D-Excess-LaA Production Directly from Biomass by Trivalent Yttrium Species.

D-lactic acid (D-LaA) synthesis directly from actual biomass via chemocatalytic conversion has shown high potential for satisfying its enormous demand in widespread applications. Here we report yttrium (Y(III))-species-catalyzed conversion of xylose and raw lignocelluloses to LaA with the highest yield of 87.3% (20% ee to D-LaA, ee%=(moles of D-LaA - moles of L-LaA)/(moles of D-LaA + moles of L-LaA) × 100). Combining experiments with theoretical modeling, we reveal that [Y(OH)2(H2O)2]+ is the possible catalytically active species, enabling the unconventional cleavage of C3-C4 in xylulose and the subsequent dehydration of glyceraldehyde to pyruvaldehyde (PRA). The distinct interactions between hydrated-PRA and [Y(OH)2(H2O)2]+ species contribute to the formation of different enantiomers, wherein H-migration via re-face attack leads to L-LaA and that via si-face attack yields D-LaA. The lower strain energy barrier is the origin of excess D-enantiomer formation.

Introduction

Valuable chemical production directly from renewable lignocellulosic biomass is of great importance in seeking a sustainable future and bio-based economy, whereas chiral chemical production is still challenging owing to the quite complicated structure of biomass (Zhang et al., 2017, Liu et al., 2017, Farrán et al., 2015, Ennaert et al., 2016). Lactic acid (LaA), as one of the top 15 platform chemicals derived from carbohydrates, has been widely used in food and pharmaceuticals, especially in the production of polylactic acid (PLA), which has shown high potential in medical and clinical applications owing to its excellent biodegradability and biocompatibility (Castillo Martinez et al., 2013, Delidovich et al., 2016). The property of PLA is greatly dependent on the ratio of D-/L-LaA. Particularly, when poly D-LaA is blended with poly L-LaA to form a stereocomplex structure, it has been pointed out that the thermal stability, mechanical performance, and hydrolysis resistance can be significantly improved (Davachi and Kaffashi, 2015, Chen et al., 2016, Nagarajan et al., 2016, Lasprilla et al., 2012). The booming of PLA industry greatly stimulates the development of LaA production (Dusselier et al., 2013). However, there is a huge gap between the production capacity and actual output of LaA, and bulk commercial supply of LaA with an excess of D-enantiomer is still unavailable (Dusselier et al., 2013).

Currently, over 90% of commercial LaA is produced by fermentation of carbohydrates with mainly L-LaA in excess, whereas it is still hard to produce D-LaA with the same productivities and purities (Baek et al., 2016, Eş et al., 2018). The high cost of enzymes and waste disposal in fermentation, in addition to the restricted feedstock, including only hexose sugars and edible di-/polysaccharides, greatly inhibit its widespread application (Desguin et al., 2014, Abdel-Rahman et al., 2013, Maki-Arvela et al., 2014). Novel chemocatalytic methods using homogeneous (Li et al., 2017, Sharninghausen et al., 2014) or heterogeneous (Coman et al., 2015, Holm and Taarning, 2010, Besson et al., 2014) catalysts for LaA production therefore have attracted increasing attention in recent years, owing to the high production capacity, widened starting resources even including non-edible cellulose and actual biomass (Wang et al., 2013, Deng et al., 2018, Dusselier and Sels, 2014, Lin et al., 2013, Gallezot, 2012, Tuck et al., 2012, Dapsens and Mondelli, 2015), easily available catalysts, as well as the avoided formation of waste salts. However, the recent advances are mainly focused on the use of non-edible cellulose. To enhance LaA productivity and meet the rapid increase of LaA demand, it is essential to take full advantages of all carbohydrates in biomass, including not only hexose but also pentose (Mäkiarvela et al., 2011). Nevertheless, pentose-based feedstock is rarely employed for LaA production. Even in the limited reports, the yield of LaA (<45%) is unsatisfactory, because the catalyst is usually not as effective as that for hexose conversion (Yang et al., 2015, Yang et al., 2016a, Yang et al., 2016b). Moreover, it is regrettable that the production of chiral LaA directly from actual biomass via chemocatalysis cannot be found in all the references we can find. In the existing reports, there is little information on the distribution of LaA enantiomers via chemocatalysis especially when pentose- and hexose-based carbohydrates are used as feedstock. Here, we achieve the production of D-excess-LaA with outstanding yield using a simple achiral yttrium (Y(III)) ion directly from pentose-based feedstock and actual biomass, that is, corn straw, eliminating the need of complex chiral catalyst in conventional asymmetric synthesis (Otocka et al., 2017, Feng et al., 2017) and the extraction of monosaccharide from raw biomass. We use 13C-C1-labeled xylose and D2O as probe molecules to gain deep insights into the active species and reaction mechanism, by combining experiments with quantum chemical modeling. Significantly, we have also proposed the origin of D-/L-enantiomer formation and explained the reason why more D-LaA is produced with the assistance of (Y(III)) ion, for the first time.

