ef-A meta-analysis of methane mitigation potential of feed additives evaluated in vitro.

A systematic literature review of in vitro studies was performed to identify methane (CH 4 ) mitigation interventions with a potential to reduce CH 4 emission in vivo. Data from 277 peer-reviewed studies published between 1979 and 2018 were reviewed. Individual CH 4 mitigation interventions were classified into 14 categories of feed additives based on their type, chemical composition, and mode of action. Response variables evaluated were absolute CH 4 emission (number of treat-ment means comparisons = 1,325); total volatile fatty acids (VFA; n = 1,007), acetate (n = 783), propionate (n = 792), and butyrate (n = 776) concentrations; acetate to propionate ratio (A:P; n = 675); digestibility of dry matter (DM; n = 489), organic matter (OM; n = 277), and neutral detergent fiber (NDF; n = 177). Total gas production was used as an explanatory variable in the model for CH 4 production. Relative mean difference between treatment and control means reported in the studies were calculated and used for statistical analysis. Robust variance estimation method was used to analyze the effects of CH 4 mitigation interventions. In vitro CH 4 production was decreased by antibodies (−38.9%), chemical inhibitors (−29.2%), electron sinks (−18.9%), essential oils (−18.2%), plant extracts (−14.5%), plants inclusion (−11.7%), saponins (−14.8%), and tannins (−14.5%). Overall effects of direct fed microbials, enzymes, macroalgae, and organic acids supplementation did not affect CH 4 production in the current meta-analysis. When considering the effects of individual mitigation interventions containing a minimum number of 4 degrees of freedom within feed additives categories, Enterococcus spp. (i.e., direct fed microbial), nitrophenol (i.e., electron sink), and Leu-caena spp. (i.e., tannins) decreased CH 4 production by 20.3, 27.1, and 23.5%, respectively, without extensively, or only slightly, affecting ruminal fermentation and digestibility of nutrients. It should be noted, however, that although the total number of publications (n = 277) and treatment means comparisons (n = 1,325 for CH 4 production) in the current analysis were high, data for most mitigation interventions were obtained from less than 5 observations (e.g., maximum number of observations was 4, 7, and 22 for nitrophenol, Enterococcus spp., and Leucaena spp., respectively), because of limited data available in the literature. These should be further evaluated in vitro and in vivo to determine their true potential to decrease enteric CH 4 production, yield, and intensity. Some mitigation interventions (e.g., magnesium, Heracleum spp., nitroglycerin, β-cyclodextrin, Leptospermum pattersoni , Fructulus Li-gustri , Salix caprea, and Sesbania grandiflora ) decreased in vitro CH 4 production by over 50% but did not have enough observations in the database. These should be more extensively investigated in vitro, and the dose effect must be considered before adoption of mitigation interventions in vivo.


