**Contents**


### **About the Special Issue Editor**

**Diomi Mamma**, Ph.D., is an Assistant Professor of Bioprocess Engineering at the School of Chemical Engineering, National Technical University of Athens, Greece. She studied Chemical Engineering at the School of Chemical Engineering, National Technical University of Athens, where she obtained her Ph.D. (2002). She teaches subjects related to bioprocess engineering. Her research interests focus on microbial production and characterization of enzymes, bioconversion of biomass to ethanol applying different fermentation strategies, and environmental biotechnology, with emphasis on the design of appropriate biological processes for the complete removal of xenobiotics.

### *Editorial* **Food Wastes: Feedstock for Value-Added Products**

#### **Diomi Mamma**

Biotechnology Laboratory, School of Chemical Engineering, National Technical University of Athens Zografou Campus, 15780 Athens, Greece; dmamma@chemeng.ntua.gr

Received: 22 April 2020; Accepted: 24 April 2020; Published: 27 April 2020

Food is a precious commodity, and its production can be resource-intensive. According to the Food and Agriculture Organization of the United Nations, nearly 1.3 billion tons of food products per year are lost along the food supply chain, and in the next 25 years the amount of food waste has been projected to increase exponentially. Food waste is produced at any stage of the supply chain, which extends from the agricultural site to the processing plant and finally the retail market. The management of food waste should follow certain policies based on the 3R's concept, i.e., reduce, reuse, and recycle [1]. Generally, food waste is composed of a heterogeneous mixture formed by carbohydrates (starch, cellulose, hemicellulose, or lignin), proteins, lipids, organic acids, and smaller inorganic parts. Currently, most food wastes are recycled, mainly as animal feed and compost. The remaining quantities are incinerated and disposed in landfills, causing serious emissions of methane (CH4), which is 23 times more potent than carbon dioxide (CO2) as a greenhouse gas and significantly contributes to climate change [2]. Valorizing food waste components could in fact lead to numerous possibilities for the production of valuable chemicals, fuels, and products [1].

The present Special Issue compiles a wide spectrum of aspects of research and technology in the area of "food waste exploitation", and highlights prominent current research directions in the field for the production of value-added products such as polylactic acid, hydrogen, ethanol, enzymes, and edible insects.

Polylactic acid (PLA) is a biodegradable polymer with great potential in replacing petrochemical polymers. The morphological, mechanical, and thermal properties of the polymer are determined by the presence of different amounts of l- and d-lactic acid monomers or oligomers [3]. The microbial production of optically pure lactic acid has extensively been studied, because chemically synthesized lactic acid is a racemic mixture. Optimizing culture conditions and selecting the LAB strains capable of producing d-lactic acid with high yield and optical purity from orange peel waste as raw material can contribute to the development of biowaste refineries. Bustamante et al. [4] evaluated six strains of the species *Lactobacillus delbrueckii* ssp. *bulgaricus* for the production of d-lactic acid from orange peel waste hydrolysate. *L. delbrueckii ssp. bulgaricus* CECT 5037 had the best performance, with a yield of 84% w/w for D-LA production and up to 95% enantiomeric excess (optical purity).

Biomethanation (methane fermentation) is a complex biological process, which can be divided in four phases of biomass degradation and conversion, namely, hydrolysis, acidogenesis, acetogenesis, and methanation. The individual phases are carried out by different groups of micro-organisms (bacteria), which partly stand in syntrophic interrelation and place different requirements on the environment. Undissolved compounds like cellulose, proteins, and fats are hydrolyzed into monomers by enzymes produced by facultative and obligatorily anaerobic bacteria [5]. The use of a microbial consortium consisting of the microbial flora of methane production and microorganisms that can degrade cellulosic biomass like *Clostridium cellulovorans* was proven efficient in degraded mandarin orange peel without any pretreatments and produced methane that accounted for 66.2% of the total produced gas [6].

Hydrogen is a noncarbonaceous fuel and energy carrier possessing higher net calorific value compared to other fuels (120 MJ/kg versus 46.7 MJ/kg for gasoline). Microbes primarily produce hydrogen via photofermentation by the purple nonsulfur bacteria *Rhodobacter* and *Rhodopseudomonas*, and during dark fermentation by strictly anaerobic *Clostridium* species [7,8]. Depending upon the availability of substrate, the selection of functional microorganisms necessary for hydrogen production is an important step. Simulation of the exchange metabolic fluxes of monocultures and pairwise cocultures using genome-scale metabolic models on artificial garbage slurry resulted in the identification of one of the top hydrogen producing cocultures comprising *Clostridium beijerinckii* NCIMB 8052 and *Yokenella regensburgei* ATCC 43003. The consortium produced a similar amount of hydrogen gas and increased butyrate (attributed to cross-feeding of lactate produced by *Y. regensburgei*), compared to the *C. beijerinckii* monoculture, when grown on the artificial garbage slurry [9].

Household food waste is a complex biomass containing various components that make it a source of potential fermentative substrates. The general scheme of bioethanol production from such complex materials involves a pretreatment step that increases the digestibility of the material—enzymatic hydrolysis—to liberate the monosaccharides and fermentation of these sugars to ethanol. In terms of cost, the most demanding step, which significantly increases the total cost of the production of bioethanol and is identified as a barrier in the further deployment of ethanol production, is enzymatic hydrolysis. If the necessary enzymes could be efficiently produced on-site, the cost could be significantly reduced. A recent study has estimated that the cellulase cost can be reduced from 0.78 to 0.58\$/gallon by shifting from the off-site to the on-site approach of cellulase production [10]. The mesophilic fungus *Fusarium oxysporum* F3 grown under solid state cultivation on wheat bran produced a multienzyme system capable of hydrolyzing the carbohydrates present in household food waste. The use of mixed-microbial cultures in bioethanol production step consisting of *F. oxysporum* solid state culture and the yeast *Saccharomyces cerevisiae* increased bioethanol volumetric productivity, compared to mono-culture of the fungus. Bioethanol production increased by approximately 23% when the mixed microbial culture was supplemented with low dosages of commercial glucoamylase [11].

Carrión-Paladines et al. [12] evaluated two *Xylaria* spp. of the dry forest areas of southern Ecuador, for ligninase and cellulase production under solid state fermentation using residues obtained from the Palo Santo essential oil extraction. The Palo Santo is considered a vital resource for the local communities of the dry forest, as different parts of the tree are used in traditional medicine, as well as for the extraction of essential oil. The essential oil extraction process generates abundant organic waste, which is commonly discarded directly into the natural ecosystems or burned. Laccase, cellulose, and xylanase activities of *Xylaria feejeensis* and *Xylaria* cf. *microceras* were generally higher than those of the control fungus *Trametes versicolor* (L.) Lloyd, furthering the understanding of the potential use of native fungi as ecologic lignocellulosic decomposers and for industrial proposes.

Beer production generates large quantities of spent yeast during the fermentation and lagering process. The spent yeast is an efficient starting material to produce yeast extract, which is generally defined as the soluble content of a yeast cell that remains once the cell wall has been destroyed and removed. The variety of different physiologically valuable substances in yeast cells offer the possibility of use as a yeast extract in different areas of the food industry. Jacob et al. [13] demonstrated that the composition of various physiologically valuable substance groups of a yeast extract depends on the biodiversity of the spent yeast from beer production, indicating that brewer's spent yeast should be carefully selected to produce a yeast extract with a defined nutritional composition.

In many cases, food wastes are difficult to utilize for the recovery of value-added products due to their biological instability or potentially pathogenic nature. Fusarium head blight (FHB), a fungal disease caused by several *Fusarium* spp., is one of the most significant causes of economic loss in cereal crops. *Fusarium* spp. produce various amounts and types of trichothecene mycotoxins, with deoxynivalenol being the major one, which are highly toxic to humans and livestock. A method to recover the nutrients from the affected cereals, without the mycotoxins, was reported by Gulsunoglu et al. [14]. The infected grains were initially fermented under solid state cultivation with *Aspergillus oryzae* and/or

*Lactobacillus plantarum*. The fermented material was provided to black soldier fly larvae, which consumed deoxynivalenol-contaminated materials and converted them in insect biomass without accumulating deoxynivalenol in their bodies. This treatment technology using black soldier fly larvae may contribute to reducing the burden of animal protein shortages in the animal feed market.

Varelas [15] compiled up-to-date information on the mass rearing of edible insects for food and feed based on food wastes. Edible insects are insect species that can be used for human consumption but also for livestock feed as a whole, parts of them, and/or protein, and lipid extract.

**Funding:** This research received no external funding.

**Acknowledgments:** The editor wish to thank our article contributors, Editorial Board members, Reviewers, and Assistant Editors of this journal, whose contributions made the publication of this Special Issue possible.

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

#### **References**


© 2020 by the author. 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/).

### *Article* **Production of D-Lactic Acid by the Fermentation of Orange Peel Waste Hydrolysate by Lactic Acid Bacteria**

#### **Daniel Bustamante 1,2, Marta Tortajada 2, Daniel Ramón <sup>2</sup> and Antonia Rojas 2,\***


Received: 31 October 2019; Accepted: 16 December 2019; Published: 18 December 2019

**Abstract:** Lactic acid is one the most interesting monomer candidates to replace some petroleumbased monomers. The application of conventional poly-lactic acid (PLA) is limited due to insufficient thermal properties. This limitation can be overcome by blending poly-D and poly-L-lactic acid. The main problem is the limited knowledge of D-lactic acid (D-LA) production. Efficient biochemical processes are being developed in order to synthesize D-LA from orange peel waste (OPW). OPW is an interesting renewable raw material for biorefinery processes of biocatalytic, catalytic or thermal nature owing to its low lignin and ash content. Bioprocessing of the pretreated OPW is carried out by enzymatic hydrolysis and fermentation of the released sugars to produce D-LA. Several strains of the species *Lactobacillus delbrueckii* ssp. *bulgaricus* have been evaluated for the production of D-LA from OPW hydrolysate using *Lactobacillus delbrueckii* ssp. *delbrueckii* CECT 286 as a reference strain since its performance in this kind of substrate have been widely reported in previous studies. Preliminary results show that *Lactobacillus delbrueckii* ssp. *bulgaricus* CECT 5037 had the best performance with a yield of 84% *w*/*w* for D-LA production and up to 95% (e.e.).

**Keywords:** added value product; D-lactic acid; LAB strains; food waste; orange peel waste

#### **1. Introduction**

Lactic acid is an important chemical and has attracted a great attention due its widespread applications in the food, pharmaceutical, cosmetic, and textile industries. Polylactic acid (PLA) is a biodegradable polymer with great potential in replacing petrochemical polymers and therefore, L-and D-lactic acids are prominent monomers of the bioplastic industry [1]. The morphological, mechanical and thermal properties of the polymer are determined by the presence of different amounts of L- and D-lactic acid monomers or oligomers [2–6]. Microbial production of optically pure lactic acid has extensively been studied because chemically synthesized lactic acid is a racemic mixture [7]. In fact, the optimization of operation conditions is very effective to achieve high selectivity to the isomer of interest [8]. Although the L-isomer has been studied in detail, information on biosynthesis of D-lactic acid (D-LA) is still limited [5,9].

PLA market demand accounts for 11.4% of total bioplastic production worldwide, approximately <sup>18</sup> <sup>×</sup> 104 metric tons per year and the PLA demandis estimated to grow by 28% per year until 2025. However, production costs of PLA are still high, mainly due to expensive fermentation media components. To overcome this problem, several residues have been employed as raw material [3,5,7,10–12]. Production of D-LA from liquid pineapple wastes [13], date juice [14], corn stover [15], hardwood pulp hydrolysate [16] and brown rice [17] has been studied. In this sense, the valorization of food waste to

useful products such as D-LA is a good alternative [1,18,19]. In particular, orange peel and pulp waste (OPW) can be used to produce D-LA after adequate pre-treatment processes [20–22].

Orange waste is the most abundant citrus waste with up to 50 million metric tons of oranges consumed every year [23]. This huge amount of waste accounts for 45%–60% of the total fruit weight, and therefore, a lot of potential applications have been studied for their valorization to date [24]. The main application of this residue is as an ingredient for cattle feed or as pelletized dry solid fuel, but its processing results in highly polluted wastewater [25]. The use of citrus waste to produce compounds of high added value, essential oils, fertilizer, pectin, industrial enzymes, ethanol and absorbents has recently been described [21,23–28]. In addition, orange waste present low levels of lignin and a large amount of sugars [27], which make it an ideal substrate for fermentation processes after the implementation of the required pre-treatment and enzymatic hydrolysis stages.

