Table 1 shows the mean value and the standard deviation across thirty cultivars for all the traits in each of the four environments. On average, the environments presented variations in grain weight, whereas the other physical characteristics did not differ significantly. The hardness index values indicated that the durum wheat samples exhibited the natural variability in very hard kernel texture for each of the considered environments. As expected, the protein concentration was shown to be highly sensitive to the growing site. In fact, in the durum wheat varieties grown in Jesi the protein content measured was greater than that grown in Foggia in both the years; these differences likely reflected the major availability of nitrogen fertilizer during grain filling and different soil fertility. The average TOAX contents of the four environments ranged from 4.5 to 4.8% d.m., Jesi09 having lower WUAX content than the other environments. As indicated in the experimental section the WEAX amount was calculated by difference (TOAX–WUAX); on average, among the four environments the range in WEAX was very limited.

There were significant correlations between protein content and the physical characteristics of the grain (r = 0.317 p ≤ 0.01; r = 0.379 p ≤ 0.001; r = 0.225 p ≤ 0.05, with hardness index, grain weight and diameter, respectively, n = 76). The reported relationships were found to be significant both in the single environments and in the whole trial. No correlations between AX fractions and hardness index were found, in agreement with Bettge and Morris [30] who suggested a minimal role of pentosans in modifying hard wheat grain hardness and with the results of Li et al. [22] in hard and spring wheat.

A group of 19 durum wheat cultivars common to the four environments was considered for ANOVA of AX. The results presented in Table 2 evidenced highly significant genotypic and environmental effects and a significant genotype x environment interaction for all the AX components. However, as previously evidenced by Finnie et al. [23] and by Gebruers et al. [31] in wheat whole meal, the contribution of this interaction to the total variability was considerably lower than that of genotype or environment. Variety had the greatest influence on the WUAX content in durum wheat samples and was important for TOAX. The results of this study indicated that great variations could be ascribed to different environmental conditions, environment being the dominant factor contributing to total variations of AX fraction content. These results are generally consistent with the results of Lempereur et al. [24] who evidenced the influence both of cultivar and environment for TOAX and WEAX fractions in durum wheat grain. As illustrated in Figure 1, different climatic conditions were registered in the four trials, especially in relation to max temperatures in the last phase of grain filling and total rainfall (see also experimental section). On average, the highest TOAX values were found in Jesi10 (4.9 ± 0.39% d.m.) and Foggia09 (4.6 ± 0.22% d.m.), two experiments with the highest amounts of rain. In general, the lowest values of WUAX were evidenced in Jesi09 (3.8 ± 0.23% d.m.). These results were in agreement with the observations of Li et al. [22], who detected the highest level of TOAX in the environments where higher amounts of rainfall occurred. The influence of water availability on AX accumulation was discussed by Dornez et al. [32], who in wheat grain evidenced significantly higher amount of WEAX during the rainy years. Other studies on common wheat have also reported environmental effects on the amount of AX fractions [31–34], the impact of genotype being more evident on structural AX changes [14]. Gebruers et al. [31] observed different variation in relation to environment climatic characteristics and concluded that the total variation in the AX levels was determined by the interaction with other factors such as agronomical input and soil type.

Varietal comparison of AX fractions based on Duncan’s test are reported in Table 3. Significant variation existed among the 19 durum wheat varieties for AX fractions. The mean values of genotypes across environments ranged from 4.3 to 5.2% d.m. and from 3.7 to 4.6% d.m. for TOAX and WUAX, respectively. The TOAX range was comparable to the results obtained by Lempereur et al. [24] with a different analysis method.

The results showed that, with regard to TOAX and WUAX contents, some cultivars were significantly superior to the others: Trionfo, followed by Alemanno, presented the highest means for TOAX and WUAX fractions, having consistently the highest TOAX content in three of the four environments analyzed (data not shown). Among the considered durum wheat varieties the WEAX range varied from 0.5 to 0.9% d.m., Creso exhibiting the highest content. Cultivars Tirex and Saragolla were the poorest for TOAX and WUAX. The WUAX appeared to vary within a larger range for some varieties in the various environments; Trionfo, Tirex, Simeto, Ciccio, Arnacoris and Imhopet presented particularly higher variability in comparison with the TOAX. The ratio WEAX/WUAX, considered a measure of extractability [24], varied from 10.9 (Arnacoris) to 23.4% (Creso) the mean value being 16.3%. As found by Lempereur et al. [24], the extractability of AX in this study also appeared to be independent of the TOAX content (data not reported).

