Naturally p-Hydroxybenzoylated Lignins in Palms

The industrial production of palm oil concurrently generates a substantial amount of empty fruit bunch (EFB) fibers that could be used as a feedstock in a lignocellulose-based biorefinery. Lignin byproducts generated by this process may offer opportunities for the isolation of value-added products, such as p-hydroxybenzoate (pBz), to help offset operating costs. Analysis of the EFB lignin by nuclear magnetic resonance (NMR) spectroscopy clearly revealed the presence of bound acetate and pBz, with saponification revealing that 1.1 wt% of the EFB was pBz; with a lignin content of 22.7 %, 4.8 % of the lignin is pBz that can be obtained as a pure component for use as a chemical feedstock. Analysis of EFB lignin by NMR and derivatization followed by reductive cleavage (DFRC) showed that pBz selectively acylates the γ-hydroxyl group of S units. This selectivity suggests that pBz, analogously with acetate in kenaf, p-coumarate in grasses, and ferulate in a transgenic poplar augmented with a feruloyl-CoA monolignol transferase (FMT), is incorporated into the growing lignin chain via its γ-p-hydroxybenzoylated monolignol conjugate. Involvement of such conjugates in palm lignification is proven by the observation of novel p-hydroxybenzoylated non-resinol β–β-coupled units in the lignins. Together, the data implicate the existence of p-hydroxybenzoyl-CoA:monolignol transferases that are involved in lignification in the various willows (Salix spp.), poplars and aspen (Populus spp., family Salicaceae), and palms (family Arecaceae) that have p-hydroxybenzoylated lignins. Even without enhancing the levels by breeding or genetic engineering, current palm oil EFB ‘wastes’ should be able to generate a sizeable stream of p-hydroxybenzoic acid that offers opportunities for the development of value-added products derived from the oil palm industry.


Introduction
In order for a lignocellulose biorefinery to function as an economically viable enterprise, it must be developed to yield value-added coproducts in the process [1,2]. Coupling the source of biomass to the waste stream of an associated commercial process is one way to create value in the overall system. The production of palm oil generates a substantial amount of empty fruit bunch (EFB) fibers that can be readily processed into sugars for biofuels production and a mixture of unfermentable components including lignin. The resulting lignin stream could then offer new opportunities for the development and isolation of value-added products derived from the oil palm industry [3]. For example, in 2013 Malaysia, the second largest oil palm producer after Indonesia, produced over 19 million tons of palm oil on a total plantation area of over 5 million ha, which also generated an estimated 70-80 million tons of biomass, mostly from oil palm EFB [4]. This biomass, which is a high-quality lignocellulosic fiber, has not been fully utilized commercially and is a largely untapped resource, although research in the biorefinery area has attempted to provide high value-added products via liquefaction [5], solvolysis [6] and pyrolysis [7].
Lignin, a stochastically generated biopolymer found in plants, forms by means of radical coupling reactions, primarily between a monolignol and the growing lignin polymer in an endwise process [8,9]. In angiosperms, coniferyl and sinapyl alcohols are the predominant monomers and, once incorporated into the lignin backbone, produce guaiacyl (G) and syringyl (S) units in the resulting polymer. A monolignol couples overwhelmingly at its β-position to the growing polymer at available 4-O-(G or S) or 5-(G only) positions. Branching is introduced by 5-O-4-or 5-5-coupling reactions between preformed oligomers, with the latter occurring only between G units and the former involving at least one G unit. Evidence continues to accumulate, particularly from studies of the lignin biosynthetic pathway using mutant and transgenic plants, that lignins can incorporate monomers beyond the traditional three monolignols. As reviewed previously [8][9][10][11], these include (but are not limited to) the novel caffeyl and 5-hydroxyconiferyl alcohols from incomplete methylation in monolignol biosynthesis (OMT-deficient plants); the immediate hydroxycinnamaldehyde monolignol precursors, coniferaldehyde and sinapaldehyde, and the hydroxy-benzaldehydes (vanillin and syringaldehyde) derived from them from incomplete reduction (CAD deficiency); and products such as dihydroconiferyl alcohol and its derived guaiacylpropane-1,3-diol in wild-type and mutant softwoods.
Relatively little is known about the nature of p-hydroxybenzoylated lignins or how they arise. From oil palm trunk, nitrobenzene oxidation produced high levels of p-hydroxybenzoic acid (9.8-14.9 %) [37], which was also the main product of alkaline hydrolysis of parenchyma milled wood lignin (MWL) [31]. Small quantities of vanillic and syringic acids were also found. The authors concluded that the phenolic acids acylated the side-chain alcohol groups of lignin subunits. High levels of p-hydroxybenzoic acid were also released from oil palm lignins [42][43][44][45][46]34]. The 13 C-NMR chemical shifts in Salix (willow) MWL samples showed good concordance with expected shifts for free-phenolic γ-phydroxybenzoate esters based on model compound studies [28]. Rather than suspecting that the lignin side-chains were acylated after the polymer had already formed, p-hydroxybenzoyl monolignol conjugates were proposed as the precursors to these structures in early research [47]; monolignol acylation was further conjectured [25] and eventually proven by the observation of novel β-β-coupling products in lignins that could only have arisen from pre-acylated monolignols [29,14,48,32]. Recent studies have delivered the transferases p-coumaroyl-CoA:monolignol transferase (PMT) in grasses responsible for lignin p-coumarate esters via p-coumaroylation of monolignols [22,23,49,50] and now an exotic feruloyl-CoA:monolignol transferase (FMT) that feruloylates monolignols prior to lignification [27]; the putative p-hydroxybenzoyl-CoA and acetyl-CoA monolignol transferases have not yet been identified.
