Sodium hydroxide (BioXtra, ≥98% (acidimetric), pellets (anhydrous)), Diphenyl ether (ReagentPlus®, ≥99%), Cyrene™ (≥98.5%), Dimethyl sebacate (99%), Dimethyl succinate (98%), Candida antarctica Lipase B (CaLB, code: L4777), chloroform (suitable for HPLC, ≥99.8%, contains 0.5%–1.0% ethanol as a stabilizer), N,N-dimethylformamide (anhydrous, 99.8%), Ethylene carbonate (98%), p-Toluenesulfonic acid monohydrate (ACS reagent, ≥98.5%), Pentaerythritol (for synthesis), Sodium bicarbonate (ReagentPlus®, ≥99.5%, crystalline), Anisole (anhydrous, 99.7%) and Vanillin (for synthesis) were purchased from Sigma Aldrich. Diethyl isophthalate (97%) was purchased from Syntree Inc. Dimethyl adipate (99%), Potassium carbonate (anhydrous, 99%), and Diethyl terephthalate (95%) were purchased from Alfa Aesar. 2-Propanol (ROTIPURAN®, >99.8%, p. a., ACS, ISO) and dimethyl sulfoxide (ROTIPURAN® ≥99,8%, p. a.) were purchased from Roth. Diethyl pyridine-2,4-dicarboxylate (PD24) was purchased from Carbosynth. Diethyl pyridine-2,5-dicarboxylate (PD24) was purchased from TCI.

Cygnet 2 and Cygnet 4 were synthesised according to a previous work (Warne et al., 2023).

Compound (1) was synthesised according to a previously published procedure (Mankar et al., 2019), and was recovered as a white powder (35% yield).

Chemical structure diagram showing a linear molecule with two aromatic rings connected by a central chain of three oxygen-containing rings. Each aromatic ring has hydroxyl and methoxy groups attached.

¹H NMR (400 MHz, CDCl₃, δ, ppm): 7.02 (d, J = 1.9 Hz, 2H), 6.96 (dd, J = 8.1, 1.9 Hz, 2H), 6.89 (d, J = 8.1 Hz, 2H), 5.66 (s, 2H), 5.39 (s, 2H), 4.89–4.81 (m, 2H), 3.90 (s, 6H), 3.87–3.76 (m, 4H), 3.63 (d, J = 11.6 Hz, 2H).

¹³C NMR (101 MHz, CDCl₃, δ, ppm): 146.5, 146.4, 130.2, 119.6, 114.2, 108.3, 102.4, 71.2, 70.7, 56.0, 32.5.

HPLC-MS-ESI: retention time: 2.20 min (MS Spectrum: m/z 137.2); 7.57 min (MS Spectrum: m/z 271.2); 9.62 min (MS Spectrum: m/z 153.2).

Three different methods were used for the synthesis of compound (2).

In dimethylformamide (DMF)

Compound (2) was first synthesised according to a previously published procedure (Mankar et al., 2019), and was recovered as a white powder (81% yield).

2. In dimethyl sulfoxide (DMSO)

This synthesis was adapted according to a previously published procedure (Mankar et al., 2019). Compound (1) (7.5 g, 18.6 mmol), ethylene carbonate (3.45 g, 39.2 mmol) and potassium carbonate (0.91 g, 6.6 mmol) were added to a 250 mL round-bottom flask with 50 mL DMSO. The reaction was heated from room temperature to 160 °C under N₂, and stirred for 2 h. On completion of the reaction, it was left to cool down to room temperature, to which NaOH solution (25 mL, 0.1 N) was added dropwise. The mixture was vacuum filtered and washed with 1 L of water, resuspended in a second L of water, and finally vacuum filtered and washed with a third liter of water. Compound (2) was then lyophilised and was subsequently recovered as a white powder (84% yield).

3. Solventless, using excess ethylene carbonate (EC)

This synthesis was adapted according to a previously published procedure (Mankar et al., 2019). Compound (1) (3.74 g, 9.3 mmol), ethylene carbonate (as specified in the text) and potassium carbonate (0.456 g, 3.3 mmol) were added to a 100 mL round-bottom flask. The reaction was heated from room temperature to 160 °C under N₂ and stirred for 2 h. On completion of the reaction, it was left to cool down to approx. 50 °C, to which NaOH solution (12.5 mL, 0.1 N) was added. The mixture was vacuum filtered and washed with water (250 mL). Compound (2) was dried under vacuum until a constant weight was obtained, yielding (2) as a white solid (80% yield).

Chemical structure diagram of a polymer segment showing two aromatic rings connected by two ether linkages and a central cyclohexane ring, with side chains containing methoxy, ethoxy, and hydroxyl groups.

