Theoretical and Applied Genetics https://doi.org/10.1007/s00122-021-03998-6 ORIGINAL ARTICLE A cryptic variation in a member of the Ovate Family Proteins is underlying the melon fruit shape QTL fsqs8.1 Cecilia Martínez‑Martínez1,7 · Maria José Gonzalo1 · Pablo Sipowicz1,2 · Manuel Campos1 · Irene Martínez‑Fernández1 · Carmen Leida3 · Mohammed Zouine4 · Konstantinos G. Alexiou5,6 · Jordi Garcia‑Mas5,6 · María Dolores Gómez1 · Pablo Tornero1 · Miguel Ángel Pérez‑Amador1 · Cristina Esteras · Belén Picó3 · Carlos Romero1 · Antonio J. Monforte1 Received: 11 September 2021 / Accepted: 8 November 2021 © The Author(s) 2021 Abstract Key message The gene underlying the melon fruit shape QTL fsqs8.1 is a member of the Ovate Family Proteins. Variation in fruit morphology is caused by changes in gene expression likely due to a cryptic structural variation in this locus. Abstract Melon cultivars have a wide range of fruit morphologies. Quantitative trait loci (QTL) have been identified underly- ing such diversity. This research focuses on the fruit shape QTL fsqs8.1, previously detected in a cross between the accession PI 124112 (CALC, producing elongated fruit) and the cultivar ‘Piel de Sapo’ (PS, producing oval fruit). The CALC fsqs8.1 allele induced round fruit shape, being responsible for the transgressive segregation for this trait observed in that population. In fact, the introgression line CALC8-1, carrying the fsqs8.1 locus from CALC into the PS genetic background, produced perfect round fruit. Following a map-based cloning approach, we found that the gene underlying fsqs8.1 is a member of the Ovate Family Proteins (OFP), CmOFP13, likely a homologue of AtOFP1 and SlOFP20 from Arabidopsis thaliana and tomato, respectively. The induction of the round shape was due to the higher expression of the CALC allele at the early ovary development stage. The fsqs8.1 locus showed an important structural variation, being CmOFP13 surrounded by two deletions in the CALC genome. The deletions are present at very low frequency in melon germplasm. Deletions and single nucleotide polymorphisms in the fsqs8.1 locus could not be not associated with variation in fruit shape among different melon accessions, what indicates that other genetic factors should be involved to induce the CALC fsqs8.1 allele effects. Therefore, fsqs8.1 is an example of a cryptic variation that alters gene expression, likely due to structural variation, resulting in phenotypic changes in melon fruit morphology. Introduction Communicated by Herman J. van Eck. Melon (Cucumis melo L.) is a crop species that belongs to the Cucurbitaceae family, which also includes impor- Cecilia Martínez-Martínez and Maria José Gonzalo have contributed equally to this work. tant cultivated species such as cucumber (Cucumis sativus L.), watermelon (Citrullus lanatus Thumb.), pumpkins and * Antonio J. Monforte amonforte@ibmcp.upv.es 4 1 Laboratoire de Recherche en Sciences Végétales, Instituto de Biología Molecular y Celular de Plantas Université de Toulouse, CNRS, UPS, Toulouse INP, (IBMCP), Consejo Superior de Investigaciones Científicas Auzeville‑Tolosane, France (CSIC), Universitat Politècnica de València, Ingeniero 5 Fausto Elio s/n, 46022 Valencia, Spain Centre for Research in Agricultural Genomics (CRAG), 2 CSIC-IRTA-UAB-UB, Edifici CRAG​, Campus UAB, Instituto Nacional de Tecnología Agropecuaria (INTA), Bellaterra, Barcelona, Spain Buenos Aires, Argentina 6 3 Institut de Recerca i Tecnologia Agroalimentàries (IRTA), Instituto de Conservación y Mejora de la Agrodiversidad Bellaterra, Barcelona, Spain Valenciana (COMAV), Universitat Politècnica de València 7 (UPV), Valencia, Spain Present Address: Universidad de Almería, Almería, Spain 13 Vol.:(0123456789) Theoretical and Applied Genetics squashes (Cucurbita spp.), gourds [Momordica charan- andromonoecious and monoecious cultivars need pollina- tia L., Lagenaria siceraria Molina. Standley, Benincasa tors to ensure proper fruit production, so the selection of hispida (Thunb.) Cogn., Luffa aegyptiaca Mill. Luffa the andromonoecious mutation does not seem to be related acutangula L. Roxb., Trichosanthes cucumerina L.] and to an increase in fruit set. It is likely that the selection other minor crops as Cucumis metuliferus Meyer ex Nau- sought rounder fruit. On the contrary, the increase of in din and Cucumis anguria L. Together with watermelon the fruit carpel number, controlled by p, has been associ- and cucumber, melon is one of the cucurbit crops with ated with rounder fruits in several mapping populations the greatest economic impact. Worldwide production has (Perin et al. 2002; Monforte et al. 2004). Wild melons increased since the beginning of the 21st Century from 20 and most traditional and modern cultivars develop three million to more than 27 million tons (FAOSTAT 2021). carpels, while five carpels are found only in a small subset Melon was domesticated independently at least twice of Indian and Oriental cultivars. Strong evidences based in Africa and Asia (Endl et al. 2018; Zhao et al. 2019), on investigations performed in cucumber (Li et al. 2016), although the cultivars that are commonly grown world- co-localization of candidate genes with the p locus in wide derived from the Asian domestication through experimental mapping populations (Pan et al. 2020) and divergent diversification from the Indian subcontinent to a genome-wide association study (GWAS) panel indicate Oriental and Occidental Eurasian regions (Gonzalo et al. that the gene p may be the melon homologue (CmCLV3) 2019; Zhao et al. 2019). One of the consequences of this of CLAVATA3 which tomato homologue (SlCLV3) controls diversification was the development of a large number of locule number in the tomato fruit (Xu et al. 2015). For landraces, traditional and modern cultivars that show an the rest of the QTLs, candidate genes belonging to the impressive diversity in fruit morphology, from the small CELL NUMBER REGULATOR (CNR/FW2.2), SlKLUH/ (30–50 g) round fruit of wild melons to the large and CYP785A/FW3.2, OVATE, SUN, TON1 RECRUTING extremely elongated fruit of flexuosus varieties (Monforte MOTIF (TRM), CELL SIZE REGULATOR (CSR/FW11.2), et al. 2014). Pitrat (2017) proposed an exhaustive horticul- YABBY, and WOX families have been proposed based on tural classification of melon that reflects such phenotypic the co-location of members from these families with diversity, defining nineteen groups, one wild (agrestis), melon fruit morphology QTLs (Monforte et al. 2014; Pan two semi‐domesticated or feral (kachri and chito) and six- et al. 2020). Among them, OVATE, CSR and SUN showed teen domesticated and cultivated ones. a larger number of co-locations (Pan et al. 2020), although The genetic control of melon fruit morphology has no functional verification has been carried out until now. been studied in several works, and quantitative trait loci Diaz et al. (2014) mapped several fruit shape QTLs in (QTL) and consensus QTLs have been defined (reviewed a segregating population derived from the cross between by Monforte et al. 2014; Pan et al. 2020). The pleiotropic the Indian landrace PI 124,112 (CALC, producing elon- effects of two genes, a and p, on fruit shape are widely gated fruit) and the Spanish “Piel de Sapo” (PS, producing known (Perin et al. 2002): a gene controlling female flower oval fruit) cultivar. The QTL with the strongest effect was sex expression, being strictly female or hermaphrodite in mapped to chromosome 8 and was named fsqs8.1. Sur- monoecious or andromonoecious varieties, respectively, prisingly, round fruits were induced by the CALC allele, whereas p controls the fruit carpel number (with three which explained the transgressive segregation observed for carpels being common for most varieties and five for a fruit shape in this population. The QTL was introgressed few other). Regarding the pleiotropic effects observed into PS by developing the introgression line (IL) CALC8- for a, fruit from strictly female flowers (in monoecious 1, which carries a segment of chromosome 8 from CALC plants) are usually much more elongated than fruit from containing the fsqs8.1 locus, while the rest of the genome hermaphrodite flowers in andromonoecious plants (Mon- was PS. CALC8-1 produced round fruits, verifying the forte et al. 2005; Abdelmohsin and Pitrat 2008). Gene a effect of fsqs8.1 (Diaz et al. 2014). Interestingly, fsqs8.1 encodes a 1-AMINOCYCLOPROPANE-1-CARBOXYLIC maps to the same genome region as the consensus QTL ACID SYNTHASE (CmACS-7), and a missense mutation FSQM8 (Monforte et al. 2014) and CmFSI8.3 (Pan et al. in the gene produces a loss of CmACS-7 protein activity, 2020). Additionally, Gur et al. (2017) showed a strong leading to the development of stamens in the female flower association of fruit shape in the same chromosome 8 (andromonoecy), which results in subsequent pleiotropic genomic region in a diverse melon germplasm collection effects making a rounder fruit shape (Boualem et al. 2008). by GWAS. These works indicated that fsqs8.1 could be Whereas wild melons are monoecious, the andromonoe- one of the major loci responsible for the enormous fruit cious mutation was selected during cultivar diversification. shape variability observed in melon germplasm. Wu et al. Traditional Indian landraces (the primary center of melon (2018) fine-mapped fsqs8.1 between