‘Cavolo Riccio’ and ‘Cima Di Rapa Di Febbraio’ Brassica Landraces as Complementary Sources of Health-Promoting Phytochemicals: The Impact of Boiling and Steaming

1. Introduction

Brassicaceae stand out as one of the primary sources of bioactive compounds in the Mediterranean diet, thanks to their high content of glucosinolates, polyphenols, vitamin C, and carotenoids, substances known for their potent antioxidant and anticarcinogenic properties [1,2,3].

There is strong evidence linking the regular consumption of these vegetables, such as broccoli (Brassica oleracea var italica Plenck.), kale (Brassica oleracea var. acephala DC.), and broccoli raab (Brassica rapa subsp. sylvestris (Lam.) Janch.), to a significant reduction in the risk of chronic and degenerative diseases [4,5,6].

This has increased growing consumer interest in “functional vegetables”, leading foods like kale to become popular as “superfoods” due to their exceptional phytochemical composition [7,8]. Indeed, kale, including its baby-leaf products recently proposed for the fresh-cut market [9], is recognized as an excellent source of essential minerals, dietary fibre, folates, phytosterols, flavonoids, vitamin C, carotenoids, and other antioxidant molecules. Among these bioactive compounds, glucosinolates represent the most characteristic secondary metabolites of the Brassicaceae family that are essential for the plant’s natural defense mechanisms and possess extraordinary nutraceutical properties in humans [10,11,12]. Through enzymatic hydrolysis, they generate bioactive compounds such as isothiocyanates and indoles, which are known for their potent anticancer, antioxidant, and anti-inflammatory properties [13,14,15,16]. These compounds not only give vegetables such as broccoli raabs their characteristic pungent and bitter flavor, but also act as protective factors against chronic and cardiovascular diseases [17,18]. While bitterness in Brassicaceae is a complex sensory trait influenced by both specific glucosinolates (e.g., epiprogoitrin and progoitrin) and phenolic compounds [19,20,21], the pungent and sharp flavor is primarily associated with aliphatic glucosinolates, such as sinigrin, and their corresponding volatile isothiocyanates [22,23]. Their presence further justifies classifying these vegetables as “superfoods” in the modern diet [8].

In this context, local Brassica crops represent an irreplaceable heritage of agricultural biodiversity, which is essential not only for the resilience of traditional cropping systems, but also for enriching the Mediterranean Diet with high-value functional foods [24,25].

Apulian curly kale, commonly known as ‘Cavolo riccio’) is a vegetable with a century-old history, deeply embedded in the agricultural and culinary traditions of the Apulia region. Although it is now considered a vegetable at risk of genetic erosion, it remains a valuable resource for local biodiversity, as it is grown primarily in small peri-urban gardens for personal consumption, particularly in the Bari metropolitan area [26]. Rich in bioactive compounds with powerful antioxidant and anti-inflammatory properties, kale contains glucosinolates, vitamins (C, E, K), and carotenoids (β-carotene, lutein, zeaxanthin) [27,28], as well as a rich profile of phytosterols, polyphenols, flavonoids, and anthocyanins [9,29]. Phytosterols are particularly valued for their ability to reduce intestinal cholesterol absorption, thereby contributing to the maintenance of healthy blood cholesterol levels and reducing cardiovascular risk [30,31].

Broccoli raab (Brassica rapa subsp. sylvestris (Lam.) Janch., known as ‘Cima di rapa’) is an ancient Mediterranean crop strongly linked to Apulia. Appreciated for the pungent flavor imparted by glucosinolates, it is a pillar of the culinary heritage and cultural identity of Southern Italy, so much so that of the 10,713 hectares cultivated in Italy, 9,151 hectares are cultivated in Southern Italy, 3,540 hectares are cultivated in Apulia [32,33,34]. ‘Cima di rapa’ represent a valuable genetic resource, and the diversity of local varieties reflects the history, dialects, and customs of the region’s various rural areas.

In addition to glucosinolates, ‘Cima di rapa’ is characterized by a nutritional profile rich in dietary fiber and phenolic compounds (such as flavonoids and flavonols), which provide high antioxidant capacity, particularly concentrated in the leaves [35]. Indeed, in local Apulian ‘Cima di rapa’ germplasm, flavonols represent the dominant phenolic class, being approximately 2.5 times more abundant than hydroxycinnamic acid derivatives, and representing the main contributors to the hydrophilic antioxidant capacity of the sprouts [36]. This antioxidant potential is highly genotype-dependent; such as the rare local landraces of marginal areas of Foggia ‘Cima di rapa dalla testa’ and ‘Cima di rapa antica’ exhibit outstanding levels of total phenols (up to 389.7 mg GAE kg⁻¹ FW) and vitamin C (up to 495.2 mg kg⁻¹ FW), resulting in a hydrophilic antioxidant activity up to 4-fold higher than common commercial cultivars [22].

Notably, both ‘Cima di rapa’ and ‘Cavolo riccio’ are included in the regional list of Traditional Agri-food Products (TAPs) of Apulia – a category legally defined by Italian Ministerial Decree no. 350/1999 as comprising products «whose methods of processing, preservation, and aging are consolidated over time» for at least twenty-five years and recognized through a voluntary, bottom-up process driven by territorial actors [37] – a status that further attests to their embeddedness in local food culture and reinforces the rationale for their valorization.

Notably, the two crops also differ substantially in the edible plant part traditionally consumed and, consequently, analyzed in this study: at commercial maturity, ‘Cavolo riccio’ is typically consumed for its leaves only [8,20], whereas ‘Cima di rapa’ is harvested and consumed as a combination of leaves and immature inflorescences [5,22]. This distinction is not trivial, as glucosinolate levels are known to vary significantly among young leaves, mature leaves, and inflorescences within the same plant [11,22], meaning that the two landraces are compared not only as different genotypes but, to some extent, as different edible products [22,32].

