Application of indirect in-package surface dielectric barrier discharge cold plasma for post-harvest storage of table grapes

Table grapes (Vitis vinifera var. Crimson Seedless) are highly susceptible to post-harvest deterioration, primarily due to weight loss, stalk browning, and spoilage by microorganisms like moulds. This study investigates the use of indirect in-package surface cold plasma within a passive Modified Atmosphere Packaging (MAP) to enhance microbial control and extend shelf-life of table grapes. Grapes were sealed in LDPE bags and treated with cold plasma for 3, 7, or 10 s (3S, 7S, 10S) and stored at 0 °C for 18 days. Ozone generation and its subsequent decomposition within the headspace were monitored. The most important physico-chemical and physiological parameters including headspace gas composition, weight loss, berry and stalk colour, texture, and microbial populations were systematically assessed at days 0, 3, 12, and 18. Cold plasma generated high initial ozone concentrations (up to 1050 ppm for 10S) that degraded within 60 min. Treatments significantly reduced yeast and psychrotrophic bacteria counts compared to controls until day 12. While headspace gas composition (O 2 and CO 2) remained stable across treatments, a significant decline in sensory score for aroma and flavour was observed in treated samples primarily due to a pungent off-odour that persisted during consumption. In-package cold plasma represents a rapid, residue-free sanitization method. Although the treatment did not significantly alter the internal quality or metabolism of the berries, results suggest an interaction with the packaging material. Therefore, while the technology is effective for microbial reduction, the specific impact of plasma on plastic properties and its subsequent influence on sensory attributes requires further investigation.

Table grapes are a fruit of increasing global economic and nutritional importance, appreciated not only for their organoleptic qualities but also for their high content of vitamins, antioxidants, and other essential nutrients, making them a valuable addition to a healthy diet. According to data from the International Organisation of Vine and Wine, global grape production in 2023 represented approximately 45.8% of total grapes production, around 31 million T. The growing consumer demand for fresh and natural products highlights the need for effective postharvest preservation strategies. However, despite their increasing commercial success and health benefits, table grapes are susceptible to rapid deterioration. Key postharvest challenges include weight loss, browning of the stalk, and, crucially, vulnerability to attack by spoilage microorganisms, particularly moulds, which severely limit their shelf-life and marketability (Palou et al., 2010; De Simone et al., 2020). To counteract these issues, the adoption of advanced preservation technologies is crucial. Modified Atmosphere Packaging (MAP) has become an established and effective technique (Piazzolla et al., 2021). Nevertheless, even with MAP, complete control of microbial proliferation, especially of specific moulds responsible for decay, remains a persistent challenge for packaged table grapes. The need for innovative solutions that are both effective, safe, and aligned with consumer expectations for “clean” products free from chemical residues is more pressing than ever (De Chiara et al., 2023; Chauhan and Rao, 2024). In this scenario, exploring additional antimicrobial agents that can be synergistically integrated with MAP has become a priority in postharvest research. Ozone (O 3) has long been recognized for its exceptional oxidizing power and broad antimicrobial properties, making it an effective disinfectant against bacteria, viruses, fungi, and spores (Xue et al., 2023). Its ability to decompose into oxygen, leaving no harmful chemical residues, makes it particularly attractive for application on fresh food products. Traditional ozone applications, however, can present limitations, such as the need for extended exposure times to achieve desired efficacy or potential safety risks for personnel due to its reactive nature if not adequately controlled. To overcome these limitations and maximize ozone’s effectiveness, an innovative approach involving its generation directly inside the package (in-package) is being developed. This methodology offers substantial advantages. Firstly, it allows for extremely rapid treatment, on the order of few seconds, making it compatible with high-speed production lines. Secondly, in-package treatment virtually eliminates the risk of post-process recontamination, as the product is already sealed within its protective environment. Furthermore, this technique enables simultaneous sanitization of the packaging’s inner surface, further contributing to the overall hygienic safety of the product. Direct in-package application is made possible by emerging technologies like cold plasma. Cold plasma, the fourth state of matter, is a partially ionized gas generated, also at atmospheric pressure, at low temperatures that produces reactive oxygen and nitrogen species (Lu et al., 2016; Patil et al., 2016). These highly reactive species are generated directly within the package headspace, where they can diffuse and effectively react with microorganisms, compromising their viability (Misra et al., 2019; Sasikumar et al., 2025). As highlighted by recent comparative studies, the efficacy of cold atmospheric pressure plasma is significantly influenced by the technology employed, with different prototypes showing varying levels of microbial inactivation and impacts on product quality (Mishra et al., 2026). While most current literature on in-package treatments focuses on Volumetric Dielectric Barrier Discharge (VDBD), where the produce is positioned directly within the discharge gap (Misra et al., 2019), this study explores the use of Surface Dielectric Barrier Discharge (SDBD). In this specific configuration, the plasma is generated along a dielectric surface, allowing for an indirect treatment where the fruit is exposed only to the diffusion of reactive species (Sasikumar et al., 2025), potentially reducing the risk of direct surface damage. This approach is crucial for delicate matrices such as table grapes, as direct plasma exposure has been shown to cause significant fruit tissue destruction and mechanical softening when excessive treatment intensities or prolonged exposure times are applied (Khalaj et al., 2023). Consequently, this study aims to investigate the application of high ozone levels, produced in very short times using an indirect in-package surface cold plasma generator, within a MAP environment. The primary objective of this research is to enhance microbial control in packaged table grapes. Simultaneously, the study evaluates the impact of this innovative approach on the physicochemical and physiological quality parameters of the product. This research seeks to make a significant contribution to the development of sustainable, efficient, and safe preservation solutions for table grapes, addressing the market’s increasing demand for fresh, high-quality products with an extended shelf-life.

Preparation of packaged table grapes and application of treatments

The grapes (Vitis vinifera var. Crimson Seedless) were obtained from Società Agricola Difesa di Canneto di Del Core Massimiliano & C. S.S., a farm locate in via Da Vinci, 69, 70,016 Noicattaro, Bari, Italy. Grapes were transported to the laboratory of the Postharvest Technology Research Unit of the University of Foggia, and approximately 200 g of small bunches of table grapes were placed in Polyethylene terephthalate (PET) trays (16 × 4 × 10 cm) and sealed in low-density polyethylene (LDPE) bags measuring 17 × 21 cm with a thickness of 20 µm with an internal volume approximately 2.5 L, in order to achieve a passive modified atmosphere packaging. The thickness was selected in accordance with literature values for gas permeability KPO 2 190,000 and KPCO 2 70,000 cm 3 [STP] μm m –2 bar –1 24 h –1 (Piergiovanni and Limbo, 2010). The in-packaging cold plasma prototype (SAIM Postharvest Technology, Fondi (LT), Italy) was used to create an enriched passive MAP with Ozone using different treatment times: 3 s (3S), 7 s (7S) and 10 s (10S). Untreated samples were used as a control (CTRL). The treatments were applied at 14 °C with the instrument set to the values of 8 kV and 20 kHz. These operating conditions were chosen based on preliminary trials to maximize ozone production (data not shown). Three packages of grapes were prepared for each treatment and each sampling day, while 3 replicates of untreated grapes were used for initial quality determinations. After treatment grapes were stored at 0 °C for 18 days and analysed after 3, 12, and 18 days.

Residual ozone measurement and headspace gas composition analysis of packages

Residual Ozone Concentration (ppm) was measured in the package headspace using an ozone analyser (Model UV100, Eco Sensors, Division of KWJ Engineering Inc., Newark, CA, United States). In addition, the headspace composition of O 2 and CO 2 (kPa) was analysed using a gas analyser (CheckMate 3, Dansensor, Ringsted, Denmark). The analysis was performed on 3 different experimental units for each treatment.

