Screening Reference Genes for Wine Grapes for Cultivation Under Low-Temperature Stress

Abstract

The harsh, cold, and dry winters in northern China necessitate burying wine grapevines (Vitis vinifera) for winter protection. In this study, we screened for stably expressed reference genes in wine grapes (V. vinifera) under low-temperature stress at 4 °C (chilling) and −15 °C (freezing). A cold-resistant line “Hanniang 1301” and its cold-sensitive parent ‘Cabernet Sauvignon’ were treated at 4 °C and −15 °C for varying durations. Nineteen candidate reference genes were selected for qPCR analysis. Gene stability under chilling and freezing stress was evaluated using the following five algorithms: Delta CT (ΔCt), geNorm, NormFinder, BestKeeper, and RefFinder. The optimal reference genes under chilling (4 °C) and freezing (−15 °C) conditions were pairs with dual-reference combinations. However, the genes selected differed between chilling and sub-freezing temperatures. For chilling stress (4 °C), EF1α-1 and EF1α-2 were the most stable. Meanwhile, for freezing stress (−15 °C), GAPDH and Actin were optimal. We identified suitable reference genes for gene expression studies in wine grapes under low-temperature stress; this establishes a theoretical foundation for optimizing reference gene selection in plants under other abiotic stresses.

1. Introduction

Grapevine (Vitis vinifera L.), a vital economic fruit crop in China, is consumed fresh or processed into raisins and wine. In northern China, field-grown vines are buried after pruning for winter protection due to harsh, dry winters. This labor-intensive practice is costly, increasing susceptibility to crown gall disease (Agrobacterium tumefaciens) through basal stem injuries [1]. Therefore, winter-hardy wine grape cultivars requiring no burial are urgently needed in production [2]. Our group previously developed the superior line ‘Hanniang 1301’ through hybrid screening from a cross of Zuoyouhong × Cabernet Gernischt × Cabernet Sauvignon. This line combines the cold tolerance of Vitis amurensis with the winemaking quality of Cabernet Sauvignon, and has survived unburied through winter in regional trials. Additionally, this cultivar exhibits excellent winemaking potential characterized by small berry size, compact cluster architecture, and distinctive coffee-like aroma in grape seeds. Collectively, it offers a suitable model system for studying cold resistance mechanisms and breeding.

While traditional physiological and morphological analyses have revealed phenotypic responses of grapevines to low-temperature stress, they remain insufficient for elucidating the core genetic regulatory networks underlying cold hardiness. Critically, identifying key genes involved in cold signaling represents the cornerstone for deciphering the molecular mechanisms of cold tolerance in Vitis vinifera [3]. Recent advances in molecular biology have accelerated the discovery and functional validation of cold-responsive genes. Quantitative real-time PCR (qPCR) has emerged as an essential methodology for the dynamic monitoring of gene expression patterns to investigate gene function and regulatory networks [4,5]. Notably, qPCR accuracy is highly contingent upon experimental standardization. Unstable reference genes compromise cross-sample comparability, whereas rigorously validated reference genes correct for technical variations in quantitative analyses [6,7].

Commonly termed housekeeping genes, reference genes (RGs) show constitutive expression across tissues and cell types, remaining stable under varying environmental or experimental conditions [8]. RG stability directly determines experimental reliability [9,10], yet no universal RG exists for all experimental systems [11]. Hence, identifying condition-specific stable RGs is a prerequisite for obtaining reliable and accurate qPCR quantification. Regarding reference gene screening, reliance on a single algorithm carries substantial limitations and risks erroneous conclusions. To ensure robust results, cross-validation using multiple algorithms is essential for comprehensive stability assessment. Each method offers distinct advantages and constraints. The ΔCt [12] method directly compares inter-sample Ct variations, enabling rapid preliminary screening but lacking group-wise differentiation and exhibiting outlier sensitivity. geNorm [13] evaluates expression variability and determines the optimal RG number yet tends to over-retain highly expressed correlated genes. NormFinder [14] incorporates group-specific variations for precise stability ranking but requires predefined sample grouping and is sensitive to sample size. BestKeeper [15] employs Pearson correlation analysis and standard deviation to quantify coordinated expression changes, demonstrating efficacy for small sample sets. RefFinder [16] integrates multi-algorithm outputs through weighted ranking. Our five-algorithm cross-validation strategy mitigates methodological biases by interrogating stability from complementary perspectives. This approach ensures the rigorous identification of condition-specific RGs, significantly enhances result reliability, and establishes a robust foundation for subsequent gene expression analyses.

