Introduction

Melon (Cucumis melo L.,) is one of the most highly diversified species in the Cucurbitaceae family which also includes other important crops such as cucumber, watermelon and squash/pumpkin. It shows a high level of morphological variability in fruit characteristics such as shape, color, texture, taste and composition. Melon is divided in two major subspecies: subsp. melo and subsp. agrestis (Kirkbride 1993). The genome of melon has been sequenced in 2012 using NGS sequencing technology and a double haploid line DHL 92, and the genome assembly was further improved using high-density SNP markers (Garcia-Mas et al. 2012; Argyris et al. 2015). With the rapid development of genetic and genomic resources, melon has become a model of choice for studying several important biological processes such as plant sex determination (Boualem et al. 2008, 2015; Martin et al. 2009), and phloem transport (Zhang et al. 2010). Furthermore, several important genes controlling fruit quality in melon have been cloned such as CmKFB gene regulating flavonoid accumulation (Feder et al. 2015) and CmPH gene determining fruit acidity (Cohen et al. 2014; Sherman et al. 2013).

Trichomes are hair-like structures which can protect plants against the damage by insects, pathogens, UV irradiation, low temperature and excessive transpiration (Hulskamp 2004; Hulskamp et al. 1999; Schellmann and Hulskamp 2005). In addition, trichomes may help plants attract pollinators and disperse seeds (Serna and Martin 2006). Trichomes exhibit high morphological variation, and can be divided into several forms, which could be unicellular or multicellular, glandular or glandless, and branched or unbranched (Werker 2000). As the representative of unicellular trichomes, the regulatory pathway of trichomes has been widely studied in the model plant Arabidopsis thaliana (Hulskamp 2004; Hulskamp et al. 1999; Schellmann and Hulskamp 2005). The initiation of trichome differentiation is regulated by GL2 (GLABRA2), which belongs to the class IV homeodomain-leucine zipper (HD-ZIP IV) family (Ishida et al. 2008; Pesch and Hulskamp 2009; Schiefelbein 2003). The GL2 was activated by an upstream MYB–bHLH–WD40 trimeric complex which is comprised by three transcription factors, GL1 (GLABRA1), which encodes an R2R3 MYB protein (Kirik et al. 2005), GL3 (GLABRA3) and its homolog EGL3 (ENHANCER OF GLABRA3), which are basic helix–loop–helix proteins (Payne et al. 2000), and TTG1 (TRANSPARENT TESTAGLABRA1), a WD40-repeat protein (Walker et al. 1999). TRY (TRIP-TYCHON), CPC (CAPRICE), and ETC (ENHANCER OF TRY AND CPC) are negative regulators, which all belong to the R3 single-repeat MYB family. They can compete with the R2R3 MYB proteins to bind the bHLH protein which will form an inactivating complex and thereby inhibit the initiation of trichome differentiation (Kirik et al. 2004; Wester et al. 2009).

The regulatory mechanisms of multicellular trichome formation are still largely unknown. The MIXTA, an R2R3-type MYB transcription factor encoding the conserved DNA-binding domain similar to GL1, could trigger long-stalked multicellular trichome formation when ectopically expressed in tobacco, while the ectopic expression of MIXTA in Arabidopsis gl1-1 mutant failed to restore the trichome phenotype (Payne et al. 1999). Moreover, over-expression of GL1 in tobacco had no effect on trichome formation, suggesting that the developments of unicellular and multicellular trichomes are likely controlled by different molecular mechanisms. This is further confirmed by the Wo gene which was identified in tomato encoding a HD-ZIP IV transcription factor and could regulate the formation of type I multicellular trichomes. Overexpression of Wo gene had no effect on trichome initiation in Arabidopsis (Yang et al. 2011). These data demonstrate that the initiation of multicellular trichomes in tobacco and tomato and unicellular trichomes in Arabidopsis may be regulated by different pathways.