Colour analysis of berries and stalks

Colour analysis was performed using a spectrophotometer colorimeter (CM-2600dG, Konica, Minolta, Japan) by measuring L, a and b, from which the parameters Chroma (C), hue angle (h°) and ΔE were calculated using the following equations:

For each bag, representing an experimental unit, 8 berries were analysed. Colorimetric parameters of the stalks were extracted by image analysis from photos, obtained in controlled light conditions, using a colour digital camera (Canon, EOS 400D, United States), and using ImageJ software (NIH, 1.53 t version, United States). Five areas of all the stalks present in the packages were sampled for each photo, from which the mean of colorimetric parameters L, a and b* were obtained, and from which the remaining colorimetric parameters (Chroma, h°, ΔE) were calculated. The analysis was performed on 3 different experimental units for each treatment at each sampling time.

Berry texture and resistance to detachment

Compression and tensile tests were performed at 20 °C using a TA-XT2 Texture Analyzer (Stable Micro Systems, United Kingdom) equipped with a 50 kg load cell. For the compression test, eight randomly selected berries were analysed for each bag. A 7.5 cm flat probe was used to compress the samples at a constant speed of 1 mm s -1 to a deformation of 5%, and the maximum force (N) was recorded. For the resistance to detachment of berry (RTD), eight berries were also analysed for each experimental unit at the same temperature of 20 °C. The grape stalk was secured to a clamp connected to the instrument, and traction was applied at a constant speed of 1 mm s -1 until detachment occurred. Also in this case, the maximum force (N) was extracted.

Sensorial analysis

The analysis was carried out by a group of 5 expert panellists (2 males and 3 females aged from 25 to 51) by evaluating aroma, flavour, sweetness, crispness, fizziness, crunchiness, sensorial resistance to detachment (SRTD), acidity, browning of stalks (BOS) and overall quality. Each of the parameters were rated using a 5-point hedonic scale, from 5 to 1, where: 5 = excellent; 4 = very good; 3 = good (limit of marketability); 2 = acceptable (limit of edibility); 1 = unsatisfactory (inedible). For each treatment, 3 bags were analyzes, testing 3 berries per bag.

Weight loss

The trays were weighted entirely with the fruit inside during the preparation stage and then re-weighed on the day of sampling. Weight loss was calculated as the difference between the initial weight and the final weight over the initial weight, and expressed as a percentage (%).

Where W l:% is the weight loss expressed as %, W i is the initial weight and W f is the final weight of grapes. The weight loss was calculated on 3 distinct experimental units per treatment at each sampling day.

Total soluble solids, pH and Titratable acidity

Total Soluble Solids (TSS), pH and Titratable Acidity (TA) were determined by analysing three bags per treatment. For each bag, three berries were pooled, pressed together and filtered through a single layer of gauze (JC NONSTE SWAB 4040, Shanghai, China), and the resulting juice was used for all physiochemical measurements. The TSS was directly measured using a digital refractometer (Atago N1, PR32-Palette, Tokyo, Japan) and expressed as °Brix. The pH was measured directly on the extracted juice, while for TA determination, 1 g of juice was diluted 1:40 with distilled water. The analysis was performed via an automatic titrator (T50 Mettler Toledo, Switzerland) by titrating against a 0.1 mol L -1 NaOH solution (Sigma-Aldrich, Burlington, MA, United States) up to a final pH of 8.1. TA results were reported as a percentage of tartaric acid.

Microbiological analysis

Under sterile conditions, 10 g of grapes were homogenized with a salt solution (0.9% NaCl) (Sigma, Milan, Italy) at a ratio of 1:10. Decimal dilutions of the homogeneous sample were made using the same diluent and inoculated into selective media to determine the specific microbial groups. Plate Count Agar (PCA) was used to enumerate the total number of mesophilic loas incubated at 30 °C for 48 h and total psychotropic load by incubating the plates for 10 days at 4 °C. Yeasts and moulds were determined by Potato Glucose Agar (PGO) selective medium supplemented with chloramphenicol (0.1 g L -1) with incubation at 25 °C for 2 and 5 days, respectively. A microbiological analysis of these microbial groups was performed for each experimental unit, expressing the results as log CFU g -1.

Ozone decomposition kinetics

The ozone decomposition kinetics was built using bags filled with 200 g of grapes stored at 0 °C; three bags were analyzed for each sampling time (0, 3, 8, 20, 25 and 30 min) and for each treatment. The ozone concentration was monitored using an ozone analyzer (Model UV100, Eco Sensors, Division of KWJ Engineering Inc., Newark, CA, United States). The experimental fata were fitted using a non-linear first-order kinetic model. In the equation n. 6, obtained by solving the equation n. 5, [O 3 ] t is the concentration of ozone (ppm) at time t; [O 3 ] 0 is the concentration of ozone (ppm) at time t = 0 min; t is the time (min) and k is the decomposition time constant (min -1) of ozone.

Then, the estimated model’s parameters of Eq. n. 6 were used to calculate the time of the exposure of grapes to ozone, using the equation n. 7 and a limit value of ozone equal to 0.1 ppm.

The fitting of data was performed by R software (R Core Team, Version 4.5.1, Austria).

Statistical analysis

Results were collected during the storage period at 0, 3, 12 and 18 days, and used for statistical analysis of the data. A two-way ANOVA was conducted to test the significance of the selected factors (treatment time and storage duration) and their interaction on the qualitative variables considered (p < 0.05) and a one-way ANOVA for individual sampling days (p < 0.05). Finally, a post-hoc test was performed, specifically Tukey’s test (p < 0.05). Specifically for the colorimetric parameters of the stalk (L, a, b*), a two-way Multivariate Analysis of Variance (MANOVA) was first performed to evaluate the global effect of the factors on the overall colour profile. Subsequently, two-way ANOVAs were conducted for each individual chromatic variable to identify which parameters contributed to the observed multivariate effects. The strategy for post-hoc comparisons was differentiated based on the interaction results: where a significant interaction was detected, a post-hoc test on the interaction (Tukey’s HSD, p < 0.05) was performed to compare all treatment-time combinations. Conversely, where the interaction was not significant, a one-way ANOVA was conducted for each individual sampling day to evaluate the treatment effect at specific time points. All statistical analyses were performed using JMP software (Version 16, SAS Institute Inc, NC, United States).

