**Catalytic Conversion of** α**-Pinene to High-Density Fuel Candidates Over Stannic Chloride Molten Salt Hydrates**

#### **Seong-Min Cho <sup>1</sup> , June-Ho Choi <sup>1</sup> , Jong-Hwa Kim <sup>1</sup> , Bonwook Koo <sup>2</sup> and In-Gyu Choi 3,4,\***


Received: 18 September 2020; Accepted: 22 October 2020; Published: 26 October 2020

**Abstract:** The synthesis of dimeric products from monoterpene hydrocarbons has been studied for the development of renewable high-density fuel. In this regard, the conversion of α-pinene in turpentine over stannic chloride molten salt hydrates (SnCl4·5H2O) as a catalyst was investigated, and the reaction products were analyzed with gas chromatography/flame ionization detector/mass spectrometer (GC/FID/MS). Overall, the content of α-pinene in a reaction mixture decreased precipitously with an increasing reaction temperature. Almost 100% of the conversion was shown after 1 h of reaction above 90 ◦C. From α-pinene, dimeric products (hydrocarbons and alcohols/ethers) were mostly formed and their yield showed a steady increase of up to 61 wt% based on the reaction mixture along with the reaction temperature. This conversion was thought to be promoted by Brønsted acid activity of the catalyst, which resulted from a Lewis acid-base interaction between the stannic (Sn(IV)) center and the coordinated water ligands. As for the unexpected heteroatom-containing products, oxygen and chlorine atoms were originated from the coordinated water and chloride ligands of the catalyst. Based on the results, we constructed not only a plausible catalytic cycle of SnCl4·5H2O but also the mechanism of catalyst decomposition.

**Keywords:** renewable fuel; high-density fuel; α-pinene dimerization; turpentine; stannic chloride molten salt hydrates

#### **1. Introduction**

Aspirations for alternative biofuel have emerged vigorously worldwide, amid increasing concern about fossil fuel sustainability and environmental pollution. In this respect, cellulosic biomass-derived chemicals such as furfural, 5-hydroxymethylfurfural, 2-methylfuran, levulinic esters, and angelica lactone have attracted considerable attention as renewable resources for the sustainable production of biofuels [1–3]. Another promising renewable chemical in order to produce synthetic fuel is turpentine. Turpentine can be obtained by either the distillation of oleoresins or the kraft pulping process as a side-product and it is mainly composed of α-pinene and lesser amounts of isomers (β-pinene, camphene, and limonene), which all belong to monoterpene hydrocarbons [4]. Because fuel properties depend on what consists of the fuel, the fact that hydrocarbons make up the majority of turpentine is an advantage of itself as the resource of biofuels [5]. It has been reported that the blends of monoterpenes and petroleum-based fuel could be utilized as drop-in fuels [6,7].

When it comes to the application of fuel to an aerospace field and ramjet engines, however, drop-in fuels are required for much higher energy density than neat turpentine. To synthesize high-energy density fuel from turpentine (α-pinene), dimerization over acid catalysts has been devised to increase density, thereby increasing energy density [5]. With the advantage of facile separation, various heterogeneous Brønsted acid catalysts have been studied to carry out the dimerization [8,9]. Because the four-membered ring of α-pinene is labile to acidic conditions, the ring is easily opened or extended to five-membered ring. This furnishes isomers including both monocyclic monoterpene hydrocarbons such as limonene and bicyclic monoterpene hydrocarbons such as camphene. Thus, the acid-catalyzed dimerization of α-pinene always accompanies the isomerization, which is followed by inevitable cross-dimerization and makes the reaction mixture consist of not only the isomers of α-pinene but also the various dimers [8]. It has been found that a dimer fraction has too high viscosity to apply it to a field where the fuel is exposed to extremely low temperatures [10]. To come up with this problem, blending the dimer faction with monoterpene hydrocarbons or the partial dimerization of α-pinene has been devised [5].

Inorganic molten salts hydrates, such as ZnCl2·3H2O, are the hydrated form of inorganic salts that easily melted like butter at an elevated temperature. A chemical reaction system exploiting the inorganic molten salts hydrates as a reaction medium has been applied to several processes, particularly biopolymer conversion. It has been reported that an acidic LiBr salt hydrates can hydrolyze both the β-1,4 glycosidic bond of cellulose [11] and the aryl ether bond of lignin [12] and co-produce boromomethylfurfural and furfural directly from lignocellulosic biomass within the biphasic system [13]. In addition to LiBr salt hydrates, ZnCl<sup>2</sup> and SnCl<sup>4</sup> salts hydrates have been used for the synthesis of isosorbide from cellulose [14] and 5-hydroxymethylfurfural from fructose, inulin, glucose, and chitosan [15–18].

Herein, we focused on the acidic nature of stannic chloride molten salt hydrates (SnCl4·5H2O). [SnCl4L2] complexes in anhydrous conditions have been reported as Lewis acid-assisted Brønsted acid catalysts when the L<sup>2</sup> are 2′ -methoxy-[1,1′ -binaphthalen]-2-ol (BINOL-Me), 2′ -isopropoxy- [1,1′ -binaphthalen]-2-ol (BINOL-*i*Pr), 2-methoxyphenol (guaiacol), or 2,6-dimethoxyphenol (syringol), because they can coordinate to the stannic (Sn(IV)) center via the vicinal hydroxy and alkoxy groups in a bidentate manner [19–22]. Therefore, we have expected that the coordinated water ligands of SnCl4·5H2O can also act as Lewis acid-assisted Brønsted acids for the catalytic conversion of α-pinene. It has been remarkable that SnCl4·5H2O can furnish the dimerized products (up to 61 wt% based on the reaction mixture) from turpentine in a solvent-less condition. However, the decomposition of the catalyst results in oxygen-containing or chloride-containing products. To overcome this limitation and ponder the improvement of the reaction system, we have discussed the mechanism of the catalytic conversion of α-pinene over SnCl4·5H2O.

#### **2. Materials and Methods**

#### *2.1. Materials*

All purchased chemicals were used without further purification. Turpentine (α-pinene ≥ 92% by GC) was purchased from Junsei Chemical Co., Ltd. (Tokyo, Japan). Stannic chloride molten salt hydrates (SnCl4·5H2O, 98%) and n-hexane (95%) were obtained via Samchun Pure Chemicals Co., Ltd. (Seoul, Korea).

#### *2.2. General Procedure for Catalytic Conversion of* α*-pinene by Stannic Chloride Molten Salt Hydrates*

We set reaction time as 1 h and considered the effect of the reaction temperature (60, 70, 80, 90, 100, and 110 ◦C) or set reaction temperature as 70 ◦C and considered the effect of reaction time (1, 3, and 5 h) because, upon the reaction temperature crossed, the melting point of stannic chloride molten salt hydrates (SnCl4·5H2O), α-pinene, quickly disappeared from the reaction mixture. In a typical experiment, 10 g of turpentine and 1 g of SnCl4·5H2O was added to 50-mL double-neck

round flask equipped with a magnetic Teflon coated stirrer, a thermometer, and a reflux condenser. After being purged with nitrogen gas, the reactor was then loaded on a preheated aluminum heating block and stirred smoothly for 1 h. The temperature of a reaction mixture was reached at the desired temperature (±3 ◦C) in less than 5 min. Upon completion of the reaction, the reactor was then removed from the heating block and immediately cooled to room temperature with a water bath. After cooling, the crude reaction mixture was diluted in 100 mL of n-hexane and filtered over a celite pad to remove metal compounds. The solvent was removed under reduced pressure. The purified product was analyzed by Agilent Technologies 7890B gas chromatograph equipped with a DB-5ms column (30 m × 250 µm, 0.25 µm thickness), a 5977A mass spectrometer detector, and a flame ionization detector (Agilent-Technologies, Seoul, Korea). The yield of products was calculated based on the below equation on the basis of the internal standard method (tridecane).