Trichomes are commonly found in melon plants covering the leaves, stems, branches, flowers, and tendrils, and the trichomes show multicellular, unbranched, glandless which are different from the unicellular trichomes in A. thaliana. There is still little study on the molecular mechanisms of trichome development in melon, but several spontaneous glabrous mutants have been characterized and studied in its close relative cucumber. The first mutant csgl1 shows no observable trichomes on leaves, stems, tendrils, and floral organs, but has obvious trichomes on the hypocotyl. Map-based cloning has revealed that CsGL1 encodes a Class I HD-ZIP TF, and the loss-function of csgl1 is due to a 2649-bp genomic DNA deletion spanning the first and second exons (Chen et al. 2014; Zhao et al. 2015). The csgl3/tril mutant exhibits a completely glabrous phenotype on cotyledons, hypocotyl, young leaves, fruits, and fruit stalks. It has been mapped on cucumber chromosome 6 and encodes a class IV HD-ZIP TF which lost its function due to the insertion of a 5-kb long terminal repeat (LTR) retrotransposon in the fourth exon of CsGL3 (Cui et al. 2016; Pan et al. 2015; Wang et al. 2016). In melon, a completely glabrous mutation was first reported by Foster (1963) which can be recognized readily making it an excellent seedling marker in breeding. Genetic analysis indicated that glabrous (gl) is controlled by a single recessive gene. However, the gl gene is still not mapped and the nucleotide sequence information for this gene remains unknown. Thus, cloning and characterizing the gl gene are necessary for understanding the mechanism of multicellular trichome formation in melon.

In this study, a completely glabrous muskmelon line NSL73046 was used to cross with a wild-type line M465 for mapping of the glabrous (cmgl) gene. Based on the resequencing of the two parental lines, Indel markers were developed in the candidate region which were then used in genotyping a large F2 mapping population. A candidate gene for cmgl was identified encoding a HD-ZIP IV transcription factor, and the loss-of-function in the mutant was due to a single base substitution resulting in a premature termination codon.

Materials and methods

Plant materials and mapping population

NSL73046 (the mutant type) is a muskmelon mutant from USDA-ARS GRIN. It displayed completely glabrous on all aboveground organs including leaves, stems, flowers, and a permanent yellow-green leaf color throughout its lifespan which have been proved to be controlled by two different recessive genes, glabrous and yellow-green plant (Foster 1963). To confirm the inheritance mode and fine mapping of cmgl gene in melon, NSL73046 was used to cross with a normal muskmelon material M465 (the wild type) to generate a large F2 mapping population. The χ 2-test for goodness-of-fit was used to test for deviation of the observed data from the theoretically expected segregation for trichome phenotype data in F2 plants.

To investigate the allelic diversity of the cmgl candidate gene in natural populations, the trichome phenotypes of a panel including 564 melon accessions with diverse genetic backgrounds were examined using CmGL co-segregating dCAPS marker. Information for these melon materials was provided in Supplementary Table S1. All the melon lines and F2 population were grown in the greenhouse at Maozhuang Research Station of Henan Agricultural University (Zhengzhou, China). Unexpanded young leaves from these plants were collected into 1.5 mL microcentrifuge tubes, lyophilized in a freeze dryer, and ground into fine powder. Genomic DNA was extracted using the CTAB method (Murray and Thompson 1980).

Scanning electron microscopy

The samples of leaves, stems and hypocotyls from NSL73046 and M465 were fixed in FAA (formaldehyde acetic acid–ethanol which contained 50% (v/v) ethanol, 5% (v/v) acetic acid and 3.7% (v/v) formaldehyde dissolved in water) at 4 °C for 24 h, then dehydrated through gradient ethanol elution [50, 60, 70, 85, 90, 95 and 100% (v/v)] and critical point dried in a Leica EM CPD030 desiccator. The dried specimens were sputter-coated with gold–palladium and observed under JSM-6360LV and JEM-2010HT scanning electron microscopes.

