Phytoene synthase 1 (Psy-1) and lipoxygenase 1 (Lpx-1) Genes Influence on Semolina Yellowness in Wheat Mediterranean Germplasm

Phytoene synthase 1 (Psy1) and lipoxygenase 1 (Lpx-1) are key genes involved in the synthesis and catalysis of carotenoid pigments in durum wheat, regulating the increase and decrease in these compounds, respectively, resulting in the distinct yellow color of semolina and pasta. Here, we reported new haplotype variants and/or allele combinations of these two genes significantly affecting yellow pigment content in grain and semolina through their effect on carotenoid pigments. To reach the purpose of this work, three complementary approaches were undertaken: the identification of QTLs associated to carotenoid content on a recombinant inbred line (RIL) population, the characterization of a Mediterranean panel of accessions for Psy1 and Lpx-1 genes, and monitoring the expression of Psy1 and Lpx-1 genes during grain filling on two genotypes with contrasting yellow pigments. Our data suggest that Psy1 plays a major role during grain development, contributing to semolina yellowness, and Lpx-1 appears to be more predominant at post-harvest stages and during pasta making.

This study was conducted to identify new allele variants and/or allele combinations significantly affecting YPC in semolina through their effect on the synthesis and degradation of carotenoid pigments in grain and semolina. For this purpose, three complementary approaches were undertaken: the identification of QTLs associated with carotenoid content on a recombinant inbred line (RIL) population, the characterization of a Mediterranean panel of accessions for Psy1 and Lpx-1 genes, and monitoring the expression of Psy1 and Lpx-1 genes during grain filling in two genotypes with contrasting YPC.

Phenotyping
The ANOVA of the yellow index (YI) data assessed in the RIL population revealed that all factors in the analysis (genotype (G), environment (E), and G × E interaction) were statistically significant (p < 0.01, data not shown). The parental lines differed significantly in YI values in the three years, with cv. "Saragolla" showing consistently higher values than "02-5B-318" ( Table 1). The RIL values for YI appeared normally distributed and showed ranges of 5.38, 6.39, and 5.86 in 2015, 2016, and 2017, respectively ( Table 1). The broad-sense heritability of YI ranged from 0.85 to 0.88 in the three testing years, highlighting that the phenotype was largely due to a genotypic effect (Table 1).

. QTL Analysis and Identification of Carotenoid Genes
As a starting point for the identification of QTL regions, the genetic map developed for the RIL population by Giancaspro et al. [35] was used as a reference for the analysis. The composite interval mapping (CIM) method was employed for QTL analysis. Putative QTLs for YI are listed in Table 2. Eight QTLs for YI (with LOD > 3) were mapped on chromosomes 2B (one), 4A (one), 4B (three), 5A (two), and 7A (one). The percentage of phenotypic variation (R 2 ) for YI explained by individual QTLs ranged from 10% to 59%. The major QTL was detected on chromosome 2B, flanked by the markers IWB12724 and IWB11333 (IWB32245 as the closest marker). This QTL was significant within the three years and across them and explained from 30% to 59% of YI variations. The QTLs on chromosomes 5A (interval IWA1258-IWB72888) and 7A (interval IWB73689-IWB25891) were statistically significant in all cases, accounting from 12% to 21% of the phenotypic variation. The QTL on chromosome 4B (IWB6397-IWB44213) was significant in two environments, accounting for 13% of YI variation. The QTLs on chromosomes 4A (IWB12722-IWB14501), 4B (IWB73832-IWB11928; IWB71402-IWB53932), 5A (IWB65257-IWB35711), and 7A (IWB49295-IWB8841) were significant in one environment with R 2 values ranging between 10 and 20%. Table 2 summarizes the location of QTLs on the genetic map, their LOD scores, and the closest markers flanking the region with the respective associated marker.
In order to verify the possible presence of candidate genes for the yellow index trait, all single nucleotide polymorphism (SNP) markers present in the detected QTL regions were studied. These SNP markers, mapped by Giancaspro et al. [35], were subjected to BLASTn analysis (based on percentage identity) to verify that the sequences have a high percentage of homology with the biosynthesis genes of the carotenoid pigments. All gene sequences used as queries were derived from Colasuonno et al. [7]. The analysis allowed for the identification of the lipoxygenase gene (Lpx) in the QTL region on chromosome 4B (IWA103 marker located in the 4B-2 QTL region, Table 2), confirming the key role of this chromosome in trait control. For the other key genes involved in the catabolic and biosynthetic pathway of the carotenoid pigment, no polymorphic markers were present in the analyzed regions.

