Extracellular cross-linking of maize arabinoxylans by oxidation of feruloyl esters to form oligoferuloyl esters and ether-like bonds


*For correspondence (fax +44 0131 650 5392; e-mail s.fry@ed.ac.uk).


Primary cell walls of grasses and cereals contain arabinoxylans with esterified ferulate side chains, which are proposed to cross-link the polysaccharides during maturation by undergoing oxidative coupling. However, the mechanisms and control of arabinoxylan cross-linking in vivo are unclear. Non-lignifying maize (Zea mays L.) cell cultures were incubated with l-[1-3H]arabinose or (E)-[U-14C]cinnamate (radiolabelling the pentosyl and feruloyl groups of endogenous arabinoxylans, respectively), or with exogenous feruloyl-[3H]arabinoxylans. The cross-linking rate of soluble extracellular arabinoxylans, monitored on Sepharose CL-2B, peaked suddenly and transiently, typically at ∼9 days after subculture. This peak was not associated with appreciable changes in peroxidase activity, and was probably governed by fluctuations in H2O2 and/or inhibitors. De-esterified arabinoxylans failed to cross-link, supporting a role for the feruloyl ester groups. The cross-links were stable in vivo. Some of them also withstood mild alkaline conditions, indicating that they were not (only) based on ester bonds; however, most were cleaved by 6 m NaOH, which is a property of p-hydroxybenzyl–sugar ether bonds. Cross-linking of [14C]feruloyl-arabinoxylans also occurred in vitro, in the presence of endogenous peroxidases plus exogenous H2O2. During cross-linking, the feruloyl groups were oxidized, as shown by ultraviolet spectra and thin-layer chromatography. Esterified diferulates were minor oxidation products; major products were: (i) esterified oligoferulates, released by treatment with mild alkali; and (ii) phenolic components attached to polysaccharides via relatively alkali-stable (ether-like) bonds. Thus, feruloyl esters participate in polysaccharide cross-linking, but mainly by oligomerization rather than by dimerization. We propose that, after the oxidative coupling, strong p-hydroxybenzyl–polysaccharide ether bonds are formed via quinone-methide intermediates.


Arabinoxylans in the primary cell walls of the Poales (grasses, cereals and related plants) are particularly interesting because they bear ferulic acid groups esterified to some of their arabinose residues (Geissmann and Neukom, 1971; Harris and Hartley, 1976; Kato and Nevins, 1985; Wende and Fry, 1996), as has also been observed for pectins in the Caryophyllales, e.g. spinach and beet (Fry, 1982, 2004; Ishii, 1994; Levigne et al., 2004). Ferulic acid, a phenolic compound, is susceptible to oxidative coupling when exposed to hydrogen peroxide plus peroxidase (Geissmann and Neukom, 1971), and it has often been suggested that the esterified diferulates (also known as dehydrodiferulates) thus formed can cross-link cell-wall polysaccharides, contributing to wall assembly and to wall tightening during maturation (Hatfield et al., 1999; Fry, 2004; Parker et al., 2005). Evidence favouring this hypothesis includes the gelation of feruloyl polysaccharides when exposed to oxidative systems (Geissmann and Neukom, 1971; Schooneveld-Bergmans et al., 1999), and the isolation of diferulate isomers from alkali hydrolysates of the cell walls of various members of the Poales and Caryophyllales (Ralph et al., 1994; Bunzel et al., 2004; Levigne et al., 2004).

Recently, larger coupling products (here described as oligoferulates) have been reported to predominate over diferulates (Fry et al., 2000), and certain specific triferulate and tetraferulate isomers have been characterized (Rouau et al., 2003; Bunzel et al., 2004, 2005, 2006; Funk et al., 2005; Barberousse et al., 2008). Such products are likely to be diverse, and it is therefore likely that any individual trimer or tetramer, such as those reported by the authors above, will be present in small quantities.

Evidence compatible with the possibility that diferulates act as covalent cross-links between arabinoxylan chains was provided by the discovery of a compound that consisted of 5,5′-diferulate ester-linked to two arabinoxylan-derived trisaccharide groups, i.e. trisaccharidediferulatetrisaccharide (where ‘’ is an ester bond), among the Driselase digestion products of bamboo polysaccharides (Ishii, 1991). Likewise, cyclic 8,8′-diferulate that was ester-linked to two arabinose groups (or to an arabinose group and a xylosyl-arabinose group) was isolated from a mild acid hydrolysate of maize bran polysaccharides (Bunzel et al., 2008). Furthermore, in dicots, the arabinan and galactan side chains of sugar-beet pectins are covalently cross-linked via esterified diferulate bridges (Ralet et al., 2005). However, in all such cases, it remains to be shown that the two sugar moieties involved were derived from different polysaccharide molecules, and thus that the diferulate bridge was interpolymeric; the alternative possibility is that diferulate forms an intrapolymeric loop (Lindsay and Fry, 2008).

Resolving this question is difficult because it has so far proved impossible to identify which wall-bound arabinoxylan molecules are covalently cross-linked, and which are single chains merely hydrogen-bonded to the cellulosic microfibrils. However, maize cells in suspension culture secrete wall-related soluble polysaccharides, notably arabinoxylans, into their medium (Kerr and Fry, 2003). The reactions of such soluble extracellular arabinoxylans in living cultures are a valuable model for the reactions undergone by insoluble arabinoxylans within the cell wall itself. Studying soluble extracellular arabinoxylans has several advantages over previous approaches to elucidating the mechanism and control of hemicellulose cross-linking. One is that it is possible to follow reactions in vivo, in real time, with live maize cells present and secreting enzymes, co-factors and inhibitors at various developmental stages during the culture cycle. Another is that it is possible to measure the cross-linking of arabinoxylans directly by the detection of changes in their molecular weight by gel-permeation chromatography (a technique not applicable to wall-bound polysaccharides). It is reasonable to assume that enzymes causing changes in soluble extracellular arabinoxylans of a maize culture have also at least partial access to wall-bound polysaccharides, so that the latter undergo reactions comparable to those occurring in the medium. Finally, the cross-linking reactions involving soluble extracellular arabinoxylans may be examined in greater mechanistic detail than is the case with wall-bound polysaccharides, e.g. by the addition of exogenous reactants or inhibitors (Kerr and Fry, 2004; Encina and Fry, 2005).

