A genomics approach to understanding the role of auxin in apple (Malus x domestica) fruit size control

Background Auxin is an important phytohormone for fleshy fruit development, having been shown to be involved in the initial signal for fertilisation, fruit size through the control of cell division and cell expansion, and ripening related events. There is considerable knowledge of auxin-related genes, mostly from work in model species. With the apple genome now available, it is possible to carry out genomics studies on auxin-related genes to identify genes that may play roles in specific stages of apple fruit development. Results High amounts of auxin in the seed compared with the fruit cortex were observed in 'Royal Gala' apples, with amounts increasing through fruit development. Injection of exogenous auxin into developing apples at the start of cell expansion caused an increase in cell size. An expression analysis screen of auxin-related genes involved in auxin reception, homeostasis, and transcriptional regulation showed complex patterns of expression in each class of gene. Two mapping populations were phenotyped for fruit size over multiple seasons, and multiple quantitative trait loci (QTLs) were observed. One QTL mapped to a region containing an Auxin Response Factor (ARF106). This gene is expressed during cell division and cell expansion stages, consistent with a potential role in the control of fruit size. Conclusions The application of exogenous auxin to apples increased cell expansion, suggesting that endogenous auxin concentrations are at least one of the limiting factors controlling fruit size. The expression analysis of ARF106 linked to a strong QTL for fruit weight suggests that the auxin signal regulating fruit size could partially be modulated through the function of this gene. One class of gene (GH3) removes free auxin by conjugation to amino acids. The lower expression of these GH3 genes during rapid fruit expansion is consistent with the apple maximising auxin concentrations at this point.