Lactic acid is produced in high amounts by lactic acid bacteria (LAB) which can do so in a homofermentative way employing the Embden-Meyerhof pathway where lactic acid is the only acid produced, or by the heterofermentative way following the phosphogluconate and phosphoketolase pathway where lactic acid is one of the products and yields of 0.5 g g−<sup>1</sup> of hexose. LABs produce either one or the two forms of lactate [4,11,29,30]. The species *Lactobacillus delbrueckii* ssp. *delbrueckii* has been reported as a homofermentative producer of D-LA using several agro-industrial residues [9]. This bacterium yields 90% D-LA from sugarcane molasses, 95% D-LA from sugarcane juice, 88% D-LA from sugar beet juice [31] and 88% D-LA from orange peel waste (OPW) [32]. Moreover, the species *Lactobacillus delbrueckii* subsp. *bulgaricus* has been used in the dairy industry to transform milk into yogurt and some strains are able to produce highly pure D-LA [33]. Therefore, lactose and whey have been widely studied as raw materials for lactic acid production [34–36], even cloning the D-lactate dehydrogenase gene in *Escherichia coli* [37]. Other studies included wheat flour, molasses, sorghum and lignocellulosic hydrolysates as feedstocks for the production of lactic acid by *Lactobacillus delbrueckii* subsp. *bulgaricus*, especially for L-LA isomer production [11,38]. This fact means that some strains of *Lactobacillus delbrueckii* subsp. *bulgaricus* could be potential candidates for D-LA production from sustainable feedstocks.

The aim of this work was to find LAB strains capable of producing D-LA with high yield and optical purity from OPW as raw material to contribute in the development of biowaste-refineries. For this purpose, several *Lactobacillus delbrueckii* ssp. *bulgaricus* strains were evaluated in comparison to the reference strain *Lactobacillus delbrueckii* ssp. *delbrueckii* CECT 286 which has been reported as a high yield producer of D-LA from biowaste and OPW hydrolysate in particular.

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

#### *2.1. Bacterial Strains, Media and Growth Conditions*

The bacterial strains employed in this study are listed in Table 1 and *Lactobacillus delbrueckii* ssp. *delbrueckii* CECT 286 was used as reference strain. The selected strains were purchased from the Spanish Type Culture Collection (CECT). After being received they were recovered in MRS medium and stored in 20% glycerol at −80 ◦C for long-term preservation. Precultures were prepared in tubes containing MRS medium with a small headspace and incubated overnight at 37 ◦C and static micro-aerobic conditions.

**Table 1.** Lactic acid bateria (LAB) strains selected for D-lactic acid production screening.


Screening of LAB strains was performed in 15 mL tubes at 37 ◦C and using a medium with sugars resembling OPW hydrolysate as follows: MRS broth plus glucose 30 g L<sup>−</sup>1, fructose 20 g L−1, galactose 5gL−<sup>1</sup> and arabinose 6 g L−1. Cultures were inoculated in duplicate with 5% *v*/*v* of preculture and were incubated in orbital shaker at 200 rpm. Aerobic and micro-aerobic conditions were tested at pH 6.2 for 40 h.

#### *2.2. OPW Hyrolysate Tolerance Assays*

Tolerance assays were performed in triplicate using selected strains and preparing a multi-well plate with 200 μL of MRS with OPW hydrolysate diluted at 50%, 85% and 100% *v*/*v* as culture medium for each condition. Precultures were prepared in MRS and inoculated at 10% of total volume. A microplate incubator spectrophotometer was used with temperature set at 37 ◦C for 45 h. The plate was shaken every hour for 5 seconds before each OD600 measurement to obtain the growth curves of the strains.

#### *2.3. Fermentation Assays*

Strains were cultured in 50 mL tubes containing MRS with 85% *v*/*v* OPW hydrolysate at pH 6.2, 37 ◦C and 45 ◦C in micro-aerobic conditions. An additional assay was done by adjusting pH at 5.8 each 24 h with NaOH 5 M. All runs started by inoculating 15% *v*/*v* of preculture and then incubated in an orbital shaker at 200 rpm for 120 h.

The experiments in the bioreactor setup were performed in 1.5 L Applikon® in batch mode with OPW hydrolysate at 85% *v*/*v* with MRS and 5 g L−<sup>1</sup> meat extract as additional nitrogen source. The OPW hydrolysate was sterilized using sterile glass fiber and cellulose acetate membrane filters with 0.2 μm of pore size, and then added to the bioreactor. Before the inoculum addition, the anaerobic atmosphere was obtained by stripping the oxygen off with a nitrogen stream. The experimental conditions were set up at 37 ◦C, 200 rpm, and pH of 5.8, adding NaOH 5 M or HCl 2 M for pH control during fermentation.

#### *2.4. OPW Pretreatments*

The substrate used in this study was OPW obtained from juice elaboration. These residues were blade-milled to a final particle diameter of around 5 mm and then, samples were subsequently stored in a freezer at −20 ◦C until use. The characterization of the raw material was performed according to the NREL procedures for determination of structural carbohydrates and free sugars, in addition to extractives [39–41], while moisture was assessed by using an infrared drying balance at temperatures between 70 and 90 ◦C until constant weight. The results obtained by applying the NREL methodology are compiled in Table 2. For D-LA production assays, OPW was milled down to 1–2 mm particle size and hydrolysis was carried out at 10% *w*/*w* of dry solid, 50 ◦C, 300 rpm and initial pH of 5.2 using enzyme cocktails with cellulases, β-glucosidase, xylanase, β-xylosidase, pectinase, and auxiliary activities (Celluclast 1.5 l, Novozym 188, Pectinex Ultra SP-L gifted by Novozymes) as described by de la Torre and colleagues [22].


**Table 2.** OPW composition analysis according to NREL protocols.

#### *2.5. Analytical Procedures*

The content of sugars and organic acids was determined by HPLC liquid chromatography (2695 HPLC with a refractive Index Detector 2414; Waters, Cerdanyola del Vallés, Spain) using a Rezex ROA Organic acid column, with H2SO4 at 2.5 mM and 0.5 mL min−<sup>1</sup> flow. The optical purity of D-LA was determined by HPLC (Agilent Technologies 1100 Series, Waldbronn, Germany) using a DAD detector, a Chirex 3126 (D)-penicillamine (250 × 4.6; Phenomenex) column working at room temperature, and a CuSO4 1 mM solution as mobile phase flowing at 1.2 mL min<sup>−</sup>1.

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

#### *3.1. Screening of LAB Strains for D-LA Production*

Lactic acid production was tested in 15 mL tubes containing 3 mL of culture resembling OPW hydrolysate for aerobic conditions and 14 mL of culture for micro-aerobic conditions to compare the behavior of the different LAB strains. Results are shown in Figure 1. *Lactobacillus delbrueckii* ssp. *bulgaricus* CECT 4005 and CECT 5038 did not produce a significant amount of lactic acid while *L. delbrueckii* ssp. *bulgaricus* CECT 5036 produced up to 14 g L−<sup>1</sup> of lactic acid racemic mixture in aerobic and micro-aerobic conditions. Furthermore, three strains, *L. delbrueckii* ssp. *bulgaricus* CECT 4006, CECT 5035 and CECT 5037 transformed sugars into lactic acid in micro-aerobic condition with D-LA enantiomeric excess in the same way as *L. delbrueckii* ssp. *delbrueckii* CECT 286. Those strains produced around 15 g L−<sup>1</sup> of lactic acid with around 75% (e.e.) of D-LA while *L. delbrueckii* ssp. *delbrueckii* CECT 286 reached 92% (e.e.) of D-LA. Therefore, those three strains were selected to study D-LA production from OPW hydrolysate in micro-aerobic conditions.

**Figure 1.** D-LA production in 15 mL tubes with MRS medium containing sugars resembling OPW hydrolysate using LAB strains selected for screening. A. Aerobic conditions. B. Micro-aerobic conditions.

Previous reports showed that lactose rather than glucose markedly increases the growth rate of *L. delbrueckii* ssp. *bulgaricus* strains [33,34]. Therefore, transport systems of sugars other than lactose are likely to vary among these strains and hence, some strains, such as *L. delbrueckii* ssp. *bulgaricus* CECT 4005 and CECT 5038, appear to have difficulties to assimilate the sugars tested in this work. Moreover, strains such as *L. delbrueckii* ssp. *bulgaricus* CECT 5035 and CECT 5037 show low yield in assays at aerobic conditions in the same way as *L. delbrueckii* ssp. *delbrueckii* CECT 286. It is known that during growth, toxic oxygen derivatives are produced for LAB strains in aerobic conditions, but the enzymes required to eliminate them seem not to be expressed in some *L. delbrueckii* ssp. *bulgaricus* strains [42]. Reducing agents may provide protection against toxic products, particularly if growth conditions are not strictly anaerobic. However, with exception of *L. delbrueckii* ssp. *bulgaricus* CECT 5036, the other strains showed higher selectivity to D-LA than *L. delbrueckii* ssp. *delbrueckii* CECT 286 in aerobic conditions and as mentioned above, *L. delbrueckii* ssp. *bulgaricus* CECT 5036 have similar results at

aerobic and micro-anaerobic conditions but produced racemic mixture in both cases. *L. delbrueckii* ssp. *bulgaricus* CECT 4005 appears to prefer aerobic conditions but yields are still low.

#### *3.2. Use of OPW Hydrolysate for D-LA Production by Selected Strains*

The OPW hydrolysates were prepared following the methodology described in Section 2.4. and developed by de la Torre and colleagues [22] obtaining a glucose yield around 60% *w*/*w* which corresponds to around 30 g L<sup>−</sup>1, and obtaining a total sugar concentration above 50 g L−1. Therefore, OPW is a good source of several monosaccharides but also have essential oils rich in limonene and containing terpenes and phenolics with some antimicrobial activity [21]. The tolerance of the strains to the substrate was tested with different concentrations of OPW hydrolysate ranging from 50% to 100% *v*/*v* diluted with MRS broth. Growth monitoring was performed in a micro-plate incubator for 48 h (Figure 2). Microorganisms grew up well at 50% *v*/*v* hydrolysate content, but the strain *L. delbrueckii* ssp. *delbrueckii* CECT 286 tolerated the hydrolysate and was able to grow even when hydrolysate content was 100% *v*/*v*. *Lactobacillus delbrueckii* ssp. *bulgaricus* CECT 4006 appears to be more sensitive to OPW hydrolysate while *L. delbrueckii* ssp. *bulgaricus* CECT 5037 was able to grow up at any OPW concentration; however, the higher the hydrolysate concentration, the higher the lag phase and the lower the growth. Differences lied on the performance of the strains, which is slightly lower when using OPW hydrolysates, probably due to the presence of essential oil components, either terpenes or phenolics. However, *Lactobacilli* are able to withstand relatively high concentrations of citrus extracts [43].

**Figure 2.** Growth curves for tolerance assays to OPW hydrolysate in microplates and microarebic conditions. (**A**) *Lactobacillus delbrueckii* ssp. *delbreckii* CECT 286. (**B**) *Lactobacillus delbrueckii* ssp. *bulgaricus* CECT 4006. (**C**) *Lactobacillus delbrueckii* ssp. *bulgaricus* CECT 5035. (**D**) *Lactobacillus delbrueckii* ssp. *bulgaricus* CECT 5037. The results were obtained as the average of three replicates and standard deviation was lower than 0.5%.

Concerning the nutritional requirements, previous studies showed that niacin, calcium pantothenate, riboflavin, and vitamin B12 were essential for the growth of *L. delbrueckii* ssp. *bulgaricus,* and that folic acid, pyridoxal, and CaCl2 were important for efficient growth [44,45]. There could be discrepancies due to

differences in medium composition or to strain-specific requirements as in the case of *L. delbrueckii* ssp. *bulgaricus* CECT 5037, which not only seems to tolerate hydrolysate, but also seems to grow with less strict nutritional requirements. Although *L. delbrueckii* ssp. *delbueckii* CECT 286 and *L. delbrueckii* ssp. *bulgaricus* CECT 5037 have shown highest robustness cultured in OPW hydrolysate, the next assays were performed using the four selected strains and inoculating the cells recovered from 15% *v*/*v* of preculture with respect to the volume of culture at 85% *v*/*v* OPW hydrolysate diluted with MRS medium and micro-aerobic conditions. The inoculum amount was increased to compare the performance of the selected strains with the maximum concentration of OPW hydrolysate during the preliminary fermentation trials.

The optimal growth temperature for *Lactobacilli* ranges from 30 to 40 ◦C, although some thermophilic strains grow well and have highly activated metabolism at temperatures around 45 ◦C [35]. The four *Lactobacillus* strains selected were cultured at 37 ◦C and 45 ◦C during 120 h to test their activity at conditions as close as possible to those of hydrolysis stage and therefore, to evaluate if the hydrolysis and fermentation stages could be done simultaneously (SSF) as a preliminary result for the future optimization and scale-up of the process. In general, the SSF process offers better yields because it avoids product inhibition and results in higher productivity [10]. Aghababaie and colleagues [36] reported that optimum temperature and pH for growth and lactate production from whey for *L. delbrueckii* ssp. *bulgaricus* were 44 ◦C and 5.7, respectively. However, the results in Figure 3 show that the strains selected in this study produced D-LA up to 90% (e.e.) in all cases, but the performance of the strains was still better at 37 ◦C using OPW hydrolysates.

**Figure 3.** D-LA production results of three *L. delbrueckii* ssp. *bulgaricus* selected in front of *L. delbrueckii* ssp. *delbrueckii* CECT 286 using OPW hydrolysate at 85% *v*/*v* and incubated at 37 ◦C and 45 ◦C to compare strains performance at different temperatures.

