The gene sequences of ldhD from L. bulgaricus ATCC11842 and the lldP from E. coli K-12 strain MG1655 were extracted from the NCBI database (www.ncbi.nlm.nih.gov/gene/) (Table S1). To prevent shortage of tRNAs for rare codons, we used a Java codon adaption tool (JCAT) to obtain the optimal sequences for K.pneumoniae (Yang et al., 2015a). The ldhD gene and lldP gene were then in vitro synthesized. Subsequently, the sequences of ldhD and lldP were linked in vector pUC18K, obtaining recombinant plasmid pUC18K-ldhD-lldP. The designed gene sequences were synthesized and verified in vitro. All the plasmid construction was performed in E. coli Trans T1.

The KG-2 strain was obtained by deleting the adhE gene and the pta gene from KG-1 by following a previously established procedure (Guo et al., 2014). The recombinant K. pneumoniae strain KG-3 was engineered by electro-transferring the plasmid pUC18K-ldhD-lldP into KG-2. Two millimeter electroporate chamber and pulser voltage selected 1.8 kV were used in this electro-transferring process. The plasmid pYYDT-C5 harboring the flavin synthesis genes (ribADEHC) from B. subtilis (Yang et al., 2015a) was transformed into E. coli WM3064, and then was transferred into a highly hydrophobic S. oneidensis strain CP-S1 by conjugation. All stains and plasmids used in this study are listed in Table 1.

The H-type MFCs were separated by Nafion 117 membranes, which were pretreated as those used in the previous study. The anode and cathode were composed of carbon cloth for anode (2.5 cm × 2.5 cm) and cathode (2.5 cm × 3 cm). The anodic electrolyte consisted of M9 buffer (Table S2) supplemented with definite carbon source and 5% (v/v) LB broth. The cathodic electrolyte was made of 50 mM KH₂PO₄, 50 mM K₂HPO₄, and 50 mM K₃[Fe(CN)₆] solution. The amplification cultured cells (OD600 = 2.0) were centrifuged, collected, and dispersed the anode chambers. To ensure consistent culture condition for different MFCs, 50 μg/mL kanamycin and 0.2 mM IPTG were supplemented in anodic electrolyte. All the MFCs were incubated at 30°C, and the voltage generation was measured by a digital multimeter (DT9205A) across the external loading a 2 kΩ resistor in MFCs.

The metabolites in the shake flasks were analyzed using a high performance liquid chromatography (HPLC) system equipped with a refractive index detector. All samples and standard solutions were diluted and filtered by a 0.22 μm filter prior to HPLC assay. To analyses glucose, xylose, lactate, acetate, and ethanol, 5 mM sulfuric acid was used as the mobile phase (0.6 mL/min) through the Aminex HPX-87H ion-exchange column (Bio-Rad) which was incubated at 45°C, and detection was performed with the Waters 2414 refractive index detector. In addition, the determination of riboflavin concentration can refer to our previous method (Li et al., 2018a).

Electrochemical analyses were performed on a CHI 1000C multichannel potentiostat (CH Instrument, Shanghai, China), when MFCs reached the pseudo-steady state. Cyclic voltammetry (CV) was conducted on a three-electrode configuration with the cathode as the counter electrode, the anode as the working electrode, and the Ag/AgCl (vs. SHE) as the reference electrode. To obtain the maximum power density and the polarization curves, linear sweep voltammetry (LSV) analysis was also performed on the CHI 1000C multichannel potentiostat.

The actually recovered coulombs were determined by integrating the current (I) over a period of batch cycle (t_b). Such that the coulombic efficiency can be evaluated as Yan et al. (2015); Li et al. (2018a,c):

where M_S (g/mol) is the molecular weight of the substrate, F is the Faraday's constant (98.485 C/mol of electrons), C is the symbol of coulomb, I (A) is the current, t_b (s) is the time period of a batch cycle, b_ES is the stoichiometric number of moles of electrons produced per mole of substrate (glucose: b_ES = 16 when acetate was used; xylose: b_ES = 12 when acetate was used), V_An (L) is the volume of liquid in the anode compartment, and Δc (g/L) is the change in substrate concentration over a bath cycle time.

To measure the composition of microbial consortium, anode carbon were sampled and assayed by colony forming unit (CFU) counting. The anode was placed in a 50 mL test tube containing 5 mL of PBS (pH = 7.2) solution, vortexed oscillation for 2 min. A series of dilutions were spread onto LB agar plates, and incubated at 30°C for 36 h. The colony of S. oneidensis and K. pneumoniae was dark orange and white, respectively.