These two Brassicaceae are typically eaten cooked, in keeping with local culinary traditions [38,39]. Raw consumption is further limited by the tough texture and the pungent and bitter flavor of these vegetables, although in recent years, the market for minimally processed, ready-to-use (IV-gamma) baby leafy greens has expanded rapidly to meet consumer demands for both convenience and health benefits, such as ‘Cavolo nero’ (Brassica oleracea (Acephala group)) baby leaves reported by [9]. While cooking softens cell walls and may enhance the extractability of certain compounds such as carotenoids and bound polyphenols, it simultaneously inactivates myrosinase, the enzyme responsible for glucosinolate hydrolysis into bioactive isothiocyanates [3,40,41,42].

The most common conventional cooking methods for many types of vegetables, including these two, are boiling and steaming [3,43,44]. Boiling is one of the most destructive cooking methods due to the combination of heat and direct contact with water; since many bioactive compounds (such as glucosinolates, polyphenols, vitamin C, and minerals) are water-soluble, they tend to dissolve and leach into the cooking water [4,11]. On the contrary, steaming is the method that best preserves both nutritional and sensory properties [4,45]. Since the vegetable is not submerged in water, there is minimal loss due to leaching, and heat transfer via water vapor is less disruptive to the vegetable’s integrity; consequently, the texture is also better preserved than when boiled [45,46]. However, to date, no study has directly compared the thermal response of ‘Cavolo riccio’ and ‘Cima di rapa’ under the same experimental conditions, nor has the genotype been explicitly considered as a modulating factor in a statistically rigorous design.

The aim of this study was to compare the phytochemical composition of two traditional Apulian Brassica landraces, ‘Cavolo riccio’ and ‘Cima di rapa di febbraio’, and to evaluate how different cooking methods (boiling and steaming, with raw material as control) affect their bioactive compound profiles. To our knowledge, this is the first study simultaneously comparing locally conserved Apulian ‘Cavolo riccio’ and ‘Cima di rapa di febbraio’ landraces under identical cultivation, processing and analytical conditions. By jointly examining genotype-related differences and thermal processing effects, this study seeks to clarify the potential nutritional complementarity of these landraces, provide evidence-based recommendations for domestic cooking practices, and support the valorization of local Apulian vegetable biodiversity.

2. Materials and Methods

2.1. Location and Cultivation

Two local populations of Brassica spp. from the DISSPA-UNIBA germplasm collection were selected for this study. Specifically, one population belonged to ‘Cima di rapa di febbraio’ (Brassica rapa subsp. sylvestris (Lam.) Janch.), whereas the other belonged to ‘Cavolo riccio’ (Brassica oleracea var. acephala DC.). Information on genebank accession ID, geographic origin, and donor is reported in Table 1. The plant material was grown at the experimental farm “P. Martucci” of the University of Bari (41°01′22.1″N, 16°54′21.0″E) on a clay loam red soil. The soil was characterized by 41.11% sand, 31.02% silt, and 27.87% clay, with a pH of 7.75, 2.35% organic matter, 1.48 g kg⁻¹ total nitrogen, 18.00 mg kg⁻¹ available phosphorus, 260.00 mg kg⁻¹ exchangeable potassium, and 10.75% total limestone [47].

Both landraces were sown on 29 September 2025 and harvested on 12 January 2026. One week before sowing, the soil was rotary tilled. One day before sowing, ammonium phosphate fertilizer was applied at a rate of 300 kg ha⁻¹, providing 54 kg N ha⁻¹ and 138 kg P ha⁻¹. Two months after sowing, ammonium nitrate was applied at a rate of 500 kg ha⁻¹, supplying an additional 171 kg N ha⁻¹.

Seeds were sown continuously along the rows and seedlings were subsequently thinned to 0.3 m within the row. With 1.0 m between rows, the final planting density was 3.33 plants m⁻².No irrigation was applied throughout the crop cycle.

Pest and disease management consisted of single applications of deltamethrin, acetamiprid, and difenoconazole, whereas weed control was performed manually after crop emergence.

2.2. Field Experiment and Agronomic Analyses

The experiment was arranged in a Randomized Complete Block Design (RCBD) with three blocks (replicates). Each experimental unit consisted of two rows of five plants each, covering a total area of 3 m². Experimental units were separated by 0.5 m to avoid border effects.

The study included agronomic, morphological, physical and chemical determinations, as detailed in the following sections. Morphological characterization was performed at commercial maturity on 10 plants per replicate using descriptors developed by the Italian Working Group on Agricultural Biodiversity (GIBA), based on the guidelines of the International Union for the Protection of New Varieties of Plants (UPOV) guidelines for kohlrabi, broccoli and broccoli raab. The recorded descriptors included plant habit, leaf colour, leaf blistering, leaf blade distal end shape, and leaf margin. Plant fresh weight (g plant⁻¹) was determined at commercial maturity by individually weighing at least three plants per replicate.

Table 1. Details of the Brassica spp. germplasm accessions evaluated in this study.

Landraces	Genebank accession ID	Species	Local name	Geographic origin	Donor
Cavolo riccio	UBGBOA001	Brassica oleracea var. acephala DC.	Colle rizze	Adelfia (BA)	Francesco Angelo De Vito
Cima di rapa di febbraio	UBGBRS118	Brassica rapa subsp. sylvestris (Lam.) Janch.	Cima di rapa di febbraio	Putignano (BA)	Vito Pesce

Table 1. Details of the Brassica spp. germplasm accessions evaluated in this study.

