Coal samples were obtained from the Argonne National Laboratory’s Premium Coal Sample Program (Argonne, IL, United States). The location, rank, and collection date for each coal sample are shown in Table 1. The collection and processing methods for these coals are detailed in Vorres (1990). To summarize, the lignite sample was collected from the Beulah-Zap coal seam in the Williston Basin (North Dakota, United States). The seam was approximately 18 ft. (5.5 m) thick at the collection area, and collection was done by accumulating core samples spaced approximately 20 ft. (6.1 m) apart (Vorres, 1990). One high volatile bituminous coal sample was collected from the Lewiston-Stockton seam of the Kanawha formation about 20 miles (32.2 km) east of Charleston, West Virginia, while the other was obtained from the Pittsburgh #8 seam approximately 60 miles (96.6 km) south of Pittsburgh, Pennsylvania. Both collection sites had a seam thickness of 6 ft. (1.8 m; Vorres, 1990). The low volatile bituminous coal sample was collected from the Pocahontas #3 seam in Buchanan County, Virginia, and the seam thickness was 6 ft. (1.8 m). The subbituminous B coal sample was collected from the Wyodak-Anderson seam approximately 6 miles (9.7 km) northeast of Gillette, WY. The seam was approximately 120 ft. (36.6 m) thick at the collection site and the sample used for processing consisted of a 6″ (15.3 cm) core spanning the thickness of the seam (Vorres, 1990). In addition to the Argon samples, a subbituminous C coal sample was collected from the Flowers-Goodale (FG) seam in the Powder River Basin in Rosebud County near Birney, Montana. Core samples from a well designated FGP-13 were extruded into a wooden trough and cleaned to remove drilling fluids. Samples were transferred to a disposable glove bag filled with N₂ where the exposed outer core was removed (Barnhart et al., 2016). Proximate and ultimate analyses of the coal samples from the Argonne National Laboratory’s Premium Coal Sample Program (Vorres, 1990) and the Flowers-Goodale subbituminous C sample can be found in Table 2. FG proximate and ultimate analysis was obtained from Barnhart et al. (2016). All analyses were determined on a dry weight basis, except for percent hydrogen, which is reported on a dry, mineral-matter free basis, and heating value, which is reported on an as-received basis.

All water and microbial samples were collected from the Birney Test Site, previously described by Barnhart et al. (2016), located near Birney (MT) in the Powder River Basin (PRB). Formation water from the FG coal bed was obtained in June 2016. Water from the FG-09 well was collected in 6-gal (22.7 l) plastic jugs after pumping two well volumes and rinsing the jugs twice with formation water before filling and storing them at 4°C until microcosm set up. Coal cores were collected during the July 2013 drilling of the FG monitoring (FGM-13) and FG pumping (FGP-13) wells. The 2-inch (5.1 cm) diameter cores were cut into approximately 12-inch long (0.3 m) sections and placed in sealed PVC tubes filled with formation water pumped from the FG-11 well, which accesses the same FG coal source and aquifer. The tubes were sealed with flexible rubber caps to allow room for gas desorption. Microbial cultures were collected from the SS-13 well, which accesses the interface between the FG coal and the overlying sandstone, in September 2015 using the down-well samplers described by Barnhart et al. (2013). The microbial cultures (slurries) from the SS-13 sampler were added to a sterile serum bottle prepared with 5 g FG coal and 45 ml of filter sterilized, reduced formation water. The serum bottle was incubated at 21 ± 1°C in the dark for 361 days prior to being used as inoculum for the microcosms described below.

