Molecular Regulation of Fruit Quality Traits in Citrus: RNA-Seq-Based Meta-Analysis

Abstract

Citrus genomes as storehouses of genetic information of immense commercial utility remain untapped for the improvement of fruit quality traits and other production-related stresses. With the rapid expansion of transcriptomic datasets, integrative meta-analysis has further aided in uncovering interspecies molecular mechanisms associated with fruit quality development. In this study, we performed a cross-project RNA-Seq meta-analysis, integrating multiple publicly available BioProjects encompassing diverse citrus species, viz., Citrus sinensis, C. reticulata, C. maxima, C. clementina, C. japonica, and C. papeda, known to dominate the morphogenetic evolution of the citrus industry. High-throughput RNA-Seq data were processed using various bioinformatics tools. A total of 15 interspecies comparisons identified 676 unique DEGs, enriched in pathways related to secondary juice yield and processing quality traits. We also established that domestication aided in metabolism, oxidative stress responses, phenylpropanoid and flavonoid biosynthesis, and hormone-mediated signaling. Multivariate analyses (PCA and heatmap visualization) highlighted distinct yet overlapping expression patterns across these citrus species. By combining differential expression, co-expression network analysis and QTL-GWAS integration, we identified 19 high-confidence candidate genes responsible for transcriptomic variation associated with measurable fruit quality traits. Genes such as LOC102612823 and LOC102607495, which co-localized with seed number QTLs on chromosome 1, represented strong candidates regulating reproductive development and seed formation, the traits that directly influence fruit texture and market acceptability. Genes linked to juice content QTLs, including LOC102611137 and LOC102612553 on chromosome 5, suggested their roles in metabolic regulations behind juice accumulation. These loci provided definitive breeding clues for enhancing the reshaping of citrus fruit transcriptomes while retaining key ancestral regulatory components.

1. Introduction

Citrus has a small genome size of 265–400 Mb [1,2]; among next-generation sequencing technologies, whole-genome sequencing of 98 citrus genomes has been performed (https://www.ncbi.nlm.nih.gov/datasets/genome/?taxon=2706, accessed on 5 December 2025), starting with the draft genome of C. sinensis [3]. The analysis of citrus species genome sequence data has untangled some of the taxonomic confusions associated with interspecific and intergeneric hybridization events [4], confirming several existing species as derivatives of three major/true species, i.e., Citrus medica L. (citron), Citrus reticulata Blanco (mandarin), and Citrus maxima (Burm.) Merrill (pummelo) [5]. Draft genome sequences have established the parentage of sweet orange and the fact that the currently cultivated mandarin has part of the pummelo genome. Despite being morphologically different, kumquat and clementine have a shared ancestry [4,5]. In addition to these genomic resources, there has been an expansion in citrus RNA-seq datasets across public repositories including NCBI SRA (National Center for Biotechnology Information—Sequence Read Archive), GEO (Gene Expression Omnibus), CNCB (China National Center for Bioinformation), and specialized citrus databases [6,7]. These expanding transcriptomic resources provide valuable insights into citrus evolutionary genomics, serving as a foundation for the identification and functional characterization of genes associated with fruit development, and streamlining metabolic pathways and nutritionally significant traits. However, despite the growing abundance of genomic and transcriptomic data for citrus species, these resources remain underutilized. The vast datasets generated from genome sequencing, pan-genome assemblies, and large-scale RNA-seq experiments are yet to be fully integrated and explored to unravel the complex biological processes. Fruit quality parameters of citrus like fruit weight, fruit shape, juice content, aroma, acidity, TSS (Total Soluble Solids), seed number, rind thickness, peel color, and pulp color are known for large field variations [8,9].

In model and crop plants, numerous meta-analyses have previously been conducted to explore gene expression dynamics under various stress conditions. For instance, large-scale RNA-Seq meta-analyses in Arabidopsis thaliana [10,11], rice (Oryza sativa) [12], and tomato (Solanum lycopersicum) [13] have successfully identified core stress-responsive genes and conserved regulatory pathways across different experimental setups. Similarly, in perennial fruit tree species, meta-analyses of transcriptome datasets have been increasingly employed to uncover the molecular mechanisms underlying diverse fruit-related traits. For instance, in strawberry (Fragaria × ananassa), large-scale RNA-seq meta-analyses have elucidated key regulatory networks controlling fruit ripening, color development, and metabolic transitions during maturation [14]. Similarly, in olive (Olea europaea), integrative transcriptomics has identified core genes associated with oil biosynthesis, fatty acid metabolism, and fruit quality traits [15]. In apple (Malus × domestica), meta-analysis of RNA-seq data has revealed conserved transcriptional responses to biotic stresses and ripening clues, highlighting shared and species-specific gene expression patterns [16]. These studies deploying transcriptomic meta-analysis aided in elucidating gene networks governing complex fruit developmental and physiological processes across fruit tree crops.

