Search
Close this search box.

Enhanced methane production with co-feeding spent coffee grounds using spare capacity of existing anaerobic food waste digesters – Scientific Reports

Reactor performance

The duplicate experimental reactors, R1 and R2, were operated for more than 26 months at increasing ratios of SCG to FW in the reactor feed, from 0 to 10% on a VS basis, and the OLR rose from 2.50 to 2.75 g VS/L·d accordingly (see Supplementary Table S2). The reactors were successfully started up and stabilized during Phase 0 fed with FW only, and they performed very similarly to each other throughout the experiment (Figs. 1 and 2). This result indicates the sound replication of the reactor experiments. The steady-state methane yield and organic removal obtained in Phase 0 (Table 1) were comparable to those reported for FW mono-digestion in other studies3,26.

Figure 1
figure 1

Methane production and organic loading profiles in the duplicate reactors R1 and R2. Methane yield was calculated per unit mass of VS (a) and COD (b) fed to each reactor. SCG co-feeding ratios are presented as percentages relative to FW on a VS basis.

Figure 2
figure 2

VS (a) and COD (b) removal profiles in the duplicate reactors R1 and R2. SCG co-feeding ratios are presented as percentages relative to FW on a VS basis.

Table 1 Steady-state performance data in each experimental phase.

Interestingly, methane yield increased with increasing SCG in the co-feeding ratio range of 1–4% of FW on a VS basis (Phases 1–3) based on both VS and COD fed to the reactors, and no further increase was observed after increasing SCG content to 10% (Phase 4) (Fig. 3). The methane yields obtained in Phases 1–4 were significantly higher than that in Phase 0 (Tukey’s test, p < 0.05). This result suggests that co-feeding small amounts of SCG (≤ 10% of FW) was not detrimental but rather beneficial to methanogenic performance under the experimental conditions. Correspondingly, the concentration of residual VFAs remained at very low levels (< 200 mg COD/L) with neutral pH throughout Phases 0–4 (see Supplementary Fig. S1). One point to note is that FW was collected on two occasions due to the long experimental period, and the reactors were fed with the first batch of FW until day 284 and afterwards with the second batch (see Supplementary Table S1). The first and second batches of FW had very similar TS and VS contents but significantly different COD concentrations, implying that the chemical composition and energy content of FW varied. The total COD-to-VS ratio decreased from 1.62 for the first batch of FW to 1.56 for the second, which explains why the methane yield per VS fed (YMVSf) did not increase significantly between Phases 1 and 2, while the methane yield per COD fed (YMCODf) did (Fig. 3). Both YMVSf and YMCODf showed an increasing trend with increasing the SCG fraction in the feed over Phases 0–3, despite the decrease in the COD-to-VS ratio of the FW, suggesting that the most efficient co-digestion of SCG and FW can be achieved by co-feeding SCG at 4% of FW (VS basis).

Figure 3
figure 3

Steady-state methane yields per unit mass of VS (a) and COD (b) fed in each experimental phase. Data presented are averages of the duplicate reactors R1 and R2. Different letters above bars indicate statistically significant differences (p < 0.05, Tukey’s test).

Both reactors failed with significant performance degradation in three turnovers of the HRT during Phase 5, which was conducted under the same conditions as Phase 4 but without supplementing the feed with trace elements (Fe, Ni, and Co; see Subsection “Reactor setup and operation“) (Figs. 1 and 2). Accordingly, a serious process imbalance occurred with a sudden accumulation of VFAs, primarily acetate and propionate, accompanied by reactor acidification, during Phase 5 (see Supplementary Fig. S1). These results agree with a previous study by the authors’ group reporting that FW digesters failed after co-feeding SCG with FW at a ratio of 1:10 (VS basis) due to a suspected lack of trace elements3. Given that the experimental reactors maintained stable operation for 188 days (nearly five turnovers of the HRT) during Phase 4 with higher methanogenic performance as compared to Phase 0, it is clear that providing additional trace elements is important for the stable co-digestion of SCG and FW, especially in a 1:10 mixture. The deficiency of trace elements in SCG-FW mixtures may be ascribed to the chelation of metal cations by anionic polymeric compounds present in SCG, such as polyphenols and especially melanoidins. Coffee melanoidins effectively chelate metals, especially iron, at low concentrations, and metal ions captured at the melanoidin core become unavailable for microorganisms27.

It is notable that the removal efficiency of protein increased significantly with the addition of SCG from less than 20% to approximately 50%, while those of carbohydrate and crude fat remained relatively stable between about 90 and 100% (Fig. 4). The markedly higher residual concentration of protein as compared to carbohydrate and fat reflects the fact that protein is the most abundant cellular component of bacteria, accounting for approximately 50–80% of dry weight28, and agrees with previous FW-AD studies reporting the significantly lower removal of protein as compared to carbohydrate and fat29,30. Although the underlying mechanism is unclear, the enhanced protein degradation appears to have contributed to the increase in methane yield in Phases 1–4 with SCG addition as compared to Phase 0 without it, which deserves further research. Accordingly, protein removal efficiency had a significant positive correlation (Pearson, p < 0.05) with the SCG fraction in the feed (r = 0.71) as well as methane yield (r = 0.90). Further, EEM fluorescence spectroscopy analysis revealed that the abundance of dissolved organic matter, especially protein-like substances, in the effluent decreased greatly across the experimental phases (see Supplementary Fig. S2).