Marker development, genetic mapping strategy and linkage analysis

A BSA strategy was used to screen for polymorphic SSR markers between two DNA pools selected from the F2 population for quick identification of molecular markers linked with the cmgl locus. The DNA pools of mutant and normal plants consisted of ten glabrous plants and 10 non-glabrous plants, respectively. 384 SSR markers evenly distributed in 12 melon chromosomes (Zhu et al. 2016) were selected to screen for polymorphisms. After initial anchor of cmgl in chromosome 8, a scaffold-based chromosome walking strategy was taken to identify more closely linked markers. New SSR markers in the target scaffolds were selected for primary mapping. At fine mapping stage, SNPs and Indels were explored based on the sequence difference by comparing the genome resequencing data between NSL73046 and M465. For Indels, only those with ≥ 3 bp differences were utilized for primer design with Primer3 (http://primer3.ut.ee/). For SNP genotyping, dCAPS markers were developed with dCAPS Finder 2.0 (Neff et al. 1998).

The PCR amplification of molecular markers and gel electrophoresis was conducted as described in Zhu et al. (2016). Linkage analysis of the cmgl locus with molecular markers was performed with the Kosambi mapping function using JoinMap 3.0 with the threshold LOD score of 4.0.

DNA sequencing, sequence comparison and gene prediction

Genome resequencing of NSL73046 and M465 was carried out following the standard Illumina protocol, and the library was used for paired-end sequencing on the Illumina Hi-Seq 2500 analyzer. After removing short reads and low quality reads, the clean reads were used for mapping to melon reference genome DHL92 (https://melonomics.net/) (Garcia-Mas et al. 2012) using software BWA (Li and Durbin 2009). SNPs and small indels detected from the alignments were called using Samtools and output was given in pileup format (Li et al. 2009). The quality score of the SNPs was assigned by Samtools to evaluate the reliability of SNP calling based on the Phred-scaled probability that the consensus is identical to the reference. The SNPs and small indels between two parental lines were detected using the GATK software tool package (DeLuca et al. 2012), and the reliable SNPs and small indels were noted and predicted using SnpEff software (Chen et al. 2009).

Expression analysis with qRT-PCR

To examine expression patterns of the candidate gene, qRT-PCR analysis was performed with tissues from NSL73046 and M465 including cotyledon, leaf, flower, fruit and tendril. Total RNA of different fresh tissues was extracted using the Plant RNA kit (Omega, USA) and reverse-transcribed using SuperScript III reverse transcriptase (Invitrogen, USA). All experiments were performed with three biological and three technical replicates. The actin gene (MELO3C008032) of melon was used as an internal control. qRT-PCR was performed using the SYBR Green PCR master mix (Applied Biosystems Inc., USA) in iCycleriQTM 5 Multicolor Real-Time PCR detection system (Bio-Rad, USA). The threshold cycle (C t) value of actin protein was subtracted from that of CmGL to obtain a ΔC t value. The C t value of control sample M465 was subtracted from the ΔC t value to obtain a ΔΔC t value. The fold changes in expression level relative to M465 were expressed as \( 2^{{ - \Delta \Delta C_{\text{t}} }} \). Student’s t test was used for comparison between NSL73046 and M465. A difference was considered to be statistically significant when P < 0.05. The primers used in this study were listed in Supplementary Table S2.

Identification of HD-ZIP IV gene in melon and phylogenetic analysis

To investigate the phylogenetic relationships of melon CmGL protein with other identified members of HD-ZIP IV family, the proteins used in this study followed those described in Pan et al. (2015) and an additional protein GhHOX3 regulating cotton fiber development was also included (Shan et al. 2014). The sequences in Arabidopsis, cucumber (Cucumis sativus), tomato (Solanum lycopersicum), maize (Zea mays), rice (Oryza sativa) and cotton (Gossypium spp.) were downloaded from NCBI GenBank (https://www.ncbi.nlm.nih.gov/genbank/). The full-length protein sequences of different species were aligned using the CLUSTX program (Chenna et al. 2003) and the alignment report was visualized by Gendoc. The phylogenetic tree was constructed using the maximum likelihood method in MEGA 5 with a bootstrap of 1000 replicates (Tamura et al. 2011).

