Pereira et al. Horticulture Research (2020)7:187 https://doi.org/10.1038/s41438-020-00411-z Horticulture Research www.nature.com/hortres ARTICLE Open Access Genetic dissection of climacteric fruit ripening in a melon population segregating for ripening behavior Lara Pereira1, Miguel Santo Domingo 1, Valentino Ruggieri1,2, Jason Argyris1,2, Michael A. Phillips3, Guangwei Zhao 4 , Qun Lian5, Yongyang Xu4, Yuhua He4, Sanwen Huang5, Marta Pujol 1,2 and Jordi Garcia-Mas 1,2 Abstract Melon is as an alternative model to understand fruit ripening due to the coexistence of climacteric and non-climacteric varieties within the same species, allowing the study of the processes that regulate this complex trait with genetic approaches. We phenotyped a population of recombinant inbred lines (RILs), obtained by crossing a climacteric (Védrantais, cantalupensis type) and a non-climcteric variety (Piel de Sapo T111, inodorus type), for traits related to climacteric maturation and ethylene production. Individuals in the RIL population exhibited various combinations of phenotypes that differed in the amount of ethylene produced, the early onset of ethylene production, and other phenotypes associated with ripening. We characterized a major QTL on chromosome 8, ETHQV8.1, which is sufficient to activate climacteric ripening, and other minor QTLs that may modulate the climacteric response. The ETHQV8.1 allele was validated by using two reciprocal introgression line populations generated by crossing Védrantais and Piel de Sapo and analyzing the ETHQV8.1 region in each of the genetic backgrounds. A Genome-wide association study 1234567890():,; 1234567890():,; 1234567890():,; 1234567890():,; (GWAS) using 211 accessions of the ssp. melo further identified two regions on chromosome 8 associated with the production of aromas, one of these regions overlapping with the 154.1 kb interval containing ETHQV8.1. The ETHQV8.1 region contains several candidate genes that may be related to fruit ripening. This work sheds light into the regulation mechanisms of a complex trait such as fruit ripening. Introduction physiological groups depending on the involvement of the Fleshy fruits are an important component of human diet hormone ethylene during ripening: climacteric fruits show and a major source of nutritional compounds. The ulti- a transient rise in respiration accompanied by the auto- mate biological function of fruit is for seed dispersal. Once catalytic synthesis of ethylene at the onset of ripening, fruit growth has been accomplished, fleshy fruits undergo leading to a peak of this hormone that, in contrast, is a series of physiological and metabolic changes to become absent in non-climacteric fruits2. It has been suggested edible and attractive to animals1. Among these changes; that there is a partial overlap between both types of sugar and organic acid accumulation, decay of fruit ripening3,4. firmness, color change and synthesis of volatiles com- In climacteric fruits, the presence of ethylene is neces- monly occur. Fleshy fruits are classified into two sary to trigger multiple responses, among them a change of texture, principally softening of mesocarp, mediated by cell wall degrading enzymes; accumulation of pigments Correspondence: Jordi Garcia-Mas (jordi.garcia@irta.cat) such as flavonoids and carotenoids in parallel with 1 Centre for Research in Agricultural Genomics (CRAG) CSIC-IRTA-UAB-UB, chlorophyll degradation, leading to changes in rind and Edifici CRAG, Campus UAB, 08193 Cerdanyola, Barcelona, Spain 2 flesh color, synthesis of sugars, organic acids and a diverse IRTA (Institut de Recerca i Tecnologia Agroalimentàries), Edifici CRAG, Campus UAB, 08193 Cerdanyola, Barcelona, Spain array of volatiles, allowing the fruit to reach an attractive Full list of author information is available at the end of the article © The Author(s) 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. Pereira et al. Horticulture Research (2020)7:187 Page 2 of 18 flavor and taste, and the formation of an abscission layer map QTLs related to fruit quality and morphology29. The in the pedicel5. Tomato is considered the biological aim of this work was to characterize ethylene production model to understand climacteric fruit ripening. Two and climacteric ripening behavior in this RIL population, enzymes of the biosynthetic pathway of ethylene, 1- which was also segregating for the climacteric behavior, aminocyclopropane-1-carboxylate (ACC) synthase and and to identify additional genetic factors that regulate this ACC oxidase, are regulated in a fine and complex way by process. transcription factors (TFs) as RIN6, NOR5, and CNR7. In addition, other TFs modulating the initiation or evolution Results of fruit ripening have been characterized, as FUL1/ Ethylene production and climacteric ripening in the Ved × FUL23,8, AP2a9, GLK210, TAGL111, and HB-112, among PS RIL population others. Mutations or overexpression of these TFs influ- The RIL population parents were Ved, a highly climac- ence ethylene production and some of its downstream teric cantalupensis line, and PS, a non-climacteric ino- effects, such as accumulation of carotenoids or changes in dorus type. Several ripening-related traits were measured texture. Recently, several studies have proved that DNA in both parents (see “Methods” section and Supplementary demethylation has a major influence in controlling the Table 1). Ved showed maximal climacteric behavior onset of fruit ripening13–15. In mutants for some of the starting around 34 days after pollination (DAP) in all TFs described above, several ripening responses are blocks: strong sweet aroma (assessed subjectively by compromised, giving rise to unripe fruits or fruits with a smelling the fruit), slight change of color from white to substantial delay in ripening when compared to wild type cream, and abscission layer formation with fruit abscission fruits. Most of these processes found in climacteric fruits in most cases (Supplementary Table 1). PS did not show are successfully achieved in non-climacteric fruits, prob- any symptom of climacteric ripening in any block. The ably through a combination of common mechanisms16 hybrid line (Hyb) presented an intermediate phenotype and different hormonal pathways17. between the parents, being always climacteric but showing Although many advances have been achieved in some of the phenotypic effects later than Ved; however, understanding climacteric ripening in tomato, knowledge the flesh firmness in Hyb fruits was lower than either of the fruit ripening process in other crops is less well- parents. These phenotypes correlate with the measured understood18–20. Melon (Cucumis melo L.) has been ethylene production levels: Ved produced a high amount proposed as an alternative model to understand fruit of ethylene (72.6 and 224.9 µL kg−1 h−1 in T3 and T4, ripening due to the coexistence of climacteric and non- respectively), PS did not produce ethylene, and Hyb climacteric varieties within the same species21. Melons showed an intermediate peak (33.3 and 37.7 µL kg−1 h−1 in show a wide range in the degree of climacteric response, T3 and T4, respectively) (Fig. 1A). Although this pattern from the maximally climacteric varieties belonging to the was consistent over two years, we observed that ethylene cantalupensis group as “Védrantais” (Ved) to the non- production (ETH) and, to a lesser extent, earliness of climacteric varieties from the inodorus group as “Piel de ethylene production (DAPE) and earliness of ethylene Sapo” (PS), which undergo virtually none22. The changes peak (DAPP), are probably environment-dependent. promoted by ethylene during melon fruit ripening are The statistics for climacteric traits in the RIL population slightly different to those described in tomato. Silencing of are presented in Table 1, where segregation was observed Aco1 in a Ved background allowed the identification of for all of them. Ethylene emission from fruit of a repre- ethylene-dependent ripening processes23. Flesh car- sentative subset of RILs was measured over time, reveal- otenoid and sugar accumulation, acidity and, partially, ing a transient peak of production (Fig. 1B). Data for both flesh softening are, unlike in tomato, ethylene- years were treated separately due to the effect of envir- independent in melon24. QTLs controlling climacteric onment on ethylene peak level and earliness; however, the ripening and ethylene production have been described in general pattern was very similar. Earliness, height, and segregating populations obtained from melon accessions shape of the ethylene peak segregated in the RIL popu- showing different ripening behaviors25–27. So far only one lation. We could observe an independent segregation, at of these QTLs, ETHQV6.3, has been cloned. An intro- least partially, between ETH and DAPE/DAPP. For gression of the non-climacteric accession PI 161375, example, RIL 159A presented an early ethylene peak but containing ETHQV6.3, into the non-climacteric PS the maximum amount of ethylene was low in comparison induced climacteric ripening26. The ETHQV6.3 under- with other lines; on the opposite side, RIL 59 showed a lying gene is CmNAC-NOR, the orthologue of the tomato considerably