Within the Citrus genus, meta-analytical approaches have been used to explore both phenotypic and transcriptomic variations. Previously, studies focused on integrating gene expression datasets related to Huanglongbing (HLB) disease, leading to the discovery of key co-expression networks and miRNA–target interactions associated with disease resistance [17]. On the other hand, recent efforts have extended to meta-analyses of physiological and agronomic traits, such as soil fertility and nutrient constraints [18], besides yield and fruit quality optimization under varying environmental and management conditions [19].

The proposed study represents the first meta-analysis of citrus fruit transcriptomes aimed at elucidating key genes and pathways involved in fruit development and fruit quality traits. We conducted a comprehensive meta-analysis of the transcriptomic studies on fruit tissues of six species, i.e., C. sinensis, C. maxima, C. japonica, C. papeda, C. reticulata, and C. clementina (Figure 1). We re-analyzed and interpreted RNA-Seq data experiments for the identification of genes commonly expressed across the species and within the species, candidate genes for various fruit traits, and genes overlapping with already established citrus fruit QTLs. Furthermore, we analyzed genes and modules involved in various networks related to fruit biology.

2. Materials and Methods

2.1. Systematic Search and Data Collection of RNA-Seq Studies

Publicly available citrus species RNA-Seq datasets were identified and downloaded from the NCBI Sequence Read Archive (SRA) (https://www.ncbi.nlm.nih.gov/sra, accessed on 5 December 2025) and China National Center for Bioinformation (CNCB) (https://ngdc.cncb.ac.cn/bioproject/browse/, accessed on 5 December 2025) (

Table 1. Transcriptomics raw data details of citrus species used for the meta-analysis.

S.No	BioProject ID	Species	No. of Samples	Stage	% Mapping *	Reference
1.	PRJNA517400	Citrus sinensis	6	Fruit (180 days; fruit tissue)	90–93%	[20]
2.	PRJNA421473	Citrus clementina	3	Fruit (180 days; whole fruit)	83–93%	[21]
3.	PRJNA421473	Citrus japonica	3	Fruit (180 days; whole fruit)	83–93%	[21]
4.	PRJNA683589	Citrus maxima	6	Fruit (~180–240 days; sarcocarp)	90–92%	[22]
5.	PRJCA025346	Citrus reticulata	3	Fruit (180 days; whole fruit)	91–92%	[23]
6.	PRJCA009572	Citrus papeda	3	Fruit (no days mentioned; small whole fruit)	89–90%	[24]

* Denotes the mapping percentage with the reference genome Citrus sinensis DVS_A1.0 (GCF_022201045.2).

). RNA-Seq datasets were available for six citrus species, i.e., C. sinensis, C. maxima, C. japonica, C. papeda, C. reticulata, and C. clementina, with a total of 24 total fruit or fruit tissues samples. These RNA-seq datasets were used for further analysis. The number of samples for each species denotes the total number of fastq files downloaded from the public databases.

2.2. Quality Control, Mapping, and Count Generation

Quality control was performed on all Genome Consortium Format (GCF) files, followed by adapters and low-quality read trimming. Trimmomatic software v.0.40 [25] was used for the trimming of five bases at the 5′ end and of low-quality bases at the 3′ end of RNA-seq reads (HEADCROP:5; TRAILING:5 parameters) for the removal of adapters found on right-ends of RNA-Seq reads (ILLUMINACLIP:TruSeq2-PE:2:30:10 parameter). Clean reads were mapped using HISAT2 v 2.2.1 [26] against the C. sinensis DVS_A1.0 (GCF_022201045.2) [4] reference genome. Mapped reads were counted using the featureCounts Subread package v 2.0.6 [27]. Gene-level read counts were generated independently for each species, based only on uniquely mapped reads.

2.3. Differential Expression Analysis and Gene Ontology Enrichment

We used the DESeq2 (1.50.2) package in R v.4.5.2 for differential expression analysis based on raw gene counts [28]. Low-expression counts (zero expression) were dropped by using (cpm_vals ≥ 1) ≥ 3 [29] and the remaining data were normalized using the median-of-ratios (DESeq2 library) method to remove technical biases and transformed using variance stabilized transformation (VST) to stabilize variance. In order to understand the expression patterns and clustering, the VST transformed counts were used to perform the PCA, followed by a heatmap with hierarchical clustering showing the top 50 gene expression pattern across all 6 species. (logFC > 2, padj < 0.001), where logFC denotes the measure of change in gene expression between two species to measure the gene upregulation or downregulation within species 1 w.r.t. species 2, and padj is the adjusted p-value as a statistical measure to ensure that no false-positive genes are included. For all six species, 15 different pairwise DE comparisons were performed, with flower plot representing significantly expressed genes. PCA, heatmap, and flower plot diagrams were generated using R v 4.5.2. For the functional annotations of these genes, GO and pathway enrichment analyses were performed on the significantly expressed unique genes using DAVID Annotation tool v.v2025_2 [30]. We submitted the Gene LOC IDs to the tool to perform gene ontology and KEGG pathway analysis with p-value < 0.05, FDR > 0.05; Fisher’s exact test was also considered. Visualization was performed in R, revealing different functions and pathways.

