Copper‐Mediated Conversion of Complex Ethers to Esters: Enabling Biopolymer Depolymerisation under Mild Conditions

Abstract Selective processing of the β‐O‐4 unit in lignin is essential for the efficient depolymerisation of this biopolymer and therefore its successful integration into a biorefinery set‐up. An approach is described in which this unit is modified to incorporate a carboxylic ester with the goal of enabling the use of mild depolymerisation conditions. Inspired by preliminary results using a Cu/TEMPO/O2 system, a protocol was developed that gave the desired β‐O‐4‐containing ester in high yield using certain dimeric model compounds. The optimised reaction conditions were then applied to an oligomeric lignin model system. Extensive 2D NMR analysis demonstrated that analogous chemistry could be achieved with the oligomeric substrate. Mild depolymerisation of the ester‐containing oligomer delivered the expected aryl acid monomer.


Introduction
The efficient use of lignocellulosic biomass for the generation of renewable chemicals and biofuelsr equires the use of the biopolymer lignin. Whilst burning lignin remainso ne viable option, its high aromatic ring content provides an opportunity for renewable aromatic chemical feedstock production.However,d epolymerising lignin to deliver the aromatic components remains challenging. An umber of approaches exist including the "lignin-first" approach. [1] Alternatively,p retreatment of the lignocellulosicb iomass can be used to generate al ignin-rich product stream that can then be depolymerised in one, [2] two [3] or sometimes more [4] steps. For example, we [3b] and others [3e, 5] have reported a-oxidation of the highly abundant b-O-4 followed by subsequentp rocessing of the b-O-4 a-OX units as a means of depolymerising lignin. Others [3c, d, 6] have used an initial g-oxidation of the b-O-4 unit as the first step in lignin depolymerisation. Whilst in the majority of the current multistep approaches ar elatively mild and controllable initial step (e.g. selective a-oxidation) is used, it is also important that the last processing step avoidsh igh temperatures and/or harsh chemical treatments. By combining mildc onditions in both steps, it is possible to isolatet he desired monomers and leave ar esidual lignin that can be used further. [7] In this context, we have recently become inspired by Bartley and Ralph's reports [8] of "zip lignins" that contain ester groups (Scheme 1A). Theirw ork was based on the introduction of chemically labile ester bonds into the lignin polymer by encoding af eruloyl-coenzyme Am onolignolt ransferase into popular trees (Populus alba and Populus grandidentata). The presence of the ester group made the woodm ore prone to depolymeri-Scheme1.A) Incorporation of monolignol ferulates into lignin introduces chemically labile esters into the polymer backbone. [8a, b] B) 18 alcohol oxidation and subsequent NHC catalysis for the depolymerisation of butanosolvderived lignin b-aryl ether linkages. [3a] C) Koskinen's experiment: [9] 0.2 mmol scale in CH 3 CN (0.1 m); Cu cat. = 10 mol %C uCl/10 mol %N aBF 4 /10 mol % BiPy (2,2'-bipyridine)/10 mol %TEMPO (2,2,6,6-tetramethylpiperidine 1-oxyl)/ 10 mol %N MI (N-methylimidazole) under an oxygen atmosphere. sation under relativelym ild, alkaline conditions. [8] An additional possible advantage of the presence of an ester in the lignin chain is that the use of mild cleavage conditions could result in the formation of smaller lignin chains which have not been condensed/ destroyed during af inal depolymerisation step. We have therefore been searchingf or ways to modify selectively the b-O-4 unit to give lignin chainst hat contain ester functional groups.
Our initial efforts in this area focused on the use of NHC catalysed-redox esterification but were ultimately derailed by the fact that the phenolic rebound reaction proved inefficient (SchemeS1). In the end we compromised and used butanol as an alternative rebound reactant leading to direct depolymerization and delivery of aromatic monomers from lignin (Scheme 1B). [3a] However, duringt he course of our NHC studies we noticedareport by Koskinen et al. [9] Their studies on the reaction of b-O-4 model compounds under CuCl/TEMPO/O 2 /NaBF 4 / BiPy/NMI/ CH 3 CN conditions included the reaction of the methylated b-O-4 model 1 ((a-OMe)-b-O-4) to give the corresponding g-aldehyde 2 as the major product along with a7% yield of the non-cleaved aryl ester 3 (Scheme1C). The formation of smalla mountso fa ldehyde 4,m ethyl ester 5 and a, bunsaturated aldehyde 6 was also reported.