markers Sca76-197 diversity) are monoecious, while andromonoecy is more and Sca76-45 on chromosome 8. Several genes were iden- common in Mediterranean and modern cultivars. Both tified in that region, including a Serine/threonine kinase 13 Theoretical and Applied Genetics (MELOC025211), a 1-aminocyclopropane-1-carboxylate DNA was extracted according to Doyle and Doyle (1990) oxidase (MELO3C025210) and a member of the OFP with minor modifications. Genotyping was carried out with family (MELO3C025206) that was previously named the KASPAR technology (Smith and Maughan 2015) by as CmOFP13 by Monforte et al. (2014). Among them, the Genotyping Service at the John Innes Center (Norwich, CmOFP13 was pointed out as the strongest candidate UK). Recombinants were transferred to the greenhouse to gene. obtain self-seed. The position of the recombinant events was The objective of the present work is to identify the established by genotyping the recombinant plants with an causal gene underlying fsqs8.1, to track the history of its array of SNP markers (Supplementary Table S2) distributed variability during melon diversification, to shed light on along the fsqs8.1 locus by KASPAR and High Resolution its effects during ovary development, and to investigate its Melting (HRM, Wittwer et al. 2003) with MELTDOC- genetic effects in several genetic landscapes. TOR™ (ThermoFisher Scientific, Waltham, MA) follow- ing the supplier recommendations. Seedlings from selected ­F3 families were genotyped with the appropriate markers to identify homozygous seedlings for CALC and PS for the Material and methods corresponding segregating recombinant introgression. The selected seedlings were transferred to greenhouse. Plants Plant material from each family were planted in the same row in the green- house, randomly distributed within each family. The fruit Diaz et al. (2014) previously developed the introgression shape index (FSI) was defined as the ratio between maxi- line CALC8-1 that carries an introgression on chromo- mum fruit length and maximum fruit diameter. The mean some 8, covering the fsqs8.1 locus, from the traditional FSI for the groups of plants with a contrasting allele com- melon accession PI 124,112 (CALC) from Bihar, India, position at the segregating introgression within each F ­ 3 fam- into the “Piel de Sapo” (PS) cultivar from Spain (Oliver ily were compared with a t-test to determine if fsqs8.1 was et al. 2001). Eduardo et al. (2005) developed the intro- present or not in the respective introgressions. gression line SC8-3, which carries an introgression with fsqs8.1 locus from the Korean cultivar “Songwan Charmi” Sequencing and genomic analysis of fsqs8.1 locus (SC) accession PI161375 in the same PS genetic back- ground. The three genotypes are andromonoecious. Large Amino acid sequences of tomato, Arabidopsis thaliana ­F 2 populations were developed from crosses between and melon OVATE family protein (OFP) members, were CALC8-1 and PS for fine mapping. SC8-3 and CALC8-1 retrieved from previous works (Monforte et al. 2014) and were also crossed to study gene interaction between aligned with CLUSTAL. Evolutionary distances were com- alleles. puted using the Poisson correction method (Zuckerkandl and An array of melon cultivars and accessions from different Pauling 1965) in units of the number of amino acid sub- world regions and belonging to several melon horticultural stitutions per site, removing ambiguous positions for each groups was obtained from the Instituto de Conservación y sequence pair. The tree was calculated with the neighbor- Mejora de la Agrodiversidad Valenciana (COMAV, Poly- joining method (Saitou and Nei 1987), and the statistical technic University of Valencia, Spain) germplasm bank support of the clusters was assessed by a bootstrap test with (Supplementary Table S1) to study associations between 2000 replicates (Felsenstein et al. 1985). All the analysis Single Nucleotide Polymorphism (SNP) markers tightly were conducted with MEGA X (Kumar et al. 2018). linked to fsqs8.1 and fruit shape. A set of five of them were MELO3C025206 was amplified from PS, CALC8-1 selected and crossed with CALC8-1 and PS to investigate and SC8-3 using the Cmofp13 primers (Supplementary interaction between fsqs8.1 and the genetic background. The Table S2) with Taq I DNA polymerase (Biotools, Madrid, cultivar ‘Mucha Nesvi’ (MN), a monoecious ameri cultivar, Spain). The PCR products were cleaned with ExoSap-IT™ was crossed with CALC8-1 to verify the epistatic interac- (ThremoFisher Scientific) and sequenced by the Sequenc- tion for fruit shape between the sex expression gene a and ing Service at the Instituto de Biología Molecular y Celu- fsqs8.1. lar de Plantas (IBMCP; Valencia, Spain) with an ABI3130 Genetic analyzer (Applied Biosystems, Foster City, CA). Positional cloning The sequences were aligned with CLUSTAL implemented in MEGA X (Kumar et al. 2018). Three thousand five hundred ­F 2 seedlings from the Illumina raw sequences from melon accessions (Zhao CALC8-1 × PS cross were screened for the markers flank- et al. 2019) were retrieved from the Sequence Read Archive ing the fsqs8.1 position Sca76-197 and Sca76-42 and the (SRA) under the NCBI BioProject with the accession ID internal marker Sca76-140b (Supplementary Table S2). PRJNA565104. Adaptors and low-quality regions were 13 Theoretical and Applied Genetics removed from the reads using cutadapt (Martin 2011). toluidine blue, and the images of ovaries were captured Clean, high-quality reads were then mapped to the melon using a Leica DM5000 microscope equipped with the LAS reference genome v4.0 (Castanera et al. 2020) using BWA- X application at the microscopy services at the IBMCP. MEM (Li 2013), and the mapping results were visualized Additionally, flowers from six days prior anthesis to with the Integrative Genomics Viewer IGV (Robinson et al. full anthesis stages were collected freshly and sectioned 2011). longitudinally. The ovary hair was eliminated to facilitate The 10 × Chromium sequencing was performed at the definition of ovary contour. Ovaries were observed the GeT-PlaGe core facility, INRA Toulouse (France). with a Leica DMS 1000 stereomicroscopy equipped with The Chromium libraries were prepared according to 10X the LAS X application for image capture at the microscopy Genomics’ protocols using the Genome Reagent Kits v2. service of the IBMCP. The samples quantity and quality controls were validated Gene expression analysis during flower development on Qubit, Nanodrop and Femto. The optimal performance was performed in the previously defined four develop- was characterized on input gDNA with a mean length ment stages in PS, SC8-3 and CALC8-1. Five independ- greater than 50 kb. The libraries were prepared from 3 µg ent replicates were analyzed. RNA extractions were per- of HMW gDNA (cut off at 50 Kb using BluePippin system) formed according to the protocol from the Aurum Total or 0.625–1.8 ng of HMW DNA. Briefly, in the microfluidic RNA Mini kit (Biorad, Hercules, CA)) and treated with Genome Chip, a library of Genome Gel Beads was com- Ambion DNAse I (RNAse-free) to eliminate DNA. cDNA bined with HMW template gDNA in Master Mix and parti- was then synthesized from 1 µg of total RNA using Pri- tioning oil to create Gel Bead-In-EMulsions (GEMs) in the meScript 1st strand cDNA synthesis kit. The expression Chromium. Each gel bead was functionalized with millions of genes was evaluated through quantitative RT-PCR of copies of a 10x™ Barcoded primer. Upon dissolution of (q-PCR) by using the TAKARA SBRY Premix ExTaq (Tli the genome gel bead in the GEM, primers containing (i) an RNAseH Plus). Supplementary Table S2 shows the cor- Illumina R1 sequence (Read 1 sequencing primer), (ii) a responding primers used for q-PCR. The q-PCR primers 16 bp 10 × Barcode, and (iii) a 6 bp random primer sequence were designed from the coding regions of each gene by were released. A read 1 sequence and the 10x™ Barcode using the Primer3 software (Untergasser et al. 2012). To were added to the molecules during GEM incubation. P5 avoid possible cross-amplification, and before any q-PCR and P7 primers, Read 2, and Sample Index were added dur- experiment, the homology of each primer was checked ing library construction. 10 cycles of PCR were applied to using the BLAST tool from MELONOMICS against the amplify libraries. Library quality was assessed using a frag- CM4.0 genome (www.​melon​omics.