Integrated Meta-QTL analysis and transcriptomic profiling reveal genomic regions for fruit quality, abiotic and biotic stress resilience in cucumber (Cucumis sativus L.)

Abstract

Cucumber is a globally significant vegetable crop whose production and market value are affected by fruit quality and resilience to diverse environmental stressors. Despite the identification of numerous Quantitative Trait Loci (QTL) over the last two decades, their direct application in breeding has been hindered by inconsistent genomic positions and broad confidence intervals. In this study, we conducted a comprehensive Meta-QTL (mQTL) analysis by integrating 647 initial QTLs from 40 independent studies published between 2003 and 2024. Using a high-density consensus map containing 9,299 markers, we projected 531 QTLs, identifying 38 robust mQTLs associated with fruit quality, biotic and abiotic stress tolerance. The identified mQTLs exhibited a significant reduction in the average confidence interval (CI) by 5.3-fold, compared to the average CI of the original QTLs and phenotypic variance explained values reaching up to 49.81% (mQTL 6.8). Our results identifying specific genomic hotspots on chromosomes 1, 3, 5, and 6 that harbor high-confidence candidate genes responsible for stress tolerance and fruit quality. Comparative analysis with seven independent genome-wide association studies validated 16 mQTL regions, confirming their stability across diverse genetic backgrounds. Biotic stress resilience was linked to immune regulators such as LRK10L2 and MLO-like protein 12, while abiotic stress tolerance was anchored by genes like NCED5 (cold), ClpB1 (heat), and MYB44-like (waterlogging). Furthermore, we identified key drivers of fruit quality, including Expansin-A4 and CNR2 for dimensions, CsWOX9 for epidermal spine initiation, and Hd3a for flowering phenology. Transcriptomic profiling provided robust expression support for these prioritized candidate genes within the target mQTL intervals. The markers linked to these genes serve as robust tools for marker-assisted selection and fine mapping, offering precise targets for the development of climate-resilient, high-quality cucumber cultivars.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12870-026-09349-1.

Introduction

Cucumber (Cucumis sativus L., 2n = 2x = 14), the fourth most globally important vegetable crop, serves as a model species for studying biological processes in the Cucurbitaceae family [1]. Originating in India’s southern Himalayan foothills, it has been cultivated for over 3,000 years and is primarily consumed as an immature fruit in salads [2]. With global production spanning two million hectares and yielding over 90 million tons annually, the cucumber is a vital economic crop [3]. Nutritionally, cucumber is highly valued for its high moisture content (approximately 95%) and low caloric density, containing minimal amounts of protein (0.6%), fat (0.1%), and carbohydrates (2.2%). It serves as a significant dietary source of essential vitamins, including thiamine, niacin, and vitamin C, as well as key minerals such as calcium, magnesium, phosphorus, and potassium [4–6]. Beyond its basic nutritional profile, the fruit is rich in dietary fiber and bioactive phytochemicals, particularly polyphenols, which contribute to its documented antioxidant, anti-inflammatory, and anti-carcinogenic properties [7–9]. Its dual role as a dietary staple and a source of health-promoting phytochemicals underscores its agricultural and nutritional significance worldwide. However, high-quality cultivation is increasingly compromised by the cucumber’s high susceptibility to environmental constraints. Productivity is threatened by a combination of biotic and abiotic factors, as well as the inherent genetic complexity of fruit quality traits [10].

Cucumber fruit quality is determined by a combination of commercial, nutrient, and flavor attributes, each influencing consumer preference and market value. Commercial quality encompasses external traits such as fruit size, shape, color, spine characteristics (density, size, color, distribution), surface wax presence, texture, firmness, and absence of defects, features critical for visual appeal and breeder prioritization [11–13]. Nutrient quality encompasses health-promoting components such as soluble solids, vitamins, and minerals. Recently, flavor quality has gained prominence as a key driver of consumer satisfaction, prompting advancements in sensory research to enhance taste profiles. Together, these traits commercial appeal, nutritional value, and flavor form an integrated framework for improving cucumber quality, balancing market demands, health benefits, and sensory enjoyment [14, 15]. Enhancing fruit architecture specifically shape, size, and pigmentation remains a primary objective in cucumber breeding programs. Key agronomic traits such as fruit weight, length, diameter, and the number of fruits per plant directly determine overall yield potential. Beyond yield, the commercial marketability of the crop, particularly for slicing cucumbers, is dictated by a suite of quality attributes. These include the fruit shape index (defined by length-to-diameter and length-to-peduncle ratios), skin characteristics such as spine morphology and color uniformity, and the presence or absence of bitterness [16, 17]. Precision in breeding for these parameters is essential to meet diverse regional consumer preferences and maximize the economic value of the harvest.

Cucumber faces diverse biotic and abiotic stresses that impair physiological functions like photosynthesis, membrane integrity, and nutrient uptake, ultimately compromising both yield and fruit quality [1]. Biotic stresses include fungal pathogens like downy mildew (Pseudoperonospora cubensis), powdery mildew (Podosphaera xanthii), Gummy stem blight (Stagonosporopsis cucurbitacearum), Fusarium wilt (Fusarium oxysporum f. sp. cucumerinum), scab (Cladosporium cucumerinum), and anthracnose (Colletotrichum orbiculare), bacterial diseases like, angular leaf spot (Pseudomonas syringae pv. lachrymans), bacterial wilt (Erwinia tracheiphila). In recent years, the impact of aphid-borne viruses such as Cucumber Mosaic Virus (CMV), Papaya Ring Spot Virus (PRSV), and Zucchini Yellow Mosaic Virus (ZYMV) has escalated, particularly in protected cultivation environments where they cause severe economic losses [18]. Simultaneously, cucumber is highly vulnerable to abiotic stresses such as waterlogging, high temperature, low light and cold injury, which disrupt critical cellular processes and impair growth. These challenges are increasingly exacerbated by climate change, as extreme temperature fluctuations and unpredictable precipitation patterns directly trigger oxidative stress and metabolic imbalances that lead to substantial yield instability [19–21]. Developing durable resistance requires a deep understanding of genetic inheritance and the molecular pathways that govern these traits. By identifying stable genomic regions and the underlying genes through targeted breeding and biotechnological interventions, breeders can synchronize stress resilience with high productivity and fruit quality.

Quantitative Trait Locus (QTL) mapping is a pivotal approach in crop improvement, aimed at linking specific genomic regions to quantitative traits of interest. The primary goal of QTL mapping is to identify and characterise the genetic markers associated with these traits, enabling breeders to select for desirable characteristics more efficiently [22, 23]. Despite the identification of hundreds of QTLs for various traits in cucumber over the last two decades, their practical application in Marker-Assisted Selection (MAS) remains significantly limited. This gap is primarily due to the large confidence intervals (CIs) associated with individual mapping studies, which often encompass hundreds of genes, making it difficult to pinpoint the exact causal variants. Furthermore, many of these reported QTLs exhibit low stability across diverse genetic backgrounds and fluctuating environments. Consequently, there is an urgent need to synthesize these fragmented findings into a high-resolution genomic framework to identify stable, narrow-interval hotspots that are reliable for breeding programs [24].

To overcome these limitations, Meta-QTL (mQTL) analysis has emerged as a superior computational strategy. It was developed by Goffinet and Gerber [25], integrates data from multiple QTL studies to identify consistent genomic regions linked to traits by constructing a consensus genetic map. This approach resolves discrepancies among individual studies, enhances mapping resolution, and refines QTL confidence intervals, enabling the prioritization of stable genomic regions and candidate genes for polygenic traits. The methodology involves aggregating heterogeneous datasets, projecting QTLs onto a unified map, and applying algorithms like Veyrieras’s to detect mQTL hotspots. By harmonizing fragmented QTL data, mQTL analysis improves reliability in localizing loci and uncovers pleiotropic regions critical for complex trait architecture.

Recently, mQTL analysis has facilitated the identification of consensus genomic regions associated with abiotic stress tolerance, biotic stress resistance, and quality traits across diverse crop species. For instance, Feizi et al. [26] delineated mQTLs governing grain quality attributes in rice, while Soriano et al. [27] conducted a comprehensive mQTL analysis in durum wheat, identifying loci linked to abiotic and biotic stress as well as quality parameters. In pearl millet, Gupta et al. [28] reported mQTLs associated with biotic stress and nutritional quality. Similarly, Abdelraheem et al. [29] characterized mQTLs that contribute to abiotic and biotic stress tolerance in cotton, underscoring the utility of mQTL frameworks for dissecting complex traits. Despite these advancements in cereals and legumes, a comprehensive meta-analysis integrating fruit quality and biotic resistance, and abiotic stress resilience in cucumber remains conspicuously lacking. To address this gap, we collected two decades of fragmented QTL data (2003–2024) to construct a high-density consensus map and delineate robust mQTLs. By validating these regions against recent GWAS studies and identifying high-priority candidate genes through in silico expression analysis, this study provides a precise genetic framework for developing the next generation of climate-resilient, high-quality cucumber cultivars.