Despite the fact that production of 3 was not the focus of Koskinen's work, this observation was potentially relevant to our quest. In as eparatep art of our lignin programme we have focussedo napretreatment protocolt hat uses butanol asa co-solvent. [10] Whilst others [11] have used butanol in pretreatments,t he existing studies did not discuss the almostf ull conversion of the b-O-4 unit to the corresponding (a-OBu)-b-O-4 unit that we observe under our conditions. This has led us to lignins that are rich in (a-OBu)-b-O-4 units analogous to the methylated unit ((a-OMe)-b-O-4) 1 used by Koskinen. We therefore decided to revisit Koskinen's work but using the (a-OBu)-b-O-4 unit rather than 1.

Results and Discussion
We have previously reported the synthesis of am odel of the (a-OBu)-b-O-4 unit 7 as am ixture of diastereomers. [10a] This study therefore startedw ith the attempted conversion of 7 under Koskinen's conditions [9] (20 hour reactiont ime, Figure 1A,e ntry 1). As expected, the outcome was very similar to that of Koskinen's with the major product being the known [3a, 12] g-aldehyde 8 (70 %, entry 1). Reassuringly, as mall quantity of the corresponding ester 9 (4 %) was also formed. It was decided to repeat this reaction starting with 8 in an attempt to generatem ore of the desired 9.A fter 20 hours (entry 2), overall conversion of 8 was low (20 %) and only small amountso f9 (5 %) were formed. At a6 0hour extended time point (entry 3), whilst conversion of 8 increased to 39 %, the amount of 9 (4 %) remained very low with increased formation of ester 11 (14 %), resulting from cleavage of the C a ÀC b bond in the b-O-4 unit, in addition to ar ange of unidentified products. Given the low yields of 9,i tw as decided to abandon the Koskinen protocol.
Baker et al. [13] have reported catalytic systemsf or the selective cleavage of lignin models using oxygen as the terminal oxidant. In parallelw ork, Baker's originalp rotocol (CuCl/TEMPO/ O 2 /Py) was being exploredb yu st oa chieveo xidative cleavage of 8 (Figures S7-S17 for am ore detailedd iscussion). Efficient conversion (100 %) of 8 in the presence of sub-stoichiometric amountso fC uCl and TEMPO in pyridine at 100 8Cu nder an oxygen atmosphere ( Figure 1B,e ntry 4) resulted in the formation of butyl ester 11 as the major product (41 %; 36 %i solated yield). Importantly in the context of the current work, an increasedy ield of the desired ester 9 (28 %) was observed compared to the Koskinenp rotocol. Aldehyde 10 (14 %) along with trace amountso fk nown [3a] enal 12 (2 %) were also formed in this reaction.
The Baker system contains TEMPO, O 2 and CuCl. Therefore it was decided to assess the role of each component in the reaction in the hope of biasing the system away from Ca-Cb bond cleavage in 8 and towards formation of ester 9.I nitial studies focussed on the use of CuCl under an oxygen atmosphere and [a] Reactions were performed on a0 .1 mmol scale. Conversions and yields were determined by 1 HNMR using 1,3,5-trimethoxybenzene as the internal standard.