The research presented in this paper seeks to answer several key questions regarding the intricacies of acylation of lignins by p-hydroxybenzoate, specifically: whether p-hydroxybenzoates exist as free-phenolic appendages on lignins, akin to the p-coumarates; whether they, like p-coumarates [51,52,23,49] and acetates [18] in most plants, primarily acylate S units; whether the various lignin substructures carry these units and thus implicate the incorporation of preformed monolignol γ-p-hydroxybenzoate conjugates into the lignification process; and whether diagnostic β-β-coupled units can be found to provide proof that lignification utilizes preformed monolignol p-hydroxybenzoate conjugates, thus, also establishing that monolignol p-hydroxybenzoate conjugates are lignin precursors. Although a survey across all palms is beyond the scope of this work, a sago palm lignin previously examined by analytical pyrolysis [35] and prior studies on coconut (Cocos nucifera) coir fibers [36] are used to suggest a degree of generality in palms. Together, such studies will help to elucidate the roles of putative genes and enzymes associated with the required transferase activity, knowledge that may clarify the poorly understood reasons for widespread lignin acylation in the plant kingdom. At the same time, it will also shed light on future paths that may be worth exploring for value-added products making lignocellulosic biofuels a more viable option; depending on the level of p-hydroxybenzoates found in a given biomass material, and the ease with which it can be cleaved and recovered from the polymer, p-hydroxybenzoic acid or its derivatives may be accessible value-added products [34,31]. A simple net search shows that p-hydroxybenzoic acid currently sells for~$US4,000/ton and has uses: as a source of esters that have antibacterial and antifungal properties, as an intermediate in pesticides and antiseptics, and for preparation of co-polyesters (including liquid crystal polymers).

General
Commercial chemicals, including solvents, were of reagent grade or better and used without further purification. Analytical thin-layer chromatography (TLC) was carried out on EM Science TLC plates pre-coated with silica gel 60F 254 ; TLC visualization was by UV. Purification by flash column chromatography was accomplished on silica gel (230-450 mesh). 4-Acetoxybenzoyl chloride pBz Cl was prepared according to a previous procedure [53]. Nuclear magnetic resonance (NMR) data and assignments for all model compounds are from welldescribed 1D and 2D NMR experiments (on a Bruker Biospin Avance 500 MHz spectrometer). High-resolution mass spectrometric (HRMS) data were acquired using electrospray ionization (ESI) on a Waters (Micromass) LCT ESI/TOF MS instrument. UV-vis analysis performed in 1-cm quartz cuvettes at λ=280 nm on a Shimadzu UV-1800.

Plant Materials
Palm Samples The African oil palm (Elaeis guineensis) fractions used in this study came from two sources. The first sample, consisting of oil palm fronds and empty fruit bunch (EFB) fibers, was supplied by the Forest Research Institute of Malaysia and is the same sample used in a previous study by Sun et al. [45] and will be denoted here as EFB-1. The second sample of oil palm EFB fibers was provided by the Ladang Tai Tak Sdn. Bhd. Palm oil mill in Kota Tinggi, Johor, Malaysia, and will be denoted here as EFB-2. These EFB fibers were steam-treated (45 psi, 130°C, 70 min) to remove the fruitlets and then shredded into fibers (4-6 in in length). Various data on both oil palm EFB samples are presented in Table 1. Isolated sago palm (Metroxylon sagu) lignin was the material previously described in a pyrolysis study [35].