Ozone decomposition kinetics and exposure time of grapes to residual ozone

Upon generation, the residual ozone concentration within the packages was found to be proportional to the treatment time. The highest initial values were recorded for the 10S samples (1055.7 ± 177.0 ppm), followed by the 7S (811.3 ± 88.6 ppm) and 3S (420.3 ± 10.0 ppm) treatments. Over the monitored period, these levels decreased rapidly across all units. In particular, the concentration in the 3S samples fell to 5.9 ± 7.1 ppm within 20 min. A similar reduction was observed for the longer treatments, with the 7S group reaching 4.8 ± 3.9 ppm after 25 min and the 10S group decreasing to 6.7 ± 0.5 ppm by the 30-min mark. The reparametrized first-order kinetic model (Equation n. 5) was used to describe the ozone concentration data as a function of time. Table 1 summarizes the fitting parameters for ozone degradation models. The first order kinetics well described the decomposition of ozone of each treated samples, showing a high correlation coefficient equal or higher than 0.97 for each treatment time and an high significance for k value and [O 3 ] 0 concentration. The estimated initial ozone concentrations were 424.6 ± 18.4 ppm for sample 3S, 882.6 ± 56.5 ppm for sample 7S, and 1050.5 ± 74.6 ppm for sample 10S. This trend is consistent with previous findings; for instance, De Oliveira et al. (2020), Alencar et al. (2011) reported similar decomposition trend of ozone over time. The high significance (p < 0.0001) and the magnitude of the k parameter reflect a rapid decomposition rate, due to the high instability of the ozone which reacts very quickly with substrate. Despite a slight change in numbers the decay rate did not change significantly among treatment, since the conditions such as temperature, quantity of grapes and size of the bags remained constant; only the treatment time varied, which implies a different initial residual concentration of ozone in the headspace of the bags among the different treatments. Although the root mean squared error (RMSE) values increase with concentration (16.65, 51.90, and 91.03 for the 3S, 7S, and 10S samples, respectively), these values remain acceptable as they occur during the initial stages at high ozone concentrations. As the process continues, the model becomes remarkably accurate: at low concentration points, the difference between measured data and predictions (residuals) drops to minimal values (as low as 0.96). This confirms that the regression is highly reliable during the final phase, ensuring an accurate estimation of the total ozone degradation time. Based on the good fit of the data to the model, an ozone concentration limit of 0.1 ppm was set to calculate the exposure time of the grapes to ozone. Regarding the estimated exposure times, sample 3S had the shortest duration at 39.8 ± 4.4 min. Sample 7S exhibited an exposure time of 50.5 ± 10.5 min, while sample 10S showed the longest estimated exposure time, reaching 57.9 ± 13.9 min. The calculation of these times has a dual purpose: firstly, it defines the precise duration of exposure of the grapes to ozone, which is directly related to antimicrobial efficacy; secondly, it ensures the safety of the process by establishing the waiting period necessary for the ozone to degrade to safe levels before the fruit is handled or consumed. However, it is crucial to highlight that the estimated exposure times found in this work are not easily comparable with those from other studies. Variables influencing these kinetics, including temperature, the type of vegetable substrate (in this case, grapes), the type of cold plasma source/ozone method production and the headspace volume, are intimately related to the specific experimental conditions of this research. In fact, the temperature of the treatment was 14 °C, limit chosen according to the Italian law concerning minimally processed product (Ministerial Regulation n. 3746 20 June 2014). Despite these specificities, the evaluation of residual ozone decomposition in this study allowed for the estimation of the exposure time of the packaged product to ozone and, moreover, enabled the calculation of a safe consumption time for products packaged with ozone-enriched MAP.

Evolution of head-space gas composition over time

The concentrations over time of O 2 (a) and CO 2 (b) in the headspace of the packaging are reported in Figure 1. The O 2 levels exhibit a downward trend for both treated and non-treated samples; however O 2 levels did not show appreciable differences in the first 12 days. On day 18, the samples treated for 3 and 10 s exhibited a decrease in oxygen by about 2 percentage points compared to the CTRL and 7S. On the other hand, a CO 2 level exhibit a growing trend for each samples. Also CO 2 concentration showed the largest differences between treatments after 18 days, with the sample 7S showing a 0.5% lower CO 2 level then the other samples. However, the difference between the samples did not result statistically significant for both O 2 and CO 2 gases. Analysis of the atmospheric composition within the packaging revealed a general decrease in O 2 levels and a concomitant increase in CO 2 levels for all samples, indicating a product respiration process. However, the observed variations were limited, with O 2 levels never dropping below 16% and CO 2 levels remaining well below 2%. This suggests that the adopted packaging conditions and the duration of the treatments employed did not induce variations in atmospheric composition significant enough to have marked physiological effects on the product. This findings are in agreements with several studies where ozone treatment did not significantly alter gas exchange or physiological metabolism of fruits and vegetables. Palou et al. (2002) reported no significant effect on the respiration rates of peaches, even with continuous gaseous ozone exposure at 0.3 ppm for up to 3 weeks at 5 °C. Similarly, Horvits and Cantalejo (2010) found no significant differences in gas atmosphere evolution in ozone-treated red bell pepper. Conversely, other research indicates that ozone can indeed modulate respiration. Gao et al. (2020) demonstrated that gaseous ozone at 14.98 mg m -3 effectively inhibited the respiration rate of Muscat Hamburg grapes, maintaining a lower rate compared to untreated controls, and Li et al. (2025) similarly suggested that ozone fumigation at 2.0 mg L -1 slowed metabolic processes in table grapes. Importantly, a reduced respiration rate in horticultural products is often associated with a decrease in microbial counts, as microbial growth significantly contributes to CO 2 production. For example, Misra et al. (2014) observed that plasma-treated strawberries showed a decreased respiration rate over time, which they attributed to reduced microbial loads, beneficial for product quality. In addition, when the treatments were less effective in controlling microbial growth, an increase in respiration was noted; Horvitz et al. (2021) reported that 0.4 ppm O 3 -treated blackberries, applied as gaseous ozone for 3 min with a fan for even distribution, had the highest CO 2 production, which the authors linked to increased microbial growth by the end of storage. In this study the ozone did not affect the gas composition of the headspace of the packages. The absence of a significant impact on gas composition reinforces the idea that product respiration was not substantially modulated by the applied treatments. However, this outcome is not necessarily negative. While some literature suggests ozone can actively reduce respiration, a maintained or unaltered respiration rate by ozone treatment could be considered beneficial for shelf-life extension, as it indicates the treatment does not induce physiological stress leading to increased metabolic activity. An example is the work of Horvits and Cantalejo (2010), in which ozone effectively contributed to quality preservation without significantly altering the respiration profile of the fruits, highlighting that ozone can be used with MAP to increase the shelf-life of red bell peppers without having a metabolic stress on the fruits and consequently without altering their quality attributes.

Global results of qualitative parameters

The results of two-way ANOVA on the effect of time of exposure to plasma, of the storage time, and of their interaction on the quality attribute of packaged grapes store up 18 days are resumed in Table n. 2. The treatment influenced each microbiological group analysed and a few chemical, physical and sensorial parameters of the grapes. Furthermore, the treatment showed a highly significant effect on aroma and flavour, consequently affecting the overall quality of the berries, as well as the physical and sensorial parameters of the stalk, with the exception of resistance to detachment (RTD). In general, storage time exhibited a significant effect on all parameters, with the exception of total soluble solids (TSS). The interaction between Treatment time and Storage time (B) resulted statistically significant only on some chemical and physical properties, without affecting microbiological and sensory attributes. Generally, interaction effects are less prevalent than the main effects of Treatment and Storage times. This suggests that although there are some interactions between the two factors, their individual effects are more important in determining product properties.