​net) and the expected ment analyzer, and libraries were quantified by q-PCR using size tested in agarose gels. The q-PCR reactions consisted the Kapa Library Quantification Kit. The libraries were per- of two step of hold at 95 °C 30 s and 40 cycles of 3 s at formed on an Illumina HiSeq3000 using a paired-end read 95 °C, and 30 s at 60 °C. The relative expression of each length of 2 × 150 pb with the Illumina HiSeq3000 sequenc- gene was determined by the comparative Ct (Cycle Thresh- ing kits. The 10 × Genomics chromium libraries were assem- old) method using C. melo ACTIN gene (MELO3C023264) bled using the 10 × Genomics Supernova software (Weisen- as the internal standard. The efficiency of amplification feld et al. 2017), using default parameters. for each amplicon was tested using increasing amounts of template cDNA, and were all roughly equivalent. The Ovary development analysis and RT‑PCR absolute value of the ΔCt slope (Ct of the target gene-Ct of the reference gene) versus serial dilutions of cDNA for For the development study, ovaries from the three geno- a given sample must be less than 0.1. The relative expres- types CALC8-1, SC8-3 and PS were collected fresh at sion of each gene was then calculated relative to a cali- different stages: E1 (carpel primordial initiated), E2 (pri- brator sample using the formula 2­ −ΔΔCt, where ΔΔCt is mordial stigmas appear), E3 (stigma and style defined and the difference between the ΔCt of each sample and the anthers with pollen) and E4 (ovule primordial initiated) ΔCt of the calibrator sample. Means of relative expression that correspond to stages 6, 8-1, 8-2, 8-3, respectively, between genotypes within each ovary development stage from Bai et al. (2004). The ovaries were fixed overnight in were compared with the nonparametric Kruskal–Wallis 4% (w/v) p-formaldehyde in 0.1 M sodium phosphate pH test. 7.2 with 0.05% (v/v) of Tween 20 at 4 °C and dehydrated in a graded ethanol series as described in Gomez et al. Arabidopsis thaliana transformation (2004, 2011). After dehydration, the samples were embed- ded in Paraffin wax (Paraplast Plus). To obtain the ovary MELO3C025206 open reading frame from both CALC8-1 sections, samples in paraffin were sectioned in a Leica was inserted into pDONR222 (Invitrogen, Waltham, MA) RM2025 microtome. The sections were stained with 0.02 by PCR with two specific oligonucleotides (GGG​GAC​AAG​ 13 Theoretical and Applied Genetics TTT​GTA​CAA​A AA​AGC​AGG​C TT​C GA​AGG​AGA​TAG​ with the contrast 2*CALC8-1-x-SC8-3-(CALC8- AAC​CAT​GAG​AAA​TCA​CAA​GTT​CCG​T and GGG​GAC​ 1-SC8-3) = 0 to determine the statistical significance of the CAC ​ T TT​ GTA​ CAA​ GAA ​ AGC ​ T GG ​ GTT ​ T CA ​ A AG​ AGG​ gene action. TGG​AGG​AGA​AGG) using a BP Gateway reaction (Inv- Regarding the interaction of the fsqs8.1 alleles with the a itrogen). Upon confirmation of the sequence, cDNA was sex expression gene, two hundred seedlings belonging to the transferred to pGWB14 (Nakagawa et al. 2007) to render ­F2 population derived from the cross CALC8-1-x- ‘Mucha MPA635 (35S:OVATE_CAL8-1), and it was introduced Nesvi’ (MN) were genotyped using fsqs8.1 SNPs markers in Arabidopsis thaliana (L.) Heynh. (ecotype Col-0) by and the perfect marker for the a gene ALU-CAPS (Boualem Agrobacterium-mediated floral-dip transformation (Clough et al. 2008) to select plants homozygous for the four possible and Bent 1998). T1s, T2s, and T3s lines were selected by allelic combinations at both genes. Plants were grown in the presence of hygromycin (Duchefa Biochemie, Haarlem, the COMAV greenhouse facilities using a completely ran- The Netherlands) in Murashige and Skoog (MS) medium domized design. The interaction between genes was con- plates (Murashige and Skoog 1962) (Duchefa). Several T2 trasted by a two-way ANOVA. lines were inspected, and four representative lines were fixed Finally, interaction of fsqs8.1 alleles with different genetic for the transgene in homozygosis. Seeds were sterilized in backgrounds (GB) was performed in F1 hybrids from the ethanol and germinated in MS for 4 days at 4 °C in the dark, crosses between CALC8-1 and PS with an array of six dif- followed by 7–8 days at 22 °C under a long day photoperiod ferent melon cultivars (Supplementary Table S1), having, (16 h/8 h). Seedlings were then transferred to soil and grown therefore, the CALC or PS fsqs8.1 alleles in heterozygous in a chamber at 22 °C/19 °C under long day photoperiod conditions. The hybrids were grown in two consecutive (16 h/8 h). Pictures from seedlings at day 9, flowers, pistils, years (2016 and 2017) at the COMAV greenhouse facilities and fruits were taken with a SMZ1270 stereomicroscope in a completely randomized design with ten replicated. The (Nikon, Leuven, Belgium), whereas those from the rosettes effect of the genetic background, fsqs8.1 allele, year and of adult plants were obtained with an Alpha ILCE-6000L their respective interactions were assessed by the ANOVA digital camera (Sony, Tokyo, Japan). model: Yijkl = 𝜇 + FSQS8.1i + GBj + Yeark + FSQS8.1 ∗ GBij Association of fsqs8.1 locus haplotypes with fruit shape in melon germplasm + FSQS8.1 ∗ Yearik + GB ∗ Yearjk + eijkl Sixty-six melon accessions (Supplementary Table S1) were where FSQS8.1i indicates the CALC or PS fsqs8.1 allele in grown in COMAV greenhouse facilities in 2014 and 2018 the hybrid in heterozygous conditions,GBj the effect of the seasons in a completely randomized design with three bio- different backgrounds, ­Yeark the effect of the year, and the logical replicates and the FSI was obtained for each one. respective two-way interactions. Statistical analyses were Accessions were genotyped with SNPs Sca76-snp192, carried out with JMP® v. 11 (SAS Institute Inc. Cary, NC). Sca76-snp155, Sca76-snp140 and the sca76-95indel-ov makers (Supplementary Table S2) located in the fsqs8.1 locus. Associations between the markers and the FSI were Results studied by ANOVA. Haplotypes were defined via visual inspection. FSI means of the haplotypes with at least five Fine mapping of fsqs8.1 accessions were compared with Tukey’s multiple compari- son test (p < 0.05). Statistical analyses were carried out with The genomic position of fsqs8.1 was previously defined JMP® v. 11 (SAS Institute Inc. Cary, NC). between markers Sca76-197 and Sca76-45 on chromo- some 8 of the melon genome (Wu et al. 2018). Accord- Interactions of fsqs8.1 in different genetic situations ing to the CM4.0 melon genome version (Castanera et al. 2020), the physical distance of those markers is 169 kb To study the gene action between the CALC and SC fsqs8.1 (Fig. 1) between positions 26,839,235 and 27,008,844, alleles five plants of each, CALC8-1, SC8-3 and the hybrid which carries five annotated genes: MELO3C025211 CALC8-1-x-SC8-3 were grown in a completely randomized (serine/threonine kinase), MELO3C025210 (gibberellin design in the COMAV greenhouse facilities. Fruits were col- 2-beta-dioxygenase 2-like), MELO3C028356 (hypothetical lected at full mature stages and the FSI index recorded. The unknown protein with some homology to a retrotransposon gene action (d/[a]), with d being the dominance deviation gag protein), MELO3C025206 (ovate family protein), and and [a] the absolute additive value) was calculated as: MELO3C025205 (RING-type E3 ubiquitin transferase). d/[a] = [ CALC8-1-x-SC8-3-(CALC8-1-SC8-3)/2]/ Interestingly, the 93 kb region between MELO3C025210 [(CALC8-1 + SC8-3)/2]. 13 Theoretical and Applied Genetics Chromosome 8 26958640 26839236 26843360 26844244 26846236 26851732 26854220 26866972 26879084 26879328 26896908 26898532 26901464 26907378 26937196 26941614 26947890 26948328 26948384 26949154 26952140 26953528 26957724 26958548 26959322 26963844 26996244 26961184 Posion PI 124112 DHL92 Repetitive sca76-45 PS MELO3C025206 MELO3C025211 SNP8 SNP4 SNP3 sca76-156b sca76-84 sca76-80 sca76-137a Sca76-197 snp-OVATE SNP1 sca76-192 sca76-191 sca76-155 Deletion-marker MELO3C28356 sca76-95indel-ov Small_deletion_end End-major_deletion Start_major_deletion Small_deletion_start MELO3C025205 Sca76-184 sca76-168 sequences sca76-140 MELO3C025210 CALC_8-1 5 cm CALC_8-1 fsqs8.1 Fig. 1  High-resolution mapping of fsqs8.1. A schematic represen- lower part. Blue dashed lines indicates the large deletions observed tation of the reference genome V4.0 (Castanera et al. 2020) based in CALC8-1. The position of fsqs8.1 based on the fine mapping on the double haploid line DHL92 is depicted in the upper part of experiments is indicated below the CALC8-1 genome. The candi- the figure, indicating the position of markers, annotated genes and date MELO3C025206 is highlighted in green. On the right, typical the repetitive sequence rich region. The length of the genes and the fruits of PI 124,112 (CALC, donor parent), PS (recipient parent) and direction of the transcription are indicated with arrows. The deduced CALC8-1 (introgression line carrying the fsqs8.1 locus from CALC genome of CALC8-1 based on re-sequencing data is depicted in the in PS background) are also shown) (color figure online) and MELO3C025206 is composed by long terminal genomic region was not of sufficient quality, the read mapping repeats (LTR) and repetitive elements, characteristics of pattern for both genotypes was similar as the one observed in retrotransposons, and only contains one annotated gene, the previous re-sequencings, confirming the presence of two MELO3C028356. deletions surrounding MELO3C025206 (Supplementary Fig. A fine mapping strategy was followed in order to reduce S1) in the CALC8-1 genome. A PCR marker was developed the number of candidate genes that might underlie fsqs8.1. to amplify a sequence within the putative large deletion; the Sequence reads previously obtained by Sanseverino et al. deletion-marker failed to amplify either CALC8-1 or CALC (2015) from PS, CALC, and SC accessions were mapped to but produced a clear amplification product in PS (Supple- the melon reference genome CM4.0 to identify SNPs for the mentary Fig. S1). These results suggested a large 40 kb dele- subsequent fine mapping efforts. The reference genome was tion, approximately from positions 26,900,000 to 26,938,000, obtained from the double haploid line DHL92 derived from and an smaller 3 kb deletion from positions 26,949,154 to the cross between PS and SC. The