Similarly to temperature, the effect of pH change on growth characteristics varied between different species of LAB and in most cases, a decrease of lactate production with a decrease of pH were observed [35]. Therefore, the strains were cultured in 85% *v*/*v* OPW hydrolysate and pH was adjusted to 5.8 each 24 h during fermentation to test their capacity of production with pH regulation. Cultures were incubated at 37 ◦C and micro-aerobiosis for 120 h. The results show that sugar consumption and yields were higher when pH was adjusted, and D-LA up to 95% (e.e.) was produced (Figure 4). *L. delbrueckii* ssp. *bulgaricus* CECT 5037 showed the best results in comparison to the other *L. delbrueckii* ssp. *bulgaricus* strains and its performance was comparable to *L. delbrueckii* ssp. *delbrueckii* CECT 286 strain using OPW hydrolysate, whose productivities were between 0.23 and 0.29 g L−<sup>1</sup> h<sup>−</sup>1, respectively. Due to the homofermentation of *L. delbrueckii* ssp. *delbrueckii* and *L. delbrueckii* ssp. *bulgaricus* [9,11], only lactic acid could be produced. Nevertheless, a small increase in ethanol concentration onwards of 48 h of fermentation was observed during pH regulation trials. The explanation for this fact, according to the literature [38,46], is that some homofermenters, when grown in limited sugar environment or in the presence of different sugars, can lead to other end products. The main difference is in pyruvate metabolism, but the homofermentation pathway is still used. Additionally, the accumulation of ethanol

in the medium (2–3 g L−1) was by far very low to change significantly the generation of the target product. Thus, D-LA continues to be the major fermentation product, and the metabolism of the strains can be considered homofermentative.

**Figure 4.** D-LA production results of three *L. delbrueckii* ssp. *bulgaricus* selected in front of *L. delbrueckii* ssp. *delbrueckii* CECT 286 using OPW hydrolysate at 85% *v*/*v* and adjusting pH at 5.8 each 24 h to evaluate strains performance with pH regulation. The results of standard deviation for the strains with respect to CECT 286 strain are: SDCECT4006 = 12.02; SDCECT5035 = 34.22; SDCECT5037 = 0.69.

#### *3.3. D-LA Production by L. delbrueckii ssp. delbrueckii CECT 286 vs. L. delbrueckii ssp. bulgaricus CECT 5037*

Preliminary scale-up assays were performed in 1.5 liter bioreactor by controlling pH at 5.8 in batch mode. Previous results showed that the performance of the strains was better under micro-aerobic conditions, so the bioreactor tests were performed under anaerobic conditions using a nitrogen stream. Cells from 15% *v*/*v* MRS preculture were inoculated in 85% *v*/*v* OPW hydrolysate with MRS and supplemented with 5 g L−<sup>1</sup> of meat extract. According to literature, the more supplemented the medium, the higher the value of final biomass and the higher the productivity of the lactic acid attainable [45,47]. Previous work showed the importance of meat extract and yeast extract in the production of D-LA, probably not due to the total amount of nitrogen but to the growth factors and vitamins contained in these extracts [32]. Fermentation was finished at 72 h (Figure 5), *L. delbrueckii* ssp. *delbrueckii* CECT 286 produced 45 g L−<sup>1</sup> of lactic acid (99.5% D-LA (e.e.)) with a yield of 86% *w*/*w* while *L. delbrueckii* ssp. *bulgaricus* CECT 5037 produced 39 g L−<sup>1</sup> of lactic acid (99.3% D-LA (e.e.)) with a yield of 84% *w*/*w*.

**Figure 5.** Growth, sugar consumption and D-LA production from OPW hydrolysate in bioreactor and batch mode. A. *Lactobacillus delbrueckii* ssp. *delbreckii* CECT 286. B. *Lactobacillus delbrueckii* ssp. *bulgaricus* CECT 5037. CDM = Cell dry weight.

The yields obtained were similar to those obtained in previous assays by adjusting pH, but the productivities were higher in this case, with values of 0.63 and 0.55 g L−<sup>1</sup> h−1, respectively. The experiments show that sugars are not completely consumed during fermentation, probably due to deficiencies in the nutritional requirements of the strains. Therefore, the D-LA production process is further optimizable using *L. delbrueckii* ssp. *bulgaricus* CECT 5037 as a promising D-LA producer from OPW hyrolysate and other sustainable feedstocks to contribute in the development of bio-waste refineries. In this regard, commercially important LA-producing LAB strains, such as *Lactobacillus* and *Sporolactobacillus* strains, are particularly useful because of their high lactic acid yield, high acid tolerance, and their ability to be metabolically engineered [9,12]. Efficient conversion of biomass to D-LA still faces considerable challenges, such as high energy demand and high enzyme cost for pretreatment of lignocellulosic biomass, inefficiency of sugar utilization by microorganisms, and undesired byproducts generated during the fermentation process [46]. Table 3 summarizes studies of D-LA production from sustainlable feedstocks such as agro-industrial residues by wild-type stains. These results indicate that OPW hydrolysate is an interesting feedstock for the production of D-LA, since the product yield is close to its theoretical value (1 g g-1) in most cases. Apart from that, the productivity value is quite high and very attractive when industrial developments are envisaged [32,48]. It is common that bioprocesses based on biomass waste give poorer results than their control experiments based on sugar mixtures resembling the hydrolysates composition. In this case, yields achieved have close values in both cases. According to the achieved purity of lactic acid (> 95%), differences were not observed when OPW hydrolysate is used, suggesting that the waste compounds do not influence D-LA purity.


**Table 3.** D-LA production from sustainable feedstocks in batch cultures by wild-type LAB strains.

\* Preliminary results of LAB screening for further optimization. SSF: Simultaneous saccharification and fermentation; SHF: Separate hydrolysis and fermentation.

The results obtained with the *L. delbrueckii* ssp. *bulgaricus* CECT 5037 strain are promising since performance of the strain was comparable to *L. delbrueckki* ssp. *delbrueckii* CECT 286 strain performance using OPW hydrolysate at the conditions tested in this work. Previous studies show that the reference strain can reach a productivity of 2.35 g L−<sup>1</sup> h−<sup>1</sup> when fermentation conditions are optimized [32]. Therefore, future work with *L. delbrueckii* ssp. *bulgaricus* CECT 5037 will be focused on optimization of fermentation methodology, including the method of inoculation of the cultures, improvement of culture media by testing low cost nutrient sources, as well as the evaluation of operational costs in developing a sustainable lactic acid production process.

#### **4. Conclusions**

Six strains of the species *Lactobacillus delbrueckii* ssp. *bulgaricus* were evaluated for the production of D-LA from OPW hydrolysate in comparison to the reference *Lactobacillus delbrueckii* ssp. *delbrueckii* CECT 286 strain. Remarkably, *Lactobacillus delbrueckii* ssp. *bulgaricus* CECT 5037 is able to tolerate the OPW hydrolysate and produce D-LA up to 95% (e.e.). The results of strain performance show a yield of 84% *w*/*w* for lactic acid production that is close to the yield of 86% *w*/*w* obtained with the reference *Lactobacillus delbrueckii* ssp. *delbrueckii* CECT 286 strain in this work and 88% *w*/*w* reported from previous works when process improvement was foreseen. Experiments will be underway to develop the process and further optimization will contribute to providing a suitable alternative to biowaste-refinery processes using OPW and other residual feedstocks as a potential substrate for valorisation.

**Author Contributions:** D.R. and M.T. conceived the research; A.R. and D.B. designed the experiments; D.B. performed the experiments; A.R. evaluated the results; D.B. wrote the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by European Commission, grant number EIB.12.007

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2019 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/).

*Article*

## **The Second-Generation Biomethane from Mandarin Orange Peel under Cocultivation with Methanogens and the Armed** *Clostridium cellulovorans*

**Hisao Tomita 1,**† **and Yutaka Tamaru 1,2,3,\***


Received: 10 September 2019; Accepted: 31 October 2019; Published: 4 November 2019

**Abstract:** This study demonstrates that the consortium, which consists of the microbial flora of methane production (MFMP) and*Clostridium cellulovorans* grown with cellulose, can perform the direct conversion of cellulosic biomass to methane. The MFMP was taken from a commercial methane fermentation tank and was extremely complicated. Therefore, *C. cellulovorans* grown with cellobiose could not perform high degradation ability on cellulosic biomass due to competition by various microorganisms in MFMP. Focusing on the fact that *C. cellulovorans* was cultivated with cellulose, which is armed with cellulosome, so that it is now armed *C. cellulovorans*; the direct conversion was carried out by the consortium which consisted of MFMP and the armed *C. cellulovorans*. As a result, the consortium of *C. cellulovorans* grown with cellobiose and MFMP (CCeM) could not degrade the purified cellulose and mandarin orange peel. However, MFMP and the armed *C. cellulovorans* reduced 78.4% of the total sugar of the purified cellulose such as MN301, and produced 6.89 mL of methane simultaneously. Furthermore, the consortium consisted of MFMP and the armed *C. cellulovorans* degraded mandarin orange peel without any pretreatments and produced methane that was accounting for 66.2% of the total produced gas.

**Keywords:** methanogenesis; cellulosic biomass degradation; consortia; armed *C. cellulovorans*

#### **1. Introduction**

Although the first-generation biofuels, which are made from corn and sugarcane, have become widespread, there is concern about competition with food supply. Without competing for food, second-generation biofuels are produced from non-edible biomass such as agricultural wastes and cellulosic substrates [1,2].

A plant cell wall is composed of cellulose, hemicellulose, lignin, pectin, etc. Cellulose is a fiber of d-glucose monomers and has strong crystalline [3]. Moreover, cellulose, hemicellulose, and lignin compose the rigid and complex structures [4]. Hemicellulose is a heteropolymer such as xylan, glucuronoxylan, arabinoxylan, glucomannan, and xyloglucan. In addition, lignin, phenol compounds, reinforces the structure of cellulose and hemicellulose, making it more difficult to degrade. Thus, since rigid and complex structures are constructed in cellulosic biomass, it is very difficult to degrade them enzymatically.

Orange juice is one of the major fruit juices and almost the same amount of orange waste as orange juice comes out as a byproduct in orange juice factories. Therefore, it has been considered that such orange wastes are available for non-edible biomass all over the world. However, d-limonene, which is included in citrus, has an extremely toxic effect on fermenting microorganisms [5,6]. Depending on the type of citrus, the orange peel contains approximately 3% limonene. Therefore, it was necessary to separate d-limonene before the cultivation or to protect microorganisms from d-limonene by encapsulation or immobilization [7,8]. Recently, it was reported that *Clostridium cellulovorans* can degrade orange wastes in the culture including d-limonene and *Clostridium beijerinckii* can carry out isopropanol–butanol–ethanol (IBE) fermentation, which is a bacterial fermentation process producing isopropanol instead of acetone on acetone–ethanol–butanol (ABE) fermentation [9–11]. Thus, the breakthrough was obtained utilizing orange wastes for second-generation biofuels without any pre-treatment.

Some *Clostridia*, such as *C. cellulovorans* and *Clostridium thermocellum*, are known to have the ability to degrade cellulosic biomass efficiently using cellulosomes and secreted non-cellulosomal enzymes [12]. Among those species, we have been studying*C. cellulovorans*, whichis amesophilic and anaerobic cellulolytic bacterium [13]. *C. cellulovorans* degrades not only cellulose but also hemicelluloses consisting of xylose, fructose, galactose, andmannose [14–16]. Whole-genome sequencing of*C. cellulovorans* and the exoproteome profiles revealed 57 cellulosomal protein-encoding genes and 168 secreted-carbohydrase-encoding genes [17,18]. Furthermore, the high degradation ability on plant cell walls has so far been reported [19]. *C. cellulovorans* grown in the culture with cellulose has large protuberances on its surface [20] and the protuberances contain cellulosomes [21]; on the other hand, *C. cellulovorans* grown in the culture with cellobiose does not have the protuberances (Figure 1a,b). *C. cellulovorans* acquires a high ability to degrade the plant cell wall with cellulosome, in other words, *C. cellulovorans* is armed with cellulosome to attack the plant cell wall. Therefore, it can be said that it is the armed *C. cellulovorans* that grow with cellulose. Although *C. cellulovorans* has high degradation ability on plant cell walls in the *C. cellulovorans* monoculture, there are challenges in that *C. cellulovorans* cannot perform high degradation ability in the co-culture or consortia of other microorganisms. Most of the reports used *C. cellulovorans* grown with cellobiose, and these challenges were most likely derived using *C. cellulovorans* grown with cellobiose. Therefore, this study focuses on utilizing the armed *C. cellulovorans* especially for the consortium of other microorganisms.

**Figure 1.** (**a**) Cell surface of *C. cellulovorans* grown with cellobiose and (**b**) cell surface of *C. cellulovorans* grown with cellulose. White bar indicates 2 μm. Modified from [21].