Results and Discussion

Catalytic Activity of Y(III)

We first investigated the catalytic activity of Y(III) for xylose conversion to LaA (Figure S2). We found that Y(III) greatly enhanced xylose conversion, giving as high as 87.3% yield of LaA with 20% ee value to D-LaA, ee%=(moles of D-LaA - moles of L-LaA)/(moles of D-LaA + moles of L-LaA) × 100 at 493 K for 0.5 h with limited by-products (Figure S3), whereas only limited LaA (3.2%) with a D-/L-LaA ratio of ∼1:1 was obtained for the blank reaction. We also investigated the catalytic efficiency of other metal ions, including Cr(III), Al(III), Pb(II), and Sn(II) (Table 1). Similarly, more D-LaA than L-LaA was obtained. This is different from the fermentation route wherein more L-LaA is usually obtained (Abdel-Rahman et al., 2013). Although Pb(II) and Sn(II) had been proved to show high efficiency for hexose conversion to LaA in the literature (Wang et al., 2013, Deng et al., 2018), they failed to effectively promote xylose conversion to LaA in this study with yields of only 41.7% and 45.0%, respectively. Y(III) outperformed other selected metal ions, achieving the highest LaA yield among these selected metal ions and those used in the literature (Table S1). To extend the application of the Y(III) system, we further employed xylan, microcrystalline cellulose (M-cellulose), and corn straw as feedstock. Yields of LaA were 72.5% (16% ee) and 57.0% (12% ee) from xylan (Figure S4) and M-cellulose, respectively, whereas 61.8% (12% ee) yield was obtained based on the amounts of cellulose and hemicellulose from corn straw (Table 1). This indicates that Y(III) exhibits multiple catalytic performances, integrating the depolymerization of hemicellulose and cellulose with the next monosaccharide conversion to yield LaA. Furthermore, we performed the separation and recycling of Y(III) ion from reaction solution by a chemical precipitation method (Note S1 and Figure S5) and found that the recycled Y(III) showed almost the same catalytic activity within three runs (Table S2).

Table 1

The Catalytic Performance of Metal Ions on the Conversion of Xylose and Biomass to LaAa

Reactant	Xylose	Xylose	Xylose	Xylose	Xylose	Xylan	M-Cellulose	Corn Straw
Catalyst	Y(III)	Cr(III)	Al(III)	Pb(II)	Sn(II)	Y(III)	Y(III)	Y(III)
Yield (%)	87.3	26.6	34.2	41.7	45.0	72.5	57.0	61.8
ee value (%)	20	17	17	23	26	16	12	12
Carbon balance (%)	70	63	61	65	65	71	67	65

Reaction conditions: 0.20 g xylose (26.7 mM) or 0.20 g M-cellulose or 1.0 g corn straw; 50 mL H2O; 6.6 mM YCl3; N2, 2 MPa; temperature, 493 K; reaction time, 0.5 h. The high-performance liquid chromatography analysis results of the mixture of D/L-lactic acid is shown in Figure S1.