We further identified the genome-wide HD-ZIP IV gene members in melon genome using the method followed by Fu et al. (2013). All the identified HD-ZIP IV proteins were further confirmed by Pfam (http://pfam.xfam.org) and SMART (http://smart.embl-heidelberg.de/). The 11 HD-ZIP IV genes identified from cucumber 9930 genome Version2 were downloaded from the Cucurbit Genomics Database (http://www.icugi.org/) (Fu et al. 2013). All full-length protein sequences of non-redundant HD-ZIP IV genes in cucumber and melon were aligned using the CLUSTX program (Chenna et al. 2003) and a phylogenetic tree was constructed using the maximum likelihood method in MEGA 5 (Tamura et al. 2011).

Results

Morphological characterization and inheritance of glabrous mutant in melon

The wild-type melon line M465 had lots of trichomes covering the hypocotyl, cotyledons, young leaves, stem, tendrils, flowers, ovaries and fruits, while the mutant line NSL73046 showed a completely glabrous phenotype on all aboveground organs (Fig. 1). Consistent with these results, we further examined the hypocotyl, leaves and stems of the mutant under a scanning electron microscope (Fig. 2) and found no trichomes on these organs. The trichomes of the wild-type plant were all multicellular consisting of three distinct cell types (head cell, stalk cell, and base cell), which was similar to cucumber (Cui et al. 2016; Pan et al. 2015; Wang et al. 2016).

Fig. 1
figure 1

Trichome phenotypic characterization of stem, female flower and leaf in M465 (a, c, e) and NSL73046 (b, d, f)

Full size image
Fig. 2
figure 2

Scanning electron microscopy images of hypocotyl, young leaf and stem in M465 (a, c, e) and NSL73046 (b, d, f). Arrows indicate the head cell (upper), stalk cell (middle), and base cell (lower) of trichome. Bars = 50 μm

Full size image

Among the F2 population of NSL73046 × M465, the segregation of wild-type and glabrous plants was in agreement with a single recessive gene underlying the trichome-free phenotype (Table 1), which was designated cmgl hereinafter. This result was also consistent with early observations by Foster (1963) who suggested that the glabrous phenotype was controlled by a single gene.

Table 1 Segregation analysis of the trichome trait in the F2 population
Full size table

Fine mapping of cmgl gene

The BSA method was employed for framework mapping of the cmgl locus. Of 384 SSR markers selected for polymorphism screening, three were polymorphic between the non-glabrous and glabrous pools, all of which were located on melon chromosome VIII. Then 96 additional SSR markers around this region were selected from a SSR primer list developed from the melon draft genome (Zhu et al. 2016) and five of them were polymorphic between the two bulks. All eight polymorphic SSR markers were used for genotyping 256 F2 plants, and the cmgl gene was mapped between markers CmSSR19480 and CmSSR19495 with a genetic distance of 0.93 and 0.01 cM from the target gene, respectively (Fig. 3a). Information about these markers used in the present study is provided in Supplemental Table S2.

Fig. 3
figure 3

Fine genetic mapping of CmGL gene in melon. a Preliminary mapping of cmgl gene using SSR markers in 256 F2 plants. b Fine mapping of cmgl gene using flanking markers in 438 F2 plants. c The cmgl gene was finally mapped between Indel2 and Indel5 marker in 1536 F2 plants. d Only one gene was predicated in the candidate region

