Granular anaerobic sludge was obtained from a wastewater treatment plant treating food industry effluent (Delft, The Netherlands). The sludge was washed, and about 2 g of volatile suspended solids (VSS) were used to inoculate vials containing 50 mL of basal medium. Medium was prepared according to the protocol previously described by Stams et al. (1993), with the following exceptions: no NaHCO₃ was added and buffering of the medium was done with 20 mM HEPES [4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid] (pH was corrected to 7.0); titanium citrate (0.2–0.3 mM) was amended as reducing agent instead of Na₂S (to avoid metal precipitation). HEPES was chosen as buffer because it does not react with metal ions (Ferreira et al., 2015). The effect of metals on aceticlastic and hydrogenotrophic activities was tested in separate assays. Aceticlastic activity was measured in the presence of 20 mM of sodium acetate and using N₂ as the headspace gas (1.5 atm). Hydrogenotrophic activity was measured using as sole substrate 1.5 atm of H₂/CO₂ (80:20; % v/v). Metals, sulfate and sulfide were added to assay vials from anaerobic stock solutions to the final desired concentrations, as detailed in the next section. All the materials (bottles, rubber stoppers) were previously washed with 3 M nitric acid to avoid any contamination with other metals. All solutions were prepared with ultrapure water. The assays were performed in the dark at 30°C and stirred at 100 rpm.

Effect of Ni and Co was assayed at low (supplementation test, Ni-L and Co-L) and high (toxicity test, Ni-H and Co-H) concentrations. Final concentrations of NiCl₂ were 2, 4, and 8 μM (Ni-L) and 2, 4, and 8 mM (Ni-H). For Co tests, CoCl₂ was added to final concentrations of 5.5, 10.5, and 25.5 μM (Co-L) and 2.5, 5, 10, 20, and 30 mM (Co-H). These final concentrations are corrected for the Ni and Co already present in the basal medium. Controls without metal addition were included. Sulfate, as Na₂SO₄, was added to the bottles with high metal concentrations; different metal to sulfate ratios were tested (i.e., 0.5:1; 1:1; 1.5:1, mol:mol). Moreover, bottles with Na₂S plus metals were also prepared (ratios 0.5:1; 1:1; 1.5:1, mol:mol). pH of Na₂S stock solution was corrected to 7.0. All conditions were tested in duplicate. A table with all the different conditions and codes for all the assays is included in Supplementary Table S1. Methane production as well as fatty acid, sulfate and sulfide concentrations were monitored over a 6-days incubation period.

Gas in bottles’ headspace was sampled using a gas tight syringe and analyzed for H₂ and CH₄ with a Compact GC^4.0 (Global Analyser Solutions, Breda, Netherlands) equipped with Carbonex 1010 column (Supelco, 3 m × 0.32 mm) followed by a Mosieve 5A column (Restek, 30 m × 0.32 mm) and a thermal conductivity detector (TCD). Argon was used as carrier gas at 0.8 mL min^-1. Standard GC settings for H₂ and CH₄ measurement were: 300 kPa; valve (injection) oven: 60°C; column oven: 100°C; TCD temperature: 100°C; filament: 175°C.

Liquid samples were analyzed for volatile fatty acids (VFA) and alcohols with an Accela HPLC (Thermo Scientific, Waltham, MA, United States) equipped with a Varian Metacarb 67H column (Agilent, 300 mm × 6.5 mm) and a refractive index detector. Column was kept at 45°C and running with 0.01N of H₂SO₄ as eluent at a flowrate of 0.8 mL/min.

Sulfate concentrations were measured in an ICS-2100 Ion-Chromatograph system (Thermo Scientific) equipped with an AS19 column (250 mm × 2 mm) using a hydroxide (gradient) solution as eluent. Sulfide was measured using the methylene blue method, as described by Trüper and Schlegel (1964).

