Thermophilic Anaerobic Digestion of Second Cheese Whey: Microbial Community Response to H 2 Addition in a Partially Immobilized Anaerobic Hybrid Reactor

: In this study, we investigated thermophilic (55 ◦ C) anaerobic digestion (AD) performance and microbial community structure, before and after hydrogen addition, in a novel hybrid gas-stirred tank reactor (GSTR) implemented with a partial immobilization of the microbial community and fed with second cheese whey (SCW). The results showed that H 2 addition led to a 25% increase in the methane production rate and to a decrease of 13% in the CH 4 concentration as compared with the control. The recovery of methane content (56%) was reached by decreasing the H 2 ﬂow rate. The microbial community investigations were performed on efﬂuent (EF) and on interstitial matrix (IM) inside the immobilized area. Before H 2 addition, the Anaerobaculaceae (42%) and Lachnospiraceae (27%) families dominated among bacteria in the efﬂuent, and the Thermodesulfobiaceae (32%) and Lachnospiraceae (30%) families dominated in the interstitial matrix. After H 2 addition, microbial abundance showed an increase in the bacteria and archaea communities in the interstitial matrix. The Thermodesulfobiaceae family (29%)remained dominant in the interstitial matrix, suggesting its crucial role in the immobilized community and the SHA-31 family was enriched in both the efﬂuent (36%) and the interstitial matrix (15%). The predominance of archaea Methanothermobacter thermoautrophicus indicated that CH 4 was produced almost exclusively by the hydrogenotrophic pathway.


Introduction
The dairy industry is one of the main sources of industrial wastewater in Europe with cheese whey (CW) and second cheese whey (SCW) making up a large part [1]. SCW is a byproduct generated from the precipitation of CW proteins by means of heat (80-90 • C) with added organic acids and salts for the production of cottage and ricotta cheeses. Similar to CW, SCW is a highly pollutant dairy waste with a significant organic load (biochemical oxygen demand (BOD) ≈ 30 g/L, chemical oxygen demand (COD) 60-80 g/L, and lactose 40-50 g/L) but with lower levels of fat (0.5-8 g/L) and protein (0.5-8 g/L) and a higher salinity (7-23 mS cm −1 ) [1].
The high organic loads of both SCW and CW represent severe disposal and pollution issues for the dairy industry and a huge opportunity for bioenergy and biochemicals production [2][3][4]. In particular, methane production by anaerobic digestion (AD) can be Processes 2021, 9, 43 2 of 16 an opportunity for small-medium dairies that cannot afford the high investment costs associated with the implementation of CW valorization technology. According to Mainardis et al., biogas dairy plants that use simple low-cost digesters, can provide most of the electricity and heat necessary for the plants, improving energy balance and reducing transport and management costs [5]. However, it is known that the high easily-fermentable organic content and low bicarbonate alkalinity of raw SCW, as well as of CW, unbalance the AD process towards an accumulation of volatile fatty acids (VFAs), and a decrease in the pH values far below the optimum value for the methanogens is often observed [6]. Different solutions have been suggested in order to control acidification and rule out the risk of failure for the AD process such as supplementation of alkalinity during the process or into the feed, use of the codigestion with a substrate having high buffering capacity (e.g., live-stock manure or slurry and sewage sludge), and use of two-stage reactor configuration [7][8][9][10]. Although the codigestion strategy can be a sustainable option for farms, its implementation is not economically and environmentally sustainable for dairies. Furthermore, when digesting CW in combination with pathogenic waste streams, health and safety issues could discourage their use. In Europe, only 7.1% of the milk produced is processed directly on farms, while the vast majority of raw milk is delivered to dairies (Eurostat, 2019). In addition, the two-stage configuration entails a higher investment cost making it economically unattractive [11].
Different one-stage reactor configurations have been used for methane production from CW. Among these, the continuously stirred tank reactor (CSTR) configuration has several advantages such as reasonable control, easy operation and cleaning, minor operating cost, and high removal efficiency [12]. However, a very low conversion per unit volume is obtained. Recently, Faisal et al. showed that the addition of high density polyethylene carriers to batch anaerobic digestion of agricultural wastes supported the formation of biofilm leading to significant increases in substrate utilization and biogas and biomethane production [13]. Similarly, Ramasamy et al. (2000) indicated that the incorporation of plastic media into a CSTR enhanced methane yield from dairy wastewaters more than 20% and without pH control [14].
Although relevant improvement in process performances were reported, no information concerning the impact of the biomass support material on the microbial communities was reported in these studies.
Recently, Treu et al. proposed the H 2 addition in CSTR used for CH 4 production from raw CW, as a strategy to manage the low pH value of the substrate [9]. The in situ H 2 addition, namely biogas in situ upgrading, is a technology generally used to increase the methane content in biogas, i.e., the introduced H 2 is combined with carbon dioxide (CO 2 ) produced in the process, to generate further CH 4 by the hydrogenotrophic methanogens. The diffusion of H 2 in the liquid phase is the limiting factor of this technology, i.e., due to the low H 2 solubility it is not bioavailable for microorganisms [15][16][17][18].
Moreover, the addition of H 2 could unbalance the equilibrium of the system. The additional H 2 represents an extra substrate for the microbial communities of bacteria and archaea involved in the AD process. It could cause a modification of the microbial community populating the biogas reactor. In particular, if hydrogenotrophic methanogens are unable to quickly remove the H 2 , its partial pressure will increase leading to disruption of the syntrophic interaction occurring between H 2 producing microorganisms (heterotrophic community as acidogenic and acetogenic bacteria) and H 2 consuming microorganisms (autothrophic bacteria as homoacetogenic bacteria and archaea as hydrogenotrophic methanogens). This condition can lead to the accumulation of propionic and acetic acids and, consequently, a decrease of pH outside the optimal range for the methanogenic archaea produces adverse effects on the whole process. In contrast, due to the removal of CO 2 , inhibition of methanogenesis could be generated by an increase in pH to values above 8.5 [19].
Little is known about how the H 2 addition could affect the microbial community responsible for the AD process, especially in concomitance with other operating parameters and reactors configurations.
The current study aimed to characterize a thermophilic (55 • C) anaerobic digestion (AD) performance and the microbial community structure, before and after in situ H 2 addition, in a novel hybrid gas-stirred reactor, namely a gas-stirred tank reactor (GSTR), fed with undiluted SCW. The GSTR hybrid reactor is the result of the CSTR modification, in which about one third of the reactor's working volume was filled with polymeric supports (high-density polyethylene, HDPE) and a gas recirculation was installed for homogenization of nutrients and as a strategy to improve H 2 solubility. The thermophilic condition was selected to exploit the acclimatization of the mixed microbial community used as inoculum and collected from a thermophilic (55 • C) anaerobic digestion plant. The microbial community of the effluent and the interstitial matrix inside the immobilized area were characterized by fluorescence in situ hybridization (FISH) and Illumina next-generation sequences (NGS) techniques. The impacts of hydrogen addition on the performance of methane production and on the modification of the microbial community structure were investigated.