In the present work, we document the timing and stability of cross-linking of soluble extracellular arabinoxylans in vivo and in vitro, and we report new observations on the behaviour of the feruloyl side chains during and after the cross-linking reactions. The results indicate that less significance should be attached to diferulate ester formation than is widely believed. We report that the formation of esterified oligoferulates is prevalent, and that ether-like (alkali-stable) bonds are produced between ferulate derivatives and polysaccharides during and (especially) after cross-linking.


Cross-linking of endogenous [3H]arabinoxylans in vivo

To track the cross-linking of endogenous hemicelluloses in vivo, we fed [3H]arabinose to a 6-day-old maize cell culture. The radioactivity was taken up by the cells within 2–6 h, and was largely incorporated into arabinose and xylose residues of polysaccharides, some of which were quickly sloughed into the medium as soluble extracellular [3H]arabinoxylans (with a small proportion of xyloglucans) of ∼0.8–1.6 MDa (Kerr and Fry, 2003). These elute within the fractionation range of Sepharose CL-2B (∼0.1–20 MDa).

At intervals after 3H-feeding, the [3H]arabinoxylans were size-fractionated on Sepharose mini-columns (Figure 1a). The [3H]arabinoxylans in 7–10-day-old cultures (1–4 days after radiolabelling) eluted as a partially included peak centred on Kav 0.5 (Figure 1a). [Kav is the elution volume of the compound relative to that of a high-molecular-weight subfraction of blue dextran (Kav 0) and to that of glucose (Kav 1).] By 11 days, about half of the radioactive material had shifted to Kav 0, indicating sudden, massive cross-linking. The other ∼50% remained at Kav 0.7–0.8. Thus, the original peak contained two main polymer populations: one comprising moderately large polymers (Kav 0.3–0.7) that cross-linked with increasing age; the other comprising smaller polymers (Kav 0.7–0.8) that did not cross-link.

Figure 1.

 Cross-linking of soluble extracellular arabinoxylans in living maize cell cultures, and its limited reversal by alkali.
(a) Cross-linking of endogenous [3H]arabinoxylans. Six-day-old cultures were fed [3H]arabinose, which was consumed within ∼2 h. On each subsequent day, soluble extracellular [3H]arabinoxylans were size fractionated.
(b) Endogenous [3H]arabinoxylans from 11-day-old cultures were incubated for 24 h in water or 0.1 m NaOH at 20 or 37°C.
(c) Cross-linking of exogenous [3H]arabinoxylans in vivo. Cumulative: [3H]arabinoxylans were supplied to 6-day-old cultures, then re-analysed when the cultures reached the ages indicated. Current: [3H]arabinoxylans were added to a culture of the age indicated, and then analysed 8 h later. The plot shows the percentage of [3H]polysaccharides found to be cross-linked. Error bars are 2σ, based on ratio of counts.

When partially cross-linked soluble [3H]arabinoxylans from 11-day-old cultures were treated with 0.1 m NaOH, they gradually reverted towards the original size distribution (Figure 1b). The addition of 0.1 m NaOH at 20°C for 24 h (sufficient for the complete hydrolysis of feruloyl esters; Fry, 1982) had little effect; however, an equivalent treatment at 37°C was more completely effective. Similar results were obtained for 12- and 13-day-old cultures (data not shown). The failure to obtain complete reversion to the original size distribution indicates that at least some of the cross-links were more alkali-resistant than feruloyl ester bonds.

Cross-linking of exogenous [3H]arabinoxylans in vivo reveals a physiological burst of extracellular cross-linking activity

To determine whether arabinoxylan cross-linking occurred in the culture medium (the alternative interpretation of Figure 1a being that older cultures began releasing wall-bound, cross-linked [3H]polysaccharides), we fed exogenous non-crosslinked [3H]arabinoxylans to maize cultures. In a ‘cumulative cross-linking’ experiment (Figure 1c), the [3H]arabinoxylans were added to 6-day-old cultures, and were then incubated for varying additional times. For the first 2 days after feeding, the [3H]arabinoxylans did not cross-link. Abruptly, however, between the ages of 8 and 9 days, about one-third of the [3H]arabinoxylan shifted to Kav 0, indicating cross-linking. (The onset of cross-linking varied between experiments, from 8 to 11 days after subculture.) Negligible 3H was lost from the culture medium during the experiment, suggesting that little wall-binding occurred. No decrease in cross-linked [3H]arabinoxylans occurred after 9 days, indicating that the cross-links were stable in vivo.

The sudden cross-linking of exogenous [3H]arabinoxylans after a 2–3-day lag period could indicate that either (i) [3H]arabinoxylan molecules could cross-link only after 2 days in the culture, or (ii) there was a sudden change in the physiological activities of the cells. To distinguish between these possibilities, we fed [3H]arabinoxylans to cultures of various ages followed by exactly 8 h of additional incubation. This experiment (Figure 1c) showed that cross-linking did not depend on the [3H]arabinoxylans being in contact with cells for more than 2 days. ‘Current cross-linking’ activity peaked in 9–11-day-old cultures, and fell to lower rates in 12- and 13-day-old cultures. These data show that cross-linking is a physiological action of the maize cells that varies with the age of the culture.

Acquisition of alkali-resistant cross-links between
[3H]arabinoxylans in vitro: effect of H2O2 dose

Figure 1 shows the formation of remarkably alkali-resistant cross-links between arabinoxylan molecules in vivo. To investigate the mechanisms responsible, we developed a cell-free cross-linking system. Dialysed culture filtrate contained not only polysaccharides but also peroxidase activity (with o-dianisidine as the substrate; data not shown). In initial attempts to generate alkali-resistant cross-links between [3H]arabinoxylans in vitro, we used dialysed culture filtrate plus 7.3 mH2O2. This rapidly led to cross-linking, but even mild alkali (0.1 m NaOH, 20°C) caused the cross-linked material to shift to Kav = 1, indicating extensive polysaccharide scission (data not shown). Such breakdown could be caused by the high concentration of H2O2 forming hydroxyl radicals, which oxidatively convert some glycosidic bonds to ester bonds (Miller and Fry, 2001), causing the polysaccharide to fragment at high pH levels. High H2O2 concentrations have likewise been reported to decrease the viscosity of polysaccharide/peroxidase mixtures (Schooneveld-Bergmans et al., 1999).

