Assessing the allelotypic effect of two aminocyclopropane carboxylic acid synthase-encoding genes MdACS1 and MdACS3a on fruit ethylene production and softening in Malus

Phytohormone ethylene largely determines apple fruit shelf life and storability. Previous studies demonstrated that MdACS1 and MdACS3a, which encode 1-aminocyclopropane-1-carboxylic acid synthases (ACS), are crucial in apple fruit ethylene production. MdACS1 is well-known to be intimately involved in the climacteric ethylene burst in fruit ripening, while MdACS3a has been regarded a main regulator for ethylene production transition from system 1 (during fruit development) to system 2 (during fruit ripening). However, MdACS3a was also shown to have limited roles in initiating the ripening process lately. To better assess their roles, fruit ethylene production and softening were evaluated at five time points during a 20-day post-harvest period in 97 Malus accessions and in 34 progeny from 2 controlled crosses. Allelotyping was accomplished using an existing marker (ACS1) for MdACS1 and two markers (CAPS866 and CAPS870) developed here to specifically detect the two null alleles (ACS3a-G289V and Mdacs3a) of MdACS3a. In total, 952 Malus accessions were allelotyped with the three markers. The major findings included: The effect of MdACS1 was significant on fruit ethylene production and softening while that of MdACS3a was less detectable; allele MdACS1–2 was significantly associated with low ethylene and slow softening; under the same background of the MdACS1 allelotypes, null allele Mdacs3a (not ACS3a-G289V) could confer a significant delay of ethylene peak; alleles MdACS1–2 and Mdacs3a (excluding ACS3a-G289V) were highly enriched in M. domestica and M. hybrid when compared with those in M. sieversii. These findings are of practical implications in developing apples of low and delayed ethylene profiles by utilizing the beneficial alleles MdACS1-2 and Mdacs3a.


INTRODUCTION
To make fresh apple fruit available year-round for consumers, the controlled atmosphere (CA) storage technology has been adapted widely in the apple industry. The technology primarily employs low temperature, low O 2 and high CO 2 in combination with an ethylene production inhibitor 1-methylcyclopropene and others. Apple fruit can be stored for 410 months under optimal CA conditions. However, physiological disorders associated with CA storage, such as injuries induced by cold and CO 2 and flesh browning induced by 1-methylcyclopropene, can cause substantial loss for storage operators. [1][2][3] Such storage disorders have been reported for major apple varieties such as 'Empire' and 'McIntosh' 2,4 and for rising cultivars such as 'Honeycrisp'. 3 A strong need for new apples of long-shelf life and improved keeping quality with few or no storage disorders exists.
The gaseous phytohormone ethylene plays an important role in climacteric fruit ripening. The shelf life and storability of apple fruit are closely correlated with their ethylene production levels. Plant ethylene biosynthesis has been well-defined in Yang cycle that involves three enzymes: S-adenosylmethionine synthase, 1-aminocyclopropane-1-carboxylic acid (ACC) synthase (ACS) and ACC oxidase (ACO). 5 The enzymes ACS and ACO have been the subject of extensive studies to better understand plant ethylene production. Studies in many plant species including tomato and apple have shown that ACS and ACO are encoded by gene families of multiple members, that is, the ACS family and the ACO family, respectively.
There are two systems of ethylene production in plants: system 1 occurs during plant/fruit growth and development; and system 2 is defined exclusively for the floral senescence and fruit-ripening stages. 6 In tomato, system 1 ethylene biosynthesis involves LeACS6, 1A and LeACO1, 3, 4; whereas system 2 uses LeACS2, 4 and ACO1,4. 7 In apple, at least five ACS (MdACS1-5) and four ACO (MdACO1-4) genes have been reported 8,9 and these genes appear to be operating similarly in the two systems for ethylene production. MdACS1 is considered a system 2 gene; and its expression is highly correlated with the ethylene production burst in ripening apples. There are two alleles for the MdACS1 gene, MdACS1-1 and MdACS1-2, and the former is often associated with high ethylene production while the latter with lower ethylene production during fruit ripening. [10][11][12][13][14] This observation has led to a marker-assisted selection strategy emphasizing on selection for allelotype (see Discussion for usage of term 'allelotype') MdACS1-2/2 for long-shelf life apples. 15 Indeed, some evidence suggests that modern apple-breeding practice has unintentionally favored selection for the MdACS1-2 allele in commercial apple cultivars, 16 presumably for fruit of low ethylene and long-shelf life.
