Microbiome-informed study of the mechanistic basis of methane inhibition by Asparagopsis taxiformis in dairy cattle

ABSTRACT Copious amounts of methane, a major constituent of greenhouse gases currently driving climate change, are emitted by livestock, and efficient methods that curb such emissions are urgently needed to reduce global warming. When fed to cows, the red seaweed Asparagopsis taxiformis (AT) can reduce enteric methane emissions by up to 80%, but the achieved results can vary widely. Livestock produce methane as a byproduct of methanogenesis, which occurs during the breakdown of feed by microbes in the rumen. The ruminant microbiome is a diverse ecosystem comprising bacteria, protozoa, fungi, and archaea, and methanogenic archaea work synergistically with bacteria to produce methane. Here, we find that an effective reduction in methane emission by high-dose AT (0.5% dry matter intake) was associated with a reduction in methanol-utilizing Methanosphaera within the rumen, suggesting that they may play a greater role in methane formation than previously thought. However, a later spike in Methanosphaera suggested an acquired resistance, possibly via the reductive dehalogenation of bromoform. While we found that AT inhibition of methanogenesis indirectly impacted ruminal bacteria and fermentation pathways due to an increase in spared H2, we also found that an increase in butyrate synthesis was due to a direct effect of AT on butyrate-producing bacteria such as Butyrivibrio, Moryella, and Eubacterium. Together, our findings provide several novel insights into the impact of AT on both methane emissions and the microbiome, thereby elucidating additional pathways that may need to be targeted to maintain its inhibitory effects while preserving microbiome health and animal productivity. IMPORTANCE Livestock emits copious quantities of methane, a major constituent of the greenhouse gases currently driving climate change. Methanogens within the bovine rumen produce methane during the breakdown of feed. While the red seaweed Asparagopsis taxiformis (AT) can significantly reduce methane emissions when fed to cows, its effects appear short-lived. This study revealed that the effective reduction of methane emissions by AT was accompanied by the near-total elimination of methane-generating Methanosphaera. However, Methanosphaera populations subsequently rebounded due to their ability to inactivate bromoform, a major inhibitor of methane formation found in AT. This study presents novel findings on the contribution of Methanosphaera to ruminal methanogenesis, the mode of action of AT, and the possibility for complementing different strategies to effectively curb methane emissions.

CO 2 over a 20-year period, making it the second-largest contributor to global warming (1).The Paris Agreement of 2015 aimed to prevent a 1.5°C increase in global temperature from pre-industrial levels.However, 2023 marked the first time that such an increase has a greater-than-not probability of occurring within the next 5 years (2), suggesting that the proposed CH 4 mitigation strategies, which thus far slow global warming by about 30% (3), are insufficient.As such, a focus on enteric CH 4 emission from livestock, one of the largest contributors of anthropogenic CH 4 emissions (4), is clearly warranted.
Livestock produces CH 4 as a byproduct of methanogenesis, which occurs during the breakdown of feed in the rumen.The ruminant microbiome is a diverse ecosystem comprising billions of microbes, including bacteria, protozoa, fungi, and archaea.While bacteria facilitate carbohydrate breakdown, methanogenic archaea work synergistically with the bacteria, acting as a hydrogen (H 2 ) sink and reducing bacterially produced substrates, ultimately resulting in methane production and emission (5).Depending on the substrates present in the feed, different methanogenesis pathways, including the hydrogenotrophic, methylotrophic, and acetoclastic pathways (6) are utilized within the rumen.While it is commonly believed that carbon dioxide (CO 2 )-utilizing methanogens like Methanobrevibacter are the dominant methane producers, recent studies demonstra ted that methylotrophic methanogens, such as Methanosphaera, also play a large role in CH 4 production (7,8).
Although the distinct methanogenic pathways utilized in the rumen are each initiated by different enzymes, they all require methyl-coenzyme M reductase (MCR) for the final step of methanogenesis.Accordingly, livestock CH 4 -reduction technologies often include the addition of methanogenesis inhibitors into livestock feed, including the MCR inhibitor 3-nitrooxypropanol (3-NOP) (9,10).In addition, supplementation with naturally occurring seaweeds such as Asparagopsis taxiformis (AT) (11) that contain halogenated CH 4 analogs such as bromoform (CHBr 3 ), capable of inhibiting the transfer of methyl groups and MCR catalytic activity (12,13), has been gaining traction.However, the effects of AT appear unstable, potentially due to the loss of bromoform functionality over time (11), and its significant reduction of enteric CH 4 emissions in dairy cattle is transient.Finally, while AT can significantly reduce methane-generating microbes such as Methanomassilicoccaceae and Methanobrevibacter in vitro (14), the consequen tial loss of a functional H 2 sink that would accompany such inhibition could alter symbiotic relationships between microbes in the rumen.As the disruption in these critical symbiotic relationships could have a negative impact on feed intake and animal productivity (15), a comprehensive assessment of the effects of AT on the rumen microbiome is clearly warranted.
To determine if the gradual resistance to the anti-methanogenic effects of AT was due to alterations in rumen microbiome diversity and gene content, we utilized 16S rRNA, real-time PCR, and shotgun metagenomics analyses to identify the most common methanogenic species and the presence of genes encoding enzymes involved in the three predominant ruminal methanogenesis pathways.Moreover, we determined if alternative H 2 sinks were engaged under conditions where methanogens were inhibited by AT and if AT had any direct impact on bacterial populations.Together, our results reveal an unexpected but significant contribution of methanol-utilizing Methanosphera toward ruminal methanogenesis, providing essential insight into the mechanistic basis of AT and the development of AT resistance mechanisms among ruminal methanogens.In addition, we identify both direct and indirect effects of AT on ruminal bacteria and fermentation pathways that could impact animal health and productivity and must therefore be considered before adopting such mitigation strategies.

