Impact of Operational Conditions on Methane Yield and Microbial Community Composition during Biological Methanation in a Hybrid Reactor System


 Background: Biogas can be upgraded to methane biologically by adding hydrogen to biogas reactors. The process is called biological methanation (BM) and can be done in-situ in a regular biogas reactor or the biogas can be transferred to a separate ex-situ upgrading reactor. The hybrid BM concept, a combination of in-situ and ex-situ BM, has received little attention, and only a few studies have been reported. The hybrid BM has the advantage of resolving the issue of pH increment during in-situ BM, while the size of the ex-situ BM reactor could be reduced.Results: In this study, the efficiency of in-situ and hybrid biological methanation (BM) for upgrading raw biogas was investigated. The hybrid BM system achieved a CH4 yield of 257 mL gVS-1 when degrading a feedstock blend of manure and cheese waste. This represented an increase in methane yield of 76% when compared to the control reactor with no H2 addition. A 2:1 H2:CO2 ratio resulted in stable reactor performance, while a 4:1 ratio resulted in a high accumulation of volatile fatty acids. H2 consumption rate was improved when a low manure-cheese waste ratio (90%:10%) was applied. Furthermore, feeding less frequently (every 48 hours) resulted in a higher CH4 production from CO2 and H2. Methanothermobacter was found to dominate the archaeal community in the in-situ BM reactor, and its relative abundance increased over the experimental time. Methanosarcina abundance was negatively affected by H2 addition and was nearly non-existent at the end of the experiment. Conclusions: Our results show that hybrid BM outperforms in-situ BM in terms of total CH4 production and content of CH4 in the biogas. The application of hybrid BM increased CH4 yield up to 42%. Furthermore, addition of H2 at 2:1 H2:CO2 ratio in in-situ BM resulted in stable reactor operation.

The characteristics of the inoculum and the applied substrates are given in Table 1. Operating parameters and performance data for the 10 L control and upgrading reactors (CR, UR) under steady-state conditions are summarized in Table 2 and 3, respectively. The experiment was conducted for 172 days and divided into six phases. Figures 2 and 3 illustrate the changes in methane yield, pH, and VFAs over the experimental period for upgrading and control reactors.
Phase I: Initial phase -without H 2

addition
In this phase, the two reactors were operated identically and showed very similar performance in terms of biogas production (241-245 mL g -1 VS ) and CH 4 yield (144-145 mL g -1 VS ) ( Table 3). The average CH 4 content of the reactors (58 to 59%) and the pH (7.9) were also similar. The total VFA content was around 18 mM, with acetic acid (AA) accounting for more than 60% of the total VFAs. The ratio of propionic acid (PA) to AA of both reactors was below 1.4, indicating a stable AD process according to [18]. The TAN concentration was around 2.5 g L - 1 . The values align well with those obtained by [19], who observed that a TAN value of 2.5g L -1 (pH 7.9) resulted in stable biogas production during thermophilic (55 o C) anaerobic digestion of cow manure.
Phase II: Initial H 2 phase H 2 was added in UR from day 64 at a ow rate of 3 mL min -1 , corresponding to a H 2 :CO 2 ratio of 2:1. As shown in Figure 2, CH 4 yield increased immediately after H 2 addition and stabilized from day 70. The average CH 4 yield of UR was 185 mL g -1 VS , which was approximately 27% higher than the average CH 4 yield of CR (Table 3). A similar observation was reported by Treu et al., [20] where H 2 addition into a CSTR at a 2:1 ratio resulted 13% increase in CH 4 yield. The pH of UR increased from 7.94 to 8.10, while the pH of CR remained the same as in phase 1. BM resulted in a rise in pH due to the removal of CO 2 from the liquid phase. Bicarbonate ions (HCO 3 -) are produced during the AD process when CO 2 reacts with OH in the liquid phase, contributing to the buffering capacity of the reactor. Addition of H 2 to the system resulted in CO 2 consumption and thus loss of buffering capacity [14]. Similar ndings have been reported in previous studies [11,20,21]. Total VFA levels in UR rose to more than double the amount in phase I. In contrast to our study, Treu et al., [20] reported relatively low and stable VFA levels after H 2 addition.
In CR, the average AA concentration was 21 mM, while in UR, it was 36 mM. PA levels were slightly higher in both reactors than in phase 1. TAN concentrations were also elevated, with 2.57 g L -1 for CR and 2.77 g L -1 for UR. The H 2 consumption rate of UR was calculated to be 25%, corresponding to a CH 4 production rate of 0.04 mL L -1 d -1 .

