A natural mutation-led truncation in one of the two aluminum-activated malate transporter-like genes at the Ma locus is associated with low fruit acidity in apple

Abstract

Acidity levels greatly affect the taste and flavor of fruit, and consequently its market value. In mature apple fruit, malic acid is the predominant organic acid. Several studies have confirmed that the major quantitative trait locus Ma largely controls the variation of fruit acidity levels. The Ma locus has recently been defined in a region of 150 kb that contains 44 predicted genes on chromosome 16 in the Golden Delicious genome. In this study, we identified two aluminum-activated malate transporter-like genes, designated Ma1 and Ma2, as strong candidates of Ma by narrowing down the Ma locus to 65–82 kb containing 12–19 predicted genes depending on the haplotypes. The Ma haplotypes were determined by sequencing two bacterial artificial chromosome clones from G.41 (an apple rootstock of genotype Mama) that cover the two distinct haplotypes at the Ma locus. Gene expression profiling in 18 apple germplasm accessions suggested that Ma1 is the major determinant at the Ma locus controlling fruit acidity as Ma1 is expressed at a much higher level than Ma2 and the Ma1 expression is significantly correlated with fruit titratable acidity (R ² = 0.4543, P = 0.0021). In the coding sequences of low acidity alleles of Ma1 and Ma2, sequence variations at the amino acid level between Golden Delicious and G.41 were not detected. But the alleles for high acidity vary considerably between the two genotypes. The low acidity allele of Ma1, Ma1-1455A, is mainly characterized by a mutation at base 1455 in the open reading frame. The mutation leads to a premature stop codon that truncates the carboxyl terminus of Ma1-1455A by 84 amino acids compared with Ma1-1455G. A survey of 29 apple germplasm accessions using marker CAPS₁₄₅₅ that targets the SNP₁₄₅₅ in Ma1 showed that the CAPS_1455A allele was associated completely with high pH and highly with low titratable acidity, suggesting that the natural mutation-led truncation is most likely responsible for the abolished function of Ma for low pH or high acidity in apple.

Introduction

Organic acids, many of which are intermediates in metabolic processes, play significant roles in fruit growth, maturation, ripening and softening. The level of organic acids greatly affects the taste and flavor of fruit, and consequently its market value. The major determinants of fruit acidity include malic acid, citric acid and tartaric acid. In mature apple fruit, malic acid is the predominant organic acid although other organic acids such as citric acid, fumaric acid and quinic acid are detectable (Zhang et al. 2010). Apple fruit varies widely in pH and titratable acidity (TA) levels. However, the acceptable range for dessert apple fruit is often measured within a range of 3.1–3.8 in pH or 3.0–10.0 mg/ml in TA, beyond either end of which, fruit acidity is either too high or too low for fresh consumption (Brown and Harvey 1971; Nybom 1959; Visser and Verhaegh 1978).

Inheritance of high pH or low TA in apple fruit was attributed to a recessive gene in early studies (Brown and Harvey 1971; Nybom 1959; Visser and Verhaegh 1978). The acidity locus was mapped to linkage group (LG) 16 and designated as Ma (malic acid), where Ma is noted for the dominant low pH or high acidity allele and ma for high pH or low acidity allele (Maliepaard et al. 1998). In other species, major genes or quantitative trait locus (QTL) similar to Ma in controlling fruit acidity include acitric in citrus (Fang et al. 1997), SS in pomegranate (Jalikop 2007) and pH in sweet melon (Lerceteau-Köhler et al. 2012), where low acidity is also inherited recessively. The major gene D in peach, however, acts differently with low acidity being dominant over high acidity (Boudehri et al. 2009) although both peach and apple are members of the Rosaceae family. In tomato, complex and multiple QTLs are reported in conditioning fruit acidity levels (Fulton et al. 2002).

The primary role of the Ma locus in determining fruit pH and TA in apple was also demonstrated in QTL studies as a major QTL was consistently detected on LG 16 (Kenis et al. 2008; Liebhard et al. 2003; Xu et al. 2011). In addition to the Ma QTL, multiple minor QTLs of significant effect on acidity were identified in these studies. Although the minor QTLs are less consistent, the notion that the Ma locus and minor QTLs collectively determine fruit acidity levels is widely accepted. Consistent with this notion, a recent report finds that a mixed model of a major gene and polygenes fits best in explaining the apple acidity variation in a complex breeding population among four models (mixed, Mendelian, polygenic and environmental) tested (Iwanami et al. 2012).

Malate metabolism in fruit cells may involve several pathways according to recent reviews (Beruter 2004; Sweetman et al. 2009). Malate synthesis is considered to occur locally in fruit. The primary path is glycolysis of hexoses derived from sucrose and/or sorbitol, which are imported from leaves, in the cytosol of parenchyma cells of fruit. Depending upon developmental stages, pathways of photosynthesis in the chloroplast, the tricarboxylic acid (TCA) cycle in mitochondrion, and glyoxylate cycle in glyoxysome in fruit cells also appear to be important for malate synthesis. For degradation of malate, gluconeogenesis and the TCA cycle are likely the main pathways. It is possible that the various enzymes involved in malate synthesis and degradation, such as phosphoenolpyruvate carboxylase (PEPC), NADP-dependent malic enzyme (NADP-ME), and NAD-dependent malate dehydrogenase (NAD-MDH) and many others, may play a role in regulating malate metabolism in fruit cells, thus acidity of fruit. In addition, the vacuolar transporters, such as the vacuolar pumps, e.g. V-ATPase (Schumacher and Krebs 2010), tonoplast dicarboxylate transporter, e.g. AttDT (Emmerlich et al. 2003), and members of the aluminum-activated malate transporter1 (ALMT1) family proteins (Barbier-Brygoo et al. 2011), e.g. AtALMT9 (Kovermann et al. 2007) and AtALMT6 (Meyer et al. 2011), may also play critical roles in determining fruit acidity as they can regulate the malate accumulation in and release from the vacuole in plant cells.

