateral buds are formed, which results in very strong buds easily breaking dormancy and growing into spurs during the following season. Due to the competition for nutrients among the axillary organs, fruit spurs rather than long branches are formed (Looney and Lane 1984). However, it seems more likely that the preferential formation of spurs compared with lateral shoots is caused by a lack of growth-promoting factors such as GA. Watanabe et al. (2004, 2006, 2008) measured the IAA concentration of the central and axillary shoots (arising from just below the previous year’s pruning cut) of 3- to 5-year-old columnar trees grafted on seedling rootstocks using gas chromatography–mass spectrometry–selected ion monitoring (GC–MS–SIM) and found the central leader to have more IAA than the lateral branches. Furthermore, in July, total IAA was higher in axillary shoots of columnar type apple trees than in axillary shoots of 1- or 2-year-old branches of 38-year-old normal type McIntosh trees on seedling rootstocks, which might be correlated with the observation that the former still grow vigorously until October, whereas the latter already cease growing in July. No significant differences were found with respect to the total IAA content in axillary shoots of columnar and normal type apple trees during the entire growth season. Unfortunately, since Watanabe et al. (2004, 2006, 2008) compared 3- to 5-year-old columnar trees of the varieties Maypole and Tuscan (grown on seedling rootstocks) to normal McIntosh trees at 38–40 years of age (also grown on seedling rootstocks), the differences might be a result of age rather than of growth habit. Furthermore, the number of plants analyzed was low (maximum of n = 15). Therefore, it would be interesting to compare IAA levels within a high number of columnar and non-columnar trees of the same age. Since it is usually the IAA transport within the shoot and not the absolute IAA level that regulates growth processes, measuring the IAA movement within the stem of columnar apple trees would also be of great interest. In summary, the results indicate a higher IAA content in columnar apple trees, which would be in agreement with higher apical control.

Focusing on CKs, in the soybean hypocotyl section bioassay, McIntosh Wijcik was found to have significantly higher levels of zeatin-like growth substances than other McIntosh strains (Looney and Lane 1984). In GC–MS-SIM experiments of the same plants as mentioned above, the endogenous concentration of zeatin riboside, the predominant CK associated with bud burst in apple, was found to be higher in both apical and lateral shoots of columnar trees during the course of a year (Watanabe et al. 2004, 2006). This could explain the high number of spurs. However, like in the IAA measurements, the authors again only compared a low number of 3- to 5-year-old columnar trees, which grow vigorously, with 38- to 40-year-old normal trees, which are at the end of their growth period, so further studies comparing a statistically significant number of trees of the same age would be needed. The ratio of isopentenyl adenosine to total CK was highest in lateral shoots and buds of normal trees in July and of columnar ones in November, so it is probably associated with the onset of winter dormancy (Watanabe et al. 2008).

Growth of normal apple trees after exogenous application of CKs follows an optimum curve: the higher the amount of CKs given, the more vigorously does the plant grow when growth is defined as the number of shoots initiated which develop to shoots longer than 1 cm or as culture weight gain for in vitro cultures. If the concentration exceeds the optimum level, plant growth is more and more inhibited. In vitro cultures of columnar apple trees show a similar reaction of culture weight gain to exogenously applied CKs, and the optima for 6-benzyladenine (BA) (5 μM) and thidiazuron (3 μM) are similar to the ones of normal type apple trees (Lane and Looney 1982). However, at very low BA concentrations, normal type apple cultures on modified Murashige and Skoog medium grow better, whereas columnar apple cultures on modified Murashige and Skoog medium show a significantly higher tolerance to supra-optimal concentrations of CKs: the maximum concentrations at which columnar and normal type plants still grow are 25 (Sarwar et al. (1998): 50 μM) versus 5 μM for BA, 60 versus 25 μM for kinetin, 25 versus 20 μM for 2-isopentyladenine, and 40 versus 25 μM for thidiazuron, respectively (Lane and Looney 1982; Sarwar et al. 1998). To perform shoot regeneration from leaf explants, higher concentrations of CKs are needed for columnar apple plants than for normal ones (Sarwar and Skirvin 1997). The authors concluded that either columnar plants can metabolize excess levels of CKs or they compensate for it by adjusting the level of other growth regulators (Lane and Looney 1982). Another explanation would be that columnar apples do not take up excessive amounts of CKs, because they have thicker cell layers than standard type apple trees (Sarwar et al. 1998).

Taken together, the levels of CKs seem to be higher in columnar than in normal apples and columnar apple trees seem to have an altered CK metabolism. Watanabe et al. (2008) hypothesized that the large leaf area of columnar apple trees might contribute in part to the increased production of CKs. However, as CKs are predominantly produced in the roots and to a smaller extent in young leaves only, this explanation seems rather unlikely.