Phase III: Increased stirring speed
In phase III, the stirring speed of both reactors was increased from 80 to 140 rpm (day 79) in an attempt to improve the transfer of H 2 to the liquid phase in UR.
As shown in Figure 2, the CH 4 yield from UR decreased signi cantly as the stirring speed increased. The CH 4 yield of UR was reduced from 185 (day 78) to 126 mL g -1 VS (day 85) for UR. The decrease in CH 4 yield of UR was corroborated by the accumulation of acetate (67 mM on average), which was nearly double of what was measured in phase II ( Figure 3b). Besides, the propionate concentration was slightly increased from 9 to 13 mM. These observations could indicate that parts of the microbial community were negatively affected by the higher share forces at 140 rpm.
Regardless of the fact that the total CH 4 yield decreased as the stirring speed increased, the H 2 consumption rate in UR increased from 25% to 46%. This observation was in agreement with our previous study [22]. The rate of CH 4 production from H 2 and CO 2 conversion was increased from 0.04 to 0.08 mL L -1 d -1 . For the CR, the CH 4 yield was reduced from 143 to 131 mL g -1 VS . Ghanimeh et al., [23] observed a decrease in CH 4 yield when stirring speed was increased from 80 to 120 rpm. No AA accumulation was observed in the CR, whereas the PA level was slightly higher than in phase II (12 mM) ( Figure 3a and Table 3).
The pH in both reactors was higher than in phase II, with pH of 8.15 and 8.28 for CR and UR, respectively. The elevated pH in UR can be attributed to greater CO 2 consumption in the liquid as a result of the increased H 2 gas-liquid mass transfer rate at higher stirring speeds and thus higher BM activity [1].
Phase IV: Change of feedstock blend ratio On day 86, the stirring speed was again reduced to 80 rpm (return to Phase II conditions), and the CH 4 yield rose signi cantly until it reached a plateau from day 90 ( Figure 2). From day 92 the CW fraction was increased from 10% to 20% on day 93 (Phase IV), resulting in an OLR of 0.78 g VS L -1 d -1 . The CH 4 yield increased in both reactors, with maximum values being 195 mL g -1 VS (CR) and 276 mL g -1 VS (UR) (Figure 2). After day 102, however, the CH 4 yield gradually decreased until it reached a stable period around day 111. During the stable period, the average CH 4 yields of CR and UR were 142 mL g -1 VS and 204 mL g -1 VS , respectively ( Table 3). The average CH 4 yield of CR measured in this study was lower than that measured by Comino et al., [24] (similar feedstock blend, 80% CM:20% whey), despite the fact that both studies had comparable CH 4 content (53%). Longer HRT (41 days) and higher OLR (3.33 g VS L -1 d -1 ) were used by Comino et al., which may explain the difference in performance. The average CH 4 content of UR was 39%. The H 2 consumption rate was around 17%, which was 31% lower than the consumption rate when CW fraction was set at 10%. The total VFA content of CR was slightly higher towards the end of phase IV (Figure 3a), while the total VFA content of UR was relatively stable (Figure 3b). The pH of both reactors was lower than in phase III, with an average pH of 7.91 for CR and 8.11 for UR. Increased CW ratio to 20% resulted in higher TAN values (both reactors) compared to phase II, suggesting more thorough CW degradation as TAN is a product of protein degradation.