¹H NMR (400 MHz, CDCl₃, δ, ppm): 7.06 (d, J = 1.9 Hz, 2H), 7.02 (dd, J = 8.2, 1.9 Hz, 2H), 6.93 (d, J = 8.2 Hz, 2H), 5.42 (s, 2H), 4.87 (dd, J = 11.6, 2.5 Hz, 2H), 4.12 (dd, J = 5.3, 3.9 Hz, 4H), 3.92 (p, J = 4.5, 2.8 Hz, 4H), 3.89 (s, 6H), 3.84 (ddd, J = 11.5, 8.9, 2.0 Hz, 4H), 3.66 (d, J = 11.6 Hz, 2H), 2.59 (t, J = 6.3 Hz, 2H).

¹³C NMR (101 MHz, CDCl₃, δ, ppm): 149.8, 148.7, 131.9, 119.0, 114.5, 109.5, 102.1, 71.5, 71.2, 70.7, 61.3, 55.9, 32.6.

HPLC-MS-ESI: retention time: 7.87 min (MS Spectrum: m/z 315.2); 9.32 min (MS Spectrum: m/z 197.2); 9.40 min (MS Spectrum: m/z 153.2); 12.27 min (MS Spectrum: m/z 493.4).

8 x 10⁻⁴ mol of diester and an equimolar amount of (2) was added to a 25 mL round bottom flask with 2 g of solvent, and 10% by weight of monomers of CaLB (51.1–57.8 mg). The flask was heated to 85 °C and stirred at 400 rpm for 6 h, at which time the system was placed under vacuum (20 mbar) for a further 90 h. Reactions that used anisole as the reaction solvent were performed under reflux, and under N₂ for the first 6 h.

Upon completion of the reaction, CHCl₃ (unless otherwise specified) was added until the polymer was fully dissolved (5–25 mL), and the mixture was filtered through cotton to remove the enzyme. The flask was washed with approx. 1 mL of CHCl₃ a further three times and all fractions were collected. The solvent was removed from the mixture via a rotary evaporator and transferred to a 50 mL Falcon tube containing ice cold methanol (35 mL) to precipitate the polymer. The mixture was vortexed and then centrifuged at 3,700 rpm for 10 min at 4 °C. The supernatant was removed, and 20 mL of ice-cold methanol was added, repeating the washing step a further two times.

Following the protocol reported in the literature, the diol monomer (2) was synthesised in two steps from vanillin (Mankar et al., 2019) (Scheme 1). The first double acetalization step with pentaerythritol was both non-hazardous and performed in a green solvent, therefore no changes were made to this procedure and structure (1) was obtained in 35% yield confirmed by NMR and HPLC-ESI-MS (Supplementary Figures S1–S3).

The second step involving hydroxyethylation of (1) with ethylene carbonate (EC) was first performed in N,N-dimethylformamide (DMF) as per the original procedure (Mankar et al., 2019), obtaining a yield of 81% (NMR and HPLC-ESI-MS seen in Supplementary Figures S4–S6). DMF is considered a solvent of concern due to acute toxicity and reprotoxicity issues (N,N-Dimethylformamide anhydrous, 2024), and industrial use was recently restricted in the EU (Sherwood et al., 2024). Hence, dimethyl sulfoxide (DMSO) was selected as a greener replacement as it is also a dipolar aprotic with a similar boiling point. The reaction proceeded in DMSO, but there was a significant amount of residual solvent present in the final product. DMSO is known to induce unfolding of enzymes (Zhang, 2022a), so should be removed to avoid any negative effect on CaLB’s activity in the following polycondensation reaction. Despite washing (2) with several liters of water and performing multiple rounds of lyophilization, approximately 2% by weight of DMSO remained. As a result, we investigated the synthesis of (2) under solvent free conditions, as EC can act as both reagent and solvent due to its low melting point (35 °C–38 °C) (Kinage et al., 2008). (2) was synthesised under these conditions using different excesses of EC; 0.15 excess similar to the solventless procedure, and 1.5 M excess. ¹H-NMR analysis of the products (Figure 2) shows the results of these syntheses compared with (2) synthesised in DMF (Figure 2A).

The use of a small excess of EC resulted in poor conversion, visible in Figure 2B. The ratio of peak a compared to peak 6 in Figure 2B. Indicates a conversion of around 70%, but it is likely that the partially hydroxyethylated compound will also be present, so the actual yield of (2) will be much lower. Comparing this to Figure 2C, (2) synthesised with 1.5 M excess EC has a much better conversion (>99%). Interestingly, further etherification of (2) was not observed despite the excess of EC, as no additional peaks were observed around 3.9 ppm. As a result, use of excess EC in place of DMF as a solvent can be considered a viable method to synthesise (2). Unreacted EC could also potentially be recovered from the aqueous waste and reused, which would also decrease the toxicity of the waste stream.