Sample Preparation and Lignin Isolation
All fibrous oil palm samples were dried at 50°C in an oven for at least 72 h before being ground. Two-step grinding was performed in a Wiley mill using a 2-mm followed by a 1mm mesh. Removal of extractives from the samples was carried out by successive extractions with water, ethanol, acetone, and chloroform. Extractions were accomplished by refluxing the material in Soxhlet equipment for at least 8 h with each solvent. Prior to analysis, extracted samples, frequently referred to as whole-cell-wall preparations, were dried for 2 weeks in a desiccator under vacuum using P 2 O 5 as drying agent, which was replaced regularly. The dried materials were ball-milled using a stainless steel vibratory ball mill for 1.5 h (with 0.5 h on, 0.5 h off, to minimize heating). The ballmilled materials (50 g) were submitted to enzymatic digestion with crude cellulases (Cellulysin, Calbiochem) for polysaccharide degradation [25,54,55]. As with aspen and poplar, the enzymatic treatment was particularly efficient at removing polysaccharides, leaving just 22.2 and 20.9 % of the cell wall from the EFB and frond samples, respectively. Samples of the enzyme lignins obtained (10 g) were then extracted with dioxane and water (96: 4) in complete darkness at room temperature for 2 days [56]. The soluble fractions were obtained by filtration and lyophilized overnight. The isolated lignins were washed extensively with water and 6 mM EDTA aqueous solution (pH 8). This removed as much of the remaining low-molecular-weight saccharides and metal ions as possible [25]. The washed Björkman lignins were lyophilized again overnight and stored. The yields of isolated EFB lignin used in a previous study by Sun et al. [45] were 19.9 % of the original cell wall material. As seen in Table 1, the AcBr lignin contents, measured with ε 280 =17.9, as previously described [57,58], of the EFB-1 fibers were 22.7±0.4 %, giving a MWL yield on a lignin basis of~88 %. The pBz content of the MWL was 4.2±0.1 %. For the EFB-2 sample, the AcBr lignin content was 18.3±0.2 %.
Isolated lignins were acetylated using acetic anhydride and pyridine as previously described [59]. Solutions of the acetylated lignins in EtOAc were washed with 6 mM EDTA to more rigorously remove metal ion contaminants following methods described previously [25].

NMR of Lignins
Spectra have been run over many years on a variety of Bruker instruments, including DP 360 and Avance 400, 500 and 600 MHz instruments, using standard experiments and conditions [13]. The 3D total correlation spectroscopy-heteronuclear single-quantum coherence (TOCSY-HSQC) experiment was run as described previously [60,55].
Derivatization Followed by Reductive Cleavage (DFRC) Method Adapted for the Analysis of Monolignol p-Hydroxybenzoate Conjugates The derivatization followed by reductive cleavage (DFRC) method [61,62] was adapted for HPLC analysis, as the monolignol p-hydroxybenzoate conjugates were found to be incompatible with GC-MS analysis. The modified procedure is detailed below.
In a 2-dram vial equipped with a polytetrafluoroethylene (PTFE) pressure-release cap and containing a stir bar was suspended each extract-free EFB whole-cell-wall sample (50 mg) in a 20 % solution of acetyl bromide in acetic acid (5 mL). The suspension was heated at 50°C and allowed to stir for 3 h. The solvent was removed on a SpeedVac (Thermo Scientific SPD131DDA, 50°C, 35 min, 1.0 Torr, 40 Torr/min, RC lamp on). The crude film was suspended in absolute ethanol (1 mL), and the ethanol was then removed on the SpeedVac (50°C, 15 min, 6.0 Torr, 30 Torr/min, RC lamp on). To the residue was added a mixture of dioxane:acetic acid:water (5/4/1v/v, 5 mL) and nano-powder zinc (250 mg). The sample was sonicated to ensure that the biomass film was dissolved and the zinc was in suspension. The reaction was stirred in the sealed vial in the dark at room temperature for 16-20 h. Additional nano-powder zinc was added to the vial as required to maintain a fine suspension. The reaction was quenched with a mixture of dichloromethane (DCM, 6 mL), saturated ammonium chloride (15 mL), and internal standard (ISTD, diethyl 5-5′-diferulate diacetate, DEDF, 54.0 μg). After isolating the organic fraction using a separatory funnel, the aqueous phase was extracted with DCM (3×10 mL). The combined organic fractions were dried over anhydrous sodium sulfate, filtered, and the solvent evaporated in vacuo. The free hydroxyl groups were acetylated by adding pyridine/acetic anhydride (1/1v/v, 5 mL) and allowing the mixture to sit in the To remove most of the polysaccharide-derived products, the crude DFRC product was dissolved in ethyl acetate (EtOAc, 0.15 mL), diluted with hexanes (0.15 mL), and loaded onto an SPE cartridge (Supelco Supelclean LC-Si SPE tube, 3 mL, P/N: 505048). The products were eluted with hexanes:ethyl acetate (1:1, 8 mL), and the solvent was removed on a rotary evaporator. The product was dissolved in ethyl acetate and filtered through a 0.2-μm PTFE filter into a high-performance liquid chromatography (HPLC) vial. Analysis of the purified products was performed on an HPLC-PDA-ESI-MS (Shimadzu LC-MS 2020) with an amino stationary-phase column (Phemonenex Luna NH 2 , 100 Å, 250 mm×4.60 mm×5 μm). The mobile phase was hexanes and EtOAc with 0.1 % triethylamine, running at 0.7 mL/min, holding at 30 % EtOAc for 1 min then ramping at 1.2 %/min up to 95 % EtOAc. The column eluent was split 80:20 between a PDA detector (265-400 nm) and an ESI-MS detector operating in positive-ion mode. External synthetic standards of the predominantly trans-isomers were used for determination of the retention time, calibration of the PDA response, and authentication of each compound by MS, Table 2. The DFRC data is also summarized in Table 1.