2.4. Cross-Validation Through Fruit QTL Trait Mapping

To identify the candidate genes, we searched for DEGs that co-localized within QTLs previously identified for fruit-related traits through GBS-based GWAS and QTL-associated fruit trait mapping [31,32,33]. For the cross-validation, we mapped the unique IDs using the NCBI batch entrez for locating the coordinates of those genes with a window of (+−) 1 Mb used to find overlapping genes. The length of all nine chromosomes was taken from a C. sinensis draft genome article for building a pseudochromosome map [3]. Genes differentially expressed in any of the fruit quality traits in at least one citrus species were identified by searching the literature, to be used as evidence of their role in the fruit developmental process.

2.5. WGCNA

To identify biologically relevant co-expression modules, gene expression data from six species were subjected to weighted gene co-expression network analysis (WGCNA) [34]. Total genes were filtered with row-wise expression sums ≥ 50 [29] for network analysis. Network parameters were tuned using the pick Soft Threshold function, based on scale-free topology merit and mean connectivity, and a signed network was built to retain the orientation of gene-to-gene correlations. The network was built using the Blockwise Modules function with a minimum module size of 100 genes. Module–trait correlations were tested by correlating module eigengenes (MEs), which represented the primary expression pattern of each module, with species as a quantitative trait. Hierarchical clustering was employed to define module membership and intramodular connectivity to calculate module membership (kME) in order to define key regulatory modules for further analysis [35].

2.6. Functional Enrichment of Key Modules

To identify the biological roles of key modules, GO enrichment was performed using a hypergeometric test and categorized GO terms into biological process (BP), Molecular function (MF), and cellular component (CC) [36]. GO annotations were obtained from the NCBI gene2go file for citrus species, and the description of the enriched terms was added using the http://www.geneontology.org (accessed on 29 January 2026) database. Enrichment was done using clusterProfiler v.4.10.0 [37], and terms with a False Discovery Rate (FDR) lower than 0.05 were considered.

3. Results

3.1. Summary of Meta-Analysis of RNA-Seq Data

Transcriptomic data from six species following the criteria previously described (Section 2.1) were downloaded and analyzed. The samples were from 180–220-day-old fruit and in some cases, whole mature fruit; in other cases, we merged the tissue-wise data into a single fruit (Table 1). After quality control and data cleaning, the RNA-Seq datasets were mapped to the C. sinensis DVS_A1.0 (GCF_022201045.2) reference genome [3], as a complete genome for better annotation. The projection of these DEGs into a two-dimensional space (PCA) revealed six sample clusters, each corresponding to a different species. PCA further revealed a clear species-level separation among six citrus species, with PC1 (54% variance) reflecting major evolutionary and domestication-related transcriptional divergence, while PC2 (20% variance) captured finer metabolic specialization (Figure 2). Wild and semi-wild species clustered distinctly from domesticated mandarins and sweet oranges, while hybrid species occupied intermediate positions, highlighting the combination of ancestry, domestication, and lineage-specific regulatory adaptations.

By measuring the expression profiles of the top 50 differentially expressed genes (Supplementary file S1) across 24 samples, a distinct transcriptomic landscape for the six species was observed. The gene expression showed distinct variation across species, quantified on the scale of +2 to −2, representing upregulation and downregulation, respectively, with 0 representing the mean expression level. The heatmap revealed that all six species have different expression patterns in the fruits (Figure 3).

3.2. Fruit Trait-Related Genes

A variety of gene expressions was observed across different combinations (

Table 2. Summary of pairwise transcriptomic comparisons among citrus species: differentially expressed genes and dominant metabolic trends (↑ indicates the upregulated genes whereas → indicates possible response/phenotype in plants).

