Mandarin fruits (‘Kiyomi’ and ‘Succosa’) were harvested from a commercial orchard in Liangshan Yi autonomous prefecture, Sichuan province, China, ensuring uniform size and absence of visible injuries. Fruits harvested in September (Sep), October (Oct), November (Nov), and December (Dec)were transported to the laboratory, where pulp samples were frozen, homogenised in liquid nitrogen, and stored at −80°C for subsequent analyses. Three replicates, each consisting of six fruits, were analysed.

‘Kiyomi’ and ‘Succosa’ fruits at each stage were selected to determine the total soluble solid (TSS) and titratable acidity (TA). At each development stage, more than 18 fruits were used for quality assessment, with three replicates. TSS (%) and TA (%) were measured using a digital acidity metre (Pocket PAL-BXIACID1, ATAGO, Tokyo, Japan) following the manufacturer’s instructions.

Glucose, fructose, and sucrose levels were determined, as described earlier (Liu et al., 2019). Briefly, 2 g of pulp was homogenised with 10 mL of ddH₂O, incubated for 15 min at 80°C, centrifuged, and filtered. The resulting supernatant was analysed on an Agilent 1260 HPLC system (Agilent Technologies) with a refractive index detector using an Innoval NH₂ column (4.6 mm × 250 mm, 5 μm, Agela Technologies, Shanghai, China). The mobile phase comprised acetonitrile: water (80:20, v/v) with a flow rate of 1 mL min⁻¹.

Citrus SPS genes were identified following established methods (Hu et al., 2015; He et al., 2016). Genomic sequences from various citrus species (Citrus clementina v1.0, Citrus grandis ‘Wanbaiyou’ v1.0, Poncirus trifoliata v1.0, and Citrus sinensis v2.0) were obtained from the Citrus Pan-genome to Breeding Database (CPBD: http://citrus.hzau.edu.cn/index.php). Arabidopsis SPS gene members and their protein sequences were sourced from the Arabidopsis Information Resource (TAIR: https://www.arabidopsis.org/browse/genefamily/index.jsp). HMMER software version 3.0 was utilised to identify C. sinensis SPS (CsSPS) genes. Furthermore, the MapGene2Chrom software (MG2C_v2.1) was employed to generate a chromosome location image of CsSPSs, and TBtools software was used to display the exon/intron structure of all CsSPSs.

Conserved motifs of SPS proteins were identified using the multiple EM for motif elicitation (MEME software) (Brown et al., 2013). The optimal width of each motif ranged from 6–20, with a maximum of six motifs to search and default values for other parameter settings (Bailey et al., 2006). To ensure the inclusion of SPS domains, all candidate SPSs were validated using the National Center for Biotechnology Information (NCBI) Conserved Domain Database (CCD) to ensure that they contained the SPS domains. Additionally, the S6PP domain was predicted through multiple sequence alignment using BioEdit.

Multiple sequence alignments of citrus and Arabidopsis thaliana SPS protein sequences were performed using Molecular Evolutionary Genetics Analysis (MEGA) version 6.0, with 1,000 bootstrap replications, pairwise deletion, and Poisson model. Subsequently, neighbour-joining phylogenetic trees were constructed.

RNA-Seq data were retrieved from published studies (Terol et al., 2019; Feng et al., 2021; Zhang et al., 2021). The data that support the findings of this study have been deposited in the NCBI BioProject database under accession numbers PRJNA636131, PRJNA517400 and PRJEB12880. RNA-Seq data analysis was performed as described previously (Pertea et al., 2016). Briefly, the analysis included preprocessing for quality using FastQC, trimming low-quality reads (q < 20) and adapters using Trimmomatic, alignment to the C. sinensis genome using HISAT2 with default parameters, and assembly of mapped reads using StringTie. Fragments per kilobase per million mapped fragments (FPKM) was used to represent the gene expression levels. CsSPS expression profiles were extracted from the RNA-Seq data.