Syntheses of γ-p-Hydroxybenzoylated β-Ether Model Compounds
As we determined by NMR and DFRC that p-hydroxybenzoates acylated essentially only S units, the prime model compounds of interest were the etherified and the free-phenolic S β-ether γ-phydroxybenzoate models 7 and their peracetates 8. These model compounds were prepared following the synthetic route shown in Scheme 1.
Preparation of 4-Acetoxy-3,5-dimethoxyacetophenone 1 To a solution of the starting acetophenone (3,5-dimethoxy-4-hydroxy-acetophenone, 51.0 mmol, 10.0 g) in pyridine (30 mL) was added Ac 2 O (30 mL). The solution was allowed to stir overnight at room temperature. Upon confirmation by TLC that the starting material had been fully consumed, the reaction mixture was diluted with EtOAc (300 mL) and washed successively with 1 M HCl (3 × 150 mL), sat. NaHCO 3 (1×150 mL), and sat. NaCl (1×100 mL). The organic layer was dried over Na 2 SO 4 and concentrated in vacuo. Residual AcOH was removed by azeotropic distillation with toluene (3×50 mL), followed by removal of the toluene with MeOH co-distillation (2×50 mL). Recrystallization from 20 % EtOAc in Et 2 O afforded 1 (11.0 g, 91 %) as a pale yellow crystalline solid: 1  Bromination of compound 1 (preparation of compound 2) Compound 1 (46.0 mmol, 11.0 g) was dissolved in EtOAc (460 mL), and the reaction solution was cooled to 0°C. Pyridinium tribromide (55.4 mmol, 20.0 g) was charged into the reaction flask incrementally over 10 min, and the reaction was allowed to stir for 2 h, with the temperature gradually warming from 0°C to room temperature (23°C). If any starting material remained detectable by TLC at that point, additional portions of pyridinium tribromide (in increments of approximately 3 mmol, or 1 g) were added until TLC analysis showed full conversion. The reaction was then quenched with a solution of sat. NaHCO 3 (400 mL) and, after separation, the aqueous layer was extracted with DCM (1×100 mL). The combined organic layers were then washed with 1 M HCl Table 2 HPLC-PDI-ESI-MS elution parameters and observed mass-ion-ratios for parent or fragment ions from synthetic standards CA, SA, CA-pBz, SA-pBz, and internal recovery standard, DEDF (see Fig. 4   Procedure for 4-acetoxybenzoylation of the γ-OH in the β-O-4-linked dimers 4 (illustrated for preparation of compound 5 SG-pBz ) Alcohol 4 SG (1.3 mmol, 508 mg) was dissolved in dichloromethane (13 mL), and the resulting solution was cooled to 0°C. DMAP (0.33 mmol, 40 mg) and Et 3 N (3.9 mmol, 0.54 mL) were charged into the reaction mixture, followed by 4-acetoxybenzoyl chloride (pBzCl; 1.6 mmol, 318 mg). The reaction was allowed to warm to room temperature gradually with continued stirring for approximately 2 h. Once TLC showed no detectable starting material remaining, the mixture was poured into 1 M HCl (50 mL). After separation, the aqueous layer was extracted with DCM (1×30 mL), and the combined organics were washed with sat. NaCl (1× 50 mL). They were then dried over Na 2 SO 4 and concentrated in vacuo. Purification by flash column chromatography (50 g SiO 2 , gradient elution from 5 to 50 % EtOAc/hexane) gave ester 5 SG-pBz (547 mg, 76 %) as a white foam: 1  Typical procedure for reduction of α-ketones 5 (illustrated for preparation of the compound 6 SG-pBz ) Ester 5 SG-pBz (0.43 mmol, 250 mg) was dissolved in EtOAc (4.4 mL) and MeOH (0.4 mL), and the reaction mixture was cooled to 0°C. NaBH 4 (0.86 mmol, 33 mg) was charged to the flask and allowed to stir continuously for approximately 1 h. After confirming complete consumption of the starting material by TLC, the organics were diluted with EtOAc (25 mL), washed with 1 M HCl (1×30 mL) and sat. NaCl (1×30 mL), dried over Na 2 SO 4 , and concentrated in vacuo. The resulting clear, viscous oil 6 SG-pBz was used crude in the final deprotection, with an assumed 100 % yield; products 6 SG-pBz , 6 SS-pBz , 6 SG-OAc , and 6 SS-OAc were formed and used in the same way.