INTRODUCTION
Decreasing enteric CH 4 emission to improve animal efficiency has been a research focus since early studies reported up to 12% of gross energy intake losses by CH 4 production (Czerkawski, 1969;Moe, 1981;Johnson and Johnson, 1995).More recently, the rising interest of governments and the society in climate change has directed researchers to better understand rumen methanogenesis and develop strategies to decrease greenhouse gas (GHG) emissions by livestock, especially enteric CH 4 in ruminants (Hristov et al., 2013;Congio et al., 2021;Arndt et al., 2022).The effectiveness of mitigation strategies, however, is inconsistent and data are, in some cases, controversial.For example, Hegarty et al. (2021) classified a low to medium agreement for ef-A meta-analysis of methane mitigation potential of feed additives evaluated in vitro.
ficacy of most enteric CH 4 mitigation strategies tested in vivo.
In vitro systems are considered as a preliminary step in the investigation of novel nutrition and rumen manipulation interventions, and they can be used to screen large number of treatments and doses in a short period of time, compared with in vivo experiments (Flachowsky and Lebzien, 2012;Hristov et al., 2012;Vinyard and Faciola, 2022).Ruminal fermentation and digestibility data generated by in vitro techniques, however, do not usually accurately represent in vivo responses, as demonstrated by studies evaluating grass and corn silages differing in plant maturity (Macome et al., 2017a;Macome et al., 2018) and N fertilization (Macome et al., 2017b), as well as dietary starch sources and levels (Hatew et al., 2015).On the contrary, a strong relationship (adjusted R 2 = 0.94) for CH 4 production data between in vitro and in vivo systems was reported by Danielsson et al. (2017), and accurate predictions of in vivo CH 4 production using an in vitro gas production system have been reported by Ramin and Huhtanen (2012).Differences in fluid and particle dilution rates, feed substrate to rumen volume ratio, and the lack of absorption of fermentation end products (Hristov et al., 2012) are some of the reasons why in vitro data are more variable than, and in some cases not representative of, in vivo experimental data.Nevertheless, when CH 4 mitigation strategies are subjected to an in vitro evaluation, it is reasonable to assume that differences from control, although not representative in absolute terms, would be unidirectional with in vivo effects.This has been described, for example, in a meta-analysis by Brandao et al. (2020), in which the relationship between independent and dependent variables used in their models were similar for data collected from omasal sampling technique and dual-flow continuous culture system, even though the magnitude of the measured responses was different.
Considering the feasibility of in vitro systems to preliminarily investigate nutritional interventions that may decrease GHG emissions in ruminants, the objective of the current study was to perform a systematic metaanalysis of in vitro studies to identify feed additives with a potential to reduce enteric CH 4 emission.Our goal was to reveal mitigation strategies that effectively decreased in vitro CH 4 production, had no negative effect on ruminal fermentation and nutrient digestibility, and have not been extensively studied in vivo.

Literature Search
A comprehensive search of the literature was conducted to identify experiments evaluating CH 4 mitigation interventions in vitro.Databases of the Commonwealth Agricultural Bureau International (CABI), the EBSCO Discovery Service, and the Web of Science were searched, and data were compiled for this metaanalysis.The search was conducted in February 2019 using the terms in vitro in combination with "methane," "fermentation," or "gas production."The abstract content of the publications retrieved by the searched criteria (n = 1,199) was reviewed and publications were selected for further consideration if they included in vitro measurements of CH 4 production, a clearly defined treatment and control, and multiple experimental replications (at least 2 or more replicates for each treatment within a study).Additional publications were searched whenever a citation in a manuscript identified a reference not listed in the searched database.Only peer-reviewed manuscripts published in English were selected for this meta-analysis.

Exclusion Criteria and Studies Included in the Database.
A PRISMA diagram (Moher et al., 2009) of the flow of data collection for the meta-analysis is presented in Figure 1.After the initial search and screening, 1,059 publications including multiple experiments were assessed for eligibility.From those, 513 studies were excluded because of the following reasons: abstract in English but full article in other language (n = 98); lack of control (n = 155); data published as abstract only (n = 121); data reported as figures (n = 37); incomplete methodology (n = 13); publication was not peer-reviewed (n = 9); error terms were not reported (n = 16); number of observations was not clear (n = 26); treatments could not be defined (n = 38).The complete database consisted of 546 publications from 1979 through 2018, containing CH4 mitigation interventions related to dietary formulation (n = 269 publications) and feed additives supplementation (n = 277 publications).The database is available at The Pennsylvania State University's ScholarSphere repository (https:// scholarsphere.psu.edu/resources/fdfe07ea-d631-459c-80c1-ddd9efb3dfc0;Martins et al., 2023a).Given the extent of the database, CH4 mitigation interventions related to dietary formulation were removed, and the current meta-analysis was focused on the in vitro mitigation effects of feed additives only.A list of the 277 publications is provided in Supplemental File (  S1; https://scholarsphere.psu.edu/resources/89508fb6-e603-41a3-8daa-41aa579449e5;Martins et al., 2023b).
Data Extraction and Classification of Mitigation Interventions.Individual CH 4 mitigation interventions were identified and classified into 14 categories of feed additives based on their type, chemical composition, and mode of action, as follows: antibodies, chemical elements, chemical inhibitors, direct fed microbials, electron sinks, enzymes, essential oils, flavonoids, macroalgae, organic acids, plant extracts, plants inclusion, saponin, and tannins.Plants inclusion, although not necessarily being used as feed additives in the original publications, was included in the current meta-analysis because they were mostly non-conventional forage sources containing different bioactive compounds (e.g., tannins, saponins, and polyphenols), which, could have a potential to modify ruminal fermentation.Response variables considered in this meta-analysis were CH 4 production (number of treatment means comparisons = 1,325); total VFA (n = 1,007), acetate (n = 783), propionate (n = 792), and butyrate (n = 776) concentrations; acetate to propionate ratio (A:P; n = 675); digestibility of DM (n = 489), OM (n = 277), and NDF (n = 177).Each study was examined individually to assign treatments as control (i.e., baseline condition) and treatment (i.e., an intervention aimed at reducing enteric CH 4 emission).In studies where more than one treatment dose was used, estimates (i.e., treatment means) and error terms across doses were averaged and compared with the control mean.Treatment means and error terms in studies with a factorial arrangement (i.e., inclusion of multiple and different treatments in the same study) were analyzed separately (not averaged) and compared with the control.

