agronomy Genetic Mapping and Identiﬁcation of the Gibberellin 3-Oxidase Gene GA3ox Leading to a GA-Deﬁcient Dwarf Phenotype in Pumpkin ( Cucurbita moschata D.)

: Plant height is an important indicator in the ideal plant model and contributes to optimizing yield and lodging resistance. The emergence of a dwarﬁng phenotype provides an opportunity for plant height improvement. In a previous study, we identiﬁed a dwarf mutant Si1 in pumpkin ( Cucurbita moschata D.) obtained by ethyl methane sulfonate (EMS) mutagenesis of the inbred line N87. Phenotype identiﬁcation for Si1 revealed a decrease in cell size and shorter internodes than those of wild type. Genetic analysis revealed that the dwarf mutant trait was controlled by a single recessive gene, CmaSI1 . By bulked segregant analysis (BSA) and subsequent ﬁne mapping, we mapped the CmaSI1 locus to a 463 kb region on chromosome 8 that contained 28 annotated genes in the F 2 population. Only one nonsynonymous single nucleotide polymorphism (SNP) in CmoCh08G006170 was obtained according to whole-genome resequencing of the two parents. CmoCh08G006170 , a homolog of Arabidopsis gibberellin 3-beta hydroxylase ( GA3ox ), is a key enzyme in the regulation of bioactive gibberellins (GAs). RNA-seq analysis and qRT-PCR showed that the expression level of CmoCh08G006170 in stems of Si1 was changed compared with that of wild type. The dwarf phenotype could be restored by exogenous GA 3 treatment, suggesting that Si1 is a GA-deﬁcient mutant. The above results demonstrated that CmoCh08G006170 may be the candidate gene controlling the dwarf phenotype. This study provides an important theoretical basis for the genetic regulation of vine length and crop breeding in pumpkin.


Introduction
The ideal plant model has a significant impact on the growth of crops, and plant height is one of the important indicators in the plant model. Moderate dwarfing better optimizes crop yield and prevents lodging resistance, which saves land resources by improving the yield per unit area [1,2]. In the middle of the 20th century, semidwarf rice varieties significantly increased the yield per unit area of rice, which was known as the first "Green Revolution" in the history of rice breeding [3]. Since then, dwarfing and semidwarfing traits have been widely introduced into important crops and have effectively alleviated the trade-off between high yield and easy lodging [4,5]. The breeding and promotion of dwarfing varieties have greatly improved productivity and the progress of human society.
Many dwarfing mutants have been found in important food crops such as maize, wheat, and rice [6,7]. Among them, more than 90 dwarfing genes in rice, 26 in wheat, and 20 in maize were identified by using dwarfing materials, such as predominant semidwarf1 (sd1) [8], shortened basal internodes (SBI) [9], reduced height genes Rht-B1b, Rht-D1b and candidate gene CmoCh08G006170, which encodes a gibberellin 3β-hydroxylase (GA3ox), served as the most likely candidate gene for the dwarf phenotype of Si1. Nonsynonymous single nucleotide polymorphisms (SNPs) were detected, and the expression level of CmoCh08G006170 in the stem of Si1 was changed compared with the wild type (WT) N87. This study provides new genetic resources and a theoretical basis for understanding the regulatory mechanisms of dwarfism, which is conducive to improving pumpkin breeding.

Plant Materials
Si1, the dwarf mutant, was obtained from an EMS mutagenized population of the pumpkin inbred line N87. N87, a high-generation inbred line, was bred by the Hunan Vegetable Research Institute, Changsha, Hunan, China. It demonstrates strong growth, has short gourd-shaped fruit, a sweet taste, and high resistance to powdery mildew and blight in the field [30]. The specific steps of EMS treatment were reported by Min et al. (2021) [31]. The plant materials were grown in the experimental field for observation and sampling. For the genetic analysis of the dwarf mutant genes, different cross combinations were constructed. Si1 (P 2 ) was crossed with N87 (P 1 ) to generate an F 1 population, and the F 1 population was self-pollinated to produce F 2 progeny; BC 1 P 1 and BC 1 P 2 progeny were then generated by BCs of F 1 × N87 and F 1 × Si1, respectively. The dwarf ratios of the F 1 and segregation populations were analyzed by a chi-squared test using SAS software (SAS Institute, 2001).