Total RNA was isolated from different tissues, as described previously (Liu et al., 2022). Specific primer pairs for CsSPSs amplification were designed using the Primer Express software (Applied Biosystems, Foster City, CA, USA). The specificity and amplification efficiency of the primers were validated using BLASTN against the sweet orange genome. Relative gene expression values, with the citrus β-actin gene as the internal reference gene, were calculated using the 2^−ΔΔCt method (Livak and Schmittgen, 2001). The sequences of RT-PCR primers are displayed in Supplementary Table S5 . All expression data were processed using the Z-score standardisation method.

Weighted gene co-expression network analysis (WGCNA) (v1.71) in R was used to construct the co-expression networks (Langfelder and Horvath, 2008). Among 29,138 genes, 16,961 with a sum FPKM < 1 across all samples were removed. The remaining genes were used for the WGCNA. The one-step network construction and module detection function were conducted using an unsigned topological overlap matrix (TOM), a soft-thresholding power b of 14 (R2 > 0.9), a minimal module size of 30, and a branch merge cut height of 0.25. The co-expression network of candidate CsSPSs was visualised using Cytoscape (version 3.6.1) (Shannon et al., 2003).

The WRKY response cis-acting elements (W-box) in the promoter regions (2000-bp upstream of the initiation codon) of CsSPS1, CsSPS2, CsSPS3 and CsSPS4 using PlantCARE online software (https://bioinformatics.psb.ugent.be/webtools/plantcare/html/) were estimated ( Supplementary Table S4 ). Afterwards, CsSPS2 promoter were inserted independently into the pGreen II 0800-LUC double-reporter vector, and the CDSs of CsWRKY2, CsWRKY20, CsWRKY28 and CsWRKY32 were inserted into the constructed pGreen 62-SK vector driven by the 35S promoter as the effector, using the primer sequences listed in Supplementary Table S7 . The constructed effector and each reporter plasmid were co-transformed into tobacco leaves using A. tumefaciens strain GV3101(psoup-p19). Plasmids containing CsWRKYs and promoter were combined at a 10:1 ratio (v/v) and then infiltrated into tobacco leaves using needleless syringes. At 2 days after infiltration at 21 ^◦C, the LUC and REN activities were measured using a Dual-Luciferase Assay kit (Promega, USA) on a Luminoskan Ascent Microplate Luminometer (Thermo, USA). The results were calculated using the LUC to REN ratio. Six biological repeats were assayed for each combination.

All data are presented as the mean (± standard deviation [SD]) of a representative experiment. Significant differences between samples were determined using ANOVA followed by Tukey’s test. The heatmaps were plotted using R studio software using the pheatmap package. The correlation analysis was performed using R studio software. Figures were created using GraphPad Prism (GraphPad Software, CA, USA).

To comprehend the dynamics of sugar and acid development in citrus, we assessed TSS and TA in two citrus varieties (‘Kiyomi’ and ‘Succosa’). As citrus fruit matured, the pulp colour deepened ( Figure 1A ), TSS content exhibited an evident upward trajectory ( Figure 1B ), and TA content significantly decreased ( Figure 1C ). Notably, the accumulation stages of sugar and acid varied among different citrus varieties. ‘Succosa’ attained higher TSS levels and lower TA content earlier than ‘Kiyomi’, which maintained stability. In contrast, ‘Kiyomi’ demonstrated a faster TSS accumulation rate and significantly higher TA content compared with that of ‘Succosa,’ despite lower TSS levels.

As the primary sugars in citrus fruits, the formation of soluble sugars, including sucrose, fructose, and glucose, were analysed for quality assessment during development ( Figure 2 ). We identified sucrose as the predominant component of soluble sugar in citrus pulp, followed by glucose and fructose. All three sugar components increased with citrus fruit development, with sucrose exhibiting the most substantial increase. ‘Succosa’ displayed higher sucrose content and consistent glucose and fructose levels compared with that of ‘Kiyomi.’ However, ‘Kiyomi’ exhibited a faster accumulation rate of sucrose, glucose, and fructose, aligning with the TSS results.