Statistical Analysis.
Treatment and control means, error terms, and number of observations extracted from each study were compiled in a database using Microsoft Ⓡ Excel (version 16.64; Microsoft Corporation).The error terms CV, LSD, relative standard deviation, relative standard error, SE, and SEM were converted to SD using SAS (SAS, v9.4; SAS Institute).Effect size estimates and corresponding sampling variances were obtained using the "metafor" (version 4.2-0) and "robumeta" (version 2.1) packages in RStudio (version v2023.03,R Foundation for Statistical Computing), following the methodology detailed in the meta-analysis by Dijkstra et al. (2018).Briefly, the mean difference (MD) of the response variables were calculated as the difference between treatment mean and its respective control mean using the "escalc" function in the "metafor" package.The magnitude of response variables and the units of CH 4 production reported varied greatly from study to study; therefore, relative MD [MD expressed as a fraction (in %) of observed control mean] was the effect size in further analysis.Relative MD was checked for normality using the "boxplot" function, and extreme values were excluded.Considering that studies included in the current meta-analysis contained multiple treatment groups sharing a common control group, the robust variance estimation (RVE) method (Tanner-Smith et al., 2016) was used to analyze statistically dependent effect sizes.Random-effects models were fitted using the "robu" function in the "robumeta" package (Fisher et al., 2023) to estimate between-study variance and heterogeneity statistics.The RVE random-effect model included the effect of mitigation interventions (i.e., categories of feed additives and individual CH 4 mitigation interventions within categories) for all response variables, except for CH 4 production.A RVE mixed-effect meta-regression model was constructed by including relative total gas production (i.e., treatment total gas production ÷ control total gas production) as an explanatory variable for the effects of mitigation interventions on CH 4 production.Estimated effect size for CH 4 production obtained from interventions with less than 4 degrees of freedom were omitted from the present study, following the package recommendation (Fisher et al., 2023).Statistical differences were considered significant at P ≤ 0.05 and tendencies at 0.05 < P ≤ 0.10.

RESULTS
To facilitate the interpretation and discussion of the data in the current manuscript, only tendencies and statistically significant results (i.e., P ≤ 0.10) from feed additives decreasing CH 4 production by more than 20% were presented.A complete list of individual CH 4 mitigation interventions with more than 4 degrees of freedom (n = 170) is available in the Supplemental Data set S1 (https: / / scholarsphere .psu.edu/resources/ 89508fb6 -e603 -41a3 -8daa -41aa579449e5; Martins et al., 2023b).A summary of the main statistical parameters for random and mixed-effect models evaluating feed additives categories is described in Table 1.It should be noticed that total gas production was a significant (P = 0.02) explanatory variable, and its inclusion clearly changed the estimates for relative MD of CH 4 production across mitigation categories, as it can be observed by the discrepancy between estimates presented in Table 1 and Figure 2. Additionally, due to diversity of the data, there were high heterogeneity (i.e., I 2 statistic) and between-study variance (i.e.,  2 ) across the models evaluating feed additives in the current study.Forest plots with the relative MD ± SE, number of treatment and control mean comparisons, and P-values are presented in Figures 2, 3, 4 and 5 summarizing the effects of the feed additives categories on CH 4 production, total VFA concentration, A:P, and DM digestibility.Estimated effect size or 95% CI are described in text to help readers better interpret the results, where appropriate.