Background
The hormonal control of fruit growth and development has been well established across many different plants. One hormone, auxin, has been shown to control the initial growth and expansion of tissues following fertilisation [1,2] and inhibit ripening. Early work with strawberry and other fruits proposed a mechanism whereby auxin produced by the developing seed regulated fruit growth by controlling firstly cell division and secondly cell expansion. As the seeds subsequently mature, auxin concentrations drop, acting as a signal for ripening to proceed. Supporting this mechanism is the observation that applied auxins can induce parthenocarpy in fruits such as tomato [3], fruit size in peach [4], cell enlargement in cherry [5] and delay ripening in strawberry [1]. Developmental regulation by the principal auxin in higher plants, IAA (Indole Acetic Acid), is achieved through the coordination of complex processes: auxin metabolism (involving biosynthesis, conjugation and catabolism), auxin transport (long distance and polarised auxin transport) and auxin signalling (perception, transduction and response). The balance of synthesis, breakdown, conjugation and transport is tightly regulated, leading to auxin homeostasis [6].
De novo auxin synthesis in plants results from multiple pathways dependent or independent of tryptophan [7,8]. IAA can be conjugated to amino acids, sugars and methyl esters. Enzymes that conjugate IAA to amino acids are encoded by members of the group II of the GH3 (Gretchen Hagen 3) family of auxin-induced genes [9]. Very little is known about the role of GH3 genes during fruit development. However, it has recently been shown in grape that GH3.1 plays a role in the formation of IAA-Asp late during development, coinciding with the onset of ripening [10]. Release of IAA from IAA conjugates is achieved by hydrolytic cleavage [11]. Auxin transport from sites of synthesis to target cells is complex and highly regulated, playing a crucial role in both establishing and changing homeostasis. Auxin is transported both passively through the vasculature and actively through transporters [12]. The most characterised auxin transport family is the efflux carrier PIN family. PIN proteins are vital for normal plant development. Mutations in the PIN1 gene lead to pin-like organs with no development of flower parts in Arabidopsis thaliana (Arabidopsis) [13] and members of the PIN family are highly expressed early during tomato fruit development, suggesting a role during fruit set [14].
The current model for auxin perception and signalling involves two types of receptors [15,16]: the Auxin Binding Protein 1 (ABP1), located at the plasma membrane, and the Transport Inhibitor 1/Auxin signalling F-Box family (TIR1/AFB), a set of nuclear receptors [17][18][19]. ABP1 is involved in very early auxin responses leading, for example, to the modification of ion fluxes [20]. ABP1 has been shown to be essential for plant life (a mutation in ABP1 in Arabidopsis is lethal) and is important for both cell division and cell expansion [21][22][23]. However, the details of the pathway going through ABP1 are poorly understood. In tomato, the diageotropica (dgt) mutant displays many auxin-related developmental defects and fruit development is dramatically altered, with a reduced fruit size [24]. DGT encodes a cyclophilin, known to act as signalling intermediate, and was shown to use ABP1 as an extracellular receptor for auxin-dependent cell growth response [25]. The signalling pathway involving TIR1 is now well characterised and explains most of auxin-regulated gene expression [16]. The three families of early auxin responsive genes, Aux/IAA, GH3 and SAUR (Small Auxin Up Regulated), contain a binding motif to the ARF transcription factor (Auxin Response Factor). At low auxin concentrations, a heterodimer of an ARF and an Aux/IAA protein represses transcription. At higher auxin concentration, auxin will bind to TIR1/AFB, an F-box protein that is part of an SCF complex (Skp1/Cullin/F-box), and triggers the degradation of the repressor Aux/IAA through the 26S proteasome. This will ultimately release the ARF transcription factor to modulate expression of early auxin response genes. In fleshy fruits, most of our knowledge involving the ARF-Aux/IAA complex during fruit development comes from studies in tomato. SlARF7 is expressed in placental and ovule tissues and down-regulated soon after pollination. Silencing of the SlARF7 gene leads to parthenocarpic fruit development, showing that SlARF7 functions as a negative regulator of fruit set [26]. Similarly, silencing of the SlIAA9 gene expression also confers parthenocarpy [27]. SlARF4 (also known as DR12) seems to play a role later in fruit development, as its expression increases throughout tomato fruit development, with the highest levels in early red-stage fruit. Down-regulation of SlARF4 leads to pleiotropic phenotypes including darkgreen immature fruit, enhanced firmness and unusual cell division in the fruit pericarp, which is thicker than in wildtype (WT) fruits [28,29].
While many fleshy fruits are carpel derived, the fruit from Malus x domestica (apple) is unusual, as it is derived from the hypanthium, a tube of fused sepals, petals and anther derived tissue. However, like other fruits, apple development can be separated into periods of cell division, cell expansion, maturity, and ripening [30]. While there have been a few studies on auxin content in apple [31,32], there is little research reported on the role of auxin in apple fruit development at the molecular level. There are a large number of different cultivars of apples showing a range of different flowering times, maturity times and times to ripen. One cultivar, 'Royal Gala', is a naturally occurring sport of the 'Gala' cultivar. It is a mid-season apple, and its growth and development has been well characterised. 'Royal Gala' has been the subject of a number of genomics studies, including a large-scale expressed sequencing tag (EST) sequencing project [33] and a microarray study of the fruit development [30] and fruit ripening [34]. It is readily transformable, with transgenic apples for ACO1 suppression [34,35], MYB10 overexpression [36], and POLYGALACTURONASE 1 [37] being reported. Recently a parent of 'Royal Gala', 'Golden Delicious', has had its genome sequenced [38].
Here we have investigated the role of auxin on apple fruit development and assessed the expression of genes involved in homeostasis, transport and signalling of auxin. The location of auxin-related genes in the genome sequence of apple was compared with QTLs for fruit weight, which is linked to fruit size.

Changes in auxin content over fruit development
Apple fruit have previously been given four distinct developmental stages following pollination, consisting of Stage 1 (cell division), Stage 2 (cell expansion), Stage 3 (fruit maturity) and Stage 4 (ripening) [30]. In 'Royal Gala' apples, Stage 1 takes 0-25 days after full bloom (DAFB), rapid growth (Stage 2) covers 20-60 DAFB, after which the fruit continue to grow at a slower rate as the fruit matures (Stage 3), with the ripening process starting around 132 DAFB (Stage 4), with tree-ripe eating apples at 146 DAFB [30].
To investigate the role of auxin during fruit development, the free IAA amounts were measured at representative times (14,45,90 and 132 DAFB) during the different fruit development stages in 'Royal Gala' fruit cortex and seed. IAA concentrations in the seed showed a large increase during fruit development, reaching a maximum concentration of 19 ng/g fresh weight (FW) ( Figure 1A). The IAA concentrations in the fruit cortex were considerably lower than in the seed, ranging from 0.6 to 1 ng/g FW, with a significant increase during cell expansion ( Figure 1B). This is consistent with the literature, showing high auxin concentrations in the seeds of tomato and strawberry, compared with those in the fruits [2,39].