Plant Phenotyping and Morphological Characterization
Pumpkin is an infinitely growing crop, its vines can reach 4-5 m. The first twenty nodes of the main vine were chosen to exhibit plant height in this study; the growth period of pumpkin includes the periods of clustering stage and vine elongation stage. Generally, pumpkins enter the vine elongation period from the sixth node and the length of the internodes tends to stabilize after the eighth node. So we chose the main vine of the eighth node as the internode length measurement object. The main vines containing wild-type and mutant internodes were measured in two places: the eighth node and the first twenty nodes. Other traits were also measured, including leaves and stem diameter. Each indicator was measured as the average of ten plants grown under the same conditions, and three replicates were used for each measurement. The error bars represent the SEs from three independent experiments. The data were analyzed by ANOVA using SAS software in the study. For morphological characterization, the eighth internode was cut into 3-mm pieces and fixed in a formalin-acetic acid-alcohol (FAA) solution (70% alcohol: 3.8% formalin: glacial acetic acid = 18:1:1) for 24 h, dehydrated with a different concentrations of ethanol (75%, 85%, 90%, 95%, and 2 × 100%), infiltrated with 100% xylene and embedded in paraffin. The samples were cut into 10-µm slices by an RM2016 microtome (Leica, Wetzlar, Germany) and mounted on glass slides. Then, they were stained with 1% toluidine blue for 5 min and washed with deionized water and ethanol. Finally, the sections were observed and imaged using an Olympus BH-2 photomicroscope (Olympus, Tokyo, Japan).

Bulked Segregant Analysis (BSA)
Leaf samples were collected from 20 randomly selected normal-vine individuals and 20 randomly selected dwarf individuals from the F 2 population to construct two bulks: a normal pool and a dwarf pool. Genomic DNA was isolated from fresh samples using the Plant Genomic DNA Kit (Tiangen, Beijing, China) following the manufacturer's instructions. The quality of genomic DNA was examined on a 1.5% agarose gel, and the purity was checked by a Nanodrop2000 spectrophotometer (Thermo Scientific, Wilmington, NC, USA). After quality control, two bulked DNA samples and two parent samples were selected for 30× resequencing. Then, the DNA fragments were subjected to purification, end-repair sequencing, and joint and PCR amplification and sequenced on the Illumina HiSeq 2500 platform. The filtered high-quality sequences were aligned and mapped to the Pumpkin Genomics Database (http://www.cucurbitgenomics.org/organism/9) (accessed on 1 July 2022) using Burrows-Wheeler Aligner (BWA) for subsequent variation analysis [32]. SNPs and INDELs were detected using the mutation analysis software GATK [33]. According to a previous study, SNP-index and sliding-window analyses were performed [34].

Marker Development and Fine Mapping of the CmaSI1 Gene
According to the BSA-Seq results, KASP markers were designed based on target candidate region resequencing data for the two parents, and the KASP assay was performed following the protocol of LGC Genomics (Berlin, Germany), as reported previously [35]. An Applied Biosystems Viia 7 real-time PCR system (Applied Biosystems, Foster City, CA, USA) was used to detect allele-specific fluorescence, and all genotypes were called with Viia 7 software, v1.0. The KASP marker design strategy was as follows: 300-400 bp between each SNP, with SNPs preferentially located in coding regions. The primers used here are listed in Supplementary Table S1.

Candidate Gene Analysis
Based on the reference genome, the information in the mapping region of these genes was downloaded, including the gene ID, gene position, and functional annotation. These genes were first screened to determine whether they contributed to the dwarf phenotype. From the resequencing data of the two parents, we focused on key SNPs that caused stoploss, stop-gain, or nonsynonymous mutations or splice-site variants as candidate SNPs. In addition, EMS mutations usually present as C-T or G-A SNPs. The primers used to amplify the genomic DNA or cDNA of the candidate genes were designed using Primer 5.0. The products of gene amplification were sequenced and aligned using DNAman software. The sequences of all polymorphic markers and primers used to amplify the candidate gene are listed in Table S1.
To analyze the sequence conservation for GA3ox across the amino acid sequences for Cucurbitaceae species, including cucumber, C. maxima, silver-seed gourd, wax gourd, bottle gourd, wild cucumber, and long cucumber, were downloaded from the Cucurbit Genomics Database. A neighbor-joining tree was generated with 1000 bootstrap replicates using MEGA 6.0. Quantitative real-time PCR (qRT-PCR) was carried out using a PrimeScript RT Reagent Kit (Tiangen, Beijing, China) in conjunction with QuantStudio 6 Flex (Life Technologies Corporation, Carlsbad, CA, USA) as described in the manufacturers' protocols. Three biological and three technical replicates were included for each experiment. The relative expression level of each gene was calculated using the 2 −∆∆Ct method [36] using the expression of actin as an internal control [37].