Determination of Active Species

Considering the high efficiency of Y(III), we conducted several control experiments with HCl as catalyst and xylose as feedstock to identify the active catalytic species in Y(III) system. With similar pH value (pH = 5.5) to YCl3 aqueous solution, HCl showed little activity for LaA formation (total yield ≤ 1.4%) but sharply increased furfural (FF) yield (Figure S6B), accompanied by a low carbon balance (Figure S7). Similarly, by the addition of equivalent HCl to the Y(III) system, LaA yield sharply declined from 87.3% to 57.1%, and continuously increasing the HCl amount to 120 mM gave only 3.6% yield of LaA (Figure S8, Table S3). We therefore consider that Y(III) predominantly contributes to the effective promotion of LaA yield, whereas the functions of both H+ generated from YCl3 hydrolysis and Cl− are relatively weak. We also conducted a series of experiments with Y(III) in neutral and alkaline media (Figure S9). It was found that the yield of LaA sharply decreased by increasing the pH value from 5.5 to 10.5, and only 10.7% yield of LaA was obtained at pH = 10.5, possibly attributed to the formation of Y(OH)3 precipitate. Therefore, the highest LaA yield was achieved at pH = 5.5 when Y(III) was autohydrolyzed in neutral aqueous solution. ESI-MS (electrospray ionization mass spectrometry) analysis of YCl3 aqueous solution exhibited obvious peaks of hydrolyzed Y(III) species [Y(OH)2(H2O)n]+ (n = 0, 2, 4) (Figure S10A). Further introduction of xylose (Figure S11) to YCl3 aqueous solution also gave [Y(OH)2(H2O)n]+ intermediates, which combined with other chemicals such as xylose and FF (Figures S10B and S12), and ESI-tandem MS analysis gave obvious fragment peaks of [Y(OH)2(H2O)n]+ (Figure S13). Theoretical computation also demonstrated that the formation of these intermediates was feasible under the reaction conditions (Figure S10G). HCl addition could suppress the hydrolysis of Y(III), which was proved by the gradually disappearing peak of [Y(OH)2(H2O)n]+ by ESI-MS analysis with increasing HCl amount (Figure S14). All these results suggest the important function of [Y(OH)2(H2O)n]+ species for xylose conversion. The results of theoretical computation indicated that the stabilization energy of Y(III) hydrolyzing to [Y(OH)2(H2O)2]+ (26.9 kcal mol−1, Note S2) was the highest among these species [Y(OH)2(H2O)n]+ (n = 0–4) (Figure S15). It is therefore plausible to assume that [Y(OH)2(H2O)2]+ is the active species in xylose conversion to LaA.

The Performance of Active Species on Xylose Conversion to PRA

We then attempted to reveal the performances of active species on LaA formation from xylose. In the process of xylose conversion to LaA with the assistance of Y(III), we observed the successive formation of xylulose and triose (dihydroxyacetone [DHA], glyceraldehyde [GLA], and pyruvaldehyde [PRA]) intermediates. We therefore speculate that LaA formation may pass through xylose isomerization to xylulose and the consequent C-C cleavage to trioses. It is reported that the bottleneck in LaA production from pentose is the selective splitting of C-C bond, and two distinct C-C cleavage mechanisms have been proposed in the literature (Yang et al., 2015, Holm et al., 2012). One involves C2-C3 cleavage in aldopentose via a retro-aldol condensation reaction yielding GLA and glycolaldehyde. This is a widely accepted pathway. The other involves C3-C4 cleavage in ketopentose via a retro-aldol condensation giving DHA and glycolaldehyde. To reveal the performance of [Y(OH)2(H2O)2]+ on C-C cleavage in xylose, we employed 13C-C1-labeled xylose as substrate and D2O as solvent. In the presence of Y(III), almost all labeled 13C was transferred into the methyl carbon atom of LaA after reaction (Figures 1C and S16), whereas no 13C was found in glycolic acid (Figure S17). Without catalyst, several by-products with labeled 13C such as formic acid (Figure S18) and FF (Figure S19) were detected after reaction (Figure 1B), whereas no 13C-labeled LaA was observed. This demonstrates that [Y(OH)2(H2O)2]+ can facilitate C3-C4 cleavage in xylose under the reaction conditions in this work (Figure 2, Scheme S1), which is different from the result in the literature wherein C2-C3 cleavage predominantly contributes to LaA formation from xylose (Yang et al., 2015, Yang et al., 2016a, Yang et al., 2016b).