However, we achieved the production of alkali-stable cross-links in vitro by taking two precautions: (i) multiple low doses of H2O2 were supplied [comparable to the Zutropfverfahren (‘drip procedure’) of Freudenberg et al. (1963) and Barakat et al. (2007)] rather than a single high dose; and (ii) any H2O2 remaining after cross-linking was scavenged with dithiothreitol prior to alkali treatment. With these precautions, cross-links were consistently formed in vitro that remained stable in mild alkali (0.1 m NaOH at 20°C for at least 48 h; Figure 2a): a degree of alkali resistance that is incompatible with ester cross-links alone. Nevertheless, most of the cross-links formed under these conditions could eventually be cleaved by concentrated alkali (Table 1). The latter conditions were more severe than are required to cleave any known carboxyl esters, and we conclude that relatively stable, ether-like cross-links (Enoki et al., 1983) had formed.

Figure 2.

 Cross-linking of arabinoxylans in a cell-free system and limited reversal by alkali.
(a) Non-crosslinked [3H]arabinoxylans in dialysed culture filtrate were supplied with six 5.87-μm doses of H2O2 at hourly intervals, then hydrolysed with 0.1 m NaOH at 20°C for 0–48 h (as indicated) and size fractionated.
(b) Graded acquisition of alkali stability. The same protocol was followed as in (a), but between one and three doses of H2O2 were supplied at daily intervals: (i–iii), not subsequently alkali-treated; (iv–vi), subsequently treated with 0.1 m NaOH for 24 h.
(c) De-esterification prevents cross-linking. Non-crosslinked [3H]arabinoxylans in culture filtrate were de-esterified, slightly acidified, and then incubated with (i) no additives, (ii) H2O2 or (iii) H2O2 plus non-de-esterified, non-radioactive culture filtrate (CF). (iv), as (ii) but without de-esterification; (v), as (ii) but de-esterified only after H2O2 treatment.

Table 1.   Effect of severity of alkali treatment on cleavage of cross-links formed between [3H]arabinoxylans by repeated low doses of H2O2in vitro
NaOH concentration used (m)Percentage of [3H]polysaccharides found to be cross-linked after 24 h treatment with NaOH at
  1. Non-cross-linked [3H]arabinoxylans in dialysed 8-day culture filtrate were cross-linked by the addition of six aliquots of 5 μm H2O2 at hourly intervals. Any remaining H2O2 was then quenched with 10 mm dithiothreitol before hydrolysis with NaOH for 24 h at either 20 or 37°C and subsequent size fractionation on Sepharose CL-2B.


Interestingly, the acquisition of this alkali resistance required no other additives than H2O2, but developed more slowly than the cross-links (Figure 2b). Thus, the first cross-links to form are esters (susceptible to mild alkaline conditions), with the majority of the ether-like bonds forming only after more extensive oxidation.

Prior removal of feruloyl residues prevents arabinoxylan cross-linking

As maize arabinoxylans are feruloyl polysaccharides (Kerr and Fry, 2003), it is likely that their extracellular cross-linking involves oxidative coupling of the feruloyl residues. To test this idea, we de-esterified non-cross-linked [3H]arabinoxylans with 0.1 m NaOH, and then attempted to cross-link them in vitro (Figure 2c). Deferuloylated [3H]arabinoxylans lost all detectable ability to cross-link (Figure 2c.ii), even if culture filtrate containing non-radioactive feruloyl-arabinoxylans were also added to the reaction mixture (Figure 2c.iii), so that the [3H]arabinoxylans could potentially act as oxygen nucleophiles in ether formation (see Discussion). Control experiments with feruloyl-[3H]arabinoxylans confirmed the H2O2-induced in-vitro cross-linking (Figure 2c.iv), and the partial alkali resistance of the cross-links that were formed (Figure 2c.v).

Cross-linking of endogenous arabinoxylans in vivo involves the formation of oligoferuloyl esters and ether-like bonds, but few diferuloyl esters

To investigate the molecular changes occurring in esterified ferulic acid groups that enable arabinoxylan cross-linking, we fed 6-day-old cultures with [14C]cinnamate (a ferulate precursor). The [feruloyl-14C]arabinoxylans thus produced were collected 7–13 days after subculture, and were size fractionated. Arabinoxylan cross-linking in this culture began at 11 days, and was complete by 13 days (data not shown). The total [14C]arabinoxylans present (freed of any low-Mr (relative molecular mass) material) in the culture filtrate at each sampling time were: 0.24, 0.25, 0.20 and 0.26 kBq ml−1, at 7, 8, 11 and 13 days, respectively. This indicates that the majority of the incorporation and secretion of 14C was complete within the first 24 h.

The soluble extracellular [14C]arabinoxylans were mildly alkali-hydrolysed (cleaving ester bonds), acidified and partitioned against ethyl acetate (Figure 3). Negligible material became insoluble. Free ferulic, diferulic and oligoferulic acids, released by ester hydrolysis, were partitioned in the ethyl acetate phase because they are relatively hydrophobic. In samples collected at 7–11 days into the culture cycle, the great majority of the radioactive material was indeed hydrophobic, but by 13 days 60% of the radioactive material obtained was hydrophilic (Figure 4a). This indicates that by the last time-point a large proportion of the phenolic moieties were attached to hydrophilic (glycan) material via alkali-stable bonds. Thus, the phenol–polysaccharide bonds present in 13-day-old cultures included relatively alkali-resistant (i.e. not ester) linkages, in addition to the remaining ester bonds.

Figure 3.

 Summary of fractionation scheme for [feruloyl-14C]arabinoxylans.
The tabulated ‘recovery’ data are approximate values for typical 7-day-old cultures (non-crosslinked arabinoxylans) and typical 13-day-old cultures (heavily cross-linked arabinoxylans) in in-vivo experiments, e.g. those shown in Figure 4.