Mitigation Effects of Individual Interventions within Feed Additives Categories
Five-hundred and 4 individual CH 4 mitigation interventions within feed additives categories were identified in the current study, but only 70 decreased in vitro CH 4 production by more than 20%.Direct fed microbials, electron sinks, essential oils, plant extracts, plants inclusion, and tannins were the only categories that contained individual CH 4 mitigation interventions producing estimates with 4 or more degrees of freedom in the statistical model.A summary of these variables will be presented in the current section of the manuscript.A complete description of statistically significant CH 4 interventions (n = 170) can be accessed in the Supplemental Data set S1 (https: / / scholarsphere .psu.edu/resources/ 89508fb6 -e603 -41a3 -8daa -41aa579449e5; Martins et al., 2023b).
Enterococcus spp.decreased CH 4 production by 20.3% and was the only effective individual mitigation intervention identified in the direct fed microbials category.Total VFA and acetate concentrations were not A robust variance estimate mixed-effect meta-regression model was constructed by including relative total gas production (i.e., treatment total gas production ÷ control total gas production) as an explanatory variable in the final mixed-effect model for CH 4 production. 3 Final mixed-effect model contained all categories included in the random-effect model.Estimate effect size of the relative mean differences (%) ± SE, number of observations and P-values are described in Figure 2.

Mitigation Effects of Individual Interventions with Limited Data
As previously indicated, estimated effect size obtained from interventions with less than 4 degrees of freedom were omitted from the present study; however, we believe it is important to highlight the mitigation effect of individual interventions that resulted in a large decrease