Li et al. [10] screened ten candidate RGs across six cranberry (Vaccinium macrocarpon) cultivars under varying experimental conditions. SAND proved optimal for different tissue types, while PP2A or RH8 showed superior stability across cultivars. Chen et al. evaluated seven RGs in luffa (Luffa cylindrica) under diverse abiotic stresses. EF-1α showed maximum stability under high-temperature, low-temperature, and ABA treatments. Meanwhile, UBQ was optimal under salt stress. For H₂O₂ and drought conditions, TUB was the most stable reference gene [17]. Ebrahimi et al. screened for stable RGs in fenugreek (Trigonella foenum-graecum) leaves under salt, low-temperature, and high-temperature stresses and elicitor treatments, including titanium dioxide nanoparticles, cold plasma, 24-epibrassinolide, and melatonin. Optimal quantitative normalization required not a single gene but a triple-gene combination of EEF1α, β-TUB, and GAPDH [18]. Sun et al. identified Actin and GAPDH as an optimal dual-reference combination for Kobresia littledalei across tissues under salt stress, drought stress, ABA, and GA treatments [19]. Mao et al. validated FaGAPC2 and FaADPrf1 as a stable reference pair for octoploid strawberry (Fragaria × ananassa ‘Octaploid’) in diverse tissues, organs, and fruit developmental stages [20]. Zong et al. identified condition-dependent optimal RG in Miscanthus under abiotic stresses. FBOX and EF1α formed the most stable pair for cadmium and salt-stressed leaves. Conversely, FBOX and PP2A were optimal for salt-stressed roots and PEG-treated leaves [21]. In pear (Pyrus spp.), Wang et al. showed cultivar-specific reference gene stability through the transcriptome analysis of fruit development. Pbr028511, Pbr038418, and Pbr041114 showed maximal stability in ‘Cuiguan’, ‘Housui’, and ‘Xueqing’ cultivars, respectively [22]. Dong et al. [23] identified the following distinct optimal reference gene combinations for taro (Colocasia esculenta) under specific experimental conditions: for corm development stages: ACY-1 and PIA2; for diverse tissues: COX10 and Armc8; for drought stress: Armc8, COX10, and CCX4L.

Reference gene expression stability varies across species, tissues, and treatments due to transcriptional differences. However, reports on reference gene screening in grapevines are limited. To date, most prior studies have focused on domestication [24,25]. Using qPCR, Chavan et al. [26] identified UBC17, RLI, and ZNF as the optimal reference set during flowering and fruit development. Stable reference gene(s) for diverse grape genotypes under low-temperature stress remain unreported.

In this study, we selected the cold-resistant superior line ‘Hanniang 1301’ and its cold-sensitive parent ‘Cabernet Sauvignon’ to screen stable RGs under chilling (4 °C) and freezing (−15 °C) stresses using qPCR. We aimed to identify optimal reference gene(s) for cold tolerance studies, validating the screening results using cold-responsive VvCBFs expression. The findings can establish a foundation for functional studies of cold resistance genes in grapevine under low-temperature stress, encompassing chilling or freezing.

2. Materials and Methods

2.1. Plant Materials

Two wine grape genotypes were studied, namely, the cold-resistant line ‘Hanniang 1301’ (Shangri-La Vineyard, Luolong County, Qinhuangdao, China; 39.79° N, 119.15° E) and its parent, the cold-sensitive cultivar ‘Cabernet Sauvignon’ (Shigezhuang Experimental Station, Hebei Normal University of Science and Technology, Qinhuangdao, China; 39.77° N, 119.14° E). Grapes were managed under the following identical practices: y-shaped bilateral cordon training (east–west orientation), 0.7 m × 3.0 m spacing, and yield control at 12–15 t/ha in sandy loam soil. The region has a temperate continental monsoon climate (low–mountain hills) with 186 frost-free days, 2745–2809 h of annual sunshine, 10.2–11.0 °C mean temperature, and 638–725 mm of annual precipitation.