genomic region including 26,952,139 (Fig. 1) in the CALC fsqs8.1 locus. the fsqs8.1 locus in DHL92 corresponds to the PS genome. We obtained seven highly informative recombinants in the Twenty-five variant calls were identified; however, no vari- fsqs8.1 locus after two rounds of recombination screening of ant calls were found in the 49 kb region between the posi- 3500 ­F2 CALC8-1xPS seedlings, with recombination points tions from 26,898,531 to 26,947,890, corresponding to the among the annotated genes (Supplementary Table S3). The previously mentioned LTR-rich region (Fig. 1). A visual progeny analysis of those recombinants showed the fsqs8.1 inspection of PS, SC and CALC read mappings with IGV locus expanded from positions 26,937,197 to 26,958,640. (Robinson et al. 2011) showed that no CALC Illumina reads The only coding sequence in that region corresponded to mapped to this genomic region, except for some low-quality MELO3C025206 (Fig. 1), strongly suggesting that this gene read mappings corresponding to repetitive sequences, while is underlying the fsqs8.1 QTL. BLAST analysis showed that mapping reads were observed for PS and SC (Supplementary MELO3C025206 encodes for a member of the ovate fam- Fig. S1). Furthermore, a second smaller region lacking CALC ily proteins, previously named by Monforte et al. (2014) as Illumina reads at 5' of MELO3C025206 was also observed. CmOFP13. The direction of the translation was reverse to In order to verify that this pattern was also present in the IL the reference genome, i. e., the sequence up to 26,937,197 CALC8-1 genome, the IL and PS were re-sequenced using was downstream to MELO3C025206, and sequence up to 10X Chromium technology. Although the assembly in this 26,958,640 upstream. 13 Theoretical and Applied Genetics Ovary development and gene expression in A. thaliana (Wang et al. 2016), tomato (Rodriguez et al. 2011) and other plant species (Dangwal and Das 2018). The final melon fruit shape is usually determined pre-anthe- Monforte et al. (2014) identified the melon gene members sis during ovary development (Perin et al. 2002; Eduardo belonging to this family. A phylogenetic analysis was carried et al. 2005). Ovary shape was studied in three genotypes: out with A. thaliana, tomato and melon OFPs (Fig. 3, Sup- PS, CALC8-1 and SC8-3. SC8-3 is an IL that carries an plementary Fig. S3). CmOFP13 clustered with AtOFP1 and introgression in the fsqs8.1 locus from SC (Eduardo et al. SlOFP20, and it is likely a homolog of these genes. More 2005). As fruits from SC8-3 are very elongated (Eduardo importantly, AtOFP1 and SlOFP20 induce organ shorten- et al. 2007), it is likely that the three genotypes represent ing in their respective species (Hackbusch et al. 2005; Wu an allelic series for fsqs8.1 ranging from elongated (SC8- et al. 2018). Therefore, phylogenetic analysis reinforced the 3), oval (PS) to round (CALC8-1). Differences in ovary putative role of CmOFP13 in the regulation of organ shape shape at anthesis and pre-anthesis were observed among in melon. them (Fig. 2a, Supplementary Fig. S2), CALC8-1 showed CmOFP13 coding sequences from SC8-3, CALC8-1 a clear rounder shape four days before anthesis (DBA) and and PS CmOFP13 were obtained with Sanger technology. SC8-3 an elongated shape even at 6 DBA. To find the ovary Only one synonymous mutation was found between PS and developmental stage when the differences in ovary shape CALC8-1. The same mutation was also present in SC8-3 appeared, ovaries were collected in four early developmen- plus two additional synonymous mutations (Supplemen- tal stages: E1 (carpel primordial initiated), E2 (primordial tary Fig. S4). On the other hand, sequence variations were stigmas appear), E3 (stigma and style defined and anthers observed in the surrounding upstream and downstream non- with pollen) and E4 (ovule primordial initiated). Differences coding regions (Supplementary Figure S5). In the down- in ovary shape were already evident for SC8-3 at E2, with stream sequence, 24 SNPs were found in PI161375 and elongated ovaries compared to PS and CALC8-1, whereas the deletion of approximately 40 kb at 3 kb from the stop CALC8-1 showed rounded ovaries compared to PS at E4 codon of CmOFP13 coding sequence in PI124112 compared (Fig. 2b). Regarding ovary length and width, PS showed with the reference sequence (PS). In the case of upstream strong growth increase between E2 and E3 compared with sequence, 123 and 114 SNPs were found in PI161375 and CALC8-1, suggesting that there was already an inhibition PI124112, respectively. The 3-kb deletion in PI124112 is of the growth in CALC8-1 at this stage that ended up with located at −724 nucleotides from the transcription start point the variation in shape observed at E4 stage. (according the annotation of accession XP_016903000 in Next, we wanted to correlate the expression of CmOFP13 www.​ncbi.​nlm.​nih.​gov). Therefore, these variations may with the variation of ovary shape through the ovary develop- affect the promoter sequences of CmOFP13 and most prob- mental stages among PS, SC8-3 and CALC8-1. Expression ably the expression profile of the gene but not preventing of CmOFP13 between the E2 and E4 stages showed different it completely. Thus, lack of non-synonymous mutations patterns among genotypes (Fig. 2c). CmOFP13 was residu- in the coding regions, the existing sequence variability in ally expressed in SC8-3 at E2 and low levels at E3, whereas the surrounding regions and the differences of gene expres- in PS the expression was low (although higher than SC8-3 in sion between PS and CALC8-1 strongly suggested that the E2) in those stages, but, on the contrary, the expression was changes in organ shape induced by CmOFP13 in CALC8-1 high in CALC8-1 in those stages. Finally, the expression was might be due to changes in gene expression, not differences similar among genotypes in E4. Therefore, the correlation in protein function between alleles. between CmOFP13 and ovary shape was observed: lack of The sequence variation in CMOFP13 across melon expression in E2 for elongated shape, low expression in E2 germplasm was studied in 1175 re-sequenced acces- and E3 with oval shape and high expression in E2 and E3 sions (Zhao et al. 2019). Three synonymous mutations with round shape. Figure 2d depicts representative ovaries were identified in the CMOFP13 coding region (Fig. 4a). at E4, where the three different ovary shapes were evident Haplotypes could be constructed for 1024 of them, show- among all three genotypes. ing a certain geographical distribution of the haplotypes (Fig. 4b). Haplotype A was characteristic of the China Characterization of CmOFP13 gene sequence and Far East regions (what overlaps with the distribution and surrounding genomic region of ssp. agrestis cultivars). Haplotype B (PS haplotype) was the most common haplotype in the Europe, America, CmOFP13 has a 999-nucleotide open reading frame, encod- Middle and Near East in regions, overlapping with the ing for a protein compromised of 331 amino acids with no distribution of ssp. melo cultivars. Haplotype C (CALC introns. CmOFP13 includes an ovate sequence motif, there- haplotype) was characteristic of India and CALC, although fore likely being a member of the OVATE family proteins was also present in America. Haplotype D had a very low (OFP). This family has been extensively studied especially frequency. In general, the haplotypes seemed to be mainly 13 Theoretical and Applied Genetics a 1 cm SC8-3 PS CALC8-1 b * * * * * * * * * * * * * * c b a a a SC8-3 c PS CALC8-1 a a b a Ovary developmental stage d 500µm SC8-3 PS CALC8-1 13 Theoretical and Applied Genetics ◂Fig. 2  Developmental analysis and q-PCR of SC8-3, PS and (Supplementary Fig. S7). Interestingly, two independent CALC8-1 ovaries. a Ovaries from the three genotypes at anthesis. b replicates of CALC (MS-884 and MS-1155 in Zhao et al. Ovary length, width (both in µm) and shape index for the three geno- types through ovary development stages: E1 (carpel primordial initi- 2019 coding) and MR-1 (ME-1018, a modern cultivar with ated), E2 (primordial stigmas appear), E3 (stigma and style defined CALC in its pedigree, Thomas 1986) showed the CALC8-1 and anthers with pollen) and E4 (ovule primordial initiated). Aster- read pattern, confirming the presence of the deletion in that isks (*) indicates means of CALC8-1 or SC8-3 ovaries statistically genomic region. Most of the accessions showing the deletion significant different (p < 0.05) from PS means c Accumulation of MELOC025206 transcripts estimated by qRT-PCR between geno- read pattern belonged to India or had an Indian origin in the types and developmental stages. Different letters indicate statistically pedigree (MR-1). These results suggest that the origin of this significant differences among genotypes within developmental stages structural variation occurred in an ancestral Indian acces- (p < 0.05). d Paraffin sections of representative ovaries from the three sion, and it has not been dispersed with the diversification genotypes at E4 of melon cultivars, except its introduction into the MR-1 cul- tivar. Even though we did not have access to seeds of these distributed geographically, nicely reflecting the history of accessions to evaluate their fruit, the Germplasm Resources melon diversification, with a primary center of diversifica- Information Network (https://​www.