Response to starvation and microbial community composition in microbial fuel cells enriched on different electron donors

Summary In microbial fuel cells (MFCs), microorganisms generate electrical current by oxidizing organic compounds. MFCs operated with different electron donors harbour different microbial communities, and it is unknown how that affects their response to starvation. We analysed the microbial communities in acetate‐ and glucose‐fed MFCs and compared their responses to 10 days starvation periods. Each starvation period resulted in a 4.2 ± 1.4% reduction in electrical current in the acetate‐fed MFCs and a 10.8 ± 3.9% reduction in the glucose‐fed MFCs. When feed was resumed, the acetate‐fed MFCs recovered immediately, whereas the glucose‐fed MFCs required 1 day to recover. The acetate‐fed bioanodes were dominated by Desulfuromonas spp. converting acetate into electrical current. The glucose‐fed bioanodes were dominated by Trichococcus sp., functioning as a fermenter, and a member of Desulfuromonadales, using the fermentation products to generate electrical current. Suspended biomass and biofilm growing on non‐conductive regions within the MFCs had different community composition than the bioanodes. However, null models showed that homogenizing dispersal of microorganisms within the MFCs affected the community composition, and in the glucose‐fed MFCs, the Trichococcus sp. was abundant in all locations. The different responses to starvation can be explained by the more complex pathway requiring microbial interactions to convert glucose into electrical current.

. Schematic of a hydraulic loop containing two microbial fuel cells (MFC).     . Results from cyclic voltammetry carried out in the end of the MFC experiment. MFC3 had been operated with acetate feed and MFC7 with glucose feed. The potential was scanned between +0.2 and -0.8 V vs SHE in three cycles at a rate of 1 mV/s. The results from the third cycle are shown. Cathodic current is negative. Figure S8. Prolonged enrichment of the biological electrodes from MFC3 and MFC7 under anaerobic conditions in a solution containing hydrogen ions and carbon dioxide as the only possible electron acceptors. The electrode potential was first controlled at -0.65 V vs SHE and then at -0.8 V vs SHE. Cathodic current is negative. Figure S9. Heatmap showing the most abundant taxa categorized based on class. On the vertical axis, the abbreviation c__ refers to class and p__ refers to phylum. The labels on the horizontal axis show the samples type, the MFC, and the type of electron donor used as feed (ac refers to acetate and glu refers to glucose-feed). The numbers in the heatmap show percentage relative abundance in the sample. Table S1. Raup-Crick dissimilarity values calculated based on 1-C0. A value >0.95 indicates that the difference in community composition between two samples is significantly higher than could be explained by random assembly. A value < -0.95 indicates that the difference is significantly smaller than could be explained by random assembly (see Chase, et al., 2011). Table S2. Raup-Crick dissimilarity values calculated based on 1-C1. A value >0.95 indicates that the difference in community composition between two samples is significantly higher than could be explained by random assembly. A value < -0.95 indicates that the difference is significantly smaller than could be explained by random assembly (see Chase, et al., 2011 Appendix S1: Description of the gas-diffusion cathodes The base layer of the cathode was a carbon paper (AvCarb P75T with 30% wet proofing, Fuelcellearth.com). The air-facing side was painted with a mixture 20 μL/cm 2 water, 6 mg/cm 2 graphite powder (200 mesh, Alfa Aesar), and 12 mg/cm 2 PTFE. The liquid-facing side was painted with a mixture of 30 μL/cm 2 propanol, 0.15 mg/cm 2 carbon black (Cabot Black Pearls 2000), and 0.05 mg/cm 2 PTFE. The painted electrodes were sintered at 350°C for 20 min. A stainless-steel wire mesh pressed against the air-facing side of the gas-diffusion electrode served as current collector. A glass fiber filter (Munktell) pressed against the liquid-side of the cathode served as a separator between the cathode catalyst layer and the bulk liquid.

Appendix S2: Microbial community analysis methods
Wet lab procedures DNA was extracted from samples using the FastPrep for soil DNA extraction kit. For the anode, about 2 cm 2 of carbon cloth was cut into 4 pieces and put into the extraction tube. For the cathode, which contained more biomass, about 1 cm 2 cut into four pieces was used for each sample. Samples from the inoculum and bulk liquid were centrifuged and a portion of the biomass pellet was added to the extraction tubes.
The bioanode and biofilm samples taken from MFC2 and MFC6 were divided into six different 2 cm 2 regions. DNA was extracted and sequenced from each separate region. We call the six samples from the same electrode or biofilm for extraction replicates.
The extracted DNA concentration was normalized to 20 ng/μL. PCR amplification of the V4 region of the 16S rRNA gene was carried out with the primer pair 515'F (GTGBCAGCMGCCGCGGTAA) and 806'R (GGACTACHVGGGTWTCTAAT) (Caporaso, et al., 2011, Hugerth, et al., 2014. In PCR tubes, 17 μL Accuprime Pfx supermix and 1 μL each of the forward and reverse primers (10 μM), and the DNA template were mixed. PCR reactions were done in duplicates, which were then pooled. The PCR program consisted of activation (95°C, 5 min); 30 cycles of denaturation (95°C, 20 sec), annealing (50°C, 20 sec) and elongation (68°C, 60 sec), followed by a final elongation (68°C, 10 min). The products from all samples were purified, normalized to equal concentration and pooled prior to sequencing on an Illumina MiSeq using Reagent Kit V3.

Bioinformatics
Raw sequence reads were processed in Usearch (version 10). Paired-end reads were merged with the fastq_mergepairs command allowing a maximum of 50 mismatches or at least 80% identities in the alignment. The merged reads were quality filtered using a maximum expected error (maxee) cutoff of 1 and a sequence length between 200 bp and 270 bp. The quality filtered reads were used as input to the Unoise algorithm to generate sequence variants . A minimum abundance threshold of 4 was specified in the unoise3 command. This means that sequence variants were discarded if they were represented by fewer than 4 quality filtered reads across all samples. A count table was constructed using the otutab command by mapping the merged reads to the sequence variants with a nucleotide identity threshold of 99% required for a match. Taxonomic classification was done with the sintax algorithm  using the Midas database (v. 213) (McIlroy, et al., 2015). The sequences in the Midas database were first trimmed to the V4 region by local alignment to the primer pair used in the study.