However, early-ripening cultivars showed faster fruit softening, regardless of their MdACS1 allelotypes. 10 This is consistent with the observation that the polygalacturonase gene (MdPG1) involved in softening of fruit flesh is expressed irregularly among apple cultivars of identical MdACS1 allelotypes. 12 Therefore, there are other factors also affecting fruit shelf life in addition to MdACS1. Interestingly, findings in a recent report have suggested that allele variations of another ACS gene (U73816), 17 designated MdACS3a (AB243060), are an essential factor regulating apple fruit ripening and shelf life. 18 There are two natural mutant alleles of the wildtype allele MdACS3a: One is the functional null allele MdACS3a-G289V, arising from a point mutation that leads to an amino-acid substitution from G 289 to V 289 at an active region for the MdACS3A enzyme activity, resulting in a functionally inactive enzyme. In melon, a similar point mutation in a conserved active region of an ACS gene led to andromonoecy, a common sexual system in angiosperms characterized by carrying both male and bisexual flowers. 19 This is an excellent example demonstrating that point mutations in conserved active regions of an ACS enzyme could confer a major phenotypic variation in plants. The other, a transcriptionally null allele Mdacs3a, is characterized by non-detectable mRNA. Moreover, combinations of Mdacs3a and MdACS3a-G289V alleles, regardless of whether they are homozygous or heterozygous, are highly associated with lower ethylene production and long-shelf life. In the six apple varieties/selections of the two null alleles studied, all showed low ethylene production and long-shelf life, irrespective to their MdACS1 allelotypes and early, mid or late physiological maturation dates. 18 Furthermore, the expression of MdACS3a is fruit tissue specific and detectable only during the transition from system 1 to 2 ethylene biosynthesis. 8,9,18 These observations suggest that MdACS3a acts as a main regulator for the transition, and is thereby crucial in regulating the fruit-ripening process. 18 In a more recent report, however, the allelotypes of MdACS3a were demonstrated to affect the ripening initiation of latematuring cultivars only, but not the early-or mid-maturing cultivars. 20 To better assess the roles of MdACS1 and MdACS3a, two approaches were taken in this study. The first approach was to estimate the allelotypic effect of the two genes by evaluating fruit ethylene production levels and softening rates in 97 diverse Malus accessions and 34 progeny from 2 controlled crosses. The second approach was to examine how variations in their allelotypic effect were associated with the frequency changes of the MdACS1 and MdACS3a alleles in M. domestica and M. hybrid as compared with those in M. sieversii, the major progenitor species of domestic apples, in 952 Malus accessions covering 53 Malus species. Allelotyping (see Discussion for usage of term 'allelotyping') of MdACS1 and MdACS3a was conducted using an existing marker for MdACS1 and two CAPS (cleaved amplified polymorphic sequence) markers specifically developed here to detect alleles ACS3a-G289V and Mdacs3a.

Plant materials
Two sets of Malus accessions were used in this study, which have been planted and maintained in the Malus germplasm repository of the US Department of Agriculture (USDA) in Geneva, New York. The first set included a total of 952 accessions, covering 53 Malus species (Supplementary Table S1). Among them, Malus domestica of 508 accessions, M. hybrid (the breeding selections derived from crosses between M. domestica and other Malus species) of 146 and M. sieversii (the major progenitor species of M. domestica) of 78 were most commonly represented (Supplementary Table S1). The second set comprised 34 halfsib progeny selected from 2 interspecific crosses GMAL4592 ('Royal Gala' × PI613978) and GMAL4593 ('Royal Gala' × PI613981). 'Royal Gala', a widely grown apple cultivar (M. domestica), has an allelotype MdACS1-2/2 and MdACS3a/MdACS3a-G289V for genes MdACS1 and MdACS3a, respectively. PI613978 and PI613981 are among the elite selections of M. sieversii collected from Kazakhstan, 21 and they have the same allelotypes for the two ACS genes, that is, MdACS1-1/1 and MdACS3a/MdACS3a-G289V. Population GMAL4592 was used in one of our previous studies. 22 Both GMAL4592 and GMAL4593 were planted on their own seedling roots in 2004.