Many studies on methane (CH4) fermentation have been reported in a wide range of study fields [22]. Methane fermentation using cellulosic biomass is also second-generation biomethane. Since methane production is carried out by the complex microbial flora including methanogens, the second-generation biomethane process needs the consortium to be constructed with microbial flora of methane production (MFMP) and microorganisms which can degrade cellulosic biomass, such as *C. cellulovorans*. The consortium of *C. cellulovorans* with MFMP (CCeM) can degrade sugar beet pulp, which is the residue in a sugar refinery factory, and ferments biogas included methane simultaneously [23]. However, the relict sugars in sugar beet pulp were possible to help *C. cellulovorans* to survive and coexist with MFMP, and the same CCeM could not carry out degrading the purified cellulose and producing methane.

In the present study, we investigated the degradation ability on cellulosic biomass of CCeM that was consistent with the armed *C. cellulovorans* (ACCeM).

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

#### *2.1. Materials*

Mandolin oranges were purchased at a grocery store in Japan in 2017. Flavedo and albedo, hereafter called removed peel, were cut into strips with scissors just after removing from a mandolin orange. The strip size was approximately 10 mm in length and 2 mm in width (Figure 2). The peel was used as fresh, not dried, and ground. Furthermore, the peel was not treated by any chemicals. The dried weight of the removed peel was measured, it contained 71.6% water. The substrate concentration of removed peel, the purified celluloses, such as Avicel (Sigma, St. Louis, MO, USA) and MN301 (MACHEREY-NAGEL, Düren, Deutschland), and cellobiose (Sigma, St. Louis, MO, USA) was 0.5% (*w*/*v*) of the dry weight. Avicel is crystalline cellulose powder that is industrially refined from natural cellulose, and the particle size of Avicel is less than 50 μm. MN301 is also industrially refined cellulose powder, and 80% of the particle size is less than 160 μm.

**Figure 2.** (**a**) Removed peel before cutting. (**b**) Strips of removed peel in the medium.

#### *2.2. Microorganism and Culture Condition*

The medium was partially modified by *Clostridium cellulovorans* medium [11]. One litter medium contained 4 g of yeast extract, 1 mg of Resazurin salt, 1 g of L-cysteine-HCl, 5 g of NaHCO3, 0.45 g of K2HPO4, 0.45 g of KH2PO4, 0.3675 g of NH4Cl, 0.9 g of NaCl, 0.1575 g of MgCl2.6H2O, 0.12 g of CaCl2.2H2O, 0.85 mg of MnCl2.4H2O, 0.942 mg of CoCl2.6H2O, 5.2 mg of Na2EDTA, 1.5 mg of FeCl2.4H2O, 0.07 mg of ZnCl2, 0.1 mg H3BO3, 0.017 mg of CuCl2.2H2O, 0.024 mg of NiCl2.6H2O, 0.036 mg of Na2MoO4.2H2O, 6.6 mg of FeSO4.7H2O, and 0.1 g of p-aminobenzoic acid and was adjusted to pH 7. *C. cellulovorans* 743B (ATCC 35296) was used and anaerobically cultivated in 0.5% (*w*/*v*) cellobiose, Avicel and MN301 at 37 ◦C stationary for 19 h, for 4 days and 2 days respectively. The MFMP was obtained from methane fermentation digested liquid in January 2017 at Gifu in Japan. The MFMP was anaerobically cultivated in *Clostridium cellulovorans* medium with 0.5% (*w*/*v*) glucose (Wako) and 0.25% (*w*/*v*) cellobiose, or 0.5% (*w*/*v*) ryegrass leaves at 37 ◦C for 19 h stationary. The ryegrass leaves were obtained in May 2017 at Aichi in Japan, and were dried and grained to a powder.

#### *2.3. Data Deposition*

The sequences of MFMP that are reported in this paper have been deposited in the DNA Data Bank of Japan (DDBJ) (accession 104 no. DRR160954).

#### *2.4. Measurement of Total Sugar Concentration*

The total sugar concentration was measured from the precipitation after centrifugation which was 20,000 rpm at 4 ◦C for 5 min. The precipitation was washed by phosphate buffered salts and 5N NaOH. The total sugar concentration of the washed precipitation was measured by the phenol-sulfuric acid method as d-glucose equivalents.

#### *2.5. Gas Concentration*

The produced gas volume of the head space in the culture vial was collected and measured by a syringe (Terumo, Tokyo, Japan). The concentrations of CH4, H2, and CO2 were measured by a gas chromatograph GC-8A (Shimadzu, Kyoto, Japan) with a Thermal Conductivity Detector (TCD) and a column SINCARBON ST (full length 6 m, inner diameter 3 mm; Shinwa, Kyoto, Japan). The column temperature was 200 ◦C. Argon was a carrier gas and set at a flow rate of 50 mL/min. The injection volume of each sample was 5 mL. The concentration of CH4 was also measured by a gas chromatograph GC-2010plus (Shimadzu, Kyoto, Japan) with a capillary column Rt-Q-BOND (30 m, inner diameter. 0.32 mm; RESTEK, Centre, PA, USA). The oven temperature was 250 ◦C and the column temperature was 150 ◦C. Nitrogen was the carrier gas and set at a flow rate of 1.21 mL/min. The injection volume of each sample was 0.5 mL.

#### *2.6. Organic Acid Concentration*

The concentration of organic acids was measured by high-performance liquid chromatography (HPLC), CBM-20A, LC-20AD, CTO-20AC, SPD-20A, and DGU-20A3 (Shimadzu, Kyoto, Japan) with a UV detector and a column KC-811 (300 mm × 2 mm, inner diameter. 8 mm; Showa Denko, Tokyo, Japan). The column temperature was 60 ◦C. The Bromothymol blue (BTB) post-column method was used. The eluent was 2 mM perchloric acid, and the flow rate was 1.0 mL/min. The reagent was 0.2 mM BTB and 15 mM disodium hydrogen phosphate, and the flow rate was 1.2 mL/min at the wavelength of 445 nm. The injection volume of each sample was 20 μL.

#### *2.7. Cell Growth*

Cell growth was measured by Lumitester PD-20, LuciPac Pen and adenosine triphosphate (ATP) eliminating enzyme (Kikkoman Biochemifa, Tokyo, Japan). It is known that integrated intracellular ATP concentration correlates with cell growth [24]. Cell growth was estimated by measuring ATP

concentration of 0.1 mL of cell culture according to the manufacturer's instruction and was expressed by the relative light unit (RLU) value.

#### *2.8. Statistics*

The data were analyzed for statistical significance using Welch's *t*-test. The difference was assessed with a two-sided test with an α level of 0.05.

#### **3. Results**

#### *3.1. Degrading Cellulose and Removed Peel under Cocultivation with Methanogens and Non-Armed C. cellulovorans*

*C. cellulovorans* was pre-cultivated in media containing 0.5% (*w*/*v*) of cellobiose, which meant *C. cellulovorans* was not armed with cellulosome. Anaerobic batch cultivations of *C. cellulovorans*, CCeM, and MFMP were carried out in a 50-mL medium containing 0.5% (*w*/*v*) of Avicel and removed peel at 37 ◦C without shaking.

According to measured cell growth of precultures, inoculation amounts for monocultures of *C. cellulovorans* and MFMP were decided so that initial RLU values of each monoculture came close to 2000. RLU values of the preculture of *C. cellulovorans* and MFMP were 45,310 and 50,494, respectively. The inoculation volume was decided as 2 mL for 50 mL monoculture; it was 26 times the dilution so that the initial RLU value of monoculture of *C. cellulovorans* and MFMP were 1743 and 1942, respectively. CCeM was inoculated 2 mL each from both precultures so that concentrations of cell growth against substrate became the same as the monoculture.

RLU values increased for 1 day of cultivation and then cell growth was observed in all cultures (Figure 3a,b). Interestingly, RLU profiles of CCeM and MFMP were quite similar, even CCeM included *C. cellulovorans* and MFMP both. The total sugar concentrations were significantly reduced in *C. cellulovorans* monocultures with Avicel and removed peel after 11 days of cultivation, it was demonstrated that *C. cellulovorans* could degrade Avicel and removed peel. However, the total sugar concentrations in CCeM with Avicel and removed peel were almost the same as that of MFMP; *C. cellulovorans* could not degrade Avicel and removed peel in cultures included MFMP (Figure 3c,d). It was thought that *C. cellulovorans* did not grow in CCeM because cell growths of CCeM and MFMP were almost the same. Methane in the head space was not detected in CCeM with removed peel and that was slightly detected in CCeM and MFMP with Avicel (Figure 3e,f).

**Figure 3.** *Cont*.

**Figure 3.** Cultivation of *C. cellulovorans*, cellobiose and MFMP (CCeM) and microbial flora of methane production (MFMP) with Avicel. (**a**) Cell growth, where *C. cellulovorans* (-\_open circle), CCeM (-\_closed circle) and MFMP (Δ\_open triangle) are included; (**c**) total sugar concentration; (**e**) gas production of H2 (hatched bar), CH4 (closed bar), and CO2 (open bar) are included. Cultivation of *C. cellulovorans*, CCeM, and MFMP with removed peel. (**b**) Cell growth, where *C. cellulovorans* (-\_open circle), CCeM (-\_closed circle), and MFMP (Δ\_open triangle) are included; (**d**) total sugar concentration; (**f**) gas production of H2 (hatched bar), CH4 (closed bar), and CO2 (open bar) are included. Values with error bars are mean ± SE of three independent samples. SE means a standard error. An asterisk indicates a significant difference (*p* < 0.05).

#### *3.2. Cellulose Degradation and Methane Production under Cocultivation with Methanogens and the Armed C. cellulovorans*

*C. cellulovorans* was pre-cultivated in media containing 0.5% (*w*/*v*) of MN301 for 2 days, which meant *C. cellulovorans* was armed with cellulosome. Anaerobic batch cultivations of the armed *C. cellulovorans*, ACCeM, and MFMP were carried out in a 40-mL medium containing 0.5% (*w*/*v*) of MN301 at 37 ◦C without shaking. RLU values of the preculture of the armed *C. cellulovorans* and MFMP were 6786 and 38,538, respectively. RLU value of the armed *C. cellulovorans* culture did not increase over 10,000 like the non-armed *C. cellulovorans*. Therefore, inoculation amounts for monocultures of the armed *C. cellulovorans* and MFMP were decided so that initial RLU values of each monoculture became close to 1000. Inoculation volumes of the armed *C. cellulovorans* and MFMP were decided as 7 mL and 1.2 mL, respectively. The armed *C. cellulovorans* was 6.69 times dilution and MFMP was 40.17 times dilution so that initial RLU values of monoculture of the armed *C. cellulovorans* and MFMP were 1014 and 959, respectively. ACCeM was inoculated in 7 mL of the armed *C. cellulovorans* preculture and 1.2 mL of MFMP preculture so that concentrations of cell growth against substrate became the same as the monoculture. Cell growths in the armed *C. cellulovorans* monoculture, ACCeM culture, and MFMP monoculture were measured for 7 days of cultivation. RLU values in ACCeM culture and MFMP monoculture rapidly increased more than 100,000 for 10 h of cultivation. The RLU value in the armed *C. cellulovorans* monoculture increased for 2 days of cultivation and then decreased slowly. Interestingly, the RLU value in ACCeM culture was higher than that in MFMP monoculture after 2 days of cultivation. It suggested that the armed *C. cellulovorans* grew for a couple of days of cultivation and coexisted with MFMP (Figure 4a). The total sugar concentration in the armed *C. cellulovorans* monoculture decreased to 0.38 mg/mL from 5 mg/mL for 7 days of cultivations; it indicated that 92.4% of MN301 was degraded by the armed *C. cellulovorans* for 1 week (Figure 4b). The total sugar concentration in ACCeM culture did not decrease for 3 days of cultivation, however it rapidly decreased from 4 days of cultivation. It also suggested that the armed *C. cellulovorans* survived for 3 days using brought in cellulosome while adapting to coexisting MFMP. Total sugar concentration in ACCeM culture decreased to 1.08 mg/mL for 7 days of cultivation, therefore it was demonstrated that ACCeM could degrade 78.4% of MN301 for 1 week. The total sugar concentration in MFMP monoculture did not decrease for 27 days of cultivation, it indicated that MFMP did not have a capability to degrade MN301. The total sugar concentrations in the armed *C. cellulovorans* monoculture and ACCeM culture was significantly low than that in MFMP for 7 days of cultivation (Figure 4c). Total gas volume in the ACCeM culture rapidly increased for 10 h of cultivation, and the total gas volume kept for 1–7 days of cultivation. Interestingly, the methane proportion in the total gas increased for 1–7 days of cultivation, even though total gas volume did not increase. Furthermore, methane volume continuously increased for 21 days of

cultivation. Finally, methane volume increased 6.89 mL for 1–21 days of cultivation, therefore it revealed that methane production rate from MN301 was 0.014 mL/h (Figure 4d). Organic acid concentrations in the culture supernatant of ACCeM were measured. Lactic acid concentration was slightly detected for 1 day of cultivation and was not detected after 2 days of cultivation (Figure 4e). Formic acid concentration was temporally increased at 1 day of cultivation, and then it was not detected after 2 days of cultivation. Acetic acid concentration increased for 6 days of cultivation that was the same term when the total sugar concentration reached 78.4%. Acetic acid concentration turned to decrease from 8 days of cultivation. On the other hand, propionic acid concentration increased but did not turn to decrease. It was suggested that MN301 was degraded by *C. cellulovorans* and acetic acid and methane were produced, and methane was continuously produced utilizing acetic acid after cellulose was degraded. Moreover, it suggested that acetic acid was converted to methane in preference to propionic acid.