$$\text{Yield of product } i \text{ (\%)} = \frac{\text{Weight of product } i \text{ after reaction}}{\text{Initial weight of } \alpha - \text{pinene}} \times 100 \tag{1}$$

#### **3. Results and Discussion**

#### *3.1. Catalytic Conversion of* α*-Pinene over Stannic Chloride Molten Salt Hydrates*

As previously mentioned in the introduction, what we have expected was that the Lewis acid-assisted Brønsted acidic nature of stannic chloride molten salt hydrates (SnCl4·5H2O) would work as a catalyst. In accordance with this expectation, the catalytic amount of SnCl4·5H2O was active for the conversion of α-pinene in a solvent-less condition. As shown in Figure S1, the reaction products could be divided into four main groups (monomeric hydrocarbon, heteroatom-containing monomeric, dimeric, and trimeric product regions) on the basis of retention time in gas chromatography/flame ionization detector (GC/FID) chromatogram and molecular ion peaks in mass spectra. Figure S2A,B show the representative mass spectra of dimeric and trimeric hydrocarbons including molecular ion peaks at *m*/*z* 272 [C20H32] <sup>+</sup> and *m*/*z* 408 [C30H48] <sup>+</sup>, respectively.

Figure 1A shows that α-pinene, which is the main constituent of turpentine, was consumed precipitously with increasing reaction temperature and the content thereof eventually stands at 0% in the reaction mixture heated above 90 ◦C. For the cases of dimeric and trimeric products, the contents increased approximately four-fold and eight-fold when the reaction temperature increased from 60 to 110 ◦C, standing at 60.9% and 15.4%, respectively, at 110 ◦C. There was a large discrepancy in the conversion of α-pinene and yields of dimeric and trimeric products, which denotes the presence of other monomeric products formed from α-pinene. Remarkably, our results differ from those of Nie et al. who suggested that, when phosphotungstic acid supported on MCM-41 was applied, the isomerization of α-pinene is the main reaction pathway at 100 ◦C and the dimeric products are formed mainly from the isomers at 160 ◦C [23]. As shown in Figure 1B, even at 70 ◦C of reaction temperature, 100% conversion of α-pinene can be achieved by increasing the reaction time. Given that the dimeric products were produced without α-pinene after 3 h of reaction, it can be thought that the monomeric products from α-pinene also participated in the dimerization [8].

Unfortunately, several mass spectra involving an unexpected molecular ion peak at *m*/*z* = 290 (Figure S2C) were extracted from the dimeric product region in Figure 1B. Given the difference between the *m*/*z* of molecular ions, the molecular ion having *m*/*z* = 290 could be described as [C20H34O]+, which means that oxygen-containing dimeric products, possibly dimeric alcohols or ethers, were also produced during the reaction. In the dimerization of monoterpenes over acidic catalysts, dimeric products having *m*/*z* = 290 as a molecular ion have often been reported. Nie et al. suggested a dimeric alcohol product in the dimerization of crude turpentine [23]. Moreover, heteropoly acid-catalyzed conversion of camphene furnished diisobornyl ether in a moderate yield and good selectivity [24].

Appl. Sci. 2020, 10, x FOR PEER REVIEW 3 of 9

stirred smoothly for 1 h. The temperature of a reaction mixture was reached at the desired temperature (±3 °C) in less than 5 min. Upon completion of the reaction, the reactor was then removed from the heating block and immediately cooled to room temperature with a water bath. After cooling, the crude reaction mixture was diluted in 100 mL of n-hexane and filtered over a celite pad to remove metal compounds. The solvent was removed under reduced pressure. The purified product was analyzed by Agilent Technologies 7890B gas chromatograph equipped with a DB-5ms column (30 m × 250 μm, 0.25 μm thickness), a 5977A mass spectrometer detector, and a flame ionization detector (Agilent-Technologies, Seoul, Korea). The yield of products was calculated based on the below

Weight of product i after reaction

As previously mentioned in the introduction, what we have expected was that the Lewis acidassisted Brønsted acidic nature of stannic chloride molten salt hydrates (SnCl4·5H2O) would work as a catalyst. In accordance with this expectation, the catalytic amount of SnCl4·5H2O was active for the conversion of α-pinene in a solvent-less condition. As shown in Figure S1, the reaction products could be divided into four main groups (monomeric hydrocarbon, heteroatom-containing monomeric, dimeric, and trimeric product regions) on the basis of retention time in gas chromatography/flame ionization detector (GC/FID) chromatogram and molecular ion peaks in mass spectra. Figure S2A,B show the representative mass spectra of dimeric and trimeric hydrocarbons including molecular ion

, respectively. Figure 1A shows that α-pinene, which is the main constituent of turpentine, was consumed precipitously with increasing reaction temperature and the content thereof eventually stands at 0% in the reaction mixture heated above 90 °C. For the cases of dimeric and trimeric products, the contents increased approximately four-fold and eight-fold when the reaction temperature increased from 60 to 110 °C, standing at 60.9% and 15.4%, respectively, at 110 °C. There was a large discrepancy in the conversion of α-pinene and yields of dimeric and trimeric products, which denotes the presence of other monomeric products formed from α-pinene. Remarkably, our results differ from those of Nie et al. who suggested that, when phosphotungstic acid supported on MCM-41 was applied, the isomerization of α-pinene is the main reaction pathway at 100 °C and the dimeric products are formed mainly from the isomers at 160 °C [23]. As shown in Figure 1B, even at 70 °C of reaction temperature, 100% conversion of α-pinene can be achieved by increasing the reaction time. Given that the dimeric products were produced without α-pinene after 3 h of reaction, it can be

+

Initial weight of α-pinene <sup>×</sup> <sup>100</sup> (1)

equation on the basis of the internal standard method (tridecane).

3.1. Catalytic Conversion of α-Pinene over Stannic Chloride Molten Salt Hydrates

and m/z 408 [C30H48]

Yield of product i (%) =

+

3. Results and Discussion

peaks at m/z 272 [C20H32]

**Figure 1.** Effect of reaction temperature and reaction time on the conversion of α-pinene and yields of dimeric and trimeric products. The reaction was carried out (**A**) for 1 h and (**B**) at 70 ◦C. suggested a dimeric alcohol product in the dimerization of crude turpentine [23]. Moreover, heteropoly acid-catalyzed conversion of camphene furnished diisobornyl ether in a moderate yield and good selectivity [24].