Microbiological profile evolution during postharvest storage

Grape sample microbial loads for each group of microorganisms are reported in Figure 2. Initial mesophilic load value (2a) was 4.58 ± 0.22 log CFU g -1. At day 3, a general decline was recorded across all samples. CTRL decreased to 3.10 ± 0.17 log CFU g -1, while treated samples showed lower counts, with 10S reaching the minimum value of 2.00 ± 0.00 log CFU g -1. Subsequently, microbial loads increased across all groups. By day 18, although the treated samples maintained numerically lower counts compared to the control, no statistically significant differences were observed (p > 0.05). Pinto et al. (2020) observed that ozone pre-treatment on various berries significantly reduced mesophilic counts by 0.3–1.3 log CFU g -1, confirming its efficacy regardless of the modified atmosphere packaging applied. In this study, similarly, an initial high mesophilic load of around 5 log CFU g -1 was observed in grape samples, which subsequently decreased across all treatments by day 3. This initial reduction aligns with findings by Ziuzina et al. (2020), who reported that static Atmospheric Cold Plasma treatment, applied for 2.5 min at 100 kV in an in-package continuous process, reaching 200–400 ppm of O 3, effectively reduced mesophilic bacteria on strawberries and spinach. However, the subsequent microbial growth observed after day 3, with treated samples maintaining lower loads (around 3.5 log CFU g -1) compared to controls (near 5 log CFU g -1) by day 12, contrasts with the observation of Ziuzina et al. (2020), who reported significantly inhibited growth in continuous ACP-treated strawberries (around 800 ppm of O 3), even under temperature abuse conditions. This suggests that while the treatments applied to grapes effectively reduce initial microbial counts, they may not offer the same sustained proliferation control as seen with ACP on other produce. The duration of treatment could be a critical factor (Sasikumar et al., 2025): while our treatment involves a maximum exposure of plasma in the head space of 10 s, without considering the residual ozone present in the packaging after treatment, other ACP studies, such as Ziuzina et al. (2020) on strawberries, utilized significantly longer treatment times, up to 2.5 min of continuous treatment, allowing for a prolonged interaction between the reactive species generated by the plasma and the microbial cells thereby maximizing inactivation. Indeed, Mishra et al. (2024) found that optimised cold plasma initially reduced the total aerobic count to below the limit of detection; however, subsequent regrowth narrowed the gap compared with the control samples by the end of the 28-day storage period. These findings may be attributed to the greater distance between the plasma source and the product: the antimicrobial effectiveness of cold plasma treatment is strongly influenced by the distance between the emitter and the target surface (Niemira et al., 2018). Results related to psychrotrophic bacteria are shown in Figure 2b. The initial concentration was 2.83 ± 0.90 log CFU g -1. At day 3, a reduction was observed, with CTRL recorded at 2.20 ± 0.35 log CFU g -1 and all treated samples remaining at the lowest detectable level of 2.00 ± 0.00 log CFU g -1. While no statistical significance was found at this stage (p > 0.05), a clear divergence occurred at day 12. Here, CTRL increased markedly to 3.81 ± 0.29 log CFU g -1, whereas the treated samples remained significantly lower (p < 0.05). Specifically, the treated samples clustered in the same statistical group with values of 2.16 ± 0.28 log CFU g -1 for 3S, 2.28 ± 0.49 log CFU g -1 for 7S, and 2.20 ± 0.35 log CFU g -1 for 10S. By the end of the trial at day 18, CTRL reached 4.03 ± 0.11 log CFU g -1, while treated samples increased but remained lower: 3S reached 3.59 ± 0.66 log CFU g -1, 7S reached 3.48 ± 0.44 log CFU g -1, and 10S recorded 3.00 ± 0.00 log CFU g -1. Although the statistical analysis did not detect significant differences at this final step (p > 0.05), the STD bars of CTRL and 10S did not overlap, pointing towards a possible influence of the treatment. Microbial proliferation is the primary cause of postharvest quality loss in table grapes, making its inhibition essential for extending storage life. Although it is widely established that reduced O 2 levels combined with elevated CO 2 can significantly slow microbial growth, no substantial differences in headspace gas composition were recorded among the treatments in this study. Consequently, the observed antimicrobial effects must be attributed specifically to the plasma treatment and the action of its generated reactive species, rather than to passive atmospheric variations. In fact, the effectiveness of ozone, a key reactive species of cold plasma, against a broad spectrum of microorganisms has been well-documented in various treatments on fruits. The ability of the plasma treatments to maintain nearly constant psychrotroph levels until day 12 confirms the antimicrobial efficacy of ozone against this group, aligning with Horvitz et al. (2021) and Horvitz and Cantalejo (2010), who reported stable reductions in psychrotrophic counts on blackberries and peppers using gaseous ozone. However, a notable divergence in our study is the “convergence of values” at day 18. While Horvitz et al. (2021) maintained low counts throughout storage, our study shows a resumption of growth. This suggests that while ozone effectively inhibits initial psychrotroph proliferation, the sustained efficacy may vary depending on different parameters like initial psychrotroph concentration, the type of product and the way of application of ozone.

The assessment of mould growth is illustrated in Figure 2c. Counts started at an initial value of 2.56 ± 0.07 log CFU g -1 at day 0. By day 3, a significant divergence was observed (p < 0.05). CTRL increased to 3.00 ± 0.00 log CFU g -1, while the treated samples showed reduced levels, clustering in the same statistical group. By day 12, the counts increased across all conditions. CTRL reached 3.57 ± 0.38 log CFU g -1, whereas 10S remained lower at 3.00 ± 0.00 log CFU g -1. Although statistical analysis showed no significant differences at this stage (p > 0.05), the STD bars of CTRL and 10S did not overlap, implying a potential efficacy of the treatment. At the end of the storage period on day 18, CTRL peaked at 4.96 ± 0.46 log CFU g -1. This value was significantly higher (p < 0.05) than 3S and 7S, which grouped together with values of 3.83 ± 0.16 log CFU g -1 and 3.80 ± 0.17 log CFU g -1, respectively, while 10S recorded an intermediate value of 4.30 ± 0.52 log CFU g -1. Furthermore, despite the recorded mould loads, no visible mould development was observed on the grapes in any of treated groups until the end of storage period. Conversely, only the control samples showed a minimal onset visible mould, characterized by approximately one single spot per bag at day 18. The initial suppressive effect aligns with Admane et al. (2018) and Gabler et al. (2010), who reported decay control in grapes using ozone. Furthermore, Cold Atmospheric Plasma (CAP) can trigger spore apoptosis through ROS accumulation (Cao et al., 2023). Although statistical significance fluctuated during storage, the treatment’s impact is best evidenced by the delay in visible mould growth and the fact that, by day 18, certain treatments (3S and 7S) still maintained a significant reduction of over 1.1 log CFU g -1 compared to the control (p < 0.05). The efficacy is typically attributed to membrane damage and the inhibition of biosynthesis pathways, such as ergosterol, leading to cell inactivation (Li et al., 2025). As shown in Figure 2D, the initial yeast population was 4.28 ± 0.25 log CFU g -1. At day 3, a general decrease was observed across the board. CTRL dropped to 3.00 ± 0.00 log CFU g -1, while the treated samples showed lower values. Specifically, 7S reached the minimum limit of 2.00 ± 0.00 log CFU g -1, proving to be significantly different from CTRL (p < 0.05). By day 12, a clear statistical separation emerged (p < 0.05). CTRL rose to 4.30 ± 0.23 log CFU g -1, whereas all treated samples clustered in statistical group with significantly lower counts: 2.66 ± 0.56 log CFU g -1 for 3S, 2.94 ± 0.12 log CFU g -1 for 7S, and 3.19 ± 0.16 log CFU g -1 for 10S. By day 18, CTRL peaked at 4.73 ± 0.26 log CFU g -1. Although treated samples maintained numerically lower counts (averaging 4.00 log CFU g -1), no statistical significance was observed (p > 0.05), indicating a narrowing of the gap between groups at the end of storage. Finally, the observed reduction in yeast levels on treated grapes in this study, exhibiting a significant decrease by day 3 and sustained efficacy until day 12, aligns with findings from other research on ozone’s antimicrobial properties. Specifically, Gao et al. (2020), utilizing a 14.98 mg m -3 ozone concentration, reported effective reductions in yeasts, on ‘Muscat Hamburg’ grapes. Similarly, Sitoe et al. (2025) found that ozone gas at 30.10 mg L -1 (with exposure times of 3, 5, or 10 min) drastically lowered grape yeast concentration, often below detection limits, demonstrating superior initial and persistent control for up to 21 days. Overall, the global microbial trends indicate that while the plasma treatment provides an effective initial decontamination, particularly evident in psychrotrophs and the delay of visible mould, its long-term efficacy could be limited by a number of factors. The convergence of values observed towards day 18 suggests that a short exposure may cause sublethal injuries rather than total inactivation; consequently, once the residual effect of the reactive species decreases, the robust cell walls of yeasts and established fungal colonies allow for microbial regrowth. These differences in sustained efficacy may be attributed to several factors, including the modality of ozone application, the specific grape cultivar (Gabler et al., 2010; Ozkan et al., 2011), and the physiological state and the initial contamination level of microorganisms. As previously documented, ozone and plasma-generated species exhibit greater effectiveness when applied during early infection stages or on spore germination, rather than on fully developed colonies (Gabler et al., 2010; Li et al., 2025; Sharpe et al., 2009). Furthermore, as previously highlighted, several technical parameters can significantly influence treatment efficacy, such as the power settings of the plasma source, the composition of the gas carrier, and the specific geometry of the electrode system, which collectively determine the concentration and penetration depth of the generated reactive species.