Coal samples were received from the Argonne National Laboratory’s Premium Coal Sample Program in glass ampules. The glass ampules and the FG coal core (depth 384–385′) were opened in an anaerobic chamber (less than 5% H₂ and less than 5% CO₂), where it was dried and crushed. Crushed coal was sieved to an effective size of less than 850 μm. Prior to use in the microcosms, all coal samples were exposed to UV sterilization in a biosafety cabinet; during UV sterilization, the coal samples were briefly exposed to oxygen. Treatments were prepared in an anaerobic chamber by adding 1 g of each coal sample to autoclaved Balch tubes. To obtain a final liquid culture volume of 10 ml, each microcosm received 1 ml of previously collected enrichment culture described above and a balance of prepared formation water (9 ml for unamended treatments and 8 ml for amended treatments). Formation water was filtered (0.2-μm bottle top filters), sparged overnight with anoxic 5% CO₂ (balance of N₂) and treated with 1 mM sulfide (Na₂S^●9H₂O) to remove residual oxygen. Resazurin was added to the reduced formation water to a concentration of 1 mM as a redox indicator. Microcosms were prepared in an anaerobic chamber by adding 1 g of each coal sample to autoclaved Balch tubes. To obtain a final liquid culture volume of 10 ml, each microcosm received 1 ml of previously collected enrichment culture described above and a balance of prepared formation water (9 ml for unamended treatments and 8 ml for amended treatments). One ml of an amendment suspension was added as appropriate. The amendment, a dried and crushed microalgal concentrate (Chlorella sp. strain SLA-04), was added to a final concentration of 0.1 g/l, as previously described (Davis et al., 2018b). Other than the microalgal concentrate, the microcosms received no other amendment. The Balch tubes were sealed with autoclaved butyl rubber stoppers and aluminum crimp seals before being removed from the anaerobic chamber. After removal, the headspace of the Balch tubes was replaced with a 5% CO₂ gas mixture (balance of N₂). Amended and unamended microcosms containing 1 mm borosilicate glass beads (GB) were set up as coal-free substrate controls. For each treatment (coal and glass beads), triplicate treatments were prepared. Treatments were incubated in the dark at room temperature (21 ± 1°C) for the duration of the 116-day experiment.

Headspace gases (CH₄ and CO₂) were analyzed using an SRI Instruments (Torrance, CA, United States) Model 8,601 Gas Chromatograph (GC) equipped with a thermal conductivity detector (TCD) interfaced with PeakSimple Chromatography software. Ultra-high purity helium carrier gas and a Supelco HayeSep-D packed stainless-steel column (6 feet × 1/8” O.D) were used for separation. One ml of headspace gas was sampled from each microcosm and manually injected; the carrier gas had a pressure of 8 psi and the oven and TCD temperatures were 40°C and 150°C, respectively. To prevent creating a negative pressure in the tubes, 1 ml of anoxic 5% CO₂ (balance N₂) was injected to replace the sample volume removed. Reactors were sampled approximately every 2 weeks for the duration of the 116-day experiment.

On day 116, one replicate of each treatment was destructively sampled for DNA analysis, according to the protocol described in Davis et al. (2018b). Briefly, the coal or glass beads (GBs) and liquid fractions were separated by decanting the liquid fraction into a 15-ml Falcon conical centrifuge tube. The coal or GBs were transferred to separate 15-ml Falcon tubes. One ml of filter sterilized, 10% sodium dodecyl sulfate (SDS) was added to each gram of coal or GBs, and tubes were placed in a 70°C water bath for 30 min. The previously transferred liquid fraction in Falcon tubes was centrifuged, the supernatant decanted and discarded to leave approximately 2 ml with the resulting pellet. Both sample fractions were stored at -80°C until extraction. Just prior to DNA extraction, the coal/SDS mixture was heated in a 70°C water bath for an additional 30 min. Sample DNA was extracted using the FastDNA Spin Kit for Soil (MP Biomedical, Solon, OH) according to the manufacturer’s instructions with minor protocol changes, as reported in Davis et al. (2018b). After extraction, the DNA was prepared for polymerase chain reaction (PCR) amplification using the OneStep^™ PCR Clean Up (Zymo Research, Irvine, CA). Extracted DNA was quantified using a Qubit fluorometer using the dsDNA HS Assay Kit (Thermo Fischer Scientific, Carlsbad, CA, United States). The 16S rRNA genes were PCR-amplified for 30 cycles with DreamTaq PCR Master Mix (Thermo Fischer Scientific, Carlsbad, CA, United States) and an annealing temperature of 55°C for 30 s using the universal prokaryotic primers Pro341F (5’-CCTACGGGNBGCASCAG-3′) and Pro805R (5’GACTACNVGGGTATCTAATCC-3′), which amplify the V4 region of the 16S rRNA genes of bacteria (Takahashi et al., 2014). Additionally, the archaea specific primers 751F (− CCGACGGTGAGRGRYGAA) and 1204R (− TTMGGGGCATRCIKACCT) were used to amplify the 16S rRNA gene of archaea (Baker et al., 2003).