Figure 2A,B show the main hydrocarbon product distribution in the monomeric hydrocarbon region in Figure S1B as a function of the reaction temperature. These monomeric hydrocarbons could be divided into two groups based on the tendency in the content thereof, according to the reaction temperature. Limonene, terpinolene, α-terpinene, γ-terpinene, and camphene except for tricyclene presented a sudden decrease in their yield above 90 ◦C, meaning their participation in dimerization. On the other hand, the contents of menthenes and *p*-cymene increased with increasing reaction temperature until 100 ◦C. These have been reported to form by the disproportionation of *p*-menthadienes [25]. In our results, the similar contents of menthenes and *p*-cymene also support the disproportionation mechanism. Meanwhile, as the same as the case of dimeric products, heteroatom-containing monomeric products were also furnished (Figure 2C). Non-negligible amounts of borneol, bornyl chloride, and fenchyl chloride were only produced as bicyclic compounds during the reaction. Figure 2D–F show the distribution of monomeric products with increasing reaction time at 70 ◦C of the reaction temperature. The considerable amounts of limonene (18.9%) and camphene (10.5%) were produced when α-pinene still remained in the reaction system whereas their yields decreased after the conversion of α-pinene was complete, which denotes their participation in dimerization. Figure 2A,B show the main hydrocarbon product distribution in the monomeric hydrocarbon region in Figure S1B as a function of the reaction temperature. These monomeric hydrocarbons could be divided into two groups based on the tendency in the content thereof, according to the reaction temperature. Limonene, terpinolene, α-terpinene, γ-terpinene, and camphene except for tricyclene presented a sudden decrease in their yield above 90 °C, meaning their participation in dimerization. On the other hand, the contents of menthenes and p-cymene increased with increasing reaction temperature until 100 °C. These have been reported to form by the disproportionation of pmenthadienes [25]. In our results, the similar contents of menthenes and p-cymene also support the disproportionation mechanism. Meanwhile, as the same as the case of dimeric products, heteroatomcontaining monomeric products were also furnished (Figure 2C). Non-negligible amounts of borneol, bornyl chloride, and fenchyl chloride were only produced as bicyclic compounds during the reaction. Figure 2D–F show the distribution of monomeric products with increasing reaction time at 70 °C of the reaction temperature. The considerable amounts of limonene (18.9%) and camphene (10.5%) were produced when α-pinene still remained in the reaction system whereas their yields decreased after the conversion of α-pinene was complete, which denotes their participation in dimerization.

Figure 2. Effect of reaction temperature and reaction time on the yields of (A,D) p-menthadiene isomers, (B,E) other monomeric hydrocarbons, and (C,F) heteroatom-containing monomeric products. The reaction was carried out (A–C) for 1 h and (D–F) at 70 °C. **Figure 2.** Effect of reaction temperature and reaction time on the yields of (**A**,**D**) *p*-menthadiene isomers, (**B**,**E**) other monomeric hydrocarbons, and (**C**,**F**) heteroatom-containing monomeric products. The reaction was carried out (**A**–**C**) for 1 h and (**D**–**F**) at 70 ◦C.

#### *3.2. Rationalization of the Formation of Monoterpene Isomers and Dimeric Hydrocarbons* Appl. Sci. 2020, 10, x FOR PEER REVIEW 5 of 9

There have been a few papers on the SnCl4·5H2O-catalyzed organic reactions. Based on the crystal structure of SnCl4·5H2O that consists of [SnCl4(H2O)2] linked by hydrogen bonding with additional water molecules [26], the mechanism of those reactions has been explained in the manner of both Lewis and Brønsted acid-base interaction. When Lewis basic heteroatom-containing reactants are used, their coordination to stannic (Sn(IV)) center is considered to be an important step of the reaction. Such examples involve β-amino alcohol synthesis by epoxide ring-opening [27] and β-amino carbonyl compound synthesis by the Mannich-type reaction [28]. On the other hand, in the absence of heteroatom in reactants, the Brønsted acidity of the coordinated water ligands is considered to be important. For example, the hydration of alkyne has been explained by the initial protonation of reactants by water, which is activated by the coordination to the stannic center [29]. Such Lewis acid-assisted Brønsted acids were also developed for enantioselective polyene cyclization [19,20] and enantioselective protonation of silyl enol [21,22]. Given α-pinene has no heteroatom that can directly interact with the stannic center in a Lewis acid-base manner, the catalytic conversion of α-pinene by SnCl4·5H2O should follow the second explanation above. 3.2. Rationalization of the Formation of Monoterpene Isomers and Dimeric Hydrocarbons There have been a few papers on the SnCl4·5H2O-catalyzed organic reactions. Based on the crystal structure of SnCl4·5H2O that consists of [SnCl4(H2O)2] linked by hydrogen bonding with additional water molecules [26], the mechanism of those reactions has been explained in the manner of both Lewis and Brønsted acid-base interaction. When Lewis basic heteroatom-containing reactants are used, their coordination to stannic (Sn(IV)) center is considered to be an important step of the reaction. Such examples involve β-amino alcohol synthesis by epoxide ring-opening [27] and β-amino carbonyl compound synthesis by the Mannich-type reaction [28]. On the other hand, in the absence of heteroatom in reactants, the Brønsted acidity of the coordinated water ligands is considered to be important. For example, the hydration of alkyne has been explained by the initial protonation of reactants by water, which is activated by the coordination to the stannic center [29]. Such Lewis acidassisted Brønsted acids were also developed for enantioselective polyene cyclization [19,20] and enantioselective protonation of silyl enol [21,22]. Given α-pinene has no heteroatom that can directly interact with the stannic center in a Lewis acid-base manner, the catalytic conversion of α-pinene by

The possible mechanisms for the conversion of α-pinene to isomerized products (tricyclene, camphene, limonene, terpinolene, α-terpinene, and γ-terpinene), disproportionated products (menthenes and *p*-cymene), and dimerized products by various acid catalysts have been proposed in the previous papers [8,23]. In this work, we focused on a plausible mechanism involving interaction between α-pinene and the catalyst, which is shown in Scheme 1. SnCl4·5H2O should follow the second explanation above. The possible mechanisms for the conversion of α-pinene to isomerized products (tricyclene, camphene, limonene, terpinolene, α-terpinene, and γ-terpinene), disproportionated products (menthenes and p-cymene), and dimerized products by various acid catalysts have been proposed in the previous papers [8,23]. In this work, we focused on a plausible mechanism involving interaction between α-pinene and the catalyst, which is shown in Scheme 1.

Scheme 1. A plausible catalytic cycle of SnCl4·5H2O (A) proposed for the synthesis of limonene (2) and dimeric hydrocarbons (C20H32) from α-pinene (1). **Scheme 1.** A plausible catalytic cycle of SnCl<sup>4</sup> ·5H2O (**A**) proposed for the synthesis of limonene (**2**) and dimeric hydrocarbons (C20H32) from α-pinene (**1**).

First, the protonation of α-pinene (1) is thought to be promoted by the Lewis acid-assisted Brønsted acid activity of the coordinated water ligands in the Sn(IV) complex (A) via TS1 (Scheme 1). Since there is no polar solvent that can solvate the generated ions, an intimate ion-pair between carbocation and Sn(IV) ate complex should be a reaction intermediate (Int1) [19,30]. To form limonene (2), the main isomerized product from compound 1, the pinanyl moiety in the intimate ion-pair Int1 First, the protonation of α-pinene (1) is thought to be promoted by the Lewis acid-assisted Brønsted acid activity of the coordinated water ligands in the Sn(IV) complex (A) via TS1 (Scheme 1). Since there is no polar solvent that can solvate the generated ions, an intimate ion-pair between carbocation and Sn(IV) ate complex should be a reaction intermediate (Int1) [19,30]. To form limonene (2), the main isomerized product from compound 1, the pinanyl moiety in the intimate ion-pair Int1 should suffer

from a ring-opening rearrangement, which generates a carbocation/Sn(IV) ate complex ion-pair (Int2). In this intimate ion-pair Int2, the abstraction of a proton from the carbocation moiety by the hydroxo ligand via TS2 regenerates stannic complex A and furnishes compound 2 as the product of a catalytic cycle. If the insertion of another compound 1 into the intimate ion-pair Int2 is prior to the proton abstraction step, a carbocation/Sn(IV) ate complex ion-pair (Int3) appears for dimeric hydrocarbons (C20H32) as the product of the catalytic cycle. Along with the case of compound 2, we can construct distinct routes to complete the catalytic cycle involving the possible combination of carbocation/Sn(IV) ate complex ion-pairs, which are able to explain the various isomers and dimeric hydrocarbons from compound 1. ion-pair (Int2). In this intimate ion-pair Int2, the abstraction of a proton from the carbocation moiety by the hydroxo ligand via TS2 regenerates stannic complex A and furnishes compound 2 as the product of a catalytic cycle. If the insertion of another compound 1 into the intimate ion-pair Int2 is prior to the proton abstraction step, a carbocation/Sn(IV) ate complex ion-pair (Int3) appears for dimeric hydrocarbons (C20H32) as the product of the catalytic cycle. Along with the case of compound 2, we can construct distinct routes to complete the catalytic cycle involving the possible combination of carbocation/Sn(IV) ate complex ion-pairs, which are able to explain the various isomers and dimeric hydrocarbons from compound 1.