Materials and methods

Data curation of the cucumber QTLs and search strategy

We conducted a comprehensive bibliographic search on the Web of Science (https://mjl.clarivate.com/search-results), Google Scholar (https://scholar.google.com/), and PubMed databases (https://pubmed.ncbi.nlm.nih.gov/) that reported QTLs linked to abiotic, biotic and fruit quality traits in cucumber. The search was executed using boolean operators with the following exact keyword combinations: (“cucumber” OR “Cucumis sativus”) AND (“QTL” OR “quantitative trait loci” OR “QTL mapping”) AND (“fruit quality” OR “fruit length” OR “fruit diameter” OR “downy mildew” OR “powdery mildew” OR “gummy stem blight” OR “cold tolerance” OR “heat tolerance” OR “salt tolerance” OR “waterlogging”). Studies were included if they: (i) investigated mapping populations derived from Cucumis sativus L. (e.g., F2, F2:3, RIL, BC, or DH), (ii) targeted agronomic traits belonging to fruit quality, biotic stress, or abiotic stress categories and (iii) provided genomic linkage maps with anchored molecular markers. Studies were excluded if they lacked common flanking or anchor markers required for alignment to the reference maps. Peer-reviewed original research articles published in English and Japanese were considered, with duplicate studies across databases identified and manually removed by cross-referencing authors, publication years, parentages, and journal names. Multiple QTLs detected within the same mapping population evaluated across multiple independent environments or distinct seasons were treated as independent statistical entries. This approach captures environmental stability and environmental interactions, which are critical for delineating robust mQTL hotspots. To ensure data homogeneity and phenotyping reliability across the compiled studies, stringent selection criteria were applied. Studies were exclusively included if they were published in peer-reviewed journals with clearly defined replicated experimental designs (greenhouse or field trials) and utilized standardized screening scales (e.g., specific disease severity index scoring for downy and powdery mildew).

A total of 44 independent studies from 2003 to 2024 that included different mapping populations i.e., F_2:3, RILs (Recombinant inbred lines), BC (Backcross) and DH (Double Haploids) were initially identified for preliminary screening. After careful reviewing, 4 studies were excluded because they lacked common anchoring markers required for map integration. Consequently, a finalized set of 40 independent studies containing 647 original QTLs was utilized for the downstream meta-analysis (Supplementary Table S1).

Each QTL mapping study was systematically reviewed to extract essential genetic parameters and metadata i.e., (i) the specific parental lines and mapping population, (ii) size of the population (iii) linkage group and chromosome number, (iv) the peak position and closely linked flanking markers, (v) the peak LOD scores, and (vi) the percentage of phenotypic variance explained (PVE/R²) by each distinct QTL. For QTLs where LOD scores and PVE were not reported, default values of 3 and 10% were assigned, respectively. When the peak position was missing, the midpoint between the two flanking markers was treated as the position. In cases where the confidence interval was not reported in the studies, it was estimated using the following equations for RILs, F_2:3 and BC populations [30, 31].

For RILs populations,

For F2:3 and BC populations,

where P refers to the size of the population and R² refers to the PVE.

Development of a comprehensive consensus map

The R package called ‘LPmerge’ [32] was used to combine the maps into a single consensus map following the steps described recently [33]. The following high marker density and reliable order genetic maps were employed as the reference maps in this study: (i) Ren et al. [34] designed cucumber linkage map which spans in total 573 cM with 995 SSRs, (ii) Yang et al. [35] A 1,681-locus cucumber consensus map, spanning 730.0 cM. iii) Miao et al. [36] 248 microsatellite loci into seven linkage groups spanning 711.9 cM. The consensus map was constructed by integrating data from independent QTL studies along with a reference maps. Marker order conflicts across independent linkage groups were resolved by setting the maximum interval parameter in LPmerge ranging from 1 to 4. The best consensus map was identified by selecting the one with the lowest Root Mean Square Error (RMSE) and the shortest length. To evaluate the structural accuracy and marker colinearity of the developed Cucumber 2025 Consensus Genetic Map, the consensus genetic positions (cM) of the integrated SSR markers were plotted against their corresponding physical positions (Mb) on the cucumber reference genome assembly (Chinese Long v3.0). The marker coordinates were determined by blasting the respective SSR flanking primer sequences against the reference genome using BLASTN with an E-value threshold of < 10^− 5. The relationship between genetic and physical positions across all seven chromosomes was visualized using a grid-scatter plot (Supplementary Fig. S1).

QTL projection and meta-analysis

Individual genetic maps and QTL details were formatted as text files and imported into BioMercator V4.2.3 [37] software (https://mybiosoftware.com/biomercator-genetic-maps-qtl-integration.html) for QTL projection. Projection accuracy was governed by two default thresholds: a flanking marker P-value homogeneity of 0.25 and a minimum map distance ratio of 0.50. Any initial QTLs failing to meet these parameters were excluded from further analysis.

Meta-analysis of the projected QTLs was performed for individual cucumber chromosome, using algorithms in BioMercator v4.2.3 [25, 38]. Two distinct approaches were employed based on number of projected QTLs per chromosome: Goffinet and Gerber [25] method was applied when the count of projected QTLs per chromosome was 10 or less than 10 and identifying the optimal model via the lowest Akaike Information Criterion (AIC) score. while the Veyrieras et al. [38] approach was utilized for QTL counts exceeding 10, where the optimal model was validated by achieving the lowest score across a majority (at least three) of five criteria: Bayesian information criterion (BIC), AIC, corrected AIC, AIC3, and average weight of evidence (AWE). Upon determining the most robust model, the position and confidence interval (CI) for each mQTL were defined. Finally, the representative LOD score and R² of each mQTL were estimated by averaging the corresponding metrics from its underlying original QTLs. A detailed workflow of the mQTL analysis pipeline implemented in this study is presented in Fig. 1.

Fig. 1.

Schematic workflow illustrating the integrated mQTL analysis and transcriptomic exploration strategy in cucumber

Candidate gene discovery within the mQTLs and their comparison with GWAS-MTAs

For candidate gene identification in the mQTL region, the primer sequences of the flanking markers corresponding confidence interval were retrieved and physical location was accessed by doing BLASTN search against cucumber (Chinese Long) reference V3 genome. Genes within this segment were annotated using the OmicsBox tool (v. 3.4) [39]. Using a cut-off of 1e-3, gene sequences were aligned against the National Center for Biotechnology Information (NCBI) non-redundant protein database (nr v5) using BLASTP to identify putative functions and homologous sequences. Subsequently, the InterProScan results were merged with the BLAST results [40]. Following this, we performed Blast2GO Mapping and Annotation using the default parameters. To further validate the identified mQTL regions, findings were compared with seven independent genome-wide association studies (GWAS) published between 2018 and 2025, focusing on abiotic stress, biotic stress, and fruit quality in cucumber. Because these studies utilized different reference genome assemblies, all significant marker-trait association (MTA) positions from earlier versions were converted to the 9930 v3.0 assembly coordinates using the UCSC liftOver (http://hgdownload.soe.ucsc.edu/admin/exe/linux.x86_64/liftOver). This remapping was performed using a high-precision chain file to ensure genomic consistency across datasets, thereby enabling direct spatial overlap of mQTL intervals with GWAS-derived MTAs.

Expression analysis of genes in the mQTL region

RNA-Seq datasets for fruit length, downy mildew, powdery mildew, and cold tolerance were retrieved from NCBI BioProject under the accession numbers PRJNA1005009, PRJNA544259, PRJNA321023, and PRJNA867170, respectively. Raw data were checked for quality and adapter trimming using FastQC 0.11.7 [41] and Trimmomatic 0.27 [42]. The resultant high-quality reads were mapped to the cucumber 9930_V3 reference genome using a splice-aware aligner HISAT2 2.1.0 [43] and StringTie was used to count gene-level fragments per kilobase per million mapped reads (FPKM) [44]. The differentially expressed genes (DEGs) were analyzed using the DESeq2 (Version 1.50.2) package with significance defined by thresholds of FDR ≤ 0.1, p < 0.05, and log₂ FC ≥ ± 1. Expression patterns were visualized using the pheatmap and tidyverse R packages. Complete metadata details for each BioProject including genotypes, tissues, treatments, and sample counts are provided in Supplementary Table S2.