Figure 4.

 Changes in phenolic moieties of arabinoxylans during cross-linking in vivo.
(a,b) [14C]Cinnamate was fed to 6-day-old maize cultures, then at various ages the soluble extracellular polysaccharides were hydrolysed with mild (a) or severe (b) alkali. Partitioning of 14C between the ‘hydrophilic’ (aqueous) and ‘hydrophobic’ (ethyl acetate) phases is reported.
(c–f) Paper chromatography of hydrophilic products from (a). External markers ran as shown in (f); SF, solvent front.
(g, h) TLC of hydrophobic products from (a). Markers: DFA, 5,5′-diferulic acid; Fer, ferulic acid.
In (c)–(f), each loading is equivalent to 1 ml of culture filtrate. Error bars are 2σ, based on the ratio of counts.

The conclusion that these resistant linkages involved polysaccharides was confirmed by paper chromatography of the hydrophilic fraction (Figure 4c–f). Much of the 14C-labelled hydrophilic material at each time-point was chromatographically immobile (47, 46, 68 and 78%, at 7, 8, 11 and 13 days, respectively), indicating that the phenolic moieties were covalently bound to polysaccharides rather than to mono- or oligosaccharides (which migrate in the chromatography solvent, and would be expected to migrate even more rapidly with phenolics attached; Fry, 2000).

Resisting 0.1 m NaOH for 1–2 days indicated ether-like bonds. The resistance of some of the phenolic–polysaccharide bonds to 6 m NaOH at 37°C for 24 h indicated even greater alkali stability: in the 13-day samples, 6 m NaOH caused more of the 14C-labelled material to be released in hydrophobic form than did 0.1 m NaOH (Figure 4b). As 6 m NaOH does not cause any measurable degradation of arabinoxylan or xyloglucan backbones (Thompson and Fry, 1997), we conclude that the severe alkali had cleaved some of the phenolic–polysaccharide bonds that had resisted the milder alkali. Nevertheless, the 11–13-day-old cultures still yield ∼30% of the radioactivity in the hydrophilic (i.e. polysaccharide-bound) form. We deduce that at least the later formed phenolic–polysaccharide bonds are far more alkali-resistant than esters. Phenolic–polysaccharide ether bonds are the probable explanation.

TLC of the hydrophobic products of mild alkali hydrolysis (Figure 4g) revealed a strong ferulic acid band, except in the 13-day-old cultures, where it disappeared and was largely replaced by chromatographically immobile material, probably comprising oligoferulic acids. Radiolabelled ferulate residues (Figure 4h) changed in parallel with total ferulate residues (Figure 4g). In addition, minor radioactive bands are visible, probably including various isomers of [14C]diferulic acid, and probably also including p-coumaric acid. These either remain unchanged or diminish by 13 days after subculture.

Thus, the dimerization of ferulic acid residues was not primarily responsible for arabinoxylan cross-linking in cell cultures. Instead, the data indicate that the formation of larger coupling products played a major role.

Cross-linking of arabinoxylans in vitro with limited formation of diferuloyl esters

As an independent way of testing the mechanism of arabinoxylan cross-linking, we used the cell-free system. Maize culture filtrate (containing non-crosslinked [feruloyl-14C]arabinoxylans plus naturally occurring peroxidases) was dialysed and thereby freed of low-Mr solutes, including H2O2 (Kärkönen and Fry, 2006) and peroxidase inhibitors (Encina and Fry, 2005). During dialysis, the arabinoxylans remained non-crosslinked (data not shown).

Freshly dialysed culture filtrate had the characteristic absorption spectrum of a feruloyl-carbohydrate ester (Fry, 1982), with a peak at ∼324 nm at pH 4.7 (Figure 5a), undergoing a bathochromic shift to ∼375 nm at pH 10.0 (Figure 5b). Dialysed culture filtrate (pH 4.7) was supplied with 0–16 doses of 6 μm H2O2 (calculated to provide 0–4.8 mol H2O2 per mol feruloyl residues) at hourly intervals. After each successive dose of H2O2, the original absorbance peaks diminished (Figure 5), indicating oxidation of the feruloyl residues. Theoretically, 0.5 mol H2O2 can oxidize 1 mol of feruloyl residues to dimers:

Figure 5.

 Changes in absorption spectra of feruloyl-arabinoxylans during oxidation by the incremental addition of H2O2in vitro.
Dialysed culture filtrate from a 3-day-old culture, containing 31 μm feruloyl residues, was treated with 0–16 doses of H2O2 at hourly intervals to give the final H2O2:ferulate ratios indicated. Absorption spectra were then prepared at pH 4.7 and 10.0. The highest dose of H2O2 (dashed line) gave a slightly turbid solution.

where ΦH2 is a feruloyl residue and HΦ–ΦH is a diferuloyl residue. However, the absorbance continued to fall with further additions of H2O2 up to ∼2 mol per mol ferulate. This indicates the continued consumption of H2O2 well beyond the quantity theoretically required for complete oxidation of ferulate to diferulate. This is explained by further oxidation of the dimers, consuming additional H2O2. In the extreme case, oxidation of all the ferulate to form a single macromolecule could be achieved by 1 mol H2O2 per mol ferulate:


where n is large. As the decrease in absorbance continued even beyond 1 mol H2O2 per mol ferulate, we conclude that additional, uncharacterized oxidation reactions also occurred. When more than 4 mol per mol ferulate was added, the solution became slightly turbid, as shown by the rise in absorbance at all wavelengths (Figure 5, dashed line), indicating some insolubilization.

Size fractionation showed that even a low dose of H2O2 (∼0.1 mol per mol ferulate) caused most of the [14C]arabinoxylans to cross-link (Figure 6b). In similar experiments, [14C]arabinoxylans were treated with various doses of H2O2, and were then alkali-hydrolysed, acidified and partitioned. High doses of H2O2 led to the formation of 14C-labelled products that remained hydrophilic even after treatment with 0.1 m NaOH (Table 2) or 1.0 m NaOH (data not shown), as observed by long-term cross-linking in vivo (Figure 4a,f).