In November 2022, one-year-old dormant, disease-free canes of ‘Hanniang 1301’ and ‘Cabernet Sauvignon’ with a uniform diameter and no mechanical damage were collected during the dormancy phase. After surface cleaning with water, the canes were cut into approximately 20–30 cm segments. The cut ends were wax-sealed and stored in moist sand at 4 °C for the stress treatments.

2.2. Low-Temperature Treatments

For the freezing stress (−15 °C) low-temperature treatments, in March 2023, dormant canes were cooled to −15 °C at 4 °C/h in a programmable freezer. Phloem tissues were collected after 12, 24, or 48 h of exposure, flash-frozen in liquid N₂, and stored at −80 °C. Three independent canes per treatment were pooled, homogenized, and equally aliquoted into three technical replicates for downstream analysis. For the chilling stress treatments (4 °C), in June 2023, canes were hydroponically grown in climate chambers (25 °C; 12 h photoperiod of an intensity of 100 μmol·m⁻²·s⁻¹) until new shoots emerged. Whole shoots were harvested after 3, 6, or 12 h at 4 °C, flash-frozen in liquid N₂, and stored at −80 °C. Three independent plants per treatment were pooled, homogenized, and equally divided into three biological replicates for subsequent analysis.

2.3. RNA Extraction and cDNA Synthesis

Total RNA was isolated from grapevine canes using a Plant Total RNA Rapid Extraction Kit (Sangon Biotech, Shanghai, China, Cat# B518631). RNA concentration and purity (A260/A280) were measured with the ultramicrospectrophotometer (BioDrop μLite, BioDrop Ltd, Cambridge, United Kingdom). Integrity was confirmed using 1.5% agarose gel electrophoresis. cDNA was synthesized from 1 μg of DNase-treated RNA using the PrimeScript™ RT Reagent Kit with gDNA Eraser (TransGen Biotech, Beijing, China).

2.4. Primer Specificity Validation

Primers for 19 candidate RGs [27,28,29] (Supplementary Table S1) were designed with Primer 5 [30] and analyzed for specificity/amplicon size via Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/, accessed on 20 October 2023). Conventional PCR was conducted using Taq DNA polymerase as follows: 94 °C for 2.5 min; 30 cycles of 94 °C/30 s, 55 °C/30 s, and 72 °C/20 s; 72 °C for 2 min. Amplicons were electrophoresed on 1.5% agarose gels (220 V, 15 min). Primers producing single bands of expected sizes without primer dimers underwent qPCR validation. Only primers with single-peak melting curves, no non-specific products, and no amplification in no-template controls were selected.

2.5. qRT-PCR Analysis

cDNA was diluted serially (1-, 5-, 10-, 20-fold) in nuclease-free water to optimize template concentration, and based on the results, the final concentration for qPCR was determined to be 150 ng/μL. qPCR was performed using PerfectStart™ Green qPCR SuperMix (TransGen Biotech) on a Bio-Rad CFX Connect instrument (Bio-Rad Laboratories, Inc., Hercules, CA, USA). Each reaction included three technical replicates and no-template controls. The three-step protocol comprised initial denaturation at 94 °C for 30 s (1 cycle), 40 cycles at 94 °C for 5 s (denaturation), 55 °C for 15 s (annealing), and 72 °C for 10 s (extension)

2.6. Data Analysis

Five algorithms were used to assess the stability of all candidate genes. The Delta CT method [12] assesses reference gene stability by quantifying variations in Ct values across samples. It calculates deviations between individual gene Ct values and the sample mean Ct value, with lower variation indicating higher expression stability. geNorm software (http://www.qbaseplus.com/, accessed on 20 December 2024) [13] evaluates RG through M-value (average expression stability measure) and V-value (pairwise variation coefficient). This algorithm identifies the most stable gene combination and determines the optimal number of RGs through iterative stability analysis. BestKeeper (https://www.gene-quantification.de/bestkeeper.html, accessed on 20 December 2024) [15] assesses reference gene stability through the following three parameters calculated from Ct values: correlation coefficient (r), standard deviation (SD), and coefficient of variation (CV). Genes with significantly higher r (p < 0.05) with lower SD and CV values were identified as having optimal expression stability. Like geNorm, the NormFinder (https://www.moma.dk/software/normfinder, accessed on 20 December 2024) [14] algorithm assesses the stability of candidate RGs by analyzing variance in Cq values. NormFinder calculates a stability measure (SV) for each gene, where a lower SV indicates higher expression stability. RefFinder [16], a web-based platform (https://blooge.cn/RefFinder/, accessed on 20 December 2024), integrates Delta CT, geNorm, NormFinder, and BestKeeper outputs to assign stability ranks. Genes with lower rank values show superior expression stability.