Typical procedure for deprotection of phenolic acetates (illustrated for preparation of the compound 7 SG-pBz ) Acetates were removed essentially as in the recently described method [53]. Benzyl alcohol 6 SG-pBz (0.32 mmol, 175 mg) was dissolved in DMF (1.3 mL). Hydrazine acetate (1.1 mmol, 103 mg) was charged to the flask, and the reaction was allowed to stir at room temperature for 1 to 3 h. The solution was diluted with EtOAc (50 mL) and washed with H 2 O (5× 15 mL) and NaCl (1×10 mL). After drying over Na 2 SO 4 , the organics were evaporated under reduced pressure. The resulting crude oil was purified by flash column chromatography (10 g SiO 2 , gradient elution using either 10 to 70 % EtOAc/hexane or 1 to 10 % MeOH in DCM) to afford 7 SG- Typical procedure for peracetylation (illustrated for preparation of compound 8 SG-pBz ) Compound 7 SG-pBz (10.0 mg, 20 μmol) was charged into a small amber vial, followed by addition of a solution containing equal parts pyridine and acetic anhydride (500 μL each). The resulting solution was allowed to stir overnight, at which point, all organic solvents were removed in vacuo to yield 8 SG-pBz as a clear, colorless oil: 1 H NMR (CDCl 3 ) δ 7.98-7.94 (m, 2.7H, 2 C,e /6 C,e , 2 C,t /6 ,C, γOAc e ); 13 C NMR (acetone-d 6 )  Note that compounds 8 SG-OAc and 8 SS-OAc , if they are the desired final products, are obviously directly available simply by reducing and acetylating compounds 4 as previously reported [64].
Syntheses of model compounds 9-11 Model compounds for 9a, 10a, and 11a (Fig. 5) in which the 4-O-etherification is simply via a 4-O-methyl group were synthesized from sinapyl 4-acetoxybenzoate and sinapyl alcohol by peroxidase-H 2 O 2 oxidation, followed by separation of the individual products, methylation (MeI-K 2 CO 3 , acetone), and acetylation (Ac 2 O-Py). Analogous model compounds for 9b, 10b, and 11b were obtained in the similar way starting from sinapyl acetate and sinapyl alcohol as previously described [14].

Results and discussion
The following sections address the various issues posed in the introduction regarding lignin p-hydroxybenzoylation in palms.

The Nature of p-Hydroxybenzoates and Acylation Regiochemistry
Determining whether p-hydroxybenzoates are in fact freephenolic entities acylating the γ-positions of lignin sidechains is most readily deduced from solution-state NMR of isolated lignins, analogously to similar determinations for pcoumarates [25]. Björkman milled wood lignins (MWLs) were prepared from the oil palm (Elaeis guineensis) EFB fibers, fronds, and trunk fractions; the EFBs had already been shown to be particularly rich in p-hydroxybenzoates [34]. Simple saponification of EFB whole-cell-wall (WCW) samples indicated that the EFB-1 and EFB-2 samples contained 1.09 and 1.05 wt% p-hydroxybenzoic acid or, based on AcBr lignin contents of 22.7 and 18.3 %,~4.8 and 5.8 % of the lignins ( Table 1). NMR of the underivatized (i.e., un-acetylated) isolated lignin was made difficult due to the high levels of iron in this sample, as previously reported [65]; unfortunately, we were not able to sufficiently effect its removal by complexation with EDTA. Nevertheless, 13 C NMR (Fig. 1) readily revealed the presence of p-hydroxybenzoates, assumed at this point to be on lignins. It is worth noting that these signals have sometimes been misinterpreted as arising from phydroxyphenyl lignin units or p-coumarates [43]. Acetylation of the isolated lignin facilitated removal of the metal contaminants. The observed changes in chemical shifts of the aromatic carbons following acetylation indicated that phydroxybenzoates were present overwhelmingly in their free-phenolic form on the lignin, Fig. 1c-d. The sago palm lignin, used in a previous study in one of our labs [35], had similar features, Fig. 1a; in fact the p-hydroxybenzoate level was highest on this particular lignin fraction. Also striking, and not observed in previous studies, was the high level of natural acetylation of the EFB lignin, Fig. 1b-c. The presence of natural acetates on lignins is frequently missed for a variety of reasons, including that lignins are routinely peracetylated to improve NMR properties, extraction conditions may hydrolyze the acetates [35], and lignin purification steps involving acetic acid can introduce artifactual acetates. Circumventing some of these issues, solid-state NMR spectra of various palm fractions indicate the presence of acetates, Fig. 1b. Furthermore, acetates are clearly observable in spectra from other published reports, where they have typically been attributed to polysaccharides [66,42]; historically, the contribution from lignin may have therefore been overlooked.
The gradient-enhanced 2D HSQC NMR experiment provides a highly detailed view of the molecular structure of plant cell walls with minimal processing of the sample [67]. The fast relaxation caused by the metals lead to extremely poor spectra, Fig. 2a-b, but the chemical shift data are consistent with acylation of the lignin. The p-hydroxybenzoate (pBz) groups are revealed in the aromatic regions of Fig. 2b, along with the G and predominant S aromatics. The side-chain region, Fig. 2a, offers good evidence that lignin γ-hydroxyls, and not the alternative α-hydroxyls, are acylated; acylated β-ether α-H/C correlations, for example, would be seen at~6.2/76 ppm. After peracetylation of the lignins, followed by successive extraction into chloroform and washing with EDTA in order to remove metal contaminants, higher-resolution spectra, Fig. 2c-d, are obtained. At this point, the nature of the acylation (i.e., γvs. α-regiochemistry) is masked, and distinguishing acylation patterns indicative of p-hydroxybenzoates from those of acetates is less direct; this issue is somewhat ameliorated by the considerably higher resolution of the spectra. The shifts in the p-hydroxybenzoate HSQC correlations (especially pBz 3/5 ) following acetylation (Fig. 2d) concur with earlier observations regarding the 1D 13 C spectra, namely, that p-hydroxybenzoates on lignins exist in their free-phenolic form and are not etherified; were they etherified, the pBz group would not become acetylated and their correlations would be essentially unchanged from those in the unacetylated lignins (Fig. 2b).
p-Hydroxybenzoates are therefore confirmed to be freephenolic pendant entities acylating the γ-hydroxyls of lignin side-chains. Additional proof is established below.