Figure 6.

 Changes in phenolic moieties of arabinoxylans during cross-linking in vitro.
(a–d) Dialysed culture filtrate from 8-day-old cultures, containing endogenous peroxidase and non-cross-linked [14C]arabinoxylan, was treated with multiple 6-μm doses of H2O2 at 10-min intervals to give the final H2O2:ferulate ratios indicated. Cross-linking was monitored by size-fractionation.
(e) Total [14C]arabinoxylans were analysed as in Figures 3 and 4; the hydrophobic fraction is shown. Centre of image, autoradiogram; peripheries of image, external markers stained with I2 vapour: Caff, caffeate; Cinn, cinnamate (detected under ultraviolet); o-Cou, ortho-coumarate; p-Cou, para-coumarate; DFA, diferulate; p-HBA, para-hydroxybenzoate; Sali, salicylate; Sin, sinapate.

Table 2.   Hydrophobicity of radioactive material released by alkali hydrolysis of [feruloyl-14C]arabinoxylans that had been cross-linked in vitro
H2O2 (moles per mole of feruloyl residues)Partitioning of products without alkali treatmentPartitioning of products after 24 h in 0.1 m NaOH at 20°C
Aqueous phaseEthyl acetate phaseAqueous phaseEthyl acetate phase
  1. aNon-crosslinked control sample.

  2. The hydrolysates were acidified before being partitioned between water and ethyl acetate. (Additional data from the experiment shown in Figure 6a–d).


Figure 6e shows chromatography of the hydrophobic fraction obtained after hydrolysis in 1.0 m NaOH. The autoradiogram is typical of five similar experiments. In the 0 H2O2 lane there is a strong [14C]ferulic acid band, which grows fainter with increasing doses of H2O2, and is replaced partly by [14C]diferulates, but mainly by chromatographically immobile [14C]oligoferulates (Figure 6e). In addition, especially at high doses of H2O2, some radioactivity partitions into the aqueous phase (Table 2), and is therefore not represented in Figure 6e. The great majority of this hydrophilic material was immobile on paper chromatography (data not shown; similar to Figure 4f). Repetition of the experiment with 0.1 or 0.3 m NaOH instead of 1.0 m gave similar results (data not shown).

The results show that H2O2-dependent cross-linking of feruloyl-polysaccharides in vitro gives products similar to those observed, presumably with endogenous H2O2, in vivo. Specifically, extensive cross-linking generates ferulate oxidation products that mild alkali cannot release from the polysaccharide as hydrophobic products. Furthermore, the limited quantity of hydrophobic material that alkali does release exhibits chromatographic properties indicative of high-Mr oligoferulates.


Advantages of radiolabelling

The use of radiolabelled arabinoxylans allowed us to monitor quantitatively, and with great sensitivity, both the polysaccharide Mr and the fate of the ferulate ester groups. Furthermore, the behaviour of a given ‘cohort’ of polysaccharide molecules (synthesized, or added, at a known time) was distinguishable from that of the molecules already present and those synthesized later. We used 3H to specifically label the pentose residues of polysaccharides, and used 14C to label the feruloyl side chains. The maize arabinoxylans have roughly one feruloyl side chain per 50 pentose residues (Fry et al., 2000).

Advantages of cell-suspension cultures

Cell cultures have several important advantages over the whole plant for in-vivo isotopic studies of the fundamental processes of apoplastic metabolism: (i) the homogeneity of cell type, with little tissue variation; (ii) the lack of gradients (of nutrients, O2 or exogenous precursors) in the extracellular solution; and (iii) the very large volume and ease of sampling of the apoplast (Lorences and Fry, 1991), which is a cell-wall–culture-medium continuum, into which some wall polysaccharides, enzymes, substrates such as H2O2 and the currently unidentified inhibitor of cross-linking (Encina and Fry, 2005), will diffuse, whereas they would have been confined to the cell wall and any non-air-filled extracellular spaces in planta.

Soluble extracellular arabinoxylans differ from wall-bound arabinoxylans in having greater mobility, greater conformational freedom, greater accessibility to enzymes and probably a greater degree of solvation with water molecules, instead of hydrogen bonding with other polymers. The first three might be expected to increase the degree of ferulate polymerization, and therefore raise the trimer/dimer ratio. Dimers are only faintly seen as a TLC band in (Figure 4g, h). Despite these differences, soluble extracellular arabinoxylans have the great advantage of being a much easier target for experimentation than wall-bound polysaccharides, which require harsh chemical treatments to solubilize them: treatments that would destroy many of the chemical bonds presently under study (Fry et al., 2000).

Physiological control of arabinoxylan cross-linking

Arabinoxylan cross-linking is relatively sudden, and rapid, mainly occurring within a 24-h period between two time-points of a feeding study (Figure 1). This behaviour is characteristic of a biological process dependent on the controlled synthesis, secretion and action of enzymes. The cross-linking of exogenous radiolabelled arabinoxylans while remaining in solution strongly suggests that the enzymes responsible are only loosely bound to the cell wall. We did not observe any abrupt increases in peroxidase activity (data not shown), suggesting that it is the action of the enzyme rather than its activity (see Fry, 2004) that determines the rate of arabinoxylan cross-linking. Enzyme action can be promoted by the controlled release of a co-factor or short-lived co-reactant, or by the removal of an inhibitor. As the cross-linking can be greatly promoted in cell-free culture filtrates by the addition of H2O2 (Geissmann and Neukom, 1971 and Schooneveld-Bergmans et al., 1999), we propose that the cross-linking enzyme is a peroxidase, and that it is enabled to act in vivo during a transient period characterized by rapid H2O2 synthesis, and the absence of low-Mr inhibitors of the type described by Encina and Fry (2005).

Observed behaviour of ferulic acid residues during
arabinoxylan cross-linking

Maize arabinoxylans are feruloylated polysaccharides, from which ferulic acid (recovered in our analyses as a ‘hydrophobic’ product; Figure 3) can be released by mild alkali. Many feruloylated polysaccharides can be cross-linked by peroxidase plus H2O2, and it has been widely speculated that this cross-linking results from the formation of diferulate diester bridges between polysaccharide molecules (e.g. Figure 7, reaction a). Diferulates are indeed released from the cell walls of maize, and many other poalean and caryophyllalean plants, by mild alkali (Fry et al., 2000; Fry, 2004; Levigne et al., 2004; SJB and SCF, unpublished data). When [feruloyl-14C]arabinoxylans form diferuloyl ester bridges, we would expect the 14C to still be releasable as hydrophobic products ([14C]diferulates) by treatment with mild alkali.