3. Results

3.1. RNA Quality Analysis

Following RNA extraction, the total RNA integrity was assessed by agarose gel electrophoresis (Figure 1a, 4 °C of chilling; Figure 1b, −15 °C of freezing). The sharp and intact ribosomal RNA bands observed indicate minimal degradation and high RNA integrity. Spectrophotometric analysis (Table S2) showed RNA concentrations ranging from 80 to 500 ng/μL and A260/A280 ratios between 1.8 and 2.1 for almost all samples, confirming sufficient quantity and purity for subsequent experiments.

3.2. Specificity Validation of Primers for Candidate RG

The specificity of primers for the 19 candidate RGs was evaluated using agarose gel electrophoresis (Figure 2a). Except for GAPDH-2 and AP47, amplification products for all genes appeared as single, distinct bands within the expected size range of 100–200 bp (Supplementary Table S1), with no evidence of non-specific amplification or primer dimers; this confirms that the primers for the remaining 17 candidate genes meet the requirements. Melt curve analysis (Figure 2b) showed a single distinct peak for these 17 genes. No fluorescence signal was detected in the negative controls. Therefore, the primers for these 17 candidate RG exhibited excellent specificity and are suitable for further analysis.

3.3. Analysis of Ct Values of Candidate RGs

The expression stability of candidate RGs was primarily assessed based on their Cq values obtained from the qRT-PCR analysis. As shown in Figure 3 and Tables S3 and S4, the Cq values for 18S ranged from 11.5 to 15.5, which was lower than the range from 20.8 to 33.2 observed for the other 16 candidate genes. This low average Cq value for 18S indicates its high abundance, likely due to its high copy number. Analysis of the Cq value distribution showed distinct patterns under different low temperatures. When chilling at 4 °C, the Cq values for EF1α-1 showed lower variability (more clustered) and a lower mean than those of UBQ-2, which showed greater dispersion; this suggests that EF1α-1 expression was more stable under chilling conditions. Meanwhile, under the freezing treatment at −15 °C, GAPDH Cq values showed lower variability and a lower mean. Meanwhile, UBQ-1 Cq values were more dispersed; this indicates that GAPDH expression was more stable under freezing conditions.

3.4. Candidate Reference Gene Stability Evaluation Using the ΔCq Method

Stability evaluation using the ΔCq method showed each candidate gene’s mean SD of ΔCq values (Stability Value, SV) (Figure 4). Since a lower SV corresponds to higher expression stability, the relative stability rankings of the 17 candidate genes under chilling stress were determined as follows, from most to least stable: EF1α-1 (SV = 0.669) > EF1α-2 > Actin > MDH > GAPDH > Tubulin-2 > EF1α-3 > UBC > EF1γ > Tubulin > SAND > UBQ-3 > PP2A > UBQ-1 > 18S > TIP41 > UBQ-2 (SV = 1.325). Therefore, EF1α-1 exhibited the highest expression stability (lowest SV), while UBQ-2 showed the lowest stability (highest SV) under chilling conditions.

Similarly, for samples stored at −15 °C, GAPDH had the highest stability with the lowest SV (0.518). Meanwhile, EF1α-3 had the lowest stability with the highest SV (0.898). The stability ranking under freezing stress was, from most to least stable, GAPDH > Actin > EF1α-2 > TIP41 > EF1α-1 > UBQ-2 > SAND > 18S > PP2A > Tubulin > MDH > Tubulin-2 > UBQ-3 > EF1γ > UBQ-1 > UBC > EF1α-3. Derived from the ΔCq method, these results identified EF1α-1 and GAPDH as the most stable candidate RGs under chilling and freezing stress, respectively, based on SV magnitude.

3.5. Stability Assessment of Candidate RG Using the geNorm Algorithm

The geNorm algorithm determines the optimal number of RGs required by calculating the pairwise variation (Vn/n + 1). A Vn/n + 1 value below the threshold of 0.15 indicates that the inclusion of ‘n’ genes is sufficient (n ≥ 2). Conversely, a Vn/n + 1 value above 0.15 suggests that ‘n + 1’ genes are necessary. The V2/3 values under chilling and freezing conditions were below this threshold (0.15) (Figure 5a). Therefore, combining the two most stable genes is recommended for normalization under both temperature regimes.