should suffer from a ring-opening rearrangement, which generates a carbocation/Sn(IV) ate complex

#### *3.3. Rationalization of the Formation of Heteroatom-Containing Dimeric Products* 3.3. Rationalization of the Formation of Heteroatom-Containing Dimeric Products

As previously mentioned, from α-pinene, stannic chloride molten salt hydrates (SnCl4·5H2O) can produce dimerized products that are high-density fuel candidates. There is one limitation of this reaction system, however, which involves forming not only the dimeric hydrocarbons that show molecular ion peaks at *m*/*z* = 272 [C20H32] <sup>+</sup> but also the oxygen-containing dimeric products, such as possibly dimeric alcohols or dimeric ethers, that show molecular ion peaks at *m*/*z* = 290 [C20H34O]+. Although the inclusion of oxygen-containing species increases the overall density of the fuel mixture [5], which matches the purpose of augmenting the energy density of renewable fuel, this is not suitable considering the synthesis of hydrocarbons for drop-in fuel. Therefore, to exclude the production of oxygen-containing species from the reaction system catalyzed by SnCl4·5H2O, we have started the rationalization of how such species are formed. As previously mentioned, from α-pinene, stannic chloride molten salt hydrates (SnCl4·5H2O) can produce dimerized products that are high-density fuel candidates. There is one limitation of this reaction system, however, which involves forming not only the dimeric hydrocarbons that show molecular ion peaks at m/z = 272 [C20H32] + but also the oxygen-containing dimeric products, such as possibly dimeric alcohols or dimeric ethers, that show molecular ion peaks at m/z = 290 [C20H34O]<sup>+</sup> Although the inclusion of oxygen-containing species increases the overall density of the fuel mixture [5], which matches the purpose of augmenting the energy density of renewable fuel, this is not suitable considering the synthesis of hydrocarbons for drop-in fuel. Therefore, to exclude the production of oxygen-containing species from the reaction system catalyzed by SnCl4·5H2O, we have started the rationalization of how such species are formed.

.

From the fact that the formation of bornyl chloride (3), one particular chlorine-containing side-product, the decomposition of Sn(IV) complex (A) during the reaction can be postulated (Scheme 2, left circle). In other words, to form compound 3, the protonation of α-pinene (1) by stannic complex A and successive ring enlargement by Wagner-Meerwein rearrangement occur first, furnishing a carbocation/Sn(IV) ate complex ion-pair (Int4). This intimate ion-pair Int4 then suffers from the dissociation of chloride ligand, producing both compound 3 and decomposed Sn(IV) complex (B). From the fact that the formation of bornyl chloride (3), one particular chlorine-containing sideproduct, the decomposition of Sn(IV) complex (A) during the reaction can be postulated (Scheme 2, left circle). In other words, to form compound 3, the protonation of α-pinene (1) by stannic complex A and successive ring enlargement by Wagner-Meerwein rearrangement occur first, furnishing a carbocation/Sn(IV) ate complex ion-pair (Int4). This intimate ion-pair Int4 then suffers from the dissociation of chloride ligand, producing both compound 3 and decomposed Sn(IV) complex (B).

Scheme 2. A plausible mechanism of catalyst (A) decomposition suggested based on the production of bornyl chloride (3, left circle) and oxygen-containing dimeric products (C12H32O, right circle). **Scheme 2.** A plausible mechanism of catalyst (**A**) decomposition suggested based on the production of bornyl chloride (**3**, left circle) and oxygen-containing dimeric products (C12H32O, right circle).

As previously mentioned, for the conversion of α-pinene (1) to occur, carbocation/Sn(IV) ate

As previously mentioned, for the conversion of α-pinene (1) to occur, carbocation/Sn(IV) ate complex ion-pairs (Int1-4) should be reaction intermediates (Schemes 1 and 2). We considered that these intermediates are essential clues for the explanation of how the oxygen-containing species are formed. While bornyl chloride (3) is formed when the dissociation of chloride ligand (Sn-Cl bond) occurs in the intimate ion-pair Int4, monoterpene alcohols can be formed by the dissociation of hydroxo ligand (Sn-O bond). At this point, the monoterpene alcohols can occupy an empty coordinate site of the stannic center as an alcohol ligand, which generates ligand exchanged Sn(IV) complex (C) (Scheme 2, right circle). Since the same Lewis acid-assisted Brønsted acidic water ligands work as both catalysts and reactants, this alcohol ligand can also protonate another compound 1, furnishing a carbocation/Sn(IV) ate complex ion-pair (Int5) via TS3. From this intimate ion-pair Int5, the dissociation of the remained alkoxo ligand can cause the formation of dimeric ethers (C20H32O) and decomposed Sn(IV) complex (D).

Based on the above rationalization, we have concluded that, although the reaction system catalyzed by the Lewis acid-assisted Brønsted acid activity of SnCl4·5H2O can produce high-density fuel candidates from α-pinene, to synthesize dimeric products as only hydrocarbons, the water ligands must be switched. Instead of them, monodentate alcohols or phenol ligands, which will form alkoxo or aryloxo ligand after the protonation of α-pinene, respectively, also should be avoided since these ligands can furnish undesirable ether products. Therefore, we have considered bidentate 2-alkoxyphenol ligands such as 2-methoxyphenol (guaiacol) and 2,6-dimethoxyphenol (syringol), given they can be tightly bonded to the metallic center with their methoxy group even after the protonation of α-pinene [19–22]. To prove this hypothesis, further study should be followed.

#### **4. Conclusions**

To synthesize renewable high-density fuel from monoterpenes, we investigated whether the catalytic amount of stannic chloride molten salt hydrates (SnCl4·5H2O) can furnish dimeric products from α-pinene in a solvent-less condition. SnCl4·5H2O would work as a catalyst for the dimerization of α-pinene under milder conditions than previously reported ones. This conversion was thought to be promoted by a Lewis acid-assisted Brønsted acidic nature of the coordinated water ligands on the stannic (Sn(IV)) center. However, we can also observe the unexpected heteroatom-containing products of which oxygen and chlorine atoms resulted from the decomposition of the catalyst. Considering the proposed catalytic cycle and the mechanism of the catalyst decomposition, we suggested bidentate ligands to minimize the above problem.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2076-3417/10/21/7517/s1. Figure S1: GC/FID chromatograms of (A) the neat turpentine used in this study and (B) the reaction mixture after 1 h of reaction at 100 ◦C. Figure S2: Representative mass spectra of (A) dimeric hydrocarbons, (B) trimeric hydrocarbons, and (**C**) oxygen-containing dimeric products.