In apple, the pattern of malate accumulation and degradation is similar in developing fruits of several high/medium acid varieties studied, i.e. malic acid level significantly increases in young fruit (around 4 weeks after full bloom) and then progressively decreases through maturity although the total content per fruit increases along with fruit development (Beruter 2004; Hulme and Wooltorton 1957; Ulrich 1970; Zhang et al. 2010). Several recent studies have attempted to identify candidate genes and/or enzymes that may be associated with the acidity variations in apple fruit. Using a cDNA-AFLP-based approach, a gene designated Mal-DDNA (DQ417661) of unknown function previously appeared to be associated with low acid in a population segregating for fruit acidity (Yao et al. 2007). Direct profiling of expression patterns and enzyme activities of genes putatively involved in malate metabolism, including MdPEPC (EU315246, for PEPC), MdcyME (DQ280492, for NADP-ME) and MdVHA-A (EF128033, for subunit A of vacuolar H⁺-ATPase), found that there were differences between low and high acid genotypes (Yao et al. 2009). Involvement of genes encoding NADP-ME (GD254910, degradation of malate) and NAD-MDH (GD254856, synthesis of malate) in malate accumulation and degradation was also reported in a cDNA microarray analysis of 1,536 genes (Soglio et al. 2009). Moreover, a gene encoding NAD-MDH (DQ221207) has been functionally demonstrated to be involved in malate synthesis in apple (Yao et al. 2011). Overall, these data suggest that the genes and/or enzymes studied above may contribute to the variation of fruit acidity.

However, a detailed analysis of a low acid variety Usterapfel and its high acid mutant (Beruter 1998, 2004) indicated that key enzymes in malic acid metabolism, PEPC, NAD-MDH and NADP-ME, may not play a key role in determining the difference in fruit acidity because there was no difference in the catalytic activity of these enzymes between the two contrasting genotypes. Examining the localities of these genes in the apple genome (Velasco et al. 2010) appeared to support that these enzymes and genes involved in malate metabolism may not be Ma because none of those studied above, including Mal-DDNA, is on chromosome 16 where the Ma gene resides.

To uncover the genes underlying Ma, we had defined the Ma locus to a region of 150 kb encompassing 44 predicted genes on chromosome 16 in the Golden Delicious genome in a previous study (Xu et al. 2011). In this study, we report the identification of two aluminum-activated malate transporter (ALMT)-like genes, Ma1 and Ma2, as strong candidates of Ma. We show that the Ma region is reduced to a genomic segment of 65 kb containing 19 predicted genes in Golden Delicious by developing three new markers and analyzing two more populations. In two bacterial artificial chromosome (BAC) clones that are distinguishable with haplotype ma and Ma from apple rootstock G.41, the Ma region harbors 12 predicted genes, including Ma1 and Ma2, although it spans over 71 kb in haplotype ma and 82 kb in haplotype Ma. We further show that the expression of Ma1 is significantly correlated with fruit acidity levels, whereas Ma2 is expressed constantly at low levels across high and low acidity fruit. Finally, we show that a single nucleotide mutation in the open reading frame of Ma1 that leads to truncation of Ma1 by 84 amino acids is perfectly associated with high pH and highly with low TA in 29 apple germplasm accessions studied.

Materials and methods

Plant materials and fruit pH and TA evaluation

Four half-sib F₁ populations of interspecific crosses were used to further narrow down the Ma locus, namely GMAL 4590, GMAL 4592, GMAL 4595 and GMAL 4596 (Table 1). The seed parent of the four populations is Royal Gala (Mama), a widely grown apple cultivar (Malus × domestica Borkh.). The pollen parents are elite clones of M. sieversii (i.e. of fruit size close to cultivated apple) collected from Kazakhstan (Forsline et al. 2003), including PI 613971 (Mama), PI 613978 (mama), PI 613988 (Mama) and PI 613979 (Mama), respectively. M. sieversii has been proven to be the major progenitor species of M. × domestica (Velasco et al. 2010). The four F₁ populations were derived from controlled crosses made in 2002 and planted on their own seedling roots in 2004 in the USDA-ARS Apple Germplasm Repository, Geneva, NY, USA. Populations GMAL 4590 of 216 individuals and GMAL 4595 of 222 genotypes were used in a previous study (Xu et al. 2011), but 36 and 23 individuals that did not bear fruit in 2010 from the two crosses, respectively, were not included previously. These individuals bore fruit in 2011 and were added in this study (Table 1). Populations GMAL 4592 (155 genotypes) and GMAL 4596 (215 genotypes) were used for the first time. Overall, there are 724 fruiting individuals in a total of 808 genotypes in the four populations (Table 1).

Table 1 List of populations segregating for fruit pH

Evaluation of fruit maturity and fruit acidity (pH paper estimates and instrumental measurements of pH and TA) was conducted similarly as described previously (Xu et al. 2011). Briefly, fruit maturity was determined via starch test that corresponds to Cornell Starch Index 4.0–6.0 (Blanpied and Silsby 1992). For pH estimates, pH paper (Hydrion Papers, pH 3.0–5.5, Micro Essential Laboratory Inc., Brooklyn, NY, USA) was applied onto the fruit cuts at maturity in the orchard. For pH and TA instrumental measurements, fruit juice samples were prepared by pooling 5–10 fruits per genotype at maturity. The pooled juices were then measured with a pH meter (Accumet AB15, Fisher Scientific, Pittsburgh, PA, USA), and subsequently, an autotitrator (Metrohm 848 Titrino Plus and Metrohm 869 Compact Sample Changer, Herisau, Switzerland). Evaluation for most genotypes was conducted either in 2010 or 2011. But for the informative recombinants between markers CH05c06 and CH02a03 or CH05a09 (Figs. 1a–d, 2), pH meter-based measurements were obtained in both years if fruits were available.

Fig. 1

Fine genetic and haploid specific physical maps of the Ma locus on chromosome 16. Fine genetic maps of Ma in PI 613988 (a), Royal Gala (b), PI 613971 (c) and PI 613979 (d). The number between the markers stands for the number of informative recombinants found in the interval. The solid vertical lines indicate the position of mapped markers, and the broken vertical lines are for positions of the presumed markers. e Physical map of the Ma region (a Genome Brower snapshot from the GDR website) in Golden Delicious (GD). The Ma region of 65 kb between markers CN889255SNP and 12514.266 is shown with a red solid bar. The labeled contigs indicate the source sequences, from which the markers were developed. f Predicted genes in the Ma region of GD. There are 19 predicted genes, which are conveniently labeled with #10–28, respectively. g A sequenced clone BAC21 of G.41 covering the Ma region. The numbers show the physical locations of the corresponding genes predicted in GD. h A sequenced clone BAC3 of G.41. i A list of the 19 genes predicted. Genes not present in G.41 (in purple): MDP0000375685 (#19), MDP0000258718 (#23) and MDP0000357895 (#26). Genes outside of the Ma region in G.41 (in grey): MDP0000250967 (#24) and MDP0000157412 (#27). Genes spliced alternatively (in blue): MDP0000241811 (#17) and MDP0000141005 (#18). Genes with duplicated IDs (in orange): MDP0000134560 (#22) and MDP0000139500 (#28). Candidate genes of Ma (in black): 12 genes, including MDP0000252114 (Ma1) and MDP0000244249 (Ma2) (colour figure online)