high ethylene peak around 43 DAP, with a NOR28, suggesting that common mechanisms are delay of 8–9 days compared to Ved (Fig. 1A, B). Trans- involved in climacteric ripening control in both species. gressive segregation was observed for ETH in some RILs, A recombinant inbred line (RIL) population from a which was consistent between years, as well as for DAPP cross between Ved and PS was developed and utilized to and DAPE. Pereira et al. Horticulture Research (2020)7:187 Page 3 of 18 Fig. 1 Ethylene production in planta. A Ethylene emission rates in Ved, PS and Hyb attached fruits in T3 and T4 blocks. B Ethylene emission in a representative subset of RILs in T3 and T4 blocks Most of the RILs produced aromatic fruits that formed (Fig. 2A, B); the color change in these cases was subtle abscission layers, but only half of them changed their and the clearest effect was observed in the region near rind color during ripening. The change of color in the the pedicel, where green color turned to yellow or slight fruit rind, mainly attributable to chlorophyll degrada- orange. The Hyb clearly turned from white rind to tion, manifested itself differently depending on the line orange-yellow when ripe (Fig. 2C). Other fruits were (Fig. 2). Some melons, with white rind when immature, green when immature and during the ripening process turned to a cream-slight orange color when ripe became bright yellow (Fig. 2D, F). In some cases, we Pereira et al. Horticulture Research (2020)7:187 Page 4 of 18 Table 1 Basic statistics for climacteric ripening traits and ethylene production in the RIL population Trait (units) Code Mean SD Median Range Production of aroma ARO 0.69 0.34 0.75 0–1 Earliness of production of aroma (DAP) EARO 43.05 6.68 41.67 30.6–60 Chlorophyll degradation CD 0.46 0.38 0.50 0–1 Earliness of chlorophyll degradation (DAP) ECD 41.39 6.55 40.00 27.5–58 Earliness of abscission layer formation (DAP) EALF 43.93 7.32 42.50 28–60 Abscission ABS 1.25 1.04 1.00 0–3 Harvest date (DAP) HAR 51.07 7.82 50.80 29.5–62 −2 Flesh firmness (kg cm ) FIR 3.05 1.62 2.80 0.74–7.14 Maximum ethylene production (µL eth kg−1 h−1) ETH 31.05 52.69 13.03 0–286.22 Earliness of ethylene production (DAP) DAPE 38.23 5.18 37.50 27.5–50 Earliness of ethylene peak (DAP) DAPP 42.21 5.89 42.00 32–58 Width of ethylene peak (days) WEP 3.71 1.82 3.50 0–9 Fig. 2 Immature (left) and ripe (right) fruit rind for Ved, Hyb, and several RILs. A Ved. B RIL59. C Hybrid (Hyb). D RIL177A. E RIL8. F RIL160 observed a striking pattern in mottled rinds, where the subtle and/or partial abscission layer formation (Sup- spots changed color when ripe (Fig. 2E, F). Concerning plementary Fig. 1B) and marked scar formation in the the abscission layer formation, the RIL population abscission zone (Supplementary Fig. 1C). Only 31% of included lines that did not abscise (Supplementary the climacteric lines presenting abscission layer fell from Fig. 1A), lines with complete fruit abscission (Supple- the plant, and even the controls Ved and Hyb did not mentary Fig. 1D) and intermediate phenotypes with always show abscission, which, according to our results, Pereira et al. Horticulture Research (2020)7:187 Page 5 of 18 Fig. 3 A major QTL ETHQV8.1 mapped in chromosome 8. A For maximum ethylene production (ETH), earliness of ethylene production (DAPE), width of ethylene peak (WEP) and earliness of ethylene peak (DAPP) in T3 and T4. B for the rest of climacteric traits seems to be highly dependent on environmental factors correlated with all the other phenotypic effects, positively as temperature and light. with abscission (ABS) and chlorophyll degradation (CD) Generally, ethylene production is measured in detached and negatively with flesh firmness (FIR) and with earliness fruits, precluding the precise phenotyping of other of ripening (DAPP, DAPE, earliness of production of ripening-related traits. The method used in our study aroma (EARO), earliness of chlorophyll degradation allows the fruit to follow the physiological process of (ECD), HAR), but slightly correlated to ETH. ripening without any alterations. Therefore the pheno- typing of downstream effects of ethylene can be observed QTL mapping non-invasively in planta. The most common, and gen- We performed two complementary mapping experi- erally earliest climacteric symptom was sweet aroma ments, the first one for each individual block indepen- production, which appeared on average four days after the dently and the second one using the mean of the starting of DAPE and in many RILs before DAPP. When a phenotyped blocks for each RIL. In the first experiment, change of rind color was induced, it appeared almost we mapped 74 QTLs for all traits on almost all melon simultaneously with ethylene production; the first or chromosomes (Supplementary Table 2A). The QTLs are second day of detectable ethylene production, chlorophyll represented in Fig. 3 and Supplementary Fig. 3, indicating degradation started to be appreciable and in approxi- for each trait the LOD scores and the genetic position mately three days the color change was complete. Com- across the chromosomes. monly, abscission layer formation was the last trait A major QTL, ETHQV8.1, was mapped in chromosome exhibited, around five days after ethylene detection. 8, with a maximum LOD around the position 86.49 cM. However, we could observe that depending on the RIL, ETHQV8.1 was detected for ETH, DAPE and DAPP in the earliness and the penetrance of the phenotype varied, both years with significant LOD scores (e.g. 6.33 for ETH with some lines that fell from the plant on the first or in T4 (Fig. 3A)). Other traits associated to the earliness of second day of ethylene production and others that climacteric ripening, such as EARO, earliness of abscis- remained with a subtle abscission layer for weeks. sion layer formation (EALF) and HAR, also showed Surprisingly, the correlation between ETH and the consistent significant QTLs in this region (Fig. 3B). Traits appearance and earliness of climacteric symptoms was not more related to the presence/absence of climacteric high (Supplementary Fig. 2). Earliness of production of ripening, as ARO, CD, and ABS, presented significant aroma (EARO), DAPE, DAPP, and harvest date (HAR) LOD scores only in some of the blocks. In all cases, the were highly and positively correlated, in agreement with Ved allele underlying ETHQV8.1 intensified the climac- our observations. Production of aroma (ARO) was highly teric effect, meaning higher and earlier ethylene Pereira et al. Horticulture Research (2020)7:187 Page 6 of 18 Table 2 QTLs for climacteric ripening and ethylene production traits detected using the mean of phenotypic values for all the blocks Chr QTL ID LODa R2 Additive Genetic Physical Marker 1 Genetic Marker 2 Genetic effectb position positionc position of position of (cM) (pb) marker 1 marker 2 2 DAPPQV2.1 2.4 14.8 2.41 58.28 8113174 chr02_3049784 45.69 chr02_14910035 69.54 FIRQV2.1 5.0 22.9 0.81 51.84 6015657 chr02_3362365 48.11 chr02_8113174 58.28 FIRQV2.2 5.0 22.7 0.80 63.53 11199109 chr02_8371192 59.28 chr02_12578264 65.4 ABSQV2.1 2.9 13.9 -0.40 63.79 11199109 chr02_8371192 59.28 chr02_15771889 72.24 3 CDQV3.1 2.4 11.7 0.14 103.55 28250248 chr03_28128535 101.84 chr03_29211378 107.6 5 ABSQV5.1 2.7 12.8 -0.37 4.23 606581 chr05_182820 0 chr05_910750 8.89 6 CDQV6.1 3.8 17.8 0.17 68.66 7435564 chr06_6598284 62.51 chr06_7589164 73.06 ABSQV6.1 2.7 13.2 0.38 145.7 33856412 chr06_34435507 137.68 chr06_33533854 148.56 7 HARQV7.1 2.8 13.5 -2.97 13.28 1046645 chr07_835787 9.54 chr07_1439538 15.16 EAROQV7.1 2.8 15.1 -2.61 13.28 1046645 chr07_835787 9.55 chr07_1517405 18.09 ABSQV7.1 2.7 13.2 0.38 13.28 1046645 chr07_868662 10.32 chr07_1657550 18.86 8 EAROQV8.1 10.4 45.5 -4.51 86.49 9634968 chr08_9446475 83.23 chr08_9941727 87.61 ABSQV8.1 4.4 20.2 0.48 86.49 9634968 chr08_9446475 83.23 chr08_9941727 87.61 EALFQV8.1 10.2 50.5 -5.25 86.49 9634968 chr08_9446475 83.23 chr08_9941727 87.61 ECDQV8.1 5.8 34.5 -3.80 86.49 9634968 chr08_9446475 83.23 chr08_9941727 87.61 HARQV8.1 17.2 58.9 -6.20 86.49 9634968 chr08_9446475 83.23 chr08_9941727 87.61 DAPPQV8.1 7.1 37.8 -3.59 86.49 9634968 chr08_9446475 83.23 chr08_9941727 87.61 DAPEQV8.1 7.5 39.2 -3.19 86.49 9634968 chr08_9446475 83.23 chr08_9941727 87.61 AROQV8.1 5.2 23.7 0.17 86.49 9634968 chr08_9446475 83.23 chr08_10225136 88.87 CDQV8.1 4.5 20.9 0.18 87.49 9941727 chr08_9446475 83.23 chr08_10225136 88.87 FIRQV8.1 6.1 27.2 -0.86 86.49 9634968 chr08_9446475 83.23 chr08_17287431 88.87 ETHQV8.1 6.6 30.8 29.70 86.49 9634968 chr08_9446475 83.23 chr08_17287431 88.87 FIRQV8.2 6.3 27.9 -0.92 97.6 27868043 chr08_27517915 94.04 chr08_29813774 102.97 10 FIRQV10.1 2.8 13.5 0.61 10.97 1222326 chr10_748430 4.24 chr10_1736076 19.75 FIRQV10.2 2.7 13.0 0.60 33.07 3356770 chr10_2259449 25.07 chr10_4144573 46.6 11 DAPEQV11.1 2.5 15.3 2.00 77.68 31321262 chr11_30708102 67.35 chr11_31585050 81.97 a Presented the QTLs with LOD > 2.4, and QTLs with a value of LOD > 3 are shaded in gray b Sign of the additive effect of the Ved allele c Physical position in version v.3.6.1 of the melon genome production as well as earlier aroma production, chlor- the Ved allele intensified the climacteric behavior, and an ophyll