Full size image

To further narrow down the candidate region harboring the cmgl locus, the two parental lines NSL73046 and M465 were re-sequenced using Illumina Hi-Seq 2500. After trimming low quality reads, 83,532,402 (25.06 Gb) and 89,589,250 (26.88 Gb) clean reads were obtained for NSL73046 and M465, respectively. Alignment of these clean reads against the reference genome identified 159 high quality SNPs and 78 Indels in the candidate gene region between two parental lines. Six Indels with difference larger than 4 bp were selected for marker development, and four of them, CmSSR19480, CmSSR19495, Indel1 and Indel6, were polymorphic between two parental lines, which were used for genotyping 438 F2 plants. This allowed to further narrow the cmgl gene down to a 50 kb region in scaffold00007 with a genetic distance of 0.3 and 0.1 cM to Indel1 and Indel6, respectively (Fig. 3b). The two flanking markers Indel1 and Indel6 were genotyped in an extended F2 population containing 1536 plants, and 8 recombinant plants were identified. Two new polymorphic markers, Indel2 and Indel5, were developed inside this 50 kb region and used for genotyping these recombinant plants. The cmgl gene was finally mapped in the region of 11.58 kb between Indel2 and Indel5 (Fig. 3c). Searching the melon genome database (https://melonomics.net/) and manual annotation, only one candidate gene, MEL03COO7377, was identified in this region supporting it is the only candidate gene for the cmgl locus.

We examined the expression level of CmGL with qRT-PCR in the cotyledon, leaf, flower, fruit and tendril of both parental lines. The CmGL gene was expressed in all these organs with the highest expression detected in the leaf (Fig. 4) and the expression level in leaf in the mutant line was significantly lower than that in the wild-type.

Fig. 4
figure 4

The expression analysis of CmGL gene in the wild-type and mutant line using qRT-PCR in cotyledon, leaf, flower, fruit and tendril. Asterisk represent statistically significant difference (P < 0.05) between NSL73046 and M465

Full size image

Gene annotation, sequence alignment and allelic diversity of CmGL

The length of genomic DNA of MEL03COO7377 in the melon reference genome DHL92 was 5367 bp, which was predicted to contain ten exons and the full-length CDS was 2166 bp. The genomic sequence of the candidate gene was compared between two parental lines NSL73046 and M465 by aligning the NGS resequencing reads to the reference genome, and ten SNPs were identified between the wild-type line and mutant line. Among of them, six SNPs were located in the introns, while four SNPs were located in the exons 4, 8 and 10, respectively. The single base substitution in the fourth exon changing from C to A resulted in a premature termination codon in the mutant line, which led to two truncated ORFs predicted in the same genomic region (Fig. 5a). Based on this SNP difference, a dCAPS marker, dCPAS1, was developed which showed co-segregation with glabrous plants in the F2 population (Fig. 3c). The deduced amino acid sequence of CmGL contained 721 amino acids and was predicted to encode a HD-ZIP IV transcription factor containing a conserved HOX domain (from 48 to 111 aa) and the START domain (from 237 to 459 aa) (Fig. 5b), while they were separated into two truncated ORFs in the mutant line. BLASTP search in NCBI showed that it shared a high identity to cucumber HD-ZIP IV protein, CsGL3, with 97.6% identity in CDS level and 99.6% identity in amino acid level. CsGL3 has been shown to regulate trichome initiation in cucumber (Cui et al. 2016; Pan et al. 2015; Wang et al. 2016). A phylogenetic tree was also constructed using CmGL and other 31 HD-ZIP IV proteins identified from cucumber, Arabidopsis, tomato, maize, rice, and cotton. The results revealed that CmGL showed a closer relationship to cucumber CsGL3, cotton GhHD1 and tomato WO (Supplementary Figure S1). These proteins are involved in trichome initiation and cell fate determination, suggesting CmGL in melon may have similar functions in regulating multicellular trichome development. Furthermore, 12 non-redundant genes encoding HD-ZIP IV proteins were identified in melon genome including the candidate gene CmGL, and they were used together with 11 HD-ZIP IV proteins identified in cucumber genomes (Fu et al. 2013) for multiple sequence alignments. A phylogenetic tree was constructed to unveil the relationship of HD-ZIP IV genes in cucumber and melon, and CmGL showed the closest relationship to CsGL3 suggesting that they were orthologous genes (Supplementary Figure S2).