Phase V: Feeding frequency
In phase V, the CW fraction was reduced to 10% and the feeding frequency was changed to once every 48 hours (instead of once per 24 hours). In term of CH 4 yield for CR, no changes were observed, while CH 4 yield for UR was gradually reduced until a stable period was achieved (day 134). The average CH 4 yield for CR was 139 mL g -1 VS and 194 mL g -1 VS for UR. The CH 4 yield of UR in phase IV was slightly higher than in phase II (feeding every 24 hours). The H 2 consumption rate was higher than phase II (24 h feeding) when the reactor was fed every 48 hours (25% vs 32%). The increased CH 4 yield and H 2 consumption rate in UR could be attributed to enrichment of hydrogenotrophic methanogens in less frequent feeding. According to Piao et al., [25], reducing feeding frequency tended to increase the abundance of H 2 -utilizing methanogens. In the Piao study, the abundance of hydrogenotrophic methanogens increased from 45% to 53% when feeding frequency was reduced from every 24 hours to every 48 hours. The average total VFA content for CR and UR were 26 and 50mM, respectively. The pH of both reactors was slightly lower than in phase II.
Phase VI: Increased H 2 :CO 2 ratio Substrate feeding was changed to once daily starting on day 141, and the H 2 ow rate was increased to 6 mL min -1 , equivalent to a 4:1 H 2 :CO 2 ratio (Phase VI). The increased H 2 :CO 2 ratio initially boosted CH 4 yield in UR with a maximum at day 151. However, the yield fell after day 163. The average CH 4 yield in this period was 165 mL g -1 VS , about 11% lower than the value in phase II (H 2 :CO 2 ratio = 2:1). Despite the lower CH 4 yield, the H 2 consumption rate was doubled (54%) compared to phase II (25%) due to the increased H 2 :CO 2 ratio, which probably stimulated H 2 -consuming anaerobic microbes.
AA accumulated toward the end of the phase, reaching a maximum concentration of 84.5 mM. The increase in AA levels may be explained by the inhibition of acetoclastic methanogens (e.g. Methanosarcina) caused by high H 2 partial pressure [26] or by the enrichment of particular microbial pathways such as homoacetogenesis (Wood-Ljungdahl pathway) [6]. PA content was also increased from 15 to 18 mM when the H 2 :CO 2 ratio was increased. The rise in total VFA content coincided with a drop in pH from 8.01 to 7.91. For CR, the CH 4 yield remained consistent throughout phase VI, with an average of 134 mL g -1 VS . The average total VFA concentration was 21 mM, with a pH of 7.82. AA concentration accounted for 58% of the total VFA content. The TAN concentration was 2.65 g L-1, which was similar to the value observed in phase II (2.57 g L-1).
In-situ vs. hybrid con gurations A hybrid con guration was tested at the end of the experiment (after day 172). An additional 2 L reactor lled with packing materials was used as an ex-situ biogas upgrading reactor (HR) for the biogas from UR ( Figure 1b). Initially, the operating parameters of UR were adjusted to the same as in phase II with a H 2 :CO 2 ratio of 2:1.
When the hybrid setup was used instead of an in-situ (phase II), 39% extra CH 4 was obtained ( Figure 4). The average CH 4 yield rose from 185 to 257 mL g -1 VS . Furthermore, the H 2 consumption rate increased by twofold compared to in-situ (phase II), and the average CH 4 content increased from 40% to 63% (Tables 3   & 4). The CH 4 content without considering H 2 from hybrid system was around 80%. When compared to the control reactor (Figure 4), the hybrid con guration resulted in a 76% higher CH 4 yield, while in-situ con guration resulted in 27% more CH 4 ( Figure 4). HR had an average pH of 8.07 and an AA concentration of approximately 4.12 mM. The TAN concentration of HR was around 1.09 g L-1.
The H 2 :CO 2 ratio was increased to 4:1 after a stable condition was observed. The average CH 4 yield fell from 257 mL g -1 VS to 234 mL g -1 VS (approximately 9% less CH 4 ). The average CH 4 content was reduced from 63 to 51%. Nonetheless, the H 2 consumption rate (62%) was slightly higher than at the 2:1 H 2 :CO 2 ratio (60%), indicating that acetate-oxidizing bacteria had the capacity to consume more H 2 to produce acetate, as observed in phase VI. Compared to in-situ con guration (phase VI), about 42% extra CH 4 was measured and approximately 75% more CH 4 was produced when compared to control ( Figure 4). The concentrations of AA and TAN were equivalent to those found at a 2:1 H 2 :CO 2 ratio.
Compared to Corbellini et al., [15] our study resulted in lower upgraded CH 4 content of in-situ BM. This may be attributed to differences in reactor working volume, as a larger working volume (6L) was used in the present study compared to 3L in [15]. Our ndings were more comparable to those of [17], who used a 9L working volume for in-situ testing. Furthermore, when a 4:1 H 2 :CO 2 ratio was added to UR in our study, AA accumulation (> 4 g L -1 ) was observed, leading to a decrease in pH, while VFA level observed in [15] was maintained at 2 g L -1.
To prevent process instability in in-situ BM reactor, we propose that the amount of H 2 added to the in-situ reactor should be kept at a relatively low H 2 :CO 2 ratio (e.g. 2:1). This will minimize the increase in pH caused by bicarbonate removal as well as the possible inhibition of some anaerobic bacteria that are sensitive to high H 2 partial pressure. Our study discovered residual H 2 in the in-situ and hybrid BM reactors, indicating that further optimization is required. A pressurized reactor may be a solution. Increased operating pressure enhances the solubility of gases and decreases bubble size. Smaller bubble size is bene cial since it maximizes the contact area between bacteria and gaseous substrates while slowing gas up ow through the reactor [1,27]. Previous research found that increasing reactor pressure during in-situ and ex-situ BM resulted in improved conversion e ciency [28,29]. A very high CH 4 concentration (> 98%) in the biogas was reported when reactor pressure was set between 5 and 15 bars for a 5 m 3 ex-situ CSTR [30]. Additionally, the design of the ex-situ reactor used in our study can be improved, for example, by using a long column design like trickle-bed reactor.