Effect of auxin treatment on apple fruit growth
As there was an increase in auxin in the cortex tissue during Stage 2 (the rapid cell expansion phase), we assessed the effect of injecting three auxin concentrations within the physiologically active range for auxins (10 -6 M) into fruit at the beginning of Stage 2 (30 DAFB). The growth of each apple was assessed by recording the diameter of the fruit at injection and after two weeks of subsequent growth. During the two-week period, all fruit showed an increase in fruit size (Figure 2A). The two lowest auxin concentrations (10 -7 M and 10 -5 M) caused a significant increase in fruit diameter compared with the control (Figure 2A), while the highest concentration (5.10 -5 M) appeared to inhibit fruit growth. The increased fruit growth observed with the 10 -7 M and 10 -5 M treatments corresponded to a greater increase in the cell size ( Figure  2B,C) compared with that of control apples. During the early stages of apple development, there is a natural selfthinning event. Apples typically have 5 florets per cluster, which for commercial purposes are thinned to 1-2 fruit per cluster, depending on the localised fruit load. Two to three apples were chosen per cluster for injection and the rest were hand thinned. During the two-week treatment, the control showed a 32% fruit drop. When injected with auxin at 10 -7 M and 10 -5 M, a higher degree of fruit retention was observed, with only 14% fruit dropped in the 10 -5 M treatment. Additional auxin promoted fruit drop, with 48% fruit abscised from the 5.10 -5 M concentration ( Figure 2D).

Genomic screening of auxin-related genes in the apple genome
The apple draft genome [38] was screened for six classes of auxin-related genes. These included the receptor-like genes ABP1 and TIR1/AFB, the transcriptional control genes ARF and Aux/IAA, and the auxin homeostasis genes, the PIN genes and GH3-like genes. All six classes of genes searched were well represented in the apple genome ( Table 1, full list in Additional file 1). The apple genome has been shown to have undergone a genome duplication, so the numbers of genes were compared with the numbers of auxin-related genes in the recently published woodland strawberry genome [40] (a related diploid Rosaceae species) (Summarised in Table 1, and a full list in Additional file 2). In each gene family, there are approximately twice the numbers of auxin-related genes in apple than in strawberry.
A phylogenetic analysis of the six classes of genes using the predicted protein sequence from apple, strawberry, tomato and Arabidopsis was performed for each class of genes (Figures 3, 4 Additional files 2, 3). In each of the phylogenetic clusters, the majority of the apple genes were contained in subclades consisting of a single strawberry gene. These subclades typically had two apple duplicated genes for each strawberry gene, with the occasional subclade showing a single apple gene, or three apple genes per strawberry gene. This 2:1 ratio of genes was robustly observed in the ABP1, TIR1/AFB and ARF class of genes, with Aux/IAA, GH3 and PIN showing a less robust pattern with some strawberry and apple genes showing no corresponding related sequences. The predicted location of the two apple duplicates genes were often found on homeologous chromosomes identified in the apple genome [38] (chromosomal locations are given in Additional file 1), with duplicated genes found on non-homeologous chromosomes occurring 18% of the time. Because of this tight phylogenetic relationship between the strawberry genes and two apple genes, when possible both apple homologues were named after the already annotated strawberry genes, for example ARF1 of strawberry clustered with genes in apple that were labelled ARF1 and ARF101. This nomenclature was not followed if the gene had previously been published or released in GenBank. With these genes, the existing name was kept (Additional file 1). While the gene families from the two Rosaceae species, strawberry and apple, were tightly aligned with each other, there was considerable divergence from the Arabidopsis genes. For example, there is evidence of clade expansion in the ARF genes in Arabidopsis including the AtARFs 12, 14, 15, 20, 21, 22, 23 ( Figure 4). The TIR1-like proteins also suggests small family expansion in Arabidopsis ( Figure 3). Small subclades containing only proteins from strawberry, apple and tomato were also observed.

