Glucanocellulosic ethanol: the undiscovered biofuel potential in energy crops and marine biomass

Converting biomass to biofuels is a key strategy in substituting fossil fuels to mitigate climate change. Conventional strategies to convert lignocellulosic biomass to ethanol address the fermentation of cellulose-derived glucose. Here we used super-resolution fluorescence microscopy to uncover the nanoscale structure of cell walls in the energy crops maize and Miscanthus where the typical polymer cellulose forms an unconventional layered architecture with the atypical (1, 3)-β-glucan polymer callose. This raised the question about an unused potential of (1, 3)-β-glucan in the fermentation of lignocellulosic biomass. Engineering biomass conversion for optimized (1, 3)-β-glucan utilization, we increased the ethanol yield from both energy crops. The generation of transgenic Miscanthus lines with an elevated (1, 3)-β-glucan content further increased ethanol yield providing a new strategy in energy crop breeding. Applying the (1, 3)-β-glucan-optimized conversion method on marine biomass from brown macroalgae with a naturally high (1, 3)-β-glucan content, we not only substantially increased ethanol yield but also demonstrated an effective co-fermentation of plant and marine biomass. This opens new perspectives in combining different kinds of feedstock for sustainable and efficient biofuel production, especially in coastal regions.

An increasing worldwide demand for energy combined with decreasing fossil energy resources not only fosters climate change 1 but also explorations for fossil energy in sensitive ecosystems 2 . A key strategy to mitigate climate change is to substitute fossil by renewable energy sources. Because liquid fuels play a predominant role in the transportation sector, second generation biofuels from lignocellulosic feedstock reveal a high potential in substituting fossil fuels 3 . Restrictions in the production of ethanol from biomass mainly derive from the plant cell wall's recalcitrance, which is primarily determined by cellulose crystallinity but also lignin and hemicellulose content 4,5 .
To improve second generation ethanol production, we tried to identify and increase the content of cell wall polymers that are easily degradable and contain readily fermentable residues. Hence, polymers that only consist of glucose would represent an optimal substrate for ethanol fermentation with efficient microorganisms like yeast (Saccharomyces cerevisiae). Apart from the major (1, 4)-β-glucan cell wall polymer cellulose 6 , only two other cell wall polymers consist entirely of glucose: (1, 3)-β-glucan, known as callose in plants 7 , and (1, 3;1, 4)-β-glucan, a mixed-linkage glucan found in plants of the order Poales, in horsetail (Equisetum spp.), and in bryophytes 8 . (1, 3)-β-glucan is important to maintain the vascular system, for pollen development, and cell plate formation in growing tissue as well as for defense responses 7 . Mixed-linkage glucan can serve as an energy storage and has a growth-related function in vegetative tissues of grasses 9 . Because of their biological function, the abundance of these β-glucan polymers has been considered low in lignocellulosic biomass 10 . Therefore, these polymers have not been targeted to improve saccharification. To test whether an undiscovered potential of β-glucan processing to ethanol would exist, we determine the (1, 3)-β-glucan and (1, 3;1, 4)-β-glucan content in lignocellulosic biomass of crops representing a major source of lignocellulosic feedstock from agriculture in temperate climates: barley (Hordeum vulgare), wheat (Triticum aestivum), and maize (Zea mays); as well as in the model plants Arabidopsis (Arabidopsis thaliana) and Brachypodium (Brachypodium distachyon), and the emerging, perennial grass Miscanthus (Miscanthus x giganteus). This low-input energy crop with high biomass yields in temperate climates has been proposed for sustainable lignocellulosic feedstock production 11 .