RNA-Sequencing and Data Analysis
The young stems containing wildtype and mutant internodes at the five-leaf and one-heart stages were harvested and stored at −80 • C for further analyses. Total RNA was extracted using an EasyPure Plant RNA Kit (Tiangen, Beijing, China), and the RNA purity was assessed utilizing an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). All libraries were sequenced on an Illumina HiSeq 4000 platform (Illumina). Clean data were obtained by removing low-quality sequences, adaptor-contaminated sequences, and sequences with ambiguous base reads accounting for more than 5% of the raw data. The Hisat2 (http://ccb.jhu.edu/software/hisat2/index.shtml) (accessed on 1 July 2022) was used to align the clean reads with the Pumpkin Genomics Database (http://www.cucurbitgenomics.org/organism/9) (accessed on 1 July 2022). The fragments per kilobase per transcript per million mapped reads (FPKM) values were calculated and used to estimate the sequencing depth and gene length on the mapped read counts. Genes with an adjusted p-value (padj) < 0.05 and |log (fold change)| ≥ 1 were assigned as differentially expressed. Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were enriched by differentially expressed genes (DEGs) if the p values were ≤0.05.

Treatment with Exogenous GA 3
GA is known as a key plant hormone contributing to plant height, and exogenous GA treatment at different concentrations can restore dwarfing phenotypes. Based on the above result, candidate gene CmoCh08G006170 is a key enzyme in the regulation of bioactive gibberellins (GAs). To determine whether dwarfing phenotypes of Si1 seedlings are related to gibberellin, exogenous GA treatment at different concentrations was conducted by spraying the Si1 seedlings. The germinated seeds were incubated in a greenhouse at 25.0 ± 1 • C with a 16 h/8 h light/dark cycle. At the two-leaf stage, the treated leaves were sprayed with 50 and 100 µM GA 3 , while the control seedlings were sprayed with water. Three biological replicates were sprayed three times every two days. The height from cotyledons to the growing point was measured 30 days after the final treatment.

Distinct Phenotypes of the Dwarf Mutant
The dwarf mutant Si1 was much shorter than wildtype (WT) N87 ( Figure 1A). The total length of the first twentieth node of Si1 (50.87 ± 0.88 cm) was reduced to 25.3% of that of the WT (200.68 ± 3.49 cm), and the eighth internode of Si1 (12.36 ± 0.40 cm) was reduced to 24.2% of that of the WT (50.94 ± 1.11 cm) (Table S2). These results showed that the length of the internodes of the dwarf mutant was approximately one-quarter of that of the WT, which could be visibly distinguished. However, there were no visible differences in leaf or stem diameter values ( Figure 1B,C), which indicated that the mutant mainly changed the length of the main vine.  To further analyze the cellular differences in the internodes of N87 and Si1, the cellular structure in the longitudinal section was observed and evaluated using optical microscopy ( Figure 2A,B). From the microscopy images, the length of parenchyma cells in Si1 (72.9 ± 5.4 μm) was shorter than that in N87 (103.6 ± 8.3 μm), and the width of the cells in Si1 (63.2 ± 4.8 μm) was shorter than that in N87 (80.9 ± 6.7 μm), indicating that the mutant had aberrant cell elongation ( Figure 2C). Additionally, the number of parenchyma cells in Si1 (64.5 ± 8.4 μm) was significantly increased compared with that in N87 (42.1 ± 5.8 μm) ( Figure 2D), suggesting that the reduced cell volume may be the main reason for the reduced internode length. To further analyze the cellular differences in the internodes of N87 and Si1, the cellular structure in the longitudinal section was observed and evaluated using optical microscopy (Figure 2A,B). From the microscopy images, the length of parenchyma cells in Si1 (72.9 ± 5.4 µm) was shorter than that in N87 (103.6 ± 8.3 µm), and the width of the cells in Si1 (63.2 ± 4.8 µm) was shorter than that in N87 (80.9 ± 6.7 µm), indicating that the mutant had aberrant cell elongation ( Figure 2C). Additionally, the number of parenchyma cells in Si1 (64.5 ± 8.4 µm) was significantly increased compared with that in N87 (42.1 ± 5.8 µm) ( Figure 2D), suggesting that the reduced cell volume may be the main reason for the reduced internode length. To further analyze the cellular differences in the internodes of N87 and Si1, the cellular structure in the longitudinal section was observed and evaluated using optical microscopy (Figure 2A,B). From the microscopy images, the length of parenchyma cells in Si1 (72.9 ± 5.4 μm) was shorter than that in N87 (103.6 ± 8.3 μm), and the width of the cells in Si1 (63.2 ± 4.8 μm) was shorter than that in N87 (80.9 ± 6.7 μm), indicating that the mutant had aberrant cell elongation ( Figure 2C). Additionally, the number of parenchyma cells in Si1 (64.5 ± 8.4 μm) was significantly increased compared with that in N87 (42.1 ± 5.8 μm) ( Figure 2D), suggesting that the reduced cell volume may be the main reason for the reduced internode length.