Landraces	Genebank accession ID	Species	Local name	Geographic origin	Donor
Cavolo riccio	UBGBOA001	Brassica oleracea var. acephala DC.	Colle rizze	Adelfia (BA)	Francesco Angelo De Vito
Cima di rapa di febbraio	UBGBRS118	Brassica rapa subsp. sylvestris (Lam.) Janch.	Cima di rapa di febbraio	Putignano (BA)	Vito Pesce

2.3. Sample Preparation and Cooking Treatments

After harvesting, and following agronomic analyses, both crops were transported to the laboratories of University of Bari in a refrigerated bag and were processed by removing non-edible parts. For nutritional analyses, samples were collected from each of the three field replicates (blocks). Each replicate consisted of a composite sample obtained from three individual plants, with equal amounts of tissue collected from each plant (leaves only for ‘Cavolo riccio’; leaves and inflorescences for ‘Cima di rapa di febbraio’, both harvested at commercial maturity). Samples were then divided into 150 g aliquots for each replication and preparation method (raw, boiled and steamed). Aliquots intended for raw analysis were put immediately in a freezer at -18 ± 1 °C; whereas cooking (boiled and steamed) was carried out on fresh aliquots according to times determined in a preliminary experiment. To obtain samples for chemical analysis, each replicate of both raw and cooked vegetables was freeze-dried.

2.3.1. Boiling

The two Brassica samples were boiled in distilled water (99 ± 1 °C) using a leaf weight/water volume ratio of 1:7 (w/v). The samples were immersed upon boiling and cooked for 5 minutes. After boiling, the samples were drained and placed into freezer bags, ensuring the material was spread thinly to maximize the surface area for rapid cooling. Following a two-hour pre-cooling period in a refrigerator (temperature to 3 ± 1 ◦C), the samples were transferred to a freezer and maintained at -18 ± 1 °C.

2.3.2. Steaming

Steaming was performed using a steam cooker (VC 101 630, Tefal, Milan, Italy). The samples were placed on the steamer tray and cooked with steam at atmospheric pressure (99 ± 1 °C) for 10 minutes. Following steaming, the samples were drained and transferred into freezer bags, ensuring the material was spread in a thin layer to maximize the surface area for rapid cooling. After a two-hour pre-cooling period in a refrigerator (3 ± 1 °C), the samples were stored in a freezer at -18 ± 1 °C.

2.4. Physical Analysis

To determine weight change, samples were weighed individually before each cooking treatment and reweighed after being drained using a colander. Results were expressed as the percentage weight gain relative to the initial raw weight.

For dry matter measurement, plant samples (both raw and cooked) were kept in a forced-air oven at 65 °C until a constant weight was achieved. Results are expressed as the percentage of residual dry matter after water removal, calculated relative to the total weight of the fresh sample.

2.5. Chemical Analyses

2.5.1. Glucosinolate

GLSs were extracted twice from 0.5 g of freeze-dried plant tissue using 5 mL of ethanol at 75 °C for 15 min under stirring, followed by centrifugation (5000 rpm, 15 min, 4 °C). Supernatants were collected for analysis. The extracted GLSs were desulfurized and purified; the resulting solution was analyzed by HPLC as reported in Conversa et al. (2024).

2.5.2. Phytosterol

Total phytosterols were determined according to Bonasia et al. (2026). Samples (100 mg of freeze-dried tissue) underwent acid hydrolysis (0.5 mol L^-1 sulfuric acid solution in 95% ethanol) and alkali saponification (2.0 mol L^-1 sodium hydroxide solution in 95% ethanol), both performed at 55 °C for 1 h, with the addition of tert-butylhydroquinone (3 mg) to prevent thermal decomposition. Double extraction with hexane was performed; pooled extracts were dried under vacuum using a rotary evaporator (Strike 300, Steroglass, San Martino in Campania, AV, Italy), reconstituted in 1.0 mL of ethanol, and purified on a 500 mg C18 SPE cartridge (Strata C18-E, Phenomenex, Castel Maggiore, BO, Italy). Quantification was performed spectrophotometrically (Evolution 201 UV-Vis spectrophotometer, Thermo Scientific, Waltham, MA, USA) at 520 nm [49] using a stigmasterol calibration curve, with results expressed as mg stigmasterol equivalents per unit fresh weight.

2.5.3. Total Phenols

For the analysis of total polyphenols and flavonoids, freeze-dried samples (30 mg) were extracted twice using 1 mL of a water/methanol solution (20:80 v/v) in an ultrasonic bath (DU-32 Digital, Argo Lab, Carpi, MO, Italy) for 15 min at room temperature. After each extraction step, the mixture was centrifuged at 14,000 rpm for 15 min at 4 °C (ThermoFisher Scientific, Waltham, MA, USA). The pooled supernatants were collected and stored at -20 °C until analysis, which was carried out within 24 h (Bonasia et al., 2026).

Total phenols were quantified spectrophotometrically at 750 nm using the Folin-Ciocalteu assay [50]. Extracts (100 µL) were reacted with Folin-Ciocalteu reagent (0.5 mL) and 20% Na₂CO₃ (1.0 mL). Absorbance was read after 45 min at 30 °C, and concentrations were expressed as mg gallic acid equivalents per unit fresh weight based on a standard curve.

Total flavonoids were quantified using the AlCl3 colorimetric method [51]. Extracts or quercetin standards (200 µL, 1-100 mg L-1) were reacted with 10% AlCl₃ (40 µL), 1 M potassium acetate (40 µL), and water (1,120 µL). After 30 min at room temperature, absorbance was read at 415 nm. Results were expressed as mg quercetin equivalents per unit fresh weight.

2.5.4. Chlorophylls, Carotenoids and Anthocyanins

Photosynthetic pigments and total anthocyanins were quantified spectrophotometrically following Conversa et al. (2024). Lyophilized tissue (0.05 g) was extracted three times with 80% ethanol and 0.1% HCl via ultrasonication (30 min) and centrifugation (4,000 g, 15 min, 4 °C). Filtered extracts (0.22 µm) were read at 664, 649, and 470 nm for chlorophylls and carotenoids [52]. Anthocyanins were measured at 525 and 650 nm, corrected for chlorophyll [53], and calculated using a molar extinction coefficient of 26,900 L mol-1 cm-1 as cyanidin-3-glucoside equivalents.