Figure 7.

 The proposed two-step reaction of feruloylated arabinoxylans that forms ether bonds.
(a) Peroxidase-catalysed oxidative coupling to form a diferulate group, in some cases with a quinone methide structure; (b) nucleophilic attack by a sugar residue (grey box; an α-l-arabinofuranose residue is arbitrarily shown), forming a benzyl–sugar ether bond. The product (structure 5) may then undergo repeat reactions similar to (a) and (b). The molecules illustrated are intended as examples, not specific proven structures: structures 1, 2 and 6, feruloyl-arabinoxylan esters (wavy line = arabinoxylan chain); structures 3 and 7, quinone methides; structures 4 and 8, sugar residues; structures 5 and 9, proposed products of nucleophilic attack.
Inset: reaction catalysed by carboxymethyloxysuccinate lyase (EC, which is chemically comparable to reaction (b).

The radioactivity from non-cross-linked [14C]arabinoxylans was indeed largely released in hydrophobic form (mainly [14C]ferulic acid) by mild alkali. However, cross-linking coincided with the disappearance of alkali-releasable [14C]ferulate, both in the [14C]cinnamate feeding study in vivo (Figure 4) and in the cell-free system in vitro (Figure 6). There was little net formation of [14C]diferulate; often, esterified [14C]diferulates decreased simultaneously with the [14C]ferulate (Figures 4 and 6). Figure 4h differs from Figure 6e in that it shows only the beginning (at 11 days) and end (at 13 days) of a cross-linking process. Figure 6e shows a gradual cross-linking process artificially controlled by the incremental addition of H2O2. If we had taken samples at hourly intervals starting at 11 days, the in-vivo experiment might have given a pattern similar to that in Figure 6e, with the dimer band appearing and disappearing as the ferulate band decreased and the origin band increased, respectively. Unfortunately, we could not accurately predict the onset of cross-linking in a maize cell culture.

Nevertheless, even after prolonged cross-linking in vivo (Figure 4a) and after cross-linking with the highest dose of H2O2 tested in vitro (Table 2), ∼40% of the 14C-labelled material remained in a form that yielded hydrophobic products upon mild alkali hydrolysis (summarized in Figure 3). TLC of this material showed it to have RF (chromotographic mobility relative to solvent front) ≈ 0.0 (Figures 4h and 6e). As this material had partitioned into ethyl acetate, it cannot have been still attached to polysaccharides; its immobility on TLC thus indicates that it comprised oligoferulates, which are high-Mr oxidative phenolic coupling products (Fry et al., 2000; Rouau et al., 2003; Bunzel et al., 2004, 2005, 2006; Funk et al., 2005). This agrees with the conclusion of Fry et al., (2000) that ferulate trimers and larger coupling products exceed diferulates in maize cell walls. However, it is also true that diferuloyl polysaccharides are less likely to be sloughed into the medium than feruloyl polysaccharides (Lindsay and Fry, 2008).

The 14C-labelled material that alkali failed to deliver into the hydrophobic fraction turned out to be mainly of very high Mr: the material was immobile on paper chromatography in a system known for resolving nonasaccharides, and sometimes also larger oligosaccharides (Fry, 2000), especially if carrying phenolic side chains (Wende and Fry, 1996). This 14C-labelled material thus comprised [14C]ferulate derivatives linked by mild-alkali-stable bonds (i.e. not esters) to hydrophilic polysaccharides.

More severe alkali hydrolysis (6 m NaOH; Figure 4b) of cross-linked [14C]arabinoxylans delivered ∼70% of the 14C into the hydrophobic fraction (i.e. released it from its bonds to polysaccharides; Figure 4b), rather than the 40% observed with mild alkali (Figure 4a). This indicates that ∼30% of the 14C-labelled material present after cross-linking was in the form of hydrophobic oligoferulates attached to polysaccharides via ether-like bonds that can be cleaved by 6 m NaOH (and likewise by 1 m NaOH; Table 1), but is not appreciably cleaved by mild alkali. For example, Enoki et al. (1983) reported that a model p-hydroxybenzyl–glucose ether was 90% hydrolysed by treatment for 40 h in 1 m NaOH at 28°C. The arabinoxylans are not themselves appreciably labile in aqueous alkali; indeed, 6 m NaOH is conventionally used to solubilize hemicelluloses.

The maize cells studied in the present work were not lignified. In contrast, heavily lignified cell walls from mature wheat and oat internodes contain phenyl–benzyl ether bonds between ferulate and lignin. Such ferulate groups are ester-bonded via the −COOH group to polysaccharide sugar residues, and ether-bonded via the ‘phenolic’ oxygen to a benzyl position in lignin (Lam et al., 2001). The ester bond of such a ferulate moiety lacking a free phenolic −OH group was cleaved by 0.5 m NaOH at room temperature (Iiyama et al., 1990), whereas the benzyl-phenyl ether required 4 m NaOH at 170°C (Lam et al., 2001). Such highly alkali-resistant ether bonds cannot play a role in retaining ferulate within the walls of our cultured maize cells, which are not lignified.

Proposed mechanism of formation of mild-alkali-stable cross-links

Arabinoxylans freshly cross-linked by a single dose of dilute H2O2 could be converted back to the original molecular weight by mild alkali (0.1 m NaOH at room temperature; Figure 2biv). However, repeated small doses of H2O2 resulted in the arabinoxylans remaining stably cross-linked, even after prolonged treatments with mild alkali (Figure 1b and 2b.vi). These mild alkali treatments readily cleave feruloyl ester bonds: for example, the half-life of feruloyl-disaccharides in ∼0.07 m NaOH at 22°C is 6 mins (Fry, 1982). It is therefore very unlikely that diferuloyl or oligoferuloyl ester bonds would survive to an appreciable extent under the conditions used here (e.g. 0.1 m NaOH for 24 h). Other workers who used 2 m NaOH to release dimers and oligomers were working with lignified bran tissue, in which the more concentrated alkali may have helped to open up the wall structure. Similarly, we observed that arabinoxylans cross-linked in vivo by maize cell cultures varied in resistance to mild alkali, and we suggest that the development of alkali resistance depends on the physiological condition of the culture.