**Figure 4.** *Cont*.

**Figure 4.** Cultivation of *C. cellulovorans*, CCeM, and MFMP with Avicel. (**a**) Cell growth, where *C. cellulovorans* (-\_open circle), CCeM (-\_closed circle), and MFMP (Δ\_open triangle) are included; (**b**) total sugar concentration; (**c**) time course of total sugar, where *C. cellulovorans* (-\_open circle), CCeM (-\_closed circle), and MFMP (Δ\_open triangle) are included; (**d**) time course of gas production, where total gas volume (-\_closed circle) and methane volume (Δ\_open triangle) are included; (**e**) time course of organic acids concentration of CCeM, where lactic acid (closed bars), formic acid (open bars), acetic acid (closed bars), and propionic acid (dotted bars) are included. Values with error bars are mean ± SE of three independent samples. SE means a standard error. An asterisk indicates a significant difference (*p* < 0.05).

#### *3.3. Removed Peel Degradation and Methane Production under Cocultivation with Methanogens and the Armed C. cellulovorans*

It was confirmed that the armed *C. cellulovorans* could degrade cellulose and MFMP also worked to produce methane simultaneously. Anaerobic batch cultivations of ACCeM was carried out in a 20-mL medium containing 0.5% (*w*/*v*) of removed peel, which was an actual agricultural waste, at 37 ◦C without shaking. The armed *C. cellulovorans* was pre-cultivated in media containing 0.5% (*w*/*v*) of MN301 for 2 days. According to measured cell growth of precultures, inoculation amounts for monocultures of the armed *C. cellulovorans* and MFMP were decided so that initial RLU values of each monoculture closely became 1000. RLU values of the preculture of the armed *C. cellulovorans* and MFMP were 3448 and 38,538, respectively. Inoculation volumes of the armed *C. cellulovorans* and MFMP were decided at 8.5 mL and 0.8 mL, respectively. The armed *C. cellulovorans* was 3.44 times dilution and MFMP was 36.6 times dilution so that the initial RLU values of ACCeM were 1002 for the armed *C. cellulovorans* and 1052 for MFMP. The degradation of removed peel was observed for 5 days of cultivation (Figure 5a). Gas production started immediately after inoculation and continued for 4 days of cultivation. However, methane was not detected for these early 4 days of cultivation (Figure 5b). Methane began to be detected after 12 days of cultivation. Interestingly, the increment of methane was 4.7 mL for 12–26 days of cultivation against that the total gas was 7.1 mL for the same term, therefore methane occupied 66.2% of the increased gas. This methane concentration is good performance as the fuel for a biogas power generation. Furthermore, methane productivity was 0.014 mL/h for 12–26 days of cultivation, which was equal to that of MN301 substrate. Acetic acid concentration in the culture supernatant increased the same amount as the MN301 substrate. However, the acetic acid concentration was 2.5 times higher than that in the culture supernatant with MN301 (Figure 5c). It suggested that there was room to convert much acetic acid to methane to improve methane production.

**Figure 5.** (**a**) CCeM culture (left), negative control (right); (**b**) time course of gas production, where total gas volume (-\_closed circle) and methane volume (Δ\_open triangle) are included; (**c**) time course of organic acids concentration of CCeM, where lactic acid (closed bars), formic acid (open bars), acetic acid (closed bars), and propionic acid (dotted bars) are included. Values with error bars are mean ± SE of three independent samples. SE means a standard error. An asterisk indicates a significant difference (*p* < 0.05).

#### **4. Discussion**

Since reducing carbon dioxide to conserve the global environment is one of the purposes to replace fossil fuels with biofuels, it is desirable to adopt the biofuel process that has a low environmental load. A sulfuric acid degradation of cellulosic biomass brings about corrosion of a reactor and needs a waste liquid treatment, and the steam explosion process requires a lot of energy to create high pressure and temperature [25,26], therefore these are not low environmental load methods. The enzymatic saccharification is environmental-friendly, because it is a mesophilic process and does not use acids. However, the degradation cost is high because the purified enzymes are expensive and used in large quantities. Instead of using expensive enzymes, the cost will be reduced to utilize microorganisms that can produce enzymes. Furthermore, since *C. cellulovorans* has many genomes related to cellulosome and non-cellulosomal [27], there is potential to improve the efficiency of the enzymatic saccharification. In addition, when the sulfuric acid process or the steam explosion process is used to saccharide cellulosic biomass, various microorganisms attached to the biomass and brought into the saccharification process are not a problem, because they are treated by high temperature and heat or acid. However, there is the key issue that the microorganisms which produce enzymes must survive and perform its saccharification ability coexisting other various microorganisms, because a lot of microorganisms are brought into the culture apparatus together with cellulosic biomass. In this respect, the present study provided a method that enables the co-culture of various microbiota and *C. cellulovorans*, which has been difficult until now. Although there are few reports only co-culture *C. cellulovorans* and one other microbe [28], many experiences can be performed from cellulose to methane using a consortium and the armed *C. cellulovorans*. Notably, this armed *C. cellulovorans* is non-GMO (non- Genetically Modified Organisms) and is not modified with genome editing, such as CRISPR/Cas9 [29], therefore anyone can easily cultivate the armed *C. cellulovorans* from a cellulosic substrate medium and use it anywhere. It is useful and powerful

that the armed *C. cellulovorans*, which anyone can use, degrades orange wastes including d-limonene and the production of methane, which accounts for 66% of the total gas volume, was produced from the consortia.

However, for stable biogas production, it is essential that the carbon, nitrogen, phosphorus, sulfur, and a carbon-to-nitrogen ratio (C/N ratio) in the culture medium have the desired values. In order for sustainable biogas production from cellulosic biomass, it is necessary to investigate in detail how these components are indispensable for microbial growth change. On the other hand, since the sulfur content contained in the cellulosic biomass is small, simplification of a desulphurization equipment in the biogas plant can be expected.

Moreover, an orange contains beta-cryptoxanthin (β-cry), especially in the orange peel [30]. β-cry is a natural carotenoid pigment, and carotenoids are known to contribute to the defense system of the human body against reactive oxygen species. Inverse associations of serum β-cry with the risk for cancer, diabetes, and liver dysfunction have been reported [31–34]. The enzymatic degradation of orange peel can extract carotenoids contained in the culture. This means orange peel, which was disposed of so far, can be converted to useful raw material to have functional foods. In addition, the enzymatic saccharification does not decompose the compound by high temperature and pressure and by acid, therefore the enzymatic saccharification is the most advanced process to extract the useful compounds.

#### **5. Conclusions**

This study revealed that *C. cellulovorans* by growing with cellulose instead of cellobiose can coexist with complex microbiota such as methanogenic microbiota. Therefore, it provided a useful method for researching the co-culture of *C. cellulovorans* with various other microorganisms. Furthermore, *C. cellulovorans* and methanogenic microflora coexisted with each other while keeping the cellulose degradation ability and the methane production ability. As a result, the biogas was produced containing 66.2% of methane using mandarin orange peel as a carbon source, and when the biogas contains 60% or more of methane, the biogas can be used directly as fuel for a gas engine.

**Author Contributions:** Conceptualization, Y.T. and H.T.; methodology, Y.T. and H.T.; validation, Y.T. and H.T.; resources, Y.T.; data curation, H.T.; writing—original draft preparation, H.T.; writing—review and editing, Y.T.; supervision, Y.T.; project administration, H.T.; funding acquisition, Y.T.

**Funding:** This research received no external funding.

**Acknowledgments:** We would like to thank Daimasa Engineering Co. LTD for sample preparation and assignment. **Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 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/).

### *Article* **Developing a Microbial Consortium for Enhanced Metabolite Production from Simulated Food Waste**

**Nathan D. Schwalm III 1,\*, Wais Mojadedi 2, Elliot S. Gerlach 1, Marcus Benyamin 2, Matthew A. Perisin <sup>1</sup> and Katherine L. Akingbade 1,\***


Received: 25 September 2019; Accepted: 22 November 2019; Published: 27 November 2019

**Abstract:** Food waste disposal and transportation of commodity chemicals to the point-of-need are substantial challenges in military environments. Here, we propose addressing these challenges via the design of a microbial consortium for the fermentation of food waste to hydrogen. First, we simulated the exchange metabolic fluxes of monocultures and pairwise co-cultures using genome-scale metabolic models on a food waste proxy. We identified that one of the top hydrogen producing co-cultures comprised *Clostridium beijerinckii* NCIMB 8052 and *Yokenella regensburgei* ATCC 43003. A consortium of these two strains produced a similar amount of hydrogen gas and increased butyrate compared to the *C. beijerinckii* monoculture, when grown on an artificial garbage slurry. Increased butyrate production in the consortium can be attributed to cross-feeding of lactate produced by *Y. regensburgei*. Moreover, exogenous lactate promotes the growth of *C. beijerinckii* with or without a limited amount of glucose. Increasing the scale of the consortium fermentation proved challenging, as two distinct attempts to scale-up the enhanced butyrate production resulted in different metabolic profiles than observed in smaller scale fermentations. Though the genome-scale metabolic model simulations provided a useful starting point for the design of microbial consortia to generate value-added products from waste materials, further model refinements based on experimental results are required for more robust predictions.

**Keywords:** butyrate; *Clostridium beijerinckii*; cross-feeding; food waste; genome-scale metabolic model; hydrogen; lactate; *Yokenella regensburgei*

#### **1. Introduction**

Generation of food waste is a global problem [1,2]. Approximately 33% of food produced is wasted, contributing to 30–60% of the total solid waste produced worldwide [3]. The burden of food waste disposal extends to military environments, where it imposes an additional logistical burden on Army operations [4]. The primary waste produced by the soldier is food related trash [4], and current practices dictate that field generated waste must either be buried or burned daily [5]. An alternative to these practices is the anaerobic fermentation of food waste to commodity chemicals [1]. Anaerobic fermentation of a single carbon feedstock is frequently used in industrial-scale processes to produce biofuels [6] and commodity chemicals [7]. Significant challenges exist applying anaerobic fermentations to food waste breakdown, due to the complexity and inconsistency of the fermentation feedstock. Utilizing additional microbes that can support the primary fermentation strain as part of a microbial

consortium is a potentially promising avenue to increasing the viability of food waste to value-added chemical conversion.

Considerable research efforts are underway to harness both microbial consortia comprising multiple genotypes of the same species [8,9] and multi-species consortia [10,11] for improved production of chemicals from readily available complex substrates [12,13]. Microbial consortia enable separation of complex substrate breakdown and chemical production between different species, which can reduce the metabolic burden on individual cells and reduce the need to engineer these microbes [14,15]. Cross-feeding of metabolites from extracellular substrate breakdown [16], cellular secretion [17], or via direct cell-to-cell interactions [18,19] can support the stability of a microbial consortium. These interactions can be designed to force the dependence of auxotrophic strains on each other for growth [20] or can be a result of native metabolic byproducts from one organism acting as a substrate for another organism [21].

Hydrogen is a promising alternative fuel source to petrochemicals due to its high energy content (120 MJ/kg versus 46.7 MJ/kg for gasoline) [22]. Microbes primarily produce hydrogen via photofermentation by the purple non-sulfur bacteria *Rhodobacter* and *Rhodopseudomonas* [23], and during dark fermentation by strictly anaerobic *Clostridium* species [22]. In purple non-sulfur bacteria, hydrogen production is coupled with organic acid utilization [23]. In *Clostridium* species, hydrogen production is coupled with ferredoxin oxidation and is accompanied by organic acid production [24], which inhibits hydrogenase activity and hydrogen production [25]. Microbial consortia of *Clostridium* that generate organic acids and purple non-sulfur bacteria that use organic acids for hydrogen production are being explored [26,27], as are consortia comprising multiple *Clostridium* species [13,28]. The formate-hydrogen lyase complex, found mainly in *Proteobacteria*, produces hydrogen via the anaerobic breakdown of formate [29]. This complex is also being explored as a route for hydrogen production in *Escherichia coli* engineered to express a hydrogenase [29], and as part of microbial consortia with *Clostridium* [30].

Genome scale metabolic models (GSMMs), combined with flux-balance analysis (FBA), provide predictive frameworks not only to assess the metabolic outputs from a variable inputs, but also to assess metabolic interactions between multiple microbes. These models are built from enzymatic genome annotations that determine available reactions for each microbe, resulting in a table of reactants/products by reaction [31]. These tables are then fed to FBA to determine what metabolic outputs are possible given a set of inputs, assuming a steady-state [32]. To further limit the solution space, there is an assumption that each microbe will optimize growth (i.e., biomass production) [32]. This framework has been demonstrated to accurately recapitulate single microbe metabolisms and metabolic interactions between members of a consortium [33–37].