2.6. Statistical Analysis

Agronomic data were reported as mean ± standard error (SE) and were not subjected to inferential statistical analysis because their purpose was solely descriptive. Post-harvest quality data were analyzed using the R statistical software (version 4.5.1; R Core Team, 2025) within the RStudio IDE (Posit Team, 2023), utilizing the car package for homoscedasticity testing and the agricolae package for multiple comparison tests and standard error calculations. Before analysis, the homogeneity of variances was verified using Levene’s test. These data were subjected to Analysis of Variance (ANOVA) based on a randomized complete block design with a 2 x 3 factorial scheme (two genotypes x three consumption methods). The “replicate” (rip) factor was included in the model as a blocking variable. Mean differences were evaluated using Tukey’s HSD test with a significance level of P≤0.05. To describe data variability in the post-harvest quality results, the Pooled Standard Error (Pooled SE) was calculated, derived from the square root of the Mean Square Error (MSE) of the ANOVA model divided by the square root of the average sample size (n=3):

$$PooledSE=\sqrt{{\displaystyle \frac{MSE}{n}}}$$

3. Results

3.1. Agronomic and Morphological Characteristics

Details of agronomic characteristics and morphological traits recorded on the two Brassicaceae evaluated in this study are reported in Table 2. The evaluated germplasm showed distinct agronomic and morphological profiles. Regarding plant fresh weight, accession ‘Cima di rapa di febbraio’ showed a mean fresh weight of 159.6 ± 14.0 g per plant^-1, while ‘Cavolo riccio’ 174.5 ± 6.8 g plant^-1. Differences were also observed for morphological traits (Table 2); for instance, while both accessions exhibited an ascending habitus and medium green leaf color, they shown different leaf blade splitting (lyrate for ‘Cavolo riccio’ and lacerate for ‘Cima di rapa di febbraio’ and leaf distal end shape (acuminate and acute, respectively) (Table 2).

3.2. Dry Matter and Weight Gain of Cooked Samples

The genotype × cooking method interaction was not statistically significant for any of the parameters in Table 3. In contrast, weight gain showed a significant genotype × cooking method interaction (Figure 1).

Concerning weight gain, the interaction revealed that the effect was genotype-dependent during boiling, with ‘Cavolo riccio’ showing a 21.8% higher weight gain than ‘Cima di rapa di febbraio’. In contrast, the two genotypes did not differ significantly from each other when steamed (7.2% and 6.4%, respectively) (Figure 1).

Regarding dry matter concentration, significant differences were observed for both main factors (genotype and cooking method). On average, ‘Cavolo riccio’ showed a dry matter value of about 21% higher than ‘Cima di rapa di febbraio’. As for the treatments, boiling caused a dry matter reduction of about 31%, compared to raw and steamed samples (Table 3 - Effects of different consumption methods of ‘Cima di rapa di febbraio’ and ‘Cavolo riccio’ on dry matter and phytochemical concentration (polyphenols, phytosterols, and flavonoids).

3.3. Polyphenols, Flavonoids and Phytosterols

A similar independent trend for both factors was observed for total polyphenols. ‘Cavolo riccio’ had significantly higher values of this parameter than ‘Cima di rapa di febbraio’ (+17.4%). Regarding cooking effect, boiling reduced the content by 59.2% compared to the average of the other two treatments (Table 3).

Concerning flavonoids, no statistically significant differences were found between the two landraces; however, the cooking method significantly affected their content. Compared to the raw sample, steaming and boiling caused a flavonoid reduction of 27.8% and 41.6%, respectively (Table 3). Finally, regarding phytosterols, no statistically significant differences were found either between the two landraces or among the cooking treatments, with an average value of 6.8 mg 100 g^-1 FW (Table 3).

Table 3. Effects of different consumption methods of ‘Cima di rapa di febbraio’ and ‘Cavolo riccio’ on dry matter and phytochemical concentration (polyphenols, phytosterols, and flavonoids).

	%	mg 100 g-1 FW	mg 100 g-1FW	mg 100 g-1FW
	Dry matter	Polyphenols	Phytosterols	Flavonoids
Landraces
‘Cima di rapa di febbraio’	10.3 b	93.1 b	7.0	61.3
‘Cavolo riccio’	12.4 a	109.3 a	6.6	69.6
Methods
Uncooked	13.0 a	133.0 a	6.0	103.3 a
Boiling	8.8 b	51.5 b	4.3	29.7 c
Steaming	12.3 a	119.2 a	10.2	63.5 b
Significance(1)
Landraces	***	*	ns	ns
Methods	***	***	ns	***
Landraces*Methods	ns	ns	ns	ns

(1) Significance of F: ns, *, **, and ***, indicating non-significant and significant at P≤0.05, 0.01, and 0.001, respectively. Different letters within the same column indicate statistically significant differences among treatments according to Tukey’s HSD post-hoc test (p ≤ 0.05). Significant interaction is reported in the Figure 1

Table 3. Effects of different consumption methods of ‘Cima di rapa di febbraio’ and ‘Cavolo riccio’ on dry matter and phytochemical concentration (polyphenols, phytosterols, and flavonoids).

	%	mg 100 g-1 FW	mg 100 g-1FW	mg 100 g-1FW
	Dry matter	Polyphenols	Phytosterols	Flavonoids
Landraces
‘Cima di rapa di febbraio’	10.3 b	93.1 b	7.0	61.3
‘Cavolo riccio’	12.4 a	109.3 a	6.6	69.6
Methods
Uncooked	13.0 a	133.0 a	6.0	103.3 a
Boiling	8.8 b	51.5 b	4.3	29.7 c
Steaming	12.3 a	119.2 a	10.2	63.5 b
Significance(1)
Landraces	***	*	ns	ns
Methods	***	***	ns	***
Landraces*Methods	ns	ns	ns	ns

(1) Significance of F: ns, *, **, and ***, indicating non-significant and significant at P≤0.05, 0.01, and 0.001, respectively. Different letters within the same column indicate statistically significant differences among treatments according to Tukey’s HSD post-hoc test (p ≤ 0.05). Significant interaction is reported in the Figure 1

Figure 1. Weight gain (g) in ‘Cima di rapa di febbraio’ (CMR) and ‘Cavolo riccio’ (CVR) with different cooking methods: boiled (B) and steamed (S). Mean values were obtained by averaging the weight of the samples weighed after being drained using a colander. Data were analyzed using a randomized complete block design (RCBD) ANOVA model (n = 3). Different letters indicate statistically significant differences among treatments according to Tukey’s HSD post-hoc test (p ≤ 0.05).