Which Lignin Units are p-Hydroxybenzoylated?
Next, we needed to establish which units in lignins were acylated by p-hydroxybenzoates. Consistent with observations on p-coumarates and acetates, were p-hydroxybenzoates largely on S units, and were they on a range of lignin unit types? In this case, NMR of the acetylated lignin derivatives could answer both of these questions as, fortuitously, the αprotons in acetylated γ-p-hydroxybenzoylated β-ether units B were rather well resolved from the normal acetylated β-ether units A (Fig. 2c).
The 1D proton projection on the HSQC spectrum in Fig. 2c indicates how well the α-protons in the major β-ether units are resolved by proton NMR. The γ-p-hydroxybenzoylated βether units B have α-proton/α-carbon correlations at 6.20/ 76.8 ppm (threo) and 6.19/75.7 ppm (erythro), whereas those from the normal acetylated units A are at 6.06/76.8 ppm Fig. 1 1D 13 C NMR spectra of various palm lignins. a Solution-state spectrum from a sago palm MWL isolated as described previously [35]. b Solid-state (200 MHz) spectrum and c solution-state spectrum of MWL isolated from the oil palm empty fruit bunches (EFB) previously characterized [34,43]. d Solution-state spectrum of acetylated EFB isolated lignin (Ac-MWL). The solution-state spectra were run at 400 MHz (100 MHz 13 C); the acetylated MWL is in acetone-d 6 , underivatized MWLs in 9:1 acetone-d 6 :D 2 O. Red-colored peaks labeled pBz are for carbons in the p-hydroxybenzoyl moiety. The red lines between plot d and its axis denote the chemical shifts from model compound 8 SS-pBz . S syringyl unit. The side-chain regions are quite disperse but generally labeled α, β, and γ. γ-OpBz signifies p-hydroxybenzoylated γ-alcohols, γ-OAc is (naturally) acetylated γ-alcohols. Cyan-colored peaks in b-c indicate the natural γ-acetylation of these lignins; no acetates appear in the sago palm lignin in a presumably due to the isolation method [35]. Spectra of lignins isolated from other parts of the oil palm plant (trunk and fronds, not shown) are almost indistinguishable from those shown in c-d (threo) and 6.02/75.4 ppm (erythro). Initially, the α-1 H-13 C correlation contours were not sufficiently resolved, but running higher-resolution spectra restricted to acquiring just the side-chain region resulted in sufficient resolution to readily distinguish the various correlations (Fig. 2c).
It is apparent from the aromatic region of HSQC spectra that the EFB lignin was S-rich, Fig. 2b, d. An S:G ratio of 83:17 can be derived from the data in Fig. 2d by volume integration of the S 2/6 and G 2 aromatic proton/carbon correlations. In support, the side-chain region (Fig. 2c) revealed very low levels of phenylcoumaran (β-5) and dibenzodioxocin (5-5/β-O-4) units that must involve G units-their contours appear below the levels plotted.