Some of the phenotype characteristics of columnar trees such as stunted growth and dark green leaves resemble characteristics of GA-deficient mutants (Koornneef and Veen 1980; Talon et al. 1990; Sun and Kamiya 1994; Peng and Harberd 1997). Thus, GAs constitute another class of phytohormones that has attracted the attention of researchers focusing on columnar growth. Looney and Lane (1984) summarized their findings on GA levels in columnar apple trees as follows: in the dwarf pea bioassay, shoot tip extracts of columnar apple seedlings did not promote growth as much as those of normal type seedling, indicating lower GA-like activity. Low polar GAs were also found in actively growing shoot tips of McIntosh Wijcik when examined using silica gel partition column chromatography and the dwarf rice bioassay. However, polar GAs are not necessarily bioactive (Atzorn and Weiler 1983). After GA3 was applied exogenously, columnar seedlings showed a greater percentage of growth increase, but still did not reach the height of their normal type counterparts. The conclusion was that low GA levels probably correlate with dwarfing of McIntosh Wijcik rather than its spurriness, and dwarfing is a phenotype characteristic independent of compact growth (Looney and Lane 1984).

With regard to ABA, lower levels of free cis-ABA were found in actively growing shoot tips including five expanded terminal leaves of young columnar progeny of McIntosh Wijcik crosses with three different non-columnar varieties grafted on M7 than in their normal type siblings (Lee and Looney 1977). These data were statistically significant on a per shoot tip basis, whereas data on a fresh weight basis had a tendency toward lower levels, but no statistical significance was achieved. The bourse buds of McIntosh Wijcik trees also had less free and conjugated ABA than those of standard McIntosh on a fresh weight basis, even though, due to their larger size, the total ABA amount per bud was higher (Looney and Lane 1984). ABA in general is higher per fresh weight in rapidly elongating shoots (Feucht et al. 1974), so the lower ABA levels are probably a consequence—not the cause—of the slower growth of fruit spurs compared with lateral shoots (Looney and Lane 1984).

In the seeds and early seedlings of a progeny of controlled crosses of McIntosh Wijcik with a non-columnar variety, levels of ABA and GA were similar for both varieties, indicating that the hormonal differences that characterize the compact seedlings are probably established at a later stage of development (Lee and Looney 1978).

Taken together, these results suggest that columnar apple varieties do show differences in phytohormone levels, but most of them are fairly subtle, and only hypotheses of their correlation with the phenotype can be made. Due to the high IAA levels and lower IAA/CK ratio of shoots, the apical dominance of columnar trees is higher than that of normal type trees and thus they do not produce long axillary shoots. Fruit spurs are produced because of high levels of CKs in combination with low levels of ABA, which favors bud break, but slows extension growth. Only in some occasions (e.g., when the central leader is damaged) are a few spurs able to overcome the apical dominance and grow out. The lower levels of GA might inhibit long extension growth and are related to the dwarfing of most columnar trees.

Transcriptome analyses of columnar type apple trees

Another approach to unravel the function of Co is the analysis of transcriptional changes in columnar compared to normal type apple trees. Taking into consideration of the fundamental phenotypic changes, it can be expected that the expression levels of several genes are altered, but it is difficult to decide on one specific pathway which is definitely influenced, so the method of choice is a whole transcriptome study.

Three recent studies have analyzed the transcriptomes of columnar and normal type apple trees via RNA-seq. Zhang et al. (2012) collected new lateral shoots of 4-year-old columnar and standard seedlings of the progeny of Fuji × Telamon. They used three time points between May and July 2010 and pooled the material of the three time points to obtain two samples, one from columnar and one from non-columnar apple. They performed a total RNA extraction and mRNA purification, and generated about 4 million reads for each sample in an Illumina HiSeq 2000 next generation sequencing run. 80 % of the reads were mapped to the apple genome (Velasco et al. 2010). Assembling yielded about 57,000 non-redundant unigenes with contigs having an N50 of about 420. On comparing differential gene expression based on reads per kilobase per million mapped reads values, 5,237 genes were found to be differentially expressed by more than twofold, 2,704 being upregulated and 2,533 being downregulated in the columnar versus the normal type apple. Of the differentially expressed genes, 15 % were involved in the biosynthesis of secondary metabolites, and 24 % had a role in metabolic pathways, some of them being key players in GA, IAA and BR biosynthesis. Unfortunately, no specific genes or their precise regulation were mentioned. In addition, 287 genes involved in pathways crucial for the regulation of plant architecture were identified, most likely based on the function of their homologs identified by BLAST searches (although the authors did not mention how they identified their function), but again no values for their expression were given. Among these 287 genes, 31 were mapped to chromosome 10, and 25 were Gibberellic-Acid Insensitive, Repressor of Gibberellic-Acid Insensitive and Scarecrow (GRAS) transcription factors like DELLAs, which play a role in the gibberellin signal transduction. Some of these genes of interest are intended to be transferred into Gala apple trees via Agrobacterium-mediated transformation.