S.No	Comparison (A vs. B)	DEGs	Genes Up (Species A)	Genes Up (Species B)	Dominant Metabolic Trend
1.	C. clementina vs. C. japonica	6	3	3	C. clementina: terpene synthases, defense genes ↑ → aroma and disease resistance
2.	C. clementina vs. C. reticulata	216	203	13	C. clementina: volatile and phenylpropanoid genes ↑ → aromatic complexity
3.	C. clementina vs. C. maxima	162	147	15	C. clementina: flavonoids ↑; C. maxima: sugar and growth ↑
4.	C. clementina vs. C. papeda	75	45	30	C. clementina: aroma and flavor genes ↑; C. papeda: stress response ↑
5.	C. clementina vs. C. sinensis	81	66	15	C. clementina: terpene synthase and acid metabolism ↑; C. sinensis: pigment and sugar genes ↑
6.	C. japonica vs. C. maxima	137	123	14	C. japonica: defense and flavonoid genes ↑; C. maxima: fruit size and sugar ↑
7.	C. japonica vs. C. papeda	93	55	38	C. japonica: secondary metabolism ↑; C. papeda: stress tolerance ↑
8.	C. japonica vs. C. reticulata	205	187	18	C. japonica: aroma and acidity ↑; C. reticulata: sweetness ↑
9.	C. japonica vs. C. sinensis	69	56	13	C. japonica: volatile metabolism ↑; C. sinensis: sugar and carotenoid ↑
10.	C. maxima vs. C. papeda	179	20	159	C. papeda: defense/stress ↑; C. maxima: sugar metabolism ↑
11.	C. maxima vs. C. reticulata	60	34	26	C. maxima: growth ↑; C. reticulata: flavor and pigment ↑
12.	C. maxima vs. C. sinensis	53	8	45	C. maxima: cell expansion ↑; C. sinensis: pigment and sugar ↑
13.	C. papeda vs. C. reticulata	249	218	31	C. papeda: flavonoid/limonoid defense ↑; C. reticulata: aroma/sugar ↑
14.	C. reticulata vs. C. sinensis	39	4	35	C. reticulata: aroma volatiles ↑; C. sinensis: carotenoids and sweetness ↑
15.	C. papeda vs. C. sinensis	118	88	30	C. papeda: stress and bitter compounds ↑; C. sinensis: sugar and carotenoid ↑

); some were repeated in a few, while others were specific to a particular species, throwing up interesting details. Among these comparisons, we found 676 unique genes differentially expressed. The highest number of DEGs (the list of DEGs is given in Supplementary file S2) were identified in C. papeda versus C. reticulata (251) and the lowest in C. clementina versus C. japonica (6) (Figure 4).

3.2.1. C. papeda

In C. papeda, a total of 297 genes were observed to be upregulated across the comparisons, displaying the highest number of upregulated genes overall, with a transcriptional profile strongly enriched in defense- and stress-related functions. Key upregulated genes included linoleate 13S-lipoxygenase 2-1 (LOC102626288), caffeic acid 3-O-methyltransferase–like protein (LOC127902074), and Kunitz trypsin inhibitor 5-like protein (LOC102622572), all of which were associated with biotic stress responses and secondary metabolism. Genes involved in lipid transport and cell wall remodeling, such as non-specific lipid-transfer protein-like genes (LOC102624990) and pectinesterase-related genes (LOC102627822), were also observed to be prominently upregulated. Such expression profile is consistent with papeda’s wild ancestry and supported the retention of ancestral stress-adaptive regulatory programs.

3.2.2. Citrus reticulata

In C. reticulata, 72 upregulated genes were observed across comparisons, showing a compact yet biologically meaningful set of upregulated genes. Key genes including chlorophyllase-1 (LOC127901308), indicating active chlorophyll turnover, and LOB domain-containing protein 42 (LOC102625682), suggesting the developmental regulatory functions. The upregulation of Peter Pan-like protein (LOC102611140) and terpene synthase genes (LOC102621988) further highlighted their roles in growth regulation and aroma biosynthesis. These expression patterns reinforce the importance of C. reticulata as a central progenitor species contributing regulatory and fruit-quality traits to modern citrus hybrids.

3.2.3. C. sinensis

C. sinensis was observed to display 99 upregulated genes across comparisons, with a moderate but functionally focused set of upregulated genes, largely associated with fruit color and aroma. Major upregulated genes included anthocyanidin glucosyltransferase (LOC102627603) and terpene synthase genes (like LOC102621988), reflecting selection for pigmentation and aroma. Signaling-related genes such as receptor-like kinases (LOC102621729) were also upregulated, indicating refined regulatory control. Compared with wild relatives, C. sinensis exhibited fewer stress-related DEGs, suggesting that domestication narrowed its transcriptional focus toward traits directly relevant to fruit development and consumer preference.

3.2.4. Citrus maxima

C. maxima showed 74 upregulated genes considering interspecies comparisons, a smaller but distinct set of upregulated genes, indicative of targeted regulatory specialization. The most prominent were calcium-binding protein CML29 (LOC102611176) and cytochrome P450 family genes (like LOC102607400) (

Table 3. Putative gene regulators and species-specific expression patterns associated with major citrus fruit characteristics.