Figure 1. Weight gain (g) in ‘Cima di rapa di febbraio’ (CMR) and ‘Cavolo riccio’ (CVR) with different cooking methods: boiled (B) and steamed (S). Mean values were obtained by averaging the weight of the samples weighed after being drained using a colander. Data were analyzed using a randomized complete block design (RCBD) ANOVA model (n = 3). Different letters indicate statistically significant differences among treatments according to Tukey’s HSD post-hoc test (p ≤ 0.05).

3.4. Carotenoids, Anthocyanins, Chlorophylls

Regarding the pigments reported in Table 4, the genotype×cooking method interaction was not statistically significant for any of the analyzed parameters; therefore, the effects of the landrace and the cooking treatment were evaluated independently.

Concerning chlorophylls, significant differences between landraces were found for total chlorophyll, with ‘Cima di rapa di febbraio’ showing significantly lower values (-36.3%) than ‘Cavolo riccio’. Regarding the impact of cooking, different behaviors were observed for the chlorophyll fractions. For chlorophyll b, cooked samples showed significantly lower levels (-52.7%) compared to the raw material. For chlorophyll a and total chlorophyll, no significant differences were found between raw and steamed samples. The marked loss observed after boiling (-40.8% for chlorophyll a, -46.9% for total chlorophyll) was the major reduction recorded among cooking (Table 4).

Carotenoid concentrations, significant differences were observed for both main factors.

‘Cima di rapa di febbraio’ showed significantly lower concentrations of these pigments (-47.8%) compared to ‘Cavolo riccio’. As for the cooking effect, cooked samples showed significantly lower levels of carotenoids (-37.2%) compared to the raw material (Table 4).

A similar trend was observed for anthocyanins, which were significantly affected by both genotype and cooking method. ‘Cima di rapa di febbraio’ exhibited a marked reduction (-71.7%) compared to ‘Cavolo riccio’. Regarding the thermal treatments, cooking methods exhibited significant reductions compared to the raw material, with an overall decrease of -74.8% in cooked samples (Table 4).

3.5. Glucosinolates

Regarding the glucosinolates reported in Table 5, a marked and statistically significant genotype×cooking method interaction was observed for individual glucosinolates compounds (epiprogoitrin, gluconapoleiferin, and glucobrassicin). Glucosinolates with non-detectable (ND) levels were imputed using the limit of detection (LOD) value.

The two landraces contain different individual glucosinolates, determining distinct qualitative profiles (Table 5). In ‘Cima di rapa di febbraio’, the identified compounds – listed according to Table 5 – were epiprogoitrin, gluconapoleiferin, glucobrassicanapin, and glucobrassicin. Conversely, in ‘Cavolo riccio’, only sinigrin and glucobrassicin were detected above the quantification limit (Table 5). Glucobrassicin showed a significant interaction between genotype and cooking method. In ‘Cima di rapa di febbraio’, glucobrassicin content was significantly reduced by boiling, decreasing from 1.1 μmol kg^-1 FW in raw samples to values below the quantification limit (0.52 μmol), whereas no significant differences were observed between raw and steamed samples (Figure 2). In ‘Cavolo riccio’, glucobrassicin content was significantly lower in boiled samples (9.4 μmol kg^-1 FW) than in raw samples (23.0 μmol kg^-1 FW), corresponding to a reduction of 59.1%. Steamed samples (15.6 μmol kg^-1 FW) showed intermediate values and did not differ significantly from either treatment (Figure 2 and Table 5).

For epiprogoitrin and gluconapoleiferin, cooking significantly affected their concentration only in ‘Cima di rapa di febbraio’, as both compounds were present at trace levels in ‘Cavolo riccio’. In ‘Cima di rapa di febbraio’, boiling caused a marked decrease in epiprogoitrin and gluconapoleiferin contents (-76%, both), compared with the raw samples, whereas steaming maintained concentrations comparable to those of the raw control (Table 5, Figure 3 and Figure 4). Furthermore, the two landraces differed markedly in the occurrence of specific glucosinolates. Sinigrin was detected only in ‘Cavolo riccio’, whereas glucobrassicanapin was found almost exclusively in ‘Cima di rapa di febbraio’, highlighting the distinct glucosinolate profiles of the two genotypes (Table 5).

Total glucosinolate content was significantly affected by both landrace and cooking method (Table 5). ‘Cima di rapa di febbraio’ showed a markedly higher total glucosinolate content (119.7 μmol kg⁻¹ FW) than ‘Cavolo riccio’ (19.2 μmol kg⁻¹ FW). Boiling significantly reduced total glucosinolate content (25.9 μmol kg⁻¹ FW) compared with raw samples (95.6 μmol kg⁻¹ FW), while steamed samples showed intermediate values (86.9 μmol kg⁻¹ FW) and did not differ significantly from either uncooked or boiled samples (Table 5).

4. Discussion

4.1. Landraces

The obtained results highlight a marked nutritional and nutraceutical complementarity between the two local Brassica landraces. ‘Cavolo riccio’ is characterized by a higher dry matter content (12.4%) (+21.1% compared to ‘Cima di rapa di febbraio’), in line with the findings of Gonnella et al. (2019), in which ‘Cavolo riccio’ displayed 31% higher dry matter than ‘Cima di rapa’. Similarly, ‘Cavolo riccio’ showed an elevated concentration of total polyphenols (109.3 mg 100 g^-1FW), (+17.4% compared to ‘Cima di rapa di febbraio’), in line with Kapusta-Duch et al. (2016), who ranked ‘Cavolo riccio’ as the Brassica vegetable with the highest polyphenol content. This high accumulation of phenolic compounds is strongly supported by literature evaluating the diversity of Apulian kale [26] and kale baby leaves [9], which documents an exceptionally high phenolic and flavonoid profile in this crop compared to other leafy vegetables. Conversely, broccoli raab exhibit a lower, yet functionally significant, phenolic content, as previously observed in local ecotypes by Conversa et al. (2020). The two landraces also differ markedly in their glucosinolate profile. The metabolite profile of ‘Cima di rapa di febbraio’, characterized by high levels of glucobrassicanapin (69.6 μmol kg-1 FW) is in agreement with Mazzeo et al. (2019) who identified this compound as a distinctive mark of Southern Italian ‘Cima di rapa’. The glucosinolate composition of ‘Cavolo riccio’, dominated by sinigrin and glucobrassicin, is consistent with findings by Korus et al. (2014), who reported these as the predominant aliphatic and indole glucosinolates, respectively, in kale leaves.