Krost et al. (2012) also compared gene expression levels of columnar versus normal type apple trees using RNA-seq. They collected shoot apical meristems of spurs of columnar Procats 28 (P28), which has a Telamon ancestor, and of branches of normal A14-190-93 (A14), isolated the total RNA and purified the mRNA followed by mRNA amplification. May 2009 and September 2009 were used as time points for A14 and P28, respectively. 454 as well as Illumina sequencings were performed on these samples, and about 250,000 reads were obtained for each 454 library as well as about 80 million reads for each Illumina library. They conducted BLAST searches of the raw sequences against all annotated Malus x domestica proteins (MDPs) and compared those to UniProtKB. Subsequently, differential expression was determined and differentially expressed genes were grouped into distinct categories. Genes of categories representing light reactions, mitochondrial electron transport, lipid metabolism and cell wall modification (expansins and xyloglucan endotransglucosylases/hydrolases) were significantly downregulated in the columnar variety, whereas another group of genes involved in cell wall modification, terpenoid and tryptophan synthesis (the precursors of IAA biosynthesis) were upregulated. Genes involved in DNA synthesis, RNA processing and protein synthesis were downregulated, which correlates with reduced growth of the columnar plants, whereas those involved in transport and protein modification were upregulated. Considering phytohormone metabolism, genes of biosynthesis and signal transduction of IAA and jasmonates were induced. The opposite was reported for genes associated with GA biosynthesis and signal transduction. These results agree with the findings on phytohormone levels described in the previous section. Furthermore, there were hints to a cell cycle arrest in G2 in columnar apples, which would result in a lower number of cells and would thus explain the stunted growth. In summary, results suggested an alteration in cell wall and cell membrane formation resulting in smaller cells which in combination with the cell cycle arrest would lead to short spurs instead of long branches. Functions associated with membrane integrity such as transport, photosynthesis and mitochondrial electron transport also seemed to be changed, which were considered to be consequences of columnar growth.

In a second gene expression study, Krost et al. (2013) narrowed down the genes of interest to those involved in phytohormone biosynthesis, signaling and transport. At the same time, they expanded the amount of data to six Illumina data sets, totaling almost 500 million reads, from three different time points (May 2009, September 2009 and July 2010) and validated their data with results of a microarray chip hybridized with RNAs of four collection dates between April and May as well as of quantitative real-time PCR carried out on RNA isolated from in vitro cultures of P28 and A14. It is doubtful whether in vitro cultures represent a suitable model for the study of transcriptional changes associated with altered tree architecture since they do not reach the same age and architectural complexity as trees, and in vitro-grown leaves have been shown to have altered methylation patterns compared with field-grown leaves, which might influence gene expression (Li et al. 2002). However, since the in vitro cultures were only used to confirm results that had already been obtained by RNA-seq and microarray analyses of in vivo material in this study, they provide the possibility to test the transferability of the results to plants grown under tightly controlled conditions. Of 619 genes found to be significantly differentially regulated by Krost et al. (2013), 16 were detected to be involved in the regulation of all major phytohormone groups, as can be expected for a phenotype affecting many different aspects of plant growth. An integration of the results suggested that an increase of IAA levels together with a higher basipetal IAA transport (due to upregulation of Auxin Resistant and Pin-formed 1) is responsible for the high level of apical dominance and apical control of columnar apple trees. This is overcome by the elevated level of CKs when spurs are formed; however, they cease their elongation growth early. Additionally, the complex interplay between growth-promoting factors (IAA, CKs and apolar GAs) and growth-inhibiting factors (JAs and XTHs), which all showed upregulation in columnar trees in this study, is responsible for the achievement of normal plant height. Another 16 genes showing constant differential regulation in at least two of the three gene expression studies conducted were analyzed with regard to their chromosomal location. Five of them were found to reside in chromosome 10, indicating a statistically approved enrichment of differentially regulated genes in the Co gene vicinity. Four of them are associated with phytohormonal regulations, so Krost et al. (2013) drew the conclusion that the region is most likely covered by a selective sweep due to the influence of the Co gene.

Further time course gene expression experiments will be necessary to validate and increase our knowledge about the function of the Co gene product. It would also be helpful to expand these analyses to other tissues and developmental stages, yielding a comprehensive picture of the physiological state of the plant.

Conclusion

While a number of individual genes and signaling molecules regulating plant architecture have already been identified in model plants, it has not yet been possible to create a concise picture of the whole process since some key players still remain elusive and knowledge about the interaction of the individual pathways is scarce. The columnar growth habit of apple is a natural mutant showing altered plant shape and thus might be the key to an important factor in the regulation of tree architecture. During the past few years, a lot of progress in the understanding of the columnar growth habit has been made. It is caused by the dominant allele of the Columnar gene on chromosome 10 which probably influences IAA, CK and GA metabolism and signal transduction leading to stunted growth and nearly absent branching. Molecular markers created with the help of the apple genome sequence (Velasco et al. 2010) have successfully been used to fine map the Co locus to 18.51–19.09 Mb on chromosome 10. Since in-depth analysis of this region is already underway, disclosure of the identity of Co can be anticipated for the near future. Additional comparisons of growth regulator concentrations in columnar and normal type apple trees as well as in-depth transcriptome analyses would facilitate the discovery of Co function and affected signal pathways.