Substrate Characteristics
In Italy, 4.9 Mt of cheese whey are produced annually and about 1 Mt is used to produce ricotta, a typical soft cheese of the Mediterranean region, generating a second cheese whey (SCW) called "scotta". Moreover, the manufacturing of ricotta cheese is also widespread in Latin America and in USA, where it is referred to as requeson and ricottone, respectively.
SCW is periodically collected at Formaggi Boccea, a small dairy factory located in Rome, Italy, providing a thermophilic (55 • C) biogas plant fed with SCW. The wastewater was stored at −20 • C and thawed before use. Undiluted SCW was fed into the GSTR reactor. Due to the fluctuations of the residual organic load derived from the production process, slight variations in its composition were observed. The main physical and chemical characteristics of the SCW are presented in Table 1. Table 1. Main characteristics of second cheese whey (SCW) used in this study.

Reactors Setup and Operation
The experiment was carried out in a novel hybrid gas-stirred tank reactor (GSTR) with a working volume of 49 L, at thermophilic temperature (55 ± 1 • C) and atmospheric pressure. Figure 1 shows the scheme of the reactor and the overall equipment design. Reactor mixing and homogenization of nutrients was ensured by continuous gas recirculation. A vacuum pump took the biogas from the reactor headspace and injected it into the bottom of the GSTR. Biofilm media carriers were used as packing material (Scubla MBBR 800, HDPE) with a specific surface area of 800 m 2 /m 3 . The immobilized area was completely submerged in the middle part of the reactor and occupied a volume of 15 L. Media carriers were enclosed in a mesh bag to prevent washing out in the outlet stream. Biogas coming from both gas recirculation and gas outlet lines were dehumidified by a chiller and the con-Processes 2021, 9, 43 4 of 16 densed water was returned into the digester. The GSTR was inoculated with thermophilic (55 • C) sludge obtained from the Formaggi Boccea biogas plant. Undiluted second cheese way (SCW) was fed in a continuously way from the bottom using a peristaltic pump. The hydraulic retention time (HRT) was 15 days, and the organic loading rate (OLR) was on average within a range of 2.18 ± 0.14 g COD L −1 d −1 and 2.4 ± 0.12 g COD L −1 d −1 . 800, HDPE) with a specific surface area of 800m 2 /m 3 . The immobilized area was completely submerged in the middle part of the reactor and occupied a volume of 15L. Media carriers were enclosed in a mesh bag to prevent washing out in the outlet stream. Biogas coming from both gas recirculation and gas outlet lines were dehumidified by a chiller and the condensed water was returned into the digester. The GSTR was inoculated with thermophilic (55 °C) sludge obtained from the Formaggi Boccea biogas plant. Undiluted second cheese way (SCW) was fed in a continuously way from the bottom using a peristaltic pump. The hydraulic retention time (HRT) was 15 days, and the organic loading rate (OLR) was on average within a range of 2.18 ± 0.14 g COD L −1 d −1 and 2.4 ± 0.12 g COD L −1 d −1 . Initially, a biogas recirculation flow rate of 59 L Lr −1 d −1 was applied. When a stable AD process was achieved, pure H2 (≥99%) obtained from an electrolyzer (DBS, model PG-H2 100), was continuously injected into the reactor from the bottom by the biogas recirculation line. In order to avoid an increase in hydrogen partial pressure, the stoichiometric ratio H2/CO2 used in this experimental phase was lower (2.7:1 UP1 and 2:1 for UP2, UP for upgrading phase) than needed for hydrogenotrophic methanogenes (4:1). The starting H2 flow rate was 1.8 LH2 Lr −1 d −1 and gas recirculation flow rate was maintained at 59 L Lr −1 d −1 .This experimental phase is denoted as Upgrading 1 (UP1). Thereafter, the H2 flow rate was decreased to 1.32 LH2 Lr −1 d −1 and gas recirculation flow was increased to 118 L Lr −1 d −1 . This experimental phase is indicated as UP2. Both UP1 and UP2 lasted for 30 days.