Genetic Analysis of the Dwarf Mutant
To analyze the inheritance pattern of the dwarf mutant, Si1 was crossed with its original parent N87. The phenotype of all reciprocally crossed F 1 plants was normal; thus, the normal trait was dominant over the dwarf trait in the two lines. In the F 2 population, 56 of 244 individuals displayed the dwarf phenotype and 188 were normal, and the chisquare test confirmed this segregation ratio to be 3:1 (χ 2 = 0.55 < χ 2 0.05 = 3.84). The BC 1 P 1 population contained 80 individuals, with 80 normal individuals and no dwarf individuals. For BC 1 P 2 , 61 individuals presented the dwarf phenotype while 63 individuals were normal, which fit a segregation ratio of 1:1 (χ 2 = 0.03 < χ 2 0.05 = 3.84) ( Table 1). These results indicate that the dwarf trait is controlled by a single recessive gene, which was named CmaSI1.

Fine Mapping of the CmaSI1 Locus by BSA-Seq and Linkage Analyses
Whole genome resequencing (WGR) through the BSA-Seq approach was used to identify the genomic region controlling the dwarf phenotype. A total of 273,422,085 raw reads were obtained from the dwarf pool (21.62× average read depth with 93% coverage), and 273,422,085 raw reads were obtained from the normal pool (24.23× average read depth with 93% coverage). In addition, the Q20 values (i.e., reads with average quality scores >20) were all >90%, and the Q30 values (i.e., reads with average quality scores >30) were all >96%. These results indicate that the accuracy and quality of the sequencing data were sufficient for further analysis. To identify the genomic regions associated with dwarfism, the SNP index of each SNP locus was calculated between the two bulks using high-quality SNPs. The results revealed that there was only one candidate interval responsible for dwarfism; it was located in the 2,098,346-6,401,791 bp region on chromosome 8 ( Figure 3A). To further narrow the candidate region harboring the CmaSI1 locus, eight SNPs within the 4.4 Mb candidate region were designed as KASP markers and showed polymorphism between the two parents. The marker information is shown in Table S1. Then, a total of 512 recessive individuals from the F2 population were subsequently used for CmaSI1 fine-mapping. Finally, the CmaSI1 gene was mapped between SNP3675880 and SNP4123114, which each had one recombinant. The markers were physically located in a 463 kb region from 3,675,880 to 4,123,114 bp on chromosome 8 ( Figure 3B). between the two parents. The marker information is shown in Table S1. Then, a total of 512 recessive individuals from the F 2 population were subsequently used for CmaSI1 finemapping. Finally, the CmaSI1 gene was mapped between SNP3675880 and SNP4123114, which each had one recombinant. The markers were physically located in a 463 kb region from 3,675,880 to 4,123,114 bp on chromosome 8 ( Figure 3B).