Fig. 2

Key informative recombinants identified from populations GMAL 4590, 4592, 4595 and 4596 and their marker genotypes. ^a 4590 = GMAL 4590; 4592 = GMAL 4592; 4595 = GMAL 4595; 4596 = GMAL 4596. Marker genotype linked to the low pH (high acid) allele Ma in coupling phase is shown in grey, and those linked to the high pH (low acid) allele ma in black; ^b pH data for these recombinants were measured by a pH meter in 2010 (for GMAL 4595) and 2011 (for the rest)

To examine the association between fruit acidity and the mutation at base 1455 in gene Ma1, pH and TA of mature fruit were evaluated for 29 representative apple cultivars and accessions (Table S1, including three progeny from GMAL 4595) grown in the USDA-ARS Apple Germplasm Repository, Geneva, NY, USA.

Marker development and genetic mapping

New simple sequence repeat (SSR) markers linked to Ma were developed using the same strategy as described in Xu et al. (2011). Briefly, DNA sequences of contigs between the two existing markers 532.669-2 and 20159.150-1 (Xu et al. 2011) were downloaded from the Genome Database for Rosaceae (GDR, http://www.rosaceae.org/), and analyzed for the presence of potential SSRs markers using the web-based program BatchPrimer3 (http://probes.pw.usda.gov/batchprimer3/index.html) (You et al. 2008). Genomic DNA isolation, PCR and SSR analyses were conducted as described previously (Xu et al. 2011).

For single nucleotide polymorphism (SNP) marker development, we targeted expressed sequence tags (ESTs) that are present in the region between the two markers 532.669-2 and 20159.150-1. The presence and segregation of SNP were determined by direct sequencing of the PCR products amplified from the five parents and the informative recombinants between markers CH05c06 and CH02a03 or CH05a09.

CAPS₁₄₅₅ is a cleaved amplified polymorphic sequence (CAPS) marker targeting base 1455 in the open reading frame of gene Ma1. The PCR program includes 2 min at 98 °C, 35 cycles of 10 s at 98 °C, 15 s at 55 °C and 90 s at 72 °C, and a final 5 min at 72 °C. PCR were conducted in a volume of 20 μl, which includes 1× PrimeSTAR^® MAX DNA Polymerase (R045A, Takara/Clontech, Mountain View, CA, USA), 0.5 mM of each primer and 30 ng of genomic DNA. Restriction digestion was performed at 37 °C for overnight in a volume of 20 μl that contains 10 μl PCR products, 2 U of BspHI (New England Biolabs, Ipswich, MA, USA), 1× NEBuffer 4. Gel analysis of CAPS₁₄₅₅ was conducted with agarose gels of 1.5 % (w/v).

Mapping of markers in relation to the Ma locus was conducted with the informative recombinants between SSR markers Hi22f06 and CH02a03 or CH05a09, which were identified from the four populations described above. The informative recombinants, as explained previously (Xu et al. 2011), refer to individual trees developed from zygotes that combined a parental-type gamete of an allele of ma (non-recombinant) with a recombined gamete near the Ma locus. Recombinants derived from zygotes that include a parental-type gamete of an allele of Ma (non-recombinant) are considered non-informative in this study. This is because the strong dominance effect of allele Ma from the parental-type gamete would make the effect of allele Ma or ma from a recombined gamete difficult, if not impossible, to detect by pH or TA values.

Identification of BAC clones and sequencing

The BAC library was constructed from G.41, an apple rootstock developed from an interspecific cross Malling 27× Robusta 5 (Cummins et al. 2006). The mature fruits of G.41 are small (2–3 cm in diameter) and have astringent taste (not edible) and high acidity (pH 3.1, TA = 13 mg/ml), suggesting that G.41 has a genotype of MaMa or Mama. The BAC library was constructed by Amplicon Express (Pullman, WA) using a restriction enzyme/vector combination of MboI/pECBAC1. It has a total of 41,472 clones with an average insert size of 120 kb, which provides approx. 6.6× coverage of the apple genome. The library was pooled at two levels with a total of five dimensions. The first level is the nine super pools, each of which comprises 12 plates containing a total of 4,608 (12 × 384) clones. The second level is the nine sets of matrix pools, and each set was pooled from the 12 plates associated with one of the nine super pools. One set of matrix pool includes eight matrix plate pools (P:1.2.3, P:4.5.6, P:7.8.9, P:10.11.12, P:1.5.9, P:2.6.10, P:3.7.11 and P:4.8.12) pooled from three of the 12 individual plate pools, eight matrix row pools (R:A.B.C.D, R:E.F.F.G, R:I.J.K.L, R:M.N.O.P, R:A.E.I.M, R:B.F.J.N, R:C.G.K.O and R:D.H.L.P) pooled from four of the 16 individual row pools, ten matrix column pools (C:1.2.3.4.5.6, C:7.8.9.10.11.12, C:13.14.15.16.17.18, C:19.20.21.22.23.24, C:1.7.13.19, C:2.8.14.20, C:3.9.15.21, C:4.10.16.22, C:5.11.17.23 and C:6.12.18.24) pooled from six or four of the 24 individual column pools, and ten matrix diagonal pools (D:1.2.3.4.5.6, D.7.8.9.10.11.12, D.13.14.15.16.17.18, D.19.20.21.22.23.24, D:1.7.13.19, D:2.8.14.20, D:3.9.15.21, D:4.10.16.22, D:5.11.17.23 and D:6.12.18.24) pooled from six or four of the 24 individual diagonal column pools. Different from a common individual column pool, which is pooled from the same column across a stack of 12 plates, an individual diagonal column pool comprises 12 varying columns on the diagonal line from the stack of 12 plates. For example, diagonal column pool D1 is pooled from column (C) 1 in plate (P) 1, C2 in P2… and C12 in P12, and pool D2 is from C2 in P1, C3 in P2… and C1 in P12.