​ars-​grin.​gov/) provides tion in India, a divergent diversification to the Western and fruit evaluation data for nine of them. The FSI of those nine Eastern regions and certain admixture due to migrations accessions ranged from round to elongated (Table 1), again and breeding (Gonzalo et al. 2019). European cultivars suggesting the non-direct association of the genomic dele- have a highly diverse fruit morphology, and in fact, they tions with the variation in FSI. In fact, CALC, the accession have been classified in several horticultural groups based donor of fsqs8.1 round allele, produces elongated fruit (Diaz on this diversity (Pitrat 2017). However, a very low haplo- et al. 2014). types diversity has been found among them. Therefore, the nucleotide variation at CMOFP13 coding sequence locus Functional validation of CmOFP13 in Arabidopsis could not be related, at least directly, to the fruit morphol- thaliana ogy variation in melon germplasm. To further investigate the possible association between Thus, to verify whether CmOFP13 is indeed a functional fsqs8.1 locus variability with FSI, 63 accessions from the homolog of AtOFP1 and SlOFP20, playing a role in the con- COMAV’s germplasm bank representing diverse worldwide trol of ovary and fruit shape, we over-expressed the allelic melon accessions were evaluated in two different experi- variant from CALC8-1 in A. thaliana, under the control of ments. FSI ranged from 0.8 (round fruit) to 4.5 (highly elon- the 35S constitutive promoter from the plant pathogen Cau- gated) in the 2014 and 2018 experiments, respectively (Sup- liflower Mosaic Virus (CaMV). Four transgenic lines were plementary Fig. S6). Accessions showing very different FSI finally selected based on single locus segregation in hygro- among experiments were removed. Accessions were geno- mycin (Fig. 5A). All four lines showed a similar phenotype, typed with SNPs located in the fsqs8.1 locus, but no SNP although they had differences in penetration, with two mild was significantly associated with FSI. Eight haplotypes were (L1 and L8) and two strong (L6 and L11) phenotypes. All found. However, B, D, E and G haplotypes presented four transgenic seedlings showed cotyledons with shorter peti- or less members, so they were removed for haplotype-based oles, and a protuberant proximal blade, which modified the association analysis, remaining A, C, F, PS haplotypes. FSI normal oval shape of the Col0 accession to a kidney-shaped mean for haplotype C (FSI = 1.04, 1.14 in 2014 and 2018 cotyledon. This phenotype continued during development, experiments, respectively) was significantly lower (rounder with round and curled leaves, resulting in compact plant fruit) than the A and PS haplotypes means (ranging between rosettes. More importantly, the overexpression of CmOFP13 1.37 and 1.69). Most accessions with the haplotype C belong provoked a shortening of flower whorls, including pistils that to the horticultural class cantalupensis, which characteristi- resulted in shorter fruits. Pistil morphology was measured cally produce round fruit (Pitrat 2017). Therefore, the asso- in the strong L6 line (Fig. 5b). Pistil length was reduced to ciation was not conclusive, as it could be due to the genetic 75%, whereas pistil width was increased by 43% of the Col-0 structure of melon cultivars. wild type. These changes resulted in a 50% reduction of The presence of the deletions observed in the CALC the ratio length to width. Overall, these phenotypes strongly and CALC8-1 genomes was also assessed in the 1,175 re- resembled those described upon overexpression of Class 1 sequenced accessions. Raw sequence reads of 25 acces- OVATE genes from A. thaliana and tomato (Hackbusch sions (Supplementary Table S4) showing missing data for et al. 2005; Wang et al. 2007, 2011; Wu et al. 2018). There- variants between chromosome 8 position 26,900,00 and fore, a higher expression of CmOFP13 caused reduced ovary 26,942,000 (Fig. 1) were retrieved and mapped de novo to and fruit length in transgenic A. thaliana plants, which in the CM4.0 melon genome. All mappings showed the same agreement with the relationship found between the variation pattern as previously observed with CALC and CALC8-1 in expression of endogenous CmOFP13 and the variation 13 Theoretical and Applied Genetics Fig. 3  Phylogeny of Arabidop- sis thaliana, tomato and melon ovate family protein genes based on Poisson distances (Zuckerkandl and Pauling 1965). Bootstrap values were based on 2000 replicates. Evolutionary analyses were conducted in MEGA X (Kumar et al. 2018). In order to simplify the figure, the nodes that do not include informative members for the current report are col- lapsed. The full phylogenetic tree is depicted in Fig. S2 a CmOFP13 haplotyes b Haplotype Posion in chromosome 8 World Region A B C D Total America 4 45 16 65 26,430,253 26,430,346 26,430,436 Central/South_Africa 10 4 12 26 haplotype China/Indochina 321 235 40 596 Europe 1 34 4 1 40 Far_East 21 20 8 49 Former Soviet Union 2 24 3 29 India/Maldives 16 31 72 1 120 A C A A Middle_East 3 38 11 52 B C G G Near_East 1 21 9 31 C C G A North_Africa 1 11 4 16 D T G A Total 380 463 175 6 1024 Fig. 4  Haplotypes found for CmOFP13 among 1175 re-sequenced 2020). CALC8-1 shares the same haplotype as PI 124,112 (c), while melon accessions (Zhao et al. 2019). a SNPs defining the four hap- PS shows the B haplotype. b Distribution of the four haplotypes lotypes and their position on the melon genome v4.0 (Castanera et al. through accessions belonging to different world regions 13 Theoretical and Applied Genetics Table 1  Characteristics of accessions showing the large deletion in fsqs8.1 locus according GRIN-USDA passport data. a Country of origin USA India Japan Israel Turkey Iran Number of accessions 1 19 1 1 1 1 b Fruit shape Round Oval Long Number of accessions 2 3 4 1a indicates the number of accessions that belong from different countries. 1b depicts the number of accessions producing different fruit mor- phology a Seedling b Rosee Flower Pisl Silique Fig. 5  Alteration of organ shape due to overexpression of CmOFP13 siliques. Scale bars are depicted in each stage. b Comparison between in Arabidopsis thaliana. a Images of A. thaliana wild-type plants ovary length, width and the length/width ratio between Col-0 and (Col0) and transgenic plants expressing CmOFP13 from CALC8-1 L6 transgenic line. (*) Indicates statistically significant differences showing a mild (L1 and L8) or strong (L6 and L11) phenotype at the (p < 0.001) 9-day-old seedling stage, adult rosette, flowers, pistils, and mature in fruit shape in CALC8-1, PS and SC8-3. We interpreted mid-value from SC8-3 and CALC8-1 was statistically that CmOFP13 was the functional homolog of AtOFP1 and significant, with the hybrid being rounder than expected SlOFP20, controlling ovary elongation in melon. (p = 0.002, Fig. 6a). Therefore, the CALC allele had a partial dominant interaction over the SC allele. Genetic effects of fsqs8.1 We next studied the epistatic interaction between the a gene, which controls sex expression (Boualem et al. 2008) Three different types of allelic interactions for fsqs8.1 were and whether it had pleiotropic effects on FSI (Perin et al. investigated: dominance between CALC and SC alleles, 2002) with fsqs8.1 (Diaz et al. 2014). The interaction was epistatic interaction with the a gene and interaction with evaluated in an ­F2 between CALC8-1 and ‘Mucha Nesvi’ different genetic background. For the dominance analysis, (MN), a monoecious ameri cultivar that produces elongated CALC and SC alleles were confronted in the hybrid CALC8- fruit. Homozygous plants for the four genotypic combina- 1-x-SC8-3. The gene action for FSI (d/[a]) was 0.43, and the tions at these two loci were selected at seedling stage with contrast of FSI means between the hybrid and the expected AluCAPS (Boualamen et al. 2008) and fsqs8.1 SNPs. The 13 Theoretical and Applied Genetics Fig. 6  Interactions of fsqs8.1 a in different genetic situations. a Comparison of CALC8-1, 1.8 SC8-3 and CALC8-1-x-SC8-3 hybrid fruit shape indexes. The Fruit Shape Index blue bar indicates the expected 1.5 mid-parent mean, and the asterisk the significance of the contrast (p = 0.0002). b Interac- 1.2 tion between a and fsqs8.1 in an F ­ 2 population from the cross between CALC8-1 and ‘Mucha Nesvi’ (MN, a monoe- 0.9 cious traditional cultivar). The means for the four homozygous allele combinations for the two loci (CALC/MN) for fsqs8.1, b ANDRO/MONO for andromo- noecious/monoecious (a), 2.0 are shown on the left and the significance of the gene effects Fruit Shape Index and interactions according two-way ANOVA on the right. 