Identification of Candidate Genes for the CmaSI1 Gene
According to the fine mapping and the annotation information from the reference genome, 28 genes were predicted, and the physical positions and gene annotations of these genes are shown in Table S3. Among them, CmoCh08G006140 encodes an auxin-responsive element, CmoCh08G006170 encodes a gibberellin 3-beta hydroxylase and CmoCh08G006180 encodes an ABC transporter protein [38,39]; all three of these may contribute to the dwarf phenotype. From the resequencing data of the two parents, we focused on seven SNPs that caused stop-loss, stop-gain, or nonsynonymous mutations or splice-site variants as candidate SNPs ( Table 2). The seven SNPs were located on different genes, and only one SNP caused a nonsynonymous mutation (Chr08:3,779,377) in CmoCh08G006170. Furthermore, the SNP on CmoCh08G006170 was confirmed by Sanger sequencing, and the corresponding KASP marker cosegregated in the F 2 population ( Figure 3C,D). The mutation is a single nucleotide substitution from G to A that occurs in the second exon, leading to an amino acid substitution at codon 229 (Glu→Arg). CmoCh08G006170 encodes a gibberellin 3-beta hydroxylase (GA3ox), which is a key enzyme in the regulation of bioactive GAs. Phylogenetic analysis indicated that the CmoCh08G006170 gene has a close relationship with seventeen homologous genes and shares a common ancestor with Carg23157 from a silver-seed gourd, suggesting that Cla015407 was evolutionarily conserved within the Cucurbitaceae family ( Figure 4B). Sequence analysis further revealed that CmoCh08G006170 and its homologs are significantly conserved ( Figure 4C). Nucleotide site 688 in CmoCh08G006170 in all homologous genes presented a G encoding the highly conserved amino acid Glu. This result revealed that the SNP in the CmoCh08G006170 gene may be the critical functional mutation.

Analysis of the Expression of GAs Biosynthetic Pathways and Regulatory Genes
To investigate the molecular mechanisms that underlie the Si1 dwarf, a comparative transcriptome analysis was used to identify DEGs between Si1 and WT. DEGs were screened based on set criteria (q value ≤ 0.05 and |log2FC| > 1), and totals of 976 DEGs (752 upregulated and 224 down-regulated) were detected. For GO enrichment terms, DEGs were highly represented in the biological process category by the terms 'oxidation reduction process', 'xyloglucan metabolic process', and 'negative regulation of peptidase activity'; in the cellular category, the most represented terms were cell periphery, 'apoplast' and 'cell wall'. In addition, 'oxidoreductase activity', 'oxidoreductase_activity', and 'oxidoreductase activity' were among the most commonly represented molecular function terms. For KEGG pathways, the pathway 'phenylpropanoid biosynthesis' was the most common term, followed by 'biosynthesis of secondary metabolites', 'metabolic pathways', and 'plant hormone signal transduction' pathway.

Analysis of the Expression of GAs Biosynthetic Pathways and Regulatory Genes
To investigate the molecular mechanisms that underlie the Si1 dwarf, a comparative transcriptome analysis was used to identify DEGs between Si1 and WT. DEGs were screened based on set criteria (q value ≤ 0.05 and |log2FC| > 1), and totals of 976 DEGs (752 up-regulated and 224 down-regulated) were detected. For GO enrichment terms, DEGs were highly represented in the biological process category by the terms 'oxidation reduction process', 'xyloglucan metabolic process', and 'negative regulation of peptidase activity'; in the cellular category, the most represented terms were cell periphery, 'apoplast' and 'cell wall'. In addition, 'oxidoreductase activity', 'oxidoreductase_activity', and 'oxidoreductase activity' were among the most commonly represented molecular function terms. For KEGG pathways, the pathway 'phenylpropanoid biosynthesis' was the most common term, followed by 'biosynthesis of secondary metabolites', 'metabolic pathways', and 'plant hormone signal transduction' pathway.
The synthesis of bioactive gibberellin is a complex process, and several DEGs that encode functional enzymes were well characterized from RNA-seq, including ent-copalyl diphosphate synthase (CPS), ent-kaurenoic acid oxidase (KAO), GA20ox and GA3ox ( Figure 5). GA20ox (CmoCh18G012300, CmoCh16G001660, CmoCh02G010360 and CmoCh15G014590) and GA3ox (CmoCh08G006170) were significantly up-regulated. Among them, GA3ox (CmoCh08G006170) was the candidate gene controlling the dwarf phenotype. Lester (1977) reported that GA3ox can be regulated by feedback and feedforward paths of active GAs, the low level of active GAs in vivo increases the transcription level of GA3ox [40]. We imply that a nonsynonymous mutation in CmoCh08G006170 may influence its molecular function causing reduced active GAs. Furthermore, qRT-PCR results revealed that the expression level of CmoCh08G006170 in stems of Si1 was significantly up-regulated compared with that in wild-type N87 ( Figure  6). However, the expression of CmoCh08G006170 in leaves and roots showed no difference The synthesis of bioactive gibberellin is a complex process, and several DEGs that encode functional enzymes were well characterized from RNA-seq, including ent-copalyl diphosphate synthase (CPS), ent-kaurenoic acid oxidase (KAO), GA20ox and GA3ox ( Figure 5). GA20ox (CmoCh18G012300, CmoCh16G001660, CmoCh02G010360 and CmoCh15G014590) and GA3ox (CmoCh08G006170) were significantly up-regulated. Among them, GA3ox (CmoCh08G006170) was the candidate gene controlling the dwarf phenotype. Lester (1977) reported that GA3ox can be regulated by feedback and feedforward paths of active GAs, the low level of active GAs in vivo increases the transcription level of GA3ox [40]. We imply that a nonsynonymous mutation in CmoCh08G006170 may influence its molecular function causing reduced active GAs. Furthermore, qRT-PCR results revealed that the expression level of CmoCh08G006170 in stems of Si1 was significantly up-regulated compared with that in wild-type N87 ( Figure 6). However, the expression of CmoCh08G006170 in leaves and roots showed no difference between Si1 and N87. The result demonstrated the accuracy of the transcriptome analysis in the present study.
between Si1 and N87. The result demonstrated the accuracy of the transcriptome analysis in the present study.