RESULTS
Twenty Holstein cows were randomly assigned to four [control, low 0.25% AT (LAT), high 0.5% AT (HAT), and oregano (O), another inhibitor of methanogenesis] treatments in a replicated 4 × 4 Latin square design in which each dairy cow was rotated between the four treatments in four periods.Toward the end of each period, enteric CH 4 emissions were measured, and the results were published (13).The effects of AT supplementation at 0.5% of dietary dry matter (HAT) on dairy cattle in a Latin square design inhibited enteric CH 4 emissions by 55% in periods 1 and 2, but the inhibitory effect of AT gradually declined by periods 3 and 4 (11).In contrast, LAT and oregano supplemen tation had no effect on enteric CH 4 emissions in this study (11).To determine the effects of these treatments on the rumen microbiome, a source of enteric CH 4 formation, ruminal samples collected using the stomach tube method during the prior study were isolated by filtering through three-layered cheesecloth, extracted for genomic DNA, and processed for 16S rRNA and metagenomics (metaG) in the current study.

Sequencing information
To assess the diversity within the rumen archaeal community that could be contribu ting to methanogenesis, 930,776 raw partial 16S rRNA sequences were acquired from the rumen samples of 79 cows that were fed either a control diet (20), high AT (20), low AT (19), or oregano (20).Following rigorous quality filtering and denoising, 869,942 sequences were retained, with read counts per sample ranging from 2,490 to 20,134, identifying 192 unique amplicon sequence variants (ASV).Likewise, in the rumen bacterial community, 5,159,552 raw partial 16S rRNA sequences were obtained from the same 79 samples.After quality filtering and denoising procedures, 3,345,832 sequences were retained, with read counts per sample varying from 17,524 to 71,193, identifying 25,082 unique ASV.To assess AT-induced changes in microbiome function, we utilized shotgun metagenomic analysis.The use of the Illumina HiSeq platform generated 534,618,479 sequences across 60 samples.After quality filtering, approximately 11% of reads were removed, resulting in 473,192,284 high-quality reads, with sample sequences ranging from 3,756,824 to 12,019,902.Taxonomy assignment using Kraken2 revealed that approximately 96% of sequences were attributed to bacteria and roughly 2% each to archaea and eukaryotes.Notably, viruses constituted a minor fraction of the data set.

Changes in methanogenic archaeal communities in response to AT supple mentation
To understand the mechanistic basis of AT (0.5% of dry matter intake) in curbing enteric CH 4 emissions, we first investigated its effects on the ruminal methanogens that produce methane.First, methanogenic communities in the rumen of dairy cows over four experimental periods were compared at the community level (Fig. 1A), as assessed by the presence of commonly present populations (weighted UniFrac) and unique populations within each treatment (unweighted UniFrac; Fig. S1).Overall methanogenic communities differed by treatment, period, and their interaction in both the weighted and unweigh ted UniFrac analysis based on permutational multivariate analysis of variance (P < 0.05; PERMANOVA test, Table 1).Further pairwise comparisons revealed significant variations in community composition for certain combinations of treatments and periods (Table 1).In period 1, the LAT treatment exhibited a significant difference compared to the control group.Similarly, in period 2, the control group was significantly different from the HAT, LAT, and oregano groups.While oregano and LAT treatments displayed significant differences in community composition compared to the control group in period 3, no differences between treatments were noted in period 4. When considering the unweighted UniFrac matrix, significant differences were observed between the control and HAT groups in periods 1, 2, and 3, but the differences faded away by period 4. Thus, while large variations among commonly present populations by treatment and period interactions were observed, the less abundant methanogen populations were consistently reduced by HAT treatment during times at which reduced enteric CH 4 emission was observed.
Second, we compared the taxonomy of individual methanogens using 16S rRNA sequencing analysis.We found that the archaeal genera Methanobrevibacter and Methanosphaera dominated the microbial community, collectively accounting for approximately 99% of the observed sequences (Fig. 1B).We observed a significant increase in the relative abundance of Methanobrevibacter in both the HAT and LAT treatment groups compared to the control during period 1 (P < 0.05).However, this difference became less pronounced in subsequent periods, with no statistically signifi cant variations observed during period 2 or period 3 (P > 0.05).In contrast, the relative abundance of Methanosphaera decreased significantly in the HAT group during period 1 (P < 0.05), indicating a strong inhibitory effect of AT.However, during period 2, we observed a restoration of the relative abundance, although with considerable variation (P < 0.05).By period 3, there were no significant differences between treatments (P > 0.05), suggesting a potential adaptation to the high AT conditions.The Oregano treatment inhibited Methanobrevibacter but Methanosphaera was not affected in periods 1 and 2. When HAT was effective in inhibiting methanogenesis (in periods 1 and 2), methanolutilizing Methanosphaera were almost completely eliminated, while no changes in Methanobrevibacter were noted.Interestingly, Oregano, which did not reduce CH 4 in our study, did not alter Methanosphaera compared to control, further highlighting the correlation between decreased CH 4 emissions and reductions in Methanosphaera.
Finally, we conducted a comprehensive analysis of the most common methanogenic species within Methanobrevibacter and Methanosphaera using shotgun metagenomics data (Fig. 1C).Among the Methanobrevibacter species, we identified six species, among which Methanobrevibacter ruminantium, Methanobrevibacter ollyae, and Methanobrevi bacter millerae were the most dominant.M. ruminantium and M. olleyae showed a slight increase in the HAT group, although this increase did not reach statistical significance (P > 0.05).In contrast, the remaining four Methanobrevibacter species displayed a reduction, with M. millerae and M. YE315 showing significant (P < 0.05) decreases in cows in the HAT group.Furthermore, both Methanosphaera species exhibited reduced abundance in the HAT group during period 1, confirming the inhibitory effect of AT on these species.While 16S rRNA analyses revealed that hydrogenotrophic methanogens were minimally impacted or were increased by AT, the disappearance of Methanosphaera in the initial phase of the study again supported their importance in AT-associated reduction in enteric CH 4 emissions.Consistent with this idea, prolonged feeding of AT beyond 5 weeks (into period 2) led to a rebounding effect in Methanosphaera, which was followed by the loss of AT function in reducing enteric CH 4 emissions.