Amplicons were checked by agarose gel electrophoresis with GelRed DNA stains (Biotium, Freemont, CA, United States). Library preparation for Illumina MiSeq was carried out following Illumina’s standard protocol “16S Metagenonic Sequencing Library Preparation” prior to being loaded for sequencing on the MiSeq v.3 platform (San Diego, CA, United States). To summarize, following PCR cleanup, purification, and indexing, DNA concentration was determined using the PicoGreen stain assay (Quant-IT, Invitrogen, Carlsbad, CA, United States). DNA concentrations were normalized and pooled with a 12.5% PhiX control library. Sequence reads were analyzed using the MiSeq standard operating procedure of the Mothur software package (Kozich et al., 2013). Forward and reverse reads were joined into contigs using QIIME (Caporaso et al., 2010). The sequences were aligned using SILVA (Quast et al., 2012). The aligned reads were quality filtered and chimeras were removed before OTUs and phylotypes were classified with an 80% confidence using the RDP database (Wang et al., 2007; Haas et al., 2011). Mothur version 1.38.1 was used to calculate species richness using Inverse Simpson Index (Supplementary Tables S1, S2).

A Generalized Linear Model was fitted to the total amount of methane produced, maximum methane production rate, total carbon dioxide production, and carbon dioxide production rate with a factor for coal sample (rank) and a factor for the presence or absence of algal amendment using the statistical software Minitab v.18 (Minitab LLC, State College, PA). A value of p of less than 0.05 was used to determine statistical significance for all statistical analyses. Microbial community statistical analyses were performed in R v3.5.3 (R Development Core Team, 2019). Heatmaps of bacterial and archaeal relative abundances of combined OTUs based on phylotype at a minimum of 2.5% in at least one sample were generated using the CRAN packages Heatplus (Ploner, 2022), Vegan (Jari Oksanen et al., 2013), RColorBrewer (Neuwirth, 2022), and ggplots (Wickham, 2016). Principal Coordinate Analyses (PCoA) using Bray–Curtis Dissimilarity distances and non-transformed, combined OTU (based on phylotype) relative abundances (>2.5%) were used to explore differences in bacterial and archaeal community composition between microcosms. The environmental fitting function found in the CRAN package vegan was used to correlate community composition and coal sample metadata.

Methane production was observed in all unamended microcosms, except for the negative control treatment containing glass beads instead of coal (Figure 1A), suggesting that coal-to-methane conversion occurred in all coal samples. Methane production was first detected on day 38 in all unamended coal treatments. The duration of methane production varied among microcosms, with Lignite producing methane until day 75, SubC until day 92, SubB until day 56, HV Bit (Pittsburgh) until day 75, HV Bit (Stockton) producing until day 75, and LV Bit until day 116. At the time of destructive sampling, cumulative methane production differed among microcosms. The unamended Lignite, SubC, SubB, HV Bit (Pittsburgh), HV Bit (Stockton), LV Bit, and GB microcosms produced 31.6 ± 8.8 μmol CH₄/g coal, 59.2 ± 6.0 μmol CH₄/g coal, 35.7 ± 2.1 μmol CH₄/g coal, 26.9 ± 2.7 μmol CH₄/g coal, 32.4 ± 6.2 μmol CH₄/g coal, 19.3 ± 2.3 μmol CH₄/g coal, and 0.0 ± 0.0 μmol CH₄/g coal, respectively.