Analytical Methods
H2, CH4, CO2, N2, and O2 percentages in biogas were analyzed by online Micro Gas Chromatograph Varian (GC4900). Two columns of 10 meters were used. The first, MS 5A, had a stationary phase with molecular sieves capable of separating the permanent gases of low molecular weight, H2, CH4, N2, and O2. The second column was a Poraplot U, used as a stationary phase divinyl benzene able to separate CO2. Argon was used as the carrier gas. The biogas and H2 flow rates were monitored by digital online flow meters (EL-Flow select series, Bronkhorst High-Tech B.V, Ruurlo Netherland).The flows of the different Initially, a biogas recirculation flow rate of 59 L L r −1 d −1 was applied. When a stable AD process was achieved, pure H 2 (≥99%) obtained from an electrolyzer (DBS, model PG-H2 100), was continuously injected into the reactor from the bottom by the biogas recirculation line. In order to avoid an increase in hydrogen partial pressure, the stoichiometric ratio H 2 /CO 2 used in this experimental phase was lower (2.7:1 UP1 and 2:1 for UP2, UP for upgrading phase) than needed for hydrogenotrophic methanogenes (4:1). The starting H 2 flow rate was 1.8 L H 2 L r −1 d −1 and gas recirculation flow rate was maintained at 59 L L r −1 d −1 .This experimental phase is denoted as Upgrading 1 (UP1). Thereafter, the H 2 flow rate was decreased to 1.32 L H 2 L r −1 d −1 and gas recirculation flow was increased to 118 L L r −1 d −1 . This experimental phase is indicated as UP2. Both UP1 and UP2 lasted for 30 days.

Analytical Methods
H 2 , CH 4 , CO 2 , N 2 , and O 2 percentages in biogas were analyzed by online Micro Gas Chromatograph Varian (GC4900). Two columns of 10 m were used. The first, MS 5A, had a stationary phase with molecular sieves capable of separating the permanent gases of low molecular weight, H 2 , CH 4 , N 2 , and O 2 . The second column was a Poraplot U, used as a stationary phase divinyl benzene able to separate CO 2 . Argon was used as the carrier gas. The biogas and H 2 flow rates were monitored by digital online flow meters (EL-Flow select series, Bronkhorst High-Tech B.V, Ruurlo Netherland).The flows of the different gases were calculated on the basis of the percentage compositions of the individual gas present in the biogas. Volatile fatty acids (VFAs), lactic acid, alcohols, and sugars were analyzed by a HPLC Thermo Spectrasystem, equipped with a UV detector (λ = 210 nm) and a refraction index detector, using the isocratic method of analysis at 75 • C with Column Rezex ROA-Organic Acid H + (8%), size 300 × 7.8 mm Phenomenex, USA.

Calculation
Performance and efficiency of the methanation process were expressed as the methane evolution rate (MER), i.e., the rate of CH 4 production from the injected H 2 , and H 2 conversion efficiency (η H 2 ). The MER (Equation (1) expresses the increase in the specific CH 4 production rate (L CH 4 L r −1 d −1 ) caused by H 2 injection as compared with the CH 4 production rate in the control condition. It is calculated as follows (Equation (1)): where UPs are the upgrading periods (with H 2 addition) and AD is the control period (without H 2 ) The H 2 gas-liquid mass transfer rate, namely r t (L H 2 L r −1 d −1 ), and the efficiency of H 2 utilization, namely η H 2 (%), were calculated according to Equations (2) and (3), respectively: where H 2 in flow rate and H 2 in output gas are the volumetric H 2 flows entering and leaving the reactor. It was assumed that all the H 2 transferred to the liquid phase was ultimately converted to CH 4 or used for other microbial metabolic pathways or employed for microbial growth [15,17]. The H 2 rate converted to biomethane (L L r −1 d −1 ) was calculated according to Equation (4): rate to biomethane = 4 × (CH 4 in UP output gas − CH 4 in AD output gas) (4) where 4 is the stoichiometric coefficient according to Equation (4) H 2 + CO 2 = CH 4 , CH 4 in UP output gas (L CH 4 L r −1 d −1 ) is the rate of CH 4 produced in UP experimental phases, and CH 4 in AD output gas (L CH 4 L r −1 d −1 ) is the rate of CH 4 produced in AD phase.

Sample Collection for the Microbial Community Analysis
Samples for the microbial community analysis were collected at the end of AD and UP phases from both the liquid medium and the interstitial matrix between the HDPE supports. After the AD phase, the GSTR was opened to collect the interstitial matrix; its dark-grey color and the presence of flocs indicated the formation of a biofilm. Before the UP phases, the GSTR was flushed with nitrogen to restore the anaerobic condition and was operated under the same previous operating conditions until N 2 was no longer detected. Samples were collected in duplicate, aliquoted (10 mL) and stored at −20 • C for different uses (i.e., DNA extraction and fluorescence in situ hybridization (FISH) analysis). They are recognized as EF-AD or EF-UP2 (AD, anaerobic digestion phase; UP2, H 2 upgrading at the end of UP2 phase; EF, effluent,) and IM-AD or IM-UP2, (IM, interstitial matrix).