Before engineering an improved utilization of (1, 3)-β-glucan-enriched biomass, we developed an equation to estimate the increase in biomass saccharification after optimized (1, 3)-β-glucan hydrolysis: where A(P x ) describes the relative amount of the glucan polymer P x , f (s) the glucose saccharification factor of the biomass B before optimization and of the glucan polymer P x after optimized hydrolysis, and f (s) i the increase in glucose saccharification after optimization. Based on equation (1), we expected an improved saccharification only if f (s) Px > f (s) B . We analyzed the maize leaf biomass broth after dilute-sulfuric acid pretreatment, hydrolysis with the cell wall-degrading enzyme cocktail Accellerase 1500, and subsequent fermentation with a non-adapted, laboratory yeast strain. Here, we detected relative high contents of laminaribiose and -triose (Fig. 2a), which we distinguished from putative glucotrioses deriving from possible (1, 3;1, 4)-β-glucan degradation using a refractive index detector coupled to an HPLC-system ( Supplementary Fig. 4). Due to their chemical composition and their relatively small size, we considered these (1, 3)-β-glucan degradation products as a potential, unused glucose source for fermentation. Therefore, we initiated experiments for optimizing (1, 3)-β-glucan hydrolysis and usage of (1, 3)-β-glucan degradation products for ethanol production. In a first step, we changed the yeast strain during fermentation to increase laminaribiose and -triose utilization during fermentation. The application of the yeast strain CEN.PK113-13D (CEN) that has been used to develop strains for optimized biomass fermentation 13 significantly improved laminaribiose utilization during fermentation, but did not effected laminaritriose utilization (Fig. 2a). The heterologous expression of the bacterial laminaribiose ABC transporter (LBT) from Clostridium thermocellum 14 in the yeast strain CEN ( Supplementary Fig. 5a,b) further improved the laminaribiose and laminaritriose utilization (Fig. 2b), resulting in only residual amounts of laminaribiose and -triose in the fermentation broth (Fig. 2a). However, the utilization of laminaritetraose, long-chained (1, 3)-β-glucan, or oligomers deriving from (1, 4)-β-glucan and (1, 3;1, 4)-β-glucan hydrolysis was not facilitated ( Supplementary Fig. 5c). Because of the efficient utilization of these two (1, 3)-β-glucan oligomers by CEN+ LBT, we considered laminaribiose and -triose as direct contributors to the overall glucose saccharification. Hence, the generation of this yeast strain represented a decisive step in engineering optimized (1, 3)-β-glucan utilization from β-glucan-enriched biomass.
Our results from saccharification and fermentation of maize and Miscanthus biomass revealed a direct correlation between the (1, 3)-β-glucan content of the feedstock and an increased ethanol yield. Hence, we concluded that a further (1, 3)-β-glucan enrichment in biomass would result in increased ethanol yields. To test this hypothesis, we followed two strategies: i) increasing the (1, 3)-β-glucan content in potential feedstock for sustainable biomass production using a biotechnological approach; and ii) identifying new sources of (1, 3)-β-glucan-enriched biomass that could be used in our adapted fermentation process.
To increase the (1, 3)-β-glucan content in feedstock for sustainable biomass production, we overexpressed the GFP-tagged (1, 3)-β-glucan synthase gene PMR4 (POWDERY MILDEW RESISTANT4) from Arabidopsis in Miscanthus (line 35S:PMR4-GFP). PMR4 overexpression in Arabidopsis increased (1, 3)-β-glucan content at infection sites but not in unchallenged tissue 16 . In contrast, we observed a constitutive increase in (1, 3)-β-glucan content in 35S:PMR4-GFP Miscanthus lines, which was proportional to the relative PMR4 expression level and reached a maximum of 8.5% in leaf tissue ( Supplementary  Fig. 10a,b). This result suggests different regulatory mechanisms of (1, 3)-β-glucan biosynthesis in Miscanthus and Arabidopsis, which would also explain the strong differences in their overall (1, 3)-β-glucan content (Fig. 1). As expected from Arabidopsis 16 , PMR4-GFP was localized at the plasma membrane whereas single GFP of a transgenic Miscanthus control line was detectable in cytosolic strands ( Supplementary  Fig. 10c). We predicted an increase in saccharification efficiency of about 16% in the 35S:PMR4-GFP line with the highest (1, 3)-β-glucan content of 8.5% after optimized hydrolysis, which was relatively close to our experimental results showing a saccharification increase of 14.5%. (Fig. 3b). The improved saccharification of this Miscanthus line resulted in an increase in ethanol production of 20% compared to non-optimized Miscanthus wild-type biomass processing (Fig. 3c). These results revealed a previously undiscovered potential in the layered architecture of maize and Miscanthus leaf cell walls that contain an atypically high content of (1, 3)-β-glucan, which was unleashed by engineering optimized enzymatic hydrolysis and yeast fermentation. Moreover, (1, 3)-β-glucan enrichment represents a new target in breeding energy crops for improved second generation ethanol production.