in the absence of TEMPO. Resultss howed that TEMPO was not essential for the formationo f9 (c.f. Figure 1B,e ntries 4a nd 5) with high conversion (88 %) of 8 to 9 (29 %) and 11 (40 %) being achieved with no TEMPO in the reaction. Extending the reactiont ime to 15 hours at 100 8C( entry 6) led to as light increase in the production of 9 (38 %a t9 5% conversion of 8) but more than half of the reactionp roducts still resulted from CaÀCb bond cleavage. These conditions also efficientlyc onverted the S-G (a-OBu)-b-O-4 aldehyde 13 to the corresponding ester 14 (entry 7, 35 %a t1 00 %c onversion of 13)w ith 15-17 also being formed. In contrast, when these conditions were appliedt ot he isomeric G-S (a-OBu)-b-O-4 aldehyde 18 am uch poorer outcomew as obtained with an overall conversion of 18 of only 18 %a nd no observable formation of the analogous aryl ester (entry 8). Butyl ester 11 (16 %) was the only identifiable product from this reaction. To the best of our knowledge, this is the first time that ar eactivity differencea safunction of the methoxy-substitution pattern in the b-O-4 unit has been observedu nder this type of reactionconditions.
In the absence of both TEMPO and CuCl ( Figure 1B,e ntry 9), efficient conversion of 8 (88 %) was stillo bserved at 100 8Cb ut none of the desired ester 9 was formed with the major product resulting from elimination of butanol to give 12.T his result confirmedt hat the presenceo fC uCl is essential for the formation of 9 under the Baker conditions. One possible mechanism (Scheme 2) involves the formation of the enolate [9,14] 19,w hich could be converted to form the corresponding radical. The radical could be converted to the hydroperoxidei ntermediate 20. Subsequentd eformylation would be expected to give 9.P ossible reasons for the apparent lack of reactivity of the G-S model 18 mayb et hat the combination of the two radicalsi sm ore difficultd ue to increased steric constraints around a b-carbon centred radical.
Whilst the presence of CuCl is not essential for the formation of butyl ester 11 ( Figure 1B,e ntry 9), the yield of 11 is significantly decreased in its absence, indicating that 11 may be formed by two alternative pathways (Scheme S2). In several of the reactions ( Figure 2B,e ntries 5, 6a nd 7), as mall amount of aldehyde 10 (or 15 when 13 was used as the substrate) was formed in the presence of CuCl. Koskinen proposed that under their conditions, the formationo f10 from 1 is likely via 1,4-addition of H 2 Ot oe nal 12 followed by ar apid retro-aldol reaction [9] although other possible mechanismsc ouldb ec onsidered (Scheme S3 and Figure S23).
Returning to the originalB aker reaction conditions (Figure 1B,e ntry 4), it wasdecided to rerun this reactionint he absence of oxygen to explore the contribution that TEMPO made to the production of 9 from 8 (FigureS24). As no oxygen was present and hence no chance of establishing ac atalytic cycle involving oxygen,a ne xcess of TEMPO was used. Under as carefullyc ontrolled conditions as could be achieved in the absence of ag love box ( Figure S25 for am ore detailed discussion), the model (a-OBu)-b-O-4 aldehyde 8 was efficiently converted to ester 9 using 1.5 equivalents of TEMPO, this time as the major product (62 %, Figure2,e ntry 1). Increasing the number of equivalents of TEMPO led to af urtheri ncrease in the yield of 9 (74 %, entry 2a nd 77 %, entry 3) and these significant improvements in the formation of the desired ester were also observedf or S-G model 13 (entry 4). Very low conversions were again observed when the G-S model 18 and the S-S model 21 (entries 5a nd 6) were used.
The benzylic oxidation and subsequentr etro-aldol reaction often proposed for these types of reactionsw hen native b-O-4 substrates are used [5, 9, 13c] is blockedi ns ubstrate 8 unless the a-butoxylated lignin undergoese limination of butanol and Scheme2.One possible mechanism for the formation of 9 under cat. CuCl/ O 2 /pyridine conditions(based on literature findings [9, 14b] ). Genericn ucleophile Nu:c ould be water or pyridine. subsequenta ddition of extraneous water.A nother possible mechanism could involve the formationo fe nolate [9,14] 19 (Figure 3C)a nd formation of the corresponding radical 23.T he adduct 22 could then be formed by radical coupling of 23 with TEMPO. Subsequentd eformylation, in an analogous manner to the reaction of 20 (Scheme 2), would convert 22 to 9.