At the end of the assays, 2 mL aliquots were withdrawn from the vials and frozen at -80°C for further DNA extraction. DNA was extracted using the FastDNA^TM Spin Kit for Soil DNA Extraction (MPBio, Santa Ana, CA, United States). After purification, DNA was amplified and representative samples of each tested condition were chosen for sequencing in Miseq platform. Barcoded amplicons targeting the V4 region of the 16S rRNA were generated using a 2-step PCR strategy in order to reduce the impact of barcoded primer on the outcome of the microbial profiling (Berry et al., 2011). The 10–20 ng of DNA was used as template in the first PCR reaction (50 μL) which contained 10 μL HF buffer (Finnzymes, Vantaa, Finland), 1 μL dNTP Mix (10 mM; Promega, Leiden, Netherlands), 1 U of Phusion^® Hot Start II High-Fidelity DNA polymerase (Finnzymes), 500 nM of each primer (UniTag1-515f (GAGCCGTAGCCAGTCTGC-GTGYCAGCMGCCGCGGTAA) and UniTag2-806r (GCCGTGACCGTGACATCG-GGACTACNVGGGTWTCTAAT)(Walters et al., 2016). The selection procedure of the UniTags is described elsewhere (Tian et al., 2016). PCRs were performed with a SensoQuest Labcycler (Göttingen, Germany) using an adaptation of the cycling conditions of Caporaso et al. (2012). The cycling conditions for the first step consisted of an initial denaturation at 98°C for 3 min, 25 cycles of: 98°C for 10 s, 50°C for 20 s, and 72°C for 20 s, and a final extension at 72°C for 10 min. The size of the PCR products (∼330 bp) was confirmed by agarose gel electrophoresis using 5 μL of the amplification-reaction mixture on a 1% (w/v) agarose gel containing 1× SYBR^® Safe (Invitrogen, Carlsbad, CA, United States).

Five microliters of the first PCR reaction were used as template for the second PCR reaction (100 μL), which contained 20 μL HF buffer, 2 μL dNTP Mix, 2 U of Phusion^® Hot Start II High-Fidelity DNA polymerase and 500 nM of UniTag1 (forward) and Unitag2 (reverse), each. UniTag1 and Unitag 2 were appended with an 8 nt sample specific barcode at the 5′end (Ramiro-Garcia et al., 2016). The cycling conditions consisted of an initial denaturation at 98°C for 30 s, followed by five cycles of: 98°C for 10 s, 52°C for 20 s, and 72°C for 20 s. A final extension at 72°C for 10 min followed the cycles. The incorporation of the specific barcodes (PCR product of ∼350 bp) was confirmed by gel electrophoresis.

The final PCR products were purified with HighPrep^TM (Magbio Genomics, Gaithersburg, MD, United States) using 20 μL of Nuclease Free Water (Promega, Madison, WI, United States) for elution and quantified using a Qubit in combination with the dsDNA BR Assay Kit (Invitrogen, Carlsbad, CA, United States). The purified products were mixed together in equimolar amounts to create a library pool and sent for sequencing on the Miseq platform (GATC Biotech AG, Konstanz, Germany). Data was analyzed using NG-Tax, a validated pipeline for 16S rRNA analysis (Ramiro-Garcia et al., 2016). Sequence data have been deposited in the European Nucleotide Archive, accession number [ENA: PRJEB20620].

A modified Gompertz model (Eq. 1) was used to fit the data from CH₄ measurements:

in which, M (mmol/L) is the cumulative methane production, P (mmol/L) is the methane production potential, R_m (mmol/L.h^-1) is the rate of production, λ (h) is the lag phase time, e is exp (1), and t (h) is time (Lay et al., 1997). An example of data fit is given in the Supplementary Figure S1.