Expression Profile of Lpx Genes on Leaves
The relationships between Lpx and yellow index were analyzed in the leaf tissue of the bread wheat accession "02-5B-318" and the durum wheat cv. "Saragolla", characterized by low and high YIs, respectively. The Lpx genes mapped on chromosomes 4A, 4B, 5A, and 5B were considered in the expression study in order to analyze each homoeologous Lpx gene. Total RNA was extracted from kernels, and quantitative real-time PCR (qPCR) was carried out with genome-specific primers. Significant statistical differences were observed in the three homoeologous forms in the two parental lines. High expression levels and significantly different expression values (p < 0.05, t-test) were detected between genotypes "02-5B-318" and "Saragolla" (0.67 and 1.58, respectively) for the Lpx gene on chromosome 5B, indicating that the Lpx allele present in cv. "Saragolla", the parent with a a high YI, was more active in the accumulation of yellow pigment that the allele present in "02-5B-318", the parent with a low YI ( Figure 1). Furthermore, significant differences (p < 0.05) were also observed for Lpx genes located on homologous group 4, showing the positive contribution to the trait of the "02-5B-318" parent, with normalized expression data of 1.0 for both the A and B isoforms compared to 0.32 and 0.47 for the 4A and 4B genes of cv. "Saragolla".

Characterization of the Mediterranean Panel for Psy1-A1 and Lpx-B1
A previous study of our team identified the allele composition of the Psy1-A1 gene as a reliable indicator of semolina b* value [30]. To gain more insight into the genomic background of this gene, allele variants of Psy1-A1 were assessed in the whole population, analyzed herein, and the results are shown in Table 3. Similar allele frequencies were found for the alleles Psy1-A1a and Psy1-A1o, but Psy1-A1l predominated (Table 4). Semolina b* values were significantly lower for genotypes harboring the Psy1-A1a allele in comparison to the ones carrying the alleles Psy1-A1l or Psy1-A1o, which showed similar values (Table 4).  gene on chromosome 5B, indicating that the Lpx allele present in cv. "Saragolla", the parent with a a high YI, was more active in the accumulation of yellow pigment that the allele present in "02-5B-318", the parent with a low YI ( Figure 1). Furthermore, significant differences (p < 0.05) were also observed for Lpx genes located on homologous group 4, showing the positive contribution to the trait of the "02-5B-318" parent, with normalized expression data of 1.0 for both the A and B isoforms compared to 0.32 and 0.47 for the 4A and 4B genes of cv. "Saragolla". qPCR was performed on the Lpx genes mapped on chromosomes 4A, 4B, 5A, and 5B in leaves of the bread wheat accession "02-5B-318" (low YI) and the durum wheat cv. "Saragolla" (high YI). ANOVA and Fisher`s least significant difference (LSD) tests were used to underline significant differences between the genotypes, p < 0.05.