Illumina Sequencing
Genomic DNA was extracted usinga Gene MATRIX Bacterial and Yeast Genomic DNA Purification Kit (EURxLtd. Gdansk Poland), in accordance with the manufacturer's recommendations. The quantity and quality assessment of the extracted DNA were performed using spectrophotometry (Eppendorf BioPhotometer plus). All DNA samples were stored at −20 • C until use.
Library preparation was performed utilizing a 460bp amplicon corresponding to the hypervariable V3-V4 region of 16S rRNA using universal primers (S-D-Bact-0341F and S-D-Bact-0785R for bacteria and archaea domains). Sequencing was performed at the Department of Agricultural Sciences, University of Naples Federico II (Portici, Italy) using the Illumina MiSeq platform. From the generated reads, demultiplexing, quality filtering, trimming, merging, and operative taxonomical units (OTUs) picking were performed using QIIME pipeline [20].
The taxonomical assignment was performed with the Greengenes database at a 97% similarity. The bioinformatics analysis of all sequences generated more than 220,000 reads, with an average of 26,800 counts per sample. The results reported are focused on the most abundant members of the microbial community having a relative abundance ≥ 0.5%. Raw sequences data were deposited at the Sequence Read Archive (SRA) under the BioProject PRJNA681387 with the accession numbers SAMN1695146 and SAMN16951461. Shannon-Weaver (H) and Pielou's evenness (E) indices were calculated using the relative abundance of sequences obtained from NGS sequencing at the family level.

Total Microbial Abundance and Bacteria and Archaea Detection
DAPI fluorescent staining and fluorescence in situ hybridization (FISH) were used to evaluate the microbial abundance and bacteria and archaea relative abundance, respectively. The collected samples were fixed (4% w/v paraformaldehyde fixative solution for 6 h, at 4 • C), as described in Amann, 1990, and stored (−20 • C) [21] until use. In order to detach microbial cells from inorganic particles, a cleaning sample procedure was performed immediately after thawing and before analysis, using the density gradient medium OptiPrep™ (Axis-Shield PoCAs, Oslo, Norway) [22]. Total microbial abundance (cells mL −1 ) was determined in the effluent and interstitial matrix by direct cell counting after DAPI staining (4 ,6-diamidino-2-phenylindole) (1 µg mL −1 ) [22] using the epifluorescence microscopy AXIOSKOP 40 (Carl Zeiss, Germany) equipped with a ZEISS HXP 120v Light Source and 1000× magnification.
The FISH analysis (in triplicate) was carry out using samples collected from effluent and interstitial matrix focusing on the relative abundance (%) of archaea and bacteria. Analyses were performed as described in Amann et al., 1995 [23] and oligonucleotide probes for the Bacteria (EUB338 II,III) and Archaea (ARC915) domains were used [24].
The oligonucleotide probes (50 ng µL −1 ), labeled with carboxyfluoresce in (FAM) or indocarbocianine (Cy3) dyed at the 5 -end, were purchased from MWG AG Biotech, Germany. All the hybridizations were carried out in combination with DAPI staining to estimate the proportion of cells targeted by the specific probes out of the total cells. Slides were mounted with a few drops of Vecta Shield (Vector Laboratories, USA) and seen using Zeiss epifluorescence microscope. The average of cells for each target probe (15 fields for each slide) was used to evaluate the percentage of the positive signals versus DAPI-positive cells. Microbial abundance of bacteria and archaea in the samples (cells mL −1 ) was calculated using the relative abundances obtained by FISH technique and the microbial abundance detected after DAPI staining.