Figure 4
figure 4

Carbohydrate, crude fat, and protein removal efficiencies in each experimental phase.

Microbial community analysis

The HTS analysis of 16S rRNA genes identified 66 archaeal and 2,270 bacterial ASVs from a total of 3,920,971 reads (150,807 ± 32,454 reads/sample) retrieved from the inoculum sludge and reactor samples. The taxonomic affiliations and relative abundances of major ASVs (≥ 3% of the total reads in at least one bacterial or archaeal library) in each library are presented in Table 2. Nearly all archaeal sequences (> 95% in all archaeal libraries) were assigned to three methanogenic genera: Methanobacterium, Methanospirillum, and Methanothrix (Fig. 5a). In terms of 16S rRNA gene concentration, acetotrophic Methanothrix was the most abundant in the reactors throughout the experiment (≥ 54.0%). Methanothrix accounted for approximately 55% of the archaeal sequences yielded from the inoculum sludge, with most of the remainder belonging to Methanobacterium. The relative abundance of Methanothrix increased greatly to 66.4–77.6% during Phases 0–3 with stable reactor operation, while that of Methanobacterium decreased accordingly. However, the relative abundance of Methanothrix decreased markedly during Phases 4 and 5 (≤ 64.0%) with increasing the SCG co-feeding ratio to 10% of FW (VS basis). These results show that acetotrophic methanogenesis, especially by Methanothrix, was the primary route for methane production across the experimental phases but the contribution of hydrogenotrophic methanogenesis increased after adding more SCG (> 4% of FW on a VS basis). This change corresponds to the buildup of residual soluble COD and VFAs along with significant performance degradation over Phases 4 and 5 (see Supplementary Fig. S1), given that hydrogenotrophic methanogens often dominate acetotrophic ones under imbalanced conditions due to their greater resistance to VFA inhibition31,32. Interestingly, the relative abundance of Methanospirillum increased as the SCG fraction in the feed increased, and it became the dominant hydrogenotrophic methanogen over Methanobacterium during Phases 4 and 5. While the exact cause remains uncertain, this dominance shift could be associated with the accumulation of VFAs, as previous studies have observed the thriving of Methanospirillum as the major methanogen under VFAs-enriched or organic overloading conditions33,34.

Table 2 Relative abundance and taxonomic affiliation of major archaeal and bacterial ASVs (> 3% relative abundance at least one library).
Figure 5
figure 5

Taxonomic distribution of retrieved archaeal (a genus level) and bacterial (b phylum level; c family level) 16S rRNA gene sequences. Sequences with relative abundance less than 3% in all samples were classified as “Others”. INO, Inoculum.

Most bacterial sequences (> 94% in all bacterial libraries) were assigned to ten major phyla (≥ 3% relative abundance in at least one bacterial library), while 1.0–4.8% of the total bacterial reads were unclassifiable even at the phylum level for each sample. The bacterial communities in the reactors underwent considerable changes in composition with the addition of SCG. The inoculum bacterial community was dominated by ASV B1 (51.3% of the total reads) assigned to Ca. Saccharibacteria (Table 2), whose members can hydrolyze various organic compounds in wastewater treatment systems35,36. However, its relative abundance decreased over the experimental phases and became less than 0.1% during Phases 4 and 5 in both reactors, which may be related to the considerable reduction in VS and COD removal efficiencies during these phases (Fig. 2).

Meanwhile, the relative abundance of Ca. Cloacimonetes (ASVs B2 and B5) and Bacteroidetes (ASVs B3, B7, B8, B16, B17, B18, B20, and B21) increased after co-feeding SCG (Fig. 5b). Members of these phyla can ferment amino acids, sugars, and alcohols into VFAs37. Both ASVs B2 and B5 were assigned to the genus Ca. Cloacamonas including H2-producing syntrophs oxidizing propionate to acetate and CO2, and members of this genus have been reported to participate in the degradation of cellulosic materials during AD38,39,40. Therefore, these Ca. Cloacamonas-related bacteria likely contributed to the degradation of SCG fibers, corresponding to the increase in their relative abundance after co-feeding SCG (Table 2 and Fig. 5b). The increment of Ca. Cloacamonas, and another hydrogenic genus Petrotoga (ASV B6) producing H2 from sugar fermentation41, over Phases 2–5 could be related to the increased proportion of hydrogenotrophic methanogens in the later phases fed with greater amounts of SCG (Fig. 5a). The dominance shifts between H2-producing populations, for example, from ASVs B5 to B2 between Phases 3 and 4, seem to reflect the differences in their tolerance and response to the accumulation of VFAs and H2 (see Supplementary Fig. S1). The relative abundance of Porphyromonadaceae (ASVs B3, B8, B17, and B20) increased with the addition of SCG over Phases 1–4. Members of this family can ferment carbohydrates and proteins into various organic acids42. The increased relative abundance of the aforementioned bacteria likely resulted from the addition of SCG (Fig. 5c).