Characterization of the Mediterranean Panel for Psy1-A1 and Lpx-B1
A previous study of our team identified the allele composition of the Psy1-A1 gene as a reliable indicator of semolina b* value [30]. To gain more insight into the genomic background of this qPCR was performed on the Lpx genes mapped on chromosomes 4A, 4B, 5A, and 5B in leaves of the bread wheat accession "02-5B-318" (low YI) and the durum wheat cv. "Saragolla" (high YI). ANOVA and Fisher's least significant difference (LSD) tests were used to underline significant differences between the genotypes, * p < 0.05. Table 5 allowed the identification of five combinations (haplotypes) between allele variants of Lpx-B1.1, Lpx-B1.2, and Lpx-B1.3 genes ( Table 3). Three of them correspond to those previously described by Verlotta et al. [32], i.e., haplotypes I (Lpx-B1.1b and Lpx-B1.3), II (Lpx-B1.1a and Lpx-B1.2), and III (Lpx-B1.1c and Lpx-B1.2), but two new ones with a low frequency in the population, named haplotypes IV and V following the nomenclature of Verlotta et al. [32], were found for the first time in the present study (Table 6). Haplotype IV, which showed the combination of Lpx-B1.1b and Lpx-B1.2, was identified in three landraces from Spain, Cyprus, and Morocco, whereas haplotype V, formed by the combination of Lpx-B1.1a and Lpx-B1.3, was detected in four landraces, two French, one Syrian, and one Spanish (Table 3). Alignments were made between the obtained amplicons and the corresponding sequences for Lpx-B1.1a, Lpx-B1.1b, Lpx-B1.1c, Lpx-B1.2, and Lpx-B1.3, showing no presence of new genes or allele sequences within the population evaluated (data not shown).

Primer pairs outlined in
Large variability was identified in haplotypes I and II, with b* values ranging from 12.38 to 25.78, and 10.32 to 23.49, respectively. These haplotypes were the most frequent within the Mediterranean panel ( Table 6). The results of the ANOVA and the comparison of the mean b* values of the five haplotypes showed no statistical differences among them (Table 6). Interestingly, all modern genotypes harbored haplotype II (Table 3). When the genotypes carrying haplotype II were segregated in "haplotype II landraces" and "haplotype II modern cultivars", significant differences appeared between them, as the latter encompassed the cultivars with higher levels of yellowness ( Figure 2).      To further dissect the population into smaller groups and get more information on the combinations between allele variants at different loci, we named the combinations between the three Psy1-A1 allele variants and the five Lpx-B1 haplotypes identified in this study. Eleven different combinations were obtained (named 1 to 10), and since eight of the nine modern genotypes harbored Genotypes from the Mediterranean panel were grouped according to their Lpx-B1 haplotype and a distinction was made between modern cultivars and landraces. One-way ANOVA and Tukey's post hoc test were performed to find significant differences (p < 0.05).
To further dissect the population into smaller groups and get more information on the combinations between allele variants at different loci, we named the combinations between the three Psy1-A1 allele variants and the five Lpx-B1 haplotypes identified in this study. Eleven different combinations were obtained (named 1 to 10), and since eight of the nine modern genotypes harbored combination 5, the modern/landrace distinction was applied to it ( Table 7). As expected, groups carrying Psy1-A1a (1 and 4) showed the lowest b* values. Allele combinations 2, 4, and 5 harbored approximately 80% of the population. By separating genotypes carrying combination 5 into modern cultivars and landraces, modern cultivars exhibited the highest b* values, which were significantly higher than the b* values of landraces from allele combination 5. The highest levels of yellowness (b* values over 19.0) were recorded for allele combinations 3 and 8 and in modern cultivars, without significant differences between them (Table 7).