Previous polymer synthesis reactions with (2) have focused on the production of co-polymers (Mankar et al., 2019), illustrating the different mechanical and thermal properties that can be achieved with different amounts of (2) incorporated into the polymer. This work investigates the use of enzymes in this reaction, focusing on the synthesis of homopolymers and the effect that various diesters have on the molecular weight. Enzymatic polycondensations of (2) were carried out with both aliphatic and aromatic diesters for a total of 7 different polyesters (Figure 3). The aliphatic diesters of acids with a chain length from C4 to C10 including dimethyl succinate (DMSu), dimethyl adipate (DMA), and dimethyl sebacate (DMSe), and the aromatic monomers diethyl terephthalate (DET), diethyl isophthalate (DEI), diethyl pyridine-2,5-dicarboxylate (PD25) and diethyl pyridine-2,4-dicarboxylate (PD24) were used. Initial reactions were performed in diphenyl ether (DPE), an aprotic solvent commonly used in literature polycondensations catalysed by CaLB (Mahapatro et al., 2003; Pellis et al., 2020). The polymerization results can be seen in Table 1 and Figure 4 and the SI (Supplementary Figures S7).

Interestingly, the DMSu-(2) polymer had the highest number average molecular weight observed in this work, at 21.8 kDa (M_n). For polymers based on aliphatic diesters, there was a trend of increased M_n with decreased chain length of the diester. M_w however, was extremely similar, showing that these three aliphatic polyesters differ mainly in terms of molecular weight distribution, seen in Figure 4C. There is a precedent for use of DMA leading to higher molecular weight polymers when compared to DMSe (Pellis et al., 2018), but the DMSu based polymer having such a high molecular weight is unusual. However this previous work has a significant difference; reactions were performed under solventless conditions (Pellis et al., 2018), which has been shown to affect molecular weight (Mahapatro et al., 2003). For polycondensations conducted in solvent, solubility of both monomers as well as the polymer must be considered. It is possible that the DMSu-(2) polymer is particularly soluble in DPE, resulting in higher molecular weights. The increased reactivity of shorter chain diesters may also be a factor, as it has been noted in previous literature that the rate of acid catalysed esterification reactions decreased with increasing chain length (Liu et al., 2006). Steric hindrance was a notable factor in this previous work, which would also apply here.

For those polycondensations with two aromatic/semi-aromatic monomers, molecular weights are much lower, as expected. There are several examples in the literature of CaLB catalysed polycondensations utilizing aromatic monomers only (Komarova et al., 2024; Pellis et al., 2016; 2019; Salum et al., 2024), many of which resulted in low molecular weight polymers (M_n < 4,000 Da) (Pellis et al., 2016; Salum et al., 2024). This is generally attributed to the low reactivity of such monomers under the reaction conditions used, or poor solubility of the semi-aromatic polymers in the reaction media (Douka et al., 2018). Here an aromatic diester (either PD24, PD25, DEI or DET) was polymerised with a semi aromatic diol (monomer (2)), so this effect should be even more pronounced. It is important to highlight how solventless conditions are unsuitable for polycondensation of aromatic monomers as they often have high (>100 °C) melting points to be used in enzymatic reactions.

Another trend of note is that use of para-substituted diesters yielded higher molecular weight polymers compared to meta-substituted diesters, for both phthalic and pyridinedicarboxylic acid-based structures (Figure 4B). For PD24 and PD25, the opposite was observed when these monomers were polymerised with aliphatic diols (Komarova et al., 2024; Pellis et al., 2019). This was thought to be due to a difference in solubility of the synthesised polymers, so it is possible that PD25-(2) polyesters are simply more soluble in DPE compared to PD24-(2) polyesters. The same is true of DEI and DET based polyesters, as a previous work attributed low molecular weights of synthesised poly (1,6-hexanediol terephthalate) compared to poly (1,6-hexanediol isophthalate) to poor solubility of the former in the reaction media (in this case toluene) (Mezoul et al., 1996).

Overall, molecular weights were high, but the limiting factor seemed to be solubility of the polymers in DPE. By 96 h the reaction mixes for the DMA, DMSu and DMSe based polymers were highly viscous, hindering stirring, but it is possible that reaction solvent volume is a parameter that can be further optimised to further increase molecular weights. However, another issue was the precipitation of DPE together with the polymer in methanol (MeOH), during the workup stage despite their miscibility, which further complicates the workup.

Despite this, enzymatic polycondensations may be more suitable for producing polyesters based on (2) when compared to chemical catalysis. In a previous work, synthesis of polyesters from (2) and dimethyl terephthalate using a dibutyltin oxide (DBTO) catalyst resulted in polyesters with a M_n of 1.2 kDa (Mankar et al., 2019), whereas here a M_n of up to 8.7 was achieved for the same polymer. Of course, it must be taken into consideration that enzymatic polycondensations are run for much longer; 96 h in this work compared to 16 h in the work of Mankar et al. (2019). The low molecular weight in the DBTO catalysed polycondensation was attributed to steric hindrance from the bulky (2) monomer affecting esterification when conversion approached 100%.