S.No	Fruit Character	Likely Molecular Drivers	Observed Expression Pattern
1.	Bitterness (limonoids)	Up: UGT76B1 (LOC102606711), P450s (LOC102627430, LOC127902134)	High in C. papeda; low in C. sinensis
2.	Aroma complexity	Up: TPS, P450 (LOC102607400, LOC102618084, LOC102627625), AAT1 (LOC102614123, LOC102612840)	High in C. japonica and C. reticulata
3.	Sweetness	Up: SUS (LOC102628674, LOC102619819), invertase (LOC102613488)	Dominant in C. sinensis and C. maxima
4.	Pigmentation (orange color)	Up: PSY, BCH, CCD4 (LOC102621234)	C. reticulata → C. sinensis
5.	Antioxidant content	Up: CHS (LOC102607309), F3H (LOC107178641), peroxidase (LOC102607325, LOC102629997, LOC102613924, LOC102625757, LOC102613627)	C. papeda and C. japonica
6.	Fruit size	Up: expansins (LOC102624467), cell wall proteins (LOC127899924, LOC102613488, LOC112496066)	C. maxima and C. sinensis
7.	Acidity	Citrin (LOC102577975), citrate transport	High in C. japonica; low in C. sinensis

), implicating their roles in signal transduction and oxidative metabolism. Genes encoding lipid-transfer proteins (LOC102624990) were also upregulated, suggesting their involvement in membrane dynamics and developmental regulation. Such transcriptional pattern is consistent with C. maxima serving as a foundational citrus species, contributing specific regulatory modules to derived cultivars rather than broad transcriptional reprogramming.

3.2.5. C. japonica

Interspecies comparisons showed as many 252 upregulated genes, predominantly associated with volatile biosynthesis, pigment metabolism, and chloroplast-related processes. Notably, gamma-terpinene synthase (LOC102621988) and other terpene biosynthetic enzymes were strongly upregulated, underscoring the characteristic aromatic profile of C. japonica. Genes involved in pigment turnover, including chlorophyllase-1, chloroplastic-like protein (LOC127901308) and carotenoid cleavage dioxygenase (LOC102621234) (Supplementary file S2), were also observed highly expressed. Several cytochrome P450 family genes (e.g., LOC102627625) were upregulated, suggesting the diversification of secondary metabolic pathways. These results indicated that C. japonica maintained a distinct metabolic specialization while remaining transcriptionally integrated within the broader citrus lineage.

3.2.6. C. clementina

C. clementina exhibited one of the largest sets of upregulated genes among the analyzed species (288), indicating extensive transcriptional activation. Prominent upregulated genes included anthocyanidin 3-O-glucosyltransferase (LOC102627603) and carotenoid cleavage dioxygenase 4 (LOC102621234), highlighting enhanced flavonoid and carotenoid metabolism. Regulatory genes such as LOB domain-containing protein 42 (LOC102625682) and cysteine-rich receptor-like kinase 44 (LOC102621729) were also strongly upregulated, suggesting active developmental and signaling regulation. Additionally, the upregulation of Peter Pan-like protein (LOC102611140) pointed to coordinated control of organ growth and differentiation. Together, these patterns reflected the hybrid origin of C. clementina and its selection for fruit quality traits, including pigmentation and metabolic complexity.

3.3. Gene Ontology (GO) and KEGG Analysis

The GO enrichment of the 676 genes showed significant activity in oxidoreductase (GO:0016491), peroxidase (GO:0004601), methylation (GO:0032259), diterpenoid biosynthetic process (GO:0016102), and heme-binding functions (GO:0020037), with biological processes focused on response to heat, salt, and hydrogen peroxide. The GO in the biological process associated with DEGs was grouped in 38 terms (p-value < 0.05) (Supplementary file S3). The top enriched biological processes were regulation of DNA-templated transcription (GO:0006355), methylation (GO:0032259), diterpenoid biosynthetic process (GO:0016102), hydrogen peroxide catabolic process (GO:0042744), protein folding (GO:0006457), and response to oxidative stress (GO:0006979). On the other hand, the GO in the molecular functional process associated with DEGs was grouped into 38 terms (p-value < 0.05) (Supplementary file S3). The annotation revealed that heme-binding (GO:0020037), iron ion binding (GO:0005506), monooxygenase activity (GO:0004497), oxidoreductase activity, acting on paired donors, with incorporation or reduction in molecular oxygen (GO:0016491), and S-adenosylmethionine-dependent methyltransferase activity (GO:0008757) were identified as important enriched molecular functions. The GO in the cellular component associated with DEGs was grouped into five terms (p-value < 0.05). The cellular component in fruit tissue was observed to be restricted to plasma membrane (GO:0005886), extracellular region (GO:0005576), cell wall (GO:0005618), cyclin-dependent protein kinase holoenzyme complex (GO:0000307), and monolayer-surrounded lipid storage body (GO:0012511) (Figure 5a).

KEGG-analysis showed that DEGs were mostly enriched in the biosynthesis of phenylpropanoid secondary metabolites, flavonoid, stillbenoid, diarylheptanoid, gingerol, sesquiterpenoid, and triterpenoid; metabolism of tryptophan, linolenic acid, and alpha-linolenic acid; protein processing in endoplasmic reticulum; and MAPK signaling pathway-plant (Figure 5b).