The absolute values of glucosinolate detected in the analyzed samples were markedly lower than corresponding literature values. For ‘Cima di rapa di febbraio’, the total glucosinolates content was 119.7 μmol kg^-1 FW, a value approximately 62% lower than that estimated for a commercial broccoli raab(317.4 μmol kg^-1 FW) by Conversa et al. (2020). A similar discrepancy is evident when comparing the present results with those reported by Conversa et al. (2016), who found markedly higher glucosinolate concentrations in Brassica rapa subsp. sylvestris (Lam.) Janch. var. esculenta Hort. cultivated in conventional and organic growing systems (1149.0 μmol kg⁻¹ FW and 886.8 μmol kg⁻¹ FW respectively). This suggests that genotype, cultivation environment, and agronomic management may substantially influence glucosinolate accumulation. The glucosinolates content in ‘Cavolo riccio’ (19.2 μmol kg^-1 FW) was significantly lower than the average value of approximately 1.32 μmol g^-1 FW (equivalent to 1320 μmol kg^-1 FW) reported for local Sicilian kale landraces (Brassica oleracea var. acephala DC.) by Branca et al. (2002).

Although the absolute glucosinolate concentrations were lower than those reported in the literature, the relative distribution of individual glucosinolates remained characteristic of each crop. This indicates that the observed differences mainly concern the overall accumulation of glucosinolates rather than an alteration of the genotype-specific metabolic profile. The predominance of glucobrassicanapin in ‘Cima di rapa di febbraio’ confirms the identity of Southern Italian germplasm, where this compound is considered a distinctive marker [34]. Likewise, the presence of sinigrin and glucobrassicin as the main glucosinolates in ‘Cavolo riccio’ faithfully reflects the typical metabolic profile of kale documented in the literature [4,11]. The differences observed relative to Conversa et al. (2020) cannot be attributed to taxonomic factors, as the comparison was conducted using the same subspecies (Brassica rapa subsp. sylvestris (Lam.) Janch.); in that study, two landraces of broccoli raab were compared with the commercial variety.

Similarly, for ‘Cavolo riccio’, the reference value reported by Branca et al. (2002) refers to Sicilian landraces of (Brassica oleracea var. acephala DC.), (‘Cavolo vecchio’), belonging to the same botanical group as the genotype examined in the present study. Therefore, the large quantitative discrepancy observed between studies is more likely attributable to differences in genotype, environmental conditions, crop management, harvest stage, and analytical procedures than to taxonomic classification. Secondly, the plant part examined plays a crucial role, as glucosinolate levels vary significantly among young leaves, mature leaves, and inflorescences [42,56]. This is particularly relevant to the specific plant organs evaluated in the present study (as detailed in Materials and Methods), which comprised leaves for ‘Cavolo riccio’ [57] and both leaves and inflorescences for ‘Cima di rapa’ [36].

By analyzing the edible portions traditionally consumed for each landrace, this study provides a comparison relevant to consumers. Because these portions comprise different plant organs, however, the observed differences should be interpreted as differences between the foods as consumed and cannot be attributed exclusively to genotype.

Overall, the two landraces exhibited complementary phytochemical profiles. ‘Cavolo riccio’ was characterized by higher dry matter and polyphenol concentrations, confirming its value as a rich source of antioxidant compounds. In addition, both contained substantial amounts of phytosterols (approximately 7 mg 100 g⁻¹ FW), which are lipophilic bioactive compounds that contribute to the nutritional value of vegetables and are particularly recognized for their cholesterol-lowering effects [31]. The occurrence of phytosterols in both kale and broccoli raab is consistent with the findings of Castellaneta et al. (2023), who identified β-sitosterol and campesterol as the predominant sterols in mature tissues of both species. These values are approximately threefold higher than those reported for Apulian kale baby leaves by Bonasia et al. (2026) (~2 mg 100 g^-1 FW) and comparable to those detected in local halophytic species such as Salicornia europaea Auct. and Arthrocaulon macrostachyum (Moric.) Piirainen & G. Kadereit (~7 mg 100 g^-1 FW) (Duri et al., 2025). Conversely, ‘Cima di rapa di febbraio’ exhibited a markedly higher glucosinolate content, particularly glucobrassicanapin, suggesting a greater potential contribution of glucosinolate-derived health-promoting metabolites. Therefore, the combined valorization of these local genotypes may broaden the dietary intake of complementary classes of bioactive compounds, including polyphenols, phytosterols, and glucosinolates.

This metabolic diversity underscores the scientific and agronomic importance of recovering and valorizing Apulian vegetable germplasm [26,32]. Therefore, promoting the diversified consumption of these local cultivars bridges in situ biodiversity conservation with nutritional disease prevention and the safeguarding of the region’s food and wine identity [1,59]

4.2. Cooking Effects

4.2.1. Thermal Stability and Matrix Disruption

Cooking-induced losses of bioactive compounds in this study can be attributed to two main mechanisms: thermal degradation and disruption of the membrane matrix, which represent the primary universal effect common to both cooking methods and are therefore discussed first [3,60]; and leaching, a mechanism specific to boiling due to direct water contact, discussed in Section 4.2.2. [11,56]. It is important to note that thermal degradation is not restricted to lipophilic molecules, but acts as the main driver of loss for all highly heat-sensitive compounds - whether lipophilic (such as chlorophyll b) or hydrophilic (such as flavonoids) - for which the absence of direct water contact during steaming provides no significant protective advantage [40,42,43,45]. The losses of lipophilic compounds were governed primarily by thermal degradation and disruption of the membrane matrix in which they are embedded, rather than by leaching into the cooking water.