In algae-amended treatments, methane was detected earlier than in unamended treatments in Lignite Algae and GB Algae (methane was detected on day 16 already). The remaining amended microcosms showed detectable methane production by day 38 (Figure 1B) but at higher concentrations than the unamended treatments (cf. Figures 1A,B). The duration of methane production in algae amended treatments varied. For the Lignite Algae, SubC Algae, SubB Algae, HV Bit (Pittsburgh) Algae, HV Bit (Stockton) Algae, LV Bit Algae, and GB Algae, methane production appeared to cease on day 56, 75, 56, 56, 56, 116, and 56, respectively. At the end of the 116-day study, Lignite Algae, SubC Algae, SubB Algae, HV Bit (Pittsburgh), HV Bit (Stockton), LV Bit Algae, and GB Algae had produced 38.7 ± 2.6 μmol CH₄/g coal, 65.3 ± 2.7 μmol CH₄/g coal, 45.1 ± 4.7 μmol CH₄/g coal, 35.7 ± 3.1 μmol CH₄/g coal, 31.8 ± 1.1 μmol CH₄/g coal, CH₄/g coal, 27.0 ± 2.2 μmol CH₄/g coal, and 11.7 ± 0.5 μmol CH₄/g coal, respectively.

All unamended coal microcosms except for the unamended LV Bit enrichment (p = 0.56) produced significantly more methane than the amended and unamended GB (p < 0.05), indicating that the microbial community was able to convert some fraction of the coal substrate to methane. Methane was not observed in the unamended GB treatment, indicating that methane production from dissolved organics in the inoculum was negligible and that the source of methane in coal- or amendment-containing treatments was from coal or algal amendment. Additionally, coal rank had a statistically significant effect on total methane production in both unamended and amended microcosms (p < 0.05), although there was not an obvious statistical trend as vitrinite reflectance increased or decreased. For example, the cumulative methane production from the lowest ranked coal evaluated in this study (Lignite, R_O = 0.25%) grouped statistically with the methane production from the second highest ranked coal [HV Bit (Stockton), R_O = 0.89%]. The amended and unamended subbituminous C samples (SubC Algae, SubC, R_O = 0.40%) produced the highest amount of methane and the unamended low volatile bituminous coal sample (LV Bit, R_O = 1.68%) produced the lowest amount of methane of all the coal treatments. The remaining microcosms grouped statistically within these bounds, but without a strong correlation with vitrinite reflectance. A complete table of Tukey pairwise comparisons of cumulative methane production between treatments can be found in the Supplementary Table S3. On aggregate, the presence of algal amendment generally resulted in an increase in total methane production (p < 0.05), suggesting that methane production can be stimulated in different ranks of coal. Controlling for coal sample, no pairwise comparison (amended vs. unamended) resulted in a statistically significant difference in total methane production (Figure 2). However, comparing amended high rank samples to unamended low rank samples, comparable methane production was achieved. For example, methane production in the amended low volatile bituminous sample (LV Bit Algae) grouped statistically with the unamended subbituminous B (SubB) microcosm and both unamended high volatile bituminous microcosms (HV Bit (Stockton) and HV Bit (Pittsburgh)). This result suggests that the biogenic methane production potential of higher rank coals can be increased to be competitive with lower rank coals by the addition of small amounts of algae or similar amendments.

Maximum methane production rates, and the time interval when they occurred, differed between coal microcosms. The maximum methane production rate for the unamended Lignite enrichment was 0.77 ± 0.54 μmol CH₄/g coal/day between days 38 and 56 (Figure 3). The maximum methane production rate for the unamended SubC and SubB microcosms were 1.28 ± 0.21 μmol CH₄/g coal/day between day 56 and 75 and 0.93 ± 0.20 μmol CH₄/g coal/day between day 38 and 56, respectively (Figure 3). The maximum methane production rates for the unamended high volatile bituminous microcosms HV Bit (Pittsburgh) and HV Bit (Stockton) were 0.64 ± 0.15 μmol CH₄/g coal/day between day 38 and 56 and 0.67 ± 0.29 μmol CH₄/g coal/day between day 16 and 38, respectively. Lastly, the maximum methane production rate for the unamended low volatile bituminous enrichment, LV Bit, was 0.29 ± 0.04 μmol CH₄/g coal/day between day 75 and 92. Coal rank had a statistically significant effect on maximum methane production rate (p < 0.05), with the highest rank coal microcosm (LV Bit, R_O = 1.68%) having the lowest methane production rate. However, a clear trend between increasing coal rank and methane production rate could not be determined.