The most plausible explanation for the acquisition of alkali resistance is the formation of a quinone methide (Figure 7, structure 3), rapidly followed by nucleophilic attack by a sugar residue (structure 4), creating a p-hydroxybenzyl ether bond (structure 5). Analogous benzyl–sugar ether bonds, formed via quinone methide intermediates, have been reported in lignin structures (Yaku et al., 1981; Takahashi and Koshijima, 1988; Barakat et al., 2007). Compared with the two ester bonds (COO−) of structure 5, the benzyl ether bond is expected to be more resistant to mild alkali, albeit more alkali-labile (as in structure VIII of Enoki et al., 1983) than most other types of ether bond.

Structure 5 could comprise a mixture of four diastereomers, if formed non-enzymically (Tokimatsu et al., 1996). However, it is also possible that reaction (b) is catalysed by a novel carbon–oxygen lyase that is comparable with the Zoogloea enzyme, carboxymethyloxysuccinate lyase (EC; Peterson and Llaneza, 1974), with the reaction running in the ‘synthase’ direction (Figure 7, inset). The specificity of such a hypothetical enzyme could account for the efficient ether bonding of the quinone methide (structure 3) to a polysaccharide sugar residue, to form a high-Mr benzyl ether, in preference to some competing nucleophile such as water, a uronic acid residue or the sucrose in the culture medium (forming a benzyl alcohol, benzyl ester or low-Mr benzyl ether, respectively). It is possible that the last of these alternatives – ether bonding to sucrose – does occur to some extent, accounting for the ∼12% of the 14C in cross-linked arabinoxylans that is found to be hydrophilic but mobile in paper chromatography after hydrolysis of the ester bonds (Figures 3 and 4c–f).

If the polymeric nucleophiles (structures 4 and 8 in Figure 7) are separate chains, not previously part of the quinone-methide-bearing polysaccharides (3 and 7), then we would predict that a non-feruloylated polysaccharide could serve this role, forming a benzyl ether bond with the dimerized ferulate group. However, we did not find evidence for this possibility: deferuloylated [3H]arabinoxylans did not become incorporated in a cross-linked network of non-radioactive feruloyl-arabinoxylans being formed in the same reaction mixture (Figure 2ciii). One explanation could be that the quinone methide (e.g. structure 3) forms only an intramolecular benzyl ether bond with a neighbouring sugar residue within the same polysaccharide chain; in other words, structure 4 is originally part of structure 3.

By this mechanism, a mild-alkali-resistant p-hydroxybenzyl–polysaccharide ether bond (structure 5) could be produced (Figure 7). This would explain the observed failure of some of the radioactivity from cross-linked [feruloyl-14C]arabinoxylans to be released in hydrophobic form by mild alkali.

The single p-hydroxybenzyl–polysaccharide ether bond in structure 5 would not result in an alkali-stable cross-link between polysaccharides, as the other two bonds with polysaccharides shown in structure 5 are esters. To form a mild-alkali-stable interpolysaccharide cross-link, the process would need to be repeated: structure 5 could react with a third feruloyl-polysaccharide molecule (structure 6) to form a new quinone methide (structure 7), which would be subject to nucleophilic attack by an additional sugar residue (structure 8). The product (structure 9) would now contain two polysaccharide chains (those shown in grey boxes; Figure 7) bonded to each other via two mild-alkali-stable benzyl ether bonds. Of these two ether bonds, the newly formed one could resemble the p-hydroxybenzyl ether bond in structure (structure 5), whereas the pre-existing one could (depending on the position to which structure 6 had become oxidatively coupled) become a p-alkoxybenzyl ether bond, which would be somewhat more alkali resistant (compare with structure VIIIa of Enoki et al., 1983) than a p-hydroxybenzyl ether bond. Detailed chemical characterization of the mild-alkali-resistant bonds will require the isolation of the oligomeric fragments containing them.


In conclusion, the cross-linking of arabinoxylans occurs at the same time as the formation of esterified oligoferulates, and slightly earlier than the formation of ether-like phenol–polysaccharide bonds. Our observations provide evidence that the in-vivo oxidation of the feruloyl ester side-chains of arabinoxylans enables the formation of cross-links between the polysaccharides, but that this principally occurs via oligoferulate bridges (rather than via the widely discussed diferulates), which are subsequently reinforced by ether-like bonds. It remains to be established how closely the behaviour of soluble extracellular arabinoxylans resembles that of the less mobile wall-bound arabinoxylans. Nevertheless, the present results provide novel insight into the types of reaction, and their in-vivo kinetics, that can be orchestrated by living maize cells. There could potentially be differences between cell cultures and intact plants; however, we consider that the maize cell cultures are unlikely (in the short time that they have existed) to have evolved the novel machinery needed for any biochemical processes drastically unrelated to those occurring in planta. It follows that analyses revealing only ester-linked products, and especially those revealing only ester-linked dimers, will seriously underestimate the true extent of the polysaccharide cross-links that contribute to wall extensibility and digestibility, as well as to cell–cell adhesion.

Experimental procedures

Cell culture

Maize (Zea mays L., Black Mexican sweetcorn) cell-suspension cultures were grown under constant dim light on an orbital shaker at 25°C. Cells were subcultured fortnightly into 200 ml of fresh medium [0.47% (w/v) MS basal inorganic medium (M5519; Sigma-Aldrich, http://www.sigmaaldrich.com), 2% (w/v) sucrose and 2 mg l−1 2,4-dichlorophenoxyacetic acid, pH 4.6–4.8], in 500-ml conical flasks.