Results

Bibliography search

A literature search on QTL mapping identified 647 QTLs from 40 studies across 34 mapping populations (Supplementary Table S1). Among the 40 studies, the majority were focused on biotic stress (45%, 18) and fruit quality traits (37.5%, 15), while abiotic stress accounted for the remaining 17.5% (7) (Fig. 2A). These studies involved populations of different sizes and types. Although most studies involved RIL populations, followed by F2 and F2:3 populations for QTL analysis. The population size varied from 69 to 480. The number of QTL per trait varied from 1 for scab resistance and vine length to 103 for the downy mildew (Table 1; Fig. 2D). These QTLs were located on 7 different cucumber chromosomes. The maximum number of QTLs was available on chromosome 1 (147), followed by 6 (124) and 5 (123); the minimum number of QTLs was available on chromosome 7 (15) (Fig. 2B and C).

Fig. 2.

a Proportional distribution of studies categorized by major trait type (Abiotic stress, Biotic stress, and Fruit Quality Traits). b Stacked bar chart illustrating the frequency of QTLs mapped on each of the seven melon chromosomes (Chr 1–7), with colors representing specific traits contributing to each chromosomal count. c Heatmap visualization of QTL density across all seven chromosomes for each trait. The color gradient (from light blue to magenta) indicates the number of mapped QTLs, highlighting high-density hotspots. d Horizontal bar plot ranking individual traits by their total number of detected QTLs

Table 1.

Summary of the 647 initial QTLs collected from literature, categorized by trait type (biotic stress, abiotic stress, and fruit quality)

S.No.	Trait Symbol	Trait name	QTL No.	Trait Type
1	ALS	Angular leaf spot	2	Biotic stress
2	Aroma	Aroma compounds	34	Fruit Quality Trait
3	CMV	Cucumber mosaic virus	2	Biotic stress
4	CT	Cold tolerance	35	Abiotic stress
5	DM	Downy mildew	103	Biotic stress
6	FD	Fruit diameter	59	Fruit Quality Trait
7	FFT	Fruit flesh thickness	19	Fruit Quality Trait
8	FGCM	Mature fruit ground color	2	Fruit Quality Trait
9	FL	Fruit length	96	Fruit Quality Trait
10	FLDR	Fruit length/diameter ratio	36	Fruit Quality Trait
11	FCC	flesh chlorophyll content	2	Fruit Quality Trait
12	FC	flesh color	2	Fruit Quality Trait
13	FEC	flesh extract color	2	Fruit Quality Trait
14	FLSR	Fruit length/stalk ratio	13	Fruit Quality Trait
15	FN	Fruit number	13	Fruit Quality Trait
16	FNL	Fruit neck length	5	Fruit Quality Trait
17	FPeL	Fruit pedicel length	8	Fruit Quality Trait
18	FPL	Fruit peduncle length	8	Fruit Quality Trait
19	FSC	Fruit Skin Color	6	Fruit Quality Trait
20	FSD	Fruit spine density	26	Fruit Quality Trait
21	FSL	Fruit stalk length	9	Fruit Quality Trait
22	FSN	Fruit skin netting	2	Fruit Quality Trait
23	FT	Flowering time	23	Fruit Quality Trait
24	FW	Fruit weight	19	Fruit Quality Trait
25	GSB	Gummy stem blight	13	Biotic stress
26	HT	Heat tolerance	3	Abiotic stress
27	LL	Low light tolerance	5	Abiotic stress
28	MYSV	Melon yellow spot virus	4	Biotic stress
29	OD	Ovary diameter	5	Fruit Quality Trait
30	OL	Ovary length	8	Fruit Quality Trait
31	PM	Powdery mildew	45	Biotic stress
32	SCD	Seed cavity diameter	26	Fruit Quality Trait
33	SR	Scab resistance	1	Biotic stress
34	ST	Salt tolerance	5	Abiotic stress
35	VL	Vine length	1	Fruit Quality Trait
36	WL	Waterlogging	5	Abiotic stress
	Total		647

Consensus map and QTL projection

A new consensus genetic map, designated the ‘Cucumber 2025 Consensus Genetic Map,’ was constructed using LPmerge by integrating three high-density reference maps with the individual linkage maps from all studies included in the meta-analysis. This map carried 9299 marker loci (mainly SNP = Single Nucleotide Polymorphism; SSR = Simple Sequence Repeat; DArT = Diversity Array Technology) (Supplementary Table S3).

The consensus map revealed considerable variation in genetic length and marker density across the seven linkage groups (Table 2). In total, 9,299 markers were mapped across a total length of 821.4 cM, resulting in an average density of 11.32 markers/cM. Chromosome 1 was the longest linkage group (150.9 cM), while Chromosome 7 was the shortest (88.7 cM). Chromosome 3 contained the highest number of markers (1,966) and the maximum density (15.80 markers/cM). Conversely, Chromosome 7 harbored the fewest markers (535) and the lowest density (6.03 markers/cM).

Table 2.

Characteristics of the high-density cucumber consensus genetic map, including chromosome lengths and marker densities

LG	Markers	Length	No. of QTLs
1	1582	150.9	147
2	922	133.8	106
3	1966	124.5	71
4	1373	90.0	61
5	1453	111.1	123
6	1468	122.5	124
7	535	88.7	15
Grand Total	9299	821.4	647

The projection of QTLs onto the newly developed consensus map was highly successful across all seven cucumber linkage groups, with LG1 displaying the highest density of both initial (147) and projected (135) QTLs (Fig. 3A). Genetic metrics reveal that most QTLs are characterized by narrow CI, with 318 QTLs falling in the 0–10 cM range, suggesting high mapping precision; however, some QTLs still exhibit higher CI (> 30 cM), which may be attributed to low marker density or the specific type of mapping population used (Fig. 3B). Furthermore, the majority of PVE values are concentrated in the 0–10% and 10–20% categories, indicating that most traits in this study are controlled by multiple genes, characteristic of polygenic inheritance.

Fig. 3.

Summary of QTL projection and frequency distribution of genetic metrics. A Comparison between the number of initial QTLs retrieved from literature (blue) and the number of QTLs successfully projected onto the newly developed consensus genetic map (orange) across the seven cucumber linkage groups. B Frequency distribution of initial QTL metrics based on classified ranges. The blue bars represent the distribution of Confidence Intervals (CI), and the orange bars represent the Phenotypic Variance Explained (PVE in %)

Of the 647 initial QTLs retrieved from the literature, 531 were successfully projected onto the newly constructed consensus map (Fig. 4). The remaining 116 loci could not be integrated due to insufficient marker information or excessively large confidence intervals (CI). Following the meta-analysis, 364 QTLs were retained based on a high membership probability (> 0.9) of clustering within a specific mQTL region. However, 18 of these were subsequently excluded because they formed singleton mQTLs (supported by only a single initial QTL) or lacked validation from at least two independent studies. Consequently, 346 distinct QTLs were utilized for downstream characterization. Of these, 205 QTLs were associated with fruit quality traits, 101 QTLs with biotic stress resistance, and 40 QTLs with abiotic stress tolerance.

Fig. 4.

Distribution of QTLs projected across 7 chromosomes in cucumber for different traits related to abiotic, biotic stress and fruit quality traits

mQTL analysis

A total of 38 consensus mQTLs were mapped across the seven cucumber chromosomes, representing key genomic regions involved in the regulation of multiple abiotic, biotic, and fruit quality traits. The number of QTL involved in individual mQTL ranged maximum with 34 QTLs in mQTL_5.4, followed by mQTL_3.5 (27), mQTL_1.8 (23), and mQTL_5.2 (23) (Table 3; Fig. 5A). The number of traits controlled by individual mQTL ranged from 10 traits involved for mQTL_1.8 and mQTL_5.2 each to at least two traits in each of mQTLs (Fig. 5C). The chromosome distribution of mQTLs differed, with 9 mQTLs on chromosome 1 and 3 mQTLs on chromosome 7 (Fig. 5B). Mean PVE % for individual mQTLs ranged from 7.20% to 49.81% (mean 16.85), the logarithm of odds (LOD) ranged from 3.20 to 33.55 (mean 8.52) and the CI for individual mQTLs ranged from a minimum 0.01 cM (mQTL_5.4) to a maximum on mQTL_7.1 (11.39 cM). The PVE by the mQTL ranged from 7.20% in mQTL_1.5, followed by mQTL_7.3 (7.25), mQTL_7.1(7.30), to 49.81% in one of the major mQTL, viz., mQTL_6.8, followed by mQTL_6.2 (40.30), mQTL_1.1 (34.76).