Recently, Magnusdottir et al. created GSMMs of approximately equal refinement for 773 gut microbes and simulated all monocultures and co-cultures on Western and high-fiber diet inputs [38]. Ecological interactions between these microbes were predicted to vary based on the diet and whether the environment was aerobic or anaerobic [38]. Perisin and Sund followed up on these findings to predict whether combinations of microbes could produce specialty/commodity chemicals at higher rates than monocultures [39]. Co-cultures were identified that could overproduce chemicals of interest such as hydrogen gas, butanol, methane, formaldehyde, propionate, and urea [39].

We developed a framework to experimentally test commodity chemical production from microbial consortia predicted to synergistically overproduce an individual chemical (Figure 1). We used genome-scale metabolic models and flux balance analysis of human gut microbiota strains provided a simulated food waste to identify two species bacterial consortia with the greatest synergistic production of hydrogen gas. We then measured the gases and metabolites produced by one of the highest overproducing consortia and its component species when cultured on an artificial garbage slurry medium (AGS). Moreover, we demonstrated the cross-feeding of metabolites between the two species. Finally, we attempted to scale-up the AGS fermentation with two distinct conditions, stationary fermentation or stirred fermentation with pH control, to assess the feasibility of the

method. Experimental consortium fermentations will further improve metabolic models for future experimental testing.

**Figure 1.** Experimental framework. Genome-scale metabolic modeling (GSMM) and flux balance analysis (FBA) were used to predict the metabolic outputs of 298,378 pairwise combinations of microbial consortia provided a simulated food waste medium. The consortium of *Clostridium beijerinckii* and *Yokenella regensburgei* was predicted to have the second highest synergistic overproduction of hydrogen compared to any individual strain. Small-scale fermentations were performed using the *C. beijerinckii* and *Y. regensburgei* consortium on an artificial garbage slurry medium to simulate food waste. Growth (CFUs), metabolites, and percentage of gases produced in gas chromatography vials were measured. Spent media from cultures of *C. beijerinckii* and *Y. regensburgei* were used to cross-feed to the other species, and growth and metabolite production were measured. Small-scale experiments were used to inform design of scaled-up fermentations of monocultures and the consortium of *C. beijerinckii* and *Y. regensburgei*. The results of these experiments can be used to further improve the GSMM and FBA forming a broad framework for testing production of commodity chemicals by microbial consortia.

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

#### *2.1. Genome-Scale Metabolic Modeling*

Simulations were performed as in Perisin and Sund [39]. Genome-scale metabolic models (GSMMs) for *Clostridium beijerinckii* NCIMB 8052 and *Yokenella regensburgei* ATCC 43003 (AGORA v 1.01) were downloaded from the Virtual Metabolic Human database (http://vmh.uni.lu) [38]. Lower bounds for input exchange reactions were set to mimic Western diet food waste (Supplementary Table 1 in [39]). All simulations were performed in R v3.4.0 [40] with plots created using *ggplot2* v 2.2.1 [41]. Systems Biology Markup Language (SBML) models were first uploaded with the *sybilSBML* v 3.0.1 [42] package. Then, *sybil* v2.0.4 [43], and *glpkAPI* v 1.3.0 [44] packages were used for model manipulations and

flux-balance analysis (FBA). For monoculture simulations, flux through each model's biomass reaction was maximized. The maximum biomass flux was then used to run parsimonious enzyme usage FBA (pFBA, *mtf* algorithm in *sybil*) so that the total absolute flux was minimized. For co-culture simulations, models were combined in a similar manner to Magnusdottir et al. [38]. The COBRA [45] MATLAB script, createMultipleSpeciesModel.m found at: https://github.com/opencobra/cobratoolbox, which is based on the FBA implementation in Klitgord and Segre [46], was used as a template. This script was converted to work in R with the *sybil* package and creates a common environment for metabolites to be exchanged between models, but does not create a host compartment as in the COBRA implementation. After combining the models, input exchange fluxes were updated based on Western diet food waste as above, and pFBA was used to simulate growth, simultaneously maximizing each model's growth and minimizing the total absolute flux.

#### *2.2. Bacterial Strains and Growth Conditions*

*Clostridium beijerinckii* NCIMB 8052 (ATCC 51743) and *Yokenella regensburgei* ATCC 43,003 were obtained from ATCC. Both strains were cultivated in anaerobic conditions (5% carbon dioxide, 5% hydrogen, 90% nitrogen) at 37 ◦C in a Coy anaerobic chamber and maintained as stocks at −80 ◦C in media containing 20% glycerol. *C. beijerinckii* was routinely cultured on Clostridial Growth Medium (CGM) [47] or Luria Bertani (LB) medium (Fisher, Fair Lawn, NJ, USA), supplemented with 0.5% glucose, unless otherwise noted. Thiamphenicol (30 μg mL−1) was used to select for *C. beijerinckii* when necessary [48,49]. *Y. regensburgei* was routinely cultured on LB supplemented with 0.5% glucose or Brain Heart Infusion (BHI) medium (Oxoid, Lenexa, KS, USA). Erythromycin (40 μg mL−1) and aerobic growth at 37 ◦C were used to select for *Y. regensburgei*, when necessary [49]. The artificial garbage slurry was generated to model organic solid waste, as previously described [50], with limited modifications. Briefly, 10% (w v<sup>−</sup>1) dog food (Kibble 'n Bits® Chef Choice Bistro Oven Roasted Beef, Spring Vegetable & Apple Flavor) was blended in distilled water and autoclaved (121 ◦C, 30 min exposure). Following the first autoclave cycle, the pH was balanced to 7.0 via addition of sodium hydroxide solution and buffered to 50 mM potassium phosphate pH 6.7, prior to a second autoclave cycle. For cross-feeding experiments, *C. beijerinckii* or *Y. regensburgei* were grown in LB supplemented with 0.5% glucose for 24 h and filtered through a 0.2 μm PES membrane (Corning, Corning, NY, USA) to produce spent media. Spent media with the addition of 50 mM potassium phosphate pH 6.8 was used for an additional 24 h of culturing of either *C. beijerinckii* or *Y. regensburgei*. For experiments assessing the ability of *C. beijerinckii* to utilize exogenous lactate, the indicated amount of DL-lactic acid (Fluka, St. Louis, MO, USA) was added to cultures and neutralized with sodium hydroxide. Bacterial growth was assessed by measuring absorbance at 600 nm, for cross-feeding and lactate supplementation experiments, or plating of multiple dilutions to count colony forming units per milliliter (CFU mL<sup>−</sup>1) on selective media, for experiments using AGS.

#### *2.3. Bioreactor Fermentations*

Bioreactor fermentations were conducted in a DASGIP® parallel fermentation system (Eppendorf, Jülich, Germany). AGS was brought to pH 6.7 by addition of sodium hydroxide solution, following the addition of the 50 mM potassium phosphate and aliquoted to reaction vessels prior to autoclaving (121 ◦C, 30 min exposure) a single time. For the pH-controlled bioreactor experiment, the DASGIP® system automatically added sodium hydroxide to maintain a pH of at least 6.7. Temperature was maintained at 37 ◦C and anaerobic headspace maintained via nitrogen addition at a flow rate of 1 L min<sup>−</sup>1.

#### *2.4. HPLC Quantification of Metabolites*

Fermentation products were separated and quantified using an Aminex HPX-87H organic acid column (300 × 7.8 mm; Bio-Rad, Hercules, CA, USA) on an Agilent 1200 series HPLC (Agilent Technologies, Santa Clara, CA, USA) equipped with a multi-wavelength UV/vis detector (210 nm and 280 nm) and refractive index detector. Samples were filtered through a 0.2 μm PES membrane (Corning) and stored at 4 ◦C until applied to the column with an injection volume of 20 μL. Separation was performed using an isocratic mobile phase consisting of 3.25 mM sulfuric acid at a flow rate of 0.600 mL min−<sup>1</sup> for a total run time of 55 min, and was temperature controlled for the entire duration at 35 ◦C. Quantification of acetate, acetone, butyrate, ethanol, formate, glucose, and lactate was performed through the use of standardized concentration gradients with HPLC-grade standards obtained from various vendors (not shown). Data was processed using ChemStation (Agilent).

#### *2.5. Gas Chromatography*

Hydrogen and carbon dioxide gas production were measured using a gas chromatograph (GC; Infinicon 3000 microGC®) equipped with a thermal conductivity detector. A fixed molecular sieve 5A capillary column using argon as the carrier gas was used for hydrogen and nitrogen quantification, and a Plot U column with helium as the carrier gas was used for carbon dioxide quantification. GC data was processed using ezIQ and Diablo EZReporter software (Diablo Analytical, Antioch, CA, USA). For experiments performed in the anaerobic chamber, immediately after inoculation, 1.5 mL of each culture was transferred to a GC vial that was then sealed. For bioreactor sampling, the same parameters were utilized with samples automatically taken from the headspace every 11 min via an in-line setup. Gas production rates were derived from the GC gas percentage output using the ideal gas law, headspace flowrate (1 L min<sup>−</sup>1), and the fermentation volume (1 L). The trapezoidal rule was used to approximate total gas production from the GC production rates using numerical integration.

#### **3. Results**

#### *3.1. Co-culture Simulations Predict Higher Hydrogen Production than Both Monocultures*

A consortium of *C. beijerinckii* NCIMB 8052 and *Y. regensburgei* ATCC 43,003 was originally identified as having the second highest predicted hydrogen production among 298,378 simulated co-cultures (Table 1) [39]. These co-culture simulations represented every pairwise combination of 773 human gut microbe GSMMs [38]. Upon further examination of output exchange fluxes for the co-culture of *C. beijerinckii* and *Y. regensburgei* and their monocultures, *C. beijerinckii* was predicted to be the sole hydrogen producer (Figure 2). Compared to the monoculture, the addition of *Y. regensburgei* boosted *C. beijerinckii* biomass (Figure 2) and hydrogen production (Figure 2). The increased hydrogen production was not due only to increased biomass, because when the hydrogen flux was normalized to growth rate, the co-culture hydrogen production rate (86 millimoles per gram dry cell weight per hour) was greater than the monoculture hydrogen production (69 millimoles per gram dry cell weight per hour). The increase in co-culture output fluxes were predicted to be due to cross-feeding of lactate from *Y. regensburgei* to *C. beijerinckii* (Figure 2). The overall ecological interaction was predicted to be commensal as the growth rate of *C. beijerinckii* increased from the monoculture to the co-culture, while the growth rate *Y. regensburgei* was unchanged(Figure 2). Additional microbial consortia containing *C. beijerinckii* that were predicted to have more than two-fold greater hydrogen production than *C. beijerinckii* alone are listed in Table 1.



**Figure 2.** Predicted metabolite flux for a consortium of *C. beijerinckii* and *Y. regensburgei*. Flux balance analysis (FBA) simulations of *Clostridium beijerinckii* and *Yokenella regensburgei* monocultures (Alone), and co-cultures of *C. beijerinckii* and *Y. regensburgei* (Consortium) to predict production of (**a**) biomass, (**b**) hydrogen, (**c**) carbon dioxide, (**d**) acetate, (**e**) L-lactate, (**f**) D-lactate, (**g**) formate, (**h**) ethanol. *C. beijerinckii* Consortium indicates the contribution of *C. beijerinckii* to the metabolites in the co-culture FBA and *Y. regensburgei* Consortium indicates the contribution of *Y. regensburgei* to the metabolites in the co-culture FBA. Units of simulated flux are millimoles of metabolite produced per gram of dry cell weight per hour.

#### *3.2. Metabolite Production by a Consortium of C. beijerinckii and Y. regensburgei*

The consortium of *C. beijerinckii* NCIMB 8052 and *Y. regensburgei* ATCC 43,003 (Table 1) was predicted to exhibit the second largest increase in hydrogen production over any individual modeled bacterial strain when provided with a food waste simulant. To test for the potential of this interaction in vivo, cells were grown on an artificial garbage slurry medium (AGS) in an anaerobic chamber. After 24 h growth, *C. beijerinckii* and *Y. regensburgei* were plated on selective media to determine the number of colony forming units per milliliter of culture (CFU mL<sup>−</sup>1) for each strain. We observed an increase in CFU mL−<sup>1</sup> over the inoculum for both strains, indicating that AGS supports the growth of these bacteria independently (Figure 3a). Both strains also grew within 24 h in a co-culture, as measured by plating for CFUs on selective media (Figure 3a). The CFU count of each strain in the consortia was significantly lower than for each individual strain (Figure 3a). However, the total CFU count of both strains in the consortia was higher than for the cultures of *C. beijerinckii* alone and similar to the CFU count in the *Y. regensburgei* monocultures (Figure 3a). Viable cells were unable to be consistently recovered by plating for CFUs at subsequent times (data not shown), likely due to depletion of an essential growth factor or accumulation of a toxic byproduct in the media.