Evolution of physicochemical parameters of berries and weight loss during storage

Titratable acidity remained substantially stable compared to the initial value (1.21 ± 0.10), except for slight fluctuations observed on day 3 (data not shown). The final values, ranging between 1.17 ± 0.10 and 1.29 ± 0.08, confirm the absence of statistically significant differences among samples (p > 0.05). Starting from an initial value of 19.9 ± 0.8, the TSS content showed a stable trend with only minor fluctuations throughout the storage period (data not shown). By the end of the experiment, values settled between 19.2 ± 0.8 and 20.3 ± 0.7; importantly, no significant differences were observed among samples at any sampling time (p > 0.05). Despite some apparent fluctuations observed on day 3 (data not shown), also the pH values maintained a stable trend relative to the initial point (3.5 ± 0.2). The final measurements, falling between 3.6 ± 0.2 and 3.9 ± 0.12, further confirmed that the treatment significantly did not affect the acidity (p > 0.05). The assessment of TSS, pH, and acidity is crucial, as these parameters are closely correlated with fruit maturity and directly impact sensory quality and organoleptic characteristics. This lack of significant differences indicates that the single “in-package” treatment, even though high intensity of the treatment (8 kV, 20 kHz), did not induce a “metabolic shock” or accelerate ripening, a conclusion strongly corroborated by the internal gas composition data, which showed uniform respiration rates across all samples; in fact it only acted as a surface decontaminant without altering the physiology of the fruit. While the electrical discharge lasts only 10 s, the reactive species generated are confined within the LDPE packaging, creating a residual active atmosphere estimated to last approximately 1 hour before decaying. This treatment is sufficient for antimicrobial action but too transient to permanently shift the berry’s metabolism. These findings align with Ziuzina et al. (2020) and Artés Hernández et al. (2007), who similarly observed that single, optimized treatments ensured safety without altering TSS or pH. Conversely, our results differ from Gao et al. (2020) and Li et al. (2025), who reported that treated grapes maintained significantly higher TSS and acidity than controls with a slower respiration compared to untreated samples. The discrepancy could be explained by the protocols: Gao et al. applied ozone for 30 min daily, and Li et al. performed fumigations every 3 days. This divergence suggests that the cumulative physiological impact of cyclic applications could play a key role. The repeated, long-duration exposure employed in those studies likely functions as a chronic stressor, potentially leading to a general deceleration of metabolic activity. In contrast, our single initial treatment represents an acute event that does not appear to trigger such systemic suppression. Weight loss showed a progressive increase throughout the storage period. While significant differences among samples were detected at day 12 (p < 0.05), these variations disappeared by the end of the experiment (data not shown). At day 18, values ranged from 0.67 ± 0.87 and 0.74 ± 0.68 respectively for 10S and CTRL, reaching 1.28 ± 0.17 for the 3S (data not shown), with no statistically significant differences observed (p > 0.05). WL is a critical factor in table grape preservation, directly impacting visual appeal and consumer acceptance. Throughout the entire storage period, the weight loss remained consistently low across all experimental groups. This overall stability is primarily attributed to the packaging material rather than the treatment itself. The LDPE film utilized for the Passive MAP acts as an effective barrier to water vapor, maintaining a high relative humidity environment around the bunches that drastically limits transpiration, regardless of the plasma exposure. Notably, no visible tissue damage were observed in the treated bunches; this evidence strongly corroborates the WL data, as the maintenance of skin integrity prevented any accelerated transpiration typically triggered by surface injuries. This finding aligns with Admane et al. (2018) and Artés-Hernández et al. (2004), who similarly reported that mass loss is contained below 1% or 0.5%. Our data sharply contrast with Vlassi et al. (2018) and Palou et al. (2002), who reported a significant increase in weight loss following ozone exposure. This discrepancy is attributable to the cyclic nature of the treatments applied in those studies. It appears that repeated or daily applications exacerbate dehydration. As highlighted by Du et al. (2022), ROS and RNS seem to increase water permeability. Indeed, this phenomenon is not novel; cold plasma is extensively investigated as a pre-treatment to accelerate the drying process of fruits and vegetables. In our context of fresh storage, however, the absence of such weight loss suggest that our single 10 s or less treatment, and our exposure to residual ozone was sufficiently mild to avoid an increase in weight loss for the grapes, also using more than 1000 ppm of ozone. The grape texture exhibited a general downward trend from the initial value of 7.9 ± 2.5 (data not shown). Although a divergence occurred at day 3, with sample 3S maintaining a higher firmness compared to the others (p < 0.05), the values realigned during the final storage period. By day 18, no significant differences were detected, with final values ranging between 5.3 ± 1.4 and 6.4 ± 2.3. The treatment had a limited effect on the texture of the berries, with no significant differences observed between treated and control samples at the end of storage. This observation generally aligns with Ziuzina et al. (2020), Hegde et al. (2025) and Artés Hernández et al. (2007), who similarly reported that static cold plasma and ozone treatments, respectively, did not significantly alter the firmness of strawberries and grapes compared to untreated controls. However, divergences exist in the literature where treatments actively preserved texture better than the control. For instance, Gao et al. (2020) found that ozone treatment significantly extended grape firmness, effectively delaying softening. Similarly, Misra et al. (2014) demonstrated that plasma treatment maintained strawberry firmness significantly better than controls, but they specifically noted that this effect was tied to a high-oxygen environment inside the package. By contrast, Lacombe et al. (2015) using an atmospheric cold plasma jet device observed that treatments exceeding 60 s significantly reduce blueberry firmness. These differences highlight that effectiveness in preserving firmness depends not only on the type of technology but heavily on specific application conditions, such as atmospheric composition and exposure duration. In our study, the specific “flash” treatment applied were neutral, probably avoiding the structural modifications observed in the other studies. Starting from an initial value of 37.50 ± 4.37, the L parameter remained constant over time (data not shown). Although the values were numerically close, a significant difference was detected at day 3 (p < 0.05): specifically, the 10S sample showed significantly lower luminosity compared to the 3S and 7S samples. However, this statistical significance was not maintained in the subsequent days. By day 18, the final values were very similar: 41.17 ± 3.07 (CTRL), 39.14 ± 4.86 (3S), 39.03 ± 2.58 (7S), and 40.62 ± 4.35 (10S). The a* values showed a pattern of fluctuations relative to an initial value of 6.60 ± 3.82 (data not shown). Within this variable trend, a transient significant difference was observed at day 3 (p < 0.05), where the 10S treatment differed significantly from the 3S group. This differentiation did not persist in the following stages. The final profile at day 18 showed values of 8.12 ± 3.18 for CTRL, 5.05 ± 4.78 for 3S, 8.63 ± 2.91 for 7S, and 5.64 ± 3.89 for 10S. The b* values exhibited an upward trend relative to the initial value of 5.18 ± 2.15, despite some oscillations (data not shown). A transient significant difference was noted at day 3 (p < 0.05) between the 3S and 10S treatments. This differentiation did not persist in the following stages. Ultimately, the CTRL and 10S samples settled at 7.61 ± 2.13 and 7.86 ± 2.30 respectively, while 3S and 7S reached 8.19 ± 2.39 and 7.10 ± 3.12. Starting from an initial value of 9.21 ± 2.13, Chroma exhibited a fluctuating trend, visually peaking at day 12 before declining (data not shown). By day 18, the final values were recorded as follows: 11.50 ± 2.37, 10.70 ± 2.20, 11.47 ± 3.33, and 10.32 ± 2.55 for CTRL, 3S, 7S, and 10S respectively. Despite these oscillations, no statistically significant differences were found among samples over the time (p > 0.05). The h° value was initially 41.62 ± 25.29, followed by a series of fluctuations throughout storage (data not shown). A significant difference emerged at day 3 (p < 0.05), specifically distinguishing the 3S sample from the 10S one. However, this statistical significance was not maintained in the subsequent days. Upon conclusion of the storage period, the recorded values were 44.54 ± 16.14 for CTRL and 39.34 ± 14.02 for 7S, while 3S and 10S reached higher levels of 60.87 ± 25.57 and 56.47 ± 22.18. The ΔE values showed a temporal evolution characterized by a progressive rise followed by a decline relative to the initial value (data not shown). Despite these changes over time, no significant differences were found among the varying treatments (p > 0.05). At the end of the experiment, the final values reached were 4.96 ± 1.16 (CTRL), 4.81 ± 2.43 (3S), 3.81 ± 1.54 (7S), and 4.64 ± 1.08 (10S). The observed data for these parameters were characterized by notable fluctuations across all samples. This high variability could be attributed to the intrinsic heterogeneity of the raw material rather than the plasma treatment itself. This uneven pigmentation is a known physiological phenomenon where high temperature conditions could inhibit the accumulation of anthocyanins, leading to uneven berry coloration (Ferrara et al., 2013; Spayd et al., 2002). Consequently, the oscillations observed in our samples reflect this natural non-uniformity in pigment accumulation. This internal biological variability directly influenced the h°; since this parameter is mathematically dependent on the a* and b* values, its high variability is an inevitable consequence of the fluctuations recorded in the primary chromatic coordinates, confirming that the observed differences stem from the initial fruit quality rather than the treatment conditions. Therefore, the biological variability of the raw material remained the dominant factor, overshadowing any potential minor modifications induced by the plasma treatment. While certain studies report an effective retention of L* and h° (Admane et al., 2018; Haci and Özcan, 2018), others suggest a limited efficacy of ozone (Sitoe et al., 2025; Sharpe et al., 2009). In these instances, color variations are attributed primarily to intrinsic raw material heterogeneity or storage duration rather than the treatment itself (Horvitz and Cantalejo, 2010; Pinto et al., 2020). This interpretation mirrors our findings, where biological variability prevailed over the processing conditions.