Figure 1

13C NMR Spectra of Different Mixture and the Schematic Diagram for the flow of 13C Labeled Atoms from 13C1-xylose with/without Y(Ⅲ)

(A–D) 13C NMR spectra of (A) mixture of 13C1-xylose with D2O at room temperature, (B) liquid products of 13C1-xylose conversion in D2O without catalyst, and (C) liquid products of 13C1-xylose conversion with Y(III). (D) Schematic diagram for the flow of 13C-labeled atoms with/without Y(Ⅲ). Reaction conditions: 0.20 g (26.7 mM) 13C1-xylose; 50 mL D2O; 6.6 mM YCl3 (if catalyst was used); N2, 2 MPa; 0.5 h; 493 K.

Figure 2

Scheme of 13C1-Xylose Conversion to 13C-LaA with Y(III)

Bond lengths are reported in angstroms.

Given the cleavage of C3-C4 in xylulose, the resulting DHA can be transformed to GLA in water by means of keto-enol tautomerization. To further reveal the performance of [Y(OH)2(H2O)2]+ on triose conversion to LaA, we conducted control experiments using trioses (DHA, GLA, PRA) as substrates. We found that Y(III) addition greatly enhanced the conversion of these trioses to LaA with much higher yield (≥92%, Figure S20 and Table S4) compared with that without catalyst (≤12%). We then used D2O as solvent in place of H2O (Figures S21 and S22 and Table S5) and observed the formation of LaA with deuterated methyl group when DHA and GLA were used as substrates (Figure S21, Table S5 entries 5 and 6), in agreement with the result of Sels et al. (Pescarmona et al., 2010). Similarly, LaA with deuterated methyl was obtained from xylose conversion under the same conditions (Figure S23, Table S5 entry 2). Nevertheless, no LaA with deuterated methyl was detected to be derived from DHA and GLA without Y(III) (Table S5 entries 8 and 9). Just heat treatment of LaA-D2O solution with/without Y(III) (Table S5 entries 3 and 4) cannot induce the formation of LaA with deuterated methyl group. These results suggest that DHA/GLA undergoes a dehydration reaction to form PRA with the assistance of Y(III). A detailed time course analysis showed that LaA yield sharply increased once PRA began to decrease in the presence of Y(III) (Figure S24). In combination with the fact that PRA exhibits the highest conversion rate to LaA among these three trioses, we infer that PRA is the most advantageous triose for LaA formation. However, in the case of PRA as substrate with Y(III) as catalyst, no LaA with deuterated methyl group was observed (Figure S21C, Table S5 entry 7), implying the H-migration from C1 to C2 in PRA before LaA formation. In addition, no D atom was connected with C2 of LaA in all cases, implying that the H atom in C2 of LaA originated from the aldehyde group. This further confirms H-migration from C1 to C2 (Scheme S2). Combining with the above-mentioned experimental results, we tentatively propose that LaA formation from xylose with the aid of Y(III) may proceed through the following pathways (Figure 3A): (1) xylose isomerization to xylulose; (2) C3-C4 cleavage of xylulose to DHA and DHA isomerization to GLA, along with a side reaction involving xylose dehydration to FF; (3) DHA/GLA dehydration to PRA via ketone-aldehyde tautomerism; and (4) PRA hydration and H-migration from C1 to C2, finally leading to LaA.

Figure 3

The Proposed Reaction Pathway of Xylose to PRA and the Corresponding Energy Profiles

(A and B) The proposed reaction pathway of xylose conversion to PRA catalyzed by Y(III) species (A) and the corresponding energy profiles (B) (in kcal mol−1, all the mentioned optimized geometries are shown in Figure S25 and the corresponding Cartesian coordinates are showed in Table S8).