The addition of algal amendment also had a statistically significant effect on methane production rate (p < 0.05), with algae-amended microcosms generally exhibiting greater methane production rates than unamended microcosms (Figure 3). The interaction between the effect of algal amendment on methane production rate and coal rank was not statistically significant (p = 0.78), suggesting that algal amendment can stimulate methane production regardless of coal rank. Complete methane production rate statistical analysis results and Tukey pairwise grouping can be found in the Supplementary Table S4. The maximum methane production rate for the amended Lignite Algae enrichment was 1.30 ± 0.22 μmol CH₄/g coal/day, occurring between day 16 and 38. The maximum methane production rate for the amended SubC Algae and SubB Algae microcosms were 1.72 ± 0.51 μmol CH₄/g coal/day and 1.61 ± 0.09 μmol CH₄/g coal/day, occurring between day 38 and 56 and day 16 and 38, respectively. The amended high volatile bituminous treatments, HV Bit (Pittsburgh) and HV Bit (Stockton), had maximum methane production rates of 1.06 ± 0.10 μmol CH₄/g coal/day between day 16 and 38 and 0.91 ± 0.09 μmol CH₄/g coal/day between day 16 and 38. Lastly, the maximum methane production rates for the amended LV Bit and amended GB Algae treatments were 0.59 ± 0.11 μmol CH₄/g coal/day and 0.41 ± 0.02 μmol CH₄/g coal/day, respectively. Notably, amended treatments had earlier maximum methane production rates compared to corresponding unamended treatments, except for the HV Bit (Stockton) treatments, which had maximum methane production rates between the same sampling days (between day 16 and day 38). Maximum methane production rates in amended treatments were observed 18–37 days earlier compared to unamended treatments (Figure 3), and methane production leveled out 17–19 days earlier in Lignite Algae, SubC Algae, HV Bit (Pittsburgh) Algae and HV Bit (Stockton) Algae compared to analogous unamended treatments (Figure 1). These results suggest that the presence of the algal amendment decreased the lag time before the onset of methane production in these treatments. The SubB Algae and LV Bit Algae treatments ceased methane production approximately at the same time as the corresponding unamended treatments, but elevated maximum methane production rates were observed. These results suggest that while final methane production may not increase, the addition of algal amendment can decrease lag times before methane production commences and can increase production rates across different coal ranks.

Depending on the methanogenic pathway, inorganic carbon (CO₂) can be either a substrate or a product, and CO₂ can therefore be an indicator of the type of methanogenesis occurring in a system. Thus, the mass of CO₂ produced in the headspace of the microcosms was monitored for the duration of the experiment. Production of inorganic carbon in unamended microcosms was associated with coal rank (p < 0.05), with subbituminous coals generally producing more CO₂ than bituminous coals. After accounting for CO₂ added during sampling (through the replacement of headspace gas), unamended Lignite, SubC, SubB, HV Bit (Pittsburgh), HV Bit (Stockton), LV Bit, and GB microcosms produced 1.1 ± 1.7 μmol CO₂/g coal, 72.0 ± 4.7 μmol CO₂/g coal, 15.2 ± 2.6 μmol CO₂/g coal, 6.0 ± 1.5 μmol CO₂/g coal, 3.6 ± 5.6 μmol CO₂/g coal, 3.7 ± 1.2 μmol CO₂/g coal, and − 3.1 ± 2.1 μmol CO₂/g coal, respectively by the end of the 116-day study (Supplementary Figure S1). A “negative” production in the GB treatment headspace suggests that CO₂ was consumed. Because no methane was detected in this treatment, the consumption of CO₂ is likely the result of CO₂ dissolution into the aqueous phase and subsequent speciation into H₂CO₃, HCO₃⁻, and CO₃²⁻ as opposed to hydrogenotrophic methanogenesis; no photosynthesis is expected to have occurred in the microcosms because they were incubated in the dark. While subbituminous coals generally produced more CO₂ than higher rank coals, the lignite treatment did not follow this trend, grouping more closely with the low volatile bituminous coal (LV Bit). An increased production of inorganic carbon in the lower rank coals may suggest the increased degradation of organic matter, which is supported by generally higher amounts of methane production in these treatments.