ASV B4 was assigned to the genus Olsenella, known to degrade carbohydrates and produce lactic acid43. ASVs B9 and B10 were affiliated with the family Ruminococcaceae and the genus Atopobium, respectively, whose members are cellulolytic and commonly observed in animal guts3,16. Therefore, the bacterial populations represented by these three ASVs were potentially involved in the decomposition of SCG, although they exhibited no distinct trend in their relative abundance following the addition of SCG. Additionally, the emergence of Ruminococcaceae, Atopobium, Paludibacter, and Petrimonas under SCG feeding conditions was consistent with observations from previous studies by the authors’ group3,42, although their specific roles remain unclear.

Changes in microbial community structure and function

Clustering analysis based on the distribution of individual ASVs in the sequenced libraries revealed that co-feeding SCG in small amounts (1–10% of FW on a VS basis) had a significant influence on the evolution of the microbial community structure in the reactors. Both archaeal and bacterial cluster dendrograms were clearly divided into two main clusters: one containing the community structures of Phases 0–3 and the other containing those of Phases 4 and 5 (see Supplementary Fig. S3). This result indicates that the increase in the amount of SCG relative to FW from 4 to 10% (VS basis) between Phases 3 and 4 led to significant changes in both the archaeal and bacterial community structures. Structural changes were more pronounced in the bacterial than the archaeal communities in both reactors, which could be partly related to the less dynamic and less diverse nature of archaea as compared to bacteria in methanogenic systems44. The bacterial community maintained significantly higher diversity (Shannon index (H’) = 4.0–5.1) than the archaeal community (H’ = 1.2–2.1) throughout the experiment in both reactors (see Supplementary Fig. S4). Interestingly, the diversity of the bacterial community was greater after the addition of SCG (Phases 1–5) than before (Phase 0 and the inoculum sludge), implying that co-feeding SCG supported the growth of more diverse bacteria. Meanwhile, the archaeal community H’ decreased between Phases 0 and 1 and remained at the reduced levels during Phases 1–3, which can be attributed to the strong dominance of the community by one population (Methanothrix-related ASV A1) with 58.2–65.8% relative abundance during these phases (Table 2). Given the simple archaeal community composition, the increase in archaeal diversity during Phases 4 and 5 appears to reflect the increment of other archaeal populations than ASV A1, especially Methanospirillum-related ASVs (i.e., a less uneven distribution).

Functional potential analysis by PICRUSt2 was performed to understand the enhancement of protein removal efficiency, leading to increased methane production, with co-feeding SCG at the functional gene level. The total predicted abundance of protease genes (i.e., the sum of all protease genes) did not correlate significantly with protein removal efficiency (Pearson, p = 0.27). The predicted protease genes were individually tested to identify the genes putatively involved in the enhanced degradation of protein, and five of them showed significant positive correlations with both protein removal efficiency and the addition of SCG (Pearson, p < 0.05) (Fig. 6). Three and two of the five identified genes encode endopeptidases (endopeptidase La [EC 3.4.21.53], C-terminal processing protease [EC 3.4.21.102], and HslU–HslV peptidase [EC 3.4.25.2]) and exopeptidases (muramoyltetrapeptide carboxypeptidase [EC 3.4.17.13] and tripeptide aminopeptidase [EC 3.4.11.4]), respectively, suggesting that the enhanced protein degradation observed under SCG co-feeding conditions may be related to facilitated peptide hydrolysis. Notably, 30 of 2,270 bacterial ASVs were identified as having a significant positive correlation with both protein removal efficiency and the amount of SCG added, as well as carrying one or more of these predicted genes. Half of the ASVs belonged to two families Ruminococcaceae (10) and Syntrophomonadaceae (5) of the fermentative order Clostridiales, and six belonged to the family Spirochaetaceae. Six of the remainder were assigned to the families Alcaligenaceae (1), Bacteroidaceae (1), Porphyromonadaceae (2), Synergistaceae (1), and Thermotogaceae (1), and the other three were classifiable only at the phylum level as Bacteroidetes. Members of the families mentioned above commonly occur in methanogenic systems and involved in the fermentative degradation of organic matter, including protein, to VFAs and/or H2/CO213,45,46,47. Furthermore, the co-occurrence analysis (Pearson, p < 0.05) of major ASVs (Table 2) identified ASV B16, which was among the three major ASVs predicted to possess the protease genes of interest (B6, B11, and B16), as the most influential node with 16 edges (nine positive and seven negative interactions) with other ASVs. In fact, ASV B16 was assigned to the genus Bacteroides, belonging to the family Bacteroidaceae, and showed 100% sequence identity with a saccharolytic and proteolytic species Bacteroides pyogenes48. These results suggest that the proteolytic bacteria discussed above may have contributed to the enhancement of protein removal efficiency with the addition of SCG in the experimental reactors.

Figure 6
figure 6

Predicted protease genes (indicated by EC numbers) potentially related to the enhancement of protein degradation under SCG co-feeding conditions.