Having identified optimised conditions for the formationo f the desired esters 9 and 14 from the dimeric model compounds 8 and 13 respectively,t he next challenge came in applying this chemistry to am odel b-O-4-containing oligomer. This step is required to assessw hether it is possible using this methodology to prepare ester-containing chains with minimal initial cleavage of the b-O-4 unit. Owing to the complexity of real samples of even butanosolv lignin, it was decided to use an artificial all b-O-4-containing oligomer.A st he TEMPO-mediated ester formation reaction did not work on G-S model 18 and S-S model 21 (Figure 2, entries 5a nd 6), studies focused on the all G b-O-4 oligomer (Scheme 3) which is as implified model of as oftwood lignin such as that obtained from Douglas fir wood. [10b] Oligomer 24 (M w 2880, DP n (degree of polymerisation)8 .13, FiguresS41-S44 for am ore detailed discussion) was synthesised using the previously reported route. [3b, 17] In an extension to previous literature 24,w hich contains native b-O-4 units, wasc onverted to the corresponding a-functionalised butanosolv b-O-4 oligomer 25 (Scheme 3a nd Figure 4A) using am odified butanosolv pretreatment [10a, 12] in quantitative yield ( Figures S45-S46). The subsequent g-oxidation reaction was then carriedo ut using Dess-Martin periodinane (DMP) [3a, 12] to give the required oligomeric substrate 26 ( Figures 4B and  S47).
In brief, the key signals observed in the HSQC NMR analysis of oligomer 25 included those corresponding to the a ( 1 H/ 13 C d 4.37-4.54 ppm/80.2-83.1 ppm), b ( 1 H/ 13 C d 3.99-4.22 ppm/ 85.1-88.2 ppm) and the diastereotopic g-positions of the two diastereomers ( 1 H/ 13 C d 3.32-3.46 ppm/61.3-63.1 ppm and 1 H/ 13 C d 3.32-3.46 ppm/61.3-63.1 ppm and 1 H/ 13 C d 3.72-3.92 ppm/60.7-62.6 ppm) of the gA unit. Signals corresponding to the two endg roups (BE and EG)a sw ell as small amountso fg-ester units (gE,F igures S45-S46 for am ore detailed discussion), remaining from incomplete ester reduction duringt he synthesis of 25,w ere also present. On oxidation of 25 to 26,d isappearance of the signals correspondingt ot he galcohol positions of unit gA in 25 appeared linked to the presence of new signals corresponding to the g-aldehyde positions in the desired unit gO ( 1 H/ 13 C d 9.57-9.68 ppm/199.6-201.5 ppm and 1 H/ 13 C d 9.77-9.89 ppm/199. Figures 4B and Figure S47).
Inspired by the dimer study ( Figure 2, entry 2), an initial attempt to form the ester using oligomer 26 started by reacting 26 with TEMPO (2.0 equiv.) in the presence of CuCl (0.1equiv.) in pyridine under an Ar atmosphere for 7hours. The reaction outcome was encouragingb ut the major product contained predominantly TEMPO-adduct TA units (FigureS52). Whilst studies on thes impler model 8 suggested that ar eaction time of 7-9 hours may be ideal ( Figure 3B), it was clear that modified conditions would be required for the successful preparation of 27 containing predominantly ester units (Scheme 3). For example, when ar eaction time of 15 hours was tried (using 3.0 equiv.T EMPO and 10 mol %C uCl) signals correspondingt ot he formation of the TEMPO-adduct unit TA (minorc omponent) and the desired ester unit AE (major component) were observed ( Figure 4C andF igure S48-S51). Diagnostic changes in the HSQC analysisi ncluded the disappearance of the signals assigned to the b-( 1 H/ 13 C d 4.34-4.51 ppm/ 85.7-88.1 ppm and 1 H/ 13 C d 4.14-4.30/86.9-89.0 ppm) and gpositions of the two diastereomers ( 1 H/ 13 C d 9.57-9.68 ppm/ 199.6-201.5 ppm and 1 H/ 13 C d 9.77-9.89 ppm/199.9- Whilst the 2D HSQC analysiso f26 confirmed that the major reactiono nf orming 27 involved formation of the desired ester unit (AE), evidence for the formation of units formed by cleavage of the C a ÀC b bond (units CE and BA in Figure 4C)w ere also obtained (Figures S57 for