Impact of AT on methanogenesis pathways
Using metagenomic shotgun data, we quantified the genes that code for enzymes involved in the three predominant ruminal methanogenesis pathways (CO 2 -, methanol-, and methylamine-reducing pathways) in cows with and without AT supplementation (Fig. 2).In addition, the taxonomy of the annotated genes was tracked to assess the role of individual methanogenic lineages in methanogenesis.

CO 2 -reducing methanogenesis pathway
The CO 2 -reducing pathway catalyzes the conversion of CO 2 and hydrogen gas (H 2 ) into CH 4 .The CO 2 -reducing pathway encompasses eight steps (Fig. 2A), with steps 1-5 being unique to this pathway and steps 6-8 shared among all three methanogenic pathways.
As we can successfully identify all the genes encoding the enzymes involved in the CO 2 -H 2 O pathway through metagenomic analysis, we next assessed the gene copy number of specific enzymes within this pathway across all treatments by periods (Fig. 2).Notably, the copy number of the gene encoding the enzyme EC: 1.2.7.12 (representing step 1 in Fig. 2A), which is involved in the reduction of CO 2 to formylmethanofuran, exhibited the highest abundance among all the enzymes involved in steps 1-5.In periods 1 and 2, the gene copy number for this enzyme was significantly lower in the AT-and Oreganosupplemented cows compared to the control cows (P < 0.05 for both periods).However, in period 3, similar abundances were observed across all groups (P > 0.05), consistent with the loss of AT's effect on methane inhibition.Additionally, the gene copy number for enzymes EC: 2.3.1.101(step 2), 3.5.4.27 (step 3), 1.5.98.1 (step 4), and 1.5.98.2 (step 5) was significantly reduced (P < 0.05) in the AT-and Oregano-supplemented cows compared to the control cows in both periods 1 and 2; however, the differences between treatments were lost by period 3 where the effect of AT on enteric methane formation was lost.

Impact of AT on methanol-and methylamine-utilizing pathways
In the methanol-utilizing pathway (Fig. 2B), a notable difference was observed in the copy number of genes encoding the enzyme methanol-corrinoid protein co-methyl transferase (mtaB; EC: 2.1.1.90)when compared to the other two enzymes (mtaA; EC: 2.1.1.246and Unclassified enzyme).The gene copy number of all three enzymes exhibited a significant reduction (P < 0.05) in HAT compared to control in period 1.However, all three enzymes appeared to increase in period 2 compared to period 1 in HAT samples, indicating a rebounding effect.While the unclassified genes and mtaB did not show any significant difference, the gene copy number for EC: 2.1.1.246was still lower (P < 0.05) in HAT compared to control samples.By period 3, all genes rebounded, with no significant (P > 0.05) difference between the control and HAT groups.
Two species of Methanosphaera were predominantly associated with the methanolutilizing pathway (Dataset S1B), with Methanosphaera stadtmanae DSM 3091 accounting for 39%, closely followed by Methanosphaera sp.BMS at 33%.Other contributors included methanogens such as Methanobrevibacter smithii ATCC 35061 (13%), Methano genic archaeon ISO4-H5 (1%), and bacteria (2%).Collectively, the five archaea species represented approximately 93% of the identified methanol-utilizing archaea.A notable observation was the intriguing absence of the two Methanosphaera species, Methanos phaera stadtmanae DSM 3091 and Methanosphaera sp.BMS, within the HAT group during period 1, which is in agreement with the loss of genes coding for enzymes involved in the methanol pathway.However, as the two species began to rebound in period 2, this was also accompanied by changes in gene copy number in the methanolutilizing pathway, with the latter still being lower in HAT compared to control.By period 3, the effect of HAT was lost completely, as Methanosphaera were completely restored, and no differences were noted in either Methanosphaera species or the genes involved in the methanol-utilizing pathway between control and HAT.Oregano did not have an effect on Methanosphaera or genes associated with the methanol pathway, suggesting the latter pathway was resistant to the effects of Oregano.A reduction in both Methanos phaera and methanol-utilizing genes is consistent with a reduction in CH4 emissions by HAT in periods 1 and 2. As Oregano had no effect on either CH4 emissions or Methanos phaera and its genes, our results suggest that Methanosphaera has a major role in ruminal methanogenesis.
In the methylamine-utilizing pathway (Fig. 2C), most of the genes coding for enzymes involved in the transfer of methylamines were found in greater copy number relative to those involved in the transfer of dimethyl or trimethylamines, although the total gene copy number was much lower than those involved in other pathways.Across all animals, copies of the gene coding for the enzyme methylamine-corrinoid protein co-methyl transferase (EC: 2.1.1.248)were negligible [<10 copies per million (CPM); mean].No differences in gene copy number were noted between treatment groups, suggesting that the impact of AT on methylamine-utilizing methanogens was negligible.