Expression Levels of Psy1 and Lpx-B1 Genes during Grain Filling in Genotypes with Contrasting YI
The transcription levels of Psy1 and Lpx-B1 genes during grain filling were assessed in two landraces that showed consistently high (DW028) and low (DW011) YI values in each of the three environments (Table 3). The results showed that DW011 had no statistical difference in Lpx-B1 gene expression levels during the grain-filling period, and Psy1 had a peak at 28 days post anthesis (DPA) with a subsequent decrease in its expression ( Figure 3A). The high yellowness line (DW028) exhibited no statistical expression differences for the Lpx-B1 gene, similarly to DW011, but Psy1 showed an increase in expression from 14 to 49 DPA, peaking at 42 and 49 DPA ( Figure 3B). When the Psy1 expression levels between the low and high yellowness genotypes were compared, they did not show statistical differences until 42 and 49 DPA, when DW028 had 7.5 and 5.8 times higher expression than DW011, respectively ( Figure 3C). In the case of Lpx-B1, DW011 had 1.8 and 2.1 times higher expression than DW028 at 14 and 21 DPA, respectively, they showed no differences at 28 and 35 DPA, and the trend reverted at 42 and 49 DPA, when DW028 had 9.3 and 9.9 times higher expression than DW011 ( Figure 3D). statistical differences until 42 and 49 DPA, when DW028 had 7.5 and 5.8 times higher expression than DW011, respectively ( Figure 3C). In the case of Lpx-B1, DW011 had 1.8 and 2.1 times higher expression than DW028 at 14 and 21 DPA, respectively, they showed no differences at 28 and 35 DPA, and the trend reverted at 42 and 49 DPA, when DW028 had 9.3 and 9.9 times higher expression than DW011 ( Figure 3D).