To synthesize E-[U-14C]cinnamate, we dissolved 185 kBq l-[U-14C]phenylalanine (17.1 MBq μmol−1; GE Healthcare, http://www.gehealthcare.com) in 100 μl of 100 mm Tris (acetate, pH 8.7) containing 5 mm 2-mercaptoethanol, added yeast phenylalanine ammonia lyase (0.005 U; Sigma-Aldrich) and incubated the mixture at room temperature until essentially all phenylalanine had been converted to cinnamate (monitored by TLC). The mixture was then acidified with HCl and the [U-14C]cinnamic acid was extracted into ethyl acetate, dried and stored as an aqueous solution at −20°C.

l-[1-3H]Arabinose (148 MBq μmol−1) was custom-synthesized by GE Healthcare and purified in Edinburgh by preparative paper chromatography.

Radiochemicals were sterilized either by dissolving in ethanol and drying in the culture vessel before the addition of the cells, or by filter sterilization.

Radiolabelling of arabinoxylans

Maize cell-suspension cultures (5 ml) were incubated with sterile [3H]arabinose (50 kBq) or [14C]cinnamate (33 kBq), starting 6 days after subculture. At various intervals, cultures were harvested and cells removed by filtration through nylon gauze. The cell-free culture filtrate containing radiolabelled polysaccharides, principally arabinoxylans, was dialysed against 0.5% chlorobutanol for 24 h, and then size fractionated.

For the bulk radiolabelling of soluble extracellular arabinoxylans, to be fed back to other cultures or to be used in cross-linking assays in vitro, the procedure was as above, except that the [3H]arabinose or [14C]cinnamate feeding was always for 48 h, and the dialysis was against ice-cold deionized water.

Size fractionation

Components of cell-free culture filtrates were size fractionated by gel-permeation chromatography on a 10-ml bed volume Sepharose CL-2B column, with pyridine:acetic acid:0.5% aqueous chlorobutanol at 1:1:98 (pH 4.7) as the eluent, and blue dextran plus glucose as markers. Blue dextran (mean Mr ≈ 2 × 106) is not completely excluded by the gel, but a small subfraction of blue dextran does elute as a peak indicating Kav 0. Non-radioactive glucose was taken to indicate Kav 1. Elution of radioactive material was monitored by liquid scintillation counting in ScintiSafe 3 (formerly OptiPhase HiSafe 3; BDH, http://www.fisher.co.uk).

Cross-linking of exogenous [pentosyl-3H]arabinoxylans by living maize cell cultures

In ‘cumulative cross-linking’ experiments, 0.75 kBq of [3H]arabinoxylans solution was dried in vacuo, washed in 95% ethanol and allowed to dry again aseptically in the culture vessel before the addition of 5 ml of 6-day-old maize cell culture. The cultures were then incubated for varying times between 4 h and 7 days.

In the ‘current cross-linking’ experiments, a similar procedure was followed, except that the [3H]arabinoxylan solution was dried into a culture vessel, and not treated with ethanol. A 5-ml portion of maize cell culture (age 6–13 days) was then added and incubated for an additional 8 h.

At the end of their respective incubation periods, cultures were filtered on nylon gauze and the culture filtrate was subjected to gel-permeation chromatography, as before.

Cross-linking of arabinoxylans in a cell-free system

Non-cross-linked [3H]arabinoxylans (1.68 kBq ml−1) or [14C]arabinoxylans (0.2 kBq ml−1), present in 90 μl of dialysed 8-day-old culture filtrate, were supplemented with 100 mm acetate (Na+, pH 4.7). The initial concentration of feruloyl groups was estimated from the molar absorbance at 373 nm of an aliquot adjusted to pH 10.0 (ε373 = 28 800 m−1 cm−1; Fry, 1982). Similar aliquots were incubated in the presence or absence of H2O2 (various concentrations, added either all at once or in small quantities at the time intervals indicated). Reactions were stopped by the addition of dithiothreitol to a concentration of 10 mm. Alkali hydrolysis was carried out by incubation for 24 h at 20°C or 37°C in the presence of 0.1–1.0 m NaOH.

In some experiments, [3H]arabinoxylans were de-esterified before cross-linking was attempted. The 8-day culture filtrate, obtained as above, was treated with 0.1 m NaOH for 24 h at 20°C, which hydrolyses off the feruloyl ester groups, and was then adjusted to pH 4.7 with acetic acid and dissolved in dialysed (NaOH-untreated) 8-day culture filtrate. The solution was either not further treated or supplied with one dose of H2O2 (to a final concentration of 5.9 μm) with or without 8-day-old NaOH-untreated culture filtrate (to 50%, v/v). The mixtures were incubated at 20°C for 3 h, after which the [3H]polysaccharides were size-fractionated on Sepharose CL-2B.

Fractionation of products formed from [14C]arabinoxylans

Culture filtrate containing [14C]arabinoxylans was alkali-hydrolysed with NaOH (usually 0.1 or 6.0 m) at 20°C or 37°C for 24 h, and then mildly acidified with TFA. The solutions were shaken three times with an equal volume of ethyl acetate, and the two phases were then separated.

The ethyl acetate phase (‘hydrophobic products’) was dried, redissolved in propan-2-ol and subjected to TLC on silica gel (with fluorescent indicator; Merck, http://www.merck-chemicals.com) in benzene:acetic acid (9:1). The plate was exposed to a 360-nm UV lamp during development, so that compounds such as ferulic acid were maintained as an equilibrium mixture of E- and Z-isomers, and thus as a single spot (Fry, 1983). Aromatic compounds were visualized by their ability to absorb 254-nm UV radiation, and thus form dark spots against the fluorescent background of the TLC plate, and by autoradiography.

The aqueous phase was chromatographed on Whatman 3Chr paper in ethyl acetate:acetic acid:water (10:5:6), and developed by the descending method for ∼16 h. Phenolic markers were detected under UV, and sugar markers were stained with aniline hydrogen phthalate (Fry, 2000). Radioactivity was located by scintillation counting of paper strips in OptiScint HiSafe 3.


We thank Mrs Janice Miller for excellent technical assistance and the BBSRC (UK) for their financial support of this work. We are very grateful to Drs John Ralph and Graham Wallace for providing samples of 8,5′- and 5,5′-diferulic acid, respectively.