In the present study, the absence of significant differences in total phytosterol content among cooking methods indicates a remarkable stability of these compounds in our genotypes. This behavior is consistent with their high intrinsic thermal stability and their role as key structural components of cell membranes. Although high temperatures during boiling and steaming disrupt the lipid bilayer matrix, phytosterols likely remain retained within the tissue due to their structural robustness, lipophilic nature, and complete insolubility in water, which prevents them from being lost via leaching into the cooking medium [60]. The reduction in total carotenoids (-37.2% on average across cooked samples) is likely related to the thermal susceptibility of specific carotenoid fractions, particularly oxygenated xanthophylls, which are abundant in leafy Brassicaceae such as kale and broccoli raab [8]. These vegetables contain considerable amounts of lutein, neoxanthin, violaxanthin, and zeaxanthin in addition to β-carotene [31,42]. While β-carotene is generally considered relatively heat-stable, epoxy-xanthophylls such as neoxanthin and violaxanthin are more prone to heat-induced degradation, oxidation, and cis/trans isomerization during hydrothermal processing [43,56]. Therefore, the reduction observed in the present study is probably driven by the degradation and transformation of xanthophylls [43]. Since both methods expose tissues to high temperatures, the damage to lipophilic pigments proceeds to a similar extent, confirming that steaming offers limited additional protection for this class of compounds compared to boiling, as reported by Murador et al. (2016), who observed massive losses in both boiled (77%) and steamed (72%) kale.

The stability of chlorophyll a and total chlorophyll, during steaming compared to the raw product is consistent with the literature identifying steaming as the most conservative method, capable of minimizing both thermal degradation and leaching of pigments [38,45,56]. In line with mechanisms hypothesized in the literature but not directly tested in this study, indirect hydrothermal heating via saturated steam is expected to limit cell lysis and leaching [3,56], instantly inactivating degradative enzymes in an oxygen-depleted environment [42] and gently softening cell walls to enhance phytochemical bioaccessibility without destructive thermal stress [40,60].

On the contrary, the marked loss observed after boiling (-40.8% for chlorophyll a, -46.9% for total chlorophyll) is likely due to the conversion of chlorophyll to pheophytin, triggered by acid release from the disrupted vacuoles under prolonged heat exposure [42,45]. Since chlorophylls are hydrophobic and associated with protein complexes in plastid membranes, their loss, as for carotenoids, reflects chemical transformation rather than diffusive leaching [42,45].

However, a different pattern was observed for chlorophyll b, which, unlike chlorophyll a, showed significant reductions under both boiling and steaming. This trend is consistent with previous studies suggesting that chlorophyll b is markedly more sensitive to thermal stress and pheophytinization than chlorophyll a during boiling [42] and under certain steaming conditions [39]. The significant loss of chlorophyll b even during steaming may indicate that its chemical structure may be more susceptible compared to chlorophyll a as discussed by [42] and found by Premavalli et al. (2001) in a different cooked leaf vegetables. Conversely, Armesto et al. (2016) have observed a higher relative retention of chlorophyll b compared to chlorophyll a during steaming, indicating that matrix characteristics and cooking parameters can significantly affect pigment stability.

4.2.2. The Leaching Mechanism

Substantial reduction in dry matter observed after boiling is attributable to the high water solubility of solute compounds, which leach into the cooking water when cell membranes are disrupted by heat [39,44]. Genotype-dependent water absorption was observed during boiling, which may be hypothetically related to differences in cell wall composition and tissue density between the two landraces, as the more fibrous structure of ‘Cavolo riccio’ may facilitate greater imbibition when cell integrity is compromised by heat [3,45]. The fact that steaming did not differ significantly from raw confirms that the absence of direct water contact is the primary protective factor in this cooking method [4]. This leaching phenomenon represents a common mechanism underlying the losses observed across several classes of hydrophilic compounds in the present study, including polyphenols, flavonoids, anthocyanins and glucosinolates [11,30,40]. A possible explanation is that, during boiling, direct immersion in water establishes a concentration gradient between the intracellular environment and the surrounding cooking medium [62]. Furthermore, as highlighted by Palermo et al. (2014), this migration is highly compound-specific and governed by molecular structure; for instance, while highly water-soluble flavonoid glucuronides easily leach into the cooking water, flavonoid glycosides and acylated derivatives are less water-soluble and tend to be preferentially retained within the vegetable tissue. Heat-induced damage to cellular structures may increase membrane permeability [3], facilitating the diffusion of soluble compounds into the cooking water, a process that could be further enhanced by the convective mixing and mechanical agitation typical of boiling conditions [62]. Steaming, by contrast, exposes the tissue to water only in vapor form: no liquid-solid interface is established, and consequently no bulk diffusive flux of solutes into an external aqueous phase can occur [4,45], even though some thermal degradation of heat-labile compounds (such as chlorophyll b, and flavonoids). Consistent with this mechanism the 59.2% lower polyphenol content observed in boiled samples, relative to the average of the other treatments, is in agreement with Renna et al. (2013). Glucosinolates followed a similar leaching-driven pattern, consistent with the general mechanism described above [4,11].

The markedly lower total glucosinolate content observed after boiling is in agreement with previous reports in Brassicaceae vegetables, where reductions of up to 85% have been attributed to the combined effects of leaching into the cooking water, thermal degradation, and enzymatic breakdown ([3,4].

In the present study, steamed samples maintained higher glucosinolate concentrations than boiled samples, confirming that the avoidance of direct contact with water contributes substantially to glucosinolate retention. Overall, steaming appeared to be the most suitable cooking method for preserving glucosinolates and, consequently, the nutraceutical value associated with these compounds [4,62].