Si1 Is a GA Biosynthetic-Deficient Mutant
To test whether GAs regulate the height of the dwarf mutant, the mutant was treated with 50 and 100 µM GA 3 . After treatment with GA 3 solution, the lengths of the first eight main vines for WT, MT(H 2 O) MT (100 µM GA 3 ) were 51.16 ± 1.16 cm, 12.19 ± 0.48 cm, and 50.77 ± 0.96 cm, and there was no significant difference between the wildtype and 100 µM GA 3 solution treatments (Figure 7). When the spraying concentration was 100 µM, the vine length of Si1 and internode length were significantly restored, while spraying 50 µM GA 3 solution did not significantly increase the vine length or internode length. The results indicated that the dwarf phenotype is associated with decreased levels of bioactive GA 3 . Taken together, these results confirmed that Si1 is a GA biosynthetic-deficient mutant.

Si1 Is a GA Biosynthetic-Deficient Mutant
To test whether GAs regulate the height of the dwarf mutant, the mutant was treated with 50 and 100 μM GA3. After treatment with GA3 solution, the lengths of the first eight main vines for WT, MT(H2O) MT (100 μM GA3) were 51.16 ± 1.16 cm, 12.19 ± 0.48 cm, and 50.77 ± 0.96 cm, and there was no significant difference between the wildtype and 100 μM GA3 solution treatments (Figure 7). When the spraying concentration was 100 μM, the vine length of Si1 and internode length were significantly restored, while spraying 50 μM GA3 solution did not significantly increase the vine length or internode length. The results indicated that the dwarf phenotype is associated with decreased levels of bioactive GA3. Taken together, these results confirmed that Si1 is a GA biosynthetic-deficient mutant.

Discussion
Moderate dwarfing makes the plant type more compact and increases space utilization, which benefits rational dense planting and mechanized production. Many genes controlling dwarfing traits have been identified, and their molecular mechanisms have been analyzed. However, dwarfing traits often affect other agronomic traits, leading