For statistical comparison of different conditions, the probability distribution of the parameter vector [P R_m λ] was obtained by a Markov chain Monte Carlo sampling method (ter Braak and Vrugt, 2008), and summarized by its 95% confidence region, following the methodology described by van Mourik et al. (2014). Sampling was carried out in log-space, i.e., after a log transformation of the parameter vector. For estimating the maximum likelihood of the parameter vector we used a hybrid algorithm consisting of a global search of the genetic search routine ’GA’ in Matlab^® software with a population of 200 with a maximum of 10³ generations, followed by a gradient based search with the ‘lsqnonlin’ routine in Matlab^®, starting from the optimum found by ‘GA.’ For Monte Carlo sampling, we used three chains, 2 × 10⁴ iterations per chain, 2 × 10³ burn-in iterations, and a thinning rate of 10. We used a log-uniform prior. The confidence regions were represented with 300 samples.

An example of the 95% confidence curves obtained for two of the studied conditions (Control and Ni-H-8, hydrogenotrophic) is given in Supplementary Figure S2. Three time points (t₂₀, t₆₀, t₁₂₀, time in hours) were selected for comparison between different conditions. These time points are representative of different growth phases: the two first time points allow to compare differences in lag phase (when present) and initial methane production, while the last point represents the final cumulative methane production.

An analysis of the genomes of Methanobacterium formicicum JCM 10132 (Taxonomy ID: 1300163), Methanosaeta concilii GP6 (Taxonomy ID: 990316), Methanosarcina barkeri CM1 (Taxonomy ID: 796385) and Methanolinea sp. SDB (Taxonomy ID: 1735327) was performed using the online tool Rast¹ (Aziz et al., 2008; Overbeek et al., 2014; Brettin et al., 2015) and checked the presence of genes related with Ni/Co transport and metal resistance in these microorganisms. RcnA could not be found in Rast search for resistance subsystems. For this reason, RcnA protein sequence (NC-000913.3) (Rodrigue et al., 2005) was also searched using the online tool BLAST² and compared with the genomes of the methanogens. These methanogens were chosen as they represent the abundant archaeal genera present in our samples. Methanosarcina barkeri, was not abundant in our sludge, but it was included in the analysis because this is the best studied methanogen in what concerns metal toxicity. M. barkeri is often present in anaerobic sludges (Ziganshin et al., 2013; Venkiteshwaran et al., 2015).

Cumulative methane production values measured after 20, 60, and 120 h of incubation of the anaerobic sludge are shown in Table 1 (incubations with Ni) and Table 2 (incubations with Co). These values were obtained from the methane production curves, with deviations calculated for confidence level of 95%. Examples of methane production curves from both stimulation and toxicity assays are shown in Figure 1. In the assays with H₂/CO₂ as substrate, no VFA were detected in the medium (data not shown). Regarding the test with acetate, no other VFAs were produced and acetate consumption could be stoichiometrically correlated to the formation of methane (data not shown).

Supplementation with 8 μM Ni resulted in approximately 10% higher cumulative methane production from H₂/CO₂ (at t₁₂₀, compared to the control), but the same effect was not observed for lower amounts of Ni (i.e., 2 and 4 μM Ni). On the other hand, high concentrations of Ni (i.e., 2, 4, and 8 mM) had a detrimental effect on methanogenesis from H₂/CO_2. When compared with the controls, cumulative methane production (t₁₂₀) decreased by approximately 20% when 2 mM of Ni were added and by 40% with 4 and 8 mM of Ni. Aceticlastic methanogenesis did not benefit from Ni supplementation, and was negatively affected by high Ni concentrations (decrease of 11 and 18% on the final cumulative methane production at concentrations of 4 and 8 mM of Ni, respectively).

Supplementation of 10.5 μM of Co to hydrogen-consuming cultures was beneficial, but also for this metal, high Co concentrations (higher than 5 mM) had a negative effect (Table 2)_. Aceticlastic methanogenesis was not affected by Co supplementation, but the toxic effect of Co was noticeable for concentrations ≥20 mM; cumulative methane production (t₁₂₀) decreased by ∼11% and ∼53% for concentrations of 20 and 30 mM Co respectively.