Amended Lignite Algae, SubC Algae, SubB Algae, HV Bit (Pittsburgh) Algae, HV Bit (Stockton) Algae, LV Bit Algae, and GB Algae microcosms produced 4.6 ± 0.3 μmol CO₂/g coal, 75.8 ± 0.4 μmol CO₂/g coal, 23.6 ± 3.5 μmol CO₂/g coal, 12.8 ± 1.9 μmol CO₂/g coal, 12.6 ± 2.7 μmol CO₂/g coal, 2.6 ± 6.5 μmol CO₂/g coal, and 7.8 ± 2.5 μmol CO₂/g coal, respectively. Once again, amended subbituminous microcosms produced more CO₂ than amended high volatile and low volatile bituminous samples. Consistent with the unamended lignite treatment, the amended lignite treatment did not group with the lower rank coals and instead produced similar amounts of CO₂ as the higher rank bituminous coals. The effect of algal amendment on CO₂ production was significant (p < 0.05), while the interaction between algal amendment and rank was not (p = 0.17), suggesting that the effect of algal amendment on CO₂ production does not depend on the rank of the coal sample. Controlling for coal sample, there were no instances in which algae-amended treatments produced significantly more CO₂ than analogous unamended treatments. A complete table of Tukey pairwise comparisons of cumulative CO₂ production can be found in the Supplementary Table S5.

Maximum CO₂ production rates (Supplementary Table S6) occurred between day 0–16 for the amended and unamended coal microcosms, except for the amended and unamended lignite sample, which had a maximum production rate between day 38 and 56, and the amended low volatile bituminous treatment (LV Bit Algae), which had a maximum production rate between day 56 and 75. Unamended Lignite, SubC, SubB, HV Bit (Pittsburgh), HV Bit (Stockton), LV Bit, and GB microcosms had maximum CO₂ production rates of 0.2 ± 0.1 μmol CO₂/g coal/day, 3.3 ± 0.2 μmol CO₂/g coal/day, 0.7 ± 0.1 μmol CO₂/g coal/day, 0.3 ± 0.1 μmol CO₂/g coal/day, 0.2 ± 0.3 μmol CO₂/g coal/day, 0.2 ± 0.1 μmol CO₂/g coal/day, and − 0.4 ± 0.0 μmol CO₂/g coal/day, respectively. Similarly, amended microcosms Lignite Algae, SubC Algae, SubB Algae, HV Bit (Pittsburgh) Algae, HV Bit (Stockton) Algae, LV Bit Algae, and GB Algae produced 0.2 ± 0.2 μmol CO₂/g coal/day, 3.5 ± 0.0 μmol CO₂/g coal/day, 1.1 ± 0.2 μmol CO₂/g coal/day, 0.6 ± 0.1 μmol CO₂/g coal/day, 0.6 ± 0.1 μmol CO₂/g coal/day, 0.12 ± 0.0 μmol CO₂/g coal/day, and 0.2 ± 0.0 μmol CO₂/g coal/day, respectively. As with total CO₂ production, the maximum production rate was generally higher in subbituminous coals than in bituminous coals (Supplementary Table S6).

Lignite treatments had maximum CO₂ production rates that occurred 40 days later than subbituminous and high volatile bituminous treatments and were closer in magnitude to the low volatile bituminous treatments. Similar to maximum methane production rates, the effect of algae on maximum CO₂ production rates was statistically significant (p < 0.05), with amended treatments on aggregate generally having greater CO₂ production rates than unamended treatments. The effect of algae on maximum CO₂ production rate was not significantly dependent on the rank of coal (p = 0.15), with the grouping of treatments generally mirroring total CO₂ production and methane production. In both amended and unamended coal microcosms, maximum CO₂ production rates occurred earlier than maximum methane production rates, except for the amended and unamended lignite treatments and the amended low volatile bituminous treatment.