Measurements of fruit ethylene production and firmness
Fruit ethylene production and flesh firmness were measured for 97 of 952 Malus accessions in the first set and the 34 half-sib progeny in the second set as described previously. 23 Briefly, for each accession, at least 25 fruits were harvested at a target maturity level as determined by the starch index of 4-6 according to the Cornell Starch Chart. 24 The 25 fruits were evenly divided into 5 groups and were stored for 0, 5, 10, 15 and 20 days at room temperature (20-25°C), respectively. Each fruit was weighed then enclosed in a gas-tight container (1.2 l) and kept for 1 h at room temperature. One milliliter of gas was sampled from the headspace in the container using a BD syringe (No. 309602, BD, Franklin Lakes, NJ, USA). The gas sample's ethylene concentration was measured with a gas chromatograph HP 5890 series II (Hewlett-Packard, Palo Alto, CA, USA) equipped with a flame ionization detector. Before the gas samples were assayed, the gas chromatograph was calibrated with standard ethylene gas (NO. 34489, Restek, Bellefonte, PA, USA) at a series of concentrations-0.01, 0.1, 0.5, 1, 5, 10 and 100 p.p.m.-to obtain the linear relation between ethylene peak area and concentration. The fruit ethylene production was calculated with the following formula: Where E stands for fruit ethylene production rate in nanoliter per gram of fresh weight per hour (nL g −1 h −1 ), [C 2 H 4 ] for ethylene concentration in p.p. m., V 1 for the volume of container in mL, V 2 for the volume of fruit in mL equivalent to fresh weight (W) in grams and T stands for the time in hours kept in the container. Fruit flesh firmness was measured using a penetrometer (Fruit Tester, Wagner FTK100, Greenwich, CT, USA) with a probe of 11 mm in diameter. The probe tip was pressed vertically into the fruit pulp (after skin-disc removal) to a depth of 10 mm. For larger fruits, four skin discs were removed from opposite sides of each fruit along the equator, and for smaller fruits, three skin discs were removed at roughly equal distance. The firmness readings were expressed in kg cm −2 , and firmness loss was measured by the percentage (%) of firmness reduced at days 5 to 20 as compared with the firmness at day 0. After the firmness was measured, fruits were sliced in half along the equator, dipped into a iodine-potassium iodide (I 2 -KI) solution, and then allowed the staining reaction for 41 min before reading Cornell Starch Index. 24 Allelotyping of MdACS1 and MdACS3a Allelotyping of MdACS1 was conducted with marker ACS1 using primers ACS1-5F/R (Supplementary Table S2) as reported previously. 10,15 However, allelotyping of MdACS3a was accomplished with two CAPS markers developed in this study using an online tool for identifying appropriate restriction enzymes 25 (see Results). These two markers, named CAPS 866 and CAPS 870 , were capable of detecting the functional null allele MdACS3a-G289V and the transcriptional null allele Mdacs3a, respectively. In practice, the same primers ACS3a-289F/R (Supplementary Table S2) were used for PCR to amplify the targeted DNA fragment for both CAPS 866 and CAPS 870 . PCRs were performed with 35 cycles of 94°C for 30 s, 58°C for 30 s, 72°C for 1 min, with an initial 94°C for 5 min and a final extension of 72°C for 10 min. Each PCR reaction mix was set in 10 μL containing 20 ng genomic DNA, 0.2 mM each dNTP, 0.5 μM of each primer, 2.5 mM MgCl 2 , 2 μL 5 × PCR Colorless GoTaq Reaction Buffer and 1 U of GoTaq DNA polymerase (Promega, Madison, WI, USA). To detect alleles MdACS3a-G289V and Mdacs3a, the PCR products were restricted with enzymes BstNI and Taq α I (New England Biolabs, Ipswich, MA, USA) following the manufacturer's instruction, respectively. The restricted PCR products were assayed by electrophoresis on 1.5% agarose gel and then stained with ethidium bromide for visualization and documentation as described previously. 22 Sanger DNA sequencing The PCR products amplified by primers ACS3a-289F/R (Supplementary  Table S2) were directly sequenced using a DNA Sequencer ABI3730XL (Applied Biosystems, Foster City, CA, USA) at the Cornell University Biotechnology Resource Center (Ithaca, NY, USA). The reverse PCR primer ACS3a-289R was used for DNA sequencing. DNA sequence analyses were performed using software Sequencher 5.2 (Gene Codes Corporation, Ann Arbor, MI, USA).