4.2.3. Thermal Degradation and Condensation-Driven Losses

A partial exception to the protective effect of steaming can occur through surface condensation of vapor into a thin aqueous film, a compound-specific leaching mechanism detailed below for flavonoids and anthocyanins [2,40].

The progressive reduction of flavonoids across all three treatments, including steaming, suggests that these compounds are also sensitive to heat degradation independently of leaching, in addition to being affected by water contact. This behavior distinguishes flavonoids from both total polyphenols and glucosinolates, whose losses under steaming were comparatively limited [4,62], challenging the general assumption that steaming is invariably superior to boiling for nutrient retention. Although the mechanisms underlying this response were not directly investigated in the present study, several hypotheses based on previous literature may help explain this divergence.

First, most flavonoids in Brassicaceae tissues occur as glycosides, in which the flavonoid aglycone is bound to one or more sugar moieties via glycosidic bonds [2,20]. It has been proposed that these bonds are susceptible to acid- and heat-catalyzed hydrolysis even under the moist-heat conditions generated during steaming, potentially yielding aglycones that are themselves more prone to oxidative degradation [63].

Second, steaming is not entirely devoid of liquid water contact at the tissue surface. Condensation of steam on the vegetable surface may generate a thin aqueous film which could permit limited diffusion of highly water-soluble flavonoid glycosides, even in the absence of the extensive leaching typically associated with boiling [40].

Third, flavonoid losses may also be influenced by residual enzymatic activity, particularly polyphenol oxidase [2] and peroxidase [40], during the early stages of heating before complete enzyme inactivation occurs [44]. Differences in heat-transfer dynamics between boiling and steaming have been suggested as a factor that could influence the duration of exposure to temperatures permitting enzymatic oxidation [62].

Taken together, these hypotheses suggest that flavonoid retention may be influenced by both leaching-related and leaching-independent thermal processes, unlike total polyphenols and glucosinolates, whose retention appeared more strictly tied to water contact. Nevertheless, these proposed mechanisms should be interpreted with caution, as neither flavonoid glycoside profiles nor enzyme activities were determined in the present study. Further targeted analyses would be required to establish the relative contribution of these mechanisms to flavonoid losses during cooking. Anthocyanins were also strongly affected by leaching, showing the most drastic reduction among the hydrophilic compounds examined (-74.8% as a mean of both methods), consistent with their extreme water solubility and thermal instability, which causes near-total dispersal into the cooking water [43,64]. As vacuolar hydrophilic pigments [20], their release follows the same heat-induced membrane disruption described above. Boiling and steaming showed no significant differences in anthocyanin retention, indicating that steaming did not provide the same protective effect observed for glucosinolates, a pattern attributable – as for flavonoids – to the same steam-condensation leaching at the tissue surface [40,63]. Individual glucosinolates also differ in thermal stability, with heat resistance decreasing, a process facilitated by the same heat-induced membrane rupture discussed above [65].

4.3. Practical Implications for the Consumer

Taken together, the findings discussed above allow some practical, target-metabolite-oriented cooking recommendations to be drawn. For consumers or food operators primarily interested in preserving glucosinolates, steaming may be preferred over boiling, since boiling was associated with the lowest glucosinolate concentrations, whereas steamed samples generally showed intermediate or higher retention of total and individual glucosinolates [4,62].

Steaming is similarly preferable when the goal is to retain chlorophylls (particularly chlorophyll a) and dry matter/soluble solids, since it avoids the direct water contact responsible for the marked losses observed after boiling [38,39,44,45]. Conversely, for compounds such as flavonoids, anthocyanins, certain carotenoids and chlorophyll b, steaming did not provide a comparable protective advantage over boiling, since their losses are driven, at least in part, by leaching-independent thermal degradation and/or by localized surface leaching through steam condensation; consumers or processors specifically targeting these compounds should therefore not assume that steaming is systematically superior to boiling. Phytosterol content, being unaffected by cooking method, does not represent a determining factor when choosing between boiling and steaming. Phytosterols are recognized for their anti-tumorigenic effects and their capacity to lower blood cholesterol, along with a structural role in maintaining cell membrane fluidity and permeability, particularly relevant under oxidative stress conditions [30].

Beyond the choice of cooking method, the genotype-level complementarity described in Section 4.1. also carries practical relevance for dietary diversification. The nutritional complementarity between the two landraces emerges from the predominance of antioxidant pigments and phenolic compounds in ‘Cavolo riccio’ and the substantially higher glucosinolate concentration in ‘Cima di rapa di febbraio’, suggesting that their combined consumption may provide a broader spectrum of health-promoting phytochemicals than either landrace alone.

Since ‘Cavolo riccio’ is comparatively richer in polyphenols, carotenoids, and chlorophylls beyond their role as a visual quality indicator, chlorophylls are also associated with anti-mutagenic, anti-cancer, and anti-inflammatory properties, with early-life dietary intake reportedly contributing to obesity prevention and a reduced risk of type 1 diabetes [30], while ‘Cima di rapa di febbraio’ provides a higher glucosinolate content, alternating the two landraces within the diet, rather than favoring one over the other, may allow consumers to benefit from a broader spectrum of bioactive compounds than either Brassica alone would provide. This complementarity strengthens the case for promoting both landraces jointly, rather than positioning them as interchangeable alternatives.

5. Conclusions

This study demonstrated that the two traditional Apulian Brassica landraces possess distinct and complementary phytochemical profiles and exhibit different responses to domestic cooking, highlighting their potential as valuable sources of health-promoting compounds. These findings support their nutritional complementarity and their valorization within a diversified Mediterranean diet. Cooking method significantly influenced the preservation of phytochemicals, although the extent of these effects depended on the compound considered. Overall, steaming generally better preserved several bioactive compounds than boiling, supporting its role as a favorable household practice for maintaining nutritional quality. Future research should confirm these findings under different environmental conditions and integrate sensory evaluation to support the development of nutritionally optimized and consumer-acceptable preparation strategies.