3.4. Co-Localization of Differentially Expressed Genes with Fruit Quality QTL Regions

All previously reported fruit-related QTL regions screened to identify the candidate gene co-localization of differentially expressed genes (DEGs) with marker intervals displayed a strong association with fruit quality traits. The mapping of 676 significant DEGs onto trait-associated regions revealed 19 overlapping genes (

Table 4. Chromosome-wise distribution of annotated candidate genes associated with key fruit quality traits.

S.No	Trait	Chr,Marker/SNP ID	Gene Symbol (Position (bp))	Gene Description	Reference
1.	Fruit Weight	Chr8_CiC4368-01 0.00–1.98	LOC112495475 (475628)	Endochitinase-like	[31,32]
			LOC102627878 (490203)	Endochitinase-like
			LOC102627194 (1577166)	Pathogenesis-related protein 1-like
			LOC102625945 (1879060)	Serine carboxypeptidase-like 17
			LOC102625654 (1895275)	Serine carboxypeptidase-like 18
		Chr9_9972980 (fw/ACIDITY/NSPF) 8.97–10.97	LOC102613237 (9993386)	Glucan endo-1,3-beta-glucosidase-like	[33]
			LOC102614791 (10175928)	Glucan endo-1,3-beta-glucosidase-like
			LOC107178516 (10733325)	Wall-associated receptor kinase 3-like
2.	Juice Content	Chr5_25614659 24.61–26.61	LOC102611377 (25700538)	(E)-beta-farnesene synthase	[33]
			LOC102612253 (26390372)	Aspartic proteinase nepenthesin-1-like
3.	Seed Number	Chr1_21029403 20.03–22.03	LOC102607495 (21226403)	Transcription repressor OFP12	[31,32]
			LOC102612823 (20345254)	auxin-induced protein 22D
4.	Mesocarp Size	Chr4_16867277 15.86–17.86	LOC102621357 (16075602)	Squamosa promoter-binding-like protein	[33]
			LOC102624285 (16296295)	11-beta-hydroxysteroid dehydrogenase A
			LOC102624206 (17003043)	Probable 3-beta-hydroxysteroid isomerase
			LOC107175045 (10033462)	Loganic acid O-methyltransferase-like
5.	Segment Count	Chr6_12323301 11.32–13.32	LOC102622372 (11957966)	L-ascorbate peroxidase 2, cytosolic	[33]
			LOC127899598 (12121545)	18S ribosomal RNA
6.	Diameter of Fruit Axis	Chr4_14366349 11.32–13.32	LOC127899146 (13361977)	ABC transporter G family member 1-like	[33]

). These genes were distributed across five QTL markers and SNP-associated regions linked to fruit weight, juice content, seed number, mesocarp size, segment count, and diameter of fruit axis. The identified candidate genes were located on chromosomes 1, 4, 5, 6, 8, and 9 (Figure 6). On chr01, candidate genes LOC102612823 and LOC102607495 were mapped within a known seed number. On chr04, there were four candidate genes, one accounting for diameter of fruit axis (LOC127899146) and three for mesocarp size (LOC102612357, LOC102624285, and LOC102624286). On chr05, LOC107175045 fell within a reported mesocarp size QTL, whereas LOC102611137 and LOC102612553 co-localized with a juice content QTL, supporting their possible roles in fruit structural and compositional traits. On chr06, candidate genes LOC102622372 and LOC127899598 overlapped with a known segment number QTL. Notably, on chr08 and chr09, multiple candidate genes mapped within previously identified fruit weight QTLs. The chr08 region showed the highest gene density (LOC112495475, LOC102627878, LOC102627194, LOC102625945, and LOC102625554); while chr09 contained LOC102612337, LOC102614791, and LOC107175616 within its fruit weight QTL interval.

3.5. Identification of Co-Expression Modules

WGCNA identified 17 distinct modules from 16,253 genes, with a size ranging from 2710 to 112 genes (Figure 7a). Hierarchical clustering revealed distinct clusters based on similar expression profiles for 16,253 genes, shown in the dendrogram (Figure 7b). Ten modules were observed to be highly significant and six showing strong positive correlations (r ≥ 0.7) with highly significant p-values (p ≤ 9 × 10⁻⁶). The six modules of particular interest were the black, cyan, red, brown, yellow, and green-yellow modules. Some modules were observed to be positively correlated in one species and negatively associated in another, like the yellow, magenta and green modules, which were positively associated with sweet orange and mandarin, and negatively associated with pummelo, revealing species-to-specific gene biological variation. We thus identified six modules regulating the key gene network; each module belonged to one particular species, as shown in the module–trait relationship heatmap, revealing species-to-specific gene co-expression patterns (Figure 7c).