Impact of AT on methyl-coenzyme M reductase, the connecting point for all methanogenesis pathways
The enzyme MCR (EC: 2.8.4.1) plays a crucial role in CH 4 formation by catalyzing the incorporation of methyl coenzyme M (Co-M) and coenzyme B (Co-B), resulting in the production of a heterodisulfide and the release of CH 4 in the penultimate step (16).In this study, copies of the gene encoding MCR were found to be among the most abundant genes associated with methanogenesis pathways (Fig. 2A).
To compare the effects of LAT, HAT, and Oregano, the percent reduction in MCR gene copy number between the respective inhibitor and control was calculated (Table 2).These results revealed a significant decrease (P < 0.05) in the gene copy number of MCR in HAT compared to control, with reductions of 61% and 65% observed during periods 1 and 2, respectively.However, with the loss of AT's effect on methane inhibition, the gene copy number of MCR was increased by 19% in HAT compared to control by period 3.
As MCR is composed of three subunits, alpha (K00399), beta (K00401), and gamma (K00402), we next looked to see if AT preferentially targeted one of the three subunits to inhibit methane formation.In period 1, all three subunits were inhibited to a similar extent by AT with reductions of 62%, 61%, and 59% of K00399 (P < 0.05), K00401 (P = 0.05), and K00402 (P < 0.05), respectively, compared to the control group.Notably in period 2, despite a noticeable rebounding effect of the methanol-utilizing pathway, the gene copy number of MCR enzyme continued to be inhibited (P < 0.05), with a reduction of 66%, 60%, and 68% in HAT-supplemented cows compared to control.In contrast, in period 3, the gene copy number of these subunits rebounded (P > 0.05) in the HAT-supplemented group and was numerically higher compared to the control group.
We also investigated the archaea species associated with MCR genes (Table 3).Among the identified species, a total of six Methanobrevibacter species, including M. ruminantium M1, M. olleyae YLM1, M. sp.YE315, M. millerae SM9, M. smithii ATCC 35061, and M. sp.AbM4, along with two Methanosphaera species, M. sp.BMS and M. stadtmanae DSM 3091, collectively made up approximately 93% of the total archaea associated with MCR.As indicated in Table 3, the species with the highest percent contribution to MCR were Methanobrevibacter species, particularly M. ruminantium M1, which was higher in HAT and LAT compared to the other two groups.Methanosphaera was negligible in period 1 but proportionally increased by period 2 although the contribution was insignificant based on gene copy number.Because the activity of MCR gene transcripts is several folds higher than the corresponding genes (17,18), the MCR gene copy number may not be truly reflective of function.As such, the use of metatranscriptomics may shed more light on the contribution of individual methanogens to the MCR enzyme that catalyzes methane formation.

Hydrogenases regulating H 2 production under normal and inhibited methanogenesis
The inhibition of methanogenesis by AT resulted in a notable sevenfold increase in gaseous H 2 concentrations compared to the control, as reported by Stefenoni et al. (11) and Table 2, consistent with the loss of an H 2 sink.Therefore, we sought to ascertain whether the dynamics in H 2 concentrations following AT supplementation were associated with discernible changes in hydrogenase activity.Hydrogenases are metalloenzymes that play a pivotal role in converting H 2 to 2[H] + 2e.They are broadly categorized into [FeFe], which plays a role in sensing H 2 concentrations and H 2 production; [NiFe], which facilitates H 2 uptake; and [Fe], the function of which remains currently unknown.Within the [FeFe] category, enzymes are further differentiated into A1-A4, B, and C1-C3 groups, with the former two groups regulating H 2 production and the latter sensing H 2 concentrations.
Our findings revealed that, within the scope of this study (Fig. S2), the predominant [FeFe]A3 were inhibited in response to inhibited methanogenesis by HAT, and the spared H 2 was sensed by an increase in sensory hydrogenases such as [FeFe]B in the HAT treatment compared to other treatments, in both periods 1 and 2. However, the effect was lost by period 3. A reduction in H 2 -utilizing hydrogenases such as [NiFe]3a, 3c, and 4d, which are abundant in CO 2 -reducing Methanobrevibacter and Methanosphaera, is expected under conditions of inhibited methanogenesis.Noteworthy is the increase in copy number of [NiFe]4a in HAT compared to control, as this group of hydrogenases is associated with formate reduction.Whether this formate is fermented by bacteria or utilized by methanogens remains unknown.

Alternative H 2 sinks under inhibited methanogenesis
We next determined if the H 2 produced during AT-mediated inhibition of methanogene sis was redirected toward alternative H 2 sinks (Dataset S2), potentially engaging in direct or indirect competition with methanogens.Methanogenesis emerged as the primary H 2 sink, demonstrating a significant reduction (P < 0.05) in the HAT-treated group during period 1 and a numerical decrease in period 2, with no observable differences in period 3.In addition, respiratory hydrogenases and bifurcating hydrogenases exhibited a notable decrease (P < 0.05) in the HAT-treated groups during periods 1 and 2, although this trend was not evident in period 3 (P > 0.05).Alternative and sensory hydrogenases, along with cofactor-coupled bidirectional hydrogenases, showed a numerical tendency to decrease in the HAT group during periods 1 and 2, with no substantial change in period 3. Energy-converting hydrogenases, fermentative hydrogenases, and sensory hydrogenases displayed relatively consistent patterns across the study periods.These data imply that none of the recognized alternative H 2 sinks increased under conditions in which methanogenesis was inhibited by HAT.However, the total hydrogenases were greatly reduced in HAT compared to control and Oregano in periods 1 and 2, indicating that H 2 metabolism was perturbed in HAT treatments.Application of metatranscriptom ics may shed more insights into the functionality of different H 2 sinks and how H 2 is regulated under inhibited methanogenesis by AT.

Impact of AT on bacterial populations
Because bacteria-methanogen interactions are fundamental to the integrity and functionality of the rumen microbiome, and AT inhibited specific groups of methanogens more than others, we hypothesized that AT may have both direct and indirect effects (mediated via spared H 2 under inhibited methanogenesis by AT) on rumen bacteria and fermentation pathways leading to differences in volatile fatty acid (VFA) production.At the phylogeny level, 16S rRNA sequencing analysis across all samples revealed that the dominant bacterial phyla were Firmicutes, Bacteroidetes, and Actinobacteria, collectively accounting for 96% of the bacterial community (Dataset S3A).Within the Firmicutes phylum, the unclassified Clostridiales (1 and 2) Butyrivibrio were the most abundant, constituting more than 36% of the community.Among the Bacteroidetes phylum, the genus Prevotella emerged as the most dominant, representing 18% of the community (Dataset S3B).To compare the taxonomic composition across all cows, a total of 82 bacterial taxa were examined, each with an abundance threshold of at least 0.1% in one or more samples.Interestingly, 18 bacterial taxa exhibited significant differences (P < 0.05; Proc Mixed) in abundance between AT-supplemented cows and the control group (Dataset S3B).Among these identified taxa, nine showed an exclusive treatment effect, while three demonstrated a combined effect of treatment and period interaction.Additionally, three taxa were influenced by both treatment and period interactions.Notably, these included three taxa, namely Butyrivibrio, unclassified Eubacterium, and Moryella, known for producing the short-chain fatty acid butyrate, all of which were more abundant in cows treated with HAT compared to the control group.While shifts in bacteria following the inhibition of methanogenesis noted by period interactions with treatments were expected, the direct effects of HAT on bacteria even after the loss of its inhibitory effects on CH 4 production are noteworthy and warrant further investigation.