Evolution of physicochemical parameters of stalks during storage

The overall color profile of the stalks of the samples, characterized by L, a, and b* coordinates, was evaluated using a two-way MANOVA to determine the global influence of treatment and storage time. The analysis revealed that the colorimetric profile was significantly affected by both main factors: Time (Wilks’ λ = 0.186, p < 0.001) and Treatment (Wilks’ λ = 0.598, p < 0.001). Most importantly, a significant interaction between Time and Treatment was detected (Wilks’ λ = 0.696, p = 0.009), indicating that the evolution of the color of the samples over the 18-day storage period followed divergent patterns depending on the specific treatment applied. Since the MANOVA provides only a global assessment of the colorimetric profile, a follow-up two-way ANOVA (Table 2) was performed for each individual coordinate to better understand the contribution of each parameter to this interaction. The interaction effect was found to be exclusively driven by the a* (p = 0.0015), whereas the L* and b* parameters were influenced by the main factors independently (p > 0.05 for the interaction). These findings justify the subsequent detailed analysis of a* evolution. Specifically, for the a* coordinate, a Tukey’s HSD post-hoc test was performed directly on the two-way ANOVA interaction effect to identify significant differences among all treatment-time combinations. Conversely, for the remaining chromatic variables, where the interaction was not significant, a day-by-day comparison was conducted using one-way ANOVA followed by Tukey’s HSD test to evaluate significant differences between treatments at each sampling point. The colorimetric parameter of stalks over the time are shown in Figure 3. The L* value (3a) showed a progressive decrease compared to the initial value of 67.05 ± 7.88 recorded at day 0 for all the samples. By day 3, the effect of the treatment was already statistically significant, with the CTRL maintaining a higher L* value of 64.72 ± 3.95 compared to all treated samples, which ranged from 54.48 ± 2.58 (10S) to 58.68 ± 2.93 (7S), all clustered together. This trend continued through day 12 where the CTRL sample, remained statistically distinct from the treated groups, which exhibited lower values. At the final measurement on day 18, the CTRL sample reached a value of 55.89 ± 2.89, while the treated samples finished with lower lightness values: 49.77 ± 4.42 (3S), 45.98 ± 4.96 (7S), and 49.44 ± 5.65 (10S). Changes in a* values over time for the stalk are shown in Figure 3b. The initial value was −22.69 ± 3.74. A significant divergence occurred as early as day 3, where the treated samples exhibited a sharp increase in a* values compared to the control. Specifically, the 10S group reached a value of −6.79 ± 2.96, proving to be statistically different from the CTRL group (p < 0.05), which remained significantly greener at −15.83 ± 3.97. This gap established at day 3 persisted throughout the remainder of the storage period. By day 18, while the CTRL group remained the greenest (−14.52 ± 6.06), a substantial increase in a* values was observed across all treatments, indicating a progressive de-greening process. Notably, the 7S group reached the highest value (−3.61 ± 6.35), representing the most significant chromatic shift. Based on the interaction analysis, the 7S treatment differed statistically from both the CTRL and the 3S group (p < 0.05). The initial b* value was 30.01 ± 5.40 (3c). By day 3, most groups showed a slight increase. However, there were no statistical differences between any of the groups at this point. After this, a general decreasing trend (loss of yellowness) began. A clear separation was evident by day 12: the CTRL (30.15 ± 6.34) and 7S (30.12 ± 3.92) groups maintained significantly higher b* values than the 3S (24.34 ± 3.26) and 10S (23.88 ± 1.44) groups (p < 0.05). This divergence continued until the end of storage. On day 18, the CTRL group (26.79 ± 4.46) and 7S group (25.98 ± 2.65) retained the most yellowness. Conversely, the 3S (21.53 ± 1.55) and 10S (21.01 ± 4.72) groups showed the most pronounced decrease. At this storage time the 10S value was statistically lower than both the CTRL and 7S groups (p < 0.05). This progression suggests that the 3S and 10S treatments accelerated the loss of yellow colour, while the 7S treatment behaved similarly to the control. At the beginning of the storage period, the Chroma value was 37.70 ± 6.00 (3d). By day 3, all values had begun to decrease slightly but showed there were no statistically significant differences between any of the groups at this stage. A clear separation emerged by day 12. The CTRL (35.16 ± 7.35) and 7S (31.81 ± 4.18) groups maintained significantly higher colour saturation (p < 0.05) compared to the 3S (26.56 ± 4.22) and 10S (26.38 ± 1.99) groups, which showed a more pronounced decline. This trend continued to the end of storage on day 18, where the 3S (24.82 ± 3.52) and 10S (23.08 ± 5.08) samples showed the lowest values. Statistically, these two groups remained significantly lower in Chroma than the CTRL group (31.20 ± 2.46). This indicates that the 3S and 10S treatments significantly accelerated the loss of colour saturation. Figure 3e illustrates the variation over time of the hue angle (h°). For all samples, the h° value always remained in the second quadrant of the CIE Lab* colour space, starting at an initial value of 127.19 ± 4.29 on day 0. By day 3, a clear decreasing trend was observed, reflecting a shift from green to more yellow/brownish hue. The 10S group showed the most significant drop to 101.93 ± 4.74 lower (p < 0.05) than both the CTRL group (115.82 ± 6.31) and the 3S group (111.11 ± 7.41). This trend shifted slightly by day 12, where the CTRL group (120.73 ± 3.53) maintained a hue angle statistically higher than the 7S group (107.29 ± 7.52). At the end of storage (day 18), the 7S group experienced the most pronounced and significant decline, finishing at 96.77 ± 13.85. This value was statistically lower than all other groups, which finished clustered together and were statistically similar. This progression confirms that the control samples consistently maintained higher h° values, suggesting better preservation of the original colour. Finally, the overall colour variation is reported in Figure 3f. The ΔE showed distinct trends between the control and treated samples over time. At day 3, a statistically