DISCUSSION
The data included in the present meta-analysis was generated from all types of in vitro systems (e.g., batch culture and non-RUSITEC and RUSITEC continuous culture), performed with different types of rumen inoculums (identified by authors as filtered, liquid, liquid and solid, and strained ruminal fluid) collected from different donor species (e.g., sheep, goats, dairy and beef cattle, and buffalos).Donors were also from different breeds, raised in diverse environmental conditions, fed with different diets, and were at different physiological states.The results from this meta-analysis can Although the above-mentioned variables could interact with CH 4 mitigation interventions, the main objective of the current study was to identify potential strategies that have not been extensively evaluated in vivo.Therefore, the interactions between CH 4 mitigation interventions and diet composition, animal species, and animal physiological state were not tested and should be addressed in future analyses.The anti-methanogenic effect of some feed additives determined in the current analysis, such as the essential oils β-cyclodextrin, cinnamaldehyde and garlic, Leptospermum petersonii, Santa-lum spicatum, and Tagetes minuta, are not comparable to effects of essential oils reported in vivo (Hegarty et al., 2021).It is known that in vitro studies tend to use higher doses than what would be practical or safe to the animal (Calsamiglia et al., 2007;Benchaar and Greathead, 2011), and toxic levels for some feed additives have not yet been well stablished in vivo.As an example, the 75.7% decrease in CH 4 production by β-cyclodextrin supplementation was obtained from an average of effects ranging from −25 to −97% in a study evaluating 3 doses of supplementation (0.1, 0.2, and 0.4 mM; Mohammed et al., 2004).Data reported in Mohammed et al. (2004) supported a high potential of β-cyclodextrin to mitigate CH 4 production, but there are no toxicological data for its dietary supplementation in ruminants.A 52-wk toxicity study indicated that the non-toxic effect level for β-cyclodextrin in rats and dogs was 11 mM and 44 mM (i.e., equivalent to 654 to 1967 mg/kg per day), respectively (Bellringer et al., 1995), indicating that 0.4 mM (i.e., equivalent to 4.5 g/cow per day for a cow with a rumen volume of 100 L) could be safe to be further evaluated in vivo.In this sense, for strategies where CH 4 production was decreased by more than 20% and further evaluations have not yet been conducted, estimates should not be considered as representative of in vivo responses until more studies are conducted and non-toxic levels are well stablished.
From all mitigation interventions evaluated, chemical inhibitors category was the most effective to decrease CH 4 production without largely affecting ruminal fermentation and digestibility of nutrients, and with a relatively large number of studies (n = 16).As reviewed by Hristov et al. (2013), studies investigating supplementation of chemical inhibitors suggested a reduction of approximately 50% of enteric CH 4 production, with a possible adaptation by the rumen microbes to this class of compounds over time.A successful example of a chemical inhibitor used in vivo is 3-nitrooxypropanol (3-NOP; Dijkstra et al., 2018;Melgar et al., 2021;Kebreab et al., 2023), which has a potential to decrease enteric CH 4 emissions by up to 30% in dairy cows without negatively affecting, and even improving (i.e., milk fat concentration), animal performance (Hristov et al., 2022).In a recent meta-analysis, Arndt et al. (2022) reported 3-NOP and bromochloromethane to have the largest CH 4 mitigation effect in sheep and cattle, with no effect on DMI, fiber digestibility, MY, or weight gain (bromochloromethane data only).Despite the high mitigation effect of the category, individual interventions within chemical inhibitors category (e.g., 3-nitro-1-propionate, 3-NOP, bromochloroacetic acid, and bromoethanesulfonate) did not significantly affect CH 4 production.This result could be likely explained by the limited number of studies evaluating chemical inhibitors included in the current database (e.g., maximum of 4 observations for bromoethanesulfonate).
There is a gap in the scientific knowledge regarding long-term efficacy of most CH 4 mitigation strategies (Hristov et al., 2022), especially for those related to rumen manipulation.Hristov et al. (2013) reviewed the literature and described up to 50% decrease of enteric CH 4 production by electron sinks supplementation, which was comparable to the reduction observed for chemical inhibitors in the same report.Overall, electron sinks supplementation effectively reduced CH 4 production in the current analysis, and nitrate, nitroglycerin, and nitrophenol were the most effective individual interventions within the category; however, their negative effects on total VFA concentration and DM digestibility could be detrimental to the in vivo application of these electron sinks.Fumaric acid and nitrate were reported as effective enteric CH 4 mitigation strategies in the meta-analysis by Arndt et al. (2022), and nitrate decreased CH 4 production by 20% across different species in the meta-analysis by Congio et al. (2021).This result aligns with the 20.2% decreased CH 4 production by nitrate supplementation in the current analysis.As reviewed by Hristov et al. (2013), the adaptability of the rumen environment, the potential increase in ammonia production, and the potential toxicity caused from intermediate products of nitrate metabolism are some of the concerns regarding the use of electron sinks in ruminant nutrition.The use of different combinations of electron sinks may increase their mitigation potential, but additive effects of these compounds have not been evaluated in the present study.
Direct fed microbials and enzymes supplementation, which are some of the most traditional ruminal fermentation modifiers evaluated in the present study, did not affect in vitro CH 4 production when considering the overall category effect.Nevertheless, Enterococcus spp., as an individual CH 4 mitigation intervention, was one of the most prominent strategies identified in the current analysis.The supplementation Enterococcus spp.decreased CH 4 production without affecting total VFA concentration and digestibility of nutrients and should be further investigated in vivo.The negative effect of Enterococcus spp.supplementation on propionate concentration, however, should be addressed in future studies.It is important to note that the supplementation of direct fed microbials and exogenous enzymes, although not effective in reducing CH 4 production in the current study, could contribute to decreasing the intensity of CH 4 emissions by increasing feed efficiency, animal productivity, and nutrient digestibility.For example, enhanced animal performance with yeast products supplementation has been reported across ruminant species (Desnoyers et al., 2009), including dairy cows (Poppy et al., 2012).
Plant secondary compounds are generally classified into saponins, tannins, and essential oils categories (Calsamilgia et al., 2007), and they have been extensively investigated as rumen modifiers in vitro and in vivo (Benchaar and Greathead, 2011;Cobellis et al., 2016;Honan et al., 2022).Tannins and saponins are known to have antinutritional effects, which can be especially problematic when dietary protein is limiting animal production.The supplementation of tanniferous forages to diets containing adequate levels of nutrients, on the other hand, decreased enteric CH 4 production by 12% without affecting milk yield or weight gain, even though DM digestibility was decreased by 12% in the meta-analysis by Arndt et al. (2022).Supplementation of different sources of tannins decreased enteric CH 4 yield by up to 27%, but also dramatically decreased (i.e., 51% reduction by Leucaena spp.) DMI in the meta-analysis by Congio et al. (2021).Tannins can also decrease nutrient digestibility, which also corroborated with the slightly decreased OM digestibility (−5.7%) data presented in the current study.
The relatively low mitigation efficacy of most plant secondary compounds described in the current study corroborates with in vivo and other in vitro analyses (Hristov et al., 2013;Hegarty et al., 2021).It should be noted, however, that many of these compounds have not yet been extensively studied in vivo, and they might have additional physiological effects other than modifying ruminal fermentation.Most essential oils have a broad spectrum of activities, in some cases negatively affecting overall ruminal fermentation, and may also interact with dietary composition and animal metabolism.For example, Silvestre et al. (2022) reported up to 7.5% reduction in enteric CH 4 production in dairy cows supplemented with increasing doses of a combination of Capsicum oleoresin and clove oil, and a quadratic effect of the same blend on blood β-hydroxybutyrate concentrations.Overall, plant secondary metabolites are expected to produce less than 10% reduction of enteric CH 4 emission (Hegarty et al., 2021).Future research should attempt to evaluate different combinations of compounds and their interactions with diet and animal metabolism, as well as the additive effect of combining plant secondary compounds with other rumen manipulation strategies.