Table 3.

Detailed features of the 38 identified mQTLs, including their genomic positions, confidence intervals (CI), and associated traits

mQTLs	LG	CI (95%)	CI	Position (cM)	Left marker	Peak Marker	Right marker	Traits	N studies	NQTLs	LOD	PVE(%)
mQTL 1.1	1	13.07–13.59	0.52	13.33	SSR15108	Marker1_1514563	UW045196	LL, HT	2	5	16.88	34.76
mQTL 1.2	1	17.625–20.29	2.67	18.96	SSR03462	Marker328363	SSR16954	CT, DM, FT	3	3	8.11	24.77
mQTL 1.3	1	25.48–27.47	1.99	26.48	UW024133	SSR06534	SSR19673	FPL, FT, GSB, FL, FD	5	5	8.04	12.78
mQTL 1.4	1	34.42–36.26	1.84	35.34	SSR04195	Marker383257	SSR21054	FL, Aroma, FLDR	3	5	6.17	13.56
mQTL 1.5	1	42.53–44.54	2.01	43.54	SSR07597	UW029657	SSR17184	FLSR, FL	2	2	5.00	7.20
mQTL 1.6	1	46.19–48.35	2.16	47.27	SSR17793	UW084469	SSR23766	Aroma, FD	3	4	6.17	10.75
mQTL 1.7	1	56.06–56.59	0.53	56.33	SSR15739	IndelLTG32	Marker1_17750087	CT, FD	3	12	13.62	30.99
mQTL 1.8	1	71.93–74.41	2.48	73.17	UW375302	Marker386072	SSR01091	FW, DM, FL, PM, FD, Aroma, FLDR	10	23	5.11	15.28
mQTL 1.9	1	89.68–89.91	0.23	89.8	SSR23049	mk146	SSR17922	FN, FT, VL, Aroma, PM, FD, FL	6	16	6.83	15.19
mQTL 2.1	2	16.61–18.2	1.59	17.41	SSR11835	Marker2_853760	SSR11097	Aroma, DM, FL, FD	3	4	5.64	9.70
mQTL 2.2	2	25.4-27.46	2.06	26.43	SSR21090	SSR03084	SSR18491	Aroma, SR	2	3	5.95	13.47
mQTL 2.3	2	53.57–55.05	1.48	54.31	SSR20132	SSR20248	SSR04870	DM, CT, Aroma, FL	7	22	4.23	10.65
mQTL 2.4	2	81.705–85.05	3.35	83.38	SSR13275	SSR02322	SSR07278	SCD, FL	2	4	7.64	13.48
mQTL 2.5	2	104.3-105.39	1.09	104.89	SSR17108	Marker2_20682099	UW084446	DM, Aroma, FN, FD, GSB	6	9	3.93	12.31
mQTL 3.1	3	18.33–19.69	1.36	19.01	SSR16133	UW085255	SSR19339	FD, FL, GSB, DM, PM, WL	7	7	3.20	9.39
mQTL 3.2	3	42.51–43.19	0.68	42.85	SSR14821	Marker510939	UW084634	DM, FL, GSB, MYSV, DM, FLDR	6	8	6.89	17.66
mQTL 3.3	3	70.96–73.99	3.03	72.48	UW084172	UW085347	SSR12032	FL, ST	2	9	12.96	21.10
mQTL 3.4	3	80.63–86.47	5.84	83.55	SSR03261	SSR04446	SSR17264	FD, SCD, FL	2	3	4.05	8.97
mQTL 3.5	3	97.07–97.42	0.35	97.25	UW084839	SSR21304	SSR03777	FL, FW, SCD, FN, FSC, FLDR, DM, FD, FFT, FPL, FNL, PM, OL	7	27	7.88	12.46
mQTL 4.1	4	29.01–34.01	5.00	31.51	UW083926	Marker88985	UW130650	CT, FW, FN	2	8	3.73	10.89
mQTL 4.2	4	41.43–45.88	4.45	43.66	SSR17511	SSR03481	UW084421	OL, FSD, FL	2	5	7.13	11.92
mQTL 4.3	4	50.4-53.24	2.84	51.82	SSR13021	Marker15953	SSR02697	GSB, DM	3	4	10.65	17.25
mQTL 4.4	4	60.5-61.98	1.48	61.24	SSR13864	Marker72295	SSR26606	DM, PM	3	6	8.83	20.93
mQTL 5.1	5	19.15–21.58	2.43	20.37	SSR13693	SSR13693	UW084595	DM, FD, ALS, FL, FFT, PM	9	14	8.62	17.91
mQTL 5.2	5	33.56–34.42	0.86	33.99	SSR03341	UW085338	UW084357	PM, FCC, DM, FC, SCD, FEC, WL	10	23	11.34	25.56
mQTL 5.3	5	52.57–53.66	1.09	53.12	SSR06521	SSR17464	UW065321	OL, FL, FD, FFT, FSC, FSN, DM, PM, FN	7	14	14.74	24.64
mQTL 5.4	5	72.44–72.45	0.01	72.45	SSR18259	InDel82	UW061092	DM, PM, FL, FT, FLDR	8	34	14.67	19.85
mQTL 6.1	6	16.15–16.84	0.69	16.5	SSR01012	Marker180451	SSR03527	GSB, FN, ST	3	3	3.77	10.30
mQTL 6.2	6	23.12–27.57	4.45	25.35	SSR16020	Marker6_7115725	SSR07198	GSB, FT, FLDR	2	4	33.55	40.03
mQTL 6.3	6	37.79–38.28	0.49	38.04	SSR02385	UW083813	UW084257	PM, CMV	2	4	7.02	18.23
mQTL 6.4	6	44.81–46.83	2.02	45.82	SSR13884	SSR30353	SSR11798	FL, FD, FW	3	7	8.08	15.64
mQTL 6.5	6	56.37–61.08	4.71	58.73	SSR16710	UW083827	SSR01179	DM, FSD, FT, FL	4	5	9.25	13.68
mQTL 6.6	6	77.72–80.31	2.59	79.02	SSR17604	Marker3786	UW020559	FPL, FPL, ST, FSD, FNL, CT, FL, FN, FD, GSB, DM	9	16	7.05	16.36
mQTL 6.7	6	86.22–89.6	3.38	87.91	SSR02906	SSR07567	UW084969	GSB, DM, OL	3	3	3.40	7.63
mQTL 6.8	6	102.38-102.42	0.04	102.4	SSR00723	SSR14001	SSR05195	FSD, FL, DM	5	18	20.68	49.81
mQTL 7.1	7	14.74–26.13	11.39	20.44	SSR20042	SSR18326	SSR19291	FLDR, MYSV	2	2	4.90	7.30
mQTL 7.2	7	42.81–54.12	11.31	48.47	SSR19665	SSR11741	SSR22593	FD, FL, FN	2	3	4.80	10.83
mQTL 7.3	7	65.18–68.81	3.63	66.99	SSR05682	SSR33278	SSR10461	FFT, DM	2	2	3.25	7.25

Fig. 5.

Characterization of mQTLs and trait distribution. A. The number of initial QTLs projected into each identified mQTL. B. Frequency distribution of mQTLs identified on each of the seven cucumber chromosomes. C. The number of independent studies supporting each mQTL. D. Distribution of QTL counts associated with specific traits across the seven linkage groups (LG1–LG7)

To quantify the mapping resolution gained through meta-analysis, we evaluated the contraction of CIs across the data. The initial pool of 647 collected QTLs exhibited a mean CI of 13.78 cM and a median CI of 10.20 cM. For the 531 QTLs successfully projected onto the consensus map, the mean and median CIs were 14.56 cM and 11.67 cM, respectively. Following mQTL analysis, the 38 refined mQTL hotspots demonstrated a highly compressed genomic interval span, achieving an average mean CI of 2.58 cM and a median CI of 2.02 cM. Based on formula- Fold Reduction = Baseline Individual QTL CI/Refined mQTL CI), our meta-analysis achieved a 5.34-fold reduction in the mean confidence interval span (and a 5.05-fold reduction in the median CI).