Concurrently, the geNorm algorithm was used to evaluate the expression stability of individual candidate genes by calculating the average expression stability measure (M-value). A lower M-value signifies higher stability. Analysis of the M-values (Figure 5b) showed the following stability rankings: Under 4 °C treatment, EF1α-1 had the lowest M-value, followed by EF1α-2. Meanwhile, UBQ-2 had the highest M-value. Under −15 °C treatment, GAPDH showed the lowest M-value, followed by Actin, whereas EF1α-3 had the highest M-value.

Based on these M-value rankings, geNorm recommends the pair EF1α-1 and EF1α-2 as the optimal reference gene combination for samples under chilling stress, and the pair GAPDH and Actin for samples under freezing conditions.

3.6. Stability Evaluation of Candidate RG Using the NormFinder Algorithm

Implemented as an Excel add-in, the NormFinder algorithm uses a specific computational approach to evaluate reference gene stability. It calculates an SV for each gene, where a lower SV indicates higher expression stability. SV data analysis (Figure 6) showed distinct stability patterns under the two storage conditions. For chilling stress, the most stable genes were EF1α-1 (SV = 0.111), MDH (SV = 0.207), and EF1α-2 (SV = 0.212, ranked as second in the geNorm algorithm). Conversely, UBQ-2 was the least stable gene at this temperature. Meanwhile, under freezing stress, fully corroborated by the geNorm analysis, GAPDH (SV = 0.184) and Actin (SV = 0.185) showed the highest stability (lowest SVs), while EF1α-3 (SV = 0.568) was the least stable.

3.7. Stability Assessment of Candidate RG Using the BestKeeper Program

The BestKeeper analysis results are summarized in

Table 1. Comparison of BestKeeper stability parameters for candidate reference genes under chilling and freezing stresses.

4 °C					−15 °C
Gene	Std Dev	CV (%)	r	p-Value	Gene	Std Dev	CV (%)	r	p-Value
EF1α-1	0.388	1.770	0.940	0.001	GAPDH	0.356	1.665	0.947	0.001
GAPDH	0.524	2.303	0.940	0.001	Actin	0.504	2.238	0.937	0.001
MDH	0.522	2.208	0.979	0.001	EF1α-2	0.578	2.514	0.923	0.001
EF1α-2	0.554	2.541	0.900	0.002	Tubulin	0.679	2.812	0.944	0.001
UBQ-3	0.643	2.713	0.874	0.005	18S	0.612	4.711	0.973	0.001
Actin	0.642	2.814	0.863	0.006	UBQ-1	0.973	4.155	0.954	0.001
UBQ-1	0.733	3.008	0.832	0.010	UBQ-3	0.892	3.978	0.983	0.001
EF1α-3	0.701	2.491	0.790	0.020	MDH	0.845	3.398	0.973	0.001
SAND	0.589	2.246	0.736	0.038	EF1α-1	0.281	1.241	0.912	0.002
18S	0.679	5.384	0.720	0.044	UBQ-2	0.599	2.614	0.906	0.002
TIP41	0.481	1.934	−0.514	0.192	TIP41	0.452	1.711	0.904	0.002
Tubulin	0.536	2.289	0.490	0.217	PP2A	0.628	2.304	0.892	0.003
UBC	0.535	2.172	0.350	0.393	SAND	0.359	1.350	0.855	0.007
PP2A	0.414	1.523	−0.286	0.493	EF1γ	0.691	2.741	0.806	0.016
Tubulin-2	0.219	0.842	0.262	0.528	UBC	0.516	2.216	0.758	0.029
EF1γ	0.336	1.371	0.098	0.818	Tubulin-2	0.386	1.303	0.732	0.039
UBQ-2	0.885	3.805	0.040	0.924	EF1α-3	0.477	1.747	0.386	0.347

. Based on an evaluation of key statistical parameters, requiring a significant probability value (p < 0.05), high correlation coefficients (r), low SD, and low CV, the following candidate genes showed superior stability under each stress condition. EF1α-1 was the most suitable reference gene for chilling, with optimal values across all criteria. GAPDH, MDH, and EF1α-2 also ranked highly as strong secondary candidates. GAPDH and Actin were identified as the top two most stable genes based on their combined statistical performance for freezing.