Effect of AT on fermentation pathways
As we hypothesized that H 2 spared under inhibited methanogenesis by AT would directly and indirectly impact bacterial fermentation, we next explored variations in fermentation pathways that contribute to VFA production by tracking changes in total and individual VFA pathways and the bacteria with which they were associated across treatments and periods (Fig. 3; Fig. S3).Surprisingly, concentrations of total VFA were consistently lower in HAT compared to all other treatment groups throughout the study, indicating that the effects of HAT on VFA production were not dictated by the degree of inhibition of methanogenesis.Overall, the molar proportion of acetate was reduced and those of propionate, butyrate, and valerate increased (P < 0.05) in HAT compared to control across all periods.To determine which bacteria might be contributing to VFA production, we initially conducted a correlation analysis between the molar proportions of individual VFA and bacterial populations identified through 16S rRNA sequencing across all samples.Bacterial taxa exhibiting significant differences (P < 0.05) between the control and AT-treated groups across all sampling periods were selected for correlation analysis with fermentation parameters.Our findings revealed a positive association between specific bacterial genera from the Firmicutes phylum, such as Eubacterium, Moryella, and Butyrivibrio, and the proportions of butyrate and valerate (Fig. S3), which were significantly increased in HAT treatment compared to other treatments across all periods.Subsequently, we investigated the presence of genes encoding enzymes involved in the distinct pathways leading to butyrate formation (Fig. 4).Noteworthy disparities in gene copy number were observed between the control and HAT treatment groups for enzymes participating in the butyrate pathway.Particularly, genes encoding the enzyme EC: 1.3.8.1, responsible for the conversion of crotonyl-CoA to butyryl-CoA, exhibited a significant increase (P < 0.05) in HAT samples compared to the control group in periods 1 and 2. These findings align with the observed increase (P < 0.05) in the molar proportions of butyrate in HAT samples compared to the control group (Fig. 3).The copy number of genes encoding the enzyme EC:2.8.3.8, which is responsible for the conversion of butyryl-CoA to butyrate (BP1 pathway), and EC:2.3.1.19and EC:2.7.2.7, contributing to the conversion of butyryl-CoA to butyrate via the butanoyl phosphate (BP2 pathway), demonstrated numerical increases in the HAT treatment group compared to the control.However, these differences did not reach statistical significance.Identified bacteria associated with the butyrate pathway are detailed in Dataset S1F.

Interactions within the rumen microbiome in different periods
As the effect of HAT on CH 4 inhibition was noted in periods 1 and 2, but faded away by periods 3 and 4 and was accompanied by similar responses in Methanosphaera (except in period 2) as well as both direct and indirect effects on bacterial populations, we next performed a correlation analysis by period (Fig. S4A).As expected, CH 4 emission was negatively correlated with H 2 concentrations, propionate, and valerate, consistent with a negative correlation with rapid fermenting bacteria such as Sharpea, Butyrivibrio, Lactobacilli, Moryella, and Eubacterium and positively correlated with acetate and Ruminococcaceae bacterial lineages.As CH 4 was negatively correlated with H 2 , the responses of H 2 were opposite to those observed for CH 4 .Specifically, the most abun dant methanogen Methanobrevibacter showed a positive correlation with H 2 , butyrate, and valerate and negative associations with acetate and all bacteria associated with acetate, while Methanosphaera showed the opposite pattern.Within the abundant bacteria, there were both positive and negative associations observed in period 1.While the associations noted in period 1 were expected (with both normal and inhibited methanogenesis samples included), these patterns were completely lost in period 2 (Fig. S4B).None of the bacteria seemed to correlate among themselves or with either VFA or CH 4 emissions.This could be attributed to carryover effects of the previous treatments, plus possible adaptation mechanisms in AT treatments.Notably, in period 3 (Fig. S4C), when the effects of AT are lost (normal methanogenesis across all treatments), the rumen microbiome appeared to restore to normalcy, with CH 4 emissions positively correlated with acetate and Methanosphaera and negatively correlated with propionate, butyrate, valerate, and Methanobrevibacter.Both positive and negative associations among bacterial lineages were noted, signifying the normal function of the rumen microbiota.In period 4 (Fig. S4D), much more intense interactions among bacteria were noted, with only limited interactions with methanogens.Finally, as there were 20 animals including first and 2+ lactation cows, the cow-to-cow variation was notable as shown in the sequence plot (Fig. S5) for CH 4 emissions, changes in methanogens, and the MCR enzyme.With only five animals per treatment, and 20 animals rotated among four treatments in four periods with 28 days per period, some of the significant findings, particularly, in period 2, could not be substantiated.Overall, the positive association between CH 4 emissions and Methanosphaera and MCR enzyme was pronounced in periods 1 and 3, showing that the inhibition of CH 4 is associated with near elimination of Methanosphaera, which was reversed in period 3.