degradation, and abscission layer formation. The additive effect as high as 29.7 µL ethylene kg−1 h−1 in impact of the QTL was more obvious for the earliness of ETH was observed. Other QTLs with highly significant climacteric symptoms rather than for their presence. LOD scores were located in chromosomes 2 and 6. The Other minor QTLs were characterized in chromosomes QTL in chromosome 2 (FIRQV2.1/FIRQV2.2) is mainly 2, 3, 6, 7, 10, and 11. Ethylene production QTLs were only involved in FIR, although the other two minor QTLs with found in chromosomes 2 and 11. In some cases, QTLs lower significance were detected in the same region. The were found for only one of the climacteric traits: the QTL allele of Ved mitigated the climacteric behavior, increas- in the center of chromosome 6 affecting mostly CD and ing FIR and diminishing ABS. Using a conservative the QTL in the top of chromosome 10 influencing strategy, we could delimit the QTL in a region of around only FIR. 28 cM, between physical positions 3,049,874 and In the second mapping experiment we performed a 15,771,889 bp (Fig. 4). The QTL in chromosome 6 cor- QTL analysis with the mean values of the phenotyped related with chlorophyll degradation, with the Ved allele blocks. The results (Table 2 and Fig. 4) overlap to a high triggering this symptom. CDQV6.1 is located in a region extent with the first mapping experiment, but with fewer of 11 cM, with the maximum LOD score at 7,435,564 bp. QTL26 detected. ETHQV8.1 was detected for all evaluated Other nine QTLs (in chromosomes 3, 5, 6, 7, 10, and 11) traits, except for width of ethylene peak (WEP), with with slight effects in different aspects of climacteric LOD > 3, allowing to delimit a confidence interval of ripening were detected, although their LOD scores were 4.3 cM spanning around 500 kb, with a maximum LOD around the threshold of significance. All of them were also score in the physical position 9,634,968 bp. In all cases, detected in some of the individual-block mapping Pereira et al. Horticulture Research (2020)7:187 Page 7 of 18 Fig. 4 QTL mapping analysis performed using the mean values of the phenotyped blocks. Only chromosomes containing QTL are displayed. Each color corresponds to a different trait: ARO, light green; EARO, dark green; CD, brown; ECD, yellow; EALF, dark purple; ABS, light purple; HAR, grey; FIR, red; ETH, light blue; DAPE, turquoise; DAPP, dark blue; WEP, blue Pereira et al. Horticulture Research (2020)7:187 Page 8 of 18 experiments in similar positions and with the same Ved allele of ETHQV8.1, presenting various degrees of direction of additive effect. climacteric fruits, in comparison to those carrying the PS Due to the high effect of ETHQV8.1 in climacteric allele in homozygosis, which were non-climacteric (Sup- ripening, other minor QTLs could be masked and poorly plementary Table 3B). There is a clear association detected in our analysis. To minimize this effect, we between the genotype of ETHQV8.1 and EALF, with the separated the RIL population into two subsets, each of Ved allele decreasing the number of DAP of appearance them with ETH8QV8.1 fixed for the Ved or the PS allele of the climacteric symptom in both genetic backgrounds (VIII-Ved, n = 34 and VIII-PS, n = 55). For each subset, (Fig. 5A, B). Statistically, the values were significantly we did an interval mapping experiment using LOD > 3 as different between Ved and both PS and Hyb (p-values of threshold (Supplementary Table 2B) that revealed the 0.004 and 0.024, respectively) for the 720 family, and same QTLs but with higher significance and some new between all the possible combinations for the 414 family potential QTLs. When ETHQV8.1 is fixed for the Ved (p-values 0.03 PS-Hyb, 0.007 Ved-Hyb and <0.001 PS- allele, all the RILs are climacteric to a greater or lesser Ved). degree, so we should increase our power of detection for To further confirm the role of ETHQV8.1 in the onset of QTLs that are modulating ethylene production or their climacteric ripening, we evaluated the appearance and the physiological responses. The principal factors detected in earliness of climacteric symptoms for ILs PS8.2 and the previous analysis and in the VIII-Ved subset were in VED8.2 during the summer season in 2018 and 2019 chromosomes 2, 6, and 8 (Fig. 4); other new potential (Table 3). PS8.2, carrying an introgression of PS in the QTLs were located in chromosomes 5 and 9. For the genetic background of Ved, presented a climacteric subset VIII-PS, we observed non climacteric and climac- behavior during both years, but it resulted delayed when teric lines, but in any case the maximum ethylene pro- compared to Ved. A significant delay was observed in duction surpassed 50 µL kg−1 h−1, suggesting QTLs that chlorophyll degradation and abscission layer formation, triggered climacteric ripening later or with lower inten- especially in 2019; and also in the level of abscission, more sity. Furthermore, with this approach we detected the severe in 2018 (Supplementary Fig. 4). On the other side, same QTLs in chromosomes 2, 3, 5, and 7, and also a new VED8.2, carrying a Ved introgression in the genetic one at the top of chromosome 8 (Fig. 4). background of PS, showed varying degrees of climacteric fruits. The effect was observed both years, but more Validation and fine mapping of ETHQV8.1 clearly in 2019, when all the fruits showed a climacteric In order to validate ETHQV8.1, we used introgression phenotype, with sweet aroma production and abscission lines (ILs) with the background of one of the parental layer formation, while in 2018 only 50% of the fruits lines, Ved or PS, and an introgression of the other parent, presented this climacteric phenotype. Nevertheless, the PS or Ved, in chromosome 8, respectively. These lines Ved allele of ETHQV8.1 alone does not recover the phe- belong to a collection of ILs developed by our team (data notype of the climacteric parent completely (Supple- not shown). Initially, two IL families were used in this mentary Fig. 4), demonstrating the polygenic nature of study (720 and 414), derived from two BC3S1 segregating climacteric fruit ripening and its interaction with envir- families that, besides the region of ETHQV8.1, still con- onmental factors. tain a few additional introgressions in other chromosomes In addition, we measured the ethylene production in (Supplementary Table 3A). Two other more advanced both parental lines and the IL PS8.2. We could validate generation ILs, PS8.2 and VED8.2, carrying introgressions the effect of ETHQV8.1 in DAPE, DAPP, and ETH, in chromosome 8 in homozygosity and free of con- observing a weaker climacteric phenotype compared to taminations in other chromosomes were also included in Ved (Fig. 5C, D). In 2018, due to the extreme abscission the analysis. phenotype, we could not clearly observe the peak of We evaluated the appearance and the earliness of cli- ethylene production in Ved, but we could detect a sig- macteric symptoms for 720 and 414 IL families during the nificant delay of the ethylene peak in PS8.2, compared to 2017 summer season. The 720 family, segregating for the Ved (p-value of 0.007) (Fig. 5C). In 2019, we could clearly introgression of PS in the genetic background of Ved, observe the peak of ethylene production in all fruits presented in all cases climacteric fruits, with presence of (Fig. 5D). Statistically, the differences were significant for aroma, abscission layer, and in most cases change of color the three measured phenotypes DAPE, DAPP and ETH (Supplementary Table 3B). However, the earliness of the (p-values of 0.047, 0.0003, and <0.001, respectively). climacteric symptoms was delayed in the ILs carrying the We increased the SNP density between the flanking PS allele of ETHQV8.1, either in heterozygosity or markers of the ETHQV8.1 interval (Supplementary Table homozygosity. The evaluation of the 414 family, con- 3A, flanking markers shaded in red). The most informa- taining a Ved introgression in the genetic background of tive segregating ILs were genotyped, and three of them PS, showed clear differences between fruits carrying the had recombinations within the QTL interval (720.17, Pereira et al. Horticulture Research (2020)7:187 Page 9 of 18 Fig. 5 Evaluation of the appearance and the earliness of climacteric symptoms for ILs. A In all, 720 family, Ved background, in 2017 summer season. B In all, 414 family, PS background, in 2017 summer season, including images to show the external appearance of two individuals of 414 family with different alleles in ETHQV8.1. C Ethylene production of parental lines and the IL PS8.2 in 2018 summer season (n = 4) and D in 2019 summer season (n = 5), showing the mean of the replicates, and centered in the day of maximum production 414.1, and 414.17) (Supplementary Table 3C). Progeny and both the earliness and the intensity of abscission were seed of 414.1 and 414.17 were germinated, and a subset of strongly associated with the genotype of the line in the ILs were evaluated during the 2017 autumn season. None segregating part of the introgression (Supplementary Fig. of the 11 plants of the 414.1 family showed abscission, 5). Both families delimit ETHQV8.1 to a region of 154.1 although some fruits