Discussion
Yellowness is a major trait determining durum wheat quality for pasta making purposes. Psy1 and Lpx-1 have been pointed out as key enzymes predominately involved in carotenoid synthesis and degradation, respectively. Several studies reported many QTLs for yellow index spread all over the wheat genome [7], but to our knowledge no previous research has been reported simultaneously analyzing the two genes, the diversity of genetic materials, and expression profiles. Among them, two different QTLs were mapped on the long arm of chromosome group 7, co-localized with the phytoene synthase 1 (Psy1) and aldehyde oxidase 3 (AO3) genes, respectively [23,[37][38][39]. While the Psy1 involvement in YPC has been deeply studied, the role of the AO3 gene in carotenoid accumulation needs to be elucidated. AO isoforms are key enzymes for abscisic acid (ABA) biosynthesis [40,41]. The plant AO family is composed of proteins with a high similarity in sequences, but different subunit compositions and substrate preferences. AO isoforms have been largely characterized in Arabidopsis and are composed of four isoforms [42].
In the current study, the Lpx gene was studied in a RIL population developed by crossing two genotypes, "02-5BIL-318" and "Saragolla", which consistently differ in their yellow coloring, thanks to its association with a QTL located on chromosome 4B. The localization on chromosome 4B of the Lpx gene was confirmed even in other research studies [32,43]. As reported from other authors [32,44,45], the linkage analysis highlighted not only the connection between this gene and the QTL, but also confirmed the key function of this gene in the oxidation of carotenoids. Thus, only for the isoforms mapped on chromosome group 4 was it deduced that the genotypes with a low content of carotenoid pigments were characterized by a greater catabolic activity. Carrera et al. [4] showed how the role of lipoxygenase on carotenoid degradation occurs in the process of pasta making and not during wheat grain development, which is in agreement with our own results.
The distribution of the Lpx-B1 gene family and the Psy1-A1 alleles were studied in a large durum wheat population comprising 128 Mediterranean landraces and modern cultivars, in order to link their presence and distribution to semolina yellowness.
The allelic variant Lpx-B1.1c was present in only two of the 128 genotypes analyzed in the current study, and they had very different levels of yellowness (DW059 and DW008, Table 3), even though both carried the Psy1-A1 alleles that are related to high levels of carotenoid synthesis (Table 4, [29]), and they are supposed to have low levels of LOX activity [4,32].
Verlotta et al. [32] made an association between LOX activity and haplotype types on wholemeal extracts, simulating the pH conditions (pH 6.6) during pasta processing, in which lipoxygenases are expected to be most active. In addition, Carrera et al. [4] reported that between 8 and 16% of the total carotenoid content is lost due to LOX activity during pasta processing.
Previous studies have not found relationships between LOX activity and yellow color in semolina [31,34]. In these studies, it was reported that in the steps from whole meal to semolina, and from semolina to pasta, there is a marked reduction in YPC, especially in genotypes having high LOX activity. LOX activity apparently becomes more important at the time of wheat processing, not before, so it makes more sense not to have found a relationship between the presence of the alleles and b* values. Additionally, when LOX activity was studied with different pH, it was reported that in varieties with high LOX activity, there is apparently more than one active LOX isoform, since the decrease in activity as the pH increases is not as pronounced as for cultivars with low LOX activity. In addition, a well-defined peak of maximum LOX activity was observed, which decayed rapidly when altering the pH, suggesting that in these cases a single isoform is produced, determining the LOX activity in wholemeal durum wheat. The authors found a relationship between LOX activity in semolina and yellowness in pasta. The presence of transcripts is not entirely consistent with the LOX activity reported in their work, so post-transcriptional mechanisms could be playing a role in the modulation of LOX activity. Transcript levels also associated with LOX activity in certain durum wheat varieties may be due to Lpx-1 and Lpx-3, while in others, only to Lpx-1. In addition, Verlotta et al. [32] reported the possibility of the existence of more allelic variants for Lpx-B1.2 and Lpx-B1.3 that were not detected in the current work with the specific primers used. Further, we can speculate that the differential Lpx-1 gene expression identified during the vegetative stage in this study, but not during grain filling, probably may not affect YPC levels, but post-transcriptional mechanisms, durum wheat processing conditions, and/or other allelic variants for Lpx-1 do influence semolina yellowness.
The expression levels of Psy1 and Lpx-B1 genes were analyzed during grain filling using two Mediterranean durum wheat genotypes with contrasting levels of yellowness compiled and purified at IRTA: DW011 (low yellowness; b*= 14.18 ± 1.29) and DW028 (high yellowness; b*= 21.71 ± 0.55). Psy1 exhibited a peak of expression at 42 and 49 DPA in the high yellowness landrace that could result in higher levels of grain phenotypic yellowness ( Figure 3A,B). In the present study, no expression differences during grain filling were encountered for Lpx-B1 in both contrasting genotypes. The high yellowness genotype had 7.5 and 5.8 times higher expression levels for Psy1 than the low yellowness landrace at 42 and 49 DPA ( Figure 3C). Interestingly, our group previously studied 12 modern durum wheat genotypes that showed greater expression of Psy1-A1, which specifically peaked at 35 DPA. At this time, Psy1-A1 was 21-fold more highly expressed in the high yellowness genotypes relative to the low yellowness genotypes [46]. The results of the present study suggest that not only is the particular allele of Psy1 responsible for the determination of grain yellowness (Psy1-A1l, Table 4), but its expression levels may also play a role in this trait (Figure 3). In addition, the higher expression levels of Psy1 of the landrace studied occurred late in grain development, while the modern durum genotypes had their peaks earlier (35 DPA), which may distinctly influence grain yellowness.
Regarding DW011, it is associated with haplotype II for Lpx-B1 (Table 3) and group 4 for Lpx-B1/Psy1-A1 (mean b* value 16.22 ± 0.39; Table 7), while DW028 is categorized as haplotype I for Lpx-B1 (Table 3) and group 3 for Lpx-B1/Psy1-A1 (mean b* value 19.25 ± 0.61; Table 7). There may be a relationship between the expression levels for Psy1 in the late grain filling stages and b* value. In the case of Lpx-B1, the DW011 had 1.8 and 2.19 times higher expression levels at 14 and 21 DPA than DW028, but this last genotype exhibited 9.36 and 9.97 higher expression levels at 42 and 49 DPA than the low yellowness landrace. Since Lpx-B1 encodes for lipoxygenase, lower gene expression levels were expected in the high yellowness genotypes, in fact, if the ratio Psy1/Lpx-B1 expression levels are compared at 49 DPA, the low yellowness genotype showed a ratio of 41.75 and the high yellowness genotype had a ratio of 24.17. Considering that the b* values for DW011 and DW028 are 14.18 ± 1.29 and 21.71 ± 0.55, respectively (Table 3), our result suggests that the expression of Psy1 had a greater influence than Lpx-B1 on b* values, but protein quantification and activity assays are required to draw more definitive conclusions.
Finally, high levels of carotenoid pigments in wheat kernels have important positive implications for human health since they are antioxidant compounds and precursors of vitamin A. Identifying the role of the main carotenoid genes in durum wheat and the specific alleles present in each cultivar can allow the development of superior cultivars through marker-assisted breeding programs. Molecular markers associated with Lpx-1 are currently being effectively used for marker-assisted selection in durum wheat breeding programs to improve YPC in different parts of the world, including the United States [11] and Canada [2].