Sucrose, a vital sugar in citrus fruits, is synthesised by SPS, the rate-limiting enzyme. We analysed the CsSPS family from four representative citrus varieties (Citrus sinensis, Citrus clementina, Citrus grandis, and Pitrus trifoliat) ( Supplementary Table S1 ) and constructed phylogenetic trees to elucidate the evolutionary relationship of SPS genes between citrus and other species ( Supplementary Table S2 ). Citrus SPS genes were categorised into three families, with CsSPS1 (Cs4g_pb004370) and CsSPS2 (Cs4g_pb022560) in family A, CsSPS3 (CsUn_pb042260) in family B, and CsSPS4 (Cs9g_pb011150) in family C ( Figure 3A ). CsSPS3 demonstrated a closer relationship with Arabidopsis AtSPS3F within the B family, while CsSPS4 also exhibited a closer relationship with Arabidopsis AtSPS4F within the C family. The chromosomal distribution of CsSPSs was analysed based on the physical location data retrieved from the GCA_018104345.1_Cs2.0_genomic database on the NCBI website. Using Dual Synteny Plotter software, we assessed the syntenic relationship between SPS genes in Citrus and Arabidopsis. Our findings revealed the distribution of CsSPSs on three chromosomes: CsSPS1 and CsSPS2 on chromosome 1, CsSPS4 on chromosome 9, and CsSPS3 on an unidentified chromosome ( Figure 3B ). Synteny analysis between SPS genes in citrus and Arabidopsis identified AtSPS1F-CsSPS1 and AtSPS2F-CsSPS2 as syntenic gene pairs ( Figure 3C ).

Furthermore, an analysis of conserved motifs and domains was conducted using the MEME web server, revealing a total of six conserved motifs and three conserved domains in citrus SPS proteins ( Figure 4A, B ). The results indicate that all citrus SPS proteins share the same motifs ( Supplementary Table S3 ) and domains: a Glycos_transf domain, an S6PP domain (C-domain), and a Glycosyltransferase domain (N-domain). Notably, citrus SPS3 possesses a Glycos_transf_4_4 domain that other SPS proteins lack. Concurrently, two transcripts of CsSPS2 (Cs4g_pb022560) lack Motif 6, and one transcript even lacks the Glycos_transf domain, suggesting that these two transcription modes may not be the primary transcription mode of CsSPS2.

The exon–intron structures of CsSPSs were further analysed. Generally, most citrus SPS genes contain twelve to fourteen exons ( Figure 4C ). Each citrus SPS within the same group shares the same intron–exon structure. For instance, all citrus SPS4 genes contain 14 introns, and every citrus SPS3 gene contains 12 introns. Additionally, each of SPS1 and SPS2 contains 13 introns, except for the two specific transcription modes of CsSPS2.

To gain further insights into the functions of CsSPSs, the 1500 bp promoter regions upstream of the CsSPSs initiation transcription site were analysed using the PlantCARE website (http://bioinformatics.psb.ugent.be/webtools/plantcare/) ( Figure 4D ). The analysis revealed that cis-elements in the promoter region of all identified CsSPSs are primarily involved in responding to light, hormones, and abiotic stress. All CsSPS promoter regions feature photoresponsive elements (Box 4 and G-box), and most CsSPS promoters contain low-temperature, methyl jasmonate (MeJA), and salicylic acid (SA) responsive elements, enabling them to respond to ethylene and abscisic acid (ABA) reactions ( Supplementary Table S4 ). The diversity of cis-acting elements in the upstream promoter region of CsSPSs indicates that their function may involve various reactions, such as hormones, abiotic stress, seed, and endosperm development.

Publicly available transcriptomic data revealed distinct expression patterns of CsSPSs in different citrus tissues. CsSPS1 and CsSPS2 were predominantly expressed in ‘sinensis’ pulp, with CsSPS1 decreasing and CsSPS2 increasing during fruit maturation ( Figure 5A ). CsSPS3 and CsSPS4 exhibited high expression in leaves, with CsSPS3 also detected in flowers. However, transcriptional expression of CsSPSs in peel and root was low. Across maturing citrus fruits, CsSPSs displayed unique expression characteristics ( Figures 5B, C ). For example, the transcription level of CsSPSs in ‘Fengjie72-1’ is significantly higher than that in ‘CaraCara’, whereas the transcription level of CsSPSs in Clementina and Hemandina is similar. The transcriptional expression levels of most CsSPSs in pulp were continuously increased. CsSPS1 and CsSPS2 showed an evident upward trend in the four varieties. However, some CsSPSs, such as CsSPS3 in ‘CaraCara’, have decreased expression levels, which reflects the tissue expression specificity of CsSPSs mentioned earlier.