1.6 c The fruit shape index of fruits from hybrids between acces- sions belonging to different 1.2 melon horticultural groups with PS and CALC8-1 is depicted on the left. On the right, the 0.8 significance of the effects (GB, genetic background, fsqs8.1 PS or CALC8-1 allele at the locus, and the two years) and their c interactions from the three-way ANOVA, are shown 5 Fruit Shape Index 4 3 2 1 Cantalupensis Dudaim Flexuosus Inodorus Trigonus Vendrantais effects of both genes on FSI were highly significant, with years (Fig. 6c). Hybrids with PS ranged from slightly CALC fsqs8.1 allele inducing a round fruit and, and the round (FSI = 1.2) to very elongated (FSI = 4.2). A similar MN a allele elongated ones (Fig. 6b). On the other hand, trend was observed in the CALC8-1 hybrids, although in the interaction fsqs8.1-x-a was statistically significant, con- all cases fruit were rounder compared with their respec- firming the previous results by Diaz et al. (2014). All the tive PS hybrids. The effects of genetic background (GB) plants homozygous for the CALC fsqs8.1 allele produced and fsqs8.1 were highly significant confirming the consist- round fruit independently of their sex expression, whereas ency of fsqs8.1 effects on diverse GBs. The interaction the effect of a on FSI was only evident in plants with the GB*fsqs8.1 was also significant, due to the variability in MN alleles at fsqs8.1. the magnitude of the effect across GBs, although fsqs8.1 Finally, the interaction of the fsqs8.1 alleles with dif- induced rounder fruit in all cases. The larger effect was ferent genetic background was studied by combining observed mostly in hybrids that included accessions that CALC8-1 with six melon accessions with contrasting fruit produced elongated fruit. It is likely that there is a limita- morphology belonging to different horticultural groups tion of the roundness of the melon fruit, so the effect of (cantalupensis, charentais, dudaim, flexuosus, inodorus fsqs8.1 may be stronger in GBs that produce elongated and wild melon), and the hybrids were evaluated in two melons. 13 Theoretical and Applied Genetics Discussion expression of the gene in CALC8-1, compared with PS, resulting in rounder fruit in CALC8-1. The role of gene In the current report, we combined high-resolution map- expression variation in tomato fruit morphology was ping together with expression and functional analysis already pointed out for fw2.2, the first cloned fruit mor- to demonstrate that CmOFP13, a member of the melon phology QTL (Cong et al. 2002) and it has been observed OVATE gene family, is the causal gene underlying the in subsequent cloned morphology genes and QTLs: SUN melon fruit shape QTL fsqs8.1. The discovery of the (Xiao et al. 2008), lc (Munos et al. 2011), fas (Xu et al. OVATE gene family started with the study of the genetic 2015) and fw3.2 (Chakrabarti et al. 2013), among others. control of the pear-shaped fruit in the tomato heirloom Therefore, fsqs8.1 is another example of this conserved variety Yellow Pear (Ku et al. 1999). The pear shape of mechanism that controls organ morphology. tomatoes was already studied by early tomato geneticists The fsqs8.1 locus showed an important structural varia- (Hendrick and Booth 1907; Price and Drinkard 1908) tion, with two deletions surrounding CmOFP13. The dele- demonstrating that the pear shape was controlled by a sin- tions were confirmed in the re-sequencings of CALC8-1, gle gene, which was named ovate (Lindstrom 1926, 1927). two independent CALCs (the donor of fsqs8.1 introgres- Liu et al. (2002) identified and cloned the first OVATE sion in CALC8-1), and a small set of melon accession gene, which did not match with any known proteins at genomes. This structural variation may have affected regu- that time, thus the authors suggested the discovery of a latory sequences, which could explain the changes in the new class of regulatory genes involved in plant growth. CmOFP13 expression patterns observed between genotypes Further research demonstrated that OVATE was a member in the current report. In tomato, hundreds of structural varia- of a large gene family present in different plant species tions have been associated to changes in gene expression pat- (Liu et al. 2014, Dangwai and Das 2018). Thus, Monforte terns (Alonge et al. 2020). Concomitantly, in the same spe- et al. (2014) reported 21 OFP members of the OVATE cies, Eshed and Lippman (2019) pointed out that mutations gene family in the melon genome. The large number of in 5’ and 3’ cis regulatory elements can alter gene expres- members and the co-segregation of their position with sion, resulting in alleles with quantitative effects. In fact, fruit shape QTLs led to the hypothesis that OFP members both SlOFP20 and PpOFP1 genomic regions present struc- other than OVATE could also be involved in the control of tural variation (Wu et al. 2018, Zhou et al. 2020). In the case organ shape in different species (Rodriguez et al. 2011; of fsqs8.1, the larger deletion region is rich in transposon- Monforte et al. 2014). related sequences in the reference genome, thus it is likely Phylogenetic analysis showed that CmOFP13 is related that transposon activity has been relatively frequent in this to tomato SlOFP14, SlOF17 and SlOFP20 and the A. thal- region, and that their activity could have been responsible iana AtOFP1, but not to OVATE itself. Wu et al. (2018) for the observed structural variation. The impact of transpo- demonstrated that SlOFP20 controlled the morphology of sons on the generation of phenotypic diversity underlies the tomato fruit and potato tubers. More recently, the peach maize domestication gene tb1 (Zhou et al. 2011). Insertions member PpOFP1, also likely homologue to SlOFP20, was of retrotransposons in upstream regions of tb1 causes a dif- found to control the flat fruit shape in this crop (Zhou et al. ferential expression that is responsible for the typical single 2020). Evidences of the role of SlOFP20 homologues in branching of maize cultivars in contrast with the multiple fruit morphology have also been found in cucumber (Colle branching of the wild ancestor teosinte. Recently, the role of et al. 2017) and pepper (Tsaballa et al. 2011). Previously, transposon activity has garnered more attention. Domínguez Hackbusch et al. (2005) showed that the overexpression of et al. (2020) found associations of transposon element AtOFP1 induced shortened siliques in transgenic A. thali- insertion polymorphisms with major agronomic traits and ana plants. Even though the role of other OFP members secondary metabolites in cultivated tomato. Similarly, evi- on organ morphology cannot be ruled out (Snouffer et al. dence of the role of retrotransposons on the evolution of the 2020), these results suggest strongly that the homologues genome and modification of gene expression has also been of SlOFP20 are crucial for the establishment of organ reported in melon (Demirci et al. 2021; Sanseverino et al. shape in a large number of plant species. 2015; Yano et al. 2020). Thus, fsqs8.1 is perhaps another In tomato and peach, the effect of SlOFP20 and case of gene expression changes due to structural variation PpOFP1 is a consequence of changes in gene expression, mediated by transposon activity. not changes in protein function (Wu et al. 2018; Zhou Strikingly, the CmOFP13 coding sequence showed a very et al. 2020). Similarly, no amino acid changes were found low variation across melon germplasm, with all mutations between the CALC and PS CmOFP13 protein sequences. being silent, what suggested that the functionality of this Therefore, the differences in fruit morphology between protein is needed for proper melon development and only CALC8-1 and PS are likely caused by the early higher changes in expression patterns might be allowed. Attempts at finding an association between haplotypes in fsqs8.1 regions 13 Theoretical and Applied Genetics and fruit shape did not provide conclusive results. Previ- due to a structural variation mediated by transposon activ- ously, Gonzalo and van der Knaap (2008) reported that the ity in the surrounding genomic region, and perhaps by pro- effect of OVATE was dependent on the genetic background moting cell elongation by microtubule orientation (Zhang in tomato. Rodriguez et al. (2013) showed that SlOFP20 et al. 2020). The effects of CmOFP13 depend on the genetic was a suppressor of OVATE, and that specific allele com- background, being evident in the PS background, but not in binations between SlOFP20 and OVATE should coincide the original Indian background. Even though no associa- to produce the typical ovate elongated fruit. Moreover, Wu tions were found between SNPs or structural variation in et al. (2018) demonstrated that slTRM5 (a member of TON- the fsqs8.1 locus with fruit morphology in melon germplasm NEAU1 Recruiting Motif family interacts with SlOFP20 in this work, fsqs8.1 genomic region has been associated and can modulate the effects on tomato fruit shape. In fact, in fruit morphology in a wide range of melon germplasm Gur et al. (2017) found association between melon fruit (Monforte et al. 2014; Gur et al. 2017; Pan et al. 2020), shape and SNPs located in the fsqs8.1 region using a differ- actually the region is considered to include a consensus FSI ent GWAS panel. Therefore, the effects of OFP genes may QTL (Monforte et al. 2014; Pan et al. 2020). Other muta- depend on the genetic background, and the fact that the OFP tions should be involved in those alleles, likely regulatory family includes a large number of members may cause a mutations rather than coding mutations, or additional linked buffering effect, thereby limiting the effects of a single OFP genes. Therefore, the CALC CmOFP13 allele is a cryptic gene that would be apparent only in certain allelic combina- variant that may have arisen by a structural variation, with a tions. Based on the deletion pattern on the fsqs8.1 locus, the strong potential to become a new breeding tool for generat- CALC allele is present with low