The ratio between arabinose (A) and xylose (X) obtained with the AX fractions was also calculated, the measurement of this ratio partially characterizing the structure of the AX [14]. In addition, Martinant et al. [35] found a strong negative correlation between A/X ratio and relative viscosity measurements of the flour water extracts and, following the suggestions of Ordaz-Ortiz and Saulnier [20], variations in the A/X ratio could likely indicate different degrees of arabinose substitution in the different AX fractions. Considering the mean values across environments of the 19 durum wheat varieties, WUAX had A/X ratio values ranged from 0.71 to 0.82 (mean value: 0.78 ± 0.027; data not shown). Similar values were obtained by Barron et al. [36] in whole grains of two soft wheat cultivars. WUAX appeared mainly composed by arabinose and xylose (mean values: 38.1% and 49.0%, respectively). Small amounts of mannose were also found. As compared with TOAX, WUAX had higher level of arabinose and xylose. No significant differences in the AX fraction composition were found among the durum wheat varieties. Under the measure conditions of this study only some indications were drawn for the WEAX fraction composition, arabinose and xylose still remaining the main sugars (26.4 and 34.7%, respectively). The compositional variations likely accounted for the localization of AX fractions in the whole grain, WUAX deriving from the endosperm cell walls but mainly from the outer layer of the grain. Some other changes in monosaccharide composition likely reflected structural differences of cell wall matrix and were associated with the presence of WEAX only in the starchy endosperm (data not reported). As shown by Martinant et al. [35], TOAX content was logically negatively related to xylose percentage (r = −0.696, p ≤ 0.001), but, in contrast with the results of Martinant et al. [35], this ratio was positively related to arabinose percentage (r = 0.823, p ≤ 0.001).

As previously discussed, genotype and environment strongly affected AX concentrations of durum wheat whole grain in this study, the variations due to G × E interaction being lower than that of genotype or environment. The results, however, evidenced that the varieties did not respond similarly to different environmental conditions and, in agreement with the results of Gebruers et al. [31] in soft wheat, some varieties tended to have consistently high or low AX contents. In order to visualize the relationships between the genotypic responses and the different environments explored, principal component analysis (PCA) was used. For each AX fraction the proportion of variance explained by the principal components was calculated (Figure 2). For TOAX and WUAX the first two principal components explained about 83% of total variation, whereas PC1 and PC2 together accounted for about 69% of the original variance of the WEAX variable and PC3 explained more than 17% of the variance. The PCA of varieties and environments for AX fractions were plotted on biplots. For TOAX and WUAX (Figures 3 and 4, respectively) all the environments fell onto the positive side of PC1, and only Jesi10 was located on the negative side for PC2. The varieties which were relatively stable across environments fell close to the PC2 axis; as an example (Figure 3), Saragolla and Tirex were below average for TOAX in all environments, the contents varied from 4.2 to 4.5% d.m and from 4.1 to 4.4% d.m., respectively. The varieties which fell on the right side of PC1 were the ones with a very strong interaction with the environment, presenting their max. potentiality of TOAX accumulation in Jesi10; in fact, for these varieties the range of the observed values was from 5.1 (Duilio) to 5.8% d.m. (Alemanno). The PCA biplot of WUAX fraction (Figure 4) again evidenced that some durum wheat varieties (Iride, Latinur, Tripudio, Dylan) had low absolute values for PC2: this fact generally indicated that the responses of these cultivars to the different environmental conditions were similar, their grain presenting low WUAX contents in most of their environments. As discussed for TOAX, the relative positions in the biplot of the varieties in positive side of PC1 reflected a significant interaction with the Jesi10 environment where they presented the highest WUAX values, in particular Alemanno (5.2% d.m.). Severo and Anco Marzio, which fell on the positive side of PC1 and PC2, generally evidenced WUAX contents slightly above the mean value; finally Trionfo, that had high positive values for PC1 and PC2, exhibited the highest WUAX content in Foggia10, its position on the biplot appearing in positive intersection with this environment. The biplot from PCA for WEAX fraction visually demonstrated that the interactions of variety with the environment were greater than for TOAX and WUAX and the durum wheat varieties likely responded to different environments in different ways (data not shown).

For this study, a group of Italian commercial cultivars of durum wheat (Triticum turgidum L. var durum) was obtained from a set of agronomic trials carried out each year by the Research Unit for Cereal Quality of the Research in Agriculture Council (CRA-QCE). The trials are intended to evaluate yield and quality of durum wheat varieties grown in different sites under normal agronomic conditions. Thirty varieties were grown in two Italian areas adapt to durum wheat cultivation (Jesi, Central-Northern Italy, and Foggia, Southern Italy) in two consecutive years (2008–2009 and 2009–2010). For the comparison among different environments, nineteen out of the thirty varieties were considered. In each experiment, all cultivar were sown in 10 m² plots in randomized blocks with three replications. Nitrogen fertilization was applied for both years according to local practice (about 150 and 90 U/ha, at Jesi and Foggia, respectively), with previous crop being fallow and leguminous in the Foggia and Jesi trials, respectively. The grain samples, blend of the three replicates, were milled to pass a 0.5 mm screen using a laboratory cyclone mill (Cyclotec 1093, FossItalia).