**Author Contributions:** Conceptualization, S.-M.C. Methodology, S.-M.C. and J.-H.C. Formal analysis, S.-M.C. and J.-H.K. Investigation, S.-M.C. and J.-H.C. Resources, S.-M.C. Data curation, J.-H.C. and J.-H.K. Writing—original draft preparation, S.-M.C. and J.-H.C.Writing—review and editing, S.-M.C. and B.K. Supervision, I.-G.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by Mid-career Researcher Program in Basic Research of National Research Foundation of Korea grant funded by the Korea government (MSIP) (NRF-2016R1A2B4014222).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Biorefinery: The Production of Isobutanol from Biomass Feedstocks**

### **Yide Su, Weiwei Zhang \*, Aili Zhang and Wenju Shao**

School of Chemical Engineering and Technology, Hebei University of Technology, No. 8 Guangrong Road, Hongqiao District, Tianjin 300130, China; 201721508004@stu.hebut.edu.cn (Y.S.); zhangaili@tju.edu.cn (A.Z.); 201931505031@stu.hebut.edu.cn (W.S.)

**\*** Correspondence: 201821508004@stu.hebut.edu.cn; Tel.: +86-22-60200444

Received: 16 October 2020; Accepted: 10 November 2020; Published: 20 November 2020 -

**Abstract:** Environmental issues have prompted the vigorous development of biorefineries that use agricultural waste and other biomass feedstock as raw materials. However, most current biorefinery products are cellulosic ethanol. There is an urgent need for biorefineries to expand into new bioproducts. Isobutanol is an important bulk chemical with properties that are close to gasoline, making it a very promising biofuel. The use of microorganisms to produce isobutanol has been extensively studied, but there is still a considerable gap to achieving the industrial production of isobutanol from biomass. This review summarizes current metabolic engineering strategies that have been applied to biomass isobutanol production and recent advances in the production of isobutanol from different biomass feedstocks.

**Keywords:** isobutanol; biorefinery; metabolic engineering; biomass utilization

#### **1. Introduction**

Energy and the environment are two major issues facing the world. Due to climate change and the demand for renewable transportation fuels, the production of environmentally friendly biofuels has aroused great interest. Compared with fossil fuels, biofuels are more sustainable and highly renewable, which has attracted much attention [1–7].

In the past few years, researchers have focused on the production of biofuels from edible crops [8]. This is not a long-term solution. From the perspective of economy and sustainability, the ultimate goal is to convert low-cost non-edible biomass resources into high-value biofuels and other chemical products [9–13]. Therefore, people have proposed the biotransformation of lignocellulosic biomass, most of which is agricultural waste. With the development of biorefinery, the potential industrialization of lignocellulosic biomass processing to release monosaccharides that can be fermented and converted into high-value chemicals has become a reality. Cellulosic isobutanol already occupies a place in the fuel sector [14].

Ethanol has obvious disadvantages compared to fossil fuels, but most current research in biorefinery is still based on ethanol production. We need to expand the types of chemicals that can be produced to facilitate further development of biorefinery. Compared with ethanol, most higher alcohols have lower hygroscopicity, higher energy density, higher octane number, and properties that are closer to gasoline, so they have a higher compatibility with existing equipment and higher operating safety factors [15–17]. Isobutanol is an important industrial compound. It is used in lubricants, coatings, adhesives, automobile spray paint, and as an intermediate for the synthesis of many drugs. In addition, isobutanol derivatives are widely used in the chemical industry [15]. As a new generation of biofuel, isobutanol has many advantages (Table 1) and uses [18–20]. Therefore, developing of biorefinery to produce biomass isobutanol is in line with market needs and sustainable development.


**Table 1.** Major characteristics comparison of several common biofuels with gasoline.

Biorefinery involves many steps of biomass treatment, biomass conversion, fermentation, product purification, fermentation processes, etc. In this article, we discuss the industrial microorganisms used in biorefinery, review the metabolic engineering strategies of isobutanol production by microorganisms, and introduce the current situation and developmental prospects of isobutanol production from various biomass feedstocks.

#### **2. Research on Isobutanol Production**

The isobutanol synthesis pathway (Figure 1) starts when pyruvate is converted to 2-ketoisovalerate (KIV) by acetolactate synthase (AHAS), acetohydroxyacid reductoisomerase (AHARI), and dihydroxyacid dehydratase (DHAD), respectively. Next, KIV is converted to isobutyraldehyde by the 2-ketoacid decarboxylase (KIVD), and finally alcohol dehydrogenase (ADH) produces isobutanol (we call the pathway consisting of these five enzymes the engineered isobutanol pathway below). Microorganisms that are widely used to biosynthesize isobutanol include *Escherichia coli*, *Corynebacterium glutamicum*, *Bacillus subtilis*, and *Saccharomyces cerevisiae*. Most microorganisms do not produce, or only produce trace amounts of isobutanol on their own. Now several strategies can increase isobutanol production, such as the overexpression of key enzymes of the isobutanol synthesis pathway, the inhibition of byproduct production, cofactor engineering, and microbial robustness enhancement. We selected some representative studies to briefly introduce current microbial isobutanol production strategies.

In 2008, Liao's group [16] validated the potential of microbes to produce higher alcohols using *E. coli*. By introducing KIVD from *Lactococcus lactis* and ADH from *S. cerevisiae*, the engineered strain JCL260 successfully produced isobutanol using glucose as a substrate. Then increasing the pyruvate and 2-ketoacid concentrations, JCL260 produced 22 g/L of isobutanol, which, at 86% of the theoretical maximum, demonstrated this strategy's potential. This is the first time that researchers have used metabolic engineering strategies to produce isobutanol with microorganisms. The cytotoxicity of isobutanol causes high concentrations of isobutanol to inhibit cell growth, thus limiting the maximum titer. Liao's group [21] further combined an in situ product removal strategy (gas stripping) with isobutanol production to improve the final titer of isobutanol. The final isobutanol production reached 50 g/L after 72 h of fed-batch fermentation in a bioreactor. They also used the chemical mutagen N'-nitro-N-nitrosoguanidine (NTG) to induced sequential mutagenesis in *E. coli* and screened for a high-yielding isobutanol strain. Repairing an inhibitory mutation resulted in a final isobutanol titer of 21.2 g/L isobutanol in 99 h [22]. Another factor that limits the isobutanol titer is a cofactor imbalance. Arnold's group [23] removed the dependence of the isobutanol synthesis pathway on NAPDH in *E. coli* and achieved 100% of the theoretical maximum isobutanol production during growth in anaerobic conditions.

; **Figure 1.** Schematic diagram of microorganism isobutanol synthesis pathways. Part (**A**) shows the general pathways of isobutanol biosynthesis; Part (**B**) shows a partial cofactor balancing strategy for successful optimization of isobutanol synthesis; Part (**C**) shows the common by-product synthesis pathways of isobutanol synthesis; and Part (**D**) shows the sugar metabolism pathway using non-glucose fermentation to produce isobutanol. The dashed line indicates the source of the precursor substance for the byproduct; AHAS: Acetolactate synthase; AHARI: Acetohydroxyacid reductoisomerase; DHAD: Dihydroxyacid dehydratase; KIVD: Keto acid decarboxylase; ADH: Alcohol dehydrogenase; G6PD: Glucose-6-phosphate dehydrogenase; 6PGD: 6-phosphogluconate dehydrogenase; XR: Xylose reductase; XDH: Xylitol dehydrogenase; XI: Xylose isomerase; XKS: Xylulokinase; AI: L-arabinose isomerase; RK: L-Ribulokinase; R5PE: L-Ribulose-5-P-4-epimerase.