Allelotypic effect of MdACS1 and MdACS3a
L Dougherty et al.

Statistical analysis
Pearson's correlation analysis and one-way analysis of variance (ANOVA) of ethylene production and fruit firmness were conducted with software JMP Pro 10.0 (SAS institute, Cary, NC, USA). Significance levels in comparison of the means were determined by Po0.05 (Student's t-test).  Table S3).

Development of allelic specific markers for MdACS3a
The null allele MdACS3a-G289V is caused by a mutation from G 866 to T 866 at the 866th base in the coding sequence of MdACS3a. 18 Based on the web-based tool for single nucleotide polymorphism (SNP) analysis, 25 the mutation abolishes the recognition site CC 866 WGG of restriction enzyme BstNI ( Figure 2). To develop a CAPS marker, two primers (ACS3a-289F/R, Supplementary Table S2) were designed to amplify a DNA fragment (480 bp) covering the SNP (G 866 /T 866) specifically from MdACS3a although the three MdACS3 member genes MdACS3a (AB243060), MdACS3b (AB243061) and MdACS3c (AB243062) are of high identity in their DNA sequences. 18 The specificity of the primer pair to MdACS3a was confirmed by sequencing of the PCR products from 92 of the 97 Malus accessions (Figure 2, Supplementary Table S3). Digestion of the PCR products with BstNI yielded restriction bands as expected (Figure 3b), indicating the successful development of a CAPS marker detecting SNP G 866 /T 866 , designated CAPS 866 . Therefore, allele CAPS 866 G represents the wild-type allele MdACS3a while CAPS 866 T stands for the functional null allele MdACS3a-G289V. Development of a marker detecting the transcriptional null allele Mdacs3a was initially thought to be challenging as the null allele was reported not to show sequence variations from the wild-type allele. 18 However, sequencing analysis of the PCR products amplified by primers ACS3a-289F/R in the 92 accessions (Supplementary Table S3) not only identified the expected SNP G 866 /T 866 , but also a new SNP C 870 /T 870 ( Figure 2). Importantly, this new SNP can discriminate the two alleles of MdACS3a in 'Fuji' (Figure 2), which was known of allelotype MdACS3a/Mdacs3a. 18 Evidence from this and other studies (see Discussion) indicated that base T 870 was associated with the Mdacs3a allele. Using a similar approach, another CAPS marker, named CAPS 870 , was developed to detect SNP C 870 /T 870 using restriction enzyme Taq α I along with the same primers ACS3a-289F/R (Figure 3c). Therefore, allele CAPS 870 C corresponds to the wild-type allele MdACS3a while CAPS 870 T corresponds to the transcriptional null allele Mdacs3a.
Effect of the allelotypes of MdACS1 and MdACS3a on ethylene production and firmness loss To evaluate the effect of the allelotypes of MdACS1 and MdACS3a, the 97 Malus accessions were assayed with markers ACS1, CAPS 866 and CAPS 870 that can detect different alleles of MdACS1 and MdACS3a (Figures 3a-c). As a result, marker ACS1 identified 53, 36 and 8 accessions of Table S3 Table S3).
A series of one-way ANOVA of the fruit ethylene production and fruit firmness loss over the 20-day period within each of the three allelotype groups (Figure 4) indicated that the most differences were observed among the MdACS1 allelotypes. Allelotype MdACS1-1/1 showed significantly higher ethylene production (days 0-20) and firmness loss (days 5-20) than MdACS1-1/2 and MdACS1-2/2 allelotypes, but MdACS1-1/2 and MdACS1-2/2 did not differ in terms of ethylene production or firmness retention (Figures 4a and d). In contrast, there were no difference among the CAPS 866 allelotypes in fruit ethylene production and firmness loss (Figures 4b and e). Among the CAPS 870 allelotypes, significant difference was not detected for ethylene production, but there were differences in fruit firmness loss between allelotypes CAPS 870 C/C and CAPS 870 C/T at day 5 and between CAPS 870 C/C and CAPS 870 T/T at day 10 (Figures 4c and f). This indicated that such differences in fruit firmness loss at day 5 and 10 in the CAPS 870 allelotypes might be caused by other factors rather than their ethylene production levels.