To better understand the genetic architecture underlying the mapped traits, we also analyzed the percentage contribution of different phenotypic categories to each established mQTL. We identified 22 mQTLs were associated with fruit quality traits, while 7 mQTLs were associated with QTLs conferring tolerance to biotic stress. In contrast, only 4 mQTLs were linked to abiotic stress tolerance-related QTLs (Table 4; Fig. 6). Notably, mQTL_6.1 and mQTL_1.2 contributed equally to abiotic and biotic stress resistance and fruit quality traits, highlighting their potential genetic clustering due to tight linkage. In addition, mQTL_7.1, mQTL_7.3, and mQTL_3.1 were associated equally with both fruit quality and biotic stress tolerance traits, indicating their importance in the simultaneous improvement of these traits.

Table 4.

Contribution of the identified mQTLs showing the number of QTLs contributed by each trait category (Abiotic, Biotic, and Fruit Quality)

mQTLs	Traits	Fruit Quality	Biotic Stress	Abiotic Stress	NQTLs	Fruit Quality (%)	Biotic Stress (%)	Abiotic Stress (%)
mQTL 1.1	LL, HT	0	0	5	5	0.00	0.00	100.00
mQTL 1.2	CT, DM, FT	1	1	1	3	33.33	33.33	33.33
mQTL 1.3	FPL, FT, GSB, FL, FD	4	1	0	5	80.00	20.00	0.00
mQTL 1.4	FL, Aroma, FLDR	5	0	0	5	100.00	0.00	0.00
mQTL 1.5	FLSR, FL	2	0	0	2	100.00	0.00	0.00
mQTL 1.6	Aroma, FD	4	0	0	4	100.00	0.00	0.00
mQTL 1.7	CT, FD	1	0	11	12	8.33	0.00	91.67
mQTL 1.8	FW, DM, FL, PM, FD, Aroma, FLDR	15	8	0	23	65.22	34.78	0.00
mQTL 1.9	FN, FT, VL, Aroma, PM, FD, FL	15	1	0	16	93.75	6.25	0.00
mQTL 2.1	Aroma, DM, FL, FD	3	1	0	4	75.00	25.00	0.00
mQTL 2.2	Aroma, SR	2	0	1	3	66.67	0.00	33.33
mQTL 2.3	DM, CT, Aroma, FL	8	5	9	22	36.36	22.73	40.91
mQTL 2.4	SCD, FL	4	0	0	4	100.00	0.00	0.00
mQTL 2.5	DM, Aroma, FN, FD, GSB	5	4	0	9	55.56	44.44	0.00
mQTL 3.1	FD, FL, GSB, DM, PM, WL	3	3	1	7	42.86	42.86	14.29
mQTL 3.2	DM, FL, GSB, MYSV, DM, FLDR	3	5	0	8	37.50	62.50	0.00
mQTL 3.3	FL, ST	8	0	1	9	88.89	0.00	11.11
mQTL 3.4	FD, SCD, FL	3	0	0	3	100.00	0.00	0.00
mQTL 3.5	FL, FW, SCD, FN, FSC, FLDR, DM, FD, FFT, FPL, FNL, PM, OL	25	2	0	27	92.59	7.41	0.00
mQTL 4.1	CT, FW, FN	2	0	6	8	25.00	0.00	75.00
mQTL 4.2	OL, FSD, FL	5	0	0	5	100.00	0.00	0.00
mQTL 4.3	GSB, DM	0	4	0	4	0.00	100.00	0.00
mQTL 4.4	DM, PM	0	6	0	6	0.00	100.00	0.00
mQTL 5.1	DM, FD, ALS, FL, FFT, PM	4	10	0	14	28.57	71.43	0.00
mQTL 5.2	PM, FCC, DM, FC, SCD, FEC, WL	15	7	1	23	65.22	30.43	4.35
mQTL 5.3	OL, FL, FD, FFT, FSC, FSN, DM, PM, FN	11	3	0	14	78.57	21.43	0.00
mQTL 5.4	DM, PM, FL, FT, FLDR	9	25	0	34	26.47	73.53	0.00
mQTL 6.1	GSB, FN, ST	1	1	1	3	33.33	33.33	33.33
mQTL 6.2	GSB, FT, FLDR	3	1	0	4	75.00	25.00	0.00
mQTL 6.3	PM, CMV	0	4	0	4	0.00	100.00	0.00
mQTL 6.4	FL, FD, FW	7	0	0	7	100.00	0.00	0.00
mQTL 6.5	DM, FSD, FT, FL	4	1	0	5	80.00	20.00	0.00
mQTL 6.6	FPL, FPL, ST, FSD, FNL, CT, FL, FN, FD, GSB, DM	11	2	3	16	68.75	12.50	18.75
mQTL 6.7	GSB, DM, OL	1	2	0	3	33.33	66.67	0.00
mQTL 6.8	FSD, FL, DM	16	2	0	18	88.89	11.11	0.00
mQTL 7.1	FLDR, MYSV	1	1	0	2	50.00	50.00	0.00
mQTL 7.2	FD, FL, FN	3	0	0	3	100.00	0.00	0.00
mQTL 7.3	FFT, DM	1	1	0	2	50.00	50.00	0.00
	Total	205	101	40	346

Fig. 6.

High-density consensus map projecting 38 mQTLs associated with abiotic, biotic stress and fruit quality. The chromosome displays a marker density heat map derived from 9,299 markers. Horizontal bars to the right of the chromosomes denote 95% confidence intervals of projected mQTLs, color-coded by trait category: Abiotic Stress (blue), Biotic Stress (red), Fruit Quality (green), and Shared Traits (purple). Positions are shown in centimorgans (cM) as indicated by the ruler on the left

mQTL validated with previously reported GWA studies

The physical locations of the nearest flanking markers for each mQTL were obtained by reviewing the literature or by performing BLASTn of the primers or marker sequence against the cucumber genome. The physical intervals of the mQTL ranged from 0.09 Mb (mQTL_6.1) to ∼ 3.34 Mb (mQTL_7.2). A literature search identified 642 MTAs related to abiotic, biotic and fruit quality traits from seven GWA studies. A comparative analysis integrating seven independent GWA studies related to abiotic, biotic, and fruit quality traits revealed co-localization of multiple mQTL genomic regions with previously reported MTAs (Supplementary Table 4). A total of 16 mQTL were found to cover the physical locations associated with traits as reported by different GWA studies. Notably, the integrated dataset unveiled several exact trait matches where the GWAS-derived MTA and the mQTL shared identical phenotypic definitions. For example, mQTL_2.5 (20.42–20.76 Mb) fully encapsulated a MTA for GSB resistance located at 20.51 Mb, perfectly matching its projected internal QTL profile (GSB, DM, FD, FN, Aroma). Similarly, the major downy mildew/powdery mildew hotspot mQTL_5.1 (3.31–4.96 Mb) directly contained a distinct DM resistance MTA at 4.25 Mb, while mQTL_6.2 (5.46–6.39 Mb) bordered a GSB resistance MTA at 5.45 Mb, and mQTL_6.6 (24.30–24.64 Mb) aligned with a fruit length (FL) MTA at 24.89 Mb. Overall, these co-localization results significantly strengthen the reliability and structural validity of the identified mQTLs.

Candidate genes underlying mQTL and in silico gene expression analysis

The identified mQTLs demonstrated a promising distribution of PVE values. Out of the 38 identified mQTLs, 9 mQTLs demonstrated major genetic effects (four exhibiting a PVE > 30% and five ranging between 20 and 30%), while the majority (22 mQTLs) explained a solid 10–20% of the variance, leaving only a small fraction (7 mQTLs) below 10%. Furthermore, high-confidence mQTLs were selected for in-depth investigation and candidate gene identification based on three rigorous criteria: (i) the consolidation of at least three initial QTLs, (ii) a LOD score ≥ 5, and (iii) a PVE value > 15%. A total of 19 mQTLs fulfilled these criteria and were subjected to downstream candidate gene identification (Supplementary Table 5). To provide expression support for the candidate genes residing within these mQTLs and further investigate the molecular basis of diverse traits in cucumber, we integrated transcriptomic evidence from multiple RNA-Seq datasets. These datasets, retrieved from NCBI BioProjects, encompassed four key experimental conditions i.e., fruit length, Downy mildew infection, Powdery mildew infection, and cold tolerance.