To understand the expression of CsSPSs in citrus pulp, we investigated the transcriptional expression of CsSPSs in ‘Kiyomi’ and ‘Succosa’ ( Figure 6A ). Consistent with the transcriptome results, the transcriptional expression of the four CsSPSs in citrus pulp increased with fruit development, and the types of CsSPSs expression patterns differed in the pulp of different citrus varieties. For example, The transcription expression of CsSPS2 and CsSPS3 is dominant in ‘Kiyomi’, while the transcription expression of CsSPS3 is dominant in ‘Succosa’. Unexpectedly, CsSPS1 is considered to be the most important SPS gene in the pulp of other citrus varieties, but in the pulp development of ‘Kiyomi’ and ‘Succosa’, the relative transcriptional expression of CsSPS1 is lower than that of other CsSPSs. Simultaneously, the stable, high transcription level of CsSPS3 and the rapid increase in CsSPS4 expression levels in the two materials indicated the importance of CsSPS3 and CsSPS4 in the development of citrus pulp.

To comprehensively highlight the significance of CsSPSs in citrus sugar accumulation, we analysed Pearson’s correlation coefficients between CsSPSs expression and sugar content ( Figure 6B ). As expected, most CsSPSs were significantly and positively correlated with sugar content and its components. However, in ‘Succosa’, CsSPS1 with low transcriptional expression level and CsSPS4 with late rapid elevation demonstrated the most significant positive correlation with sugar content. Counterintuitively, despite having the highest transcriptional expression level, CsSPS3 had a low correlation with sugar content.

To further understand the possible regulatory pathways of citrus sugar metabolism, we analysed the co-expression of CsSPSs based on our previous study (Zhang et al., 2023a). The results showed that all four CsSPSs were in the blue module (4,871 genes), which was positively correlated with the total sugar content, and the expression levels of most genes in the blue module increased with fruit development ( Figure 7A ). Importantly, in the blue module, we observed 196 transcription factors ( Supplementary Table S6 ) that have a significant co-expression relationship with CsSPSs (weight > 0.15), including bHLH (15), ERF (12), bZIP (7), MADS (10), NAC (17) and MYB (18) ( Figure 7B ). It is worth noting that we found that eight WRKY transcription factors (CsWRKY20, CsWRKY47, CsWRKY32, CsWRKY2, CsWRKY65, CsWRKY3, CsWRKY28, CsWRKY74) have a co-expression relationship with CsSPSs. A positive relationship between CsWRKY47, CsWRKY3, CsWRKY28, CsWRKY74 and sugar was found in our previous study, indicating that there may be a regulatory relationship between CsWRKYs and CsSPSs.

To elaborate on the potential involvement of CsWRKYs in the sugar metabolism of citrus fruits through CsSPSs, we examined the expression of co-expressed WRKY genes in Kiyomi’ and ‘Succosa’. The results revealed a continuous increase in the transcriptional expressions of CsWRKY2, CsWRKY20, CsWRKY28, and CsWRKY32, and CsWRKY47 with the maturation of citrus pulp ( Figure 8A ).

Notably, W-box cis-elements was found in CsSPS2 promoter, but not in other CsSPSs ( Supplementary Table S4 ). The dual-LUC activity assay between CsWRKYs and CsSPS2 indicated that the co-expression of CsWRKY2, CsWRKY20 with the CsSPS2 promoter significantly enhanced the LUC to REN ratio, and co-expression of CsWRKY28 with the CsSPS2 promoter significantly reduced the LUC to REN ratio, whereas no significant LUC to REN ratios were observed in the independent co-expression of CsWRKY32 with the promoters of CsSPS2. This indicated that CsWRKY2, CsWRKY20 might act as an activator in citrus fruit of sugar accumulation through CsSPS2, while CsWRKY28 inhibits sugar accumulation through CsSPS2 ( Figure 8C ).