We measured gas production by *C. beijerinckii* and *Y. regensburgei*, either alone or as part of a consortium, as an increase in atmospheric percentage over the background atmosphere by gas chromatography (GC), after three days of growth on the artificial garbage slurry medium in sealed vials. *C. beijerinckii* produced approximately 8% hydrogen above the atmospheric background, whereas *Y. regensburgei* produced only about 1% hydrogen above the atmospheric background, when each was cultured individually (Figure 3b). The consortium produced a statistically similar percentage of hydrogen above the background (*p* = 0.89, by two-tailed Student's *t* test) to the *C. beijerinckii* monoculture (Figure 3b). Each individual strain and the consortium of the two strains produced an approximately equivalent proportional percentage of carbon dioxide to hydrogen (Figure 3b).

We quantified the organic acids produced by *C. beijerinckii* and *Y. regensburgei* individually and as a consortium from filtered extracts via HPLC. Both *C. beijerinckii* monocultures and the consortium primarily produced butyrate, consistent with previous reports [28], with the consortium producing a small but significantly larger amount than the *C. beijerinckii* monoculture (Figure 3c). As predicted by the FBA (Figure 2), the *Y. regensburgei* monoculture produced almost no butyrate (Figure 3c), but produced approximately 30 mM lactate, which was not observed in either the consortium or the *C. beijerinckii* monoculture (Figure 3d). The consortium produced substantially lower lactate than would be expected, if the *Y. regensburgei* present produced a proportional amount of lactate to that in the *Y. regensburgei* monoculture (Figure 3d). Acetate was a secondary fermentation product in all cultures, accumulating to approximately 11 mM in both the *C. beijerinckii* monoculture and the consortium, and approximately 4 mM in the *Y. regensburgei* monoculture (Figure 3e). A significant amount of formate was consumed in both monocultures, while the consortium produced a small amount of formate (Figure 3f). The consumption of formate by *Y. regensburgei* could be an indication of formate-hydrogen lyase activity [29], although low amounts of hydrogen were observed for *Y. regensburgei* monocultures (Figure 3b).

**Figure 3.** Metabolite production by a consortium of *C. beijerinckii* and *Y. regensburgei*. *Clostridium beijerinckii* and *Yokenella regensburgei* were grown as monocultures or as a consortium on artificial garbage slurry medium (AGS) for 144 h in an anaerobic chamber, either in gas chromatography vials (b) or sealed tubes (**a**,**c**,**d**,**e**). (**a**) Colony forming units after 24 h. (**b**) Normalized percent hydrogen and carbon dioxide measured by gas chromatography after 72 h. (**c**) Butyrate, (**d**) lactate, (**e**) acetate, and (**f**) formate production after 72 h and 144 h. Units for metabolites are mM as normalized to HPLC standards. The mean and standard error of the mean of four independent replicates are graphed. Asterisks indicate significant difference between the indicated samples (n.s. (not significant), *p* > 0.05; \*, *p* ≤ 0.05; \*\*, *p* ≤ 0.01; \*\*\*, *p* ≤ 0.001 by two-tailed Student's *t* test).

#### *3.3. Y. regensburgei Cross-feeds Lactate to C. beijerinckii for Butyrate Production*

We hypothesized that the lactate produced by *Y. regensburgei* was utilized as a carbon source by *C. beijerinckii*, which subsequently produced additional butyrate, because a higher concentration of butyrate was observed in the consortium than in the *C. beijerinckii* monoculture (Figure 3c). To examine this possibility in a simpler system, we grew *C. beijerinckii* or *Y. regensburgei* on sterile filtered spent media from 24 h monocultures of *Y. regensburgei* or *C. beijerinckii* that were grown on Luria Bertani (LB) medium supplemented with 0.5% glucose. No growth was initially observed of either *C. beijerinckii* or *Y. regensburgei* on the sterile filtered spent media (data not shown); however, addition of sterile 50 mM potassium phosphate pH 6.8 to filtered spent media enabled growth of both strains (Figure 4a), suggesting that decreased media pH was inhibitory to growth. Growth of both *C. beijerinckii* and *Y. regensburgei* was significantly greater on spent media derived from the opposing species (*i.e., C. beijerinckii* growth on *Y. regensburgei* spent media and *Y. regensburgei* growth on *C. beijerinckii* spent media) than observed for growth on spent media derived from the same species (i.e., *C. beijerinckii* growth on *C. beijerinckii* spent media and *Y. regensburgei* growth on *Y. regensburgei* spent media) (Figure 4a). This is consistent with the idea of cross-feeding between the two species.

**Figure 4.** Spent media cross-feeding by *C. beijerinckii* and *Y. regensburgei*. *Clostridium beijerinckii* and *Yokenella regensburgei* were grown as monocultures for 24 h on LB + 0.5% glucose in an anaerobic chamber. Spent media was sterile filtered and 50 mM potassium phosphate pH 6.8 was added. Spent media from each species was inoculated 1:100 with overnight cultures of either *C. beijerinckii* or *Y. regensburgei* and after 24 h, (**a**) OD600, (**b**) butyrate, (**c**) lactate, and (**d**) an unknown metabolite that elutes at the expected elution time for acetone, were measured. Units for metabolites are mM as normalized to HPLC standards. The mean and standard error of the mean of six independent replicates are graphed. Asterisks indicate significant difference between the indicated samples (\*, *p* ≤ 0.05; \*\*, *p* ≤ 0.01; \*\*\*, *p* ≤ 0.001 by two-tailed Student's *t* test).

*C. beijerinckii* significantly increased the concentration of butyrate in spent media from both *C. beijerinckii* and *Y. regensburgei*, similar to growth on AGS; however, this increase was substantially larger in the *Y. regensburgei* spent media than the *C. beijerinckii* spent media (~700% versus ~12% increase) (Figure 4b). *C. beijerinckii* significantly depletes the lactate from *Y. regensburgei* spent media, concurrent with the large increase in butyrate (Figure 4c), suggesting that lactate produced by *Y. regensburgei* supports additional growth and butyrate production by *C. beijerinckii. Y. regensburgei* produces a significant amount of additional lactate in both *C. beijerinckii* and *Y. regensburgei* spent media (Figure 4c), suggesting that *C. beijerinckii* may produce nutrient(s) capable of being further metabolized by *Y. regensburgei*, and that pH is likely a limiting factor for its growth in LB supplemented with glucose.

The HPLC method used for metabolite detection was previously optimized for detection of short-chain fatty acids, acetone, butanol, and ethanol from *Clostridium acetobutylicum* [51]. A peak eluting at the time of the expected acetone peak was detected in the *Y. regensburgei* spent media (Figure 4d). *Y. regensburgei* has not previously been reported to produce acetone, and it does not encode the genes necessary to produce it [52]. The *C. beijerinckii* spent media also contains a metabolite with the same HPLC elution time; however, its peak area is only ~25–30% of the peak observed in the *Y. regensburgei* spent media. *C. beijerinckii* significantly depletes the undetermined metabolite when grown on the *Y. regensburgei* spent media (Figure 4d), while *Y. regensburgei* significantly increased the concentration of the metabolite when grown on both the *C. beijerinckii* and *Y. regensburgei* spent media (Figure 4d). Moreover, *Y. regensburgei* accumulates ~1.7-fold more of the metabolite when grown anaerobically than aerobically, despite aerobic cultures growing to more than two-fold higher optical density at 600 nm (data not shown). Electrospray ionization (ESI) mass spectrometry was performed on an HPLC fraction collected from the *Y. regensburgei* spent media corresponding to the acetone peak. The major mass to charge ratio (m/z) peaks that were not identified as solvent or carrier peaks were at m/z = 171.11 in the ESI+ spectrum and m/z = 127.99 and 111.00 in the ESI-spectrum (data not shown). These peaks do not map to a known metabolite that could plausibly be produced by *Y. regensburgei* given the current understanding of its metabolism. Further experimentation is required to determine the role of this molecule in potential cross-feeding between *Y. regensburgei* and *C. beijerinckii*.

#### *3.4. C. beijerinckii Uses Lactate as a Carbon Source*

Since *C. beijerinckii* is capable of depleting lactate from *Y. regensburgei* spent media (Figure 4c), and the GSMM predicted cross-feeding of lactate from *Y. regensburgei* to *C. beijerinckii* (Figure 2), we hypothesized that lactate could support the growth of *C. beijerinckii* either with or without a limited amount of glucose. The optical density of *C. beijerinckii* cultures indeed increased compared to an un-inoculated control in LB media containing 30 mM lactic acid and 20 mM potassium phosphate, when the initial pH was controlled to 6.7 with sodium hydroxide (Figure 5a). Concomitant with the increase in optical density, a significant amount of lactate was depleted from the media, suggesting its use as a carbon source by *C. beijerinckii* (Figure 5b). Moreover, *C. beijerinckii* grew (Figure 5a) and significantly depleted lactate from the media (Figure 5b) in cultures with 5 mM glucose and 15 mM lactate within 24 h, although the amount of growth was significantly less than observed in cultures containing 10 mM glucose during the same duration (Figure 5a). In both cases, glucose was completely depleted from the media (data not shown). *C. beijerinckii* produced primarily butyrate in each of the three conditions, with the amount produced positively correlating with the amount of glucose present (Figure 5c). *C. beijerinckii* produced substantially more ethanol in the absence of lactate, exhibiting low ethanol production in both the lactate alone and mixed carbon source cultures (Figure 5d).

**Figure 5.** Lactate can support the growth of *C. beijerinckii* with or without glucose. *C. beijerinckii* was grown on LB + 20 mM potassium phosphate pH 6.8 + either 10 mM glucose, or 30 mM lactate, or 5 mM glucose and 15 mM lactate, in an anaerobic chamber. (**a**) OD600, (**b**) lactate, (**c**) butyrate, and (**d**) ethanol, were measured after 24 h from the *C. beijerinckii* cultures and un-inoculated controls (Media). Units for metabolites are mM as normalized to HPLC standards. The mean and standard error of the mean of four independent replicates are graphed. Asterisks indicate significant difference between the indicated samples (n.s., \*, *p* > 0.05; \*\*, *p* ≤ 0.01; \*\*\*, *p* ≤ 0.001 by two-tailed Student's *t* test).

#### *3.5. Enhanced Metabolite Production by the Consortium Is Condition-Dependent*

To quantify the production rate and amount of gases produced by *C. beijerinckii* and *Y. regensburgei* either alone or as a consortium, we used a DASGIP® parallel bioreactor system coupled with two gas chromatographs to perform near constant measurements of gases (every 11 min) from two independent replicates grown in one liter volumes on AGS. The experiment was performed with two distinct sets of bioreactor conditions. First, to closely mimic the closed tubes in the anaerobic chamber, the bioreactor mixtures were left stationary until fifteen seconds prior to sample collection and pH was not controlled. Second, to maintain the optimal pH for hydrogen production by the *C. beijerinckii* hydrogenase [25], the pH of bioreactor vessels was automatically controlled at pH 6.7, with continuous stirring at 100 rpm for the entirety of the experiment.

The two bioreactor experiments resulted in distinct patterns of growth and metabolite production for both the consortium and monocultures of *C. beijerinckii* and *Y. regensburgei*. In the stationary bioreactor experiments without pH control, *C. beijerinckii* grew to a higher density in monoculture than the consortium starting eight hours after inoculation and continuing for the duration of the experiment, until it was undetectable in both the monoculture and consortia by the 48 h time point (Figure 6a). By contrast, no difference was observed in the amount of colony forming units for *Y. regensburgei* between the monoculture and the consortium until the 36 h time point when there was a 2-log decrease of *Y. regensburgei* in the consortium. *Y. regensburgei* levels further decreased to undetectable levels by the 48 h time point (Figure 6a). Although a lower initial amount of *C. beijerinckii* was present in the consortium, both *C. beijerinckii* and *Y. regensburgei* were present in similar amounts from 12–24 h (Figure 6a). With pH control, *Y. regensburgei* maintained a high level of culturable CFUs in both the

monoculture and the consortium, even at 72 h (Figure 7a). With pH control, *C. beijerinckii* in both monoculture and the consortium reached an apparent peak density at 12 h, before decreasing to nearly undetectable levels by 72 h (Figure 7a).

*C. beijerinckii* monocultures and the consortium produced similar amounts of hydrogen (Figure 6b) and carbon dioxide (Figure 6c), without pH control, though variability between samples was observed. With pH control, *C. beijerinckii* monocultures produced greater amounts of hydrogen (Figure 7b) and carbon dioxide (Figure 7c) than the consortium. The total amounts of hydrogen and carbon dioxide produced were higher for fermentations with pH control than without, although differences in the kinetics between the individual bioreactors of both *C. beijerinckii* monocultures and the consortium with pH control were observed. *Y. regensburgei* produced small amounts of hydrogen (Figures 6b and 7b) and carbon dioxide (Figures 6c and 7c) throughout the experiments with and without pH control.