presented a subtle abscission layer Kb, which contains the maximum LOD peak obtained in around 55 DAP or later (Supplementary Table 3D). In the QTL mapping (Supplementary Table 3E). contrast, nine from the 15 plants from the 414.17 family presented abscission between 44 and 52 DAP. The family GWAS for climacteric ripening in a collection of melon 414.1 was segregating for a small introgression in chro- accessions mosome 8, which partially included ETHQV8.1, from the We performed an association study for aroma produc- initial position 9,446,475 to 9,603,217 bp (Supplementary tion (assessed subjectively by smelling the fruit) in a Fig. 5); the phenotype of this family ruled out that subset of 211 melon accessions corresponding to the ssp. ETHQV8.1 was located in this interval, since the subtle melo (Supplementary Table 4)30. We identified two close abscission layer observed was not associated with the association signals in the top of chromosome 8 (−log10 genotype of the lines. The family 414.17 segregated in the [P] = 7.0 and (−log10[P] = 6.0, respectively; Supplemen- upstream part of ETHQV8.1, from 9,446,475 to 9,757,323, tary Fig. 6). One of the signals coincides with the Pereira et al. Horticulture Research (2020)7:187 Page 10 of 18 Table 3 Basic statistics for climacteric ripening traits in the parental lines and the ILs (PS8.2 and VED8.2) evaluated in 2018 and 2019 2018 2019 Trait Line Median Mean SD p-value Median Mean SD p-value EARO VED 33 33.00 0.00 0.235 32 31.40 1.52 0.0659 PS8.2 33.5 34.25 1.89 34 35.00 3.46 PS 54.5 54.5 2.12 0.5 55 55 0 0.000075 VED8.2 47.5 48.2 11.2 43 44 3.32 ECD VED 33 33.25 0.50 0.595 31 31.60 2.19 0.0191 PS8.2 33.5 33.75 1.71 35 36.00 2.55 PS 41 41 9.9 0.836 55 52.6 5.37 0.00116 VED8.2 36.5 39 10.9 41 40 2 EALF VED 33 33.00 0.00 0.000105 32 32.20 2.28 0.0128 PS8.2 35 35.20 0.50 35.5 37.75 2.75 PS 54.5 54.5 2.12 0.746 55 55 0 0.0003 VED8.2 54 55.5 3.7 47 46 3.32 HAR VED 34 33.80 0.50 0.000115 36 36.40 1.67 0.000476 PS8.2 40 39.75 1.26 43 43.00 2.00 PS 54.5 54.5 2.12 0.749 55 55 0 0.0207 VED8.2 54 55.5 3.7 50 50.4 3.58 ABS VED 3 3 0 0.000423 1 1.4 0.894 0.217 PS8.2 1 1.25 0.5 1 0.8 0.447 PS 0 0 0 1 0 0 0 0.0133 VED8.2 0 0 0 1 1 0.707 FIR VED 3.4 3.28 0.314 0.0114 3.74 3.51 0.803 0.9 PS8.2 4.29 4.26 0.446 3.22 3.44 0.837 PS 4.38 4.38 0.177 0.0204 3.4 3.11 0.885 0.012 VED8.2 2.15 1.95 0.863 1.69 1.77 0.27 ETHQV8.1 region of 154.1 Kb identified by genetic been widely used and with demonstrated efficiency, it mapping. presents some disadvantages when applied to mapping populations, as it requires enough space to store con- Discussion tainers to contain the fruits during the entire ripening A gradient of climacteric ripening was observed in the period. Previous knowledge of the ripening date of each Ved × PS RIL population line is needed to adjust the harvest point. In addition, as The main approach used to determine the genetic the in planta development is interrupted, it may lead to control of climacteric fruit ripening has been the char- potential alterations in ripening. The quantitative, non- acterization of tomato ripening mutants in which the invasive method we used to monitor the ethylene peak climacteric wild type is compared with the non- allowed us the observation of ethylene responses and the climacteric mutant5. As an essential part of phenotyping timing of their appearance, and their intensity in planta31. of ripening mutants, ethylene production is evaluated The comprehensive evaluation of climacteric ripening in during the ripening process. Generally, the hormone is the Ved × PS RIL population has led to a complex and measured in detached fruits, which are collected when diverse group of phenotypes. From the two extreme they have acquired the competence to ripen but before phenotypes observed in the parental lines, the segregating the ripening process begins. Although this strategy has RIL population displayed several combinations differing Pereira et al. Horticulture Research (2020)7:187 Page 11 of 18 not only in quantity and earliness of ethylene production, chromosomes 1, 2, 3, and 1125. An IL population founded but also in ripening-associated phenotypes. Some by the same exotic parental, PI 161375, and PS, both non- ripening-associated phenotypes were not present in highly climacteric, led to the identification of QTLs for climac- climacteric lines, e.g. RIL 124 presented high production teric ripening (eth3.5) and flesh firmness (ff2.2, ff3.5, ff8.2, of ethylene, very sweet aroma, and abscission in all blocks, ff8.4 and ff10.2)33. Another study implicating the same PI but it did not change external color; whereas RIL 161375 x PS population revealed a second QTL 213 showed medium levels of ethylene, aroma, and a (ETHQV6.3) that also rescued the climacteric phenotype striking change of color to bright yellow without abscis- in the PS background when the exotic allele was present26. sion layer formation. The presence of the exotic allele in either eth3.5 or Some RILs that produce as little as 5% of the ethylene ETHQV6.3 promoted climacteric ripening, but the com- produced by Ved are phenotypically very similar, indi- bination of both conferred an earlier and more intense cating that only a minimal threshold of the hormone climacteric phenotype26. One QTL in chromosome 10 (around 2 µL kg−1 h−1) is necessary to trigger climacteric and possibly a second one in chromosome 7 were iden- ripening. However, some common characteristics were tified, using an IL population with the background of Ved observed in RILs producing a low quantity of ethylene: and introgressions from the non-climacteric Japanese they tended to ripen later, the ethylene peak was less “Ginsen makuwa” cultivar from the ssp. agrestis27. The defined in comparison with lines that produced higher present work has characterized several QTLs interfering levels of ethylene, and their phenotype was dependent on with different aspects of climacteric ripening in chromo- the environment, showing wider variation depending on somes 2, 3, 5, 6, 7, 8, 9, 10, and 11, using a combination of the block. We hypothesize that the amount of ethylene QTL mapping experiments. Several of these QTLs could produced by these RILs is very close to the minimum be allelic to the QTLs previously described (Supplemen- threshold necessary to trigger the climacteric behavior, so tary Table 5), since they co-localize in the same physical any environmental alteration that reduces it may have a positions. The minor effect of most of the described QTLs critical effect resulting in a non-climacteric fruit. fits with the hypothesis of a complex trait with polygenic Depending on the stability of the observed climacteric inheritance, displaying a continuum range of climacteric phenotypes among the RILs, we could identify three dif- intensity. In this work, the use of two phylogenetically ferent ripening patterns: climacteric lines, which had at close varieties as Ved and PS, both commercial and least one symptom of climacteric ripening in all blocks; belonging to ssp. melo, allowed to identify several QTLs unstable climacteric lines, showing an inconsistent beha- involved in climacteric ripening. vior in different blocks; and non-climacteric lines in all blocks. In general, unstable climacteric lines developed A major QTL, ETHQV8.1, is sufficient to trigger climacteric melons clearly climacteric in some blocks, with detectable ripening ethylene, aroma, and abscission layer formation, but Among the identified QTLs, ETHQV8.1 was most sig- totally non-climacteric in other blocks. A PCA accurately nificant in affecting both ethylene production and groups the RILs according with this classification and ripening-associated phenotypes. For all the recorded traits evidences the high proportion of climacteric and unstable except WEP, a highly significant QTL was mapped in an climacteric lines in comparison with non-climacteric ones almost identical interval as ETHQV8.1, around the phy- in this RIL population (Supplementary Fig. 7). sical position 9,634,968 bp of chromosome 8. In all of The phenotypic distribution anticipates the existence of these traits the Ved allele is causing a stronger climacteric multiple QTLs with variable effects modulating climac- phenotype, explaining between 20.2% and 58.9% of the teric ripening. A panel of melon accessions was previously variance. Furthermore, in the RIL subset Ved-VIII, which classified in degrees of climacteric ripening, using prin- contains only the RILs with the Ved allele in the cipally FIR and abscission (ABS) as indicative traits32. ETHQV8.1 interval, none of them is non-climacteric. It is Although their phenotyping was not as extensive as in our possible that ETHQV8.1 is allelic to Al-3, a major gene work, they suggested a non-absolute classification of described in the PI 161375 x Ved RIL population as melon varieties according to this trait. These complex and essential to trigger fruit abscission and endogenous polygenic nature of the trait was also demonstrated by synthesis of ethylene25. However, the physical position of previous works that characterized several QTLs for cli- Al-3 could not be determined as the genetic map where it macteric ripening or their associated effects. A RIL was described was exclusively based on AFLP markers, population obtained from the cross between Ved and PI precluding the comparison of the QTL positions. 