Discussion
Moderate dwarfing makes the plant type more compact and increases space utilization, which benefits rational dense planting and mechanized production. Many genes controlling dwarfing traits have been identified, and their molecular mechanisms have been analyzed. However, dwarfing traits often affect other agronomic traits, leading to clumps of plants, leaf shrinkage, and even reduced fertility, making it difficult to apply these traits in dwarfing breeding [41]. Therefore, it is particularly important to find and create dwarf germplasm resources with excellent characteristics. There are few studies on pumpkin dwarfing, and most of them have focused on C. pepo and C. maxima. Currently, the C. moschata dwarf lines bu and cga are clumping dwarfs with very short main vines, which have limited direct breeding value [29,42]. In this study, the dwarf mutant Si1 exhibited shortening of the main vine without changes to other traits, providing a new germplasm resource for the study of dwarf C. moschata and resulting in the first C. moschata dwarf gene to be reported. In subsequent research, we will try to use this material for research on variety improvement.
EMS mutation technology and BSA technology have greatly improved the efficiency of new germplasm creation and gene function research. Ethyl methane sulfonate had a mutagenic effect, and since then, this method has been used in several different plants [17,43]. This method is widely used to induce mutants and construct mutant libraries because of its relatively low price, strong operability, and reusability. In our previous research, we successfully used mutation technology to construct a pumpkin mutant library that contains rich variation in traits related to leaves, plant type, reproductive organs, fruit, and other characteristics [31]. This library not only enriched the germplasm resources of C. moschata but also laid a foundation for the study of functional genomics and the breeding of new varieties of C. moschata. In addition, combined with BSA technology and fine mapping, we quickly used a small F 2 population to locate a target gene in a 463 kb region. Because EMS mutations usually present as C-T or G-A changes, the above types were screened from a large group of mutations found between the parents. Then, we focused on SNPs that caused stop-loss, stop-gain, or nonsynonymous mutations or splice-site variants to quickly identify candidate SNPs. In our study, only one predicted nonsynonymous mutation was found in the 463 kb mapping region; this mutation regulates bioactive GAs, which may contribute to the dwarf phenotype. At present, this strategy has also been successfully used in many studies to improve the efficiency of gene function research [17,44,45].
In this study, GA treatment rescued the dwarfing phenotype, suggesting that the dwarf-induced mutation was gibberellin-dependent, which confirmed that functional mutation of GA3ox causes GA-sensitive dwarf mutants. GA-sensitive mutants are often found in genes related to the GA biosynthesis pathway, and these mutations can lead to blockade of the GA biosynthesis pathway or the inability to synthesize bioactive GA. For these mutants, the dwarfing phenotype can be restored to the wildtype phenotype by spraying GA [46]. Currently, genes encoding enzymes related to GA biosynthesis or signal transduction pathways have been cloned [47,48]. In the GA biosynthesis pathway, CPS plays an early role in transforming geranylgeranyl diphosphate (GGDP) into CDP in plastids, and KAO in the endoplasmic reticulum catalyzes the conversion of ent-kaurene acid into GA12 [49,50]. GA3ox and GA20ox play important roles in the final steps of the GA biosynthesis pathway [39,47]. GA3oxs transform GA 12 into bioactive GA and is a key gene in the last step of the GA synthesis pathway [41,47,48]. Four GA3ox homologs (AtGA3ox1, AtGA3ox2, AtGA3ox3, and AtGA3ox4) have been identified in Arabidopsis thaliana [51], with two GA3ox homologs in rice (OsGA3ox1 and OsGA3ox2), and four GA3ox homologs in cucumber and pumpkin [52,53]. Several studies have confirmed functional changes in GA3ox mutations, such as DWARF1 in maize, which can lead to dwarfism, a 13 bp deletion in a GA3ox gene causing a GA-deficient dwarf phenotype in watermelon, and mutations to two gibberellin 3-oxidase genes, AtGA3ox1 and AtGA3ox2, that affect Arabidopsis development [18,51,54]. Our study showed that a single base mutation in GA3ox may influence its molecular function causing reduced active GAs. This study provides a new molecular basis for studying gibberellin-sensitive mutations.

Conclusions
We identified a dwarf mutant Si1 in pumpkin obtained by EMS mutagenesis of the inbred line N87. By BSA and subsequent fine mapping, we mapped the CmaSI1 locus to a 463 kb region on chromosome 8 that contained 28 annotated genes in the F 2 population. Only one nonsynonymous SNP in CmoCh08G006170 was obtained according to wholegenome resequencing of the two parents. CmoCh08G006170, a homolog of Arabidopsis gibberellin 3-beta hydroxylase (GA3ox), is a key enzyme in the regulation of bioactive gibberellins (GAs). RNA-seq analysis and qRT-PCR showed that the expression level of CmoCh08G006170 in stems of Si1 was changed compared with that of wild type. The dwarf phenotype could be restored by exogenous GA 3 treatment, suggesting that Si1 is a GA-deficient mutant. The above results demonstrated that CmoCh08G006170 may be the candidate gene controlling the dwarf phenotype.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/agronomy12081779/s1, Table S1: List of primers in this study; Table S2: Morphological characterization of Si1 and WT N87; Table S3: Genes annotated in the mapping interval.