We next conducted quantum chemical computation, aiming at getting more insights into the reaction mechanism at the molecular level, as well as explaining how [Y(OH)2(H2O)2]+ species have a decisive influence on the catalytic reaction. We found that the results of quantum chemical computation were consistent with those in experiments. The addition of [Y(OH)2(H2O)2]+ significantly reduced the activation barrier of xylose isomerization to xylulose from 32.9 (non-catalytic reaction) to 17.4 kcal mol−1 (Table S6). We attribute it to a different H-migration from C2-O to C1-O of xylose aided by a water molecule between [Y(OH)2(H2O)2]+ and xylose (Figure S26), which is more effective than the conventional direct proton transfer in references (Yang et al., 2016a, Yang et al., 2016b, Román-Leshkov et al., 2010). Subsequently, [Y(OH)2(H2O)2]+ species continued to induce and accelerate C3-C4 cleavage of xylulose with an 18.8 kcal mol−1 energy barrier, passing through the similar H-shift from C4-OH to C2-O helped by a water molecule (Figure S27). 1-Ene-propanetriol (4 in Figure 3, captured in ESI-MS spectra [Figure S28]) was generated after C3-C4 cleavage, which was easily converted to DHA with an 11.3 kcal mol−1 energy barrier. The remaining C2 compound was transformed to glycolic aldehyde. The energy barrier for the next aldehyde-ketone isomerization of DHA to GLA also decreased from 32.6 (non-catalytic pathway) to 19.7 kcal mol−1 (Figure S29). Without catalyst, huge energy barrier (ΔG≠ = 78.9 kcal mol−1) was required for C3-C4 cleavage (Figure S30), and the direct breaking of C2-C3 in xylose via the conventional retro-aldol condensation also showed a 76.4 kcal mol−1 energy barrier (Figure S31). The next GLA dehydration to PRA is widely considered as the rate-limiting step in the process of conversion of trioses to LaA in the literature (Rasrendra et al., 2010), due to the high-energy barrier (59.6 kcal mol−1) without catalyst. In this study, [Y(OH)2(H2O)2]+ significantly promoted GLA dehydration to PRA through three steps: (1) –OH in [Y(OH)2(H2O)2]+ deprived H atom on C2 to form H2O; (2) C3-hydroxyl protonation and H2O leaving, yielding chemical 10; and (3) isomerization of 10 to PRA. The energy barriers of these three steps were 10.7, 7.0, and 5.9 kcal mol−1, respectively (Figure S32 and Table S7).

The Origin of D-/L-Enantiomer Formation

The following PRA conversion to LaA is commonly considered to be an intra-molecular Cannizzaro reaction, which is easier under alkaline conditions (Yang et al., 2016a, Yang et al., 2016b, Dong et al., 2016). We also obtained LaA from PRA with high yield when replacing Y(III) with equivalent amount of NaOH (Table S9). Our calculations showed that LaA formation from PRA might undergo sequential hydration of aldehyde group and H-migration from C1 to C2 (Figure 4). With the aid of [Y(OH)2(H2O)2]+, PRA hydration showed lower activation energy barrier than that via non-catalytic pathway (Table S6). The next H-migration proceeded via three steps: (1) the H transfer from C1-OH to [Y(OH)2(H2O)2]+; (2) the next H-migration from H2O in [Y(OH)2(H2O)2]+ to C2, that is, water in catalyst assists H transfer; (3) another H from C1 directly to C2 is the crucial step for the formation of different LaA enantiomers. We found that the H-migration from C1 to C2 via si-face attack of hydrated PRA could lead to D-LaA, whereas re-face attack gave L-LaA (Figure 4C). Without catalyst, the activation energy barrier of D-LaA formation was equal to that of L-LaA formation (35.1 kcal mol−1, Figures 4B and 4E). This is possibly ascribed to the similar steric hindrance of si-face attack to that of re-face attack, thus giving a racemic mixture. This is well in agreement with the experimental result. By contrast, in the presence of [Y(OH)2(H2O)2]+, hydrated PRA could interact with [Y(OH)2(H2O)2]+ via two different ways, finally yielding two distinct intermediates C-im10-D and C-im10-L, both of which could result in the formation of D- and L-LaA enantiomers via different transition states (Figures 4A and 4D). Time-of-flight (TOF) analysis (Table S10) demonstrates that the TOF value of catalytic cycle to D-LaA via C-ts9-D is much higher than that to L-LaA via C-ts9-L2, and the pathway to L-LaA via C-ts9-L also shows much higher TOF value than that to D-LaA via C-ts9-D2. We then compared these two predominant pathways and found that D-LaA formation via C-ts9-D exhibits much higher TOF value than L-LaA formation via C-ts9-L. However, we found that H-migration in C-ts9-D via si-face attack showed similar steric hindrance to that in C-ts9-L via re-face attack. To investigate the actual chirality-determining factor, we further conducted activation strain analysis (Table S11) and found that the strain energy barrier for C-ts9-D (97.0 kcal mol−1) was much lower than that of C-ts9-L (107.5 kcal mol−1), which led to lower ΔE≠ of C-ts9-D than that of C-ts9-L, although the interaction energy of C-ts9-D between catalyst and hydrated PRA was higher than that of C-ts9-L. Thus, it seems that the reaction cycle via C-ts9-D is more competitive than that via C-ts9-L, with a 0.7 kcal mol−1 lower activation energy barrier (Figure 4A). We therefore speculate that the lower strain energy between the catalyst and reactant enables the formation of a little more D-LaA with a calculated ee value of 34.3% (Figure 4E), according to Curtin-Hammett principle (Schneebeli et al., 2009), which is close to the 20% ee value obtained in the experiment. In the process of triose to D-/L-LaA, the final H-migration from C1 to C2 becomes the rate-limiting step, which is also the chirality-controlling step for xylose conversion to D-/L-LaA in this catalytic system.