161375, a non-climacteric accession, was used to identify In order to validate ETHQV8.1, we generated ILs with two genes responsible for abscission layer formation, Al-3 the PS and Ved background and with introgressions and Al-4 in chromosomes 8 and 9, respectively, and four containing PS or Ved alleles in the ETHQV8.1 interval, QTLs affecting the amount of ethylene production in respectively. The evaluation of two ILs, VED8.2 and PS8.2, Pereira et al. Horticulture Research (2020)7:187 Page 12 of 18 Fig. 6 Functional analysis of the two candidate genes of ETHQV8.1. A, B identification of conserved protein motifs using orthologs from several species (Arabidopsis, tomato, peach, kiwi, apple, papaya and pepper); the gray bars represent non-synonymous SNPs between PS and Ved. C, D Multiple protein alignment containing the most likely causal variant for each gene; in red box and letters, the aminoacid change in melon proteins between PS (before the position) and Ved (after the position) with PS and Ved background, respectively, proved the produced abscission layer and fell from the plant, sug- contribution of ETHQV8.1 to the climacteric phenotype. gesting that in the PS background, fruit abscission is less According to the polygenic control of the trait, the IL with environment-dependent. Therefore, the phenotyping of climacteric background (PS8.2) did not lose completely two progenies of recombinant individuals from the family the climacteric phenotype. However, a significant delay 414 reduced ETHQV8.1 to an interval between the posi- was observed in climacteric symptoms caused by a delay tions 9,564,672–9,757,323 bp. Using the version v4.0 of and also a decrease of the ethylene production. This effect the annotation of the melon genome34, 14 candidate is different to the one reported using mutants of the genes are located inside the interval (Supplementary CmNAC-NOR gene with in the “Charentais Mono” Table 6). Tentatively, three of them could be discarded background (cantaloupensis group), where the mutants because they are not expressed in fruit tissues in Melonet- showed a delay in ethylene production and subsequently, DB35; by functional annotation, three genes could be in climacteric symptoms appearance, but without a related to ETHQV8.1, encoding an ethylene-responsive reduction in the amount of ethylene28. On the other side, transcription factor ERF024 (MELO3C024520), a serine/ the IL with non-climacteric background (VED8.2) pre- threonine kinase CTR1-like (MELO3C024518) and a sented a smooth climacteric phenotype, but highly protein ROS1 (MELO3C024516). MELO3C024518 has six dependent on the environment. non-synonymous variants between PS and Ved. CTR1 is a To fine map ETHQV8.1, we selected two individuals negative regulator of ripening that interacts physically that had recombined inside the QTL interval and per- with ethylene receptors36. In the absence of ethylene, the formed an evaluation of their progenies in autumn 2017. ethylene receptor activates CTR1, which leads to the The climacteric behavior was less intense to the one inhibition of the downstream transduction pathway; when observed during summer, as expected, since even below ethylene is present, the ethylene receptor terminates the controlled conditions, fruit ripening is altered. The cli- activation of CTR1, thus releasing ethylene responses37. macteric fruits did not change substantially their external MELO3C024516 has 10 non-synonymous variants color and did not produce aroma, nevertheless, they between Ved and PS. ROS1 encodes a demethylase, and Pereira et al. Horticulture Research (2020)7:187 Page 13 of 18 demethylation has been reported as one of the main accessions of the melo ssp. Two close association mechanism regulating fruit ripening15,38. In tomato, the signals were detected in chromosome 8, one of them ROS1 orthologue DML2 is a demethylase governing fruit located in the 154.1 Kb that contains the above- ripening14. MELO3C024520 has not sequence variants mentioned candidate genes (Supplementary Table 7). between PS and Ved causing changes in the protein, and This result suggests that ETHQV8.1 may be important its expression in fruit is low34. for regulating climacteric fruit ripening in melon To further investigate the relevance of the non- germplasm. synonymous variants between Ved and PS within the two most likely candidate genes MELO3C024516 and Other minor QTLs interact to modulate the climacteric MELO3C024518, we obtained a phylogenetic tree of the response homologous proteins in other climacteric (tomato, peach, In addition to ETHQV8.1, other genetic factors are kiwi, apple, papaya) and non-climacteric (pepper) species, involved in the ripening process in our RIL population. In also including Arabidopsis (Supplementary Fig. 8). We general, they are not affecting the amount of produced identified the closest orthologs to MELO3C024516 ethylene, but other traits as flesh firmness or external fruit (CmROS1) and MELO3C024518 (CmCTR1) using a color. Although some of them, as FIRQV2.1/FIRQV2.2 phylogenetic analysis and selected them to study the and CDQV6.1, presented a highly significant LOD (>3.5), conserved motifs containted in these proteins (Fig. 6A, B). in most cases they have a limited effect in ripening from a Both proteins were highly conserved among species and global perspective, in contrast to ETHQV8.1. 15 significant motifs were identified in each of them. Two regions, in chromosomes 2 and 7, seemed to be CmCTR1 and CmROS1 proteins were complete and implicated in fruit ripening in a general way, since QTLs included all the motifs observed in the tomato and Ara- for different traits were detected. In chromosome 2, we bidopsis orthologs. Of the six non-synonymous SNPs identified two QTLs for FIR, which colocalize with other contained in CmCTR1 (Fig. 6A), three were located within QTLs for DAPP and ABS. Furthermore, the other two conserved motifs, potentially affecting the protein. Among QTLs for EARO and HAR are detected in the Ved-VIII them, two were conservative aminoacid replacements subset and a forth one for FIR in the PS-VIII. In all cases, (E386Q and E628D), which most likely might not impact the Ved allele was diminishing or delaying the climacteric protein function, and one (P272T) was contained in motif response (Table 2 and Supplementary Table 2B). We 8 and was non-conservative (Fig. 6C). Ved, the climacteric hypothesized that a unique QTL affecting the earliness of parent, presented a threonine at this position, as well as ethylene production is located between the physical most analyzed climacteric species (tomato, kiwi, apple, positions 3,049,784–15,771,889 bp of chromosome 2, peach); meanwhile PS presented a proline, as one of the which contains 669 annotated genes. We did not identify two orthologs of the non-climacteric pepper. However, QTLs for the qualitative traits (ARO, CD) even in the PS- the papaya ortholog also had a proline in this position. For VIII subset, suggesting that FIRQV2.1/FIRQV2.2 is CmROS1, we observed 10 non-synonymous changes and incapable to trigger the autocatalytic ethylene production two of them were contained within conserved protein by itself. Previous QTLs in chromosome 2 have been motifs (Fig. 6B). The first one, contained in motif 13, was described for FIR and ETH (Supplementary Table 5). The a non-conservative change (E250G), although the ami- second region was identified in chromosome 7, presenting noacid was quite variable in this position among species three overlapping QTLs for HAR, EARO, and ABS, with a (20% R, 20% V, 53% I, and 6.7%E). The second one, maximum LOD of 2.8 in the physical position located in motif 8, was from glycine (aliphatic) in PS to 1,046,645 bp, contributing positively to climacteric serine (polar) in Ved at position 1685 (Fig. 6D), but again ripening and explaining around 15% of the variance this position was not highly conserved. According to the (Table 2). A QTL in a similar position was mapped using non-synonymous variants observed within the parents of the PS-VIII subset for HAR, with a LOD of 3.45, so this population for the two candidate genes, their possibly this QTL is one of the factors that contribute to expression levels in the melon expression atlas35 and the trigger climacteric ripening when ETHQV8.1 is fixed for previous knowledge about their function in other cli- PS alleles. The CI for the QTL is delimitated by the macteric species, both candidate genes could be involved physical positions 835,787–1,657,550 bp and contains 129 in climacteric ripening and be the underlying gene for genes. Although a QTL for climacteric ripening in chro- ETHQV8.1. Further experiments, such as the generation mosome 7 was not described before, an IL with an exotic of knock-out mutants and the analysis of expression levels introgression in this region that was much less climacteric during fruit ripening are needed to validate either of them than the control has been reported27; however, an addi- as the underlying gene for ETHQV8.1. tional introgression in chromosome 10 that actually