Effect of AT on methanogenic isolates
While it was evident from the metagenomic data that AT was selective for Methanos phaera, with Methanobrevibacter either showing no change or increased abundance, we next sought to validate our findings using pure methanogenic isolates.In our previous study (K.S. Narayan, A. C. B. Johnson, N. Indugu, J. Bender, H. A. Stefenoni, A. N. Hristov, A. Melgar, and D. Pitta, unpublished data), we assessed the methane-emitting potential of two pure isolates: CO 2 -utilizing Methanobrevibacter ruminantium MI and methanol-utiliz ing Methanosphaera stadtmanae.We reported that Methanosphaera stadtmanae makes three times more CH 4 and grows faster (<12 h) than does M. ruminantium M1, which grows between 12 and 24 h in anaerobic Hungate tubes under laboratory conditions.To test the effect of bromoform, we added bromoform at both lower and higher concentrations than the concentration of bromoform reported in AT (approximately 65 µg/mL of media).Notably, we found that following the addition of bromoform up to 65 µg/mL, Methanobrevibacter ruminantium M1 reached its maximum CH 4 -emitting potential within 12 h, with only a marginal increase between 12 and 24 h.In contrast, in control Hungate tubes of Methanobrevibacter ruminantium M1 lacking bromoform, CH 4 emission was lower at 12 h but reached its maximum by 24 h.These data reveal that the addition of bromoform did not affect the total CH 4 emissions from Methanobrevibacter ruminantium M1, but instead accelerated its production (Fig. 5).Interestingly, increasing the dose of bromoform added to Methanobrevibacter ruminantium M1 culture tubes to 130 and 260 µg/mL, resulted in the inhibition of methane emission by 16% and 34%, respectively, at 12 h.This reduction was further reduced by 50% when measured at 24 h, revealing that the rapid changes that Methanobrevibacter ruminantium M1 undergoes when exposed to bromoform are only transient.The interaction between bromoform and individual ruminal isolates have not been previously described and hence the mechanistic basis remains unknown.In contrast, methane emissions from M. stadtmanae were gradually inhibited with increasing the dose of bromoform from 7% to 22% at 12 h and 6% to 40% at 24 h compared to control.These data clearly sup port findings derived from omic analysis that AT has selective preference for inhibiting Methanosphaera.Because the mechanisms of AT on individual methanogenic isolates are unknown, and HAT supplementation (dose at 65 µg/mL) resulted in a near elimination of Methanosphaera, other active compounds within AT that may be capable of suppress ing Methanosphaera may exist.These data highlight the need to further investigate the mechanistic basis between AT and methanogen interactions to determine the role of Methanosphaera in ruminal methanogenesis and the mechanistic impact of AT on Methanosphaera.