Similar to Miscanthus, brown macroalgae have been considered for sustainable biomass production 18 , however, without competing for arable land and food production. Because we demonstrated optimized ethanol production for both, (1, 3)-β-glucan-enriched plant and marine feedstock, we proposed our engineered biomass processing for co-fermentation of Miscanthus and bladderwrack biomass (Fig. 4a). A successful co-fermentation using equal amounts of Miscanthus and bladderwrack biomass proved the applicability of this engineered production approach (Fig. 4b). Regarding saccharification of mixed Miscanthus and bladderwrack biomass, we identified enzymatic hydrolysis as a field of further improvement. Here, the release of laminaribiose from (1, 3)-β-glucan was specifically inhibited during mixed Miscanthus and bladderwrack biomass processing compared to single biomass processing whereas inhibition did not occur for laminaritriose or glucose release in the mixed biomass approach (Fig. 4b).
Because brown macroalgae do not contain lignin and (1, 3)-β-glucan substantially contributed to improve second generation ethanol production in our study, we considered the produced ethanol as glucanocellulosic.
The effective co-fermentation opens new perspectives for sustainable and efficient ethanol production in bio-refineries, especially in coastal regions that combine the potential of offshore macroalgae aquaculture and proximate energy crop cultivation. An example for a coastal region that would fulfill these prerequisites is Schleswig-Holstein in northern Germany. Existing and planned offshore wind parks in the North and Baltic Sea would facilitate effective macroalgae aquacultures 19 , and a high potential for Miscanthus cultivation in Schleswig-Holstein was shown in our recent study 20 . Hence, these coastal regions would represent prototypic sites for future biorefineries for plant and marine biomass co-fermentation, combining short delivery distances of feedstock with a high abundance of renewable electricity for processing, which could help to promote large-scale energy transition projects like the ambitious German Energiewende 21 .

Methods
Biological material. Brachypodium (Brachypodium distachyon, inbred line Bd 21 22 ), barley (Hordeum vulgare, cultivar Golden Promise 23 ), wheat (Triticum aestivum, cultivar Nandu, Lochow-Petkus, Bergen-Wohlde, Germany), maize (Zea mays, inbred line A188 24 ), and Miscanthus (Miscanthus x giganteus) were cultivated as described in Meineke et al. 25 . Arabidopsis (Arabidopsis thaliana, wild-type Columbia) was cultivated as described in Stein et al. 26 . Naturally dried leaf material was harvested manually at its final developmental stage after senescence and 2 additional weeks of drying 25 . Biomass was additionally dried at 50 °C for 2 days in a drying oven. Washed ashore thalli of the brown macroalga bladderwrack (Fucus vesiculosus) were collected in November from the Baltic Sea shore at Eckernförde (Schleswig-Holstein, Germany, geographical position: 54 °27'57.5″ N 9°50'28.0″ E) and dried at 50 °C for 3 days. Plant and alga biomass was homogenized with a mill fitted with a 0.5 mm mesh screen prior processing. Material subject to (1, 3)-β-glucan extraction was ground in liquid nitrogen using a mortar and pestle.