Plant Material and Experimental Setup
A set of 135 F [6][7] RILs was developed by the Department of Environmental and Territorial Sciences, University of Bari, Italy (DISAAT) through a single seed descendant (SSD) method, as described by Giancaspro et al. [35], and used for QTL analysis. The parents were the bread wheat accession "02-5B-318", derived from the Chinese cv. "Sumai-3", characterized by a low YI, and the durum wheat cv. "Saragolla", characterized by a high YI. The parents and the RIL population were grown in Valenzano (Metropolitan City of Bari) for three years (2015, 2016, and 2017) using a randomized complete block design with four replications. Each plot consisted of 1-m rows, 30 cm apart, with four g of seeds sown in each plot and supplemented with nitrogen (10 g/m 2 ).

Detection of QTLs for Carotenoid Content in the RIL Population
The identification of QTLs associated with YI was carried out based on the genetic map previously developed by Giancaspro et al. [35]. QTL detection was performed in QGene 8.3.16 using composite interval mapping (CIM), as proposed by Zeng [48]. The association between marker and trait was considered significant when one or more markers showed a −log10(p) ≥ 3.0, determined by modified Bonferroni correction. The contributions of "02-5B-318" and "Saragolla" were highlighted by a positive and negative sign, respectively.
Graphical representations of linkage groups and QTLs were carried out using MapChart 2.2 software (SolarWinds, Austin, TX, USA). Genes located in the QTL regions were identified by blasting the SNP sequences from Wang et al. [49] against the annotated Triticum (NCBI) and contig (URGI) sequences.
In order to pick primer combinations to use for the Lpx gene expression analysis, the sequences and the gene models were downloaded from the Svevo [50] and Chinese Spring [51] genome databases. The Lpx genes were located on chromosome groups 4 and 5 (4A: TRITD4Av1G195350; 4B: TRITD4Bv1G010710; 5A: TRITD5Av1G200190; 5B: TRITD5Bv1G195300). Primer combinations were designed in the conserved region of the first exon for all genes, considering dissimilarities between the two homoologous copies ( Table 9). The genetic materials used for the quantitative real-time PCR analyses were collected from leaves at the seedling stage belonging to the wheat cultivars "Saragolla" and "02-5B-318", characterized by high and low values of YI, respectively. Total RNA was extracted from the grain tissue of both genotypes using the RNeasy Plant Mini Kit (QIAGEN ® ) (Germantown, MD, USA) and checked on 1.5% denaturing agarose gel. All RNA samples were the same concentration (1 µg) and were reverse transcribed into double stranded cDNA with a QuantiTect Reverse Trascriptase Kit (QIAGEN ® ). Data were normalized, as previously described by Marcotuli et al. [52], using three reference genes (ADP-RF, RLI, and CDC), which have a stability value, calculated with NormFinder software, of 0.035 [52].
Quantitative real-time PCR was carried out using Cyber ® GREEN in the CFX96TM real-time PCR system (BIO-RAD) following the protocol described by Marcotuli et al. [52]. A series of six scalar dilutions were carried out to determine the primer amplification efficiency, while the specificity of the amplicons was confirmed by the following steps: a single band on the agarose gel (2% w/v) and a single peak in the melting curves and the sequence of the amplified fragments (3500 Genetic Analyzer, Applied Biosystems).
The qRT-PCR data for genes were derived from the mean values of three independent amplification reactions carried out on the parental lines ("Saragolla" and "02-5B-318") of the RIL population.
Three reference genes were used to normalize data (cell division control AAA superfamily of ATPases, CDC; ADP-ribosilation factor, ADP-RF; RNase L inhibitor-like protein, RLI [53,54]). All calculations and analyses were performed, as reported by Marcotuli et al. [52], using CFX Manager 2.1 software (Bio-Rad Laboratories), applying the ∆∆Ct method. Standard deviations were used to normalize values, and ANOVA and LSD tests were used to underline significant differences between the genotypes.