Screening of BAC clones was conducted on the library super pools and their associated matrix pools using the Ma-linked PCR-based markers we developed. BAC clones originated from the Ma region were restricted with endonuclease BamHI and NotI (New England Biolabs, Ipswich, MA, USA) and then analyzed by pulse field gel electrophoresis (PFGE) using CHEF-DR II System (Bio-Rad, Hercules, CA, USA) for preliminary fingerprinting and size estimation. BAC sequencing was conducted using a 454 GS FLX system at Cornell Biotechnology Center and assembled with the Newbler Assembly (454 Life Sciences, Branford, CT, USA).

Gene prediction and annotation at the Ma locus

Genes predicted in the Ma region of Golden Delicious (Velasco et al. 2010) were adopted and their coding sequences (CDS) and deduced protein sequences were downloaded from GDR. Confirmation of gene annotation was carried out by searching the GenBank non-redundant protein database using the BLASTP program with a cutoff expected value of 10⁻⁹. Putative functions of the predicted genes were annotated with the GenBank accession numbers of the highest similarities and associated functions if known.

Quantitative (q) RT-PCR assay of Ma candidate genes

Total RNA from mature fruits of 18 of the 29 apple accessions (Table S1) was isolated using Spectrum™ Plant Total RNA Kit (Sigma-Aldrich, St. Louis, MO, USA) with three biological replicates. Reverse transcription reactions were carried out with 1.8 μg of total RNA using the Superscript III RT (Invitrogen, Carlsbad, CA, USA). The resulting first strand cDNA was diluted by fivefold, and then used as templates for qRT-PCR analysis, in which a Malus (Gala) actin gene/EST (EB136338) served as a reference with primers Actin F (5′-GGCTGGATTTGCTGGTGATG-3′) and Actin R (5′-TGCTCACTATGCCGTGCTCA-3′).

Two rounds of qRT-PCR were performed. In the initial round, all 12 genes predicted at the Ma locus were screened with their gene-specific primers (Table S2). Four low acid (Britegold, KAZ 96 08-17, Novosibirski Sweet and Sweet Delicious) and four high acid (Cox’s Orange Pippin, Golden Delicious, Marshall McIntosh and Winter Majetin) apple accessions were used. cDNA of each genotype was bulked evenly from the three replicates and then used for qRT-PCR. In the second round, three selected genes (Ma1, Ma2 and MDP0000141005) were analyzed in detail with all 18 apple accessions. The gene-specific primers (Table S2) for Ma1 are Ma1F (5′-CGTCATGGTGTCTGGAACAT-3′) and Ma1R (5′-CTCCATGGCAAAAACCTGTC-3′), and those for Ma2 are Ma2F (5′-TCGGAAGACGGCCTAATGGA-3′) and Ma2R (5′-TTGAAGCCGGGCAACAAACT-3′). These gene-specific primers were designed to cover the known alleles of Ma1 and Ma2 (Figs. S2, S3).

qRT-PCR was conducted using Roche (Indianapolis, IN) LightCycler 480 Real-Time PCR System. For each qRT-PCR, a final volume of 20 μl was used, which contained 5 μl of the cDNA dilutions, 0.5 μM of the forward and reverse primers, and 1× SYBR Green Master Mix (Roche Cat. # 04707516001). The qRT-PCR program included an initial denaturation step of 10 min at 94 °C, 45 cycles of amplification using 10 s at 94 °C, 30 s at 58 °C, and 25 s at 72 °C, and a dissociation stage of 5 s at 95 °C, 60 s at 60 °C, and 15 s at 97 °C. Expression quantification and data analysis were performed by LightCycler 480 Software (Version 1.5) using the comparative cycle threshold method (Pfaffl 2001). Regression analysis between the gene expression and fruit acidity variation was performed using MS Excel 2007.

Phylogenetic analysis

Phylogenetic analysis of the deduced protein sequences of the Ma candidate genes, Ma1 and Ma2, was conducted along with the members of the ALMT1 family in Arabidopsis, which sequences were downloaded from TAIR 10 (http://www.arabidopsis.org/index.jsp), using MEGA4 (Tamura et al. 2007).

Results

Delimiting the Ma locus to a 65 kb genomic segment on chromosome 16

Segregation of fruit pH in populations GMAL 4590 and GMAL 4595 had been studied previously, and the three parents Royal Gala, PI 613971 and PI 613988 had been determined of heterozygous genotype Mama (Xu et al. 2011). With additional fruiting individuals included, i.e. 36 in GMAL 4590 and 23 in GMAL 4595, the low pH (≤3.8) and high pH (≥3.9) segregation remained unchanged with the expected ratio of 3:1 (Table 1). In population GMAL 4596, fruit pH segregated similarly with the ratio 3:1 (156:42, P = 0.22), suggesting PI 613979, the pollen parent of GMAL 4596, is of a heterozygous genotype Mama as well. However, 82 low and 51 high pH genotypes were scored in population GMAL 4592, indicating a significant deviation from the 3:1 ratio (P = 0.0004). Examining the markers linked to Ma (Fig. 1) revealed that none of them segregated for the pollen parent PI 613978 while all segregated normally for Royal Gala. Moreover, the markers that segregate for Royal Gala alone predicted the segregation of pH (data not shown), suggesting that PI 613978 has a genotype of mama. Given the known genotype Mama of Royal Gala, pH is expected to segregate 1:1 in population GMAL 4592. But the observed ratio of 82:51 distorted significantly from 1:1 (P = 0.007) (Table 1).

Three new markers, including two SSRs 12514.266 and 12995.82-2, and one SNP CN889255SNP, were developed (Fig. 1a–d; Table 2) between the existing two markers 532.669-2 and 20159.150-1 that defined the Ma region previously (Xu et al. 2011). For map integration, the three new markers were assessed with a total of 52 informative recombinants between markers CH05c06 and CH02a03 or CH05a09, including 17 mapped in GMAL 4590, 7 in GMAL 4592, 19 in GMAL 4595, and 9 in GMAL 4596 (Fig. 1a–d). Out of the 52 informative recombinants, 14 were the most informative in ordering the markers (Fig. 2). SSR marker 12995.82-2 along with the existing marker 18695-28-2 cosegregated with Ma, and markers CN889255SNP and 12514.266 flanked Ma immediately to narrow the Ma locus down to a smaller genetic interval on chromosome 16 (Figs. 1a–d, 2). This genetic interval of Ma was supported with four most informative recombinants, including GMAL 4595-6-149 and GMAL 4590-1-131 between marker CN889255SNP and Ma, and GMAL 4592-4-33 and GMAL 4595-6-121 between Ma and marker 12514.266 (Fig. 2). In physical terms, the Ma interval corresponds to a genomic segment of 65 kb on chromosome 16 in Golden Delicious (Fig. 1e), which was reduced from a 150 kb region defined previously (Xu et al. 2011).