Distinguishing whether p-hydroxybenzoates are attached to S or G units requires long-range 1 H-13 C correlation experiments [13]. The heteronuclear multiple bond correlation  (360 MHz); a side-chain region, b aromatic region. The unacetylated material contained magnetic material, presumably iron compounds, that could not be effectively washed out; the spectra are therefore extremely broad, but a still clearly shows the degree to which γhydroxyls are acylated, either by acetate or p-hydroxybenzoate. c-d The EDTA-washed acetylated EFB MWL (500 MHz cryoprobe instrument); c side-chain region, d aromatic region. These much sharper spectra resolve the α-protons (and the corresponding correlations) of γ-phydroxybenzoylated B from γ-acetylated A units. e-g HSQC slices (F2-F3 dimension) from a 3D TOCSY-HSQC experiment on acetylated EFB lignin over a small proton chemical shift range in the F1 dimension: e 6.29-6.19 ppm (α-protons in γ-p-acetoxybenzoylated β-ether units B), f 6.13-6.07 ppm (α-protons in normal β-ether units A, both syn-and antiisomers), and g 6.04-5.98 (α-protons in normal β-ether units A, mainly anti-isomers). The colored symbols on the red contours in e are shifts from model compounds 8; clearly, only the SS-pBz moieties are present at significant levels in the lignin (with chemical shifts for the α-H/C correlations matching well with those from compounds syn-and anti-SS-pBz but not with those from the SG-pBz compounds; the β and γ data are not shown for the latter as they are less diagnostic). Units A are normal (acetylated) β-ether units, B are γ-p-acetoxybenzoylated β-ether units. G guaiacyl, S syringyl, pBz p-hydroxybenzoyl (HMBC) experiment can correlate a proton with carbons up to three bonds away. The simplest probe is via the resolved αprotons in β-ether units (in the acetylated lignins), which should correlate with carbons γ (3-bond), β (2-bond), 1 (2bond), and 2 and 6 (3-bond). The feature of this experiment that makes distinction of S-from G-units so elegant is that the identical 2/6 carbons in symmetrical S units resonate at about 105 ppm, whereas carbons 2 and 6 are distinctly different in the unsymmetrical G-units, resonating at 114 and 120 ppm. As shown in Fig. 3a, it is readily evident that the β-ethers in normal units A are both S (dark blue) and G (cyan), whereas it is only possible to detect γ-p-hydroxybenzoylated S units B (red). The prevalence of p-hydroxybenzoates on S over G units may be either a reflection of the control in the transferase-catalyzed acylation reaction or simply the timing of the conjugation and the levels of the two monolignols present at that time, i.e., that p-hydroxybenzoylation of sinapyl alcohol is either significantly favored over coniferyl alcohol, or that sinapyl alcohol is more prevalent at the time of acylation.
The 3-bond correlations in the HMBC data between the γprotons and the carbonyl (Fig. 3b) also provide compelling proof that the p-hydroxybenzoates acylate lignin γ-positions. The HMBC experiment can also be used to identify the variety of units that are acylated. This is achieved here by examining the γ-protons correlating with the carbonyl carbon of the phydroxybenzoate moiety (Fig. 3b). However, determining the range of products is not as easy as it was for p-coumarates, as there are insufficient model compounds, and the dispersion between the unit types is limited. For example, it is currently difficult to prove that p-hydroxybenzoates acylate the phenylcoumaran (β-5) units or hydroxycinnamyl alcohol endgroups; each of those unit types is present only at low levels in this S-rich lignin. The dispersion that is afforded to the α-protons suggests that it should be possible for a 3D TOCSY-HSQC experiment to delineate whether p-hydroxybenzoates acylate both syn-(threo-) and anti-(erythro-) isomers of the β-ether units. This is because one can take 2D HSQC planes at given proton frequencies and exploit the α-proton differences between the γ-p-hydroxybenzoylated B and normal A units noted above. Similarly, one can take 2D HSQC-TOCSY planes through the α-proton region. Such 2D HSQC slices are shown in Fig. 2eg, and HSQC-TOCSY slices are shown in Fig. 3c-e, as described in the captions. It is revealed that there are syn-and anti-isomers of the normal β-ether units (Figs. 2f-g and 3de), with anti-isomers predominating because of the S-rich nature of the lignin, as already well established [68][69][70][71][72]. It is equally clear that p-hydroxybenzoates acylate both of the βether isomers (Figs. 2c and 3c). A comparison of the α-H/C correlation peak data for the lignin versus the overlaid data from the various model compounds (Fig. 2e) indicates that, in addition to the pBz's being almost entirely on S units as noted above, the data for the two SS-pBz isomers match very well, but those for the β-O-4-G analogs SG-pBz do not. Similarly, the data match well for the same SS compounds and not (especially for the β-H/C correlations) for the SG analogs (Fig. 3c). As in the case made early on for p-coumarates acylating grass lignins [25], the presence of p-hydroxybenzoates indiscriminately on both isomers of β-ether units suggests that they arise by lignification with preformed monolignol p-hydroxybenzoate conjugates.

Determination of Regiochemistry and Quantification by DFRC
Proving γ-regiochemistry (although not its exclusivity) and the S versus G nature of the units involved with p-hydroxybenzoates is elegantly made by the DFRC (derivatization followed by reductive cleavage) assay [61,62], Fig. 4a. This is possible as DFRC advantageously cleaves lignin βethers but leaves esters fully intact [62,51,18,23,27]. Unfortunately, the acetylated coniferyl and sinapyl phydroxybenzoate products anticipated from this method do not survive GC analysis conditions, so the products were analyzed by HPLC-ESI-MS, Fig. 4b. Quantification of the DFRC-released diacetates of coniferyl alcohol (CA), sinapyl alcohol (SA), coniferyl p-hydroxybenzoate (CA-pBz), and sinapyl p-hydroxybenzoate (SA-pBz) accounted for 0.9, 2.0, 0.1, and 1.3 wt% of the EFB-1 sample ( Table 1). The monolignol component, uncorrected for recovery and losses, accounted for some 16.7 % of the lignin. The DFRC-released total S to G monomer ratio was 73:27, whereas the ratio of SA-pBz to CA-pBz was 96:4, confirming the NMR observations that the p-hydroxybenzoates are primarily located on S units. Similar data were obtained for the EFB-2 sample ( Table 1).