Psy1-A1 and Lpx-B1 Genotyping in the Mediterranean Panel
Plants were grown in the greenhouse, in triplicate, in a mixture of peat and perlite in a 2:1 ratio, without fertilization, until they reached the two to three leaves stage, in order to extract DNA with the CTAB protocol [55], with minor modifications. One hundred milligrams of leaves were crushed to powder in the presence of liquid nitrogen and 1 mL of CTAB buffer and 0.2% β-mercaptoethanol and 2% PVP were added. The mixture was incubated at 65 • C for 30 min, mixing it gently every 5 min. Then, 1 mL of chloroform and isoamyl alcohol in a ratio of 24:1 were added, mixed by vortexing, and centrifuged at 13,000 rpm for 10 min. Then, the upper phase was extracted and mixed with an equal volume of isopropanol and incubated overnight at −20 • C. Then, the mixture was centrifuged at 13,000 rpm at 4 • C for 10 min and the pellets were air dried at 37 • C, before adding 50 µL of ultrapure water (Invitrogen) (Waltham, MA, USA). Each DNA sequence was quantified by QUBIT 3.0 (Life Technologies) (Carlsbad, CA, USA) and the integrity was checked by agarose electrophoresis.
Regarding Psy1-A1, 112 out of 128 genotypes used in this study were previously characterized by Campos et al. [30]. The rest of the population was genotyped with the marker PSY1-A1_STS (Table 5) and PCR conditions used by Singh et al. [29], allowing the discrimination between the alleles Psy1-A1a, Psy1-A1l, and Psy1-A1o, using the previously extracted DNA.
In order to amplify the Lpx-B1 genes and different alleles [4,32], the primer pairs in Table 5 were employed in this study. The PCR conditions for every sequence were as follows: Lpx-B1.1a

Psy1 and Lpx-1 Gene Expression during Grain Filling
The expression of Psy1 and Lpx-1 genes during grain filling was monitored through qPCR. For total RNA extraction, 100 mg of grain tissue were used at each grain developmental stage from 14 DPA to 49 DPA, in triplicate, for genotypes DW011 (low yellowness) and DW028 (high yellowness). RNA extraction was performed based upon the Furtado [56] protocol, with some modifications. The two-step RNA extraction involved the use of TRIzol™ Reagent (ThermoFisher Scientific, Carlsbad, -CA, USA), and the NucleoSpin ® RNA Plant Kit (Macherey-Nagel, Düren, Germany). Firstly, seed tissue was pulverized in a mortar and pestle in the presence of liquid nitrogen and 1.5 mL of TRIzol Reagent™ were added. The samples were then centrifuged at 12,000× g and 4 • C for 10 min. Subsequently, the upper phase (~750 µL) was recovered, mixed with 300 µL of chloroform and agitated through inversion for 3 min, and then centrifuged for 15 min at 12,000× g and 4 • C. The upper phase was then recovered in a new tube, and from this step onwards, the NucleoSpin ® RNA Plant Kit was used according to the manufacturer's instructions. Finally, RNA was eluted with 30 µL RNAse-free H 2 O. RNA integrity was verified on 2% agarose gel and visualized with GelRed ® staining (Biotium (company name), Fermont, CA, USA), and RNA was quantified using the NanoDrop 2000 spectrophotometer (ThermoFisher Scientific (company name), Carlsbad, CA, USA). First strand cDNA synthesis was obtained from 4 µg of total RNA using SuperScript™ II (ThermoFisher Scientific (company name), Carlsbad, CA, USA) and oligo dT primers, following the manufacturer's instructions.
The relative expression analyses were carried out using the SensiFast TM SYBR ® No-ROX kit (Bioline (company name), United Kingdom). For each reaction, 50 ng of cDNA were used and tested for the expression of the genes Psy1 (primers qrtPsy1_F and qrtPsy1_R, Table 5) and Lpx-1 (primers qrtLpx_B1_F and qrtLpx_B1_R, Table 5) using the ADP ribosylation factor as a reference gene [53]. All primers were designed using the Amplifix and the NCBI primer blast, and they are further characterized in Table 5. The efficiency and CT value of each reaction were determined using LinRegPCR V2018.0 [57], using a common threshold and an individual fit. The relative expression of each gene was calculated using the Pfaffl method [58] with a calibrator value of 0, and then the relative expression was divided by the lower value in each group of data. Three biological and three technical replicates were used at each developmental stage.