Microbial community composition
Microbial analysis of the reactor feed (80% CM:20% CW) showed that Firmicutes and Proteobacteria were the two dominant bacterial phyla, accounting for approximately 50 and 18 % of the abundance, respectively (Figure 5a). Other phyla present in the feed included Actinobacteria (9%) and Bacteriodetes (8 %). Analysis of the inoculum microbiology showed that Firmicutes was the dominating phylum (71%), followed by Synergistetes (7%), Actinobacteria, and Euryarchaeota (both phyla accounted 3% abundance) (Figure 5b). Atribacteria and Thermotogae were also detected in the inoculum, but they were not found in the feed sample.
The taxonomic classi cation of the microbial community revealed that Firmicutes were the most abundant phyla n the reactors, accounting for 57 to 72% of relative abundance depending on the time points (Figure 5c). This is in agreement with the ndings of [31] where Firmicutes dominated a thermophilic biogas reactor digesting cow manure. Firmicutes engages in a variety of metabolic processes for carbohydrate and fatty acid degradation, including the Wood-Ljungdahl pathway (homoacetogenesis) and syntrophic acetate oxidation, which explains their abundance in the reactors [11]. Clostridia, which belong to the Firmicutes, was the most abundant class (representing more than 33% of all bacterial sequences). Other bacterial phyla, such as Synergistetes and Bacteriodetes, were present in both reactors at rst, but their numbers declined over time. In terms of methanogenic population, the abundance of Euryarchaeota varied over time, between 13 -33% for CR, and 18 -38% for UR (Figure 5c).
Some bacteria, such as HAW-R60, an Atribacteria phyla, was clearly negatively affected by H 2 addition (Figure 6a). Their abundance declined over time and was nearly non-existent in phase VI.Atribacteria have been found previously in thermophilic biogas reactors and are involved in hydrolysis of polysaccharides [32]. Another hydrolytic bacterium, Halocella, behaved differently, reaching highest abundance when the H 2 :CO 2 ratio was increased to 4:1 (phase VI) ( Figure   6b). Their abundance in UR increased from 6.7 (without H 2 addition) to 14.6%. The increase in stirring speed in phase II (day 79-85) seemed to negatively affect Halocella, with decreased abundance in both CR and UR. The cellulolytic bacteria Halocella belong to the class Clostridia and is responsible for cellulose degradation and produces ethanol and H 2 from lignocellulosic substrates [33]. Halocella have mainly been found in manure-based samples and their presence in thermophilic biogas reactor has been reported previously [34].
Within the domain archaea, Methanosarcina was the only detected methanogen capable of acetoclastic methanogenesis, although it can also carry out hydrogenotrophic methanogenesis [35]. Methanosarcina was clearly negatively affected by H 2 addition and disappeared from UR after 108 days (Figure 6c).
High H 2 partial pressure has previously been shown to be detrimental to Methanosarcina [36]. Furthermore, the observed accumulation of AA in UR (Figure 3b) is in agreement with inhibition of Methanosarcina.
In contrast to Methanosarcina, the hydrogenotrophic methanogen Methanothermobacter increased in abundance over time and responded positively to H 2 addition. Methanothermobacter are typical hydrogenotrophic methanogens that are commonly found in thermophilic biogas reactors [37]. As shown in Figure  6d, their abundance in UR got higher than the abundance in CR over time, suggesting that they were enriched as a result of H 2 addition. The high abundance of Methanothermobacter found in this study is consistent with previous research that found this genus to be dominant in thermophilic biogas upgrading systems [6, 14,38]. According to [39], Methanothermobacter expand rapidly when H 2 is abundant and are adaptable to different concentrations of dissolved H 2 .
Syntrophaceticus abundance increased rapidly in UR when H 2 -supplementation was initiated but was greatly reduced after day 140 when the 48h feeding regime was introduced (Figure 6e). Syntrophaceticus is a well-known syntrophic acetate-oxidizing (SAO) bacterium that was discovered in a biogas reactor that relied on the energy from acetate oxidation to produce H 2 and CO 2 [15,34]. SAO bacteria, which are syntrophic with hydrogenotrophic methanogens (Methanothermobacter in our case), can be inhibited by short or long-term H 2 addition to their living atmosphere [20,35]. Increased H 2 partial pressure can inhibit SAO from a thermodynamic perspective because syntrophic sustainability is dependent on the H 2 /formate concentration, which is usually kept low by the methanogenic partners [40]. Interestingly, our study revealed that H 2 addition at an H 2 :CO 2 ratio of 2:1 promotes the growth of Syntrophaceticus while increasing the H 2 :CO 2 ratio to 4:1 signi cantly reduces their abundance. In addition, the abundance of Syntrophaceticus of was maximum when the CW ratio was increased from 10 to 20%. Similar to Halocella, f_Hydrogenisporaceae_OTU_28, was also affected by the increased stirring speed, seen as reduced abundance after 64 h in both reactors ( Figure 6f). f_Hydrogenisporaceae_OTU_28, a member of the OPB54 class, have previously been reported to be involved in the fermentation of carbohydrates to produce acetate and H 2 [41].
Our ndings revealed that the H 2 :CO 2 ratio, stirring speed, CM:CW ratio, and feeding frequency all had an effect on in-situ BM, either on overall CH 4 production or on CH 4 production from H 2 and CO 2 conversion. However, it was only the H 2 :CO 2 ratio and stirring speed that strongly affected the microbial community pro le of the reactors.