To seek such factors, peak ethylene day, which measures ethylene peak timing, was examined ( Figure 5) as this trait was negatively correlated with fruit firmness loss at day 10 (r = − 0.258, P = 0.011) although the correlation was insignificant at day 5 (r = − 0.112, P = 0.275) ( Table 1). Encouragingly, the three CAPS 870 allelotypes showed significant difference from each other, with CAPS 870 C/T having peaked the earliest, CAPS 870 C/C intermediate and CAPS 870 T/T the latest (Figure 5a). These data appeared to suggest that the earlier peak ethylene day of CAPS 870 C/C might have contributed to its greater fruit firmness loss of CAPS 870 C/C as compared with that of CAPS 870 T/T at day 10 ( Figure 4f). However, the lowest fruit firmness loss of CAPS 870 C/T at day 5 remained to be explained. Peak ethylene day was also analyzed in the other two groups of allelotypes. In the allelotypes of MdACS1, MdACS1-1/ 1 had an earlier peak ethylene than MdACS1-2/2, but showed no difference from MdACS1-1/2 (Figure 5a). In the three allelotypes of CAPS 866 , no significant difference was observed (Figure 5a).
It was clear that the effect of MdACS1 on ethylene production and fruit firmness loss was much stronger than that of MdACS3a (Figure 4). To see if the random presence of the MdACS1 alleles might have obscured the detection of the effect of MdACS3a allelotypes (Figures 4b, c, e and f), another series of ANOVA was conducted for the MdACS3a allelotypes of five or more accessions (Figure 6) under the same background of MdACS1 allelotypes MdACS1-1/1 and MdACS1-1/2, which occurred in 53 and 36 of the 97 accessions (Supplementary Table S3), respectively. The third allelotype MdACS1-2/2 was not included in the analysis ( Figure 6) due to limited number of 8 accessions.
For CAPS 866 , the ANOVA analyses were conducted for two allelotypes CAPS 866 G/G and CAPS 866 G/T under MdACS1-1/1, as well as under MdACS1-1/2 (Figures 5b and 6a,c). This allowed us to identify that allelotype CAPS 866 G/T produced significantly higher levels of ethylene than CAPS 866 G/G at day 10 under MdACS1-1/1 (Figure 6a). For CAPS 870 , three allelotypes CAPS 870 C/C, CAPS 870 C/T and CAPS 870 T/T under MdACS1-1/1 and two allelotypes CAPS 870 C/C and CAPS 870 C/T under MdACS1-1/2 were analyzed (Figures 5c and 6b,d). The results showed that allelotype CAPS 870 T/T had significant later peak ethylene day than CAPS 870 C/C and CAPS 870 C/T under MdACS1-1/1, and CAPS 870 C/C had significant later peak ethylene than CAPS 870 C/T under MdACS1-1/2 (Figure 5c). There were no significant differences detected between the other allelotypes of CAPS 866 and CAPS 870 at a given time point (Figures 5b and 6a-d). These observations suggested that the direct effect of MdACS3a on ethylene production and firmness loss was limited, but its effect on peak ethylene day was clearly detectable through allele Mdacs3a (CAPS 870 T/T).
The analyses also provided information regarding the effect of MdACS1 under the same background of CAPS 866 (Figures 5b and  6a,c) or CAPS 870 (Figures 5c and 6b,d) allelotypes. As expected, allelotype MdACS1-1/1 had higher ethylene production (Figures 6a and c) and more firmness loss (Figures 6b and d) than MdACS1-2/2, but had similar peak ethylene day as MdACS1-1/2 (Figures 5b and c) except under the CAPS 870 C/C background (Figure 5c). These results suggested that the effect of MdACS1 on peak ethylene day was insignificant under the same background of MdACS3a, which was in disagreement with the observation that the effect of MdACS1 on peak ethylene day was significant when the background of MdACS3a was not considered (Figure 5a).

