First of all, we have observed genetic variation for cell wall bound hydroxycinnamates in the maize MAGIC population evaluated, for both the RILs and the founders agreeing with former evaluations . The correlations between cell wall components traits followed the trends previously reported in the literature [2, 24, 44]. FA, particular dimers and DFAT showed co-variation. This means that if the target for improvement is FA, individual and total dimers would be also modified. On the other hand, this would not happen for PCA, as this trait was not correlated with any other trait.
Furthemore, moderate to high heritability for hydroxycinnamate contents agreed with the results obtained in previous studies [2, 24]. From those heritabiliy values, we would expect a good response to selection, since additive effects are more important than additive × environment interaction effects. Although phenotypic selection could be effective based on those heritability estimates, a genomic selection approach could be implemented to speed up selection, in contrast to a time consuming phenotyping method.
With respect to other studies, novel genomic regions involved in hydroxycinnamate content were found, such as those in bins 10.04 (PCA), 5.06 and 7.01 (for FA), 2.08 and 7.02 (for DFA 5-5), 7.04 (for DFA 8-O-4), 5.06 (for DFA 8-5) and 5.01 and 7.02 (for DFAT). Nine of the ten SNPs associated with PCA, that correspond to three QTL, are located in bin 10.04. These QTL appear to be good candidates for selection targeting PCA. We found overlapping QTLs for FA and DFA 8-5 in the bin 5.06 and for DFA 5-5 and DFAT in bin 7.02. Those co-localizations have sense because genotypic correlations between those traits are high.
In addition, SNPs significantly associated with a trait in the current study co-localized with QTL for the same trait found by other authors. This is the case of the marker associated with PCA in bin 1.02, that co-localized in the same bin with one QTL associated with the same trait portrayed by García-Lara et al. , studying the cell wall phenolic composition of maize pericarp tissues in 163 F2:3 families; or the case of DFA 8-O-4 in bin 1.07 that co-localized with QTLs previously noted by López-Malvar et al.  using a maize diversity panel. Most interesting, in the same regions that we found QTL associated with cell wall bound hydroxycinnamates, QTLs for resistance to pests, animal digestibility,biofuel production and other cell wall components have already been reported. The aforementioned co-localizations hepls to support the fact that the cell wall influences the final usage of maize.
However, despite the evidence that an indirect selection targeting cell wall hydroxycinnamates may be a good strategy for the improvement of final use of maize related traits, is not the case from the observed in the current MAGIC population. The low percentage of the variance of those traits explained by hydroxycinnamates, added to the lack of co-localizations among SNPS associated with hydroxycinnamates and SNPs associated with final use of maize related traits evaluated in previous research using the same MAGIC population (, López-Malvar et al. 2020a,b submitted), suggest the convenience of a direct and particular breeding strategy for the improvement of saccharification efficiency, digestiblity of the organic matter or tunnel length damage resistance.
Among genes proposed as candidates for cell-wall bound hydroxycinnamateswe highlight peroxidases involved in oxidative coupling of FA to form dimers, genes that are responsible for transcriptional control of the phenylpropanoid pathway, implicated in xyloglucan and arabinoxylan synthesis, and involved in polysaccharides synthesis and modification. Finally, we also spot genes involved in gibberellin and suberin biosynthesis.
Cellulose and glucurono-arabinoxylans are the main constituents of lignified secondary walls; among the genes involved in the upstream parts of cell wall carbohydrate biosynthesis, we found several UDP-glycosyltranferases within the supporting intervals of QTL for PCA (qPCA_10_3), DFA 5-5 (qDFA5-5_7_1) and 8-O-4 (qDFA-8-O-4_1_1), and total diferulates (qDFAT_7_1). UDP-glycotransferases are enzymes involved in the elongation of carbohydrate chains using nucleotide sugar as substrates and thereby causing variation in the cell wall polysaccharides structure . For example, in rice, the gene underlying the brittle-culm-14 mutants, was a nucleotide sugar transporter that causes reduced mechanical strength by decreasing cellulose content and altering wall structure, including higher xylan extractability. In addition, we spotlight a reduced residual arabinose 3 gene, which encodes an arabinosyltransferase that adds the second arabinose residue in a β-1,2 linkage in arabinoxylan chains, as candidate for the QTL qDFA8-O-4_7_1 . Besides, FA is ester bond to arabinose residues of arabinoxylans chains, and FA dimers crosslink hemicellulose chains binding specifically to arabinose [48, 49]. Modifying arabinosyl transferase activities could be a promising strategy for modulating ferulate cross-linkages in the walls [50, 51], so we consider Zm00001d021974 a good candidate for DFA 8-O-4 content. Similarly, in the confidence interval of DFA 5-5, we found a glycosyl hydrolase (Zm00001d006940). Glycosyl hydrolases are mainly cellulases and xylanases; that modify the phenolic composition of cell walls as they cleave phenolic ester linkages. Thus, we considered Zm00001d006940, which encodes a glycosyl hydrolase, a good candidate for qDFA5-5_2_1. Besides, FA could be also bound to α-xylosil of xyloglucan (XyG)  and diferulate crosslinking could also anchor the xyloglucan chains. XyG participates in the formation of cellulose- XyG network, allowing the attachment of cellulose to other wall polymers and playing an important role in cell wall extension during plant growth . In this sense , we spot a xyloglucan endotransglucosylase/hydrolase gene that codifies for an enzyme involved in modifications of the xyloglucan for cell wall elongation as candidate for the overlapping QTL for FA and DFA 8-5.
Similarly, it has been hypothesized that GA could influence phenolic cross-linking via an effect on peroxidases . As we mentioned throughout this work, phenolic groups can be oxidatively coupled by peroxidase + H2O2  and/or by oxidases (laccases) + O2  to form dimers. The conditions that increase the number of diferuoyl crosslinking included, among others, low gibberellin supply . In this context, within the supporting intervals of qDFA8-O-4_1_1, we highlight a gibberllin 20-oxidase 2, as candidate for DFA 8-O-4 content. In the same way, we also spotlight several peroxidase genes as candidates for FA and individual dimers in QTL qFA_5_1, qDFA8-5_5_1 and qDFA5-5_2_1.
Among the genes found in the interval of qPCA_3_1, we spot MYB tf 41 which has been classified by Du et al.  as involved in the “Phenylpropanoid Pathway”. Barrière et al.  proposed ZmMYB041 as the probable gene underlying a QTL for lignin content,and this could be also involved in PCA biosynthesis because increased lignification has been commonly associatedto higher PCA concentration .
Finally, we found, in the QTL interval for PCA in chromosome 10, a gene encoding a ω-hydroxypalmitate O-feruloyl transferase (Zm00001d024864) which takes part in the esterified suberin biosynthesis pathway. Suberin is a polymeric constituent of plant cell wall, which consists of two domains that are cross-linked. Concerning the aromatic fraction of suberin, hydroxycinnamates esters, such as PCA, fortify the crosslinking between arabinoxilans and suberin fatty acids; besides PCA deposition has been associated with highly suberized tissues. Even though there is some of the esterified PCA esterified to polysaccharides, as previously mentioned, most of the PCA in grasses is ester bond to S units of lignin [60, 61].