Conclusions
The current work demonstrates the feasibility of the hybrid biogas upgrading concept and identi es some challenges that must be tackled for future process improvement. When hybrid BM was used instead of in-situ BM, it resulted in a 39% increase in CH 4 yield. Furthermore, maximum H 2 utilization (62%) was observed during hybrid BM. The co-digestion of CM and AC aided in keeping the pH of the reactor below 8.1 during in-situ BM. The addition of H 2 at a H 2 :CO 2 ratio of 2:1 resulted in stable operation of the in-situ reactor system, while at higher ratio VFAs started to accumulate resulting in pH drop. The microbial analysis revealed that Methanothermobacter, a hydrogenotrophic methanogen, dominates both the control and the H 2 reactors, with a higher abundance in the H 2 reactor. The main factors affecting the microbial community composition were H 2 addition and stirrer speed. The ndings of our study may be useful to other researchers or biogas plant operators in developing processes for enhancing BM performance and methane yields.

Inoculum and substrate
Thermophilic inoculum was obtained from two 10L CSTRs digesting cow manure (CM) collected from a cow farm in Ås, Norway. Both reactors were operated at 55 o C and 20 days of hydraulic retention time. The same CM was also used as a model substrate for the present study. To limit pH increment during in-situ BM, the CM was co-digested with acidic cheese obtained from the Food pilot plant at Norwegian University of Life Sciences (NMBU). The cheese was produced only for experimental purposes [42] and discarded once the experiment was completed. The cheese waste (CW) was collected and was stored at 4 o C until further usage. Table 1 lists the characteristics of the inoculum and substrates used in this study.

In-situ BM setup
The setup comprised of two 10L CSTRs (Control reactor, CR, and in-situ upgrading reactor, UR), each with 6L working volume. The temperature of both reactors was maintained at thermophilic condition (55 o C). Three-blade Elephant Ear impeller operated in the down-pumping mode was used for mixing at 80 rpm. Approximately 300 g of substrate (90% CM: 10% CW) were fed into the reactors every 24 hours after the same amount of e uent had been discharged.
Initially, the organic loading rate was kept at 0.83 g VS L -1 d -1 . Starting day 64, H 2 was injected into UR using a stainless-steel Mott sparger with a pore size of 2 µm, which was mounted at the bottom of the reactor. The sparger measured 12 cm in length and had a 12 mm outer diameter. The ow rate of H 2 was initially set to 3 mL min -1 (H 2 :CO 2 ratio = 2:1). To increase the contact time between anaerobic microbes and H 2 , gas recirculation was introduced from day 64. A peristaltic pump was used to recirculate the output gas at gas recirculation rates of 7.63 mL min -1 .