(1, 3)-β-Glucan extraction and determination. 20 mg of mortared and lyophilized leaf or alga biomass was destained in ethanol (96%) at 50 °C and 600 rpm for 10 min. Subsequent procedures of (1, 3)-β-glucan determination followed the description in Voigt et al. 27 . Ethanol was removed after centrifugation (2 min, 10,000 g), and the sample was dried using a centrifugal evaporator. After a washing step with H 2 O, the sample was dried again in a centrifugal evaporator. For (1, 3)-β-glucan extraction, the sample was resuspended in 400 μ l of 1 M NaOH and incubated at 80 °C and 600 rpm for 1 h. After centrifugation (10 min, 2000 g), the supernatant was used for the aniline blue fluorescence assay for (1, 3)-β-glucan determination. 5 μ l sample were mixed with 45 ml H 2 O, 5 μ l HCl (1 M), 220 μ l K 2 HPO 4 (150 mM), and 5 μ l aniline blue fluorochrome (ABF, 0.1 mg·ml −1 in H 2 O, Biosupplies, Australia). Standards ranging from 0 to 20 μ g ml −1 were generated from purified (1, 3)-β-glucan from Euglena gracilis (Sigma-Aldrich, Germany) in the same way as described for plant and alga samples. Additional standards were generated accordingly from oat and barley (1, 3;1, 4)-β-glucan deriving from the mixed-linkage beta-glucan kit (Megazyme, Ireland) to verify the specificity of ABF in staining  Treatment and fermentation of plant biomass. For pretreatment, 5 g of milled leaf biomass was mixed with 43 ml sulfuric acid (1.75% (v/v)) and autoclaved for 15 min at 120 °C. Subsequent procedures of enzymatic hydrolysis and fermentation followed the description in Meineke et al. 25 . To test the impact of the (1, 3)-β-glucanase from F. johnsoniae, 500 ng of the purified enzyme were additionally added to fermentation reactors and incubated for 24 h and 200 rpm at 37 °C. Fermentations were initiated with the inoculation of 2 ml of overnight yeast cultures of the non-adapted, laboratory strain MaV203 (MaV, Life Technologies), CEN, or CEN+ LBT (generated in this study). Amounts of glucose, laminaribiose, laminaritriose, and ethanol in fermentation supernatants were quantified with a refractive index detector on an ICS-5000 system (Dionex, USA) with a HPX 87H column (Bio-Rad, USA, mobile phase 0.005 M H 2 SO 4 , flow rate 0.6 ml·min −1 , column temperature: 50 °C, refractive index detector) as described in Meineke et al. 25  Cloning and Miscanthus transformation. We generated two vector constructs for the transformation of Miscanthus: i) overexpression of the callose synthase gene PMR4 from Arabidopsis (At4g03550) fused to GFP and ii) overexpression of single GFP, both under control of the 35S promoter. PMR4-GFP and GFP were amplified from the vector pCAMBIA-35S:PMR4-GFP 16 using primers in PCR reactions that provide DNA recombination sequences (attB sites) at their 5′ and 3′ ends (PMR4-5′ attB: 5′ -GGGGACAAGTTTGTACAAAAAAGCAGGCTATGAGCCTCCGCCACCGC, GFP-5′ attB: 5′ -GGGGACAAGTTTGTACAAAAAAGCAGGCTTGGAGATC CAAACAATGAGTAAAG, GFP-3′ attB: 5′ -GGGGACCACTTTGTACAAGAAAGCTGGG TTAAGCTTGAATTCTTATT TGTATA) for utilization with the Gateway cloning technology. PMR4-GFP and GFP were introduced into the plant expression vector pIPKb004 28 , which provided 35S promoter-driven gene expression. Generated vector constructs containing 35S:PMR4-GFP and 35S:GFP expression cassettes were transformed into Agrobacterium (Agrobacterium tumefaciens, strain GV3101). The generation of transgenic Miscanthus lines followed the principal procedure of Agrobacterium-mediated callus transformation and selection on hygromycin-containing plant cell culture medium. Resistance to hygromycin was provided by the used plant expression vector pIBKb004. A detailed description of the transformation procedure is provided in the Supplementary Information. Statistical analysis. Descriptive statistics including the mean and the standard error of the mean (SE) along with the Tukey range test for multiple comparison procedures in conjunction with an ANOVA were used to determine significant differences. P < 0.05 was considered significant.