3.8. Stability Ranking of Candidate RG Using RefFinder

RefFinder was used to generate a stability ranking (

Table 2. Comprehensive stability ranking of candidate reference genes under chilling and freezing stresses by RefFinder analysis.

	4 °C		−15 °C
Rank	Genes	Geomean of Ranking Values	Genes	Geomean of Ranking Values
1	EF1α-1	1.97	GAPDH	2.21
2	EF1α-2	2.78	Actin	2.74
3	MDH	3.22	TIP41	2.99
4	Tubulin-2	4.05	EF1α-1	3.34
5	Actin	4.12	EF1α-2	3.95
6	GAPDH	5.38	Tubulin-2	4.90
7	EF1γ	6.16	SAND	6.24
8	EF1α-3	7.36	UBQ-2	7.84
9	UBC	9.12	18S	8.44
10	Tubulin	9.21	PP2A	9.12
11	PP2A	10.03	Tubulin	10.68
12	UBQ-3	10.70	MDH	12.15
13	SAND	11.00	EF1α-3	13.10
14	TIP41	11.96	UBC	13.24
15	UBQ-1	14.21	UBQ-3	13.95
16	18S	14.49	EF1γ	13.98
17	UBQ-2	17.00	UBQ-1	15.46

). EF1α-1 was identified as the optimal reference gene, EF1α-2 ranked second in terms of stability, while UBQ-2 was the least stable gene under chilling stress. Meanwhile, under the freezing treatment, GAPDH was the optimal reference gene, followed by Actin in second place, and UBQ-1 had the lowest stability.

3.9. Validation in Selected RGs Using Target Gene Expression Analysis

To validate the reliability of the selected RGs, we analyzed the expression of VvCBF1, VvCBF2, and VvCBF3, which are key genes well-established in plant cold resistance studies. As shown in Figure 7, expression profiles of these target genes were assessed in two grapevine cultivars under chilling and freezing low-temperature stress. Gene expression was normalized using the following three distinct reference sets: (1) the most stable single reference gene, (2) the least stable single reference gene, and (3) the optimal combination of RG. This comparative analysis showed that, while the expression trends of VvCBF1, VvCBF2, and VvCBF3 across increasing cold exposure durations were largely similar regardless of the reference gene(s) used, statistically significant differences in expression levels were observed depending on the normalization strategy used.

As shown in Figure 7a of ‘Hanniang 1301’, under the chilling treatment, there were no significant differences when normalized with EF1α-1, and there was significant downregulation at 3 h with the optimal gene pair (EF1α-1 + EF1α-2) and at 6 h with UBQ-2 of VvCBF1. There was consistent expression under EF1α-1 and the gene pair, but it was divergent with UBQ-2 of VvCBF3. Similar patterns were observed for VvCBF2 and VvCBF3 in ‘Cabernet Sauvignon’ (Figure 7b).

In the cold-tolerant ‘Hanniang 1301’ under freezing treatment (Figure 7c), there were no differences across all RGs of VvCBF2 and VvCBF3. However, expression varied significantly among references at 12 h/24 h/48 h of VvCBF1. Meanwhile, the cold-sensitive ‘Cabernet Sauvignon’ (Figure 7d) showed consistent expression of VvCBF1. However, VvCBF2 GAPDH and the gene pair (GAPDH + Actin) showed concordance, diverging from UBQ-1. There were minor differences between GAPDH and the gene pair at 12 h vs. 48 h, but there were deviations with UBQ-1 of VvCBF3. These results validate the reliability and importance of the selected RG for accurate normalization.

4. Discussion

Low-temperature stress severely affects plant growth and development. There are distinct responses in grapevines (Vitis vinifera) to chilling stress above freezing versus freezing stress with subzero temperatures [31,32,33]. Chilling stress disrupts physiological and biochemical processes without intracellular ice formation, allowing potential recovery post-short-term exposure. In contrast, freezing stress induces irreversible physical damage from ice crystallization, often leading to plant mortality. Therefore, stably expressed RGs likely differ between these two stress types.

qPCR (quantitative real-time PCR) quantifies target gene expression by monitoring fluorescence signals during amplification cycles. Defined as the cycle number at which fluorescence exceeds a predefined threshold, the cycle threshold (Ct/Cq) value inversely correlates with gene expression levels. Lower Ct values indicate higher expression [34,35].