In addition, we performed a GWAS analysis for modified climacteric ripening made it difficult to assess production of aroma in a subset of 211 re-sequenced the contribution of chromosome 7. Pereira et al. Horticulture Research (2020)7:187 Page 14 of 18 CDQV6.1 controls one process exclusively, the chlor- ETHQW9.1 (Supplementary Table 2B). WEPQW5.1 could ophyll degradation that leads to a visible change of color. modulate the rate at which the ethylene production It was detected with significant LOD in the mapping increases during the onset of ripening, until reaching the experiments with three blocks (Supplementary Fig. 3C) peak; so it could be participating in promoting the auto- and with Ved-VIII subset (Supplementary Table 2B), in catalytic synthesis of the hormone, rather than interacting addition to the one with mean values (Table 2). The Ved with the physiological responses. ETHQW9.1 may be allele increased CD in a climacteric background, as the regulating the amount of ethylene produced but may not QTL was significant in the Ved-VIII subset, but we could have the ability of initiating the production itself, which not detect it in the PS-VIII subset. Mutants with a similar would explain why we could not map it using data from phenotype had been studied in tomato and pepper, the whole population. ETHQW9.1 could be allelic to a identifying a STAY-GREEN protein as responsible for second major gene implicated in fruit abscission in the their inability to degrade chlorophyll at the onset of distal part of chromosome 9 (Supplementary Table 5). On ripening39. CDQV6.1 is located between the physical the opposite side, the PS-VIII subset suggested the exis- positions 6,598,284–7,589,164 bp of chromosome 6 and tence of minor QTLs in chromosomes 3, 5, 7, and 8 that contains 106 genes. The interval did not contain CmNAC- could be provoking a slight climacteric ripening. Most of NOR (coordinates 27,663,292–27,665,351 bp of chromo- them were already mapped in very similar positions some 6), in fact no LOD peak was detected at these (Supplementary Fig. 3 and Fig. 4) except EALFQX8.3 and positions; probably the interactions between the QTL DAPPQX8.3. The Ved allele for these QTLs were accel- with the genetic background and the environment caused erating the climacteric response, but they were not the loss of detection for the effect of this gene in our detected for the qualitative traits (ARO, CD) and were population. therefore probably not sufficient to trigger ethylene QTLs for flesh firmness were mapped in chromosomes production. 10 and 8. Two QTLs, FIRQV10.1 and FIRQV10.2, were The ripening behavior of the RIL population obtained located in chromosome 10 (Table 2). Since both are by crossing a highly climacteric and a non-climacteric increasing FIR when the Ved allele is present and are close melon type suggests that climacteric fruit ripening is a from each other, probably there is a single QTL in the complex trait in melon. A major QTL ETHQV8.1 is suf- interval 748,430–1,736,076 bp (Supplementary Fig. 3E). ficient to trigger climacteric ripening, although additional This interval contains 154 genes. FIRQV8.2 was mapped minor QTL are important to modulate the climacteric in chromosome 8, around 10 Mb downstream ETHQV8.1, response. The genomic interval containing ETHQV8.1, with a maximum LOD of 6.3 in the position 27,868,043 bp which was also identified in a parallel GWAS experiment, (Table 2). A QTL for FIR in similar positions was detected contains at least three candidate genes that may be related using the Ved-VIII subset, with a maximum LOD = 3.4 to fruit ripening. Additional experiments are needed in (Supplementary Table 2B). Fruit texture has been broadly order to identify the gene responsible for the ETHQV8.1 studied in melon due to its great importance from the climacteric phenotype. This work allowed the identifica- commercial perspective: the market desires climacteric tion of a new component that contributes to this complex varieties that remain firm during the postharvest trait, which may be useful in breeding programs aimed at stage22,27,40. Unlike in tomato, the softening of fruit flesh obtaining long-shelf life melon types. in melon is partially controlled by ethylene, so the non- climacteric types also ripen entering in a phase of sugar Materials and methods accumulation and decrease of flesh firmness24. A com- Plant material parative transcriptomic analysis of climacteric and non- Plant material was grown in a greenhouse at Caldes de climacteric varieties showed that the set of enzymes Montbui (Barcelona). Plants were pruned weekly and modifying cell-wall metabolism was different in Ved pollinations were executed manually, limiting develop- (polygalacturonases, glucan endo-1,3-β-glucosidases, and ment to only one fruit per plant. The harvest point was β-d-xylosidases) and PS (fascilin-like arabinogalactanan determined by the following criteria: (a) abscission date protein)22. when the fruit abscised; (b) after seven days of the The mapping experiment that discriminates the RIL appearance of any symptom of climacteric ripening in population depending on the haplotype for ETHQV8.1 absence of abscission; (c) at 55–62 days after pollination allowed us to propose three new QTLs that were masked (DAP) when fruits were non-climacteric. in the main QTL mapping analysis and to support some The RIL population used in this work was described in of the minor QTLs already detected (Fig. 4), in addition to Pereira et al. Briefly, the population was developed from a gaining some insight about how these genetic factors cross between Ved (ssp. melo, cantalupensis group) and work. Using the Ved-VIII subset, we could detect two PS (ssp. melo, inodorus group) and contained 89 RILs. QTLs related to ethylene production, WEPQW5.1 and Five blocks (T1-T5), consisting in a unique fruit per RIL Pereira et al. Horticulture Research (2020)7:187 Page 15 of 18 and 1–3 fruits per control line (Ved, PS and the F1 (Hyb)), formation (ALF); 1, subtle and/or partial ALF; 2, almost were grown during the summers of 2015 and 201629. The complete ALF with obvious scar; and 3, total ALF gen- three blocks from 2015 (T1-T3) were sowed and planted erally with fruit abscission. The aroma production was with a delay between them of 20 days approximately. evaluated each day by smelling the fruits. The firmness of Although all the phenotypic evaluations were performed fruit flesh was measured at harvest using a penetrometer within the summer period (July-September) and the (Fruit TestTM, Wagner Instruments), in at least three conditions in the greenhouse were partially controlled, we regions of the fruit (distal, proximal and median), and the considered them as independent trials. The two blocks mean value was registered. Ripening-related traits were from 2016 (T4-T5) were planted at the same time, but phenotyped in the five blocks of the RIL population (T1- were treated as independent trials to have a homogeneous T5). ILs were evaluated in 2017, 2018, and 2019 following design. the same criteria. Two introgression line (IL) collections were obtained For GWAS analysis, climacteric ripening was recorded from the F1 of the Ved x PS cross after three backcrosses evaluating aroma production by smelling the fruits. with the recurrent parentals Ved and PS, respectively Accessions without aroma were recorded as “0” and (data not shown). Two families of ILs (720 and 414), accessions with strong aroma were recorded as “2“. containing introgressions in chromosome 8 of PS and Ved in the background of Ved and PS, respectively, were Ethylene production evaluated during the summer of 2016 (family 720, n = 18 Ethylene production in attached melon fruits was and family 414, n = 12, respectively) and the progenies of measured using gas chromatography – mass spectrometry two recombinant individuals of the 414 family (414.1, n = (GC-MS) as previously described31. Two blocks (T3 and 11 and 414.17, n = 15) during the autumn of 2017. Two T4) of RILs were evaluated for this trait. Ethylene pro- advanced ILs (PS8.2 and VED8.2), containing introgres- duction in attached fruits was measured in 66 RILs both sions in chromosome 8 of PS in the background of Ved, years, in 17 RILs only 1 year and six RILs were excluded and Ved in the background of PS, were evaluated during from the analysis due to infections or a wrong fruit set. In the summer of 2018 (n = 4) and 2019 (n = 5). addition, IL PS8.2 and both parental lines were evaluated Melon accessions for genome-wide associacion studies during the 2018 and 2019 summer seasons. (GWAS) were obtained from the National Mid-term The ethylene peak was monitored during the ripening Genebank for Watermelon and Melon (Zhengzhou, period customizing the design for each RIL. The first China), the Zhengzhou Fruit Research Institute (Chinese measurement before the onset of ripening was determined Academy of Agricultural Sciences) and the US Depart- for each RIL based on previous evaluations, due to the ment of Agriculture. Plants were grown in Zhengzhou segregation of earliness of ethylene production. For cli- (Henan province), Sanya (Hainan province) and Changji macteric lines, the atmosphere of the chamber containing (Xinjiang province) in 2015 and 2016 and were previously the fruit was measured every other day while ethylene was described30. Because of poor adaptation, some accessions undetectable and every day after ethylene detection; the were evaluated in one or two locations, and three repli- measurements were stopped when: a) fruit abscised from cates were evaluated at each location. the plant or b) ethylene production decreased during at least two consecutive days. For non-climacteric lines, the Phenotyping of climacteric ripening traits atmosphere of the chamber was examined at least every Ripening-related traits were evaluated as qualitative three days, confirming that they did not produce any (production of aroma (ARO), chlorophyll degradation detectable amount of ethylene during fruit ripening. (CD), and abscission (ABS)) or quantitative (rest of traits) For the ILs evaluated in 2018 and 2019 (PS8.2 and (Table 1). Traits were divided in four different groups VED8.2), the measurements were started at 25 DAP, and according to the physiological response to the production stopped when: (a) fruit abscised from the plant or (b) of ethylene: biosynthesis of volatiles leading to a sweet 5 days after the formation of the abscission layer. The aroma (ARO and EARO), change of color mainly due to non-climacteric parent PS was examined using the same chlorophyll degradation (CD and ECD), abscission layer criteria as in the RILs. formation in the pedicel of the fruit, provoking abscission To better characterize the ethylene peak, four traits in some cases (EALF, ABS and HAR) and softening of were defined: maximum production of ethylene in the fruit flesh (FIR). The visual inspection of melon fruits, peak (ETH), earliness of ethylene production, represent- attached to the plant, was performed daily, from ing the first day when ethylene was detectable (DAPE), approximately 25 DAP until harvest. In addition, indivi- earliness of the ethylene peak (DAPP), and width of dual pictures of the fruits were obtained weekly. ARO and ethylene peak (WEP), the latter calculated subtracting CD were recorded as 0 = absence and 1 = presence. ABS DAPP and DAPE (Table 1). For non-climacteric lines, the was recorded using an index from 0, no abscission layer earliness of the trait was considered as missing data. Pereira et al. Horticulture Research (2020)7:187 Page 16 of 18 DNA extraction and genotyping generate a phylogenetic tree in NGphylogeny.fr47. The DNA extractions were performed from young leaves algoritm MAFFT was used for the alignment48, BMGE for following the CTAB protocol41 with some modifica- the alignment curation49, and PhyML+SMS for the tree tions29. The IL families 720 and 414 and the ILs PS8.2 and inference50. Afterwards, the sequences belonging to the VED8.2 were genotyped with SNPs using the KASPar same clade as the melon candidate gene were used to SNP Genotyping System (KBiosciences, Herts, UK). identify the conserved protein motifs using MEME KASPar assay primers were designed following the pro- (MEME Suite 5.1.1)51, allowing 15 motifs at maximum. tocol of LGC Genomics (Supplementary Table 8A). The Later on we generated a protein multiple alignment in R genotyping of SNPs across the genome was performed (package msa52) with the function msaMuscle. using the high-throughput genotyping system Biomark HD, based on Fluidigm technology, with 96 × 96 chips. Statistical analyses Additional SNPs located within the flanking SNPs of All the statistical analyses and graphical representations ETHQV8.1 were genotyped by qPCR (Supplementary were obtained using the software R v3.2.353 with the Table 8B). All the SNPs used were obtained in silico from RStudio v1.0.143 interface54. the re-sequencing of both parental lines42. The PCA was performed using R package “factoextra”. The re-sequenced accessions used in GWAS are To obtain the correlation matrix among traits we calcu- described in Zhao et al.30. lated the Pearson coefficient with the R package “Hmisc” and the visualization of data was performed with “corr- Genetic map construction and QTL mapping plot”. To compute these functions the data should not The genetic map, containing SNPs and INDELs geno- contain any missing data. In order to include in the typed by genotyping-by-sequencing (GBS), was previously analysis the non-climacteric lines without any phenotypic described in29. Phenotypic data was used to perform a symptoms, EARO, ECD, EALF, DAPP and DAPE values QTL mapping analysis43 with the interval mapping pro- were imputed with the latest harvest date value (55–62 cedure in each block individually, T1-T5 for ripening- DAP), and for WEP as the maximum value of this variable related traits and T3-T4 for ethylene traits. The threshold (8 days). The lines without any data for ethylene pro- of significance was fixed at LOD > 2.544 and the 1-LOD duction were excluded of the PCA and correlation confidence interval (CI) was used to locate the QTL. analysis. QTLs were considered significant only when at least three Mean comparisons between the IL groups were per- of them were detected in similar positions and con- formed with the function “pairwise.t.test” from the tributed to ripening in the same direction. Chromosomes package “stats”. To represent the EALF in the non- that did not contain any significant QTL were not climacteric lines, we substituted the missing value by represented. 55–62 DAP, as explained above. In order to increase the statistical power, a second interval mapping analysis was performed using the mean Genome-wide association study values for each RIL (LOD threshold = 2.4). A Principal Considering the different independent domestication Component Analysis (PCA) based on a subset of RILs and events in melon, we only used C. melo L. ssp. melo variables, evaluated in five blocks and without missing accessions described in ref. 30. Only SNPs with minor data, showed that, when extracting the effect of the line, allele frequency ≥ 0.05 and missing rate ≤ 0.4 were used to the environmental effect was not associated with the year carry out GWAS. This resulted in 1,923,713 SNPs that (Supplementary Fig. 9). were used in GWAS. We performed GWAS using the To name the QTLs, the first letter corresponds to the trait Efficient Mixed-Model Association eXpedited (EMMAX) code (Table 1), followed by a “Q”, a letter identifying the program55. Population stratification and hidden related- experiment (“V”, “W”, and “X” in the mapping with the mean ness were modeled with a kinship (K) matrix in the values, the subset VIII-VED and the subset VIII-PS, respec- emmax-kin-intel package of EMMAX. The P-value tively), a number indicating the chromosome, a dot and a digit thresholds for significance were approximately 2.51 × to differentiate QTLs in the same chromosome45. 10−6. Phylogenetic analysis and multiple alignments To obtain the orthologs of the candidate genes, the Acknowledgements This work was supported by the Spanish Ministry of Economy and protein sequences from melon candidate genes were Competitiveness grants AGL2015–64625-C2–1-R and RTI2018-097665-B-C2, blasted in Plaza46, Uniprot and NCBI. The most similar Severo Ochoa Programme for Centres of Excellence in R&D 2016–2010 (SEV- proteins from climacteric (tomato, peach, kiwi, apple, 2015-0533) and the CERCA Programme/Generalitat de Catalunya to J.G.-M., L.P. and M.S.-D. were supported by a FPI grant from the Spanish Ministry of papaya) and non-climacteric (pepper) species, also Economy and Competitiveness. V.R. was supported by the European Union’s including Arabidopsis, were downloaded and used to Horizon 2020 research and innovation programme under Marie Skłodowska- Pereira et al. Horticulture Research (2020)7:187 Page 17 of 18 Curie grant agreement No 6655919. Y.X. was supported by the China 14. Liu, R. et al. A DEMETER-like DNA demethylase governs tomato fruit ripening. Agriculture Research System (CARS-25). Proc. Natl Acad. Sci. USA 112, 10804–10809 (2015). 15. Lang, Z. et al. Critical roles of DNA demethylation in the activation of ripening- Author details induced genes and inhibition of ripening-repressed genes in tomato fruit. 1 Proc. Natl Acad. Sci. USA 114, E4511–E4519 (2017). Centre for Research in Agricultural Genomics (CRAG) CSIC-IRTA-UAB-UB, Edifici CRAG, Campus UAB, 08193 Cerdanyola, Barcelona, Spain. 2IRTA (Institut 16. Seymour, G. B. et al. A SEPALLATA gene is involved in the development and de Recerca i Tecnologia Agroalimentàries), Edifici CRAG, Campus UAB, 08193 ripening of strawberry (Fragaria x ananassa Duch.) fruit, a non-climacteric Cerdanyola, Barcelona, Spain. 3Department of Biology, University of tissue. J. Exp. Bot. 62, 1179–1188 (2011). Toronto–Mississauga, Mississauga, ON L5L 1C6, Canada. 4Zhengzhou Fruit 17. Cherian, S., Figueroa, C. R. & Nair, H. ‘Movers and shakers’ in the regulation of Research Institute, Chinese Academy of Agricultural Sciences, Zhengzhou, fruit ripening: a cross-dissection of climacteric versus non-climacteric fruit. J. China. 