Reactors Performance
The daily course profiles of process performances of all experimental phases are shown in Figure 2 and the steady operation condition process performance data are summarized in Table 2.
During the AD period, starting from the 10th day to the end of this phase, a stable methane production rate of 0.79 ± 0.04 L CH 4 L r −1 d −1 was reached. The pH of the reactor was around 7, indicating a strong buffer capacity of the system, although no pH control was applied. The average CH 4 content in the biogas was 55.5 ± 0.8% and the average specific methane yield (SMY) was 0.329 ± 0.02 L CH 4 g COD −1 , which corresponded to 94% of the maximum theoretical value obtainable by the conversion of organic matter into methane (i.e., 0.350 L CH 4 g COD −1 ) [9,19]. The substrate was almost completely consumed, as confirmed by the low average value of lactose (270 ± 28 mg L −1 ) detected in the reactor effluent, indicating an efficient performance of the AD process. The acetic acid was the unique VFA found during the AD process, moving from an average value of 210 ± 5.6 mg L −1 during the first 10 days to that of 37 ± 2.4 mg L −1 at the end of AD. Moreover, in order to verify the effect of biogas recirculation on the soluble metabolite distribution along the GSTR, an HPLC analysis was also performed on the lower and upper liquid fractions. The average lactose and acetic acid concentrations of 270 ± 28 mg L −1 and 45 ± 10 mg L −1 , respectively, were measured indicating a good mixing of the reactants.  During the AD period, starting from the 10th day to the end of this phase, a stable methane production rate of 0.79 ± 0.04 LCH4 Lr −1 d −1 was reached. The pH of the reactor was around 7, indicating a strong buffer capacity of the system, although no pH control was applied. The average CH4 content in the biogas was 55.5 ± 0.8% and the average specific methane yield (SMY) was 0.329 ± 0.02 LCH4 g COD −1 , which corresponded to 94% of the maximum theoretical value obtainable by the conversion of organic matter into methane (i.e., 0.350 LCH4 g COD −1 ) [9,19]. The substrate was almost completely consumed, as confirmed by the low average value of lactose (270 ± 28 mg L −1 ) detected in the reactor effluent, indicating an efficient performance of the AD process. The acetic acid was the unique VFA  This result is among the highest SMY obtained from AD of synthetic or raw dairy wastewaters in hybrid anaerobic biofilm reactors (0.03-0.359 L CH 4 gCOD −1 ) [12,25]. However, these studies were conducted at mesophilic temperature. So far, there have been few studies on single-stage thermophilic anaerobic digestion using only dairy wastewaters [4]. Yang et al. achieved a maximum yield of 0.360 L CH 4 g COD −1 using a 5 L CSTR reactor under an HRT value of 7.5 days and using diluted CW (10 g COD L −1 ) [26]. Other authors have recorded an unstable AD process with poor thermophilic (55 • C) CH 4 production yield (0.100 L CH 4 g COD −1 ± 0.21) and a remarkable accumulation of acetic acid 10 ± 1 g L −1 using a 3 L CSTR configuration with an OLR of 2.4 g CODL −1 d −1 and an HRT of 15 days using cheese whey permeate and cheese waste powered as substrates [27]. Fernandez et al., obtained a SMY of 0.315 ± 6.6 L CH 4 g COD −1 during a thermophilic AD of deproteinized CW in a 25 L anaerobic sequencing batch reactor (ASBR) at an HRT of 8.3 days and an OLR of 4.6 ± 0.3 g COD L −1 d −1 [28]. Although different HRT and OLR values were used, the yields obtained were comparable to ours, suggesting that the GSTR, used in this study, maintained the characteristics of a CSTR for the homogenization of nutrients in the liquid phase. However, the implementation of supports for the immobilization of the biomass makes it comparable to an ASBR whose peculiar feature is to uncouple the SRT (solid retention time) from the HRT.
The UP1 started at day 30 and, after 12 days, a stable methane production process was reached and lasted until the end of the experimental phase (60 days). During the steady state of UP1, the CH 4 production rate was 0.99 ± 0.06 L CH 4 L r −1 d −1 , which represented an increase of 25% as compared with the AD phase. The calculation of the efficiency of H 2 utilization (η H 2 ), suggests that 54% of H 2 added is utilized by the microbial community.
However, the incomplete H 2 consumption led to a dilution of CH 4 concentration as compared with that obtained in AD, indeed the composition of the outflow biogas in the UP1 phase was 48.2% CH 4 , 39.8 H 2 , and 11.6% CO 2 . Considering the high unconverted percentage of H 2 observed in UP1, it was assumed that the H 2 injection rate of 1.8 L H 2 L r −1 d −1 used in UP1 was too high. Therefore, the H 2 inlet flow was decreased to 1.32 L H 2 L r −1 d −1 in the UP2 phase. Moreover, in order to increase the H 2 dissolution in the liquid phase, the gas recirculation rate was increased from 59 to 118 L H 2 L r −1 d −1 . During the steady state of UP2 (reached after day 10) the CH 4 flow rate was equal to 0.92 ± 0.08 L CH 4 L r −1 d −1 and the percentage of CH 4 in the biogas went up to 55%. The pH increased, reaching a value of 7.3-7.4. The CH 4 production rate was 17% higher as compared with AD and no accumulation of acetic acid (38 ± 5 mg L −1 ) or lactose (208 ± 10.5 mg L −1 ) were observed. The H 2 efficiency value increased to 65%, but the MER value (0.13 L CH 4 L r −1 d −1 ) was lower than the one calculated in UP2 (0.20 L CH 4 L r −1 d −1 ). According to the stoichiometric reaction 4H 2 + CO 2 = CH 4 , MER values of 0.21 and 0.24 L CH 4 L r −1 d −1 were expected, respectively. These differences could be partly explained by the use, in UP1 and UP2, of OLRs slightly lower than the one used in AD, involving in an overestimation of the CH 4 formation rate from SCW. In addition, taking into account that the amount of acetate as H 2 equivalent can be neglected, two further metabolic pathways of the H 2 added other than the production of CH 4 took place inside the GSTR reactor. For example, during the methanation process, a distinct part of both H 2 and CO 2 are metabolized to produce biomass by the archaea community [29]. The results of the FISH analysis seem to confirm this hypothesis; the Archaea abundance was more than doubled after the H 2 dispersion phase in EF-UP2 and it increased in IM-UP2. However, sulphate-reducing bacteria also utilize H 2 as a substrate to reduce sulphate to hydrogen sulphide. In this study, the hydrogen sulphide content of biogas was not monitored.
Briefly, although the purpose of this study was not to improve the efficiency of the in situ biomethanation process and, although higher CH 4 concentration values were achieved in a published paper, the above results demonstrated that it was possible to convert H 2 to CH 4 in this novel GSTR with partially immobilized biomass by using HDPE supports.

Total Microbial Abundance and Bacteria and Archaea Detection
Total microbial abundance (DAPI staining), as well as abundance in bacteria and archaea (FISH technique associated with DAPI staining) under different experimental conditions are shown in Figure 3.