Yellowness Determination
Yellow intensity was assessed in the two germplasm panels in wholegrain flour through the yellow index (YI) [11]. L*, a* and b* coordinates in the Munsell color system were taken using D65 lightning. A reflectance colorimeter (CR-400, Konica-Minolta, Japan) equipped with a filter tri-stimulate system was used for these determinations.

Statistical Analyses
Statistical analyses were performed using GraphPad Prism version 8.2.1. GenStat (version 18, VSN International Ltd., Hemel Hempstead, UK), which was used to carry out the ANOVA in the RIL population on the YI to identify how much of the variation was attributed to genotype. Broad-sense heritability (h 2 B ), was estimated, accounting the genetic variance, σ 2 G , the genotype x environment interaction variance, the residual variance, the number of environments (in this case three), and the number of replicates per line (in this case three replicates) [52,59].

Conclusions
Semolina yellowness is a key aspect determining durum wheat end products. Understanding the synthesis and degradation of carotenoid pigments is a vital aspect of durum wheat breeding programs aiming to produce superior cultivars with higher yellowness. In the current study, eight QTLs related to yellowness were identified, with different degrees of involvement in YI variation, and Lpx was linked to a QTL found in chromosome 4B, which showed lower expression in a high yellowness cultivar in comparison to a low yellowness wheat.
The Lpx-B1 characterization in a landrace and modern wheat Mediterranean population allowed the identification of novel allele combinations (haplotypes IV and V), and by adding a Psy1-A1 characterization, 11 different combinations of groups appeared, revealing different levels of YI, the modern group being the one with higher levels of yellowness, harboring Psy1-A1l, Lpx-B1.1a, and Lpx-B1.2. Besides, by comparing high and low yellowness landraces, it is possible to suggest that, during grain development, Psy1 plays a major role contributing to semolina yellowness, and Lpx is suggested to be more predominant at post-harvest stages and during pasta making. The practical applications of our work in durum wheat breeding programs relate to the use of marker-assisted selection for Psy1 alleles showing higher transcript abundance and greater semolina yellowness, whose genotypes could be specifically evaluated and selected at 42 DPA. In addition, studying whether the high yellowness Psy1 alleles and transcript abundance could be identified at earlier vegetative stages (i.e., leaves) of the plant would be very useful to reduce the selection times without waiting for grain filling. Furthermore, the utilization of biotechnological approaches to elevate the Psy1 transcripts at late grain development stages would be valuable to potentially increase semolina yellowness. Finally, this study highlights the applicability of markers linked to the Lpx-1 gene in marker-assisted selection programs.

Conflicts of Interest:
The authors declare no conflict of interest.