Expression analysis of auxin-related genes during apple fruit development
Gene expression for each of the auxin-related genes was screened using quantitative polymerase chain reaction (PCR) across flowering and at time points representing the four stages of apple fruit development (0, 14, 45, 90, 132 DAFB) (Additional file 4). Some of the homeologous genes showed very little sequence divergence at the DNA level, making it hard to select optimal qPCR primers that were specific for each gene in the homeologous pairs. Of the 108 genes tested, 25 primer pairs were predicted to  GH3 (Group II) 15 9 10 be unable to differentiate between the homeologues (Additional file 4). In these instances, the expression patterns are given with both names ( Figure 5) or marked with an asterisk (Figures 6, 7). The genes were divided into three main functional groups for analysis: receptors ( Figure 5), homeostasis ( Figure 6) and response ( Figure 7). When it was possible to differentiate the homeologous genes, the expression patterns within each pair were compared with each other. There were a considerable number of homeologous genes with different expression patterns and also instances where one of the homeologous genes was apparently switched off, for example ABP1 and ABP101 ( Figure 5). This may be because of the quality of the primer, or actual lack of expression. To address this concern, we examined the EST libraries [33] for ESTs corresponding to the 10 receptor-like genes, and used diagnostic DNA polymorphisms to identify each of the homeologues. From these sequence data, it appears that ABP1 is transcribed and ABP101 is not (confirming the qPCR data). For the TIR1/AFB class there was a similar pattern, with TIR1, AFB102, and AFB106 having ESTs, while TIR101, AFB2, and AFB6 were not represented, suggesting that the transcription of these genes is suppressed. For the apparent lowly expressed AFB5 and AFB105 there were no ESTs with which to compare. Both the ABP1 and TIR1 class of receptor genes were expressed constitutively through fruit development. In the TIR1/AFB family, AFB102 showed a Stage 1 and 2 predominant expression, while AFB106 showed a high degree of expression at Stage 4 fruit development ( Figure 5). PIN103 showed the highest expression only during the cell division (Stage 1), while PIN4 and PIN105 showed predominant expression as the fruit progressed into maturity and ripening (Stages 3-4) ( Figure 6A). Half the GH3 conjugating proteins showed high expression levels early in fruit development -cell division (Stages 0 and 1), while all were suppressed during Stages 2-3 (cell expansion), with a number increasing again as the fruit ripened ( Figure 6B).
Many of the ARF/Aux/IAA family of genes had discrete expression patterns for a single stage of fruit development (Figure 7). Half the differentially expressed genes were predominantly expressed during the early stages of fruit development, with a high proportion of Aux/IAA genes being highly expressed at full bloom. As auxin has been shown to be central to the regulation of fruit size, which is determined during cell division and cell expansion, we were particularly interested in profiles that had high expression in Stages 1 and 2. In this cluster, four ARF and two Aux/IAA had highest expression during Stage 1. While ARF212 had peak expression at Stage 2, it had a very low expression level. Only two genes had high expression during both these two stages (ARF7 and ARF106) (Figure 7), suggesting these two genes may play a role in