To evaluate the potential of sulfide (produced by SRB) as metal detoxification agent, anaerobic sludge was incubated with 8 mM of Ni or 30 mM of Co plus sulfate. Incubations with sulfate (no metal) and with sulfide (Na₂S) plus metal were also performed as controls. The results are depicted in Figures 2–4 and Table 3.

The addition of sulfate together with 8 mM of Ni or 30 mM of Co had no beneficial effect on methane production from H₂/CO₂. However, direct addition of Na₂S to assays with either Ni or Co helped to alleviated metal toxicity in most of the cases (both methane production rate and final cumulative methane concentration). This is illustrated in Figure 3, where we show methane production profiles from H₂/CO₂ in the presence of Ni, with and without sulfate or Na₂S. For Co, the effect of adding sulfide on metal toxicity mitigation is not immediate and initial rates of H₂/CO₂ conversion to methane are still in effect, although this addition has an effect on the final recovery of methane (Figure 5). For aceticlastic cultures, the addition of 15 mM to 30 mM of sulfate together with Co had beneficial effects on methane production rate and/or cumulative maximum methane production (Figure 4B). However, when assays where amended with 45 mM sulfate, methane production was even lower than in the presence of only the metal.

The archaeal community (Figure 6) was dominated by species of Methanobacterium (hydrogenotrophic methanogens) and Methanosaeta (aceticlastic methanogens), which together represented 70 to 96% of the total archaeal community in the samples. Methanolinea, another genus of hydrogenotrophic methanogens, represented about 4-16% of the archaeal community. Unclassified species of Halobacteriales (1–23%) were also detected in most of the samples. In the presence of 8 mM of Ni, with H₂/CO₂ as substrate, there was a change in the relative abundances of Methanosaeta from 40–48% to ∼21%, while Halobacteriales relative abundances increased to 20%. Relative abundances of Methanobacterium and Methanolinea members in H₂/CO₂-grown cultures showed minute variations in the presence of Ni (all range of tested concentrations). However, in the presence of 30 mM Co Methanobacterium decreased from 42–46% to 28–35%. As expected, when acetate was the supplied substrate, members of Methanobacterium decreased from ∼45% to 24–31% in the presence of the metals. Halobacteriales were absent in the assays growing with acetate, without or with low concentrations of metals, while they appeared in samples with metal plus sulfate or sulfide in relative abundances up to 16%.

Bacterial communities were dominated by Anaerolineaceae (3 to 13%), Syntrophomonadaceae (3 to 12.5%), Nitrospiraceae (2 to 18%), Geobacteraceae (5 to 19%), and Syntrophaceae (1.7 to 17.5%), while other families were present in smaller abundances (Figure 7). Carnobacteriaceae members, which in hydrogenotrophic samples represented up to 13% of the community, were not present in samples containing high concentrations of Co, Ni, sulfate, or sulfide. Relative abundances of Eubacteriaceae, which were more dominant in hydrogenotrophic incubations (up to 7%), also changed in the presence of metal, sulfate, or sulfide. Rhodocyclaceae members, specifically of the genus Azospira, were present in the hydrogenotrophic samples containing Ni and sulfate, representing 8 to 12% of the community in these samples. Desulfovibrio species could also be detected, but only in incubations with 45 mM of sulfate and Ni.

Genes encoding for CbiMNQO and NikMQO transport systems could be found in the genomes of M. formicicum JCM, M. concilii GP6, M. barkeri CM1, and Methanolinea sp. SDB. Furthermore, M. barkeri genome contains genes encoding for the NikABCDE system, including the NikA subunit. M. concilii, M. formicium, and M. barkeri possess genes encoding for NikN. In addition, all four genomes have multiple homologs encoding for CzcD, which is a known cobalt-zinc-cadmium resistance protein. The Ni and Co resistance gene from E. coli RcnA was also searched in the genomes of the methanogens but no significant hits were retrieved.