significant difference was observed between the control (CTRL) and the treated samples (3S, 7S, and 10S). At this time point, the CTRL sample showed a ΔE of 8.89 ± 3.23, whereas the treated samples exhibited much higher values, such as 17.09 ± 3.19 (3S), 17.07 ± 2.63 (7S), and 20.44 ± 0.91 (10S). This numerical gap increased as storage progressed. By day 12, the CTRL sample reached a ΔE of 13.62 ± 3.16, while the treated samples continued to show higher values, ranging from 22.70 ± 4.09 (3S) to 27.08 ± 7.55 (7S). At this stage, however, the differences between all groups were no longer statistically significant. At the final measurement on day 18, the CTRL sample reached a final ΔE of 15.51 ± 2.96. The treated samples finished with higher values: 22.24 ± 4.75 (3S), 28.75 ± 8.06 (7S), and 24.08 ± 6.87 (10S). The final ΔE for the 7S sample was almost double that of the control. Furthermore, the standard deviation error bars for CTRL and 7S did not overlap at all. Although the observed differences at day 12 and day 18 were no longer statistically significant, the numerical data showed that all treated samples (3S, 7S, and 10S) ended the observation period with higher ΔE values than the untreated control sample. The results indicate that the plasma treatment significantly accelerated the colour evolution of the rachis during storage. The analysis of colorimetric parameters reveals a marked deviation compared to the control, evidenced by total colour difference values consistently higher in treated samples from the early stages of storage. Collectively, these trends, lead to the conclusion that the treatment effectively accelerated the browning processes of the rachis compared to the natural senescence observed in the control, as can be seen in the collage of some representative samples for the different treatments during the storage period (Figure 4). This observation aligns with the findings of Gabler et al. (2010), who noted that oxidative treatments (specifically ozone) can cause “rachis injury,” characterized by dark lesions or irregular browning. However, our data diverge from other studies where the ozone/cold plasma treatments successfully preserved the visual quality of the stalks. For instance, Admane et al. (2018) observed that ozone maintained higher chlorophyll content and green colour than controls. In contrast, in the present work, the h° value of stalks for the 7S sample exhibited a marked decline, passing from vibrant green to yellow-green. Similarly, Surya et al. (2024) and Palou et al. (2002) reported superior rachis appearance in treated bunches without signs of phytotoxicity. Feliziani et al. (2014) reported that While continuous exposure to a low ozone concentration of 0.150 μL L -1 for 3 weeks was able to preserve the green appearance of the rachis better than storage in air, a prolonged treatment of about 10 weeks at 0.300 μL L -1 caused severe oxidative injury, resulting in nearly complete browning of the stem. To understand this discrepancy, it is necessary to consider the physiological mechanisms underlying rachis browning. This phenomenon is typically driven by two main factors: water loss (dehydration) and enzymatic oxidation. In our study, weight loss of the whole bunches was extremely low (<1.6%) for all groups and comparable among treatments at the end of storage. This observation suggests that dehydration may be excluded as the primary cause of the observed darkening, as highlighted by Hegde et al. (2025) where ozone-treated samples have been shown to exhibit lower physiological weight loss and significantly less rachis dehydration compared to untreated control. Since the tissues remained hydrated, the accelerated browning in treated samples may be attributable to the oxidative stress induced by Reactive Oxygen and Nitrogen Species (RONS) generated by the plasma. This abiotic stress presumably triggered the direct degradation of chlorophylls and the activation of oxidative enzymes. However, the available literature specifically addressing the impact of ozone and plasma on stalk browning and detailed colour parameter changes remains limited. While some studies touch upon overall appearance, a comprehensive understanding of the mechanisms behind the effect of oxidative treatments on colour degradation and browning in grape stalks is still developing. These discrepancies underscore the complex and varied effects of treatments on stalk colour stability, suggesting that the phenomenon should be studied more deeply, including the content of chlorophyll, the activity of enzymes, and the concentration of polyphenols involved in the browning event. The assessment of detachment resistance started with an initial value of 2.8 ± 0.9 N (data not shown). Over the course of the storage time, this parameter exhibited several fluctuations among the different groups. At day 3, CTRL was recorded at 2.8 ± 1.1 N, while the treated samples ranged between 2.7 ± 0.9 N for 7S and 3.1 ± 1.3 N for 10S. By the last sampling day at day 18, all values showed a decline: CTRL dropped to 2.0 ± 0.6 N, while 3S recorded 2.6 ± 1.4 N, 7S was at 2.4 ± 0.8 N, and 10S ended at 2.5 ± 0.8 N. Overall, despite these fluctuations, no statistically significant differences were found from the initial day to the final sampling day (p > 0.05). Despite the visual degradation of the rachis, the mechanical resistance to detachment remained statistically unchanged compared to the control. This suggests that the plasma-induced oxidative stress was primarily confined to the epidermal tissues, causing aesthetic acceleration of browning without compromising the deeper structural integrity of the pedicel’s vascular connections. Specific studies on the direct effect of plasma on mechanical detachment force are currently limited. However, our findings align with indirect observations in related studies, such as Admane et al. (2015), who observed that ozone treatments combined with modified atmosphere packaging successfully maintained rachis quality without inducing berry drop or shattering. Since the detachment force is strictly correlated with the turgidity and health of the pedicel, the absence of main effect of the plasma on resistance to detachment suggests that the plasma exposure presumably was not severe enough to trigger the formation of abscission layers or dehydrate the pedicel connection, thereby maintaining the structural integrity of the bunch comparable to the control.