Based on this trait-specific categorization, we conducted targeted in silico expression analyses to provide expression support for the underlying candidate genes. Specifically, only genes residing within the genomic boundaries of the assigned abiotic, biotic, or fruit quality mQTLs were queried against their respective, trait-matching RNA-Seq datasets. Candidate genes within these isolated regions were filtered based on significant differential expression. The normalized expression values of these DEGs were subsequently visualized using clustering heatmaps (Fig. 7). These heatmaps revealed distinct expression patterns across the different experimental conditions, allowing for the robust identification and prioritization of the most highly responsive candidate genes governing each specific trait category.

Fig. 7.

Transcriptomic profiling of prioritized candidate genes within trait-specific cucumber mQTL intervals. A Fruit Length: Spatial profiling in developing fruit tissue (15 days after flowering) comparing contrasting recombinant inbred lines RIL-67 (long fruit) and RIL-190 (short fruit) from a ‘PI183967’ and ‘XinTaiMiCi’ population. B Cold Tolerance: Temporal responses in leaf tissues of inbred line ‘JINYAN 4’ under low-temperature chilling at 0 h (control), 3 h, 12 h, and 24 h post-treatment. C Powdery Mildew: Pathogen-induced kinetics in leaf tissues of resistant segment substitution line SSL508-28 and susceptible recurrent parent D8 at 0 h (control) and 48 hpi with Podosphaera xanthii. D Downy Mildew: Interaction dynamics in leaves of resistant near-isogenic lines (NILs) DM412 and DM413 (PI 197088 introgression) versus susceptible recurrent parent HS279 at 0 dpi (Mock) and 3 dpi with Pseudoperonospora cubensis

Candidate genes with expression support identified within the refined biotic stress-associated mQTLs represented several key functional classes, including transcription factors (MYB, bHLH, WRKY, NAC, and TCP), receptor-like kinases (RLKs), and proteins involved in hormonal regulation and ubiquitination. Key genes identified through functional annotation included a cluster of Cysteine-rich Receptor-Like Kinases (CRKs), MLO-like protein 12, and Leaf Rust 10 Kinase (CsLRK10L2). Transcriptomic profiling provided strong expression support for these intervals, identifying significant differentially expressed genes directly involved in host defense. Under DM stress, prominent responsive candidates within these regions included hypersensitive-induced response protein 4, the calcium-sensor CML44, the exocyst complex component EXO70H1, and E3 ubiquitin-protein ligase RZF1, which collectively coordinate localized cell death and structural fortification. For PM resistance, significant genes identified in response to Podosphaera xanthii infection included the hydrolytic enzyme glucan endo-1,3-beta-glucosidase, major stress-responsive transcription factors (TCP7, ZAT12, RAP2-7, and ERF014), and key signaling components such as MDIS1-interacting receptor-like kinase 2 and Leaf Rust 10 disease-resistance locus receptor-like kinase. Furthermore, multiple uncharacterized loci and detoxification proteins (e.g., protein DETOXIFICATION 10) exhibited significant differential expression, reinforcing the importance of the complex RLK-mediated and metabolic signaling hubs identified in the overall mQTL analysis.

Similarly, abiotic stress-associated regions harboured genes implicated in physiological resilience to climate extremes. Notably, functional annotation highlighted putative candidate genes for heat, cold and low-light tolerance physically localized within these high-confidence intervals, including Chaperone protein ClpB1 (Hsp100), Enolase 1, WEAK CHLOROPLAST MOVEMENT UNDER BLUE LIGHT-like 1, NCED5, Ornithine aminotransferase (OAT), and DEAD-box RNA helicase 42, Protein TONSOKU, MYB44-like, and Omega-hydroxypalmitate O-feruloyl transferase. Transcriptomic profiling provided robust expression support for these specific genomic hotspots, identifying highly responsive genes associated with environmental adaptation. Analysis of cold-tolerance expression datasets revealed significant transcriptional shifts for several key regulatory and metabolic candidates within these regions. Prominent responsive genes included major stress-signaling transcription factors (TCP14, WRKY 72 A, bHLH47, and NAC100) and cyclic dof factor 3. Furthermore, the low-temperature response triggered the strong expression of structural and metabolic defenders, including hypersensitive-induced response protein-like 1, geranylgeranyl diphosphate reductase, isochorismate synthase 2, and chlorophyll a-b binding protein P4, which collectively act to maintain cellular homeostasis and photosynthetic integrity under chilling stress.

Functional annotation of fruit-quality-associated mQTLs identified candidate genes across categories of cell wall modification, proliferative regulation, sensory metabolism, and reproductive phenology. Key candidates localized within these refined genomic intervals included Expansin-A4, Cullin-1, Cellulose Synthase A (CesA2/4), RING finger protein 10, CNR2, and AGAMOUS. Furthermore, the analysis highlighted NCED6, Phytoene Desaturase, Hd3a, CsFPA, Hexokinase-3, CsWOX9, and HEC3-like as primary candidates within the fruit-related mQTL regions. Transcriptomic profiling of developing fruit tissue provided robust expression support for these genomic hubs, identifying significant differentially expressed genes actively driving longitudinal growth. Prominent responsive candidates within the targeted genomic regions included the major organ size determinant cell number regulator 5, along with structural and cytoskeletal coordinators such as villin-1, patellin-4, and beta-glucosidase 24. Other significant candidates supported by differential expression during fruit elongation included critical signaling and transcriptional regulators, namely serine/threonine-protein kinase PBL9, transcription factor SRM1, fatty alcohol caffeoyl-CoA acyltransferase, and protein-degradation machinery including specific E3 ubiquitin-protein ligases (RHC1A and XBAT31).

Pathway enrichment

A comprehensive pathway enrichment analysis was performed across all trait categories, yielding highly distinct biological profiles (Supplementary Fig. 2). For candidate genes residing within biotic stress-associated mQTL intervals, the analysis revealed a sharp overrepresentation of cell wall safeguarding pathways, specifically driven by xylanase inhibitor l-like functional classes, xylanase inhibitor C-terminal domains, and xylanase inhibitor N-terminal domains, which were accompanied by enzymatic degradation networks like the aspartic peptidase A1 family, peptidase family A1 domain, and the aspartic peptidase domain superfamily, as well as active biochemical defenses including the phytoalexin metabolic process and the toxin biosynthetic process. Concurrently, the genomic intervals governing abiotic stress tolerance demonstrated a tightly coordinated focus on cellular detoxification and redox homeostasis, dominated by glutathione S-transferase (GST) C-terminal domains, glutathione S-transferase Omega/Tau-like signatures, plant-specific glutathione S-transferases Tau C-terminal alpha helical domains, and glutathione S-transferase N-terminal domains, which were complemented by Vitamin C (ascorbate) metabolism, glutathione conjugation pathways, the thioredoxin-like superfamily, and vascular transport processes. Mirroring the complex nature of organ development, fruit quality mQTL regions involved massive gene counts dedicated to core regulatory infrastructure within the nucleus, including nucleic acid binding, macromolecule modification, and the tetratricopeptide-like helical domain superfamily, operating alongside growth-related structural components such as plant galacturonosyltransferase GAUT, receptor-like protein 23-like markers, ubiquitin homologues, and berberine/berberine-like domains.

Discussion

Cucumber productivity and fruit quality are complex polygenic traits governed by intricate genetic architectures and significant environmental interactions. mQTL analysis serves as a powerful computational strategy to synthesize fragmented genetic data from multiple independent studies, transforming disparate results into a cohesive genomic framework [38, 45]. While traditional QTL mapping has been instrumental in identifying regions associated with complex traits, its practical utility is often limited by large confidence intervals, population-specific results, and environmental instability. By projecting these diverse QTLs onto a single, high-density consensus map, mQTL analysis significantly refines the genomic landscape, narrowing broad intervals often from over 20 cM down to less than 1 cM and pinpointing stable loci that remain consistent across different genetic backgrounds and environments. This integration not only enhances the precision of marker-assisted selection but also facilitates the discovery of pleiotropic hotspots and high-priority candidate genes. Ultimately, mQTL analysis bridges the gap between raw statistical associations and functional genomics, providing breeders with the reliable molecular tools needed to develop high-yielding, resilient crop varieties.