Liao's *C. glutamicum* is widely used to produce various amino acids, so some researchers have explored isobutanol production by *C. glutamicum*. In 2010, Liao's group [24] tried to produce isobutanol with *C. glutamicum* for the first time. They overexpressed the engineered isobutanol pathway, and on this basis deleted the *PYC* gene (encoding pyruvate carboxylase) and *LDH* gene (encoding lactate dehydrogenase). This engineered strain they obtained produced 4.9 g/L of isobutanol when fermented for 96 h. Blombach's group [25] found that the main byproducts of *C. glutamicum* during isobutanol production are lactic and succinate. By knocking out the byproduct synthesis genes, restoring the redox balance in combination with heterologous expression of transhydrogenase (*pntAB* from *E. coli*), and overexpressing *adhA*, the final isobutanol titer reached 13 g/L at 48 h. Recently, Inui's group [26] has used different promoter combinations to confirm the importance of higher activity of AHAS and KDC for isobutanol synthesis in *C. glutamicum*. In combination with a cofactor strategy (altering the cofactor specificity of AHARI and ADH) and enhanced glycolytic flux strategy (overexpression of endogenous glycolytic genes and the phosphoenolpyruvate:carbohydrate phosphotransferase system

(PTS), plus introduction of the Entner–Doudoroff pathway from *Zymomonas mobilis*), 20.8 g/L isobutanol was produced at 24 h to reach 84% of the theoretical value.

Wen's group [27] proved that *B. subtilis* could be used as a cell factory for isobutanol synthesis through an isobutanol tolerance test. In 2011, they introduced the engineered isobutanol pathway into *B. subtilis*, resulting in strain BSUL03. The concentration of the isobutanol reached 2.63 g/L at 54 h. Next, they performed an elementary mode analysis (EMA) on the engineered strain BSUL03 to identify targets in the metabolic network that could be optimized, lactate dehydrogenase and pyruvate dehydrogenase complexes. By knocking out the *ldh* and *pdhC* genes, they engineered the strain BSUL05 and obtained an isobutanol titer of 5.5 g/L after 60 h of fermentation [28]. Afterwards, they performed a metabolic flux analysis and comparison of the two engineered strains. To increase the concentration of NADPH and achieve a redox balance, the engineered strain BSUL08 was obtained by knocking out *pgi* (encoding glucose 6-phosphate isomerase), overexpressing *zwf* (encoding glucose 6-phosphate dehydrogenase), and further overexpression the transhydrogenase gene (*udhA* from *E. coli*). The production performance was tested with fed-batch fermentation and the isobutanol titer reached 6.12 g/L at 60 h, which was 63% of the theoretical maximum [29].

*S. cerevisiae* can produce a small amount of higher alcohols through the biosynthetic pathway of various amino acids. It also has a natural tolerance to alcohols that is higher than other microorganisms [30]. In recent years, *S. cerevisiae* has been widely used to produce higher alcohols. In 2011, Chen's group [31] overexpressed of the genes *ILV2* (encoding acetolactate synthase), *ILV3* (encoding dihydroxy acid dehydratase), and *ILV5* (encoding diacetolactate reductase) to obtain an isobutanol yield of 3.86 mg/g glucose. This is the first record of *S. cerevisiae* being used to produce isobutanol. Based on that, Boles's group [32] truncated the N-terminal mitochondrial targeting sequence of the *ILV*s to achieve expression of *ILV2*, *ILV3*, and *ILV5* in the cytoplasm, and optimized the codons of these three genes. They studied the activities of KIVD and ADH and determined that *ARO10* and *ADH2* were the most active enzymes in the synthesis of isobutanol. The final concentration of isobutanol produced by the strain they obtained was 0.63 g/L. Then they [33] knocked out the synthesis pathway of leucine, isoleucine, 2,3-butanediol, glycerol, pantothenate, and isobutyrate, thereby increasing the carbon flux of the isobutanol metabolism pathway. Here, the isobutanol titer reached 2.09 g/L in 96 h. Ethanol has always been the largest byproduct of isobutanol production, but the complete removal of pyruvate decarboxylase activity would also arrest cell growth. Therefore, Avalos's group [34] designed two powerful optogenetic gene expression systems in *S. cerevisiae*. In the presence of light, this system induces the expression of *PDCs* and promotes cell growth; in the dark, it induces the expression of *ILV2* to enhance the isobutanol biosynthesis. By controlling light exposure during fermentation, they could control ethanol production and could increase the isobutanol titer to 8.49 g/L.

#### **3. Biomass Isobutanol Production**

#### *3.1. Isobutanol Production from Lignocellulose*

Isobutanol is considered a promising alternative to gasoline, and cellulosic isobutanol is becoming increasingly important in the wave of next-generation biofuels. Extensive research in cellulosic ethanol [35,36] and cellulosic butanol [37,38] has provided a theoretical basis sufficient for the synthesis of cellulosic isobutanol. Some researchers have carried out feasibility analyses on the industrial production of cellulose isobutanol [39,40], confirming the great potential of cellulose isobutanol. However, few studies focus on the synthesis of isobutanol from lignocellulose biomass (details are summarized in Table 2).

3.1.1. Cellulosic Isobutanol Produced by Natural Cellulose-Degrading Microorganisms

Degrading cellulose with microorganisms that naturally utilize lignocellulose and convert the resulting C5 and C6 sugars into isobutanol is the simplest way to obtain cellulosic isobutanol. To minimize the lignocellulose glycation process and production costs, the consolidated bioprocessing

(CBP) strategy was developed. The first cellulosic isobutanol was synthesized by introducing the engineered isobutanol pathway into *Clostridium cellulolyticum* [41], which utilizes cellulose naturally. The strain produced a final isobutanol titer of 0.66 g/L. The simple introduction of the isobutanol synthesis pathway resulted in very low amounts of isobutanol titer, which is not sufficient for industrial scale production. The inability of *C. cellulolyticum* to process large amounts of substrate has been reported [42]. Destroying the ability of *C. cellulolyticum* to form spores by knocking out the *spo0A* gene improved cellulose utilization, but there was little variation in isobutanol titer (even lower than the wild type when the cellulose substrate reaches 50 g/L) [43]. The biosynthesis of cellulose isobutanol was also attempted in *Geobacillus thermoglucosidasius* [44], which produced isobutanol at 0.6 g/L from cellobiose in 60 h. While unsuccessful at producing large volumes of cellulosic isobutanol, this experiment demonstrated the strong thermal stability of ALAS and KIVD, which are widely used for isobutanol synthesis.



**Table 2.** *Cont.*


**Table 2.** *Cont.* 


**Table 2.** *Cont.*

Note: The abbreviation in the upper left corner of the gene indicates that the gene is a mutation; abbreviations in the lower right corner of genes indicate microorganisms of gene origin. BS: *Bacillus subtilis*; CG: *Corynebacterium glutamicum*; CP: *Clostridium phytofermentans*; CT: *Clostridium thermocellum*; EC: *Escherichia coli*; GT: *Geobacillus thermoglucosidasius*; KP: *K. pneumoniae* KCTC2242; LL: *Lactococcus lactis*; PE: *Piromyces* sp. E2; PF: *Pyrococcus furiosus*; PS: *Pichia stipites*; SC: *Saccharomyces cerevisiae*; SS: *Sche*ff*ersomyces stipitis*; TF: *Thermobifida fusca*; TI: *Thermoactinomycetes intermedius*; XC: *Xanthomonas campestris*.