Fig. 5
figure 5

Gene prediction, sequence comparison and allele diversity of CmGL gene in melon. a The gene prediction and sequence alignment of CmGL gene between the wild-type (upper) and mutant line (lower), and four SNPs were detected in the exons. A single base substitution in the fourth exon from C to A led to a premature termination codon in the mutant. b The CmGL gene was annotated to have two conserved domains in the deduced amino acid sequences. c Allelic diversity of CmGL gene in different melon germplasms using CmGL co-segregating marker dCAPS1. The first lane represents 100 bp DNA ladder, the second lane represents the mutant line and the remaining ones represent the natural melon germplasms

Full size image

To examine allelic diversity of the CmGL gene in melon natural populations, trichomes of 564 melon accessions (Supplementary Table S1) were observed in 2015 and 2016 field trials. All of them had trichomes as found in M465 wild-type parental line except for the mutant line NSL73046. The cmgl co-segregating marker dCAPS1 was used for genotyping all 564 melon accessions (Fig. 5c). The genotypic data at this marker locus were completely consistent with the phenotyping analysis, suggesting that the SNP within the cmgl gene was a causal mutation in NSL73046 and further confirmed it was the candidate gene.

Discussion

Plant trichomes play an important role in protecting plants from damage caused by herbivores and pathogens. Different studies have revealed that there is a close relationship between trichome and resistance against aphids and viruses in many species, such as potato (Gregory et al. 1986) and tomato (Kennedy and Sorenson 1985). In melon, leaf trichome density has been also shown to be associated with the rejection of A. gossypii Glover to settle (Sarria et al. 2010). A major QTL, tric11 mapped on melon chromosome XI was identified for controlling the density of glandular trichomes, which explained between 23.8 and 58.7% of its phenotypic variation (Palomares-Rius et al. 2016). In this study, we identified a completely glabrous mutant in the melon accession NSL73046 and we show a region of ~ 12 kb harboring the cmgl locus was associated with this mutation (Fig. 3). Only one gene encoding a HD-ZIP IV transcription factor in this candidate region was identified and a dCAPS marker was developed based on the SNP in the fourth exon showing co-segregation with glabrous phenotype in the F2 plants and natural populations, further suggesting that it was the candidate gene for the cmgl locus (Fig. 5). This was the first report in melon to clone a gene for glabrous mutations, which was different from the tricl1 major-effect QTL for trichome density reported earlier (Palomares-Rius et al. 2016). The glabrous phenotype will be useful in the identification of seed purity for melon hybrid in the seeding stage for its ease to differentiate the leaf and hypocotyl in the glabrous plants with normal plants.

Many transcription factors have been identified in regulating trichomes development in the model species Arabidopsis (Hulskamp 2004; Hulskamp et al. 1999; Schellmann and Hulskamp 2005; Ishida et al. 2008; Pesch and Hulskamp 2009; Schiefelbein 2003), and the molecular mechanisms involved in the initiation and development of trichomes have been extensively investigated (Kirik et al. 2004, 2005; Payne et al. 2000; Walker et al. 1999; Wester et al. 2009). Several studies in tobacco and tomato have revealed that there may be different regulation networks controlling the development of multicellular and unicellular trichomes (Payne et al. 1999; Yang et al. 2011). However, the molecular mechanism of trichomes development in melon is still largely unknown. Further functional analysis on CmGL gene will be helpful for better understanding the regulation of multicellular trichomes development in melon.