DISCUSSION
Enteric CH 4 formation results from energyinefficient fermentation of feed by microbiota within the bovine rumen.To reduce the contribution of enteric CH 4 emission to global warming, seaweeds such as AT that are predicted to knock out methanogens have been utilized.However, the effects of AT appear transient, and bromoform, the active ingredient that suppresses CH 4 formation, appears to lose its functionality over time (11).To gain a deeper understanding of the mechanistic basis underlying the effective yet transient effects of AT as a CH 4 inhibitor, we undertook a comprehensive assess ment of AT-dependent alterations in rumen microbiome diversity and gene content.Moreover, as AT is likely to interfere with symbiotic interactions between H 2 -producing microbes and H 2 -utilizing methanogens, we identified both the direct as well as indirect effects of AT (via spared H 2 concentrations under inhibited methanogenesis) on rumen microbiota.Together, our findings provide mechanistic insight into AT-driven reduction in methane emissions and elucidation of additional pathways that may need to be targeted to maintain its inhibitory effects, while preserving microbiome health and animal productivity.
During period of AT-inhibited methanogenesis, methanol-utilizing Methanosphaera were almost completely eliminated, while only limited inhibitory effects were noted on ruminal hydrogenotrophic and methylamine-utilizing methanogens.Selective inhibition of Methanosphaera by HAT raises two important questions: first, what is the mechanistic basis underlying the effect of AT on Methanosphaera?Second, does the association between reduced CH 4 emissions and the specific elimination of Methanosphaera in HAT-fed cows suggest that Methanosphaera has a greater role in enteric methane formation than previously thought?AT contains a number of halogenated CH 4 analog components, including bromoform that likely inhibits CH 4 formation by interfering with the transfer of methyl groups by methyl transferases and MCR during Wolfe's cycle of methanogenesis.Vitamin B12 forms the central core of methyltransferases that transfer methyl group from methanol/methyl amines to coenzyme M, and bromoform likely inhibits methylotrophic methanogens via interfering with vitamin B12 synthesis.However, as diverse groups of methyl transferases are also present in hydrogenotrophic as well as methylamine-utilizing methanogens that are resistant to the effects of AT, it suggests additional, as yet unknown, mechanisms underly the inhibitory effects of AT on CH 4 formation.In support, increasing the dose of bromoform to an activated culture of Methanosphaera stadtmanae did not completely inhibit methanogenesis, suggesting that additional compounds contribute to the selective preference of AT for Methanos phaera.
Enteric CH 4 emissions significantly increase 2-4 h following feed intake in dairy cows (19), and we reported that Methanosphaera lineages dominate during this time when compared to Methanobrevibacter lineages (20), further supporting our finding that higher CH 4 formation is strongly associated with an increase in the abundance of Methanosphaera.In prior studies (18), we reported that the metabolic activity of Methanosphaera, as revealed by metatranscriptomics, is approximately fivefold higher than what was estimated from the gene content of Methanosphaeraound using metagenomics.In a report (Narayan et al., unpublished), we found that a pure isolate of Methanosphaera stadtmanae grows faster (12 vs 24 h) and produces three times more CH 4 than the same cell mass of Methanobrevibacter ruminantium.Furthermore, we (18) reported that hydrogenotrophic methanogens, including Methanobrevibacter ruminantium, were more sensitive to the MCR inhibitor 3-NOP (an analog of methyl coenzyme M) than were methylotrophic methanogens, resulting in a 30% reduction in total CH 4 emissions (21).These data, in combination with our new data revealing that AT inhibition of enteric CH 4 emissions by >50% is accompanied by the elimination of methanol-utilizing Methanosphaera with limited effect on other methanogens, suggest that Methanosphaera may have a greater share in total CH 4 formation than previously thought.These new findings lay the foundation for targeting novel mitigation strategies and identifying functionally distinct complementary inhibitors that will help curb enteric CH 4 formation.
Notably, our findings provide some insight into the transient and variable effects of AT on CH 4 emission.Specifically, in period 2, when inhibition of CH 4 emission by HAT was at 55%, Methanosphaera more than doubled (in absolute sequence numbers based on 16S rRNA sequence reads, and also metagenomic reads; Dataset S4) compared to control samples.Increased copy number of Methanosphaera to variable extents may indicate the onset of resistance mechanisms in response to AT. Halogenated compounds are either metabolized or excreted in the gastrointestinal tract of animals (22,23), and methanogens are specifically able to "dehalogenate" compounds, with bromoform being dehalogenated faster than chloroform (24).Moreover, both coenzyme M methyltrans ferases and MCR can carry out reductive dehalogenation (25).Thus, the observed doubling of Methanosphaera by the end of period 2 may reflect the development of AT resistance through increased capacity to dehalogenate bromoform.Although bromoform concentrations in the rumen were not measured, milk collected at the end of periods 1 and 2 had higher concentrations of bromide, which forms following the dehalogenation of bromoform.However, it remains to be determined if the dehaloge nation of bromoform by methanogens as speculated here or the sudden decrease of bromoform concentrations in seaweed as described in Stefenoni et al. (11) better explains the decrease of bromide in milk and may require more frequent sampling to help understand methanogen-bromoform interactions in the rumen.Furthermore, there may be other halogenated compounds such as iodoform that are as potent or more potent than bromoform for inhibiting methanogens.Regardless, Methanosphaera appears to have evolved mechanisms that help it overcome the inhibitory effects of AT, as reflected by their increased abundance toward the end of period 2.
In response to methane inhibition in HAT-supplemented cows, Stefenoni et al. (11) reported that H 2 concentrations were three-to sixfold higher in treated cows.Surpris ingly, we observed a decrease in both A1 (reduced-ferredoxin dependent hydrogenases) and A3 (bifurcating) hydrogenases in HAT compared to control.While such findings are not consistent with an increase in H 2 gas, the concentration of dissolved H 2 was not measured in this study.Previously, we reported an increase in dissolved H 2 concen tration in cows that were supplemented with the MCR inhibitor, 3-NOP.This increase was accompanied by a decrease in A3 bifurcating [FeFe] hydrogenases, with a concom itant increase in A1 ferredoxin only [FeFe] hydrogenases.We also observed a signifi cant increase in group C3 hydrogenases, suggesting that increased H 2 concentrations were sensed by H 2 -producing bacteria, that then switched their hydrogenases from A3 (bifurcating) to A1 (ferrdoxin-dependent).Such findings are in line with spared H 2 increasing the partial pressure of H 2 .Because the 3-NOP experiment was a continuous study, changes in hydrogenases followed the flux in spared H 2 concentrations under inhibited methanogenesis by 3-NOP over a 15-week period.Our inability to associate changes in free H 2 gas with shifts in hydrogenases in the current study may be due to our rotating Latin square experimental design in which different animals were assigned to HAT in each period.Because the rumen microbiome differs between cows, it is reasona ble to expect variations in hydrogenases, making it more difficult to detect patterns.It may also be that the measurement of dissolved H 2 correlates with hydrogenases rather than free H 2 gas concentrations.Finally, cows that were assigned to HAT treatment had lower dry matter intake and lower milk production, potentially explaining why we saw a decrease in both A1 and A3 hydrogenases, which reflect an overall reduction in fermentation patterns.As all of these issues could contribute to our inability to detect associations, continuous experiments for prolonged periods of feeding AT will be required to provide a mechanistic basis in connecting methane mitigation, spared H 2 , hydrogenases, and shifts in fermenting microbes in the rumen.Finally, as Methanos phaera is inhibited to a greater extent than is Methanobrevibacter in periods 1 and 2, it is possible that only a small degree of H 2 is spared (only 1 mole of H 2 is needed to reduce methanol), which agrees with the changes noted in hydrogenases.In addition, it is possible that AT directly impacts hydrogenotrophic bacteria to spare H 2 in addition to inhibiting methanogens.Complementing metagenomic data with gene expression of hydrogenases using metatranscriptomics may provide a better picture of the link between hydrogenases and actual H 2 production.
In addition to the expected indirect impact of AT on bacteria within the microbiota, we identified some direct effects that were less expected.Specifically, our bacterial analysis revealed 18 genera that were significantly increased in HAT-treated animals compared to control, even when the inhibitory effect of AT on methanogenesis was lost.Some of these bacteria include Butyrivibrio, Eubacterium, and Roseburia species, which belong to Clostriales XIV cluster, all known butyrate producers.Stefenoni et al. (11) previously reported that butyrate was significantly increased in HAT treatments compared to control.Butyrate synthesis involves the condensation of two acetate to acetoacetate and finally, acetoacetylcoA, which is then converted to crotonyl coA.A critical step in butyrate synthesis is the conversion of crotonyl-CoA to butyryl-CoA, which is catalyzed by a flavinbased electron bifurcation complex (butyryl-CoA dehydrogen ase/electron-transferring flavoprotein complex; BcdA-EtfBC) (26).This step is irreversible and may not allow H 2 -forming butyrate producers to switch fermentation patterns.In the current study, butyrate production was consistently increased, whereas acetate to propionate ratio decreased (11) with a concomitant increase in gene copies of butyryl-CoA dehydrogenase, suggesting that butyrate synthesis may be an alternate sink to inhibited methanogenesis by AT.In the terminal step of the butyrate synthesis pathway, the conversion of butyryl-CoA to butyrate can occur via two distinct pathways, one mediated via the butyrate kinase pathway (BP1) and is mostly observed in Clostridia and the other mediated via the butyryl-CoA: acetate-CoA-transferase pathway (BP2) predominant in Negativicutes (27).As we found that HAT specifically increased the genes coding for the enzyme phosphate butyryltransferase (EC 2.3.1.19)(in periods 1 and 2 with no differences in periods 3 and 4), it can be inferred that the increase in butyrate does not depend on increased acetyl CoA in the context of inhibited methanogenesis but instead occurs via the direct stimulation of certain bacterial populations such as Clostridia.It remains unclear if the AT effect is due to bromoform or other compounds in AT that may be stimulatory, which may explain why the effect was significant in periods 1 and 2, while it is only marginal in periods 3 and 4, with an overall increase in butyrate formation.In contrast, we found that the gene copies (Fig. 2D) of acetate-CoA transferase (EC: 2.8.3.8) of the BP2 pathway increased in period 3, although not significantly, and is numerically higher in AT treatment compared to control.This is also in agreement with a significant increase in butyrate molar proportions in periods 1 and 2, but only marginal increases in periods 3 and 4. Therefore, the increase in butyrate synthesis as an alternate sink to inhibited methanogenesis by AT may not be accurate but may reflect a direct effect of HAT on butyrate-producing bacteria.
While there have been studies highlighting the potential of seaweeds, particularly AT (red seaweed) to significantly curb CH 4 emission (60%-90%), risks associated with supply and side effects arising with increasing doses of AT have also been discussed (24,28).Animal health concerns with feeding A. taxiformis should be considered.As discussed in reference (29), bromoform is categorized as a "potential human carcinogenic" compound by the EPA, and its impact on animal health is unclear (30).A summary of toxicological risk to animals and humans concluded that at low inclusion levels, A. taxiformis did not cause problems for ruminant animals or consumers through their products (24).In previous work, we conducted a milk sensory panel from a study where A. taxiformis was fed to dairy cattle (11), in which consumers were unable to distinguish milk from cows fed A. taxiformis from control milk, but the difference approached a trend (P = 0.11), with 39% of participants correctly identified milk from A. taxiformis cows as different from milk from control cows.Additionally, milk from cows fed A. taxiformis had five to eight times higher concentrations of iodine and bromide, respectively.Whereas the risk of bromide toxicity is unclear, the increase in iodine concentration in milk could have significant impacts on human health, particularly in persons with abnormal thyroid function.Given the high risk and high reward of AT in tackling methane mitigation from livestock globally, understanding the mechanistic basis of AT on ruminal methanogene sis, rumen microbial metabolism, and their collective long-term impact on animal health and productivity is pivotal to launching seaweeds as feed additive to livestock.This study provides the basis for the mode of action of AT, albeit transient, and provides essential insight for further investigation assessing inhibitor interactions with methanogens, the source of enteric CH 4 formation in livestock.