The cellulose-utilizing *Clostridium thermocellum* is considered a promising producer of cellulosic biofuels, so additional experiments have attempted cellulosic isobutanol synthesis in this bacterial species [69]. Refinement gene expression systems [70–72] and gene editing techniques [73] have already been developed for *C. thermocellum*, and its basic metabolic network has been explored. Introducing the KIVD into *C. thermocellum*, optimizing the isobutanol synthesis pathway, and limiting the urea content of the medium enabled the engineered strain to ferment 5.4 g/L of isobutanol in 75 h [45]. Experiments have also identified a new pathway in *C. thermocellum* that converts KIV to isobutanol. This pathway decarboxylates KIV to isobutyryl-CoA with a ketoisovalerate ferrooxide-dependent reductase (KOR), which is converted to isobutanol by the aldehyde/alcohol dehydrogenase (ADH). Most recently, Holwerda's group [46] engineered *C. thermocellum* by eliminating the acetic and lactic synthesis pathways and performed adaptive laboratory evolution, resulting in the strain LL1043 that produced an isobutanol titer of 5.1 g/L. High titer were obtained without optimizing isobutanol synthesis, suggesting that the production of cellulose isobutanol with *C. thermocellum* remains a possibility.

Some new studies attempt to apply isobutanol production to other microbes. Lin's group [47] designed a microbial consortium that cocultured the cellulose utilizing strain *Trichoderma reesei* with an isobutanol-producing strain of *E. coli* to produce 1.88 g/L of isobutanol from pretreated corn stover. Another study [48] investigated a new AOR-ADH pathway for alcohol production. This pathway converts acetate to ethanol via aldehyde ferredoxin oxidoreductase (AOR) and ADH. The cellulolytic extreme thermophile *Caldicellulosiruptor bescii* naturally produces acetate. Heterologous expression of the AOR-ADH pathway in *C. bescii*, resulted in ethanol synthesis. Isobutanol was synthesized using switchgrass as a carbon source following isobutyrate supplementation, providing new ideas for subsequent isobutanol synthesis.

#### 3.1.2. Cellulosic Isobutanol Produced by Non-Native Cellulose-Degrading Microorganisms

Another strategy for obtaining cellulosic isobutanol is to further reform the pre-existing engineered strains for isobutanol production to utilize lignocellulose or its pretreatment products. *E. coli* and *S. cerevisiae* are the most promising cellulosic isobutanol-producing strains. Two model organisms have invested considerable research in cellulose utilization, which have been summarized in a considerable number of reviews [74,75]. They both have significant carbon catabolite repression and cannot utilize glucose simultaneously with other sugars. Xylose is the most abundant sugar in lignocellulose aside from glucose, so many efforts have been to construct industrially viable xylose-utilizing strains. To eliminate carbon catabolite repression, *E. coli* used UV mutagenized screened a new target gene [49], *Mlc*, encoding a DNA-binding transcriptional repressor. A strain with this mutant *Mlc* gene outperformed previous research in fermenting isobutanol from mixed glucose and xylose. Further optimization of the isobutanol synthesis pathway genes expression (*ilvC*, *ilvD*, *alsS*, *kivd* and *adhA*) with the biomass-inducible chromosome-based expression system (BICES) [50], achieved an isobutanol titer of 3.7 g/L from cedar hydrolysate.

*S. cerevisiae* cannot convert xylose naturally, but considerable efforts have been made to engineer a strain of *S. cerevisiae* that can [76]. There are two types of xylose metabolism (Figure 1D), one catalyzed by xylose reductase (XR) and xylitol dehydrogenase (XDH), the other by xylose isomerase (XI). However, the XR-XDH pathway requires the cofactor NADPH that is also in high demand for isobutanol synthesis, prone to cofactor imbalance. For this reason, the XI pathway is favored for isobutanol synthesis, but most attempts to express heterologous xylose isomerases in *S. cerevisiae* have failed. XylA, a xylose isomerase from *Clostridium phytofermentans*, was successfully expressed in *S. cerevisiae* by Boles' group [51]. Further, they increased xylose metabolism by overexpressing *Xks1* and *Tal1*, with the redesigned cytoplasmic isobutanol pathway, it achieved first isobutanol production by using xylose as the sole carbon source. Stephanopoulos' group engineered a strain to produce ethanol from xylose by the introducing xylose isomerase from *Piromyces sp.* E2 in *S. cerevisiae*. Overexpressing *RKI1*, *RPE1*, *TKL1*, *TAL1* and *XYL3* in this background increased xylose assimilation and facilitated adaptive laboratory evolution [52]. Boles' group [53] further optimized the use of

xylose in this strain by knocking out gene *PHO13*. With expression of the mitochondrial isobutanol pathway and inhibition of valine and acetic acid synthesis, the production of isobutanol was 3.1 g/L in 192 h of fed-batch fermentation. These experiments show that xylose promotes mitochondrial activity significantly more than glucose does and increases with increasing concentrations of xylose. Thus, the advantages of isobutanol production through the mitochondrial isobutanol pathway from xylose were demonstrated. Another xylose utilization pathway, the XR-XDH pathway, has also been used to produce isobutanol. Runguphan's group [77] determined through combinatorial screening that the xylose metabolism pathway from *Sche*ff*ersomyces stipitis* is most effective when applied to isobutanol synthesis. Incorporating the strategies identified in previous studies to enhance xylose conversion and assimilation (knockout *PHO13*, *GRE3* and overexpression of *XYL3*) and optimizing the copy number of isobutanol pathway genes produced 48.4 mg/L of isobutanol within 144 h. Adaptive laboratory evolution of the resulting engineered strain, and identification of two newly discovered mutation targets (the *CCR4*A638S and *TIF*A79S) to improve xylose utilization [54] increased the xylose ratio growth rate by 40.6%, but had little impact on isobutanol production. Using the XR-XDH pathway is likely to cause redox imbalance, and numerous studies have attempted to alleviate the cofactor imbalance by altering the cofactor specificity of the enzyme [78–80]. Jin's group [55] adjusted the copy numbers of three genes (*XYL1*, *XYL2*, and *XYL3*) in *S. cerevisiae* to improve redox balance and reduce acetate and xylitol accumulation. These results of evolutionary engineering confirmed the importance of deleting *PHO13* for xylose utilization. The final isobutanol titer was 2.6 g/L when coupled with the mitochondrial isobutanol pathway [56]. Metabolite analysis of the engineered strain revealed that the use of xylose increased valine levels in *S. cerevisiae*, which can be converted by branched-chain amino acid transaminases into KIV. This once again confirms that xylose promotes isobutanol synthesis.

*C. glutamicum* is also a promising industrial producer of isobutanol, and it is necessary to develop an engineered strain that can utilize cellulose. By heterologous expression the xylose XI and the arabinose metabolic pathways from *E. coli*, Blombach's group [57] succeeded in constructing strains that can rapidly utilize mixed sugars, including glucose. When combined with previous strategies to optimize isobutanol synthesis [25,58], 0.53 g/L of isobutanol could be produced from hemicellulose fractions.

One way to address the fermentation of mixed sugars containing xylose is to bypass the use of glucose. Cellobiose is an intermediate product of cellulose hydrolysis to glucose and has no carbon catabolite repression effects on cells. Mixed sugar fermentation using cellobiose and xylose has been reported [81,82], but no studies have addressed isobutanol synthesis. Isobutanol production using cellobiose was attempted based on a previously engineered strain from *E. coli*. Expression of β-glucosidase from *Thermobifida fusca* resulted in an isobutanol production of 7.64 g/L by optimizing gene copy number [59]. It should be noted that direct transport of cellobiose to intracellular hydrolysis is required to avoid carbon catabolite repression, and no growth of cells expressing the cytoplasmic enzyme was observed in this study. Another study [60] tested cellobionic acid, the main product of cellulose hydrolysis assisted by the use of lytic polysaccharide monooxygenase (LPMO), as a carbon source for the production of isobutanol. *E. coli* was found a naturally utilize the cellobionic acid pathway. A final titer of 1.4 g/L was produced from the cellulose hydrolysate of *Neurospora crassa* conversion, thus expanding the available carbon source for cellulosic isobutanol.