frequency in Indian germ- ing monoecious cultivars that produce round fruit. plasm, but it was not associated with fruit morphology in that germplasm. The CALC (and other accessions) genome Supplementary Information The online version contains supplemen- tary material available at https://d​ oi.o​ rg/1​ 0.1​ 007/s​ 00122-0​ 21-0​ 3998-6. may contain one or several suppressors of CmOFP13, or another gene/genes needed are not expressed at the proper Acknowledgements The authors thank Ms. Soledad Casal and Dr. time, resulting in elongated fruit. However, the introduction Gorka Perpiña for technical support, Dr. Javier Forment from the of the CALC CmOFP13 allele into the PS genetic back- IBMCP Bioinformatics Facility for bioinformatics support. This work ground induced the development of round fruit, suggesting was supported by European Union and Spanish Ministry of Economy and Competitiveness /FEDER grants AGL2015-64625-C2-1-R to AJM/ that the PS genome lacks the putative suppressors or the CR, by MCIN/AEI/10.13039/501100011033 and by “ERDF A way other needed gene/genes would be expressed at the proper of making Europe” grants AGL2017-85563-C2-1-R (BP), BIO2017- time. Similarly, the effects of the CALC CmOFP13 allele on 83138R (MAP-A) RTI2018-097665-B-C2 (AJM/CR), RTI2018- fruit shape were consistent when confronted with different 097665-B-C1 (JGM) and PROMETEO project 2017/078 and 2021/072 (to promote excellence groups) by the Conselleria d’Educació, Inves- genetic background. These experiments were also carried tigació, Cultura i Esports (Generalitat Valenciana) to BP. PS was sup- out in the hybrid PS-x-accessions genetic backgrounds (the ported by a scholarship from Bec.Ar of the National Ministry of Edu- genetic background of CALC8-1 is PS), so the putative sup- cation, Argentina. pressors might be suppressed or the other needed gene/genes would be expressed, and as consequence CALC CmOFP13 Author Contribution statement CM-M, MJG, PS, MC, IM-F, CL, MD-G, CR conducted experiments related with the fine mapping, RT- allele could induce its effects on fruit morphology. PCR and development analysis. MZ, KGA, JG-M and AJM obtained An interesting practical breeding results from this and analyzed genomic data. PT and MAP-A performed the A thaliana work is the repression of fruit elongation induced by a by experiments. CE and BP selected germplasm and performed green- CmOFP13. The development of monoecious cultivars pro- house trials. AJM conceived and designed the study. AJM, CR and MAP-A wrote the manuscript. All authors reviewed and accepted the ducing round fruits (characteristic of the common market manuscript. class cantalupensis, galia and honeydew) is desirable, as it eliminates the need for emasculation for crossing make Funding Open Access funding provided thanks to the CRUE-CSIC crosses to produce hybrid seeds. However, the effects of the agreement with Springer Nature. European Union and Spanish Ministry a gene on fruit elongation makes difficult the combination of of Economy and Competitiveness /FEDER grants AGL2015-64625- C2-1-R to AJM/CR, by MCIN/AEI/10.13039/501100011033 and by both properties. Therefore, CmOFP13 round alleles could be “ERDF A way of making Europe” grants AGL2017-85563-C2-1-R introduced in breeding programs to generate such cultivars. (BP), BIO2017-83138R (MAP-A) RTI2018-097665-B-C2 (AJM/CR), In conclusion, we have demonstrated that the gene under- RTI2018-097665-B-C1 (JGM) and PROMETEO project 2017/078 lying the fruit shape QTL fsqs8.1 is CmOFP13, a member of and 2021/072 (to promote excellence groups) by the Conselleria d’Educació, Investigació, Cultura i Esports (Generalitat Valenciana, OFP gene family, a likely homologue to the tomato SlOFP20 ES) to BP. PS was supported by a scholarship from Bec.Ar of the gene. The effects of CmOFP13 in ovary and fruit shape are National Ministry of Education, Argentina. most probably due to changes in its expression pattern, likely 13 Theoretical and Applied Genetics Availability of data and material Genome sequences from PS Cong B, Liu JP, Tanksley SD (2002) Natural alleles at a tomato fruit and CALC8-1 are available at NCBI repository BioProject ID size quantitative trait locus differ by heterochronic regulatory PRJNA729905. mutations. Proc Natl Acad Sci USA 99:13606–13611 Dangwal M, Das S (2018) Identification and analysis of OVATE family members from genome of the early land plants provide insights Declarations into evolutionary history of OFP family and function. J Mol Evol 86:511–530 Conflicts of interest Authors declare no conflict of interest. Demirci S, Fuentes RR, van Dooijeweert W, Aflitos S, Schijlen E, Hes- selink T, de Ridder D, van Dijk A D J, Peters S (2021). Chasing Open Access This article is licensed under a Creative Commons Attri- breeding footprints through structural variations in Cucumis melo bution 4.0 International License, which permits use, sharing, adapta- and wild relatives. G3 11: jkaa038 tion, distribution and reproduction in any medium or format, as long Diaz A, Zarouri B, Fergany M, Eduardo I, Alvarez JM, Pico B, Mon- as you give appropriate credit to the original author(s) and the source, forte AJ (2014) Mapping and introgression of QTL involved in provide a link to the Creative Commons licence, and indicate if changes fruit shape transgressive segregation into “Piel de Sapo” melon were made. The images or other third party material in this article are (Cucumis melo L.). PLoS ONE 9:e116098 included in the article's Creative Commons licence, unless indicated Dominguez M, Dugas E, Benchouaia M, Leduque B, Jimenez-Gomez otherwise in a credit line to the material. If material is not included in JM, Colot V, Quadrana L (2020) The impact of transposable ele- the article's Creative Commons licence and your intended use is not ments on tomato diversity. Nat Commun 11:4058 permitted by statutory regulation or exceeds the permitted use, you will Doyle JJ, Doyle JL (1990) Isolation of plant DNA from fresh tissue. need to obtain permission directly from the copyright holder. To view a Focus 12:13–15 copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. Eduardo I, Arus P, Monforte AJ (2005) Development of a genomic library of near isogenic lines (NILs) in melon (Cucumis melo L.) from the exotic accession PI161375. Theor Appl Genet 112:139–148 Eduardo I, Arus P, Monforte AJ, Obando J, Fernandez-Trujillo References JP, Martinez JA, Alarcon AL, Alvarez JM, van der Knaap E (2007) Estimating the genetic architecture of fruit quality traits Abdelmohsin ME, Pitrat M (2008) Pleiotropic effect of sex expression in melon using a genomic library of near isogenic lines. J Am on fruit shape in melon. In: Pitrat M (ed) 9th EUCARPIA Meet- Soc Hortic Sci 132:80–89 ing on Genetics and Breeding of Cucurbitaceae, INRA, Avignon, Endl J, Achigan-Dako EG, Pandey AK, Monforte AJ, Pico B, France, pp 551–555 Schaefer H (2018) Repeated domestication of melon (Cucumis Alonge M, Wang XG, Benoit M, Soyk S, Pereira L, Zhang L, Suresh H, melo) in Africa and Asia and a new close relative from India. Ramakrishnan S, Maumus F, Ciren D, Levy Y, Harel TH, Shalev- Am J Bot 105:1662–1671 Schlosser G, Amsellem Z, Razifard H, Caicedo AL, Tieman DM, Eshed Y, Lippman ZB (2019) Revolutions in agriculture chart a Klee H, Kirsche M, Aganezov S, Ranallo-Benavidez TR, Lem- course for targeted breeding of old and new crops. Science mon ZH, Kim J, Robitaille G, Kramer M, Goodwin S, McCombie 366:705–706 WR, Hutton S, Van Eck J, Gillis J, Eshed Y, Sedlazeck FJ, van FAOSTAT (2021) https://​www.​fao.​org/​faost​at/​en/#​home der Knaap E, Schatz MC, Lippman ZB (2020) Major impacts Felsenstein J (1985) Confidence-limits on phylogenies–an approach of widespread structural variation on gene expression and crop using the bootstrap. Evolution 39:783–791 improvement in tomato. Cell 182:145–161 Gomez M, Beltran JP, Canas L (2004) The pea END1 promoter drives Bai SL , Peng YB, Cui JX, Gu HT, Xu LY, Li YQ, Xu ZH, Bai SN anther-specific gene expression in different plant species. Planta (2004). Developmental analyses reveal early arrests of the spore- 219:967–981 bearing parts of reproductive organs in unisexual flowers of Gomez MD, Urbez C, Perez-Amador MA, Carbonell J (2011) Char- cucumber (Cucumis sativus L.) Planta 220: 230–240 acterization of constricted fruit (ctf) mutant uncovers a role for Boualem A, Fergany M, Fernandez R, Troadec C, Martin A, Morin AtMYB117/LOF1 in ovule and fruit development in Arabidopsis H, Sari MA, Collin F, Flowers JM, Pitrat M, Purugganan MD, thaliana. PLoS ONE 6:e18760 Dogimont C, Bendahmane A (2008) A conserved mutation in an Gonzalo MJ, van der Knaap E (2008) A comparative analysis into the ethylene biosynthesis enzyme leads to andromonoecy in melons. genetic bases of morphology in tomato varieties exhibiting elon- Science 321:836–838 gated fruit shape. Theor Appl Genet 116:647–656 Castanera R, Ruggieri V, Pujol M, Garcia-Mas J, Casacuberta JM Gonzalo MJ, Diaz A, Dhillon NPS, Reddy UK, Pico B, Monforte AJ (2020) An improved melon reference genome with single-mole- (2019) Re-evaluation of the role of Indian germplasm as center of cule sequencing uncovers a recent burst of transposable elements melon diversification based on genotyping-by-sequencing analy- with potential impact on genes. Front Plant Sci 10:1815 sis. BMC Genomics 20:448 Chakrabarti M, Zhang N, Sauvage