Hybrid BM setup.
A hybrid BM set-up where the in-situ reactor (UR) was combined with ex-situ reactor (HR) was tested at the end of the experiment (day 173 to 203). The CR was not included in this experiment. The ex-situ upgrading reactor was established using a 2 L bottle lled with 800 mL ltered and degassed inoculum (digestate from UR) and 108 g polyethylene packing materials (Hel-X biocarriers, HXF13KLL+, Christian Stöhr GmbH & Co). Once a week, 50 mL of the ltered and pasteurized CM was added to HR (nutrient supply) after the same amount of e uent had been discharged. A peristaltic pump was used to transfer the outlet biogas from the UR to the 2L bottle and inject it at the bottom via a diffuser. Figure 1a & 1b depicts the in-situ and hybrid con gurations.

Sample analysis
Gas chromatography (GC) (SRI 8160C) with a Flame Ionization Detector and N 2 as the carrier gas was used to measure the gas composition (CH 4 , CO 2 , and H 2 ). A standard biogas mixture (64% CH 4 and 36% CO 2 ) and a 10% H 2 gas mixture (with 90% N 2 ) (AGA Norway) were used for GC calibration on a regular basis. A digital pH meter (Thermo Scienti c Orion Dual Star, USA) was used to measure pH of the digestate. pH measurement was performed immediately after the digestate was discharged from the reactors to avoid CO 2 removal from liquid phase.
Digestates from the reactors were collected regularly for total solid (TS), volatile solid (VS), TAN and VFA analysis. TS, VS and TAN were measured according to the Standard Methods for Examination of Water and Wastewater (APHA, 2005). VFA samples were prepared following [22]. VFA concentration was determined using a high performance liquid chromatography (Dionex, Sunnyvale, CA, USA) with Aminex column as described previously [22].

Microbial analysis DNA sampling and extraction
The liquid e uent from each reactor was collected regularly and stored at -80 o C until DNA analysis. DNA extraction and sequencing were performed by DNASense (Aalborg, Denmark). The template DNA was extracted using the FastDNA Spin kit for Soil (MP Biomedicals, USA). The DNA extraction was performed following the manufacturer protocol except that samples were subjected to bead beating at 6 m/s for 4x40s [43]. DNA quantity and quality were assessed using gel electrophoresis with Tapestation 2200 and Genomic DNA screentapes (Agilent, USA). The Qubit dsDNA HS/BR Assay kit was used to determine the concentration of DNA (Thermo Fisher Scienti c, USA).

Sequencing analysis
Microbial community pro les were determined using 16S rRNA gene variable region V4 with primers [515FB] GTGYCAGCMGCCGCGGTAA and [806RB] GGACTACNVGGGTWTCTAAT [44]. The 25 µL PCR reactions contained (12.5μL) PCRBIO Ultra mix, 400 nM primers and up to 10 ng of extracted DNA. The PCR thermal cycling consisted of a hot start step at 95 o C for 2 min, followed by 30 cycles of 95 o C for 15 s, 55 o C for 15 s, 72 o C for 50 s, and then a nal 72 o C extension step for 5 min. For each sample, duplicate PCR reactions were performed, and the duplicates were pooled following PCR. The obtained amplicon libraries were puri ed using the standard protocol for CleanPCR SPRI beads (CleanNA, NL) with a bead to sample ratio of 4:5. The DNA concentration was quanti ed using Qubit dsDNA HS Assay kit (Thermo Fisher Scienti c, USA) and the quality was con rmed by gel electrophoresis using Tapestation 2200 and D1000/High sensitivity D1000 screentapes (Agilent, USA). The puri ed libraries were pooled in equimolar concentrations and spiked with > 10 % PhiX control.
The denatured library was sequenced on a MiSeq (Illumina, USA) using the Miseq Reagent kit V3.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and material
The nucleotide sequence dataset used this study is available in the European Nucleotide Archive (ENA).

Competing interests
The authors declare that they have no competing interests.

Funding
This work was supported by the Research Council of Norway through grants 270038 (NorBioLab) and 257622 (Bio4Fuels).

Authors' contributions
RW and SJH conceived the idea for the study. RW set up and operated the reactors, as well as collected samples and process data. RW was responsible for sample analyses and interpretation of the experimental data. RW wrote the rst draft of the manuscript, and SJH reviewed and edited subsequent drafts. Both authors read and approved the nal manuscript. Tables   Table 1 Characteristics of inoculum and substrates  CR -control reactor, UR -in-situ upgrading reactor TVFA -total volatile fatty acid, AA -acetic acid, PA -propionic acid Table 4 Performance of hybrid reactor system at different H 2 :CO 2 ratios (mean ± S.D)   Methane yield and H2 consumption at different experimental phases (I -VI). CR, control reactor; UR, in-situ upgrading reactor; H2 consumed, H2 consumption.