Total Microbial Abundance and Bacteria and Archaea Detection
Total microbial abundance (DAPI staining), as well as abundance in bacteria and archaea (FISH technique associated with DAPI staining) under different experimental conditions are shown in Figure 3. (a) Total microbial abundance detected in the effluent (EF, gray bars) and in the interstitial matrix (IM, blue bars), before (AD, anaerobic digestion) and after (UP2) H2 addition, using DAPI staining; (b) Abundance of bacteria (green) and archaea (red) in the effluent (EF) and interstitial matrix (IM), before and after H2 addition (AD and UP2, respectively) obtained by FISH technique and DAPI staining. At the bottom, the archaea are shown on a lower scale to highlight the differences.
The highest and the lowest total microbial abundances were detected, respectively, in IM-UP2 with 5.53 x 10 8 ± 1.25 x 10 7 cells mL −1 and in EF-AD with 3.7 x 10 8 ± 6.24 x 10 6 cells mL −1 (Figure 3a). In both effluent (EF) and interstitial matrix (IM), the total microbial abundance increased after H2 addition by 22% and 9%, respectively. Both IM-AD and IM-UP2 had a total microbial abundance higher than the corresponding effluent phase (EF-AD and EF-UP2) by 37% and 22%, respectively, highlighting the positive effect that the presence of immobilizing supports exerted on the microbial community. The FISH investigations showed a strong predominance of bacteria in the microbial communities of all experimental conditions (Figure 3b), while archaea contribution to the microbial communities, ranged between 2.2% and 5.2%, in line with the concentration usually detected in microbial communities during an AD process [30]. More specifically, before H2 addition the bacterial cells detected were 89% and 91% in EF-AD and IM-AD, respectively, while at the end of UP2 phase their concentration decreased to 80% and 84%, respectively. The H2 addition positively affected archaea populations, especially in EF-UP2 effluent samples, for which a concentration of 5.2% corresponding to 2.23 x 10 7 cells mL −1 of archaea was detected, with an increase of 134% as compared with EF-AD which showed a (a) Total microbial abundance detected in the effluent (EF, gray bars) and in the interstitial matrix (IM, blue bars), before (AD, anaerobic digestion) and after (UP2) H 2 addition, using DAPI staining; (b) Abundance of bacteria (green) and archaea (red) in the effluent (EF) and interstitial matrix (IM), before and after H 2 addition (AD and UP2, respectively) obtained by FISH technique and DAPI staining. At the bottom, the archaea are shown on a lower scale to highlight the differences.
The highest and the lowest total microbial abundances were detected, respectively, in IM-UP2 with 5.53 × 10 8 ± 1.25 × 10 7 cells mL −1 and in EF-AD with 3.7 × 10 8 ± 6.24 × 10 6 cells mL −1 (Figure 3a). In both effluent (EF) and interstitial matrix (IM), the total microbial abundance increased after H 2 addition by 22% and 9%, respectively. Both IM-AD and IM-UP2 had a total microbial abundance higher than the corresponding effluent phase (EF-AD and EF-UP2) by 37% and 22%, respectively, highlighting the positive effect that the presence of immobilizing supports exerted on the microbial community. The FISH investigations showed a strong predominance of bacteria in the microbial communities of all experimental conditions (Figure 3b), while archaea contribution to the microbial communities, ranged between 2.2% and 5.2%, in line with the concentration usually detected in microbial communities during an AD process [30]. More specifically, before H 2 addition the bacterial cells detected were 89% and 91% in EF-AD and IM-AD, respectively, while at the end of UP2 phase their concentration decreased to 80% and 84%, respectively. The H 2 addition positively affected archaea populations, especially in EF-UP2 effluent samples, for which a concentration of 5.2% corresponding to 2.23 x 10 7 cells mL −1 of archaea was detected, with an increase of 134% as compared with EF-AD which showed a concentration of 0.83 x 10 7 cells mL −1 (Figure 3b). The percentage of archaea in the interstitial matrix, before and after H 2 addition, was 3.1% and 3.7%, respectively, corresponding to a 43% increase in the interstitial matrix after the H 2 addition. Moreover, the presence of HDPE supports generated a favorable habitat for the microbial community, as evidenced both by the total microbial ( Figure 3a) and bacteria abundances (Figure 3b) in both experimental phases. After the H 2 addition, an increase in total microbial (Figure 3a) and archaea abundances (Figure 3b) in the interstitial matrix was observed. It can be hypothesized that these supports, promote the contact of hydrogen with microbial cells and also provide a habitat refuge in which the disturbance of the spatial disposition of AD populations with different cell leaching have been limited. Other authors [31] observed an increase in the archaea component when a porous support was offered to a mesophilic anaerobic microbial community as a refuge habitat, although the present study is among the first to refer to a thermophilic condition.

Illumina Sequencing
The microbial communities were analyzed by Illumina-based 16S sequencing. The representativeness (%) of the microbial communities' composition at the different taxonomic levels tended to decrease from the phyla (96%) toward the species (25%) levels, showing 85% (families) and 50% (genera) of representativeness. The low assignment at genera and species levels suggested the presence, in the microbiome, of numerous unexplored or undescribed taxa.

Bacteria Communities
The relative abundance of Bacteria at the phylum level is shown in Figure 4. chaea abundances (Figure 3b) in the interstitial matrix was observed. It can be hypothesized that these supports, promote the contact of hydrogen with microbial cells and also provide a habitat refuge in which the disturbance of the spatial disposition of AD populations with different cell leaching have been limited. Other authors [31] observed an increase in the archaea component when a porous support was offered to a mesophilic anaerobic microbial community as a refuge habitat, although the present study is among the first to refer to a thermophilic condition.

Illumina Sequencing
The microbial communities were analyzed by Illumina-based 16S sequencing. The representativeness (%) of the microbial communities' composition at the different taxonomic levels tended to decrease from the phyla (96%) toward the species (25%) levels, showing 85% (families) and 50% (genera) of representativeness. The low assignment at genera and species levels suggested the presence, in the microbiome, of numerous unexplored or undescribed taxa.