Jesi (altitude 97 m) presents a continental climate with a cold-humid winter and sultry-humid summer. Foggia (altitude 76 m) is characterized by a Mediterranean climate, a very warm summer and mild winter. The mean temperatures and total rainfall, referred at three seasonal periods, are reported in Figure 1. The means of minimal temperatures for the three considered periods were similar, whereas the maximum values of temperature presented differences from April to June between the two trial years, Foggia09 and Jesi09 having higher temperatures (mean values: 24.4 and 25.2 °C, respectively) in comparison with the successive year (mean values: 22.6 and 23.6 °C for Foggia10 and Jesi10, respectively). Anomalous precipitations occurred during the two trial years; moreover, total rainfall showed important differences between years and areas, with 616 and 749 mm at Jesi09 and Foggia09, respectively, and 702 (Jesi10) and 440 mm (Foggia10) in the successive year.

A single kernel classification system (SKCS 4100, Perten Istruments, Sweden) was utilized for the grain hardness (expressed as hardness index), diameter and weight determination.

All determinations were carried out in duplicate on two independent aliquots of each composite sample, blend of three replicates. The protein content was estimated with the nitrogen combustion method by using the Leco-FP 528 nitrogen analyzer [37]. Moisture content was determined at T = 120 °C with a thermo balance (Sartorius MA40-Göttingen, Germany). The arabinoxylans as total (TOAX) and unextractable (WUAX) content were analyzed by gas liquid chromatography (GLC). Following the treatment of Lai et al. [38] on rice, the cell wall material (CWM) was isolated according to the procedure of total (TDF) and insoluble (IDF) dietary fiber [39] by using an enzymatic kit for fiber determination (Bioquant, Merck, Germany). Briefly, 0.5 g of milled grains were incubated in phosphate buffer 0.08 M pH 6.00 with heat-stable α-amylase at 95–100 °C for 30 min, followed by cooling and hydrolysis with protease at 60 °C for 30 min at pH 7.5. Successively, at the same temperature, the last enzymatic step with amyloglucosidase at pH 4.5 was performed. For TOAX fraction analysis, the polysaccharide material was further precipitated with 95% ethanol (T = 60 °C) and then filtrated with the Fibertec semiautomatic system (FossItalia); this procedure found the CWM to contain 2.4–2.8% of proteins. The residue for WUAX determination was isolated by directly filtrating after the treatment with amyloglucosidase. Both residues were oven dried at T = 105 °C overnight, and, then, quantitatively recovered and hydrolysated with 10 mL of 2M TFA at T = 121 °C for two hours in a screw-capped glass tubes. Total neutral sugars were estimated by GLC [40,41], by using the following standard sugar mixture: D(−)arabinose, D(+)xylose, D-(+)mannose, D(+)glucose monohydrate (more than 99% Fluka), D-(+)galactose (more than 99% Sigma Aldrich) and internal standard β-D-allose (99%Acros Organics or 98%Alfa Aesar).

The chromatographic conditions employed were: column temperature 220 °C; injection and detector temperature 275 °C, helium flow rate 1.5 mL min⁻¹. The instrument (Clarus 600, PerkinElmer, Shelton, USA) was equipped with an autoinjector and a Restek capillary column (RTX 2330 30 m, 0.32 mm, 0.2 μm). The AX and WUAX contents were calculated as in Ingelbrecht et al. [42]: AX = [%xylose + %arabinose − (0.7 × %galactose)] × 0.88. The soluble AX was calculated by difference: TOAX–WUAX.

As reported by Li et al. [22] two independent aliquots of milled whole grains were considered statistical replicates of each grain composite sample. Analysis of variance (ANOVA) was performed with MSTATC program (Michigan State University, East Lausing, MI) by using a factorial model (mod.9) with cultivars, environment (locality and year) and variety by environment interaction; variety means were separated using Duncan’s multiple range test (p ≤ 0.01) for each environment separately and by combining the results across environments. Principal component analysis (PCA), performed with MATLAB software (R2010a version, MathWorks Inc., USA), was used to study the variations associated with genotype and environment for AX fractions. Simple correlation coefficients were also calculated.