Evolution of sensorial parameters during storage

Sensory attributes, specifically Crunchiness, Sweetness, Sourness, Fizziness, and RTDS, exhibited a consistent downward trend from their initial values (ranging from 4.80 to 5.00) throughout the storage period (data not shown). Despite this general decline, statistical analysis revealed no significant differences among the treatments at any stage (p > 0.05). Importantly, by the end of the trial (day 18), all samples reached statistically comparable levels and consistently maintained mean scores above the marketability limit (>3.0). Specifically, Crunchiness varied from 3.3 ± 0.8 (3S) to 4.4 ± 0.4 (10S), with the CTRL recorded at 3.7 ± 0.7. Sweetness scores settled between 3.9 ± 0.9 (CTRL) and 4.4 ± 0.4 (10S). Sourness values for treated samples fell within the 3.9–4.1 range, compared to 3.7 ± 0.8 for the control. Finally, Fizziness and RTDS showed the mean control values of 4.0 ± 1.0 and 3.4 ± 1.1 respectively, showing no significant deviation from treated samples, which ended with mean values ranging from 3.8 to 4.4 for Fizziness and 3.2 to 3.8 for RTDS. The evolution of BOS, flavour, aroma and overall quality of the stalks are shown in Figure 5. The initial value of BOS was 4.8 ± 0.4 (5a). At day 3, a significant difference emerged among the conditions (p < 0.05). CTRL recorded a value of 4.7 ± 0.4, while 10S showed a significantly lower score of 3.5 ± 0.5. By day 12, scores continued to decline across all groups. CTRL was recorded at 4.0 ± 0.0, while the treated samples ranged from 2.8 ± 0.8 for 10S to 3.3 ± 0.4 for 7S. Although the statistical analysis showed no significant differences at this stage (p > 0.05), the STD bars of CTRL and 10S did not overlap, indicating a potential impact of the treatment. On the final sampling day at day 18, the values further decreased. CTRL ended the trial at 3.5 ± 0.7, while the treated samples recorded scores of 2.9 ± 0.9 for 3S, 2.4 ± 0.9 for 7S, and 2.3 ± 01.0 for 10S. Despite these numerical differences, no statistically significant effects were detected (p > 0.05), though the STD bars of CTRL and 10S remained non-overlapping, pointing towards a possible influence of the treatment. The assessment of flavour scores began with an initial value of 4.8 ± 0.4 (5b). Throughout the storage period, a general decline in sensory quality was observed across all groups. At day 3, scores decreased to 4.6 ± 0.9 for CTRL, while treated samples recorded values of 3.5 ± 1.0 for 3S, 3.7 ± 1.2 for 7S, and 3.0 ± 1.0 for 10S. Despite this numerical drop, no statistically significant differences were found at this stage (p > 0.05). Similarly, at day 12, the statistical analysis confirmed the absence of significant effects (p > 0.05), with CTRL at 4.1 ± 0.7 and treated samples ranging from 3.7 ± 1.1 for 3S to 2.7 ± 1.0 for 10S. A significant divergence finally emerged on the last sampling day at day 18 (p < 0.05). CTRL maintained the highest score of 3.90 ± 0.89, significantly outperforming 10S, which dropped to 2.2 ± 1.3. Meanwhile, 3S and 7S showed intermediate values of 3.3 ± 0.7 and 2.5 ± 0.5. The initial value of aroma was 5.0 ± 0.0 (5c). A progressive decline in perceived quality was observed across all samples throughout the storage period. At day 3, a significant difference was recorded (p < 0.05); CTRL maintained a higher score of 4.2 ± 0.8, while 10S dropped significantly to 2.4 ± 0.5. During this stage, 3S and 7S showed intermediate behaviour, both sharing a value of 2.9 ± 0.9. By day 12, the statistical analysis showed no significant differences among the conditions (p > 0.05). CTRL was recorded at 3.4 ± 0.6, while treated samples showed lower average values of 3.2 ± 0.8 for 3S, 2.7 ± 0.7 for 7S, and 2.3 ± 0.7 for 10S. Similarly, on the final sampling day at day 18, no statistically significant effects were detected (p > 0.05). CTRL ended the trial at 3.7 ± 0.6, whereas treated samples ranged from 2.6 ± 0.9 for 3S to 2.2 ± 1.1 for 10S. At both day 12 and day 18, the STD bars of all groups consistently overlapped, confirming the absence of a significant treatment effect during the final phases of storage. The initial value of overall quality was 5.0 ± 0.0 (5d). Throughout the trial, a general decline in scores was observed across all conditions. At day 3, a significant difference emerged (p < 0.05), with CTRL recording the highest score of 4.7 ± 0.3. In contrast, 10S dropped significantly to 3.1 ± 0.7, while 3S and 7S showed intermediate values of 3.8 ± 0.4 and 3.6 ± 1.1. By day 12, CTRL remained at 4.0 ± 0.3, which was significantly higher than the treatments 7S and 10S (p < 0.05) with a values respectively of 2.7 ± 0.7 and 2.6 ± 0.7, while 3S maintained an intermediate score of 3.7 ± 0.5. On the final sampling day at day 18, the statistical separation was confirmed (p < 0.05). CTRL ended the study at 3.8 ± 0.8, significantly outperforming 7S and 10S, which remained clustered in the lowest statistical group with specific values of 2.4 ± 0.5 and 2.2 ± 0.8, respectively. As it was reported in Table 1, in-package plasma treatments did not affect sensorial parameters related to sweetness, sourness, crunchiness, and to the presence of frizzy taste. While minor fluctuations and a slight general decline were observed as storage progressed, these changes were not statistically different among the treatments (data not shown). Importantly, these sensory observations are fully supported by the instrumental measurements, confirming the reliability of the panel’s evaluation. The lack of significant variations in crunchiness mirrors the texture analysis results, indicating that the berry structure was preserved. Similarly, the stability in sweetness aligns with the Total Soluble Solids (TSS) measurements, where no significant changes were detected. Furthermore, the low scores recorded for fizziness correspond to the minimal CO 2 concentrations measured in the package headspace, suggesting that fermentative processes were negligible throughout the storage period. Conversely, all the other tested sensorial parameters were significantly affected by the treatments. Score for the stalk browning decreased over time for all samples, although with varying rates according to the treatment. This sensory assessment, particularly the rapid decline observed at day 3 in the most intense treatment (10S), is in agreement with the colorimetric data; this convergence suggests that the rachis tissue suffered an initial oxidative shock, a phenomenon captured by both the panellists and the instrumental readings. Existing literature is not very rich in references that evaluate the impact of ozone on grape quality. However, in our study, the most critical finding concerns the significant reduction in Aroma, Flavour, and Overall Quality scores in the treated samples. Specifically, the panellists attributed these low scores to the perception of a pungent or burnt off-odour in the headspace of the packaging. Considering that flavour is defined as the result of the interaction between gustatory sensations and the retronasal perception of aroma, the marked decline of this parameter implies that the pungent “plastic-like” or “burnt” off-flavours were not merely confined to the package headspace but were perceptible on the grapes during consumption. Moreover, Riganakos et al. (1999) and Stoffers et al. (2004), who performed oxidative treatments on packaging using electron beams, highlighted similar results regarding the smell of the plastic. Tyapkova et al., 2009 also highlighted the formation of plastic odors and pungent odors after treatment with ionizing radiation, which are also oxidizing agents. Therefore, the formation of these odors is not entirely new in the context of in-packaging treatments. To explain this phenomenon, it is necessary to consider the interaction between the packaging and the product. As highlighted by Ashraf et al. (2023), Volatile Organic Compounds (VOCs) emitted from packaging materials can significantly influence the quality of fresh produce, as these compounds can be adsorbed by the fruit matrix, altering its sensory profile. Despite these evidences, the exact origin of these compounds in the specific context of this study remains uncertain. It is currently unclear whether the production of these off-flavours stems from the direct contact between the plastic material and the plasma source, from the chemical-physical interaction between the plasma discharge and the polymer film, or from the high concentration of ozone accumulated in the headspace in contact with the plastic. Notably, this sensory defect was not highlighted by other researchers who applied different in-package plasma systems, suggesting that the issue is likely related to the specific configuration of the discharge geometry in our prototype. However, to the best of our knowledge, there is a lack of literature clarifying which of these mechanisms is predominant in causing sensory defects in plasma-treated grape packaging. Therefore, further targeted research is required to fill this knowledge gap by identifying the specific VOCs and their genesis. Understanding this interaction is a fundamental prerequisite for the industrial application of this technology; such investigation is essential not only to ensure that the antimicrobial efficacy of cold plasma does not compromise the sensory acceptability of the product but, most importantly, to guarantee the safe use of the technology and safeguard consumer health.

The plasma treatment resulted in high concentration ozone exposure, with concentrations and duration proportional to the treatment time. This exposure exerted a positive effect on microbial inhibition but negatively impacted stalk quality and specific sensory attributes, while weight loss and fruit physiology remained unaffected. Crucially, the MAP conditions did not vary across treatments, implying equal respiration activity, and the headspace gas composition remained comparable to air. This stability suggests that the observed effects can be attributed exclusively to the plasma treatment rather than to atmospheric variations. Consequently, Ozone-enriched (MAP) has shown promising results in reducing microbial growth in packaged table grapes without exhibiting any effect on the physiological processes of the fruit and minimally impacting most quality characteristics. However, this approach presents challenges, including the observed stalk browning and the development of a distinctive off-odor in the package headspace. While these initial results underscore the technology’s potential, further investigation is crucial. Future studies should focus on identifying and mitigating the source of the unpleasant odor, optimizing treatment parameters such as frequency and electrical intensity, and, critically, evaluating different types of packaging films. Exploring new films is essential, as alternative materials may prove less reactive and thus more suitable for this combination of in-packaging cold-plasma and passive MAP.