Corn straw (CS) used in this study was harvested in the suburb of Tianjin, China. CS was milled coarsely using a beater pulverizer and the fractions between 20 and 80 meshes were screened. The CS was further transformed to hydrolysates upon pretreatment and enzymatic hydrolysis. The procedure of pretreatment was carried out by following a previously published procedure with minor modifications (Li et al., 2016, 2018c). And the enzymatic hydrolysis experiments were performed in accordance with the NREL standard protocol (LAP-009). After enzymatic hydrolysis at 30°C and 200 rpm for 72 h in an orbital incubator, the mixture was centrifuged at 12,000 rpm for 5 min to separate the hydrolysates from solid residues. The hydrolysates were frozen at −20°C for subsequent sugar analysis.

As the fermenter, K. pneumoniae could utilize glucose or xylose as a carbon source to produce lactate, concurrently producing a considerable amount of by-products, i.e., ethanol and acetate (Guo et al., 2014). Therefore, to drive more glucose or xylose to synthesize lactate, we eliminated the ethanol and acetate biosynthesis pathways by knocking out the gene adhE encoding the alcohol dehydrogenase and gene pta encoding phosphotransacetylase in K. pneumoniae. However, interruption of ethanol biosynthesis would accumulate abundant intracellular reducing equivalent (NADH), causing unbalancing of intracellular redox, which further inhibited cell viability. To solve this redox unbalance and increase lactate synthesis, we expressed the ldhD gene encoding lactate dehydrogenase from L. bulgaricus, thus consuming the excess NADH. Concurrently, a lactate transporter encoded by lldP gene from E. coli was introduced in K. pneumonia to increase the lactate transportation across the hydrophobic cell membrane. Thus, as shown in Figures 2A–C, compared with the wild K. pneumonia KG-1, the recombinant K. pneumoniae strain KG-3 expressed lactate biosynthesis pathway, deleted by-products synthesis and produced a significant amount of lactate, almost in the absence of ethanol and acetate.

In addition, based on different electron transfer mechanisms, two strategies were adopted to facilitate extracellular electron transport. To improve the direct contact-based electron transfer pathway of exoelectrogens, the CP-S1 S. oneidensis mutant was used to further improve the adhesion of S. oneidensis to the carbon electrode surfaces (Figure S2). To enhance the shuttles-mediated electrons transfer, a flavin biosynthesis pathway was assembled in CP-S1 to further accelerate electron transfer. As a result, the flavins produced by the CP-S1-C5 increased by 6.4-folds compared with that of the WT S. oneidensis MR-1 (from 1.2 ± 0.1 μM to 8.9 ± 0.3 μM) (Figure 2D).

To realize a continuous and stable supply of carbon sources for exoelectrogen and achieve a maximum MFC power generation, we optimized the seeding ratio of the microorganisms and the feed-in glucose and xylose concentration, respectively. Firstly, the seeding ratio of the K. pneumoniae and S. oneidensis was optimized by a serial of orthogonal experiments to achieve a maximum MFC power generation in anolytes. The result indicated that the seeding OD₆₀₀ ratio between K. pneumoniae and S. oneidensis being 1 to 5 would reach a maximum bioelectricity generation in MFCs (Figure S1). Secondly, to ensure that S. oneidensis had a suitable continuous supply of lactate produced by K. pneumoniae for the maximum MFC power generation, we further optimized the initial feed-in glucose and xylose concentrations. Three different concentration ratios of glucose and xylose, i.e., 30 mM (glucose): 10 mM (xylose), 20 mM: 20 mM, and 10 mM: 30mM xylose were chosen to determine the optimal substrate concentration for our engineered strains. As shown in Figures 3A–C, the results showed that the feed-in ratio of 20 mM (glucose): 20 mM (xylose) reached a highest output voltage (268.5 ± 10 mV), while 10 mM: 30 mM and 30 mM: 10 mM generated a lowest MFC output voltage, 120.2 ± 7.0 mV and 210.3 ± 6.0 mV, respectively. Furthermore, as shown in Figure 3D, the coulomb efficiency of the MFC inoculated with KG-3: CP-S1-C5 and substrates (glucose: xylose = 20 mM: 20 mM) was 33.3%, which was higher than in the other scenarios.