5Shenzhen Branch, Guangdong Laboratory for Lingnan Modern Exp. Bot. 65, 4705–4722 (2014). Agriculture, Genome Analysis Laboratory of the Ministry of Agriculture, 18. McAtee, P. A. et al. The hybrid non-ethylene and ethylene ripening response Agricultural Genomics Institute at Shenzhen, Chinese Academy of Agricultural in kiwifruit (Actinidia chinensis) is associated with differential regulation of Sciences, Shenzhen, China MADS-box transcription factors. BMC Plant Biol. 15, 304 (2015). 19. Minas, I. S. et al. Discovery of non-climacteric and suppressed climacteric bud Author contributions sport mutations originating from a climacteric Japanese plum cultivar (Prunus M.P. and J.G.-M. designed studies and contributed to the original concept of salicina Lindl.). Front Plant Sci. 6, 316 (2015). the project, V.R. performed the bioinformatics analysis, L.P., M.S.-D., J.A., Y.X., 20. Elitzur, T. et al. Banana MaMADS transcription factors are necessary for fruit and Y.H. managed the plant material, performed the phenotyping and the ripening and molecular tools to promote shelf-life and food security. Plant mapping experiments, M.P. participated in the ethylene measurements, G.Z., Physiol. 171, 380–391 (2016). Q.L., and S.H. performed the GWAS analysis, L.P., M.S.-D., M.P., and J.G.-M. 21. Ezura, H. & Owino, W. O. Melon, an alternative model plant for elucidating fruit drafted the paper. ripening. Plant Sci. 175, 121–129 (2008). 22. Saladié, M. et al. Comparative transcriptional profiling analysis of developing melon (Cucumis melo L.) fruit from climacteric and non-climacteric varieties. Conflict of interest BMC genomics 16, 440 (2015). The authors declare that they have no conflict of interest. 23. Ayub, R. et al. Expression of ACC oxidase antisense gene inhibits ripening of cantaloupe melon fruits. Nat. Biotechnol. 14, 862–866 (1996). Supplementary Information accompanies this paper at (https://doi.org/ 24. Pech, J. C., Bouzayen, M. & Latche, A. Climacteric fruit ripening: Ethylene- 10.1038/s41438-020-00411-z). dependent and independent regulation of ripening pathways in melon fruit. Plant Sci. 175, 114–120 (2008). Received: 28 May 2020 Revised: 4 August 2020 Accepted: 16 September 25. Perin, C. et al. Molecular and genetic characterization of a non-climacteric 2020 phenotype in melon reveals two loci conferring altered ethylene response in fruit. Plant Physiol. 129, 300–309 (2002). 26. Vegas, J., Garcia-Mas, J. & Monforte, A. J. Interaction between QTLs induces an advance in ethylene biosynthesis during melon fruit ripening. Theor. Appl Genet. 126, 1531–1544 (2013). References 27. Perpiñá, G., Esteras, C., Gibon, Y., Monforte, A. J. & Picó, B. A new genomic 1. Giovannoni, J., Nguyen, C., Ampofo, B., Zhong, S. & Fei, Z. The Epigenome and library of melon introgression lines in a cantaloupe genetic background for Transcriptional Dynamics of Fruit Ripening. Annu Rev. Plant Biol. 68, 61–84 dissecting desirable agronomical traits. BMC Plant Biol. 16, 154 (2016). (2017). 28. Rios, P. et al. ETHQV6.3 is involved in melon climacteric fruit ripening and is 2. Lelièvre, J.-M., Latchè, A., Jones, B., Bouzayen, M. & Pech, J.-C. Ethylene and fruit encoded by a NAC domain transcription factor. Plant J. 91, 671–683 (2017). ripening. Physiol. Plant. 101, 727–739 (1997). 29. Pereira, L. et al. QTL mapping of melon fruit quality traits using a high-density 3. Bemer, M. et al. The tomato FRUITFULL homologs TDR4/FUL1 and MBP7/ GBS-based genetic map. BMC Plant Biol. 18, 324 (2018). FUL2 regulate ethylene-independent aspects of fruit ripening. Plant Cell 24, 30. Zhao, G. et al. A comprehensive genome variation map of melon identifies 4437–4451 (2012). multiple domestication events and loci influencing agronomic traits. Nat. 4. Osorio, S. et al. Integrative comparative analyses of transcript and metabolite Genet. 51, 1607–1615 (2019). profiles from pepper and tomato ripening and development stages uncovers 31. Pereira, L., Pujol, M., Garcia-Mas, J. & Phillips, M. A. Non-invasive quantification species-specific patterns of network regulatory behavior. Plant Physiol. 159, of ethylene in attached fruit headspace at 1 p.p.b. by gas chromatography- 1713–1729 (2012). mass spectrometry. Plant J. 91, 172–183 (2017). 5. Giovannoni, J. J. Fruit ripening mutants yield insights into ripening control. 32. Leida, C. et al. Variability of candidate genes, genetic structure and association Curr. Opin. Plant Biol. 10, 283–289 (2007). with sugar accumulation and climacteric behavior in a broad germplasm 6. Vrebalov, J. et al. A MADS-Box gene necessary for fruit ripening at the tomato collection of melon (Cucumis melo L.). BMC Genet. 16, 28 (2015). Ripening-Inhibitor (Rin) locus. Science 296, 343–346 (2002). 33. Moreno, E. et al. Candidate genes and QTLs for fruit ripening and softening in 7. Manning, K. et al. A naturally occurring epigenetic mutation in a gene melon. Theor. Appl Genet. 116, 589–602 (2008). encoding an SBP-box transcription factor inhibits tomato fruit ripening. Nat. 34. Ruggieri, V. et al. An improved assembly and annotation of the melon Genet. 38, 948–952 (2006). (Cucumis melo L.) reference genome. Sci. Rep. 8, 8088 (2018). 8. Shima, Y. et al. Tomato FRUITFULL homologs regulate fruit ripening via 35. Yano, R., Nonaka, S. & Ezura, H. Melonet-DB, a Grand RNA-Seq Gene Expression ethylene biosynthesis. Biosci. Biotechnol. Biochem. 78, 231–237 (2014). Atlas in Melon (Cucumis melo L.). Plant Cell Physiol. 59, e4 (2018). 9. Chung, M.-Y. Y. et al. A tomato (Solanum lycopersicum) APETALA2/ERF gene, 36. Kieber, J. J., Rothenberg, M., Roman, G., Feldmann, K. A. & Ecker, J. R. CTR1, a SlAP2a, is a negative regulator of fruit ripening. Plant J. 64, 936–947 (2010). negative regulator of the ethylene response pathway in arabidopsis, encodes 10. Powell, A. L. et al. Uniform ripening encodes a Golden 2-like transcription a member of the Raf family of protein kinases. Cell 72, 427–441 (1993). factor regulating tomato fruit chloroplast development. Science 336, 37. Binder, B. M. The ethylene receptors: complex perception for a simple gas. 1711–1715 (2012). Plant Sci. 175, 8–17 (2008). 11. Vrebalov, J. et al. Fleshy fruit expansion and ripening are regulated by the 38. Gao, Y. et al. Diversity and redundancy of the ripening regulatory networks Tomato SHATTERPROOF gene TAGL1. Plant Cell 21, 3041–3062 (2009). revealed by the fruitENCODE and the new CRISPR/Cas9 CNR and NOR 12. Lin, Z. et al. A tomato HD-Zip homeobox protein, LeHB-1, plays an important mutants. Hortic. Res. 6, 39 (2019). role in floral organogenesis and ripening. Plant J. 55, 301–310 (2008). 39. Barry, C. S., McQuinn, R. P., Chung, M. Y., Besuden, A. & Giovannoni, J. J. Amino 13. Zhong, S. et al. Single-base resolution methylomes of tomato fruit develop- acid substitutions in homologs of the STAY-GREEN protein are responsible for ment reveal epigenome modifications associated with ripening. Nat. Bio- the green-flesh and chlorophyll retainer mutations of tomato and pepper. technol. 31, 154–159 (2013). Plant Physiol. 147, 179–187 (2008). Pereira et al. Horticulture Research (2020)7:187 Page 18 of 18 40. Bianchi, T. et al. Textural properties of different melon (Cucumis melo L.) fruit 48. Katoh, K. & Standley, D. M. MAFFT multiple sequence alignment software types: Sensory and physical-chemical evaluation. Sci. Horticulturae 201, 46–56 version 7: improvements in performance and usability. Mol. Biol. Evol. 30, (2016). 772–780 (2013). 41. Doyle, J. J. Isolation of plant DNA from fresh tissue. Focus 12, 13–15 (1990). 49. Criscuolo, A. & Gribaldo, S. BMGE (Block Mapping and Gathering with Entropy): 42. Sanseverino, W. et al. Transposon insertions, structural variations, and snps a new software for selection of phylogenetic informative regions from mul- contribute to the evolution of the melon genome. Mol. Biol. Evol. 32, tiple sequence alignments. BMC Evol. Biol. 10, 210–210 (2010). 2760–2774 (2015). 50. Lefort, V., Longueville, J.-E. & Gascuel, O. SMS: smart model selection in PhyML. 43. MapQTL Version 3.0: Software for the calculation of QTL positions on genetic Mol. Biol. Evol. 34, 2422–2424 (2017). maps (Plant Research International Biometris (WU MAT), 1999). 51. Bailey, T. L. et al. MEME SUITE: tools for motif discovery and searching. Nucleic 44. Van Ooijen, J. LOD significance thresholds for QTL analysis in experimental Acids Res. 37, W202–W208 (2009). populations of diploid species. Heredity 83, 613–624 (1999). 52. Bodenhofer, U., Bonatesta, E., Horejš-Kainrath, C. & Hochreiter, S. msa: an R 45. Diaz, A. et al. A consensus linkage map for molecular markers and quantitative package for multiple sequence alignment. Bioinforma. (Oxf., Engl.) 31, trait loci associated with economically important traits in melon (Cucumis 3997–3999 (2015). melo L.). BMC Plant Biol. 11, 111 (2011). 53. R: A language and environment for statistical computing (R Foundation for 46. Van Bel, M. et al. PLAZA 4.0: an integrative resource for functional, evolutionary Statistical Computing, Vienna, Austria, 2012). and comparative plant genomics. Nucleic Acids Res. 46, D1190–D1196 (2018). 54. RStudio: Integrated development environment for R (RStudio Inc, Boston, 47. Lemoine, F. et al. NGPhylogeny.fr: new generation phylogenetic services for 2012). non-specialists. Nucleic Acids Res. 47, W260–W265 (2019). 55. Kang, H. M. et al. Variance component model to account for sample structure in genome-wide association studies. Nat. Genet. 42, 348–354 (2010).