3.6. Functional Enrichment Analysis of Key Modules

The identified co-expression modules subjected to GO-enrichment analysis to determine the functional roles of the co-expressed genes were found to be highly specialized around energy production, transport, and regulatory signaling. Across all modules, protein phosphorylation (GO:0006468) and ATP binding (GO:0005524) were consistently observed enriched as top terms, though each key module also displayed unique functional roles (Supplementary file S4).

Module-Specific Biological Insights

Brown Module (Kumquat): As many as 1785 genes dominated by ATP binding (108 genes) and DNA binding (46 genes) were covered under this module, including other biological processes like signal transduction and protein ubiquitination. It showed the highest gene count confined to the nucleus (123 genes) among the cellular components. A notable specific process identified was the photochemical reaction (GO:0015979) (Figure 8a).

Black Module (Papeda): This module, with 1107 genes, was primarily involved in protein phosphorylation and regulation of DNA-templated transcription. Other enriched terms included defense response, carbohydrate metabolic process, and response to oxidative stress. The cellular component analysis placed the majority of these genes in the nucleus (63 genes) and membrane (50 genes) (Figure 8b).

Cyan Module (Clementine): This module, with 202 genes, was observed to be unique for its strong enrichment in methylation (GO:0032259) and O-methyltransferase activity (GO:0008171). Statistical significance was found for methyltransferase processes (GO:0008171) (Figure 8c).

Green-Yellow Module (Mandarin): This module was represented by 343 genes, involved in biological processes featuring protein phosphorylation and transmembrane transport. Distinctively, it was also enriched for sterol metabolic process (GO:0016125) and ATP hydrolysis activity, in addition to other terms like protein folding and protein ubiquitination (Figure 8d).

Red Module (Pomelo): The red module, with 1277 genes, exhibited enrichment in RNA binding (GO:0003723), with 58 genes associated with this function. The cellular component profile showed nearly equal distribution between the nucleus (110 genes) and the cytoplasm (92 genes), besides other terms comprising translation, zinc ion binding, and DNA binding (Figure 8e).

Yellow module (Sweet orange): Similar to the red module, the yellow module (1403 genes) focused heavily on RNA binding and protein phosphorylation. However, the module also showed the highest cytoplasmic gene count of all modules (95 genes), with other terms observed including regulation of DNA-templated transcription and protein kinase activity (Figure 8f).

4. Discussion

Citrus is considered highly stress-sensitive amongst perennial fruit trees. Its remarkable phenotypic diversity in terms of fruit quality traits is largely attributed to interspecific and intergeneric hybridization, a process that has historically complicated citrus taxonomy, while simultaneously enriching the cultivated citrus gene pool [5,38,39]. The widespread culinary, nutritional, and cultural uses of citrus fruits further underscore the selective pressures acting on traits such as sweetness, acidity, rind thickness, and pigmentation. Recent genomic and transcriptomic investigations provided compelling evidence in favor of domestication, reshaping the regulatory landscape of fruit development and leading to pronounced shifts in gene expression associated with fruit quality traits [40,41]. In particular, changes in pathways governing organic acid metabolism, sugar accumulation, and carotenoid biosynthesis have been identified as playing pivotal roles in the diversification of citrus fruit quality [42,43]. Consistent with these findings, comparative transcriptomic analyses indicated that domestication has favored an increase in sugar content and reduced juice acidity, especially in mandarins, reflecting targeted selection for improved palatability during the fruit ripening process [40,44].

Although pummelo and mandarin show clear-cut phenotypic differences, as do any two diverse citrus species, they exhibited substantial variability not only in fruit-related traits but also across multiple crop phenological events. In our study, publicly available transcriptomic datasets from the fruit developmental stage elucidated some interesting underlying gene expression patterns associated with fruit development. The comparative transcriptional profiling across six citrus species displayed marked differences in the number and functional categories of upregulated genes, reflecting their evolutionary histories, domestication status, and metabolic specialization. These expression patterns aligned closely with phylogenetic relationships and breeding trajectories described in foundational genomics of citrus [3,44].

Papeda is known for exhibiting well developed defense-related and secondary metabolic pathways. The prominent upregulation of linoleate 13S-lipoxygenase 2-1 (LOX)(LOC102626288) suggested the activation of jasmonic acid (JA)-mediated defense signaling. Lipoxygenases catalyze the first committed step in oxylipin biosynthesis and are central regulators of biotic stress responses in plants [45]. Mandarin displayed a comparatively smaller set of upregulated genes, emphasizing their roles in developmental regulation and fruit-associated traits. The induction of chlorophyllase-1 indicates active chlorophyll turnover, a process central to fruit maturation and color transition as a function of chlorophyll degradation tightly coordinated with carotenoid accumulation during the citrus fruit ripening process [46]. The upregulation of LOB domain-containing protein-42 suggests involvement in organ boundary formation and developmental patterning. LOB domain (LBD) transcription factors further regulate lateral organ development and hormonal integration [47], influencing fruit morphology. Compared to wild relatives, sweet orange showed relatively fewer stress-associated DEGs, supporting the hypothesis that domestication reduces investment in broad-spectrum defense, while enhancing traits directly linked to fruit development [48].