Most of the studies described above are based on either industrially produced cellulose or the monosaccharides obtained from its hydrolysis as substrates, but there is a lack of studies on the direct use of cellulose or its hydrolysates to produce isobutanol. Depending on the source of cellulosic biomass, the proportions of cellulose, hemicellulose, and lignin can vary considerably [83]. The hydrolysis products will also be significantly different, depending on how the pretreatment is metabolized [84]. These hydrolysis products (weak acids, furan derivatives, phenolic compounds, etc.) toxic to cells and reduce the isobutanol titer. Several experiments have tested the ability of engineered isobutanol-producing strains designed to ferment the hydrolysis products of cellulosic biomass. Duckweed [61,85] is a fast-growing non-food crop with few growth requirements that is very easy to handle. Empty fruit bunches [62] are a large byproduct of palm oil production and

sugarcane bagasse [63] is a common industrial byproduct. Their pretreated fermentation inhibitor content is low, as can be concluded from the comparison with glucose fermentation. This resulted in these biomasses being good carbon source providers. The best isobutanol titer was obtained from *Enterobacter aerogenes* [64] use of sugarcane bagasse hydrolysates to produce 23 g/L of isobutanol by pervaporation-coupled fermentation.

#### *3.2. Isobutanol Production from Protein*

Biorefinery is focused on the industrial production of high-value compounds from biomass. Industrial waste like distillers' grains and okara waste cannot be integrated into the most existing strategies due to their primary component, proteins. For the process of protein hydrolysis to amino acids, some amino acids produce 2-keto acids, which are precursors of isobutanol. Liao's group [65,86] has pioneered the use of proteins as feedstock for higher alcohols production. In *E. coli*, they used chemical mutagenesis to screen the *YH19* strain that can utilize 13 amino acids as a sole carbon source. The restriction of ammonia assimilation, introduction of three exogenous transamines and deamines cycles, and knocking out population-sensing genes (*luxS* and *lsrA*) generated the strain YH83 [65]. Strain YH83 expresses the engineered isobutanol pathway that used algal protein hydrolysates to produce higher alcohols. Taking the more severe cofactor imbalance triggered by the use of protein biomass into consideration, Davis' group [66] performed cofactor-specific reconstructions of 2,3-dihydroxy isovalerate oxidoreductase (IlvC) and NADPH-dependent isobutanal dehydrogenase (YqhD) in the *YH83* strain to obtain an isobutanol titer of 0.2 g/L from algae protein hydrolysate. Davis' group [67] then attempted to utilize both sugars and proteins from biomass by constructing an *E. coli* microbial consortium, that is coculturing a proportional mix of strain *AY3*, which can produce heteroalcohols from proteins, and strain *BL2*, which can produce isobutanol from glucose and xylose. Protein utilization increased from 16.3% to 31.3% in a separate culture at a ratio of 1.5:1. The fermentation of distillers' grains hydrolysate under these conditions also resulted in the highest isobutanol titer of 6.5 g/L, which was higher than the 5.5 g/L achieved with strain *BL2* fermentation alone. The fermentation of other biomass like algal protein hydrolysates was also investigated and the highest isobutanol titer of 2.38 g/L was achieved when strains *AY3* and *BL2* were fermented in a 4:1 ratio, demonstrating the industrial potential of this microbial consortium.

Since *E. coli* does not naturally produce the proteases required to process protein biomass, further attempts were made in *B. subtilis*, which secretes proteases that enable growth on polypeptides. Due to the different genetic backgrounds of *E. coli* and *B. subtilis*, a series of experiments were performed to determine the importance of codY (a global regulator) deletion for inhibiting ammonia assimilation and improving branched-chain amino acid synthesis. Combined with a strategy of preventing the degradation of branched-chain amino acid by knocking out the dihydrolipoyl acyltransferase (*bkdb*) in the branched-chain 2-keto acid dehydrogenase complex, a final biofuel titer of 0.72 g/L was obtained from protein biomass [86]. Based on this, Choi's group [68] knocked out *RelA*, expressing the regulatory protein responsible, that recognizes nutritional stress to activate the ilv-leu operon [87] and further promotes branched-chain amino acids synthesis. Comparing the results of spent coffee grounds and okara wastes fermentation, the triple deletion strain expressing KIVD and ADH produced almost no isobutanol with spent coffee grounds hydrolysate as a substrate, while okara wastes hydrolysate fermented about 0.02 g/L of isobutanol in eight days. The microbial consortium strategy was also applied to this engineered strain and isobutanol was successfully obtained from the hydrolysates of watermelon rind and soybean residue by coculturing with *E. coli AY3* [88]. The highest isobutanol titer was 0.88 g/L at an *E. coli* to *Bacillus subtilis* ratio of 4:1. From these studies that used proteins to produce isobutanol, it was found that differences in the source of biomass could lead to large changes in product ratios, which require further research.

#### **4. Conclusions**

In recent years, researchers have made considerable effort to increase the yield of microbial biofuel production and continued to explore chemical synthesis from different biomass materials [13]. In addition, some achievements have been made in the production of bioethanol [14,89]. In this paper we summarized current metabolic engineering strategies applied to biomass isobutanol production and recent advances in the production of isobutanol from different biomass feedstocks.

It can be seen that in the last decade, there was a significant improvement in the production of cellulosic isobutanol while the synthesis of isobutanol from protein biomass was demonstrated. Biomass isobutanol is no longer at the theoretical stage. However, the efforts made in the broader context of biorefining were not enough. Since isobutanol is a secondary metabolite, controlling the competing pathway to maximize the conversion of pyruvate to isobutanol is difficult, and removing the largest competitor, the ethanol pathway, can lead to severe growth defects. Coupled with the fact that biomass utilization efficiencies are low, the combination of the two problems leads to very low yields. Efficient inhibition of ethanol synthesis has been accomplished using light-controlled genetic systems [34], but the in vivo metabolic backgrounds of different microorganisms are very different, and further investment in research is needed to accomplish an efficient inhibition of the competing pathways. Another issue is how to enhance the robustness of engineered microorganisms, as both isobutanol and organics from biomass hydrolysates can cause considerable cellular damage. This issue includes how to improve the redox imbalance caused by isobutanol production, as it has been shown an improved redox significantly increases strain isobutanol tolerance [90] or furfural tolerance [91–93] and thus isobutanol production. Adaptive laboratory evolutionary applications for tolerance enhancement have obtained some results, with a large number of target genes being identified. However, the tolerance mechanisms are still opaque for both isobutanol and lignocellulose-derived microbial inhibitors. Research on the application of tolerance enhancement strategies to biomass isobutanol production is also quite scarce.

In particular, it should be noted that of all the current studies on the production of isobutanol from biomass, the highest titers reach 23 g/L by pervaporation-coupled fermentation, while the highest titers of isobutanol production are currently over 50 g/L due to in situ product removal. The timely isolation of isobutanol from the medium reduces the inhibitory effect of isobutanol on cells, reduces cytotoxicity, and significantly enhances the final titer. Fermentation processes also have a considerable influence on the production of isobutanol, but the existing research on microbial isobutanol synthesis and product separation [92,93] is weakly linked. This aspect also needs to be strengthened in future research on microbial isobutanol production. We believe that the synthesis of biomass isobutanol will make significant progress with the further research on these issues.

**Author Contributions:** Conceptualization, Y.S., W.Z. and W.S.; writing—original draft preparation, Y.S. and W.Z.; writing—review and editing, Y.S.; supervision, A.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Chunhui program of the Ministry of Education (Grant No.Z2017012).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