To illuminate the underlying reason, the lactate concentrations sampled from the anolyte were measured after 24 h cultivation. The data indicated that feed-in ratio 30 mM (glucose) and 10 mM (xylose) could cause a higher level accumulation of lactate in MFCs (Figure 3E), however, the MFC output voltage was not the optimum. This result was caused by the fact that the rapidly accumulated lactate resulted in an acidic condition (pH = 5), in which the growth activity and flavins biosynthesis of cell would be inhibited, reducing the power generation capability of S. onedensis (Yong et al., 2013; Lin et al., 2016). When feeding in 10 mM glucose and 30 mM xylose in MFC, the voltage output maintained at a low level all the time. This could be explained as the substrate could be exhausted rapidly, leading to a shortage of lactate to feed S. oneidensis. However, upon feeding 20 mM glucose and 20 mM xylose, the recombinant consortium enabled higher and longer voltage output than other concentrations. In this scenario, K. pneumoniae can continuously supply lactate to S. oneidensis, and the lactate from this feed-in mix sugar concentration was neither too high to generate cellular toxicity to S. oneidensis, nor too low as substrates to feed S. oneidensis. In addition, the substrates consumption results demonstrated that the engineered strain KG-3 showed a superior performance of substrate consumption than that of the KG-1 and the KG-2 (Table S3). Meanwhile, these strains preferred to utilize glucose than xylose for the production of lactate.

The cyclic voltammetry (CV) was conducted to reveal the redox reaction kinetics at the interfaces of bacterial cells and anodes with a scan rate of 1 mV/s (Li et al., 2018b,c). As shown in Figure 4A, there were typical redox peaks of favins in the CV curves starting from around ~-0.4 V (vs. Ag/AgCl), revealing that the dominating mechanism for bioelectricity production in this microbial consortium was the favins-mediated EET. Furthermore, another catalytic current arose at ~-0.27 V (vs. Ag/AgCl) was observed in these strains, which was caused by the direct involvement of conductive c-type cytochromes. To further investigate the MFCs' performance inoculated with these microbial consortia, the linear sweep voltammetry (LSV) and the polarization curves at a low scan rate of 0.1 mV/s were obtained (Baron et al., 2009). As shown in Figure 4B, the maximum catalytic current of KG-3: CP-S1-C5 was the highest (408.2 ± 8.0 mA/m²), which was higher than the maximum catalytic current of KG-1: WT (119.6 ± 6.0 mA/m²) and KG-2: CP-S1 (167.84 mA/m²). Notably, the dropping slope of the polarization curve obtained from the engineered KG-3: CP-S1-C5 was smaller than those obtained from the other two microbial consortium (KG-1: WT and KG-2: CP-S1), implying that the internal charge transfer resistance of the MFC inoculated with KG-3: CP-S1-C5 was relatively smaller (Figure 4C). The power densities were calculated for each; the engineered KG-3: CP-S1-C5 microbial consortium obtained a maximum power density of 104.7 ± 10.0 mW/m², which was 7.2- and 2.2-folds higher than that of the KG-1:WT (12.7 ± 3.0 mW/m²) and KG-2:CP-S1 (32.7 ± 6.0 mW/m²), respectively.

To evaluate the electricity generation capacity of utilizing cellulose hydrolyzates, this synergistic microbial consortium (KG-3: CP-S1-C5) was further inoculated to the sterile or non-sterile corn straw hydrolyzates-fed MFC. As shown in Figure 5, the maximum output voltage of the sterile cellulosic hydrolyzates-fed MFCs could reach 152.3 ± 8.0 mV, the maximum current density was 72.8 ± 5.0 mA/m², and the maximum power density of the sterile cellulosic hydrolyzates-fed MFC could reach 23.5 ± 6.0 mW/m².

To further verify the stability of the engineered co-cultures, particularly over long term and in the presence of non-sterile feeds, we further used non-sterile corn straw hydrolyzates as the carbon source to evaluate the electricity generation capability. As shown in Figure S3, the output voltage of MFCs with a multi-cycle operation revealed that the engineered co-culture enabled a stable power generation. And the results of bioelectrochemical characterization also demonstrated that the engineered co-culture (KG-3: CP-S1-C5) showed a superior performance of electricity generation than that of co-culture (KG-1: WT). In all, our engineered co-culture showed stability of power generation in MFCs with either sterile or non-sterile feeds.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.