Melon and cucumber are two important crops in the genus Cucumis, (family Cucurbitaceae) that were diverged around 10 million years ago (Sebastian et al. 2010). The CmGL gene identified in this study shared a very high similarity to the cucumber CsGL3 gene (Pan et al. 2015) with 97.6% nucleotide sequence identity and 99.6% amino acid sequence identity. The cucumber csgl3 (Pan et al. 2015) and melon cmgl (this study) mutants both showed the completely glabrous phenotype suggesting they may perform similar functions. The loss-of-function mutation in melon cmgl was due mainly to a single base substitution from C to A which led to a premature termination codon and a truncated protein without the START domain, while that in the cucumber csgl3 mutant was because of insertion of a 5005 bp LTR retrotransposon (LTR-RT) (Pan et al. 2015). We checked the physical locations of these two homologous genes and found that they were located on a highly conserved syntenic block between cucumber and melon genomes (Garcia-Mas et al. 2012; Li et al. 2011; Yang et al. 2014). It is interesting to note that the LTR-RT has a much higher fraction in the assembled melon draft genome than that in cucumber; the activity of LTR-RT is also much higher in the melon lineage after the divergence (Garcia-Mas et al. 2012). Using CmGL co-segregating dCAPS marker, we examined the allelic diversity at the CmGL locus in a diverse melon panel including the wild melon and two subspecies: subsp. melo and subsp. agrestis, there was no difference among these natural germplasms, suggesting the SNP within the cmgl gene may be the result of a natural mutation rather than due to domestication or human selection.

The plant-specific HD-Zip transcription factor gene family is characterized by a highly conserved 60–61 amino acid homeodomain (HD) (Ariel et al. 2007; Mayda et al. 1999). These proteins exhibit the singular combination of a HD domain with a leucine zipper acting as a dimerization motif. They can be classified into four subfamilies, HD-ZIp I–IV, according to a set of distinctive features that include DNA-binding specificities, gene structures, additional common motifs and physiological functions (Mukherjee et al. 2009). Several HD-ZIP transcription factors have been identified in regulating multicellular trichome development. For example, the Wo gene is essential for trichome formation in tomato, which encodes a HD-ZIP IV transcription factor, and mainly regulates the formation of type I trichomes (Yang et al. 2011). Another HD-ZIP IV transcription factor OCL4, regulating the differentiation of the anther cell wall, has been also identified in inhibiting trichome development in both maize and Arabidopsis (Vernoud et al. 2009). In cucumber, CsGL1 and CsGL3 encode a HD-ZIP I and HD-ZIP IV transcription factor, respectively, and they regulate the trichomes development in different manners. The csgl1 mutant shows no visible trichomes on leaves, stems, tendrils, and floral organs except for the hypocotyl, and the role of CsGL1 in trichome development in cucumber has not been reported in Arabidopsis (Chen et al. 2014; Zhao et al. 2015). The csgl3 mutant shows a completely glabrous phenotype on the aerial parts of plants (Cui et al. 2016; Pan et al. 2015; Zhao et al. 2015). The CmGL gene identified in our study controlling the completely glabrous phenotype in melon was also encoded by a HD-ZIP IV transcription factor, and a phylogenetic tree constructed using all HD-ZIP IV proteins identified in cucumber and melon revealed that it was an orthologous gene of CsGL3 (Supplementary Figure S2). Thus, it seems that most HD-ZIP genes regulating trichome development are members of the HD-ZIP IV subfamily, which have been reported to generally regulate the processes related to epidermal cell differentiation, trichome formation, and anthocyanin synthesis (Nakamura et al. 2006). The START domain is often identified in the HD-ZIP IV subfamily proteins, but their functions in plants are still unknown; their conservation between plants and animals suggest that they may be involved in lipid/sterol signal transduction (Schrick et al. 2004). The loss-of-function in the melon cmgl mutant resulted from a truncated protein with a lost START domain, and the domain was also missing in the tril mutant in cucumber (Wang et al. 2016). In addition, the START domain was also lost in cmgl locus in the melon glabrous mutant. The role of the START domain in trichome formation in higher plants remains to be elucidated in future studies.

Author contribution statement

XS, PS, QH, XZ and XL performed phenotyping in F2 plants and fine mapping. HZ, JH and SS contributed to data processing and phylogenetic analysis. QZ and JP contributed to SEM microscopic analysis. LY, HZ and YW wrote the manuscript. All authors reviewed and approved this manuscript.