DISCUSSION
The effect of MdACS1 and MdACS3a and beneficial alleles The allelic effect of MdACS1 on fruit ethylene production and softening was significant and detectable at nearly all time points tested during the 20-day post-harvest period in the 97 Malus accessions. This was consistent with the critical role of MdACS1 reported in many other studies. [10][11][12][13][14][15][16][26][27][28][29] Since the allele frequency of MdACS1-2 was 24.5% in M. domestica, 18.8% in M. hybrid and only 0.6% in M. sieversii (Figure 8), which is the major progenitor species of domestic apples, artificial selection has clearly favored MdACS1-2 over MdACS1-1. In fact, such allele preference of MdACS1-2 over MdACS1-1 was even reported within M. domestica when the frequencies of the two alleles in apple cultivars were plotted against their time of introduction. 16 These observations are in accordance with the finding that allele MdACS1-2 is a beneficial allele associated with low ethylene and slow softening (Figures 4 and 6).
MdACS3a was regarded a main regulator for ethylene production transition from system 1 to 2. 18 The gene was also similarly shown to be an accelerator 30    progeny from the 2 controlled crosses segregating for allelotype MdACS3a-G289V/G289V (CAPS 866 T/T) under the same background of MdACS1 allelotype. Furthermore, the allele frequency of MdACS3a-G289V (CAPS 866 T) was 13.5% in M. sieversii, 13.2% in M. domestica and 5.8% in M. hybrid, providing no evidence that MdACS3a-G289V (CAPS 866 T) has been enriched in response to selection. These results were surprising as MdACS3a-G289V was shown to be a functional null allele of MdACS3a. 18 In a previous study, the two null alleles MdACS3a-G289V (CAPS 866 T) and Mdacs3a (CAPS 870 T) were concluded to affect the ripening initiation only in late-season apple cultivars, but not in early-or mid-season ones. 20 Such discrepancy in different studies regarding the roles of the two null alleles of MdACS3a, particularly MdACS3a-G289V, calls for further investigations into the role of MdACS3a-G289V. Nevertheless, alleles MdACS1-2 and Mdacs3a (CAPS 870 T) are clearly demonstrated to be beneficial for breeding apples of low or delayed ethylene profiles in this study, a first effort that simultaneously assessed the roles of MdACS1 and MdACS3a in fruit ethylene production and softening in highly diverse Malus materials.

Markers ACS1, CAPS 866 and CAPS 870
The assessment of the roles of MdACS1 and MdACS3a in apple fruit ethylene production and softening largely relied on the previously developed marker ACS1 10,11 and the two markers CAPS 866 and CAPS 870 developed in this study. Since CAPS 866 directly detects the mutation SNP G 866 /T 866 , CAPS 866 is an unequivocal marker for identifying the functionally null allele MdACS3a-G289V. 18 Marker CAPS 870 detects SNP C 870 /T 870 that does not correspond to a change in the encoding amino acid, that is, CAPS 870 detects a silent mutation in MdACS3a. Regardless of the nature of SNP C 870 / T 870 , T 870 is a genetic signature for allele Mdacs3a as the mutation was identified in 'Fuji', the very source from which the transcriptional null allele Mdacs3a was originally defined. 18 Based on the genomic DNA sequences from 'Fuji', alleles MdACS3a (JF833309) and Mdacs3a (JF833309) differ by 14 nucleotides, and of these, only 4 were within the coding sequence. 20 Sequencing of the 92 Malus accessions in this study indicated that SNP C 870 /T 870 is authentic and varying only between 2 nucleotides C 870 and T 870 (Figure 2, Supplementary Table S3). These data strongly support that CAPS 870 is a reliable marker for detecting allele Mdacs3a.
Since both CAPS 866 and CAPS 870 detect the characterized SNPs in the coding sequence of MdACS3a and can be simply performed by electrophoresis on agarose gels, the two markers are readily applicable for marker-assisted selection in apple breeding.
Since SNP C 870 /T 870 is located only four bases downstream of SNP G 866 /T 866 , markers CAPS 866 and CAPS 870 were once considered to be used as a single marker in this study. However, such usage would lead to an ambiguous scenario for allelotype G 866 T 866 /C 870 T 870 as it could be formed by a combination either between gametes G 866 T 870 and T 866 C 870 or between gametes G 866 C 870 and T 866 T 870 . To avoid such possible uncertainty, the two markers were used independently.