Figure 4

The Proposed Reaction Pathway on the Formation of Lactic Acid Enantiomers from PRA

(A–E) (A) the energy profiles for catalytic pathway; (B) the energy profiles for the key chirality-controlling step in non-catalytic pathway; (C) the graphic for the formation of chiral center in non-catalytic pathway; (D) the proposed reaction pathway for the formation of lactic acid enantiomers from PRA catalyzed by Y(III) species; (E) the relative Gibbs energy (ΔG, kcal mol−1) and the corresponding energy difference (ΔΔG, kcal mol−1) of competing transition in chirality-controlling step between catalytic pathway and non-catalytic pathway and the corresponding stereo-selectivity (ee%) (all the mentioned optimized geometries in Figure 4A are shown in Figure S33 and the corresponding Cartesian coordinates are shown in Table S8). Bond lengths are reported in angstroms.

Conclusions

Y(III) outperforms other selected metal ions, achieving an outstanding LaA yield from xylose and corn straw. [Y(OH)2(H2O)2]+, as the possible active species, primarily contributes to C3-C4 cleavage of xylose, leading to LaA formation via an unconventional pathway. The formation of D-/L-enantiomer is significantly dependent on the different interactions between hydrated PRA and catalytically active species. Water coordinated with Y(III) and the reactant itself construct a chiral environment, and the function of water for helping chiral catalysis might be popular in natural life process. The lower strain energy results in enantioselectivity favoring D-LaA. To the best of our knowledge, this is the first detailed description of the enanotioselectivity of LaA enantiomer obtained from actual raw biomass containing pentose and hexose via chemocatalytic conversion, and also the first attempt to reveal the formation mechanism of LaA enantiomers. Although more insights need to be deeply investigated in future work, we think this work will give some guidance in controlling the enanotioselectivity of LaA enantiomer directly from renewable lignocelluloses. The insights in this work provide some clues for the development of new synthetic strategies to produce chiral compounds, that is, enantioselective production of chiral molecule can be achieved by controlling the interaction of the reactants with an achiral catalyst. The information in the present work can also help us to understand the formation mechanism of chiral compounds in nature.

Limitations of the Study

The ee value of D-LaA needs further improvement, although a little more D-enantiomer can be obtained from biomass by chemocatalysis. This increases the difficulty of separation and purification of LaA enantiomers. It is therefore necessary to design chiral catalysts/ligands to improve the enantioselectivity in further work.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.