In this study, a meta-analysis was initiated with 647 initial QTLs compiled from 40 independent experiments conducted between 2003 and 2024. Of these, 531 QTLs were successfully projected onto the consensus map, whereas 116 QTLs were excluded due to a lack of sufficient shared anchoring markers or inadequate genetic map connectivity required for precise coordinate projection. The 531 projected QTLs were then subjected to rigorous statistical and biological filtering to isolate high-confidence consensus regions. First, 364 QTLs were retained because they met the strict threshold of a > 0.9 membership probability, while 167 unstable, environment-specific QTLs were discarded as background noise. Second, an additional 18 QTLs were excluded because they formed singleton clusters supported by only a single initial experiment. This stringent refinement pipeline successfully eliminated spurious and non-reproducible loci, ensuring that the final consensus mQTLs represent highly robust, multi-environment genetic hotspots. By integrating these, we identified 38 robust mQTLs that represent stable genomic regions for simultaneous improvement of fruit quality and stress resilience. The meta-analysis demonstrates significant enhancement of the genomic landscape, revealing mQTLs with a mean PVE of 16.85% and refined confidence intervals as narrow as 0.01 cM (mQTL_5.4). This precision represents a marked improvement over the wider intervals (10–20 cM) found in original mapping studies. The distribution of these genetic regions was highly uneven, with the majority clustering together on chromosomes 1, 5, and 6. This indicates that these three chromosomes carry the highest number of both the original QTLs and the newly identified mQTLs. This non-random clustering likely reflects evolutionary selection pressures during cucumber domestication, where gene complexes controlling essential traits such as disease resistance and fruit quality traits have been conserved as functional genomic blocks [46, 47]. Conversely, the scarcity of mQTLs on chromosome 7 suggests a potential bottleneck in genetic diversity or a historical lack of marker density in that region, highlighting a target for future exploration [34, 48]. The validation of 16 mQTLs against recent GWAS-derived MTAs and the confirmation of candidate gene activity through RNA-Seq further confirm that these regions are not just statistical artifacts, but functional regulatory hubs essential for the next generation of precision breeding in cucumber. By combining linkage mapping, meta-analysis, and GWAS, we established a reliable framework for identifying key genetic regions that can simultaneously improve both stress resilience and fruit quality.

Biotic stress

Globally, biotic stress significantly limits cucumber productivity, resulting in substantial yield losses estimated between 50 and 80% [49–51]. This vulnerability is driven by a diverse complex of fungal and oomycete pathogens, which represent the most critical biological threats to the crop. These infections typically manifest as widespread foliar necrosis, fruit rot, or complete vascular collapse, resulting in devastating economic consequences [52]. Furthermore, the crop exhibits high susceptibility to viral pathogens that induce systemic stunting and fruit malformation, as well as bacterial agents that promote rapid tissue degradation and vascular failure, collectively compromising marketable yield.

Cucurbit DM, caused by the obligate oomycete Pseudoperonospora cubensis, and PM, primarily induced by the fungus Podosphaera xanthii, are the two most economically devastating foliar diseases affecting cucurbit production globally [53, 54]. The present study reveals significant colinearity between DM and PM resistance loci, as the identification of major mQTLs across 1, 4, and 5 chromosomes suggests that the underlying genes may be tightly linked or share a common genetic basis. This observed linkage is consistent with previous genomic reports in cucumber [55–57], further supporting the potential for synchronized resistance breeding.

The resistance hub on chromosome 5 (mQTLs 5.2 and 5.4) and chromosome 4 (mQTLs 4.3 and 4.4) exemplifies this duality. On chromosome 5, mQTL 5.2 harbors a homolog of the Arabidopsis DMR6 (2’-deoxymugineic-acid 2’-dioxygenase), a negative regulator of immunity; its loss-of-function is known to boost salicylic acid-mediated defense. This region also contains an inactive receptor-like protein kinase (XP_004148196.1) and Omega-3 fatty acid desaturase (NP_001274404.1), suggesting a coordination of JA signaling to restrict the colonization of both P. cubensis and P. xanthii [54, 58–60]. Similarly, mQTL 5.4 contains a Nudix hydrolase (XP_004135122.1), a susceptibility factor that, when modulated, can arrest fungal effector activity [59, 61, 62]. These S-genes are complemented by dominant recognition systems. In mQTL 5.4, primary sensing is anchored by LRR-RLK kinases (XP_011655784.1), while mQTL 4.4 (aligning with the dm4.1 locus) is mediated by the Leaf Rust 10 Receptor-Like Kinase (CsLRK10L2, XP_016899874.1) [56, 61]. Further reinforcing this defense-in-depth are the G-type Lectin Receptor Kinases (XP_011653741.1) and BON1-associated protein 2 (XP_004144363.3) in mQTL 4.3, which maintain basal immunity [60, 63, 64]. Additionally, recessive alleles of nutrient transporters, such as CsAAP2A (XP_011653779.1) and sulphate transporter 1.3-like (XP_004150525.1), suggest a “nutrient starvation” strategy, restricting metabolic resources available to the pathogen [60, 63–65].

Beyond recognition, the identified mQTLs facilitate an active counterattack by genome-wide enrichment of hydrolytic enzymes. We identified a striking cluster of glucan endo-1,3-beta-glucosidases across multiple loci: mQTL 5.1 (XP_004147695.2), mQTL 1.9 (XP_004139045.1, XP_011660145.1), and mQTL 3.2 (endo-1,3;1,4-beta-D-glucanase, XP_004134157.1). These Pathogenesis-Related (PR-2) enzymes provide a conserved defense mechanism by directly hydrolyzing the beta-glucan-rich cell walls of P. cubensis and P. xanthii [66, 67]. In mQTL 1.9, this enzymatic activity is likely synergized by a calcium-dependent protein kinase 8 (XP_011660135.1), which translates pathogen-induced Ca + 2 signatures into downstream ROS production [66–68]. To prevent pathogen ingress, these enzymatic defenses are coupled with cell wall reinforcement. In mQTL 6.3, Cinnamate 4-monooxygenase (NP_001292707.1) and Cell wall integrity component 1 (XP_004139940.2) work in tandem to synthesize lignin barriers. This is supported by Dirigent Proteins (XP_004147726.1) and CsEXO70B1 (XP_004150962.1) in mQTL 5.1, which facilitate the transport of cell wall components [66, 67, 69]. Furthermore, mQTL 6.3 contains the susceptibility factor Callose Synthase 12 (PMR4) (XP_004140034.1), while mQTL 5.4 harbors pectinesterase inhibitors (XP_004135375.1) that protect the host wall from fungal degradation enzymes [70, 71].

Within the multi-pathogen resistance hub mQTL_3.2, peptide methionine sulfoxide reductase B5 (XP_004134168.1) and glutamine synthetase (XP_004134161.1) were specifically identified as primary candidate genes driving biotic stress tolerance by coordinating metabolic energy reallocation and repairing infection-induced oxidative damage [66, 67, 72]. For viral defense, this region contains DEAD-box RNA helicase 58 (XP_011650984.1), a key component in sensing viral RNA and triggering RNA interference [73, 74]. This antiviral layer is further supported by ERF054-like transcription factors (XP_031742566.1) in mQTL 6.3 and MATE efflux pumps (XP_004135382.3) in mQTL 5.4, which extrude toxins and restrict systemic viral movement [75, 76].

Abiotic stress

Abiotic stress presents a substantial barrier to global cucumber production, as this thermophilic plant is the most environmentally sensitive member of the Cucurbitaceae family. A recent study spanning 1960 to 2024 indicates that temperature fluctuations, low light, and water imbalances profoundly disrupt the crop’s physiological homeostasis, necessitating the breeding of hardy cultivars [77]. Our findings identify chromosome 1 as a major genomic hotspot for environmental adaptation, coordinating a multi-faceted survival strategy. Cold tolerance is predominantly regulated by mQTL 1.7 and mQTL 1.2. mQTL 1.7 functions as a hormonal signaling hub via NCED5 (XP_004149672.1), the rate-limiting enzyme in Abscisic Acid (ABA) biosynthesis that triggers cold-responsive (COR) gene expression [78]. This integration highlights a clear physiological cross-talk, where ABA-dependent signaling networks converge with the canonical ICE1-CBF/DREB transcriptional cascade to orchestrate systemic cryoprotection. This signaling is complemented by metabolic reinforcements in mQTL 1.2, where Ornithine aminotransferase (OAT, XP_004137365.1) drives proline accumulation as an osmoprotectant, and DEAD-box RNA helicase 42 (XP_011649205.1) acts as a molecular chaperone to maintain protein synthesis at low temperatures [79–81]. Additionally, Phosphomethylethanolamine N-methyltransferase (XP_004137342.1) and Superoxide dismutase (SOD, XP_011649383.1) maintain membrane fluidity and scavenge ROS [82, 83]. Long-term genome integrity under these stresses is further preserved by Protein TONSOKU (XP_011657766.1) in mQTL 6.6 [84, 85].