Animals and experimental design
The present study serves as an accompaniment to the animal investigation outlined in Stefenoni et al. (11).The detailed account of the animals and experimental design can be found in Stefenoni et al. (11).In summary, the study adopted a replicated 4 × 4 Latin square design, ensuring balance for residual effects.A cohort of 20 Holstein cows, comprising 4 primiparous and 16 multiparous individuals, with an average (±SD) of 2.6 ± 1.19 lactations, 95 ± 22.0 days in milk (DIM), and a starting milk yield of 42.2 ± 2.59 kg/day, were organized into five groups based on parity, DIM, and milk yield.The experimental design spanned four periods, each lasting 28 days, with a 21-day adaptation phase followed by a 7-day period for data and sample collection.Cows within each group were randomly assigned to one of the four treatments: control (basal diet without additives), 0.25% AT (LAT), 0.50% AT (HAT), or 1.77% O (oregano leaves).All cows were subjected to the same basal diet and received AT and O in a premix containing ground corn grain and wheat middlings, which was incorporated daily into the total mixed ration.The premix, stored at 4°C, was prepared biweekly, and cows received the full dose of AT and O from day 1 of each experimental period.

DNA and RNA extraction, PCR amplification, and sequencing
The genomic DNA from solid ruminal samples was extracted using the repeated bead beating and column (RBB + C) method followed by extraction with the QIAmp Fast DNA Stool Mini Kit (Qiagen Sciences; Germantown, MD, USA) as described in reference (31).The extracted genomic DNA, both the V1-V2 regions of the bacterial 16S rRNA gene and the V6-V8 regions of the archaeal 16S rRNA gene, was PCR amplified in triplicate.The bacterialspecific primers used were F27 (5′-AGAGTTTGATCCTGGCTCAG-3′) and R338 and an interaction term for treatment and period.Period served as the repeated variable, with cow and treatment designated as subjects, and an autocorrelation structure of type "ar1" was employed as the covariance structure.Subsequently, least-squares means for "Treatment, " "Period, " and their interaction were computed.
In the analysis, relative abundances were employed for the assessment of archaeal genera, while CPM values were used for the evaluation of KEGG pathway genes and enzymes.For the analysis of bacterial genera, log-transformed values were applied.In instances of multiple tests, P-values were adjusted to control the false discovery rate, using the Benjamini-Hochberg method.
Project administration, Resources, Supervision, Writing -original draft, Writing -review and editing.

FIG 1
FIG 1 Assessment of rumen archaeal diversity and composition across different treatments by period.Treatments include control (C), HAT-, LAT-, and Oregano (O)-treated cows.(A) Principal coordinates analysis (PCoA) depicting weighted UniFrac distances of 16S rRNA archaeal compositions across periods 1, 2, 3, and 4. (B) Boxplots illustrating the abundance of the most prevalent archaeal genera based on 16S rRNA analysis.(C) Boxplots display the abundance of the most prevalent archaeal genera based on metagenomics analysis.Data significance is indicated as follows: NS, no statistical significance in generalized linear model; *P < 0.05; **P < 0.01; and ***P < 0.001.

FIG 3 FIG 4
FIG 3 Effect of Asparagopsis taxiformis on rumen pH and VFA in lactating dairy cows across different treatments by period.

TABLE 1
PERMANOVA analysis for 16S rRNA archaeal amplicon sequencing data a

TABLE 2
Effect of Asparagopsis taxiformis on enteric methane yield, hydrogen emission, and genes (copies per million) encoding for methyl-coenzyme M reductase enzyme (EC: 2.8.4.1) in the rumen of dairy cows a a K00399, alpha subunit of MCR; K00401, beta subunit of MCR; K00402, gamma subunit of MCR; and C, control.

TABLE 3
Mean values of methyl-coenzyme M reductase enzyme (EC: 2.8.4.1) and the contribution of individual methanogenic archaea (copies per million) to MCR enzyme in the rumen of dairy cows under different treatments within each period