Proof that Lignin p-Hydroxybenzoates Arise via Lignification with Preformed Monolignol Conjugates
All of the aforementioned data strongly suggest, but strictly do not prove, that the p-hydroxybenzoate groups inherent on these palm lignins arise from lignification incorporating preformed monolignol p-hydroxybenzoate conjugates. As we have described previously for acetates (e.g., on kenaf lignins), there is a way to provide compelling proof [62,14]. Given that post-coupling rearomatization of β-β-coupled quinone methide lignification intermediates is the one reaction that is perturbed by the presence of an acylated γ-OH, observing novel acylated β-β-coupling products in the lignin provides convincing evidence that acylated monolignols are involved in lignification [14].
As seen in Fig. 5, evidence for all three types of products 9a-11a indicating the involvement of sinapyl p-hydroxybenzoate monomers in lignification is provided by observing sets of signature H/C correlations in HSQC spectra, correlations that well match the data from synthesized model compounds for all of the structures 9a-11a. These units include both those arising from the cross-coupling of sinapyl phydroxybenzoate with sinapyl alcohol (9a, 10a) as well as the homo-dimerization of sinapyl p-hydroxybenzoate (11a). These spectra also provide evidence (with correlations for structures 9b-11b) that sinapyl acetate is a monomer conjugate in this palm EFB lignin, as was previously observed in kenaf [14].
Evidence from metabolite profiling of poplar wood for novel β-β-coupled heterodimers between sinapyl alcohol and sinapyl p-hydroxybenzoate [29,32] suggests that it is safe to contend that all lignin p-hydroxybenzoates, across the plant taxa that have them, derive from lignification using monolignol p-hydroxybenzoate conjugates. The evidence is therefore sufficiently compelling that currently unknown transferases are responsible for producing monolignol phydroxybenzoate conjugates that are used as 'monomers' for lignification in palms (as well as in willows, poplars, and aspen). The levels of p-hydroxybenzoates on lignins should be manipulable by normal methods of selection and breeding or by misregulating such genes, once found. If p-hydroxy benzoic acid or its derived products are sufficiently valuable as commodities and coproducts generated from a biomass conversion process, a long-term strategy might be to upregulate the p-hydroxybenzoate levels on lignin in target plants. As determined here, only some 23 % (Table 1) of lignin monomers are p-hydroxybenzoylated in palm EFBs, and the level is likely to be lower in poplar and willow, so there appears to be significant potential for its increase. Other factors being equal, for example, if all of the monolignols destined for lignification were p-hydroxybenzoylated, the amount of p-hydroxybenzoic acid available would be approximately 40 wt% of the lignin or roughly 8 wt% of the EFB fiber.

Conclusions
Our data provide compelling evidence that, as for acetates on kenaf (and many other plants) lignins and p-coumarates on grass lignins, p-hydroxybenzoates acylate the γhydroxyl of lignin side-chains and are found mainly on S units. Furthermore, like their p-coumarate analogs, the phydroxybenzoate moieties are not subjected to significant levels of radical coupling, or other etherification, during lignification and therefore remain overwhelmingly as freephenolic pendant units on lignin unit side-chains and can therefore be easily cleaved off to deliver a single compound in quite pure form. As has been demonstrated for p-coumarates, and as there is no a priori reason that p-hydroxy-benzoates can not undergo radical coupling, the logical reason is that such units are preferentially subject to radical transfer reactions to the G and S monomer and polymer phenolics present in the radical-limited lignification process [12,73]. The data presented here also provide evidence that, analogously to the p-coumarates and acetates found acylating lignin units, p-hydroxybenzoate units arise from lignification using the preformed monolignol p-hydroxybenzoate conjugates and, at least in these palm lignins, are almost entirely derived from sinapyl p-hydroxybenzoate. It is now logical to contend that any of the acylation products seen in any of the native lignins is via acylation at the monomer level and that post-polymerization (or postcoupling) acylation reactions are unlikely. We therefore  Table 2 and elsewhere. The retention time, UV-vis spectrum, and mass spectrum of each compound were authenticated using external synthetic standards expect that p-hydroxybenzoyl-CoA:monolignol transferases are involved in lignification in the various willows (Salix spp.), poplars and aspen (Populus spp.), and palms (family Arecaceae) that have p-hydroxybenzoylated lignins, and that identification of the genes involved will allow p-hydroxybenzoate levels on lignin to be manipulated as recently demonstrated for p-coumarates [23]. To date, however, the in planta pathway to p-hydroxybenzoate (or its CoA derivative) remains ambiguous. If p-hydroxybenzoic acid becomes a viable coproduct from cellulosic bioenergy lignin streams, it could conceivably be upregulated or even introduced into other biomass plants. Right now, the current palm oil empty fruit bunch 'wastes' should be able to generate a sizeable stream of relatively clean p-   Fig. 5c (γ-correlations) is simply the second (lowest field, highest ppm) γ-H/C correlation from the B unit of compound 9a hydroxybenzoic acid; when available at the multiton level, industrial applications are expected to increase beyond those already in use today.