Under heat and low-light stress, mQTL 1.1 emerges as a key regulatory hub. High-temperature resilience is mediated by Chaperone protein ClpB1 (Hsp100, XP_004138316.1) and Enolase 1 (XP_004138202.1), which stabilize the cytoskeleton and refold denatured proteins, while RbohF (XP_011648418.1) generates ROS signals to trigger the systemic heat shock response [86–88]. Simultaneously, this locus optimizes photon capture in shaded environments through Chlorophyll a-b binding protein P4 (XP_004138171.1) and WEAK CHLOROPLAST MOVEMENT UNDER BLUE LIGHT-like 1 (XP_011660213.1) [89, 90]. Salt tolerance is governed by a synchronized detection-and-transport system in mQTL 3.3 and mQTL 1.7. In mQTL 3.3, the master regulator ICE1 (XP_011651587.1) and the calcium-sensor CML19 (XP_004145714.1) detect ionic imbalances to activate whole-plant defense [91–93]. This is supported by the HKT1 sodium transporter (XP_004149678.1) in mQTL 1.7, which prevents cellular toxicity by maintaining the optimal Na+/K+ ratio [94, 95]. Mechanistically, while the master regulator CsICE1 represents a shared network node that coordinates upstream low-temperature and generalized osmotic responses, the recruitment of CsHKT1 signifies a highly specialized, stress-specific ionic adaptation. This specific transporter works in conjunction with the classic Salt Overly Sensitive (SOS) pathway to physically restrict Na+ influx into photosynthetic sinks, distinguishing it from non-specific drought or osmotic adjustments. Finally, mQTL 5.2 orchestrates survival under waterlogging through structural and metabolic adaptation. Omega-hydroxypalmitate O-feruloyl transferase (XP_004148179.1) promotes the formation of root suberin barriers to prevent radial oxygen loss. This is coordinated with Transcription factor MYB44-like (XP_004148201.1), which regulates stomatal closure and ROS scavenging; notably, MYB44 has been identified as a central hub for waterlogging tolerance in other crops, including barley [96]. To counteract iron-mediated chlorosis and membrane damage in submerged conditions, this region also recruits Ferritin-3 (XP_011655010.1) and Omega-3 fatty acid desaturase (NP_001274404.1) [97–99].

Fruit quality traits

The study identified 207 QTLs associated with fruit quality, with a pronounced clustering on chromosomes 1, 3, and 6 [51]. FL and FD are regulated by a complex interplay between cell proliferation and structural expansion. mQTL 6.4 acts as a central hub for longitudinal growth, where Expansin-A4 (XP_004143121.1) provides the enzymatic engine for wall loosening, balanced by ACO4 (NP_001295778.1), which modulates growth cessation via ethylene biosynthesis [100, 101]. This elongation is further driven by hormonal gradients established by Auxin F-BOX 2 (NP_001267557.1) in mQTL 3.3 and the PIN3 efflux carrier (NP_001267715.1) in mQTL 5.3, which facilitate polar auxin transport. Additionally, DET2 (XP_011655823.1) in mQTL 5.4 supplies the brassinosteroids necessary for cell stretching.

This expansion phase is preceded by a critical proliferative window regulated by RING finger protein 10 (XP_004145702.1) in mQTL 3.3, which functions similarly to the Sf1 locus as a cell division brake. Total cell count is further determined by fw2.2 homologs like CNR2 (XP_011656262.1) and CNR5 (XP_004134741.1), while AGAMOUS (NP_001292697.1) and SOC1 set the initial carpel identity and ovary thickness [102–105]. Sink strength, essential for fueling this growth, is maintained by AGPase (XP_004134146.2) in mQTL 3.2, which regulates starch biosynthesis. Radial expansion and shape are refined by tissue-patterning factors. Scarecrow-like protein 34 (XP_004143205.1) in mQTL 6.4 regulates radial cell division, while Cellulose Synthase A (CesA2) (XP_004141979.1) in mQTL 6.6 dictates the orientation of cellulose microfibrils. Structural integrity is maintained by Pectinesterase Inhibitors (XP_004135547.1) in mQTL 1.8 and WAT1-related proteins in mQTL 6.6, which coordinate the stretching of the fruit neck and peduncle. Finally, OFP5 (XP_011658258.1) and SPL13A (XP_004148013.1) act as developmental timing regulators to determine final organ shape [106, 107].

Flowering time and fruit number are governed by a sophisticated integration of photoperiod and metabolic status. mQTL 1.9 harbors HEADING DATE 3 A (Hd3a) (NP_001292686.1), an FT ortholog that controls the vegetative-to-reproductive transition [47]. This pathway is metabolically gated by Hexokinase-3 (CsHXK3) (XP_004139044.1) and TRAUCO (XP_004139041.1), which ensure flowering initiates only when carbohydrate thresholds are met. Independent of day length, the Flowering time control protein FPA (CsFPA) (XP_004153439.1) in mQTL 6.2 promotes earliness via the autonomous pathway. Final yield potential is modulated by ARR9 (XP_004136644.1) in mQTL 3.5, which integrates cytokinin signaling to regulate ovary size, and RALF-like 19 (XP_004148669.1) in mQTL 6.6, which influences fruit set success. This reproductive effort is supported by MUTE (XP_011650991.1) in mQTL 3.2, which optimizes stomatal density to fuel developing sinks, and TONSOKU (XP_011657766.1), which maintains floral meristem competence [108, 109].

The spine density and aroma profile of the fruit is determined by genes anchored in mQTL 6.6 and mQTL 6.8. Spine initiation is driven by WUSCHEL-related homeobox 9 (CsWOX9) (XP_004134676.1) in mQTL 6.8, a positive regulator of multicellular trichomes. Spine texture and shape are further modified by Trichome Birefringence-like (XP_004134666.1) and HEC3-like (XP_004148648.2), supported by the structural framework of CesA2 [110–112]. Pigmentation and aroma are synthesized through metabolic hubs in mQTL 3.5 and mQTL 2.3. Phytoene Desaturase (XP_004136658.1) in mQTL 3.5 acts as a bottleneck in carotenoid biosynthesis, determining skin color intensity. This is complemented by NCED6 (XP_004150345.1) in mQTL 2.3, which cleaves carotenoids into volatile apocarotenoids, and ACO6 (XP_004142180.1), which triggers ethylene-dependent ripening. Additionally, 4-coumarate-CoA ligase (4CL) (XP_004135516.2) in mQTL 5.3 influences lignin-based netting and flavonoid pigmentation [113, 114].

Conclusion

The present study successfully integrated two decades of quantitative genetic data to resolve the complex genomic architecture of cucumber by refining 647 fragmented QTLs into 38 robust mQTLs, thereby narrowing the search space for key regulatory genes and reducing confidence intervals from broad genomic spans (> 20 cM) to precise sub-centimorgan regions, achieving a 5.3-fold mean interval reduction. This comprehensive meta-analysis revealed a non-random distribution of functional loci, identifying evolutionarily conserved breeding hotspots on chromosomes 1, 5, and 6 that harbor genes essential for domestication and adaptation, including architecture for biotic stress where receptor-like kinases (LRK10L2) work in concert with hydrolytic enzymes, as well as key abiotic signaling hubs (NCED5, ICE1) that coordinate physiological homeostasis. Among the novel candidates in this study, EXO70H1, DETOXIFICATION 10, MDIS1-interacting receptor-like kinase 2, and an inactive receptor-like protein kinase provide prioritized expression support for biotic stress resilience against downy and powdery mildew infections. Environmental adaptation under cold stress is highlighted by the robust expression shifts of the universal stress protein YxiE, while LONGIFOLIA 1, patellin-4, and villin-1 emerge as high-priority candidates governing fruit quality traits and longitudinal tissue elongation. Ultimately, this integrated framework provides a strategic advantage for crop improvement, offering valuable genomic resources and tightly linked flanking markers that can be deployed for downstream validation in diverse populations and future high-resolution fine-mapping efforts to develop cucumber cultivars capable of withstanding erratic climatic and biotic pressures.

Authors’ contributions

MAI and SJ conceptualized and designed the study; BSR and SS conducted the data curation and formal analysis; BSR wrote the initial draft of the manuscript; SJ and SSD reviewed and edited the manuscript. All authors contributed to the article and approved the submitted version.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information file.

Declarations

Ethics approval and consent to participate

The research did not require ethics approval.

This research does not involve any human subjects.

Consent for publication

This research does not involve any individual person’s data for publication.

Competing interests

The authors declare no competing interests.