Table 2 Primer sequences and other relevant information of markers developed in the Ma region

Haplotypes of the Ma locus

The draft sequence of the Golden Delicious genome does not provide clear haplotype information although M. × domestica is a highly heterozygous species. To understand the possible sequence variation and local genomic structure and organization that may discriminate allele Ma from ma, we identified two BAC clones BAC3 and BAC21 from the BAC library of apple rootstock G.41 using three markers 18695.28-2, 12995.82-2 and 12514.266 simultaneously. The two BAC clones were confirmed to contain not only the three makers used to screen the BAC library, but also the PCR amplicon source for marker CN889255SNP (Fig. S1a), suggesting both BAC clones cover the Ma locus completely. Based on the band patterns associated with markers 18695.28-2, 12995.82-2 and 12514.266 (Fig. S1b) and the restricted bands generated by BamHI and NotI digestions (Fig. S1c), the two BACs are clearly of different haploid origin although the genotype of G.41 could be either Mama or MaMa. The estimated sizes for BAC3 and BAC21 were 150–160 kb and 110–120 kb, respectively (Fig. S1c).

Sequencing of the two BAC clones revealed that the Ma region extends over a larger segment of 71 kb in BAC21 and 82 kb in BAC3 in G.41 (Fig. 1g, h). Sequence alignment using BLAST demonstrated that BAC21 had higher overall sequence identity with the Golden Delicious contigs than BAC3 (data not shown), suggesting that BAC21 represents a haplotype likely closer to the two haplotypes in Golden Delicious than BAC3.

Identification of Ma1 and Ma2

The 65 kb genomic region of Ma contains 19 predicted genes in Golden Delicious (#10–28 in the 44 genes listed in Xu et al. (2011), Fig. 1f, i; Table S3). Aligning the 19 predicted genes with the two BACs indicated that three genes (MDP0000375685 (#19), MDP0000258718 (#23) and MDP0000357895 (#26)) were not found in the two BAC sequences, and two (MDP0000250967 (#24) and MDP0000157412 (#27) reside outside of the Ma region defined by the two markers CN889255SNP and 12514.266 in G.41. Moreover, MDP0000134560 (#22) and MDP0000139500 (#28) are duplicated gene IDs for a single gene, and MDP0000241811 (#17) and MDP0000141005 (#18) are alternatively spliced variants from another single gene. Therefore, the two BAC clones harbor 12 predicted genes at the Ma locus (Fig. 1g–i; Tables S2, 3), which include MDP0000252114 (#16), designated Ma1, and MDP0000244249 (#11), designated Ma2. Proteins Ma1 and Ma2 are putative members of the ALMT1 family and, respectively, share 57 % (338/595) and 55 % (302/553) of identity in amino acid sequence with AtALMT9, an Arabidopsis protein known to be a vacuolar malate channel involved in maintaining the cytosolic malate homeostasis (Kovermann et al. 2007). A search of Malus EST databases in GenBank found that there are 20 EST accessions of the origin of Ma1 (Table S4) and one EST (CN929391) matching with Ma2, suggesting both Ma1 and Ma2 are expressed genes, and therefore strong candidate genes of Ma.

There are two inversions in gene orders between Golden Delicious and G.41 (Fig. 1f, g): one between genes MDP0000130613 (#10) and MDP0000244253 (#15), and the other between genes MDP0000134560 (#22) and MDP0000247199 (#25) (Fig. 1f, g). Genes Ma1 and Ma2 are physically separated by one gene MDP0000130613 (#10) in both haplotypes of G.41 (Fig. 1g, h), but by four genes MDP0000244250 (#12), MDP0000244251 (#13), MDP0000130619 (#14) and MDP0000244253 (#15) in Golden Delicious (Fig. 1f).

qRT-PCR analysis of genes predicted at the Ma locus

To investigate the expression patterns in mature fruit, genes Ma1 and Ma2 as well as the other ten genes in the Ma region were screened alongside four low acid (Britegold, KAZ 96 08-17, Novosibirski Sweet and Sweet Delicious) and four high acid (Cox’s Orange Pippin, Golden Delicious, Marshall McIntosh and Winter Majetin) apple germplasm accessions (Table S1) using qRT-PCR. Gene Ma1 was expressed at much higher levels in high acid fruit than in low acid fruit while Ma2 was expressed consistently at low levels across both low and high acid fruit (data not shown). The correlation between gene expression and TA among the eight apple accessions was highly significant for Ma1 (R ² = 0.9430, P = 0.0001) but non-significant for Ma2 (R ² = 0.0559, P = 0.5729). Among the other ten genes, MDP0000141005, which encodes a putative serine/threonine protein phosphatase 2A (PP2A) regulatory subunit A, was expressed at high levels and showed a significant correlation with fruit acidity (R ² = 0.7428, P = 0.0059). The remaining nine genes were expressed at low levels and did not show correlations with fruit acidity (data not shown), and therefore we did not analyze them further.

A more comprehensive qRT-PCR analysis of Ma1, Ma2 and MDP0000141005 indicated that the relative expression levels of Ma1 remained high and were significantly correlated with TA (R ² = 0.4543, P = 0.0021) and pH (R ² = 0.4630, P = 0.0019) in fruit of 18 apple germplasm accessions (Fig. 3a, b). In contrast, the expression of Ma2 was low and showed no correlation with TA (R ² = 0.0086, P = 0.7148) and pH (R ² = 0.0356, P = 0.4531) (Fig. 3c, d). These data suggest that Ma1 may be the major factor in determining fruit acidity and the role of Ma2 would be limited if any. The correlation of MDP0000141005 expression with acidity was reduced to a non-significant level (R ² = 0.1497, P = 0.1126 for TA; R ² = 0.0916, P = 0.2222 for pH), allowing MDP0000141005 to be excluded from subsequent analyses.