Bacteria Communities
The relative abundance of Bacteria at the phylum level is shown in Figure 4. The microbial communities during the AD phase were dominated by Synergistetes (43% and 30% in EF and IM, respectively) and Firmicutes (42% and 53% in EF and IM, respectively). Representatives of Thermotogae and Bacteriodetes phyla were also abundant in microbial communities of the AD phase, representing 9% and 4% in the EF sample and 6% and 5% in the IM sample, respectively. Minor representative phyla in the AD phase were identified only in the IM sample, i.e.,Planctomycetes (4%), Chloroflexi (1%), and OP9 The microbial communities during the AD phase were dominated by Synergistetes (43% and 30% in EF and IM, respectively) and Firmicutes (42% and 53% in EF and IM, respectively). Representatives of Thermotogae and Bacteriodetes phyla were also abundant in microbial communities of the AD phase, representing 9% and 4% in the EF sample and 6% and 5% in the IM sample, respectively. Minor representative phyla in the AD phase were identified only in the IM sample, i.e., Planctomycetes (4%), Chloroflexi (1%), and OP9 (1%). After H 2 injection, relevant changes in the microbial community structures were observed both for effluent and interstitial matrix samples. The Synergistetes and Firmicutes phyla still represented a considerable fraction of UP2 communities, although their relative abundance decreased considerably as compared with the AD samples. In particular, in the EF-UP2 samples, Synergistetes represented 25% of the bacterial community, while Firmicutes represented 19%. A different trend was observed in the IM-UP2 sample, in which the abundance of Synergistetes further decreased to 17%, while the Firmicutes accounted for 51% of the whole community. The high increase in the relative abundance of Chloroflexi phylum was the most relevant observed change affecting the microbial community structure during the UP2 phase. Representatives of this phylum accounted for 37% and 16% of the bacterial community in EF-UP2 and IM-UP2 samples, respectively. Moreover, an increase in the relative abundances of Planctomycetes (9% and 6% in EF and IM, respectively) and OP9 (5% and 5% in EF and IM, respectively) was also observed. In contrast, the relative abundances of Thermotogae and Bacteroidetes decreased, representing 3% and 1.5% in EF-UP2 and 0.5% and 1% in IM-UP2 samples, respectively.
The detailed results of the bacterial communities at the family level are shown in Figure 5.
Firmicutes represented 19%. A different trend was observed in the IM-UP2 sample, in which the abundance of Synergistetes further decreased to 17%, while the Firmicutes accounted for 51% of the whole community. The high increase in the relative abundance of Chloroflexi phylum was the most relevant observed change affecting the microbial community structure during the UP2 phase. Representatives of this phylum accounted for 37% and 16% of the bacterial community in EF-UP2 and IM-UP2 samples, respectively. Moreover, an increase in the relative abundances of Planctomycetes (9% and 6% in EF and IM, respectively) and OP9 (5% and 5% in EF and IM, respectively) was also observed. In contrast, the relative abundances of Thermotogae and Bacteroidetes decreased, representing 3% and 1.5% in EF-UP2 and 0.5% and 1% in IM-UP2 samples, respectively.
The detailed results of the bacterial communities at the family level are shown in Figure 5.    Anaerobaculaceae was the unique family within Synergistetes detected in all samples, regardless of the experimental phase or the sampling point, thus, suggesting its functional role in the anaerobic digestion process of SCW. This family was characterized at the species level as A. hydrogeniformans, a moderately thermophilic NaCl requiring fermentative bacterium, attesting that it was suitable for the saline and the milk derived substrates. The same microorganism was identified by Treu et al. (2019) [9] during thermophilic anaerobic digestion of cheese whey. Among the Firmicutes phylum, the highest number of families was identified. In particular, members of the Lachnospiraceae family were dominant in the EF-AD sample (30%) and had a lower relative abundance in the IM-AD (5%) sample. This family was not detected in the UP2 samples. Moreover, members of the Thermoanaerobacteraceae family were detected with a higher relative abundance in the IM-AD sample (8%) than in the EF-AD (2%) sample. These microorganisms were stable in EF-UP2 sample (2%) but showed a relevant decrease in the IM-UP2 sample (3%) as compared with the IM-AD sample. The genus Thermacetogenium was identified as a unique genus detected in all samples; it is characterized as an acetate-oxidizing syntrophic microorganism [32]. Members of the Thermodesulfobiaceae family were dominant in the interstitial matrix of both the AD and UP2 phases, accounting for 32% and 29% of the microbial community, respectively. Previous studies have highlighted how these families carried out an acidogenic hydrolytic activity [33,34]. Coprothermobacter was identified among the Thermodesulfobiaceae family as the dominant genus with a relative abundance of 32% and 29% in samples IM-AD and IM-UP2, respectively. It is worth noting that its presence in anaerobic digesters is often related to configurations using biofilm supports [35].
Coprothermobacter is a bacterial genus that includes anaerobic thermophilic members that are proteolytic and produce acetate, H 2 , and CO 2 . Moreover, its involvement in the syntrophic acetate oxidization (SAO) pathway during thermophilic AD processes has been previously suggested [36,37]. A new phylogenetic affiliation has been proposed for this taxa according to its overall phenotypic properties as well as to a phylogenetic analysis supporting its placement in a distinct deeply rooted novel phylum [38]. The SHA-31 family of the class of Anaerolineae was revealed among the Chloroflexi phylum as the unique member characteristic of the UP2 experimental phase. Relative abundances corresponding to 36% and 15% of the microbial community were found in EF and IM samples of UP2 phase, respectively. Several authors [39][40][41] considered members of the Anaerolineae family to be a "semi-syntrophic" microorganisms in anaerobic systems due to their involvement in heterothrophic carbohydrate degradation and in interspecies electron transfer mechanism in mutualistic cooperation with methanogens.
According to both Shannon-Weaver (H') and Pielou's evenness (E) indices, the IM-UP2 sample showed the highest diversity (H' = 2.38 and E = 0.75), suggesting that the H 2 injection had a positive impact in the presence of HDPEs supports. Moreover, a comparison of indices between EF-AD (H' = 1.62 and E = 0.55) and EF-UP2 (H' = 2.06 and E = 0.65) highlighted that microbial diversity was increased also in the effluent by the hydrogen addition. At the same time, the higher diversity in the IM-AD sample (H' = 2.01 and E = 0.68) as compared with the EF-AD sample (H' = 1.62 and E = 0.55) indicated that the immobilization strategy was also able to increase the microbial diversity.