RGs should exhibit stable expression within a species under varying conditions, although optimal candidates differ across species and experimental contexts [8]. Here, 17 traditional candidate RGs, including Actin, EF1α-1, EF1α-2, EF1α-3, EF1γ, GAPDH, MDH, PP2A, SAND, TIP41, Tubulin, Tubulin-2, UBC, UBQ-1, UBQ-2, UBQ-3, and 18S, were evaluated for stability using the ΔCt method, geNorm, NormFinder, BestKeeper, and RefFinder. The optimal reference pairs were: EF1α-1 and EF1α-2 for 4 °C (chilling stress) and GAPDH and Actin for −15 °C (freezing stress).

EF1α-1 and EF1α-2 encode eukaryotic elongation factor 1α (EF1α) isoforms. A phylogenetically conserved protein, EF1α, mediates tRNA delivery to ribosomes and maintains stable expression across tissues [36].

EF1α-1 and EF1α-2 function as translation elongation factors with specialized molecular roles and cellular protective mechanisms [37]. Under non-lethal cold stress (e.g., 4 °C), these genes exhibit minimal functional disruption and superior stability. EF1α maintains translational continuity during low-temperature stress by facilitating aminoacyl-tRNA binding to ribosomes, thereby sustaining its own expression stability. In contrast, at −15 °C, EF1α undergoes significant structural and functional alterations, rendering it unsuitable as a reference gene under such conditions. Unlike EF1α, GAPDH and Actin demonstrate relative stability during freezing stress (−15 °C). This stability does not imply complete functional immunity to freezing damage but reflects significantly lower expression variability across treatment groups compared to other candidate genes within our specific experimental system (plant genotype/tissue type/freezing protocol).

The robustness of GAPDH and Actin [38] correlates with their multifunctional roles in stress responses; GAPDH serves pleiotropic functions beyond glycolysis, acting as a stress-response protein involved in oxidative stress management, DNA damage repair, and apoptosis regulation. This functional versatility necessitates its persistent cellular expression even under extreme stress. Actin, as a core cytoskeletal component, maintains stability at −15 °C through its critical involvement in structural integrity preservation and damage repair mechanisms.

Across diverse plant species, EF-1α and GAPDH show conserved stability as RGs. In barnyard grass (Echinochloa crus-galli), EF-1α and GAPDH maintained stable expression under multiple abiotic stresses [39]. EF-1α was validated as the optimal reference for root tissues of ramie (Boehmeria nivea) under combined stresses [40]. GAPDH also showed the highest stability across all tested samples of dangshen (Codonopsis pilosula) [41].

CBFs (C-repeat binding factors) are key transcription factors in cold stress response pathways and are extensively validated across plant cold resistance studies [42]. Elevated CBF expression triggers active cold adaptation by binding to CRT/DRE cis-elements in COR (cold-regulated) gene promoters, enhancing freezing tolerance [43]. CBF accumulation concurrently inhibits plant growth. We analyzed VvCBF1, VvCBF2, and VvCBF3 expression levels to validate our reference gene selection. While expression trends of these genes remained consistent across normalization methods (Figure 7), the statistical significance of differential expression varied with the RG used.

The primary objective of this study is to address the critical challenge of gene expression normalization in grapevines subjected to both chilling and freezing stresses. The identification of stable reference genes enables the precise quantification of transcriptional responses, thereby facilitating the discovery of core regulatory genes essential for low-temperature adaptation. Our findings provide an empirical foundation for elucidating the molecular mechanisms underlying cold tolerance in Vitis vinifera. Furthermore, this work identifies candidate genes associated with cold-hardiness traits, offering potential targets for marker-assisted breeding programs. The established framework may also serve as a methodological reference for analogous investigations in other plant species.

5. Conclusions

Screening and validating 17 candidate RGs revealed differences in their expression stability in wine grapes under different low-temperature stresses. Combining EF1α-1 and EF1α-2 was the optimal reference under chilling stress conditions. Combining GAPDH and Actin was the most suitable choice under freezing stress conditions.

This study identifies stably expressed reference genes under both chilling and freezing stresses in grapevine, enabling reliable normalization for gene expression analysis during distinct low-temperature responses. Ultimately, our findings establish foundational insights for elucidating cold adaptation mechanisms and advancing cold-hardy grape breeding.