Fig. 3

Regression between fruit acidity (TA and pH) and relative gene expression (Ma1 and Ma2) in 18 apple germplasm accessions. a–d Self explainable

Allelic variations of Ma1 and Ma2

The Ma1 allele in BAC3, designated Ma1-G41, differed by eight bases from that in BAC21, designated ma1-G41 (Fig. S2, Table 3). Examining the coding sequence of MDP0000252114 showed that nucleotides at seven positions are ambiguous, i.e. M = A/C (bases 118 and 162); K = G/T (bases 834 and 1304); W = A/T (base 1011) and R = A/G (bases 1286 and 1455), presumably caused by the two different haplotypes in Golden Delicious (Fig. S2, Table 3). To distinguish the two alleles of Ma1 in Golden Delicious, we compared the sequence of MDP0000252114 with both Ma1-G41 and ma1-G41. Excluding the seven ambiguous positions, MDP0000252114 differed by one base from ma1-G41, but by six bases from Ma1-G41, suggesting MDP0000252114 is much closer to ma1-G41 than to Ma1-G41. When the seven ambiguous positions were considered, one of the two possible bases at each of the seven positions matches with the base at their corresponding positions in ma1-G41. This set of seven bases was therefore inferred to be co-present in one allele of Golden Delicious, designated ma1-GD. The other set of seven bases was concluded to be co-present in the other allele of Golden Delicious, designated Ma1-GD (Fig. S2; Table 3).

Table 3 DNA and amino acid sequence variations in the Ma1 alleles of G.41 and Golden Delicious (GD)

In the deduced amino acid sequences, Ma1-G41 and Ma1-GD diverge by seven residues while there is no difference between ma1-G41 and ma1-GD (Fig. 4; Table 3). However, both ma1-G41 and ma1-GD are truncated by 84 amino acids at the carboxyl terminus compared with either Ma1-G41 or Ma1-GD (Fig. 4). This truncation is due to a nucleotide mutation from G to A at the 1455th base (SNP₁₄₅₅) in the open reading frame, leading to a pronounced change from a tryptophan (W) codon TGG₁₄₅₅ to a stop codon TGA₁₄₅₅ (Fig. 4; S2; Table 3).

Fig. 4

Alignment of the Ma1 deduced protein sequences. MDP252114 stands for the Golden Delicious protein MDP0000252114, which combines proteins Ma1-GD and ma1-GD. Each sign “–” in MDP252114 is for two possible amino acid residues, and annotated accordingly as shown. Amino acid residues that vary are highlighted in blue. The stop codon TGA₁₄₅₅ caused by SNP_1455A leads to a truncation of 84 amino acids at the carboxyl terminus in proteins ma1-G41 and ma1-GD compared with proteins Ma1-G41 and Ma1-GD

The allelic variations of gene Ma2 were investigated similarly (Figs. S3–4; Table S5). Briefly, the Ma2 allele in BAC3 and that in BAC21 were designated Ma2-G41 and ma2-G41, respectively, whereas Ma2-GD and ma2-GD were assigned as two alleles for Golden Delicious based on the MDP0000244249 sequence of four ambiguous positions, i.e. R = A/G (base 26), W = A/T (bases 165 and 951) and M = A/C (base 1245). There are 24 different bases (17 aa) between Ma2-G41 and ma2-G41, 4 bases (2 aa) between Ma2-G41 and Ma2-GD, and 22 bases (17 aa) between Ma2-G41 and ma2-GD. The coding sequences in alleles ma2-G41 and ma2-GD are identical (Figs. S3, 4; Table S5).

Allelic association of the Ma1 and Ma2 alleles with Ma and ma

To uncover which Ma1 and Ma2 allele is associated with Ma or ma, a CAPS marker, named CAPS₁₄₅₅, was developed to target SNP₁₄₅₅ using endonuclease BspHI, which cleaves site TCATGA₁₄₅₅ in the truncated alleles ma1-G41 or ma1-GD, but not TCATGG₁₄₅₅ in the intact alleles Ma1-G41 or Ma1-GD (Table 2). Agarose gel assay of maker CAPS₁₄₅₅ in population GMAL 4595 and the informative recombinants indicated that homozygous genotype CAPS _1455G CAPS _1455G cosegregated with MaMa, CAPS _1455G CAPS _1455A with Mama, and CAPS _1455A CAPS _1455A with mama (Figs. 1, 2, 5a), suggesting the intact allele of Ma1 (Ma1-1455G), such as Ma1-G41 or Ma1-GD, is associated with the high acid allele Ma while the truncated allele of Ma1 (ma1-1455A), such as ma1-G41 or ma1-GD, with the low acid allele ma. Consequently, alleles Ma2-G41 and Ma2-G41 are associated with Ma while ma2-G41 and ma2-GD with ma.

Fig. 5

Agarose gel analysis of marker CAPS₁₄₅₅. Bands of 2,013 bp correspond to allele CAPS _1455G , i.e. the Ma1-1455G allele for high acidity. The combined bands of 1,764 and 249 bp are expected for allele CAPS _1455A , i.e. the ma1-1455A allele for low acidity. A/A = genotype CAPS _1455A CAPS _1455A , A/G = genotype CAPS _1455A CAPS _1455G , and G/G = genotype CAPS _1455G CAPS _1455G . The numbers indicate fruit pH. a Lane 1: 1 kb Plus DNA Ladder (Invitrogen, CA). Lanes 2–12: 11 progeny of GMAL 4595. b Lane 13: 1 kb Plus DNA Ladder. Lanes 14–22: nine apple germplasm accessions as shown

Together with the analyses in the haplotypes at the Ma locus and allelic variations in the two genes Ma1 and Ma2 in G.41 and Golden Delicious, the allelic associations identified here conclude that BAC3 stands for a haplotype of Ma for high acidity while BAC21 represents a haplotype of ma for low acidity, and that the allele diversity is higher for the high acidity alleles, but none or low for the low acidity alleles.