amore detailed discussion), consistent with the dimer study result (Figures S24-S25). This was exploredf urtheru sing ap reviously reported [19] 2D DOSY NMR approacht his time on 26 and 27 to determine any apparent change in the molecular weight( MW) of the lignin models. There was, on average, as mall increasei nt he diffusion coefficient on going from 26 (blue signals) to 27 (green signals, Figure 5), indicating ad ecrease in size duringt he reaction, which was consistent with the appearance of the signals observed for the CE unit in the HSQC analysis of 27 (Figure4C). On hydrolysis of 27 (Scheme 3, step D), as ignificant increasein the average diffusion coefficient of the crude product(s) was observedc ompared to both 26 and 27 ( Figure 5). Subsequent purification of the crude products by column chromatography gave the expected acid 28 (Scheme 3, 7.0wt% isolatedy ield). No attemptst oo ptimise this yield have been made. In addition, am ixture of products (16.0 wt %i solatedy ield) that was shown to contain dimeric units was isolated from the column ( Figures S59-S60 for am ore detailed discussion). The isolation of these dimericu nits (albeit as part of am ixture) was consistent with ar elatively mild final step in the depolymerisation process allowing fragments of model oligomerc hains to be obtained for potentialf uture use ( Figure S61 for am ore detailed discussion).  Figures S46-48 and S60. The chemical shifts of peaks in the aldehyde region haveb een correctedd ue to folding in the original analysis.
[a] he assignment of signals in gA was determined by comparison the 2D HSQC spectraof25 with that of the correspondingd imer model compound ( Figures S45-S46). [b] The presence of signals correspondingtot he g-ester unit (gE)w as achieved by comparison with the 2D HSQC spectrao fthe corresponding dimer model compound ( FiguresS44-S46).
[c] The signalsc orresponding to the a-protons in the benzyl alcohol (BE)e nd groupa nd the a and b-protons of the ethylene glycol (EG)end group wereassigned based on the literature. [18] After the butanosolv step (Scheme3,stepA), R( in the ethylene glycol end group) = CH 2 OH;a fter oxidation step (steps Ba nd C), R( in the ethylene glycole nd group) = CHO; [ d] In the dimer study,t he HSQC NMR spectra showed that the signal corresponding to the a-proton of ester 9 and the aprotono fT EMPO-adduct 22 wereoverlapping (FigureS50,entry E), making it hardt od istinguish AE from TA using this signal. However,t he presence of a TA unit was clear due to the presence of the aldehyde proton. In addition, HMBC analysis confirmed the formation of either AE or TA or both ( Figures S51).
[e]The formationo ft he acid unit AD is discussed in Figures S61.

Conclusions
In summary,w hilst preliminary studies based on the Koskinen protocol 21 did not enablethe efficient conversion of amodified b-O-4 unit to the desired ester-containing unit, an effective methodw as developed based using the CuCl/TEMPO system in the absence of oxygen. This methodology was not only applicable to adimeric b-O-4 model unit but also to an oligomeric b-O-4 model, leadingt oa no ligomer chain (in 27)t hat contained as ignificant number of internal ester units. Am ore detailed understanding of the key aspects of the CuCl/TEMPO/ pyridine system was obtained. For example, the study confirmed that the desired ester was formed via TEMPO-adduct 22 which was isolated and, on resubmission to the reactionconditions, was converted to the corresponding ester 9.D etailed 2D NMR (HSQC, HMBC and DOSY) analysiss howed that the reaction outcomes in both the dimeric and oligomeric models were analogous. It is envisagedt hat this methodc ould be applied to ab utanosolv lignin but challenges associated with the successful removal of oxygen and the amount of TEMPOa re likely to limit the scale on which it can be applied.