Archaea Communities
Archaea microbial communities were represented by the unique phylum of Euryarchaeota. The overall relative abundances of all methanogens were very low during the AD phase, ranging from 0.1% in EF-AD to 0.40% in the IM-AD. Subsequently, during H 2 addition, the methanogenic population increased up to 0.64% in the effluent and up to 1.42% in the interstitial matrix. Figure 6 shows the relative abundances of archaea at the family level.
Members of Methanobacteriaceae family dominated in both experimental phases, as well as in effluent and interstitial matrix, showing values of relative abundances higher than 74%. Among this family, the hydrogenotrophic Methanothermobacter thermoautrophicus was found to be highly abundant in the AD and UP2 phases. The dominance of the hydrogenotrophic pathway during anaerobic digestion of cheese whey was reported by Treu who suggested that this pathway was driven by the saline characteristics of a substrate [9]. Moreover, members of the Methanosarcina family were detected in all samples except for the EF-AD sample. In particular, the relative abundance in EF-AD was 16%, increasing in EF-UP2 (25%) and lowering in IM-UP2 (11%). Methanosarcina are known as generalist methanogenic archaea that are able to use hydrogen/dioxide carbon, acetic acid, and methylamines [42] as substrates. Therefore, addition of an extra substrate, namely hydrogen, caused the enrichment of the Methanosarcina community. Processes 2021, 9, x FOR PEER REVIEW 13 of 16 Figure 6. Relative abundances of the archaea community at the family level at the end of AD (anaerobic digestion) and UP2 (H2 injection) phases. Communities with a relative abundance ≥0.5% (in at least one sample) are reported. EF, effluent and IM, interstitial matrix.
Members of Methanobacteriaceae family dominated in both experimental phases, as well as in effluent and interstitial matrix, showing values of relative abundances higher than 74%. Among this family, the hydrogenotrophic Methanothermobacter thermoautrophicus was found to be highly abundant in the AD and UP2 phases. The dominance of the hydrogenotrophic pathway during anaerobic digestion of cheese whey was reported by Treu who suggested that this pathway was driven by the saline characteristics of a substrate [9]. Moreover, members of the Methanosarcina family were detected in all samples except for the EF-AD sample. In particular, the relative abundance in EF-AD was 16%, increasing in EF-UP2 (25%) and lowering in IM-UP2 (11%). Methanosarcina are known as generalist methanogenic archaea that are able to use hydrogen/dioxide carbon, acetic acid, and methylamines [42] as substrates. Therefore, addition of an extra substrate, namely hydrogen, caused the enrichment of the Methanosarcina community.

Conclusions
The partially immobilized gas-stirred tank reactor (GSTR) used in this study was suitable for thermophilic anaerobic digestion (AD) of undiluted second cheese whey (SCW). A methane yield of 0.329 ± 0.17 LCH4 gCOD −1 , corresponding to 94% of the maximum theoretical value, was obtained without pH control. Moreover, H2 injection (UP1 phase) increased the CH4 production rate by 25% as compared with that obtained during conventional AD (AD phase) without compromising the efficiency of organic matter removal.
Data on microbial abundance showed an increase in bacteria communities at the end of both the AD and UP2 experimental phases in the interstitial matrix. Moreover, the H2 addition (UP2) increased the archaea communities in both effluent and interstitial matrix as compared with the AD phase. From a functional point of view, we identified a common microbial "core" in all samples represented by members of the Anaerobaculaceae family, strictly correlated to the SCW metabolism. In addition, in samples collected from the interstitial matrix at the end of both experimental phases, a new microbial core was identified in members of the Thermodesulfobiaceae family. Finally, H2 addition caused the enrichment of a peculiar microbial core represented by members of the SHA-31 family. A common core of the methanogenic microbial community was identified in the hydrogenotrophic members of the Methanobacteriaceae family in AD and UP experimental phases.

Conclusions
The partially immobilized gas-stirred tank reactor (GSTR) used in this study was suitable for thermophilic anaerobic digestion (AD) of undiluted second cheese whey (SCW). A methane yield of 0.329 ± 0.17 L CH 4 gCOD −1 , corresponding to 94% of the maximum theoretical value, was obtained without pH control. Moreover, H 2 injection (UP1 phase) increased the CH 4 production rate by 25% as compared with that obtained during conventional AD (AD phase) without compromising the efficiency of organic matter removal.
Data on microbial abundance showed an increase in bacteria communities at the end of both the AD and UP2 experimental phases in the interstitial matrix. Moreover, the H 2 addition (UP2) increased the archaea communities in both effluent and interstitial matrix as compared with the AD phase. From a functional point of view, we identified a common microbial "core" in all samples represented by members of the Anaerobaculaceae family, strictly correlated to the SCW metabolism. In addition, in samples collected from the interstitial matrix at the end of both experimental phases, a new microbial core was identified in members of the Thermodesulfobiaceae family. Finally, H 2 addition caused the enrichment of a peculiar microbial core represented by members of the SHA-31 family. A common core of the methanogenic microbial community was identified in the hydrogenotrophic members of the Methanobacteriaceae family in AD and UP experimental phases. Moreover, the versatile members of Methanosarcinaceae were enriched by the H 2 addition. All these functional microbial cores obtained a balanced AD process in both experimental phases, as shown by results of the GSTR performances at the steady state. Therefore, the whole microbial communities were able to adapt to H 2 , as well as to the HDPE supports, thus, maintaining all the metabolic syntrophic relationships over the experimental time.