QTL mapping for fruit size
To assess regions controlling fruit size in apples, two mapping populations were assessed for fruit weight, as a surrogate for fruit size, over a number of successive years. Fruit weight was phenotyped for a total of 572 and 123 genotypes from 'Royal Gala' × 'Braeburn' (RGxBB) and 'Starkrimson' × 'Granny Smith' (STKxGS) mapping populations. Measurements were made over 2 and 5 years of production, and two and one sites, respectively. Analysis of variance showed that the genotypic (G) and the year (Y) effects for fruit weight were highly significant for each population (P values < 2.2e-16). For the RGxBB population, the effects of the growing environment (E; P value < 2.2e-16) and its interaction with the genotype (GxE; P value < 2.2e-16) were highly significant. Best Linear Unbiased Predictors (BLUPs) independent of year and environment were extracted for each genotype for both studied populations and were used as phenotypic data for QTL detection. Six QTL regions were identified for fruit weight using the RGxBB and STKxGS genetic maps on Linkage Group (LG) 5, 8, 11, 15, 16 and 17 ( IAA14  ARF117  ARF102  ARF101  ARF9  IAA13  IAA24  ARF109  ARF107  IAA18  IAA229  IAA4  IAA6  IAA118  ARF105  IAA16A  IAA106  ARF5  IAA222  ARF4*  ARF103  ARF112  IAA121  IAA107  ARF104 IAA1 ARF110*  For each QTL, the table displays the marker used as a cofactor for MQM analysis, the LOD score, and the percentage of genetic variation explained by each single QTL (R 2 ). When several QTLs were detected for the same trait, the global R 2 (the proportion of variation explained by the QTLs) and the interactions between QTLs are indicated.
The six QTL regions were compared with the in silico locations of auxin genes that showed predominant expression during Stages 1 and 2. ARF106 is located on LG 15 and could account for the fruit weight QTL on this linkage group. The genetic marker developed from the sequence of ARF106 had the highest LOD score (logarithm of odds) for the fruit weight QTL, in both the 'Royal Gala' and 'Starkrimson' × 'Granny Smith' genetic maps ( Table 2 and Figure 8). The QTL interval spanned an overlapping area between both STKxGS and RGxBB QTLs, of 1.92 Mb, from markers CH03b10 (35.397 Mb) to GDsnp01971 (37.346 Mb). Within this region, in addition to ARF106 (MDP0000232116), 132 other predicted gene models were found (Additional file 6). While some of these genes showed homology to Arabidopsis genes that have annotated gene ontology, which may control fruit size, such as involvement in cell division, cell cycle and signal transduction (Additional file 6), these were not studied further as they were outside the scope of this project.
The LG 8 and 15 QTLs were mapped in duplicated regions of the apple genome because of whole genome duplication [38]. The homeologous gene to ARF106 is ARF6. While expressed later in fruit development, there is still the possibility that ARF6 retains some residual fruit size control. The genetic marker developed from ARF6 mapped to the fruit size QTL but was not the strongest marker in this region (Table 2 and Figure 8).

Discussion
With the availability of the whole genome sequence for apple [38], there is now the possibility of identifying the complete gene family for different classes of genes. This powerful resource has been used to identify individual auxin gene classes (mostly ARF and Aux/IAA) in a number of plants including Arabidopsis [41], poplar [42], maize [43,44] and tomato [45]. The auxin signal has been implicated in many components of fruit growth and development, including determination of fruit size through cell expansion, as well as the control of fruit ripening, and the regulation of fruit drop [46]. Here we have presented a genomics screen in apple of different auxin-related genes, covering perception, homeostasis and transcriptional regulation, and their relative expression patterns over fruit development. These expression data provide the ground work for further studies in the role of auxin on apple fruit growth and development.

Fruit size determination
There are a number of factors that regulate fruit size in apples, those that are controlled by the genetic make-up of the apple (size potential), and those related to the environment in which the apple develops. The environmental control of size is determined by both the effects of orchard husbandry and irrigation and local effects in the apple tree, such as fertilisation success, and localised crop load. We demonstrated that in the RGxBB segregating population, the effect of the year, the environment and its interaction with the genotype were highly significant. Furthermore, a significant difference was observed in the global phenotypic variation explained by the QTLs for fruit size detected in the two segregating populations. The larger explained phenotypic variation for the STKxGS population can be explained by the fact this population was grafted on less dwarfing rootstock ('Pajam') than the RGxBB ('M.9'). In addition, the crop load for the RGxBB was regulated, while the STKxGS was not. While the environment can cause significant amounts of variation, the genetic size potential of the fruit is a major determinant of fruit size. This work and others have shown that size potential is a complex multi-loci trait in apples [47] and other fruits such as tomato [48,49]. The control of fruit size in apples has been linked to both the number of cell division steps that occur directly following pollination and to subsequent cell expansion [50]. Apples have been shown to have a range of different cell sizes across different cultivars [51], and increasing cell number and size through endo-reduplication causes a 38% increase in fruit weight [52]. Our work suggests that auxin signals through this cell division/cell expansion phase may be modulated by an ARF gene (ARF106) that is up-regulated during these developmental time points and co-locates with a stable QTL for fruit weight. ARF106 is most closely linked to the AtARF17, which is microRNA controlled and when over-expressed using a microRNA-resistant form, gives a pleotropic phenotype [53] including excess tissue growth in leaves. The corresponding ARF in tomato SlARF17 is expressed highly at fruit ripening (named SlARF13 in [45], see Additional file 2).
The balance of auxin is critical for fruit expansion. This auxin signalling is complex, as not only can the presence of auxin elicit a developmental response, but different concentrations can cause different responses [54]. Here we found that injecting different concentrations of auxin could cause an increased cell expansion, decreased fruit growth, and ultimately fruit drop [55]. The increased cell growth with lower amounts of injected auxin suggests that fruit growth is at least in part limited by auxin concentration, as application of more can enhance it, which is consistent with the observations of Percy and collaborators [56].
Extensive molecular research on auxin in Arabidopsis has now identified a number of key genes involved in the regulation of auxin content of plant tissue, and the method by which the auxin signal is converted into a developmental change. Part of the complexity of the auxin signal transduction can be explained by the regulation of response genes: by developmentally regulating both the signal transduction transcription factors (ARF) and the modulators of transcription (Aux/IAA), a complex network of regulation can be achieved. This complexity can explain how the same signalling molecule can relay different signals at different times during development [6].