Previously, an SSR marker targeting at the promoter region of MdACS3a was developed and used to allelotype MdACS3a in 103 apple varieties. 20 It was shown that three alleles (331, 353, and 359 bp) of the SSR marker corresponded to the wild-type allele MdACS3a (that is, MdACS3a-1 in ref. 20), two alleles (333 and 335 bp) to Mdacs3a (that is, MdACS3a-2) and one allele (361 bp) to MdACS3a-G289V (that is, MdACS3a-1V). This makes the corresponding relationship between the SSR marker alleles and the MdACS3a alleles somewhat indirect and inconvenient. Since the size of the SSR marker alleles frequently differ by 2 bp, an automatic DNA sequencer-based detection system is necessary, thereby requiring more sophisticated handling and analysis, compared with the agarose gel-based markers CAPS 866 and CAPS 870 . However, identical allelotypes were observed for all 19 apple cultivars used by co-insistence in both studies (Supplementary Table S4), suggesting that the SSR marker and the 2 CAPS markers are useful for allelotyping of MdACS3a. As expected, identical allelotypes for MdACS1 were also obtained for the 19 common apple cultivars between these 2 studies (Supplementary Table S4).
It should be mentioned that two degenerated CAPS (dCAPS) markers were developed to confirm alleles Mdacs3a and MdACS3a-G289V in cDNA, but the two dCAPS markers were not used for allelotyping the MdACS3a alleles. 20 Therefore, the applicability of the dCAPS markers is unknown in diverse apples.
Utility of the data Of the 952 Malus accessions, 97 were evaluated for their fruit ethylene production and softening at 5 time points over a 20-day post-harvest period (Supplementary Table S3). Although most accessions seemed to have predictable ethylene-regulated postharvest behaviors, 'Virginia Gold' (PI588778, M. domestica) was unusual as it had minimal firmness loss (comparable to 'Fuji') during the 20-day storage while producing high levels of ethylene (comparable to 'Golden Delicious'). This suggested that the slow softening (long-shelf life) character of 'Virginia Gold' is likely less dependent on ethylene production. More importantly, 'Virginia Gold' has also been shown with an excellent storability. 32 To understand the lack of ethylene-related softening in 'Virginia Gold', several preliminary experiments have been initiated by the authors. In melon, it was reported that flesh softening involved both ethylene-dependent and -independent components. 33 In tomato, the ethylene-independent aspects of fruit ripening were evidenced to be regulated by the FRUITFULL homologs. 34 It is possible that investigating fruit softening independent of or less dependent on ethylene production would lead to new knowledge for better understanding of the apple fruit-ripening process, promising an interesting research area in apple post-harvest biology.
In addition, the data set of allelotypes for genes MdACS1 and MdACS3a generated in the 952 Malus accessions would be useful for other future studies involving MdACS1 and MdACS3a, which are the only 2 apple ACS genes known to be expressed specifically in fruit and associated with apple fruit ethylene production and firmness. 8,9,13 The data set, together with three markers ACS1, CAPS 866 and CAPS 870 , would be also useful for planning new crosses for developing improved apples with low ethylene and reduced loss of firmness.
Usage of terms allelotype and allelotyping Term allelotype is defined as 'the frequency of alleles in a breeding population.' according to 'A Dictionary of Genetics'. 35 In this study, allelotype is referred to the allele composition at a specific gene locus, that is, MdACS1 or MdACS3a, in individual accessions, highly similar to term 'genotype' for a given DNA marker. Such usage of allelotype represents a drift from or an expansion for the original definition of allelotype defined in the dictionary. However, the usage offers convenience for describing allele composition at a specific gene locus. Indeed, such usage has been adapted already in literature. 14,18,20 The definition for term allelotyping in ''Encyclopedia of Genetics, Genomics, Proteomics, and Informatics' 36 reads 'Allelotyping is the determination of the spectrum and frequency of allelic variations in a population.' The usage of allelotyping in this study is largely covered by the definition, but an extension to include activities for determining allelotype (allele composition at a specific gene locus) is also practiced.

Conclusions
A substantial effort to simultaneously assess the roles of MdACS1 and MdACS3a in fruit ethylene production and softening in diverse Allelotypic effect of MdACS1 and MdACS3a L Dougherty et al.