CONCLUSIONS
In summary, the present meta-analysis characterized the in vitro effects of 170 CH 4 mitigation interventions with different efficacies.However, estimated effect sizes for most mitigation interventions were obtained from less than 5 observations, as a consequence of limited data available in the literature.The most effective category evaluated in this study was chemical inhibitors, decreasing in vitro CH 4 production by 29.2% without negatively affecting ruminal fermentation and digestibility of nutrients.When evaluating individual interventions within feed additives category, Enterococcus spp.(i.e., direct fed microbials) and nitrophenol (i.e., electron sinks) decreased CH 4 production by 20 and 27.1%, respectively, without largely affecting ruminal fermentation.Plant secondary compounds categories (e.g., essential oils, plant extracts, plants inclusion, saponins, and tannins) decreased CH 4 production by up to 18.2%.Many individual mitigation interventions within the former categories were effective in decreasing CH 4 production, but estimates were mostly generated from 1 or 2 observations.An exception was for tan-nin extracted from Leucaena spp., which data from 22 observations demonstrated an effectiveness in decreasing in vitro CH 4 production by 23.5%, without extensively affecting ruminal fermentation and digestibility of nutrients.Some of the strategies herein evaluated did not affect CH 4 production but may affect animal performance and thus decrease CH 4 emission yield and intensity.These effects should also be considered when selecting mitigation strategies to be further evaluated in vivo.Number of observations used for statistical analysis, lack of data regarding ruminal fermentation and digestibility of nutrients reported in the current analysis (see Supplemental Data set S1), and overestimation of effect of treatments due to high doses should be considered before selecting individual mitigation interventions to be tested in vivo.

Table
Martins et al.: METHANE MITIGATION STRATEGIES

Table 1 .
Martins et al.: METHANE MITIGATION STRATEGIES Number of publications (Publ.),number of mean comparisons (n), estimated effect size (Mean), standard error (SE), heterogeneity (I 2 ), and between-study variance ( 2 ) obtained from the models evaluating the effects of mitigation interventions on in vitro CH 4 production 1Mean represents the estimate effect size of the relative mean difference (%) between control (i.e., baseline condition) and treatment (i.e., a strategy aimed at reducing enteric CH4 emission).Relative mean difference was calculated as follows: Relative MD, % = [(treatment meancontrol mean) ÷ control mean].2