Auxin-related processes in fruit set
While the molecular control of auxin is best understood in the model plant Arabidopsis, which bears dry dehiscent fruit, there has been a substantial body of work in fleshy fruit species. Surprisingly, for many of the ARF proteins, the tomato orthologue (Asterid) is more closely related to the apple gene than the Arabidopsis orthologue (a Rosid, like apple). The ARF proteins, which are more similar to those in tomato (MdARF1, 101, 11, 111, 8, 108, 4, 104, 13 113, 3, 103) are all expressed at a specific stage in fruit development (Figure 7). This may suggest a strong evolutionary pressure for conservation within fleshy fruit species. This conservation is also observed in the expression patterns presented in [45], with genes such as MdARF113/ SlARF4, MdARF4/SlARF19 expressed at similar times in fruit development. The phylogenetic discrepancy is also observed in the IAA cluster (Additional File 3); however, there are currently no published genomics studies of IAA genes in tomato to make this comparison. Key auxin-related genes such as AtARF8 (FWF) in Arabidopsis and SlARF7 in tomato have been associated with fruit set [26,57]. Mutations in AtARF8 and down-regulation of SlARF7 cause parthenocarpic fruit in Arabidopsis and tomato respectively, indicating that AtARF8 is functionally equivalent to SlARF7, while it is not the closest homologue ( Figure 4) [58]. Down-regulation of SlIAA9 also leads to development of parthenocarpy and it has been hypothesised that it could work in the same pathway as SlARF7 [59]. SlARF7 clusters with ARF5 and ARF105, AtARF8 clusters with ARF17 and ARF117, while SlIAA9 clusters with IAA8, IAA27A and IAA127A (Figure 4, Additional file 3). Of these apple genes, ARF105 has highest expression at full bloom, like SlARF7 [26,45] (named SlARF9 in this later study), suggesting that this may play a similar role, while ARF5 is undetectable.

Auxin and ripening
The role of auxin in the regulation of ripening in strawberry has been well established after it was found that application of auxins through the peduncle caused a significant delay in ripening [1]. This link suggesting that auxin is a negative regulator of ripening was further  enforced by a study over-expressing a pepper GH3related gene in tomato [60]. In this study, the CcGH3 over-expressing lines matured earlier than untransformed lines and ripened earlier with the addition of exogenous ethylene. The closest homologue from of this CcGH3 in apple is GH3.1, which also shows a large increase in expression around fruit maturity ( Figure 6B). Interestingly, this gene is also expressed during the cell division stage. As well as the GH3-related genes, there is also a cluster of auxin signal transduction genes that are up-regulated at Stage 4 ( Figure 7), suggesting that an auxin signal can still be transduced at this stage. This is consistent with auxin playing a role in fruit maturation and ripening that is well known in non-climacteric fruits [1,10] and beginning to be established in climacteric fruit [61]. Another gene in tomato SlIAA3 is induced by both ethylene and auxin at fruit maturation [62]. SlIAA3 clusters with IAA6 and IAA106, neither of which is expressed at the ripening stage. However, 6 ARF/Aux/IAA genes are highly expressed at ripening, suggesting a similar role in apple. Down-regulation of another tomato auxin response factor, SlARF4 (also referred initially as DR12) [28,29], leads to unripe fruit that are firmer because of a perturbed pectin metabolism, and the fruit display an unusual cell division pattern in the pericarp. SlARF4 clusters with ARF13 and ARF113, of which ARF13 is expressed during cell expansion and maturation. This differs slightly from SlARF4, which shows increasing expression during tomato fruit development, with the highest in ethylene-producing fruits.