C, Munos S, Blanca J, Canizares Gur A, Tzuri G, Meir A, Sa’ar U, Portnoy V, Katzir N, Schaffer AA, J, Diez MJ, Schneider R, Mazourek M, McClead J, Causse M, Li L, Burger J, Tadmor Y (2017) Genome-wide linkage-disequi- van der Knaap E (2013) A cytochrome P450 regulates a domes- librium mapping to the candidate gene level in melon (Cucumis tication trait in cultivated tomato. Proc Natl Acad Sci USA melo). Sci Rep 7:9770 110:17125–17130 Hackbusch J, Richter K, Muller J, Salamini F, Uhrig JF (2005) A cen- Clough SJ, Bent AF (1998) Floral dip: a simplified method for Agro- tral role of Arabidopsis thaliana ovate family proteins in network- bacterium-mediated transformation of Arabidopsis thaliana. Plant ing and subcellular localization of 3-aa loop extension homeodo- J 16:735–743 main proteins. Proc Natl Acad Sci USA 102:4908–4912 Colle M, Weng YQ, Kang YY, Ophir R, Sherman A, Grumet R (2017) Hedrick UP, Booth NO (1907) Mendelian characters in tomato. Proc Variation in cucumber (Cucumis sativus L.) fruit size and shape Am Soci Hortic Sci 5:19–24 results from multiple components acting pre-anthesis and post- Ku HM, Doganlar S, Chen KY, Tanksley SD (1999) The genetic basis pollination. Planta 246:641–658 of pear-shaped tomato fruit. Theor Appl Genet 99:844–850 13 Theoretical and Applied Genetics Kumar S, Stecher G, Li M, Knyaz C, Tamura K (2018) MEGA X: of SUN, OVATE, LC, and FAS in the tomato germplasm and the MEGA X: molecular evolutionary genetics analysis across com- relationship to fruit shape diversity. Plant Physiol 156:275–285 puting platforms. Mol Biol Evol 35:1547–1549 Rodriguez GR, Kim HJ, van der Knaap E (2013) Mapping of two sup- Li S, Pan YP, Wen CL, Li YH, Liu XF, Zhang XL, Behera TK, Xing pressors of OVATE (sov) loci in tomato. Heredity 111:256–264 GM, Weng YQ (2016) Integrated analysis in bi-parental and Saitou N, Nei M (1987) On the maximum-likelihood method for natural populations reveals CsCLAVATA3 (CsCLV3) underly- molecular phylogeny. Jpn J Genet 62:547–548 ing carpel number variations in cucumber. Theor Appl Genet Sanseverino W, Henaff E, Vives C, Pinosio S, Burgos-Paz W, Mor- 129:1007–1022 gante M, Ramos-Onsins SE, Garcia-Mas J, Maria Casacuberta Li H (2013) Aligning sequence reads, clone sequences and assembly J (2015) Transposon insertions, structural variations, and SNPs contigs with BWA-MEM. http://arxiv.org/abs/arXiv: 1303.3997 contribute to the evolution of the melon genome. Mol Biol Evol Lindstrom EW (1926) Linked inheritance in tomatoes. Iowa State Coll 32:2760–2774 J Sci 1:3–13 SM Smith PJ Maughan 2015 SNP genotyping using KASPar assays Lindstrom EW (1927) The inheritance of ovate and related shapes of Plant Genotyping: methods Protocols 1245 243 256 tomato fruit. J Agric Res 34:961–985 Snouffer A, Kraus C, van der Knaap E (2020) The shape of things to Liu JP, Van Eck J, Cong B, Tanksley SD (2002) A new class of regula- come: ovate family proteins regulate plant organ shape. Curr Opin tory genes underlying the cause of pear-shaped tomato fruit. Proc Plant Biol 53:98–105 Natl Acad Sci USA 99:13302–13306 Thomas CE (1986) Downy and powdery mildew resistant muskmelon Liu D, Sun W, Yuan YW, Zhang N, Hayward A, Liu YL, Wang Y breeding line MR-1. HortScience 21:329–329 (2014) Phylogenetic analyses provide the first insights into the Tsaballa A, Pasentsis K, Darzentas N, Tsaftaris AS (2011) Multiple evolution of OVATE family proteins in land plants. Ann Bot evidence for the role of an ovate-like gene in determining fruit 113:1219–1233 shape in pepper. BMC Plant Biol 11:46 Martin M (2011) Cutadapt removes adapter sequences from high- Untergasser A, Cutcutache I, Koressaar T, Ye J, Faircloth BC, Remm throughput sequencing reads. Embnet J 17:10–12 M, Rozen SG (2012) Primer3-new capabilities and interfaces. Monforte AJ, Oliver M, Gonzalo MJ, Alvarez JM, Dolcet-Sanjuan R, Nucleic Acids Res 40:e115 Arus P (2004) Identification of quantitative trait loci involved in Wang S, Chang Y, Guo J, Chen JG (2007) Arabidopsis ovate family fruit quality traits in melon (Cucumis melo L.). Theor Appl Genet protein 1 is a transcriptional repressor that suppresses cell elonga- 108:750–758 tion. Plant J 50:858–872 Monforte AJ, Eduardo I, Abad S, Arus P (2005) Inheritance mode of Wang SC, Chang Y, Guo JJ, Zeng QN, Ellis BE, Chen JG (2011) Arabi- fruit traits in melon: heterosis for fruit shape and its correlation dopsis ovate family proteins, a novel transcriptional repressor fam- with genetic distance. Euphytica 144:31–38 ily, control multiple aspects of plant growth and development. Monforte AJ, Diaz A, Cano-Delgado A, van der Knaap E (2014) The PLoS ONE 6:e23896 genetic basis of fruit morphology in horticultural crops: lessons Weisenfeld NI, Kumar V, Shah P, Church DM, Jaffe DB (2017) from tomato and melon. J Exp Bot 65:4625–4637 Direct determination of diploid genome sequences. Genome Res Munos S, Ranc N, Botton E, Berard A, Rolland S, Duffe P, Carretero 27:757–767 Y, Le Paslier MC, Delalande C, Bouzayen M, Brunel D, Causse Wittwer CT, Reed GH, Gundry CN, Vandersteen JG, Pryor RJ (2003) M (2011) Increase in tomato locule number is controlled by two High-resolution genotyping by amplicon melting analysis using single-nucleotide polymorphisms located near WUSCHEL. Plant LCGreen. Clin Chem 49:853–860 Physiol 156:2244–2254 Wu S, Zhang BY, Keyhaninejad N, Rodriguez GR, Kim HJ, Chakra- Murashige T, Skoog F (1962) A revised medium for rapid growth and barti M, Illa-Berenguer E, Taitano NK, Gonzalo MJ, Diaz A, Pan bio assays with tobacco tissue cultures. Physiol Plant 15:473–497 YP, Leisner CP, Halterman D, Buell CR, Weng YQ, Jansky SH, Nakagawa T, Kurose T, Hino T, Tanaka K, Kawamukai M, Niwa Y, van Eck H, Willemsen J, Monforte AJ, Meulia T, van der Knaap Toyooka K, Matsuoka K, Jinbo T, Kimura T (2007) Development E (2018) A common genetic mechanism underlies morphological of series of gateway binary vectors, pGWBs, for realizing efficient diversity in fruits and other plant organs. Nat Commun 9:4734 construction of fusion genes for plant transformation. J Biosci Xiao H, Jiang N, Schaffner E, Stockinger EJ, van der Knaap E (2008) A Bioeng 104:34–41 retrotransposon-mediated gene duplication underlies morphologi- Oliver M, Garcia-Mas J, Cardus M, Pueyo N, Lopez-Sese A, Arroyo cal variation of tomato fruit. Science 319:1527–1530 M, Gomez-Paniagua H, Arus P, de Vicente MC (2001) Construc- Xu C, Liberatore KL, MacAlister CA, Huang ZJ, Chu YH, Jiang K, tion of a reference linkage map for melon. Genome 44:836–845 Brooks C, Ogawa-Ohnishi M, Xiong GY, Pauly M, Van Eck J, Pan YP, Wang YH, McGregor C, Liu S, Luan FS, Gao ML, Weng YQ Matsubayashi Y, van der Knaap E, Lippman ZB (2015) A cascade (2020) Genetic architecture of fruit size and shape variation in of arabinosyltransferases controls shoot meristem size in tomato. cucurbits: a comparative perspective. Theor Appl Genet 133:1–21 Nat Genet 47:784–792 Perin C, Hagen LS, Giovinazzo N, Besombes D, Dogimont C, Pitrat Yano R, Ariizumi T, Nonaka S, Kawazu Y, Zhong SL, Mueller L, M (2002) Genetic control of fruit shape acts prior to anthesis in Giovannoni JJ, Rose JKC, Ezura H (2020) Comparative genomics melon (Cucumis melo L.). Mol Genet Genom 266:933–941 of muskmelon reveals a potential role for retrotransposons in the Pitrat M (2017) Melon genetic resources: phenotypic diversity and modification of gene expression. Commun Biol 3:432 horticultural taxonomy. In: Grumet R, Katzi N, Garcia-Mas J (eds) Zhang XW, Wu JL, Yu Q, Liu RY, Wang ZY, Sun Y (2020) AtOFPs Genetics and genomics of the Cucurbitaceae. Springer, New York, regulate cell elongation by modulating microtubule orientation pp 25–60 via direct interaction with TONNEAU2. Plant Sci 292:110405 Price HC, Drinkard AW (1908) Inheritance in tomato hybrids. Va Zhao GW, Lian Q, Zhang ZH, Fu QS, He YH, Ma S, Ruggieri V, Agric Exp Station Bull 177:17–53 Monforte AJ, Wang PY, Julca I, Wang HS, Liu JP, Xu Y, Wang Robinson JT, Thorvaldsdottir H, Winckler W, Guttman M, Lander ES, RZ, Ji JB, Xu ZH, Kong WH, Zhong Y, Shang JL, Pereira L, Getz G, Mesirov JP (2011) Integrative genomics viewer. Nat Bio- Argyris J, Zhang J, Mayobre C, Pujol M, Oren E, Out D, Wang technol 29:24–26 JM, Sun DX, Zhao SJ, Zhu YC, Li N, Katzir N, Gur A, Dogi- Rodriguez GR, Munos S, Anderson C, Sim SC, Michel A, Causse M, mont C, Schaefer H, Fan W, Bendahmane A, Fei Z, Pitrat M, Gardener BBM, Francis D, van der Knaap E (2011) Distribution Gabaldon T, Lin T, Garcia-Mas J, Xu YY, Huang SW (2019) A comprehensive genome variation map of melon identifies multiple 13 Theoretical and Applied Genetics domestication events and loci influencing agronomic traits. Nat Q, Korban SS, Han YP (2020) A 1.7-Mb chromosomal inversion Genet 51:1607–1615 downstream of aPpOFP1gene is responsible for flat fruit shape in Zhou LL, Zhang JY, Yan JB, Song RT (2011) Two transposable ele- peach. Plant Biotechnol J 19:192–205 ment insertions are causative mutations for the major domes- Zuckerkandl E, Pauling L (1965) Molecules as documents of evolution- tication gene teosinte branched 1 in modern maize. Cell Res ary history. J Theor Biol 8:357–366 21:1267–1270 Zhou H, Ma RJ, Gao L, Zhang JN, Zhang AD, Zhang XJ, Ren F, Zhang Publisher's Note Springer Nature remains neutral with regard to WH, Liao L, Yang QR, Xu SL, Ogutu CO, Zhao JB, Yu ML, Jiang jurisdictional claims in published maps and institutional affiliations. 13