**Figure 6.** Metabolites produced by the *C. beijerinckii* and *Y. regensburgei* consortium in stationary bioreactors. *Clostridium beijerinckii* and *Yokenella regensburgei* were grown as monocultures or as a consortium on AGS for 48 h in 1 L bioreactor vessels. Vessel contents were only stirred at 100 rpm for 15 s prior to collection of each sample. (**a**) Colony forming units, (**b**) hydrogen, (**c**) carbon dioxide, (**d**) butyrate, (**e**) lactate, (**f**) acetate quantified at the indicated time points. Each replicate is graphed independently. Units for gas production are moles of gas produced per liter of fermentation volume (mol L<sup>−</sup>1), derived from the GC gas percentage output using the ideal gas law, headspace flowrate, and the fermentation volume. Units for metabolites are mM as normalized to HPLC standards.

**Figure 7.** Metabolites produced by the *C. beijerinckii* and *Y. regensburgei* consortium in bioreactors with stirring and pH control. *Clostridium beijerinckii* and *Yokenella regensburgei* were grown as monocultures or as a consortium on AGS for 144 h in 1 L bioreactor vessels, with the pH maintained at 6.7 via the addition of sodium hydroxide and constant stirring at 100 rpm. (**a**) Colony forming units, (**b**) hydrogen, (**c**) carbon dioxide, (**d**) butyrate, (**e**) lactate, (**f**) acetate quantified at the indicated time points. Each replicate is graphed independently. Units for gas production are moles of gas produced per liter of fermentation volume (mol L<sup>−</sup>1), derived from the GC gas percentage output using the ideal gas law, headspace flowrate, and the fermentation volume. Units for metabolites are mM as normalized to HPLC standards.

*C. beijerinckii* monocultures produced primarily butyrate to a higher level than the consortium both with pH control (Figure 7d) and without pH control (Figure 6d). *Y. regensburgei* monocultures produced primarily lactate in both conditions (Figures 6e and 7e). However, the amount of lactate produced by *Y. regensburgei* with pH control was lower than the amount of lactate produced by the consortium (Figure 7e), and the amount of lactate produced by *Y. regensburgei* without pH control was substantially lower than in any other experiment (Figure 6e). *C. beijerinckii*, *Y. regensburgei*, and the consortium produced similar levels of acetate (Figures 6f and 7f) and ethanol (data not shown) as secondary fermentation products in both the experiments with and without pH control.

#### **4. Discussion**

Anaerobic fermentation of food waste is an attractive alternative to other disposal methods, to reduce waste volume and recover lost energy [1–3]. Here, we developed a framework to experimentally test metabolic modeling predictions of microbial consortia for their ability to produce commodity chemicals (Figure 1). We have established that a consortium of *C. beijerinckii* and *Y. regensburgei* is capable of producing increased amounts of commodity chemicals from a simulated food waste medium, compared to monocultures of either species. Genome-scale metabolic modeling predicted that co-cultures of *C. beijerinckii* and *Y. regensburgei* would result in synergistic overproduction of hydrogen as a result of lactate cross-feeding from *Y. regensburgei* to *C. beijerinckii*. We observed similar hydrogen gas production by *C. beijerinckii* and the consortium in stationary conditions (Figures 3b and 6b) and overproduction of the commodity chemicals butyrate (Figure 3c) and lactate (Figure 7e) by the consortium in distinct experimental conditions. Moreover, we demonstrated that *C. beijerinckii* is capable of using exogenous lactate as a carbon source (Figure 5) and that *Y. regensburgei* cross-feeds lactate to *C. beijerinckii* (Figure 4). Differences in the metabolites produced by the consortium of *C. beijerinckii* and *Y. regensburgei* during experiments with the artificial garbage slurry medium (Figures 3, 6 and 7) exhibit the difficulty in fermentation scale-up, but also demonstrate that opportunities exist to modulate fermentation conditions for varied commodity chemical output.

#### *4.1. Experimental Implementation of Genome-Scale Metabolic Modeling Predictions*

The consortium of *C. beijerinckii* and *Y. regensburgei* was predicted by GSMM and FBA to have the second highest overproduction of hydrogen gas compared to any individual species from models of 773 gut microbiota species [38,39], leading to its selection for further experimentation. Similar levels of hydrogen were observed for *C. beijerinckii* and the consortium of *C. beijerinckii* and *Y. regensburgei* in small-scale experiments performed in an anaerobic chamber (Figure 3). This experimental setup prevented quantification of gas from more than a single time point. GC vials from the experiment all had the same headspace volume, but slight variations in growth may have caused different pressures for any given vial at time of headspace analysis. It is possible to measure the pressure in a given vial; however, this was not done, as it is difficult to obtain a pressure measurement without releasing gas from the GC vial, which could potentially affect the GC concentration measurements. Continuous GC monitoring over time was performed from 1 L bioreactor cultures (Figures 6 and 7). For stationary cultures without pH control, which most closely mimicked the small-scale experimental design, the hydrogen production by the consortium was similar to the amount of hydrogen produced by *C. beijerinckii* monocultures (Figure 6b). While the consortium did not produce similar amounts of hydrogen to *C. beijerinckii* monocultures grown with pH control, which exhibited the highest amounts of hydrogen produced from any experimental condition tested (Figure 7b), the FBA modeling did not account for such conditions. Manipulating simulated conditions with FBA modeling could provide additional avenues to improve commodity chemical production.

#### *4.2. Y. regensburgei Cross-feeding to C. beijerinckii*

In small-scale fermentations of the AGS, substantially lower lactate was produced by the consortium than would be expected if the *Y. regensburgei* present produced a proportional amount of lactate to that in the *Y. regensburgei* monoculture (Figure 3d). This suggested that the predicted cross-feeding of lactate from *Y. regensburgei* to *C. beijerinckii* (Figure 2) may be occurring when the consortium was grown in AGS. *Clostridium butyricum* [53] and *Clostridium saccharoperbutylacetonicum* [54] were previously described to metabolize acetate and lactate to butyrate, while several other *Clostridium*, including *C. beijerinckii*, were proposed to breakdown lactate [55]. Here, we show that *C. beijerinckii* can metabolize a limited amount of lactate, with or without a limited amount of glucose (Figure 5). The proposed mechanism for lactate metabolism in *Clostridium* is based on the lactate oxidation pathway in *Acetobacterium woodii*, which couples a flavin adenine dinucleotide (FAD)-dependent lactate

dehydrogenase with an electron flavoprotein complex to convert a reduced ferredoxin, lactate, and two oxidized nicotinamide adenine dinucleotides (NAD) to an oxidized ferredoxin, pyruvate, and two reduced nicotinamide adenine dinucleotides (NADH) [53]. A similar mechanism is likely used by *C. beijerinckii*, as a genomic locus with significant sequence homology to the locus proposed to breakdown lactate in *C. butyricum* [53] is found in *C. beijerinckii* (*Cbei\_2884*-*Cbei\_2888*) using the Basic Local Alignment Search Tool (BLAST) [56]. While FBA predicts that multiple species can cross-feed lactate to *C. beijerinckii*, the ability to cross-feeding of lactate to *Clostridium* may be somewhat limited, as lactic acid bacteria, which produce large amounts of lactate, are often detrimental to efforts to produce hydrogen from complex microbial consortia [55]. Moreover, a pilot experiment attempting to cross-feed lactate from the lactic acid bacterium *Lactobacillus fermentum* to *C. beijerinckii* was unsuccessful, as *L. fermentum* produced a significant amount of lactate (~30 mM), but no change in lactate or butyrate levels was observed when *C. beijerinckii* was cultured in the spent medium (data not shown).

#### *4.3. Varying Growth Conditions to Control Metabolic Output*

While small variability in metabolic output was observed between replicates with small-scale cultures (Figure 3), a substantial amount of variability was observed between all individual replicates containing *C. beijerinckii* for the bioreactor experiments (Figures 6 and 7). Part of this variability may stem from differences in handling of the inoculums for the bioreactors versus the small-scale cultures. *C. beijerinckii* is sensitive to oxygen, and inoculation of the bioreactors required brief exposure to oxygen as the inoculum was transported from an anaerobic chamber to the anaerobic environment of the bioreactors. This exposure coupled with slight differences in the growth phase or culture density of the *C. beijerinckii* inocula could have biased individual bioreactors to be more or less favorable for *C. beijerinckii* growth. The AGS is another potential source of variability, as the precise composition of the main component could have varied from vessel to vessel. The pH control also could have introduced variability between replicates, because noticeably different amounts of sodium hydroxide were automatically added to each vessel, which differentially affected the pH and osmolality of the medium. Both pH and sodium concentration can have profound effects on *Clostridium* metabolic output [57,58].

In the experiments with pH control, the consortium of *C. beijerinckii* and *Y. regensburgei* produced a drastically different metabolic profile than in other experiments (Figure 7). After an initial lactate decrease, which corresponded to an increase in *C. beijerinckii* colony forming units (Figures 7a and 6e), the consortium cultures produced more lactate than *Y. regensburgei* monocultures. This increase in lactate corresponded to a decrease in *C. beijerinckii*, while *Y. regensburgei* levels remained near constant. Thus, *Y. regensburgei* likely dominated the metabolite profile of the cultures, and *C. beijerinckii* may have either shifted its metabolism to produce additional lactate or cross-fed a metabolite to *Y. regensburgei* increasing its lactate production. Modifying the metabolic output of a consortium by either controlling or not controlling the culture pH is a mechanism that could be used to increase the agility of a designed microbial consortium for a future application.

#### *4.4. Microbial Consortia with Distinct Mechanisms of Chemical Overproduction*

Modeling predicted the consortium of *C. beijerinckii* and *Y. regensburgei* to produce increased *C. beijerinckii* biomass (Table 1), due to the cross-feeding of lactate from *Y. regensburgei* to *C. beijerinckii* that was experimentally observed (Figure 4). Other consortia that were predicted to produce an increased amount of hydrogen compared to *C. beijerinckii* monocultures (Table 1) likely have different mechanisms that could contribute to commodity chemical overproduction. For example, *Cellulosimicrobium cellulans* can break down cellulose and xylans [59], complex polysaccharides common in plant material [60], which are inaccessible to *C. beijerinckii*. In contrast to the consortium of *C. beijerinckii* and *Y. regensburgei*, FBA predicted that the biomass of both *C. beijerinckii* and *Cellulosimicrobium cellulans* would increase in a co-culture compared to monocultures (Table 1), perhaps due to increased available carbon and cross-feeding of metabolites from *C. beijerinckii* to *Cellulsimicrobium cellulans*. Similarly, members

of the *Capnocytophaga* genus can metabolize complex polysaccharides [61] and carbon dioxide [62], which could explain the predicted hydrogen overproduction by a consortium of *C. beijerinckii* and *Capnocytophaga sputigena* (Table 1). Integrating genome-scale metabolic model improvements and testing additional consortia with different predicted interactions could enable discovery of pathways to enhanced commodity chemical production.

#### **5. Conclusions**

Genome-scale metabolic models and flux-balance analysis predicted several microbial consortia expected to produce a significantly greater amount of hydrogen than any individual species [39] (Table 1). We focused on the consortium of *C. beijerinckii* and *Y. regensburgei* that was predicted to have the greatest increase in biomass and hydrogen production by *C. beijerinckii*. The flux-balance analysis predicted the increase in hydrogen and biomass based on the cross-feeding of lactate to *C. beijerinckii* (Figure 2). We established that cross-feeding of lactate from *Y. regensburgei* to *C. beijerinckii* can occur (Figure 4), and that exogenous lactate is capable of supporting the growth of *C. beijerinckii* (Figure 5). We were unable to demonstrate that growing *C. beijerinckii* as part of a consortium with *Y. regensburgei* increased the production of hydrogen above the levels observed for *C. beijerinckii* monocultures (Figure 3). The consortium was capable of producing more butyrate (Figure 3c) or lactate (Figure 7e) than individual monocultures, depending on the growth conditions. Examining the effects of adding another species that is predicted to have a different role in promoting hydrogen production than cross-feeding of lactate to the consortium of *C. beijerinckii* and *Y. regensburgei*, or testing that species with *C. beijerinckii* alone could provide better improvement of hydrogen production from the artificial garbage slurry than was observed. The framework presented here can be used to screen large numbers of possible microbial combinations by first using genome-scale metabolic modeling and flux-balance analysis to predict consortia with an increased likelihood to convert waste to commodity chemicals and then testing the highest producing consortia experimentally for their ability to produce chemicals of interest from readily available materials.

**Author Contributions:** N.D.S., M.A.P., and K.L.A. conceptualized the research; M.A.P. performed the GSMM and FBA, and wrote the original draft of the Introduction, Materials and Methods, and Results section for the GSMM. W.M. performed the HPLC and GC analysis and wrote the original draft of the Materials and Methods section for the HPLC and GC. N.D.S. and E.S.G. performed the bioreactor fermentation experiments and edited the paper. M.B. performed the analysis of the bioreactor GC data. N.D.S. performed all other experiments and wrote the original draft of the remainder of the paper.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors would like to acknowledge members of the Combat Capabilities and Development Command Army Research Laboratory for useful discussions and Yue Li of the University of Maryland for performing mass spectrometry.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2019 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/).

*Article*