Conclusions
This work has provided a genomics study of auxin regulation in apples, with many auxin-related genes changing through fruit development. The complexity of expression patterns of these genes suggests a complex role of auxin regulation in apple fruit development. Exogenously applied auxin during the end of cell division/early cell expansion phase can increase fruit size, showing that auxin is at least in part one of the limiting factors controlling cell expansion. The role of one auxin response gene, ARF106, which maps to a size-related QTL, needs to be further investigated to determine if this gene plays a role in apple fruit size regulation.

Selection of genes in the apple genome
Auxin-related genes were selected by using BLASTP search of known Arabidopsis auxin-related genes against predicted apple protein sequences within the 'Golden Delicious' whole genome sequence [38]. The predicted MDP numbers were mapped to chromosome location using the apple Malus x domestica GBrowse from the Genome Database for Rosaceae (http://www.rosaceae.org/gb/ gbrowse/malus_x_domestica/) (Additional file 1). The protein sequences were compared with strawberry and Arabidopsis genes based on phylogeny. For each gene family, the longest highly conserved alignable protein sequence regions were used as detailed below: ABP1: the whole protein without the leader peptide. TIR1/AFB: The F-box domain and the leucine rich repeat domains. ARFs: The DNA binding domain. AUX/IAA domains I-IV. PINs: Transmembrane domains. GH3: The whole protein. Apple auxin-related genes were named firstly on published Gen-Bank accessions and secondly on the nearest strawberrynamed gene in the strawberry predicted protein sequences V2. Alignments and phylogenetic trees were generated using the Geneious Pro™ 5.4 (Biomatters). Multiple alignments were performed using MUSCLE and phylogenetic trees were built using neighbour joining with 1000 bootstraps. Sequences from the moss Physcomitrella patens were used to root the trees. Accession numbers of proteins from Arabidopsis, strawberry and tomato are given in Additional file 2.

Auxin injection in apples and assessment
Developing apple fruit were injected with different concentrations of IAA (Sigma-Aldrich, UK), dissolved in a 0.1% ethanol solution, 30 days after full bloom. Control apples were injected with a 0.1% ethanol solution. For each treatment, 100 μL were injected through the calyx using a syringe, into a minimum of 50 apple fruit; the apple diameter was measured on its equator, and the fruit tagged. Fifteen days after injection, the apples were harvested and the apple diameter was re-measured to establish the growth rate. For each treatment, the cortex tissue adjacent to the calyx from five representative apples was assessed using freeze fracture scanning electron microscopy (SEM). CryoSEM was performed using a Polaron PP2000 Cryo Transfer system (Quorum Technologies, Ringmer UK) attached to an FEI Quanta250 Scanning Electron Microscope (FEI Hillsboro OR). Blocks of apple tissue about 4 × 6 × 2 mm were placed in aluminium sample holders, held in brass transfer shuttles, using a mixture of colloidal graphite and OCT™ compound (Sakursa Finetek, Zoeterwoude, NL) as adhesive, so that a portion of the apple protruded from the surface. These were frozen in liquid nitrogen slush. Samples were transferred under vacuum to the PP2000 preparation stage, which was held at -150°C and the apple tissue fractured using a cooled metal blade or probe. The fractured surface was sputter-coated with gold/palladium (60 sec) and transferred to the SEM for observation on a stage cooled to -150°C using an accelerating voltage of 15 kV. Cell size was measured by counting the number of whole cells in the fracture window from each of the five treatment samples.