Although DPE has been proven to be an excellent solvent for this reaction, it is also highly environmentally toxic and derived from petroleum. In addition, DPE contamination in the final polymer is significant, despite multiple washing steps (Supplementary Figures S8–S14). Due to these reasons, efforts were made to replace DPE as the reaction solvent, and several alternatives were tested in the polycondensation of DMA and (2).

The initial choice was Cyrene and other Cygnet derivatives, as solvents that are both bio-based and have previously been shown to be very suited for enzymatic polycondensation reactions (Warne et al., 2023). However, initial tests showed that (2) had limited solubility in Cyrene, Cygnet 2 and Cygnet 4, and indeed, polycondensations in these media resulted in low molecular weights (Table 2). Instead, anisole was selected as a solvent analogous to DPE; due to its lower boiling point it can be removed more readily during workup, and unlike DPE it is not an environmental hazard. Anisole has also been successfully used as reaction media for enzymatic polycondensations in the literature (Salum et al., 2024).

Minor changes were required in the procedure; due to the lower boiling point and hygroscopicity, the reaction was conducted under reflux and under N₂ for the first 6 h. A large proportion of anisole was lost to evaporation over the course of the reaction but still resulted in a polymer with a M_n of 21 kDa (Table 2). Slow evaporation of the anisole over 96 h would concentrate the monomers, possibly increasing mass transfer and therefore reaction rate.

Although the reaction of DMA and (2) in anisole was initially promising, it was repeated several times and the issue of reproducibility in terms of molecular weights can clearly be seen in Table 3. Although the initial polycondensation resulted in a polyester with a M_n of 21 kDa, subsequent polycondensations led to polymers having much lower molecular weights.

The variability in molecular weights is considerable and is reflected in the large standard deviation in the M_n, M_w and DP values. There is some variation in conversion, but despite reactions 1 and 3 both having a conversion >99%, molecular weights are still very different. Here, although the initial transesterification of (2) and DMA is similar in the early stage of the reaction and almost all of (2) reacts, the subsequent chain elongations proceed very differently. The reason for this is likely due to the used reaction medium. As previously mentioned, anisole is hygroscopic and even a small amount of water present may result in the hydrolysis of the polymer chains. Additionally, use of a reflux condenser means any water present is removed more slowly. Notably, the first synthesis performed in anisole from the freshly opened bottle (that is, fully anhydrous) gave the highest molecular weight by far (21.0 kDa M_n). Another potential cause is the reflux condensers used. Several different condensers were used, which had a noticeable effect on the evaporation rate of anisole. Use of waterless condensers resulted in extremely fast evaporation of anisole, and the matrix had visibly hindered stirring at a much earlier timepoint compared to those reactions run with conventional condensers. Faster evaporation of anisole could reduce mass transfer and significantly reduce the final molecular weights. The inconsistency in molecular weights of polyesters synthesised in anisole is likely to be due to both factors, the presence of water and variability in the evaporation rate of anisole.

Although there are more factors to consider when using anisole as a reaction solvent, the difficulties mentioned above are not insurmountable. To ensure anisole is completely anhydrous it should be kept under an inert atmosphere and in the presence of molecular sieves. An efficient condenser should also be used to minimize anisole evaporation over the course of the reaction. Finally, vacuum strength is a key parameter that should be further investigated for this reaction; ideally it would be sufficient to readily remove the MeOH byproduct whilst keeping all the low boiling organic solvent in the reaction flask.

Overall, although anisole showed some promise in enzymatic polycondensation reactions, under the reaction conditions used in this work, it is unsuitable for consistently producing high molecular weight polymers. This is mainly due to its lower boiling point and hygroscopicity, but the reaction conditions can be further optimised to minimize solvent loss, as well as maintain an anhydrous system. Although more suitable, polycondensations in DPE are not without their issues as the main problem with this synthesis is the precipitation of DPE together with the polymer during the workup stage, despite its solubility in methanol. Anisole was initially selected for its similarity to DPE but there are other potential solvent candidates that may have similar effects without the need of an anhydrous system, such as ethoxybenzene, propoxybenzene or tert-butoxybenzene. However, Salum et al. compared anisole and ethoxybenzene (among others) in the enzymatic polycondensation of PD24 and 1,8-octanediol, and found that use of anisole resulted in higher M_n (Salum et al., 2024). Yet these reactions were conducted under 360 mbar vacuum, meaning the methanol by-product is removed less rapidly, and for a shorter time. A systematic investigation of enzymatic polycondensations in lower bp media is needed with regards to vacuum strength, reaction time, and temperature, to fully exploit these solvents.