Association of the mutation-led truncation in Ma1 with low fruit acidity in apple germplasm

To see how SNP₁₄₅₅ may explain the acidity levels in the other apple germplasm, a set of 29 (Table S1) representative apple germplasm accessions were analyzed with marker CAPS₁₄₅₅ (Figs. 5b, 6). Genotype CAPS _1455A CAPS _1455A is associated either exclusively with high pH (7/7, Fig. 6a) or tightly with low TA (7/9, Fig. 6b). Genotypes CAPS _1455G CAPS _1455G and CAPS _1455G CAPS _1455A , however, are associated either completely with low pH (22/22, Fig. 6a) or highly with high TA (20/22, Fig. 6b). These data indicate a complete or highly tight association between the mutation-led truncation in Ma1 (ma1-1455A) with low acidity in these apple accessions.

Fig. 6

Survey of marker CAPS₁₄₅₅ genotypes and their association with fruit pH (a) and TA (b) in 29 apple germplasm accessions (Table S1). A/A = genotype CAPS _1455A CAPS _1455A , A/G = genotype CAPS _1455A CAPS _1455G , and G/G = genotype CAPS _1455G CAPS _1455G

Discussion

Delimiting the Ma locus to a 65 kb genomic segment and identification of two ALMT-like genes Ma1 and Ma2

By developing three new markers and analyzing two additional populations, we delimited the Ma locus between markers CN889255SNP and 12514.266. The genetic interval was supported by four recombinants with GMAL 4595-6-149 and GMAL 4590-1-131 between marker CN889255SNP and Ma, and GMAL 4592-4-33 and GMAL 4595-6-121 between Ma and marker 12514.266 (Figs. 1a, 2) among the 52 informative recombinants identified. The Ma locus between markers CN889255SNP and 12514.266 corresponds to a homologous genomic segment of 65 kb in Golden Delicious, enabling us to reduce the number of candidate genes of Ma from 44 identified previously (Xu et al. 2011) to 19 in the present study.

Since the draft sequence of the apple genome does not provide clear haplotype specific information (Velasco et al. 2010), we identified two BAC clones of different haploid origin from apple rootstock G.41, BAC3 and BAC21, which completely cover the Ma locus. Sequencing the two BAC clones revealed that the Ma locus spanned 71 kb in BAC3 and 82 kb in BAC21. A more detailed analysis showed that out of 19 predicted genes in Golden Delicious, three were not present in the two BACs and two were beyond the Ma interval. In the remaining 14 predicted genes, two were duplicated, leading to 12 predicted genes for Ma in both BACs, including Ma1 and Ma2 (Fig. 1g–i; Table S3). Although the draft sequence of the apple genome is of high quality (Velasco et al. 2010), the local general structure of the Ma locus revealed by the two sequenced BACs from G.41 may be more representative. Given the limited number of genes in the Ma locus and the putative functions of ALMT genes in maintaining the malate homeostasis in plant cells, e.g. AtALMT9 (Kovermann et al. 2007) and AtALMT6 (Meyer et al. 2011), Ma1 and Ma2 are considered to be strong candidate genes of Ma.

Putative function of Ma1 and Ma2 as vacuolar malate channels/transporters in apple fruit

The first member of the ALMT1 family unique to plants is TaALMT1 that confers wheat tolerance to soil aluminum toxicity (Sasaki et al. 2004). TaALMT1 protein facilitates malate efflux from root apices and is localized on the plasma membrane (Yamaguchi et al. 2005). The counterpart of TaALMT1 that shows similar aluminum tolerance function includes AtALMT1 in Arabidopsis (Hoekenga et al. 2006), ScALMT1-M39.1 and ScALMT1-1135.1 (a hybrid gene) in rye (Collins et al. 2008), and BnALMT1 and BnALMT2 in rape (Ligaba et al. 2006). The Arabidopsis genome encodes 14 ALMT1 genes, which are distributed in four of the five clades in the ALMT1 family (Barbier-Brygoo et al. 2011). Phylogenetic analysis of the deduced protein sequences of Ma1 and Ma2 together with the 14 Arabidopsis ALMT1 proteins showed that the two apple proteins belong to clade 2 that includes five members AtALMT3–6, 9 (Fig. 7).

Fig. 7

Phylogenetic analysis of Ma1 and Ma2 proteins. The 14 members AtALMT1-14 of the AtALMT1 family were retrieved from TAIR 10 (http://www.arabidopsis.org/). The protein sequences were aligned with ClustalW and the trees were constructed with the MEGA4 program (Tamura et al. 2007) using the neighbor joining method. To test the phylogeny, “bootstrap samples of 1000” was set during the analysis. The tree is drawn to scale and the evolutionary distances are in the units of the number of amino acid substitutions per site. Naming system of the clades as described previously (Barbier-Brygoo et al. 2011) is adapted here

AtALMT9 is a vacuolar membrane protein functioning as a vacuolar malate channel for maintaining cell malate homeostasis (Kovermann et al. 2007), differing from AtALMT1 (Yamaguchi et al. 2005) and AtALMT12 (Meyer et al. 2010), which are plasma membrane proteins. AtALMT9 is expressed in all organs, but its expression in leaves is specifically in mesophyll cells. AtALMT6, another member in clade 2 that has been characterized recently, is expressed in guard cells of leaves as well as in flower organs and stems, but not in roots (Meyer et al. 2011). The AtALMT6 protein is also targeted to the vacuolar membrane, and it functions as a malate influx or efflux channel that is highly regulated by vacuolar pH and cytosolic malate (Meyer et al. 2011). It has been shown that low malate content in low acid fruit is the result of a restricted ability to accumulate malate in apple parenchyma cells (Beruter 2004). As members of clade 2, Ma1 and Ma2, especially Ma1, are likely vacuolar malate channels/transporters with primary function in maintaining malate homeostasis by regulating the malate levels in vacuole and cytosol in the parenchyma cells of apple fruit, thereby controlling fruit acidity levels.