Auxin content measurement
IAA was extracted from 'Royal Gala' cortex (4 replicates) and seed (2 replicates) at different times during fruit development. Tissue samples were homogenised and IAA was extracted with 80% (v/v) methanol containing 250 mg l -1 butylated hydroxytoluene. 10 ng of [ 13 C 6 ]IAA, internal standard was added to extracts and left at 4°C for 24 h. The extracts were then filtered through Whatman no. 1 filter paper. Samples were reduced in volume to less than 1 mL under vacuum at 35°C and an aliquot was loaded onto a Sep-Pak C18 cartridge in 0.4% acetic acid. IAA was eluted with 50% methanol in 0.4% acetic acid. The eluate was dried and taken up in 1% acetic acid. Samples were then analysed using a Waters Acquity H-series UPLC coupled to a Waters Xevo triple quadrupole mass spectrometer. A Waters Acquity UPLC BEH C18 column (2.1 × 100 mm × 1.7 μm particles) was utilised. The solvents were 1% acetic acid in water (Solvent A) and acetonitrile (Solvent B) at a flow rate of 0.25 mL/min, with a linear gradient from 80% A:20% B to 50% A:50% B at 4.5 mins, followed by re-equilibration to starting conditions for 3 mins. Five μL of each sample was injected. The mass spectrometer was operated in positive ion electrospray mode with a needle voltage of 2.4 KV, and selected reaction monitoring was used to detect IAA and 13 C 6 IAA. The ion source temperature was 150°C, the desolvation gas was nitrogen at 1000 L/h, the cone gas flow was 50 L/h and the desolvation temperature was 300°C. The MS/MS transitions monitored were m/z 176.2 to 130.1 for IAA and 182.2 to 136.1 for 13 C 6 IAA. Cone voltage was 18 V and collision energy was 18 V. Dwell time was 161 ms per channel.
Data were analysed using MassLynx software. IAA and 13 C 6 IAA eluted at 3.74 mins under these conditions.

Gene expression analysis
'Royal Gala' apple fruit were harvested in the year 2006-07 from the Plant & Food Research Orchard, Hawke's Bay. For flowers, whole flowers were sampled; all fertilised apples had seed removed before harvesting. As the apples developed, sections of tissue from at least 10 apples were harvested (containing skin, cortex and core) into liquid nitrogen. RNA extraction and cDNA was generated as described in [30]. Gene expression was measured using quantitative PCR (qPCR) using Power SYBR ® green probes (Applied Biosystems, UK). Primers were used that gave a single melting peak for each of the genes assessed (Primer sequences can be found in Additional file 4). Quantitative PCR was conducted across two instruments: ARF and TIR gene expression were performed on a Roche LightCycler ® 480™ with the set up according to [63], whereas ABP1, PIN, GH3 and Aux/IAA were assessed using an ABI PRISM ® 7900 HT Sequence Detection System (Applied Biosystems) according to the method described in [64]. In all cases, actin was used as a reference gene.

Mapping fruit size
Two F 1 progenies were used to study the fruit size. The first contained a duplicated population of 590 seedlings from a 'Royal Gala' and 'Braeburn' cross (RGxBB), grafted onto 'M.9' rootstock, located at two locations in New Zealand (PRF research orchards in Havelock North and Motueka). The second contained a duplicated population of 123 seedlings from a 'Starkrimson' and 'Granny Smith' cross (STK×GS) grafted on the semi-dwarfing rootstock 'Pajam 1' located in France (Melgueil INRA Montpellier Experimental station). For the RGxBB seedlings, two seasons of apples were assessed. Apples were thinned to a low crop load of 4 fruit per cm 2 of trunk size and a minimum of five representative fruit sizes were weighed at harvest. For the STK×GS population, no thinning was undertaken. Five seasons of total apple crop on each tree were harvested weighed and counted, and the average weight of apple calculated per year.
Analysis of variance and linear models (in R software v.2.9.2 -R Development Core Team, 2009 [65]) were used to assess genetic and environmental regulation of fruit size at each site over the number of seasons measured. QTL analyses were performed using the RGxBB parental genetic maps and on the STKxGS consensus map [66,67]. JoinMap 3.0 [68] was used for constructing linkage maps. Fruit weight QTL intervals for each population were defined based on the peaks LOD-1 and LOD-2.
The list of predicted gene transcripts present within the QTL interval was extracted from GDR (http://www. rosaceae.org/projects/apple_genome) and we present the results of pairwise comparison of the Malus x domestica genome predicted genes against the Arabidopsis TAIR10_pep_20100802 database using BLASTP with an EXP cut-off < 1e-30.

Additional material
Additional file 1: List of auxin-related genes in apples. Table of Predicted Apple genes by MDP number [38], designated names and chromosome location (Gene and protein sequences can be obtained from GDR: http://www.rosaceae.org).