Papeda and kumquat showed the highest number of upregulated genes, supporting their roles as wild/ancestral species with active defense and secondary metabolism. Recent comparative citrus genomics and pan-genome analyses indicated that wild and semi-wild citrus taxa maintained broader transcriptional plasticity and enriched the stress-related gene repertoires relative to domesticated cultivars [44]. Clementine behaved as a transcriptionally active hybrid, consistent with evidence that interspecific citrus hybrids exhibited expanded regulatory variation and allele-specific expression patterns derived from parental genomes [5]. In contrast, domesticated species showed very few significant upregulated DEGs. Citrus species like C. sinensis, C. reticulata, and C. maxima rarely showed significant upregulation under strict criteria, consistent with canalized domesticated transcriptomes [49]. Comparative transcriptomic studies in citrus and other perennial crops demonstrated that artificial selection narrowed expression variability, particularly in defense-related pathways, while enhancing pathways directly linked to fruit developmental traits [39,50].

In our study, GO analysis was done for 676 genes and six modules identified in the WGCNA; the most enriched GO terms in biological processes were identified as regulation of DNA templated transcription, protein ubiquitination, and protein phosphorylation. Enriched terms in molecular function were protein binding, heme binding, protein kinase activity, and transcription regulator activity. The black module (papeda) and brown module (kumquat) were dominated by kinase-mediated signaling and transcription factor activity, consistent with the high biotic and abiotic stress tolerance reported in wild citrus relatives [51,52]. On the other hand, the cyan module (clementine) was dominated by epigenetic regulation, RNA processing, and post-transcriptional shifting towards a balance between developmental plasticity and defense response. The pattern in the green-yellow module (mandarin), the red module (pomelo), and the yellow module (sweet orange) was dominated by post-transcriptional regulation, underscoring the domestication-driven optimization of physiological processes associated with growth, development, and fruit quality [53].

The integration of GWAS- and QTL-based mapping with differential expression analysis enabled the robust identification of candidate genes underlying the key fruit quality traits. The 19 co-localized genes span multiple chromosomes and traits, highlighting the polygenic nature of fruit development and composition. As many as 19 candidate genes identified through DEG–QTL/GWAS co-localization provided a prioritized resource for targeted genome editing aimed at improving citrus fruit quality. A gene like LOC102607495, annotated for transcription repressor OFP12, is linked to the seed number. OFP is an ovate family protein reported to control the pear shape of the fruit, and it is also established as determining the number of seeds per silique in Brassica napus [54]. An in-depth study of this gene in citrus species offered a strong possibility of generating seedless citrus species. Genes involved in cell wall remodeling and defense responses (e.g., glucan endo-1,3-β-glucosidases, chitinases, PR proteins, and wall-associated receptor kinases) could be edited using CRISPR/Cas9 to fine-tune cell expansion and tissue integrity, thereby modulating fruit weight and mesocarp thickness. Likewise, transport- and metabolism-related genes (e.g., ABC transporters and oxidative stress enzymes) could be edited to alter organic acid accumulation and juice content. Importantly, genome editing focusing on cis-regulatory regions instead of coding sequences could be used to mimic favorable natural alleles identified in QTLs, enabling precision breeding, besides preserving elite cultivar backgrounds.

5. Conclusions

Our studies established meta-analysis as a unified transcriptional framework for understanding molecular mechanisms guiding citrus fruit quality developments by linking cross-species gene expression patterns with genetically defined fruit trait loci. Wild species such as C. papeda and C. japonica exhibited extensive upregulation of defense- and stress-related genes, whereas domesticated species showed more canalized expression profiles emphasizing fruit quality traits. The integration of DEG data with previously reported fruit quality QTL and GWAS regions helped in identifying 19 candidate genes regulating fruit quality traits. A molecular tool like WGCNA put forth as many as 17 gene modules, of which six modules showed significantly positive correlations with certain citrus species, revealing candidate regulatory networks associated with citrus fruit quality traits. Such an integrative framework highlighted the embedded power of RNA-Seq meta-analysis for citrus fruit biology, prioritized candidate genes, and regulatory modules for precision breeding and genome editing aiming at fruit quality improvement. These candidate genes, identified through the current study, provide ready-made actionable targets for CRISPR/Cas-based modulation of regulatory regions for multi-trait and multi-environment genomic prediction of fruit quality, without compromising the elite performance traits of cultivars.