Haplotypes of Ma and allelic association of the Ma1 and Ma2 alleles with Ma and ma

Sequencing of the two BAC clones from apple rootstock G.41 provided the first view of the Ma locus at the DNA sequence level with distinction between haplotypes Ma and ma. The difference between the Ma (BAC3) and ma (BAC21) haplotypes is significant in both size (82 vs. 71 kb) and the coding sequences of predicted genes (Fig. 1g–h). In the Ma1 and Ma2 sequences, the alleles (Ma1-G41 and ma1-G41, and Ma2-G41 and ma2-G41) are clearly distinguishable. This made it possible to infer their allelic counterparts (Ma1-GD and ma1-GD, and Ma2-GD and ma2-GD) in genes MDP0000252114 and MDP0000244249 of Golden Delicious, respectively. Comparison of the allelic sequences of Ma1 and Ma2 revealed that there are no variations in the deduced amino acid sequences in alleles (ma1-G41 and ma1-GD, and ma2-G41 and ma2-GD) associated with ma for low acidity, whereas the variations are considerable for alleles (Ma1-G41 and Ma1-GD, and Ma2-G41 and Ma2-GD) associated with Ma for high acidity. A similar trend exists in the entire Ma region when the sequences at the Ma locus between Golden Delicious and G.41 were compared as BAC21 is much closer to Golden Delicious than BAC3. One possible explanation is that the natural or human selection of fruit acidity has mostly acted upon the high acid allele Ma rather than the low acidity allele ma due to its recessive nature, leading to a greater diversity in high acidity allele Ma. Whether or not the high diversity among the Ma alleles plays a role in large fruit acidity variations in different apple cultivars would be of great interest for future investigation.

One of the most important findings of this work is the discovery of the mutation at the 1455th base of Ma1, which turns the tryptophan (W) codon TGG₁₄₅₅ in Ma1-G41 into a stop codon TGA₁₄₅₅ in ma1-G41, leading to a premature termination and truncation of ma1-G41 by 84 deduced amino acids at the C terminus. The presence of the mutation in Golden Delicious is confirmed with the ambiguous base R₁₄₅₅, which stands for G₁₄₅₅/A₁₄₅₅ in Ma1 (Fig. S2). In view of the dramatic implication of this mutation and the critical role of the C-terminus in regulating the function and activity of TaALMT1 in wheat (Furuichi et al. 2010; Ligaba et al. 2009), marker CAPS₁₄₅₅ was developed to target SNP₁₄₅₅. Analysis using marker CAPS₁₄₅₅ showed that it segregates in a codominant fashion and accurately predicts genotypes MaMa, Mama and mama in population GMAL 4595 and the informative recombinants (Figs. 1, 2, 5a). Moreover, the marker shows a perfect association with pH and a highly tight association with TA in 29 apple accessions studied (Figs. 5b, 6). Overall, these data strongly suggested that SNP₁₄₅₅ is critical in determining the function of the Ma1 alleles.

It should be pointed out that the plant materials used in this study are restricted in M. sieversii, M. domestica and some of its hybrids. Since there are at least 23 species in Malus (Robinson et al. 2001), understanding the role of the Ma locus and SNP₁₄₅₅ in the remaining species would be an interesting extension of this work.

Expression of Ma1 and Ma2

Compared with Ma2 expression in mature fruit, Ma1 expression is much higher (Fig. 3). This trend appeared to be consistent with the number of ESTs identified for the two genes in the Malus EST database of 336,017 accessions in GenBank. There are 20 EST accessions for Ma1 (Table S4) and one for Ma2, i.e. CN929391 derived from pre-opened floral bud of Royal Gala. The tissue source for the 20 Ma1 ESTs includes fruit (9 accessions), flower (3), leaf (3), root (3), stem xylem (1) and bud (1) from nine apple varieties, such as Royal Gala (6), GoldRush (5), Granny Smith (2), M.9 (1, rootstock) and others. Therefore, in addition to higher expression levels, Ma1 is also evidenced to be expressed in a wider range of organs than Ma2, suggesting a broader role of Ma1 in apple.

Significant correlations between gene expression and fruit acidity were observed for gene Ma1 but not for Ma2 (Fig. 3). This suggests that Ma1 is the major factor in determining fruit acidity levels. Since alleles Ma1-1455G and ma1-1455A are associated with Ma and ma, respectively, the strong positive correlation between Ma1 expression and fruit acidity would suggest that transcripts of Ma1-1455G be more readily detected than those of ma1-1455A. Examining the presence of SNP₁₄₅₅ in the 20 ESTs of Ma1 supported this reasoning. SNP_1455G appeared in all seven ESTs (CO723101, CX024250, CN494439, GO547092, GO509271, GO562003 and HM641023) that span over base 1455 in Ma1 while SNP_1455A was not detected (Table S4). It appears, therefore, that both SNP₁₄₅₅ and expression levels of Ma1 are important in apple fruit acidity. To elucidate the role of Ma2, more dedicated studies are needed.

Gene MDP0000141005 encodes a putative serine/threonine protein phosphatase 2A (PP2A) regulatory subunit A and its expression was initially found to be correlated with fruit acidity. MDP0000141005 was excluded in allelic variation analysis since the correlation became non-significant when 18 apple accessions were analyzed. We examined the coding sequences of MDP0000141005 in BAC3 and BAC21 of G.41, which did not confer variations in the amino acid sequences. Although PP2A is involved in many plant processes (Ahn et al. 2011; Leivar et al. 2011; Skottke et al. 2011), its subunit genes, including regulatory subunit A, have been used as reference genes for qRT-PCR analysis in plants (Czechowski et al. 2005; Navascues et al. 2012; Obrero et al. 2011). The constitutive expression of the PP2A regulatory subunit A gene and the inconsistent correlation between the MDP0000141005 expression and fruit acidity make it unlikely the gene responsible for fruit acidity variation.

In conclusion, we discovered two ALMT-like genes, Ma1 and Ma2, at the Ma locus of 65–82 kb containing 12–19 predicted genes that controls fruit acidity levels in apple. Expressions of Ma1 and Ma2 contrast sharply in the 18 apple germplasm accessions studied. Ma1 was expressed at much higher levels than Ma2 in mature fruit, especially in those of high acidity. Moreover, the Ma1 expression is significantly correlated with fruit acidity, whereas the Ma2 expression remains at low levels regardless of fruit acidity variations. These data suggest that Ma1 is the major determinant at the Ma locus controlling fruit acidity. Sequencing of clones BAC3 and BAC21 that cover the two distinct haplotypes at the Ma locus allowed us to determine specific alleles of both Ma1 and Ma2 for high or low acid phenotype. A single nucleotide mutation at base 1455 in the open reading frame of Ma1 led to a premature stop codon TGA₁₄₅₅, which truncates the carboxyl terminus of Ma1 by 84 amino acids. A survey of 29 apple germplasm accessions using marker CAPS₁₄₅₅ targeting SNP₁₄₅₅ found that the CAPS_1455A allele is associated completely with high pH and tightly with low TA, suggesting that the natural mutation-led truncation is most likely responsible for the abolished function of Ma for low pH or high TA in apple.